Insights Into Formation and Growth of Colloidal Multielement Alloy Nanoparticles in Solution through In Situ Liquid Cell TEM Study

The nucleation and growth of nanoparticles are critical processes determining the size, shape, and properties of resulting nanoparticles. However, understanding the complex mechanisms guiding the formation and growth of colloidal multielement alloy nanoparticles remains incomplete due to the involvement of multiple elements with different properties. This study investigates in situ colloidal synthesis of multielement alloys using transmission electron microscopy (TEM) in a liquid cell. Two different pathways for nanoparticle formation in a solution containing Au, Pt, Ir, Cu, and Ni elements, resulting in two distinct sets of particles are observed. One set exhibits high Au and Cu content, ranging from 10 to 30 nm, while the other set is multi‐elemental, with Pt, Cu, Ir, and Ni, all less than 4 nm. The findings suggest that, besides element miscibility, metal ion characteristics, particularly reduction rates, and valence numbers, significantly impact particle composition during early formation stages. Density functional theory (DFT) simulations confirm differences in nanoparticle composition and surface properties collectively influence the unique growth behaviors in each nanoparticle set. This study illuminates mechanisms underlying the formation and growth of multielement nanoparticles by emphasizing factors responsible for chemical separation and effects of interplay between composition, surface energies, and element miscibility on final nanoparticles size and structure.


Introduction
[17] Co-reduction of metals from their ionic solution mixture, is a simple and effective wet-chemistry process to synthesize alloy nanoparticles. [18,19]22] The mechanism of radiation-induced synthesis of colloidal metal nanoparticles has been studied for many years but only limited to mono and bimetallic systems (Note S1, Supporting Information).[28] First, due to the large difference in reduction potential and rate for the different metal ion precursors, simultaneous reduction may be problematic and particularly challenging for the preparation of multielement alloys (Notes S1-S3, Supporting Information).Additionally, it is critical to control the nucleation and growth processes to prevent the formation of separate monometallic nanoparticles of each element (Notes S4-S7, Supporting Information). [29]Consequently, regulating the alloy nanoparticle composition is difficult due to the difficulty of controlling the thermodynamics and kinetics of different metals during the nucleation and growth stages.Despite all the challenges, the fast co-reduction of metals in the radiation-induced synthesis method has the potential to suppress de-mixing of atoms and overcome the miscibility gaps of elements. [24]In fact, the radiation-induced co-reduction and formation of colloidal multi-element alloy nanoparticles can be a prospective bottom-up room temperature synthesis method analogous to fast quenching of atoms from high temperatures in carbothermal shock synthesis of high entropy alloys with the advantage of having higher yield of production. [30]here have been few studies on the mechanism of wetchemistry low temperature synthesis for multielement alloy nanoparticles despite their importance.An autocatalysis mechanism has been reported based on the evidence showing of Pd-rich early composition of Pt-Ir-Pd-Rh-Ru and Ni-Pd-Rh-Ir-Pt nanoparticles. [31,32]The autocatalyzed growth particles along the {111} facets to facilitates the reduction of metal salts is a suggested pathway, although results show that the pathway is system related and can change depending on the selection of elements. [32]The colloidal route is quite challenging for a system containing elements from different transition metal series (3d, 4d, and 5d) with large atomic radii disparities up to 20%, since all the constituting elements differ structurally and chemically.Providing a reduction temperature high enough to decompose all the metal salts is argued to be necessary for having a homogeneous and nearly equimolar ratio of elements. [21]Although, the control over the composition is more difficult in systems containing metals with higher oxidation states undergoing reduction in multiple steps.On the other hand, ultrasmall size Ir-Pd-Pt-Rh-Ru nanoparticles (1.32 nm) with homogenous composition synthesized via room temperature co-reduction of metal salts by a strong reducing agent were reported by Minamihara et al. [33] In this case, the ultra-small multielement nanoparticles with homogeneous distribution did not undergo any autocatalyzed reduction and growth.In addition, the decomposition temperature of the metal salts was not an effective factor since synthesis occurred at room temperature.The high reducing power of the used reducing agent (E 0 = −3.10V) is claimed to accelerate the simultaneous reduction of all metal salts and the nucleation of small metal clusters at room temperature.The ratio of metal ion to thiolate polymer ligands when considered as a variable was also shown to affect the size and composition of alloy nanoparticles. [34]verall, the nucleation and growth of multielement nanoparticles pose several scientific challenges, including (i) the nucleation process as the mechanism that drives the formation of these particles are not clear; (ii) understanding how to control the composition and structure of multielement nanoparticles as it requires a detailed understanding of the interplay between the constituent elements and their interactions with each other; and (iii) the nucleation and growth starts at time and length scales that are too small to be characterized using conventional techniques.In situ studies of electron beam-induced formation of metal nanoparticles from the precursor solution in the liquid cell transmission electron microscopy (LC-TEM) is one of the powerful methods to reveal the nucleation and growth mechanisms. [35,36]In this study, using LC-TEM, the mechanisms of formation and growth of colloidal multielement alloy nanoparticles are investigated and important thermodynamic and kinetic considerations that govern their occurrence are discussed.Moreover, the evidence of factors leading to chemical segregation of specific elements in the mix-ture and the effect of constitutional elements of alloy on size and structure of the product nanoparticles are highlighted.

In Situ TEM Observation of Colloidal Nanoparticles Formation In Liquid Cell
An equimolar aqueous mixture of HAuCl  1a,b show the setup for the liquid flow holder where the solution of mixed metal salts was viewed under the electron beam.Colloidal particles formed inside the liquid cell under electron beam irradiation showed two distinct patterns of nucleation and growth.Figure 1c,d show a sequence of TEM snapshots (colored images extracted from Movie S1 and S2, Supporting Information) where the liquid solution flows into the imaging window.Upon reaching the center of the imaging area and after the reduction of metal ions in reaction with solvated electrons in solution, small particles nucleate and grow (Note S1, Supporting Information).The first set of formed particles shown in Figure 1c (Figure S1 and Movie S1, Supporting Information) are detectable as individual particles which rapidly grow to the size that are separated from the rest of the ionic solution.These observations are analogous to the precipitates forming in a metal salt solution upon adding a reducing agent in the batch synthesis of colloidal nanoparticles.We refer to these rapidly formed large particles as primary particles to differentiate them from the particles forming at the later stage.The secondary particles follow a different formation and growth behavior shown in Figure 1d (Figure S2 and Movie S2, Supporting Information).In this pathway, a cloud-like network of nano-colloidal particles appears and spreads within the ionic solution inside the liquid cell where individual particles are not detectable clearly at low magnifications and the resolution limit of cell thickness.

