Insights into the Synthesis Mechanisms of Ag-Cu3P-GaP Multicomponent Nanoparticles

Metal–semiconductor nanoparticle heterostructures are exciting materials for photocatalytic applications. Phase and facet engineering are critical for designing highly efficient catalysts. Therefore, understanding processes occurring during the nanostructure synthesis is crucial to gain control over properties such as the surface and interface facets’ orientations, morphology, and crystal structure. However, the characterization of nanostructures after the synthesis makes clarifying their formation mechanisms nontrivial and sometimes even impossible. In this study, we used an environmental transmission electron microscope with an integrated metal–organic chemical vapor deposition system to enlighten fundamental dynamic processes during the Ag-Cu3P-GaP nanoparticle synthesis using Ag-Cu3P seed particles. Our results reveal that the GaP phase nucleated at the Cu3P surface, and growth proceeded via a topotactic reaction involving counter-diffusion of Cu+ and Ga3+ cations. After the initial GaP growth steps, the Ag and Cu3P phases formed specific interfaces with the GaP growth front. GaP growth proceeded by a similar mechanism observed for the nucleation involving the diffusion of Cu atoms through/along the Ag phase toward other regions, followed by the redeposition of Cu3P at a specific Cu3P crystal facet, not in contact with the GaP phase. The Ag phase was essential for this process by acting as a medium enabling the efficient transport of Cu atoms away from and, simultaneously, Ga atoms toward the GaP-Cu3P interface. This study shows that enlightening fundamental processes is critical for progress in synthesizing phase- and facet-engineered multicomponent nanoparticles with tailored properties for specific applications, including catalysis.

M ulticomponent nanostructures are promising materials for several applications, 1−5 including catalysis, 6 photovoltaics, 7 bioimaging, 8 biosensing, 9 and phototherapy. 10 For photocatalytic applications, combining semiconductors with metals 11 or other semiconductors 12 can result in efficient charge carrier generation and separation, which are crucial steps in the catalytic reaction. 13 Besides the present phases' identities, the facets forming the multicomponent photocatalytic materials' interfaces/surfaces are essential for their functionality. 14 Consequently, nanostructures' properties can be tailored through phase 15 and facet 16 engineering, enabling the creation of advanced materials.
Common approaches for synthesizing multicomponent photocatalysts are based on solution-based routes, usually including capping ligands. 12,14,17 Capping ligands can block active sites at the crystal surface facets that are consequently not accessible for reactants. 18 On the other hand, several reports identified the controlled ligand adsorption on the crystal surface facets as promising to enhance the catalysts' performances and selectivities. 19−22 Besides removing capping ligands by nanostructure treatment after the synthesis 23 or capping ligand-free solution-based approaches, 24 another strategy is synthesizing nanoparticles by gas-phase methods without applying capping ligands to obtain clean crystal surface facets. 25−27 Those approaches (will) help evaluate the capping ligands' impact on the catalysts' performances by creating capping ligand-free starting points. 28 Although controlling the faceting of catalysts is challenging without capping ligands, progress in this field is needed. 29 Moreover, understanding the formation mechanisms of multicomponent nanostructures is crucial for gaining control over the synthesis, which enables designing phases and surface/interface facets. The formation mechanisms are usually reconstructed by characterizing the products after the synthesis. However, the indirect determination of the formation mechanisms involves the risk of misinterpreting the results or, at the very least, making interpretations more challenging.
Therefore, in situ approaches, including the atomic-scale visualization of nanostructures during the entire synthesis process, are essential to progress in this field. 30−34 Moreover, catalysts' surface alterations have been observed due to the interactions of catalysts with different environments making in situ characterizations crucial for understanding the actual contribution of the designed facets to the catalysts' performances. 35−41 During the last decades, the metal-assisted growth of semiconductor nanostructures has developed as a promising tool enabling high control over crystal properties, including crystal structure, 42−44 composition, 45−47 growth direction, 48−50 morphology, 48,51,52 surface faceting, 53−55 doping profile, 56−58 and defect structure. 59−61 The seed material's role was limited to the capability to nucleate and selectively grow the semiconductor nanocrystal. 62−64 Few reports considered the seed material as an active component of the synthesized nanostructure. 65−68 We propose that the metal-assisted growth approach is promising for synthesizing advanced multicomponent nanostructures. Therefore, a deeper understanding of the involved formation mechanisms is essential to exploit its full potential.