Elemental Composition Analysis of The Formed Colloidal Nanoparticles
The annular dark field scanning TEM (ADF-STEM) imaging and energy-dispersive spectroscopy (EDS) characterizations are performed at dry state on the surface of a single open microchip after disassemply of the liquid cell.Figure 2 shows the ADF-STEM images of the formed nanoparticles along with corresponding EDS analysis of the elemental distribution map and composition of the final product nanoparticles at the end of the process.Interestingly, the two different sets of larger primary and ultrasmall secondary formed particles exhibit distinct elemental compositions.In Figure 2a,b, it is evident that Au is notably concentrated in the larger, separated 25 nm particle considered the primary particle, while Pt, Ir, and Ni elements are concentrated in smaller aggregated secondary particles and only Cu is evenly distributed among all the formed particles.Figure 2c displays the low-angle annular dark field (LAADF) STEM images of a single 25 nm primary Au-rich alloy nanoparticles.The STEM-EDS elemental map in Figure 2d indicates that the primary nanoparticles are alloy nanoparticles with a near-homogenous distribution of elements without chemical segregation in a single particle.The quantitative elemental composition analysis obtained from EDS spectra in Figure 2e reveals the primary nanoparticle composed of 80.3 at.%Au, 18.2 at.%Cu, and 1.5 at.% other three elements (Pt, Ir, and Ni) are mainly Au-Cu binary alloys.(Figure 2e).Additional Au-Cu primary nanoparticles with corresponding STEM-EDS analysis and average composition are shown in Figure S3 (Supporting Information).The HAADF-STEM image with higher resolution in Figure 2f shows an area with aggregation of a high number of small nanoparticles with sizes of ≈1-4 nm.In Figure 2g, the STEM-EDS elemental mapping of the particles shown in the selected area represents the distribution of elements in the secondary nanoparticles.The quantitative EDS analysis of the multielement alloy nanoparticles indicates an elemental composition of 38.7 at.%Pt. 36.1 at.%Cu, 17.1% Ir and 8.1% Ni.Additional multielement secondary particles with corresponding STEM-EDS analysis are shown in Figure S4 (Supporting Information).The EDS analysis maps on single secondary particles demonstrate a random and disordered compositional distribution of Pt, Cu, Ir, and Ni elements even in particles as small as 1-2 nm (Figure S4b, Supporting Information).The quantitative EDS analysis on the composition of several individual nanoparticles and different selected regions with the agglomeration of secondary nanoparticles indicates that secondary nanoparticles do not have uniform composition and there is variation mainly in the Pt and Cu at.% (Figure S4b-d, Supporting Information).Overall particles show three types of Pt-rich (Pt 51±9 at.%,Cu 24±5 at.%,Ir 18±3.5 at.%,Ni 6±2 at.% and Au 1±0.5 at.%),Cu-rich (Cu 46±4 at.%,Pt 31±4at.%,Ir 17±3 at.%,Ni 5±1 at.% and Au 0.4±0.3at.%) and near equimolar Pt-Cu (Pt 40±1.5 at.%,Cu 38±2 at.%,Ir 16.5±2 at.%,Ni 5±1.5 at.% and Au 1±0.5 at.%) compositions and the calculated average composition is Pt 41±8 at.%,Cu 36±9 at.%,Ir 17±0.5 at.%,Ni 5±0.5 at.%andAu 0.5±0.2at.%. (Figure S4d, Supporting Information) Additionally, multielement colloidal nanoparticles were fabricated in an ex-situ batch solution synthesis through the coreduction of mixed metal ion precursor using NaBH 4 as a reducing agent (Figure S5a,b, Supporting Information).The compositions of product colloidal nanoparticles were characterized by the STEM-EDS analysis for comparison.The result reveals a quite similar composition distribution in product nanoparticles where Au is segregated in larger particles surrounded with clouds of ultra-small nanoparticles composed of Cu, Pt, Ir, and Ni elements (Figure S5c-h, Supporting Information).

Crystal Structure and Size Evolution Analysis
In Figure 3a, four different TEM snapshots (1.00, 1.67, 2.33, and 3.00 seconds) and the enlarged images of the selected region representing the formation of primary nanoparticles are illustrated.To gain a better understanding of their overall growth patterns, the size (by diameter) distribution of the particles was determined by measuring the number and diameters of detected particles at each frame, as shown in Figure 3b,c.The TEM snapshots obtained at this level of magnification possess the resolution that enables us to identify particles larger than 1 nm.Through the monitoring of the identified particles, it was observed that the smallest size of the detected particles undergoing growth in size over time exceeds 2 nm.Consequently, our analysis focused on The EDS spectra for quantitative analysis of the elemental atomic composition of the primary particle shown in (c) which specifies it as Au-rich binary Au-Cu alloy nanoparticle.f) HAADF-STEM image of agglomerated secondary nanoparticles with sizes of 1-4 nm.g) STEM-EDS elemental mapping of a selected area in (f) showing a homogeneous distribution of elements.The scale bar is 5 nm.h) The EDS spectra for quantitative analysis of elemental atomic composition of secondary particles shown in the selected area in (f) which specifies them as quaternary Pt-Cu-Ir-Ni alloy nanoparticles.
quantifying and measuring particle count and size, with a defined lower limit of 2 nm for particle size.The outcomes of the analysis reveal an increase in both average size and number of particles over time.After 1 second, there were 19 detected particles with an average size of 3.92 nm, which increased to 30 particles with an average size of 5.41 nm after 3 seconds.Figure 3d presents snapshot sequences that track the growth of seven individual particles in less than 3 seconds.The change in size of these selected particles over time is measured and plotted in Figure 3e.The plot reveals that the initial sizes of all seven particles, when detected as single particles, were ≈3 nm.They exhibited rapid growth and reached sizes of ≈7-10 nm in less than 2 seconds.The growth of a single particle over time occurs by either atomic addition based the classical growth or by autocatalytic surface growth when metal ions are attached to the particle's surface, reduced, and converted to surface atoms.The observed rapid growth of the individual primary particles suggests a facilitated surface interaction for Au-Cu alloy nanoparticles.The results of the ex-situ synthesized colloidal nanoparticles in Figure S6a-c (Supporting Information) display a similar behavior where Au-Cu nanoparticles have reached sizes > 5 nm within the short time of the reaction synthesis and show (fcc) face centered cubic crystal structure (lattice constant size of a = 4.12 Å).
The growth of the Au-Cu primary particles further proceeded through the coalescence of particles.Figure 4a (Movie S3) shows the enlarged snapshots of three particles similar in size that undergo coalescence and merge into a single triangular-shaped particle over a span of 4 seconds.The process begins with particles moving towards each other and necking between two of three particles, followed by the thickening of the neck and the formation of a dumbbell-shaped particle.Subsequently, another necking mechanism and thickening of the neck occurs between the third particle and the other two joined particles.The thickening of the neck is attributed to the diffusion of atoms on the surface of particles after the neck has formed. [37]The diffusion and relocation of surface atoms lead to the reshaping and minimizing of the surface energy of the new, larger particles.The change in contrast within the coalesced particles indicates recrystallization through interparticle alignment to form a single crystalline structure.Hence, the coalescence and intraparticle ripening growth mechanism is responsible for the formation of the final larger Au-Cu particles shown with HAADF-STEM images in Figure 4b (Note S6, Supporting Information).
Figure 5a provides enlarged TEM snapshots of the selected region, illustrating the formation of the secondary nanoparticles at four different time points (1.00, 1.5, 2.33, and 3.00 seconds).It demonstrates that as time proceeds, the cloud-like structure expands in the liquid, leading to the formation of a greater number of small clusters.However, individual particles larger than 2 nm in size are not observed, and the size of the identified small particles (< 2 nm) does not grow considerably throughout the observation period.Figure 5b shows a HAADF-STEM image of the region with aggregated secondary particles obtained after the process was completed.The enlarged area reveals that the sizes of these particles fall within the range of 1-2 nm.This indicates that the secondary particles remain stable following the observed formation and growth, without undergoing further changes in the liquid cell during the in situ process.Information about the early stages of atomic clusters formation through the aggregation of atoms, as well as the transformation of atomic clusters into nanoparticles, can be derived from the ADF-STEM images shown in Figure 5c.These images depict the coexistence of crystalline nanoparticles (> 2 nm) alongside non-crystalline nanosized (< 2 nm) clusters, sub-nanometer clusters, and few-atom clusters.They provide insights into the initial phases of nonclassical nucleation, where few-atom clusters come together to form sub-nanometer clusters (Note S4, Supporting Information).These sub-nanometer clusters subsequently continue to grow as additional atoms join them, eventually evolving into disordered atomic clusters with sizes less than 2 nm.The evidence also suggests that crystallization of nanoparticles occurs when their size exceeds 2 nm.Similar results were observed in the case of ex situ synthesized multielement nanoparticles (Figure S6d,e, Supporting Information).The non-crystalline clusters, ranging in size from sub-nanometer to 2 nm, are found in an aggregated state, forming cloud-like structures.In contrast, when the particles exceed 2 nm in size, a crystalline structure becomes observable.
In Figure 6a, the high-magnification HAADF-STEM image of small Pt-Cu-Ir-Ni nanoparticles with sizes ranging from 1 to 4 nm, reveals their observable crystalline structure.The particles appear in an aggregated state, with multiple crystalline particles of varying sizes and crystallographic orientations attached by amorphous grain boundaries.At the surfaces of these particles, disordered bonded atoms can be observed that also serve as a linkage between them.However, the subsequent stages of coalescence through interparticle ripening to merge the particles are not observed.Furthermore, small metal clusters, with sizes around 1 nm, are also visible in the image where one of these clusters is in direct contact with the crystalline particle, as indicated by an arrow in Figure 6a.This observation suggests that the growth of multielement nanoparticles may also proceed through the diffusion and addition of atoms originating from attached small metal clusters (Note S6, Supporting Information).
Notably, there is a contrast difference among the individual atoms in the HAADF-STEM images.The contrast in HAADF-STEM is associated with the atomic number (Z) of the elements within the nanoparticles. [38]Figure 6b displays the Fast Fourier Transformation (FFT) pattern of the HAADF-STEM image shown in Figure 6a where the measured d-spacings of 0.224, 0.194, 0.137, and 0.118 nm, are corresponding to (111), ( 200), (220), and (311) crystalline planes, respectively.This implies that the multielement secondary nanoparticles have a face-centered cubic (fcc) structure with a lattice constant of 3.88 Å, which is close to the lattice constants of platinum (3.97 Å) and iridium (3.87 Å) crystals.The FFTs of the two selected particles in Figure 6a show their single crystalline fcc structure at [01 1] (Figure 6c) and [001] (Figure 6d) crystal view directions.The FFT analysis of the ex-situ synthesized multielement nanoparticles also reveals a similar fcc crystal structure with measured lattice constant size of a = 3.85 Å.
Figure 6e provides a visual representation highlighting the aggregation of multiple small crystalline nanoparticles interconnected by amorphous grain boundaries.Additionally, in Figure S7 (Supporting Information), the presence of non-crystalline bonded atoms on the surface of particles is depicted.These observations coupled with the detection of a chlorine (Cl) peak in the EDS spectra of the secondary nanoparticles (Figure 2h), suggests that the slow surface interaction and presence of charged metal ions on the surface of particles can contribute to their slower rate of surface growth.This phenomenon, in turn, inhibits direct crystal-to-crystal contact, thereby impeding growth through coalescence and merging, which would otherwise lead to the formation of a larger single particle.
The overall process of colloidal nanoparticles formation, where the presence of both primary and secondary particles is observable together within the imaging window, is shown in Figure S8 (Supporting Information) (Movie S4, Supporting Information).The analyzed data presenting the size and count measurements of the total nanoparticles formed over time (Figure S8b, Supporting Information) reveals a similar process and the formation of two distinct groups of particles within three main stages.First, the nucleation and individual surface growth of the primary particles occur .Subsequently, the coalescence of the primary particles takes place (Figure S8c, Supporting Information), and last, the nucleation and growth of the secondary particles occur.