Metal phosphides' potential as photocatalysts has been discussed in the literature, 69−71 and heterostructures using earth-abundant metal phosphides, such as GaP-Ni 2 P 72 and Cu 3 P-Au, 73 have been investigated for their performances in photocatalytic water splitting. Moreover, metal cocatalysts, including Ag, can enhance catalysts' performances by plasmon−exciton coupling. 74−76 Therefore, this study focused on combining GaP, Cu 3 P, and Ag in a multicomponent nanostructure.
In a gas-phase synthesis approach, we enlightened the dynamic processes following the supply of Ga-and Pcontaining precursors to Ag-Cu 3 P nanoparticles. The experiments were performed inside an environmental transmission electron microscope (TEM) with an integrated metal−organic chemical vapor deposition (MOCVD) system. The dynamic processes were captured by high-resolution transmission electron microscopy (HRTEM) images/movies. A detailed data analysis revealed the fundamental processes involved in GaP nucleation and growth using Ag-Cu 3 P nanoparticles as seed materials.
A topotactic reaction initiated the GaP nucleation at the Cu 3 P phase's surface, leading to rearrangement processes in the multicomponent nanoparticle. As a result, the phases connected and formed a stable GaP growth front. Counterdiffusion of Cu + and Ga 3+ cations was involved in the growth process, with the Ag phase acting as a transport medium for Ga and Cu atoms. The formation and alteration of interfaces were critical steps in synthesizing Ag-Cu 3 P-GaP nanoparticles. The presented results will allow for a deeper understanding of the fundamental mechanisms involved in synthesizing complex multicomponent nanoparticles and help progress in designing nanomaterials with applications in different fields, including catalysis.

RESULTS AND DISCUSSION
Ag-Cu 3 P Nanoparticle Synthesis. The sample was prepared by depositing Ag-Cu nanoparticles with diameters of ∼30 nm on a microelectromechanical system (MEMS)based heating chip using a home-built spark ablation system to generate and select nanoparticles with specific mobility diameters. 26 Subsequently, the sample was transferred to the environmental TEM. Inside the environmental TEM, the sample temperature was increased to 420°C, and phosphine (PH 3 ) was supplied to initiate a selective chemical reaction with the Cu phase to form Ag-Cu 3 P nanoparticles ( Figure 1).
We chose a slightly adapted Ag-Cu 3 P nanoparticle synthesis approach to the recently reported procedure by our group, 27 excluding the pretreatment under a H 2 atmosphere. A Ag-Cu 3 P nanoparticle with both phases oriented in one of their zone axes and the heterointerface aligned parallel to the direction of the electron beam was selected for further investigations (Figure 1a).
The HRTEM image in Figure 1a was acquired at 90 s before trimethylgallium (TMGa) was added to the precursor supply. The power spectrum corresponding to the HRTEM image in Figure 1a was overlaid with matching simulated electron diffraction patterns of the Ag (red, space group: Fm3m) 77 and Cu 3 P (blue, space group: P6 3 cm) 78 phases ( Figure 1b). Power spectra of the phases are shown as insets in Figure 1b and allow associating the Ag and Cu 3 P phases to specific regions in the HRTEM image in Figure 1a, highlighted by color-coded labels.
It is worth mentioning that the selected Ag-Cu 3 P nanoparticle's interface was formed by Ag{110} and Cu 3 P{1120} facets. Recently, we reported Ag-Cu nanoparticles with Ag{111}/Cu{111} interfaces acting as templates stabilizing Ag-Cu 3 P nanoparticles with interfaces formed by Ag{111} and Cu 3 P{1010} facets after initiating the chemical reaction via the supply of PH 3 . 27 Under certain conditions, such a nanoparticle had the potential to rearrange into an Ag-Cu 3 P nanoparticle with the thermodynamically more favorable interface formed by Ag{110} and Cu 3 P{1120} facets.