Discussion
To explain the formation of two distinct sets of particles with such compositions and size differences from a mixed multielement solution, the underlying factors at each stage of particle formation need to be considered.First, it is important to identify the reasons behind the primary formation of Au-Cu nanoparticles and the secondary formation of Pt-Cu-Ir-Ni nanoparticles.
[28] The reduction of metal ions is the first stage with a crucial role on controlling the outcome of colloidal synthesis of alloy nanoparticles. [39,40]The simultaneous reduction of different elements to atoms generates an ideal equimolar feed for nucleation of the homogeneous multielement cluster.In contrast, the selective reduction of elements due to their different nobility and reduction potential may lead to formation of chemically phase segregated particles similar to bilayer core-shell structures in bimetallic systems. [24,41,42]The sequential reduction of elements does not always follow their relative nobility, but also depends significantly on the rate and mechanism of the reduction reaction.(Note S3, Supporting Information). [43,44]It has been observed in other multielement system studies that Pd is reduced earlier than the more noble metals such as Pt and Ir, leading to the formation of an initial Pd-rich seed. [31,32]The reduction of multivalent ions Mn + to M 0 involves multi-step reactions with different thermodynamics and kinetics at each step. [22][47] Among the five metallic ions in the solution, Au is the most noble element, and Pt is the second most noble element, and their reductions are thermodynamically more favored compared to the reduction of other metal ions in solution (Table S1, Supporting Information). [48]However, the reduction of [AuCl 4 ] -(Au 3+ ) ions is reported to be kinetically faster compared to [PtCl 6 ] 2-(Pt 4+ ) ions potentially due to the higher electronegativity of Au. [49] Therefore, the faster and more readily facilitated reduction of [AuCl 4 ] -in solution is expected to lead the primary formation of Au-Cu nanoparticles.On the other hand, the multi-step and slower reduction of [PtCl 6 ] 2-and IrCl 3 account for the delayed and secondary formation of multielement nanoparticles. [45,46,50]The presence of Cu in primary particles and its high atomic concentration in the secondary particles indicate that it has a more kinetically favored reduction compared to the other two noble metals, Pt, and Ir.Additionally, the negative reduction potential of Ni (Table S1, Supporting Information) makes it the least stable among the reduced elements, and it is more inclined to undergo selective oxidation and dissolution in the solution.
Subsequently, after the reduction steps, different rates and pathways of nanoparticle formation were experimentally observed for each set of particles.While primary Au-Cu particles quickly appeared within the solution as individual particles larger than 2 nm in size, the secondary nanoparticles formed through an increase in the number of clusters smaller than 2 nm.The presence of stable sub-nanometer and < 2 nm non-crystalline clusters in the final state of secondary multielement nanoparticles suggests that they undergo a multi-step non-classical nucleation process, starting with the formation of pre-nucleation small atomic clusters.The pre-nucleation species are known to act as intermediate states that lower the surface energy barrier during nucleation and somehow tie classical and non-classical theories together (Notes S4-S5, Supporting Information). [51,52]articles with higher bulk surface energy are more likely to undergo multi-step nucleation and form more stable prenucleation clusters. [53]Assuming the surface energy of multielement alloy is correlated with the surface energies of its constituent elements, Pt-Cu-Ir-Ni alloy has a higher surface energy compared to Au-Cu alloy particles (Figure S9a, Supporting Information).Therefore, the lower surface energy of Au-Cu can potentially accelerate the nucleation of crystalline Au-Cu nanoparticles, whereas the nucleation of crystalline Pt-Cu-Ir-Ni alloy nanoparticles faces more significant surface energy barriers (Notes S4,S5, Supporting Information).
Furthermore, it is essential to discuss the factors that determine the different elemental selectivity of primary and secondary particles during their early stages of formation, as well as the specific elemental compositions of each set of particles.
In a multielement system and during the multi-step reduction of metal ion complexes, premixing of metal elements and formation of metal-metal bond in pre-nucleation clusters are likely to happen. [54]Therefore, it can be expected that depending on different prenucleation species and premixing of metal elements, alloy nanoparticles with different compositions are formed in solution. [53]Despite their importance, little is known about the structure and function of prenucleation ionic complexes of metals. [55]The kinetics and mechanisms of reduction reactions of different metal ion complexes may depend on various factors and synthesis conditions and is challenging to be ruled out even in single-element systems. [55,56]Nevertheless, by comparing different properties of the metal elements in our system, we are suggesting some of the factors that may be correlated with the composition selection of resulting nanoparticles during the reduction steps (Figure S9b, Supporting Information).Au and Cu are the only elements in the mixture with stable oxidation state of 1+ while the minimum stable oxidation state of Pt and Ni is 2+ and for Ir is 3+ (Figure S9b, Supporting Information).This factor becomes particularly relevant when considering that in multi-step reduction processes of metal ions, the rate-limiting step typically involves reducing the metal from its lowest stable oxidation state to its zero-valent state.The similar valance electron configuration of Au (5s 1 5d 10 ) and Cu (4s 1 3d 10 ) and consequently similar metal-ligand geometry with high active surface area (linear [AuCl 2 ] 2-and [CuCl 2 ] -) may lead to preferential premixing of Au and Cu metal ions and formation of initial Au─Cu bond in Au n Cu m Cl x bi-metallic prenucleation complexes (Notes S2,S3, Supporting Information). [57,58]In addition, the presence of Cu + and its quick one-step reduction to zerovalent metal (Cu + /Cu 0 ) near or on the surface of Au cluster (Au n ) could be another reason for its better mixing with Au among other ions is solution.
Additionally, in this system, apart from the overall phase separation into two sets of particles, there is no significant phase segregation observed in individual particles, such as core-shell structures.Individual nanoparticles exhibit a nearly homogeneous but non-equimolar distribution of elements, indicating their solidsolution alloy structure.The theory of nanoalloy formation in small metallic clusters points out that the favored mixing of any two metal elements over forming a segregated structure depends on factors such as, the relative strength of metal-metal binding (or their mixing enthalpies), crystal structure, atomic size and the surface energies of individual metals . [59,60]The large negative mixing enthalpy between highly miscible Au and Cu elements, along with their significant size difference, serves as the main driving force toward their complete mixing and facile interatomic diffusion.(Figure S9b,c, Supporting Information) In this regard, a rapid and spontaneous mixing of Cu atoms in small Au nanoparticles at room temperature is reported before. [61]Among the four other elements, Pt and Cu have also a large negative mixing enthalpy and are highly miscible.Ir is immiscible with Pt and Cu elements, and only the formation energy of Ir─Ni bond is negative (Figure S9c, Supporting Information).Therefore, the lower composition of Ir compared to Cu and Pt can primarily be attributed to its energetically unfavorable bond formation with Pt and Cu.The slower reduction and rapid oxidation of intermediate partially reduced states of Ir in the solution further contribute to its lower composition in the final nanoparticles.Additionally, the weak binding energy and the similarity in size between Ir and Pt can result in slower interatomic diffusion.Consequently, Ir atoms that are reduced or attached to the particle surface may undergo oxidation before they have a chance to diffuse into the particles.The high concentration of Ni ions in the region without formed particles suggests that its lower concentration is primarily driven by the reduction steps (Figure S4a, S5f, Supporting Information).Since Ni is the only element with a negative reduction potential, it is more likely to sacrificially oxidize and dissolve in the solution.However, its homogeneous distribution within particles and ultra-small clusters suggests that it is present in the early stages of nucleation.(Figure S4b, Supporting Information).
Moreover, it is necessary to address the difference in growth behavior and the significant size disparity between the two sets of particles.The reduction kinetics of precursors affect not only the structure and composition of the initial seed but more importantly their growth patterns.There may be two different reduction pathways involved for the metal ion precursor during the colloidal synthesis.The dissolved metal ions can either be directly reduced to atoms in solution or adsorb on an existing particle surface and autocatalytically reduce to atom.In the early stages, the direct reduction of metal ions in solution is taking place which is responsible for forming the first atomic clusters.As the synthesis progresses, a combination of both solution reduction and surface reduction can occur, with one pathway potentially dominating the particle's growth at different stages.Due to the lower activation energy barrier and the associated autocatalytic process, precursor reduction primarily occurs on the surface of the seed particles.The direct solution reduction is notably slower compared to the surface reaction.While the rate of the precursor reduction in solution in early stages effectively regulates the structure and composition of the nanoparticles during the nucleation steps, the surface reduction predominantly modulates the growth of the nucleated nanoparticles. [62,63]Hence, in order to identify the major underlying factors leading to different rate of particles growth, it is important to compare the surface growth activity of two types of particles.

Density Functional Theory (DFT) Calculations
DFT simulations were employed to gain deeper insights into the molecular-level mechanisms underlying the growth of nanoparticles.Particularly, Au-Cu nanoparticles exhibited significantly faster growth rates and achieved larger sizes compared to the multielement nanoparticles composed of Pt, Cu, Ir, and Ni.To elucidate the factors contributing to this difference in growth behavior, a series of ab-initio calculations were conducted.Initially, the computational investigation focused on the growth of Au-Cu nanoparticles from [AuCl 2 ] -and [CuCl 2 ] -precursors.Subsequently, various precursors were used to study the growth of multielement nanoparticles.Experimental results presented in Figure 2 indicated that Au-Cu nanoparticles comprised 80% Au and 20% Cu, whereas the multielement nanoparticles had a composition of Pt (38 at%), Cu (37 at%), Ir (17 at%), and Ni (8 at%).Following these experimental findings, diverse nanoparticles were created, and DFT calculations were performed on their surfaces, as illustrated in Figures S10,S11 (Supporting Information).Figure S10 (Supporting Information) demonstrated the creation of Au nanoparticles, followed by Au 0.80 Cu 0.20 nanoparticle and the cleaved (111) surface, where the nanoparticle growth was investigated.Furthermore, the distribution of elements within the Au 80 Cu 20 nanoparticle was shown on the far right, revealing random Cu atom dispersion throughout the nanoparticle.The stability of different Au 80 Cu 20 nanoparticles with varying Cu distributions was explored, and it was observed that as long as Cu atoms were evenly and randomly distributed throughout the nanoparticles, the ground-state energy difference among these nanoparticles, as obtained from DFT calculations, remained negligible.Consequently, a representative surface from such nanoparticle was cleaved to study Au and Cu precipitation.Similarly, for the multielement nanoparticles, efforts were made to create experimentally comparable configurations for DFT calculations.Several multielement nanoparticle structures with varying positions of constituent elements were considered, and it was found that the ground-state energy of each nanoparticle was very similar as long as the atomic percentage was maintained.Therefore, a representative multielement nanoparticle was selected for DFT calculations.As the surface and subsurface composition around the nanoparticles resembled each other closely, a representative slab was generated by cleaving the first four layers of the nanoparticle, as shown in Figure S11 (Supporting Information).Additionally, the distribution of elements on such a cleaved (111) surface is presented on the right side of Figure S11 (Supporting Information).
After creating the corresponding nanoparticle and surface, we investigated the precursor adsorption and subsequent dissociation processes over Au 0.80 Cu 0.20 and Pt 0.38 Cu 0.37 Ir 0.17 Ni 0.08 (111) surfaces using DFT calculations.Figure 7a illustrates the DFT results of the interactions between [AuCl 2 ] -and [CuCl 2 ] -precursors with the Au 0.80 Cu 0.20 (111) surface, along with the resulting charge density.Initially, both precursors were placed on the surface composed of Au and Cu atoms.After optimization to find the ground-state energy, the precursors dissociated, giving rise to individual Au and Cu atoms on the surface.This dissociation occurred rapidly, with the Cl atoms detaching from their respective metals, resulting in an energetically favorable atomic configuration.Interestingly, the adsorbed Au and Cu atoms promptly formed a new layer on top of the (111) surface, with bond lengths similar to those in the bulk fcc structure.This observation suggests that the growth of Au-Cu nanoparticles occurs rapidly and without significant barriers, consistent with experimental measurements as depicted in Figure 3.A comparison was made between the dissociation of two [AuCl 2 ] -and two [CuCl 2 ] -fragments over the (111) surface (Figure S12, Supporting Information).Analyzing these DFT results, it was observed that [AuCl 2 ] - dissociated more rapidly (Figure S12a, Supporting Information) than [CuCl 2 ] -(Figure S12b, Supporting Information) fragments.The optimization led to the formation of two Au atoms on the (111) surface, following the fcc structure growth.A similar situation was observed for two [CuCl 2 ] -fragments, with one Cu atom assuming a position consistent with the fcc structure, while the other exhibited a significantly stretched Cl bond with a changed position.This comparison indicated that Au deposition occurred faster than Cu deposition, providing an explanation for the observed 80/20 ratio between Au and Cu in most of the experimentally observed nanoparticles.
Next, we compared the DFT results of Au and Cu precursor dissociation on the Au-Cu nanoparticle surface to the dissociation of different precursors on multielement alloy surfaces.Figure 7b depicted the results of [CuCl 2 ] -and [PtCl 4 ] 2-precursors' stability on the multielement (111) surface, consistent with the experimental composition of Pt (38 at%), Cu (37 at%), Ir (17 at%), and Ni (8 at%).The interaction of [CuCl 2 ] -and [PtCl 4 ] 2-with the multielement (111) surface showed a starkly different behavior compared to the Au-Cu (111) surface.After optimization, there was only a small rearrangement of precursors on the surface, with no dissociation or stretching of M-Cl bonds observed.The charge density iso-surface profile on the right in Figure 7b confirmed that all bonds remained intact, with a smooth charge density profile among all interacting atoms.Further DFT calculations were performed for a single [PtCl 4 ] 2-compared to four [PtCl 4 ] 2-fragments (Figure S13, Supporting Information).Despite the significant increase in coverage from one to four fragments, the additional interaction among [PtCl 4 ] 2-did not lead to dissociation and single metal atom deposition.The analysis of charge density on the right of Figure S13 (Supporting Information) supported this observation.Despite extensive efforts to search for energetically favorable adsorption locations for precursor dissociation on the multi-element surface, no such positions were found.This indicated that the growth of multielement nanoparticles would proceed through a different scenario compared to Au-Cu nanoparticles growth.
In addition to the initial surface growth of individual particles up to a certain size, further particle growth occurs through aggregative and coalescence mechanisms.Coalescence takes place when colloidal particles come into contact.The colloidal stability of particles in a colloidal solution depends on a balance between electrostatic repulsion and Van der Waals attraction between the particles.As particles grow in size, the electrostatic re-pulsion force decreases, while the Van der Waals attraction force increases. [64]In the case of Au-Cu nanoparticles, a rapid and complete reduction of ions at the particle surface revealed by DFT calculations leads to their swift enlargement, typically exceeding 5 nm in size.This size increase enhances Van der Waals attraction while weakening electrostatic repulsion, resulting in colloid instability and particle aggregation.The coalescence process initiates with the neck formation between adjacent particles, followed by a rapid intraparticle ripening and eventual fusion of the particles (as illustrated in Figure 4a).The relatively simple composition of Au-Cu bimetallic alloy nanoparticles, consisting of elements with rapid interatomic diffusion and high miscibility, facilitates the rapid intraparticle ripening and recombination of crystalline particles.A schematic illustrating the proposed mechanism for the formation of primary Au-Cu nanoparticles is presented in Figure 8a.
Conversely, the growth of small multielement alloy nanoparticles is impeded by coalescence.This can be attributed to their lower surface reduction activity, as indicated by DFT results, which results in smaller particle sizes and reduced Van der Waals attraction forces.Moreover, the HAADF-STEM images in Figure 6 and Figure S7 (Supporting Information) illustrate the surface structure of multielement particles and reveal the presence of an amorphous grain boundary between the aggregated nanoparticles.This, in combination with the DFT calculation results and detection of chlorine (Cl) in EDS analysis of multielement nanoparticles, supports the hypothesis that unreduced metal ions and ligands may be firmly attached to the particle surfaces.The presence of these ions and ligands on the particle surfaces is likely responsible for providing the necessary charge and electrostatic repulsion, effectively preventing the coalescence of small colloidal particles and their merging into larger particle.In addition, the higher surface energy of these particles make them more amenable to stabilization through interaction with charged ligands and stabilizer additives.Consequently, direct contact between crystalline nanoparticles is inhibited.Furthermore, due to the random distribution of elements in multielement nanoparticles, the atomic composition can vary significantly between two neighboring particles.The slower atomic diffusion in the distorted crystalline lattice of multielement alloy nanoparticles, which consist of elements with different atomic sizes, further hinders interatomic diffusion. [65,66]This complex compositioninduced slow interatomic diffusion serves as an additional obstacle to interparticle ripening and coalescence when particles are in close proximity.A schematic illustrating the proposed mechanism for the formation of secondary Pt-Cu-Ir-Ni nanoparticles is presented in Figure 8b.
Figure 8c provides a schematic overview of the overall nanoparticles formation process within the system.Initially, Au-rich Au-Cu particles (primary particles) form from the equimolar mixture of five metal salts in the solution.At the next stage, the mixture of metal salts which is nearly depleted of Au ions and enriched with the other four metal ions, serves as the precursor for the formation of secondary particles.While the secondary particles appear in the remaining precursor solution, the already formed and grown Au-Cu particles with sizes larger than 5 nm undergo further conversion into larger particles through coalescence events.Eventually, the final particles exhibit two distinct phases.
It is important to note that the electron beam dose rate can impact the nucleation and growth behavior of nanoparticles and alter the experimental condition.The effect of electron beam dose on the growth behaviors of nanoparticles can be substantial and is often studied in the context of electron microscopy techniques.Radical concentrations depend on dose rate which can control the reduction reaction condition and nucleation rate of nanoparticles.Changes in the electron dose can also alter the mass transfer rate to crystals and determine their growth mechanism, with diffusion-limited growth occurring at high electron doses and reaction-limited growth occurring at low electron doses. [69]he dose rate can also influence the final size of the nanoparticles by controlling the diffusion or attachment growth mode.At low radiation doses, an excess of unreduced metal ions compared to nuclei allows for the formation of larger ionic cluster before reduction and aggregation, resulting in larger nanoparticles.Conversely, higher doses consume most metal ions during nucleation, leading to smaller-sized nanoparticles. [70]This difference in size due to radiation dose variation applies to bimetallic nanoparticles as well, suggesting changes in nucleation and growth rates. [71,72]Moreover, radiolysis generates strong reducing and oxidizing agents whose relative concentrations determine whether growth or etching occurs.As the ratio of the reducing agent to the oxidizing agent rises with the dose rate, etching is anticipated at lower doses, while growth is expected at higher doses. [73]n bimetallic systems, dose rate can also impact on the chemical phase structure of the final nanoparticles.The formation of alloy or core-shell clusters depends on the competition between electron beam induced reduction and collision of atoms and inter-metal electron transfer rates.If both ions are incompletely reduced, electrons gradually transfer from the less reactive to the more reactive ion.This leads to the initial formation of a core with the more reactive metal, followed by the reduction of the less reactive metal ions on the core to form the shell.However, at high dose rates, complete reduction of all ions might occur before inter-metal electron transfer, favoring the formation of nanoalloys instead of core-shell clusters. [23]Generally, high dose rates favor alloying, while low dose rates promote the segregation of metals into core-shell structures within clusters.However, the specific dose rate determining nanoalloy formation instead of core-shell clusters varies significantly based on the combination of metals in the solution.Therefore, it not only depends on the dose rate but also the characteristics of elements and their interactions.For example, the alloying can occur even at low dose rate for binary systems like Au-Cu and Pt-Ni while some combination like Au-Ag and Ag-Pd require higher dose rate. [25]n STEM, dose control is straightforward, whereas in TEM, the number of electrons delivered to the sample is significantly higher.The electron dose rate can be manipulated by changing magnification, beam current, or accelerating voltage. [74]In our study, the accelerating voltage, beam current and magnification were maintained at 200 kV, 15 nA and X150K to avoid changing the electron dose rate.The estimated electron dose rate based on the current density (pA cm −2 ) measured at the phosphor screen was 300 electrons Å −2 s which is considered a high dose rate suitable for providing enough strong reducing agent concentration for a reduction and growth condition as well as alloy formation.Although the dose rate remains unchanged in our experiment, however, the cumulative dose of electron in solution is changing in the beginning of the irradiation also slightly during the entire process due the nature of flowing liquid stream inside the cell.Hence, this may also influence the induction time for the nucleation of different clusters and different growth behaviors.At relatively lower concentration of generated electrons in the beginning, similar to low dose conditions when there are excess unreduced ions, the formation of larger bimetallic ions before their complete reduction and aggregation in nuclei cluster can be a possible cause resulting in larger bimetallic nanoparticles.Additionally, the primary bimetallic clusters formed are composed of Au and Cu which in known to favor alloy structure even in low-dose condition.On the hand, with the continuous high dose flux of electron, concentration of reducing agent increases leading to sudden reduction of remaining metal ions and nucleation of small-size secondary clusters.

Conclusion
In this in situ study of colloidal nanoparticles synthesis from a precursor solution of mixed metal salts, we have elucidated the formation and growth mechanism of multielement alloy nanoparticles with composition and size variation.In the multimetal salt solution containing Au, Pt, Ir, Cu, and Ni elements, two diverse types of formation and growth pathways occur resulting in formation of two distinct sets of particles.The primary formed particles are 10-30 nm Au-Cu bimetallic alloy nanoparticles, and the secondary formed particles are 1-4 nm Pt-Cu-Ir-Ni multielement alloy nanoparticles.The Au-Cu particles were initially nucleated in solution.The individual nanoparticles detected with size > 2 nm grow rapidly up to 5-10 nm in size.The DFT calculation indicates a spontaneous reduction of Au and Cu ions can happen on the surface of Au-Cu nanoparticles and their swift enlargement in solution happens dominantly through the autocatalytic surface growth.Subsequently, coalescence events merge them into larger particles, ultimately reaching sizes of 10-30 nm.The multielement Pt-Cu-Ir-Ni nanoparticles, in contrast, nucleate more slowly and appear as secondary formed particles.Their non-classical nucleation pathway begins with intermediate atomic clusters that evolve into crystalline nanoparticles once they surpass 2 nm in size.Surface growth through autocatalytic surface reduction of ions is notably sluggish for these multielement alloys, resulting in minimal size increase for detected small clusters or nanoparticles.Furthermore, the particles' surfaces are likely to be covered with adsorbed charged ion complexes and nanoparticles appear agglomerated with amorphous interfaces, which impedes further growth by coalescence.The electrostatic interactions on charged surfaces are inclined to limit direct surface-to-surface contact of nanoparticles.Moreover, slow interatomic diffusion associated with the multielement structure may contribute to hindering their intraparticle ripening, merging, and recrystallization.In summary, this study offers valuable insights into the mechanisms that govern the formation and growth of multielement nanoparticles.Additionally, it sheds light on the factors responsible for the chemical separation of specific elements within the mixture and underscores how the elements and composition of the alloy influence the size and structure of the final nanoparticles.
In Situ Liquid Cell TEM Experiment: An equimolar aqueous mixture of HAuCl 4 .3H 2 O, H 2 PtCl 6 .xH 2 O, IrCl 3 .xH 2 O, CuCl 2 .2H 2 O and NiCl 2 .6H 2 O with 0.01 M concentration was used as the precursor solution.A solution of polyvinylpyrrolidone (PVP) was prepared as stabilizer and mixed with precursor solution at constant concentration of 0.5 wt.%.A Protochip Poseidon select liquid holder integrated with microfluidic tubing was used to inject the precursor solution into the liquid cell with the constant rate of 50 μl hr −1 (Figure 1a).The microchips were fitted with 50 nm thick electrontransparent silicon nitride (SiN x ) with 550 × 20 μm 2 viewing window and 150 nm liquid spacer.In situ TEM observation was carried out on a JEOL ARM200CF Microscope operating at 200 kV.Videos were acquired using an Orius SC200D CCD Gatan camera with 2048 × 2048 image size and 6 frames per second at magnification of ×150000.The estimated electron dose rate based on the current density (pA cm −2 ) measured at the phosphor screen was 300 electrons Å −2 s.
Image Analysis: Extracted frames from the in-situ captured videos were analyzed using ImageJ software. [67,68]TEM Characterization: The crystalline phase structure, élémental composition and distribution of the formed nanoparticles were characterized by HAADF-STEM-HAADF and STEM-EDS.At the end of the process liquid cell was opened, excess liquid were removed and the remaining formed particles on the surface of a single SiN chip were gently washed with DI water to remove all the excess Cl ions and other unreacted ions in solution.Afterwards the liquid cell chip was dried in the air before the assembly and insertion for STEM and EDS analysis of the formed particles.STEM-HAADF and STEM-LAAF imaging with 512 × 512 pixels scanning resolution and STEM-EDS analysis were performed at 8c probe size on a JEOL ARM200CF scanning electron microscope operating at 200 kV.
Density Functional Theory (DFT) Calculations: In this study, the Vienna Ab Initio Simulations Package (VASP) code was employed for conducting DFT calculations, utilizing the generalized-gradient approximation (GGA) and the PBE (Perdew, Burke, and Ernzerhof) functional to account for exchange-correlation effects.For systems with an odd number of electrons, non-spin-polarized calculations were performed, while for systems with even numbers of electrons, spin-polarized calculations were executed.A cutoff energy of 450 eV was selected for all calculations, as further increasing the cutoff energy only leads to minor changes in the ground state energy while significantly increasing the computational time.Therefore, 450 eV was considered the optimal value, striking a balance between ground state energy accuracy and computational efficiency.Structural optimizations continue until the forces acting on atoms were below 0.01 eV Å −1 , ensuring a thorough search for the most stable configurations.The criterion for energy change during optimization was set at 0.1 eV.To create the nanoparticles and slabs used in the calculations, the Atomistic Tool Kit (ATK) was employed, providing a reliable and efficient method for constructing these complex structures with high accuracy.ATK facilitates the generation of NPs and slabs, enabling precise control over their composition and atomic arrangement.

Figure 1 .
Figure 1.In situ formation of metallic colloidal nanoparticles by subjecting a solution containing mixed metal salts to electron beam irradiation within a liquid cell TEM setup.a) Schematic of flowing liquid cell TEM holder configuration.b) Schematic of the process of particle formation from a mixture of salts containing Au, Cu, Pt, Ir, and Ni metal cations inside the liquid cell under the electron beam.c) Series of TEM snapshots from Movie S1 (Supporting Information) showing the dynamic evolution of primary colloidal nanoparticles as they form and grow within the liquid solution.d) Series of TEM snapshots from Movie S2 (Supporting Information) showing dynamic formation and spread of a cloud-like network of secondary colloidal nanoparticles within the liquid solution.The scale bar is 50 nm.

Figure 2 .
Figure 2. The ADF-STEM characterization and EDS elemental composition analysis of in situ formed particles inside the liquid cell.a) HAADF-STEM image of a primary particle surrounded by secondary particles.b) STEM-EDS elemental mapping of the area selected in (a) showing a 25 nm size nanoparticle with Au-rich composition surrounded by aggregation of small multielement nanoparticles.The scale bar is 20 nm.c) LAADF-STEM image of a single primary particle formed in the liquid cell.d) STEM-EDS elemental mapping of the 25 nm nanoparticle in (c) showing the distribution of elements.The scale bar is 10 nm.e) The EDS spectra for quantitative analysis of the elemental atomic composition of the primary particle shown in (c) which specifies it as Au-rich binary Au-Cu alloy nanoparticle.f) HAADF-STEM image of agglomerated secondary nanoparticles with sizes of 1-4 nm.g) STEM-EDS elemental mapping of a selected area in (f) showing a homogeneous distribution of elements.The scale bar is 5 nm.h) The EDS spectra for quantitative analysis of elemental atomic composition of secondary particles shown in the selected area in (f) which specifies them as quaternary Pt-Cu-Ir-Ni alloy nanoparticles.

Figure 3 .
Figure 3.In situ TEM observation of the primary colloidal nanoparticles formation.a) TEM snapshots of the in-situ formation of primary nanoparticles at four selected time stamps (1.00, 1.67, 2.33, and 3.00 s) (The scale bar is 50 nm) and the enlarged images of selected areas (The scale bar is 20 nm).b) The plots showing the result of particle size (by diameter) and count analysis at four selected time stamps (1.00, 1.67, 2.67, and 3.00 s shown in (a)) during the growth of identified particles with the minimum size limit of 2 nm; and c) the plot showing the change in the total number of particles and their average sizes (by diameter) over 3 seconds of their formation (from the results shown in (b)).d) TEM snapshot series tracking the growth of seven individual particles marked as P 1 -P 7 in solution; and e) the plotted particle size (by diameter) over time for seven particles.

Figure 4 .
Figure 4. Crystal structure and size evolution of the primary particles at the final growth stage.a) TEM timelapse sequences tracking the coalescence of three particles and formation of a larger particle with near triangular morphology.The scale bar is 10 nm.b) HAADF-STEM images of four large primary Au-Cu nanoparticles.The scale bars are 10 nm.(The EDS mappings of particles are shown in Figure S3, Supporting Information).

Figure 5 .
Figure 5.In situ TEM observation of secondary colloidal nanoparticles formation.a) TEM snapshots showing the in-situ formation of secondary nanoparticles at four selected time stamps (1.00, 1.5, 2.33, and 3.00 s) (The scale bar is 50 nm) and the enlarged images of selected areas (The scale bar is 20 nm).b) HAADF-STEM image of the secondary particles in their dry state after the liquid-phase process was completed.The enlarged region highlights particle sizes within the range of 1-2 nm.c) HAADF-STEM images showing the initial stages of particles formation, starting from few-atom clusters, progressing to disordered atomic clusters, and ultimately converting into crystalline nanoparticles measuring 2 nm in size.

Figure 6 .
Figure 6.Crystal structure of the secondary particles in the final growth stage.a) High-magnification HAADF-STEM image of small secondary Pt-Cu-Ir-Ni nanoparticles.b) FFT pattern of all the particles in (a) with fcc crystalline structure.c) FFT pattern of a selected particle from (a) with fcc crystal structure at zone axis [001].d) FFT pattern of a selected particle from (a) with fcc crystal structure at zone axis.[01-1] e) HAADF-STEM image of multiple small crystalline nanoparticles aggregated with amorphous grain boundaries that prevent the direct surface-to-surface contact between nanoparticles.

Figure 7 .
Figure 7. DFT calculation results of precursors stability over Au-Cu nanoparticles a) and over the multi element b) consisting of Pt, Cu, Ir, and Ni nanoparticles.The left images show the initial structure, the middle images the optimized structure and the far-right images illustrate the charge density iso-surface of 0.3 eV Å -3 .The yellow spheres represent gold atoms, the orange spheres are copper atoms, the white spheres are platinum atoms, the dark blue spheres are iridium atoms, the green spheres are Ni atoms, and the blue spheres are chlorine atoms.

Figure 8 .
Figure 8. Schematic illustration of proposed mechanism for the formation of a) primary Au-Cu nanoparticles, b) secondary Pt-Cu-Ir-Ni nanoparticles and c) two different sets colloidal alloy nanoparticles by co-reduction in solution.