The adapted synthesis procedure, lacking the pretreatment under an H 2 atmosphere, could have led to supplying PH 3 to Ag-Cu nanoparticles with rough heterointerfaces. 27 That circumstance could have caused the formation of the thermodynamically more favorable Ag{110}/Cu 3 P{1120} interface in the Ag-Cu 3 P nanoparticle presented in Figure 1. However, this study does not focus on the statistical analysis of Ag-Cu 3 P nanoparticle interfaces formed by the supply of PH 3 to Ag-Cu nanoparticles with rough heterointerfaces. Instead, this observation suggests potential approaches to further develop facet engineering in Ag-Cu 3 P nanoparticles via the presented procedure.
Alterations in the Ag Phase Caused by Adding TMGa to the PH 3 Supply. Subsequently, TMGa was added to the precursor supply (timestamp: 0 s), leading to Ag-Cu 3 P nanoparticle alterations ( Figure 2). The HRTEM image of the Ag-Cu 3 P nanoparticle (timestamp: 120 s) in Figure 2a was overlaid with the Ag phase's outline (red dotted line) obtained by the HRTEM image in Figure 1a. Comparing the Ag phases' projected shapes before and after adding TMGa to the PH 3 supply suggests a considerable impact on the Ag phase. Although the area of the Ag phase's projected shape increased significantly, conclusions regarding the Ag phase's volume should be carefully drawn when working with two-dimensional projections of nanostructures. Simultaneously, no substantial changes were observed in the Ag phases' power spectra and the Cu 3 P crystal.
Such alterations in the power spectra could be expected if Ga atoms accumulated in the Ag phase. However, a detailed analysis in Figure S1, comparing the Ag phases' power spectra before and after adding TMGa to the PH 3 supply, reveals no significant Ag spot alterations. Moreover, calculations show that such alterations are not expected for this specific system due to a minor decrease in the unit cell volume upon increasing the Ga concentration in the Ag phase. 79 Therefore, the total volume of the Ag phase by adding Ga atoms to a fixed number of Ag atoms could increase without any significant alterations in the power spectrum. Based on the observed dynamic processes, the accumulation of Ga atoms in the Ag phase accompanied by increasing its total volume without significantly altering the volume of its unit cell was identified as the most probable process.
Subsequently, the Ag phase's stepwise surface alteration was observed (Figure 2b−f). The white arrows in Figure 2b indicate the regions in the HRTEM image where the first Ag surface changes occurred (timestamp: 125 s). The surface changes were initiated at the Ag-Cu 3 P nanoparticle's triplephase boundary. Therefore, the zoomed-in HRTEM image of the addressed area raises the question of the type of atom/ atoms that was/were located at the Ag phase's surface ( Figure  2c).
The surface alterations were also observed in other regions of the Ag phase, indicated by white arrows in Figure 2d, and started at the Ag-Cu 3 P nanoparticle's triple-phase boundary ( Figure 2d). The associated zoomed-in HRTEM image reveals that the surface atoms formed similar patterns to those observed in Figure 2c ( Figure 2e). Finally, the surface alterations expanded to other regions of the Ag phase indicated by the white arrow in Figure 2f.
GaP Nucleation. The TMGa supply was subsequently increased while keeping the PH 3 supply constant to facilitate nucleation of the GaP phase (timestamp: 262 s, Figure 3). Initially, no changes were observed in the Ag-Cu 3 P nanoparticle, with the altered Ag phase's surface (timestamp: 332 s, Figure 3a). However, GaP nucleation occurred approximately at 72 s after increasing the TMGa supply (timestamp: 334 s, Figure 3b). The GaP nucleus formed at the Cu 3 P phase's surface and replaced a part of the Cu 3 P crystal.
The GaP nucleus was initially not in contact with the Ag phase. However, it expanded along its [112] direction (see GaP phase's power spectrum, Figure 3f) until the three active phases were in contact with each other (timestamp: 345 s, Figure 3c). The HRTEM image in Figure 3c was overlaid with the Cu 3 P phase's blue dotted outline before the GaP nucleation ( Figure 3a), revealing that the GaP nucleation was accompanied by the removal of Cu 3 P from the GaP-Cu 3 P interface and a growing Cu 3 P crystal in the regions outside the marked area. The white arrow in Figure 3c indicates a step in the Cu 3 P crystal moving toward the Ag-Cu 3 P-GaP triple-phase boundary as the GaP crystal expanded.
Further GaP growth led to the formation of a new interface between the Ag and GaP phases, as indicated by the white arrow in Figure 3d (timestamp: 403 s). Nevertheless, the region highlighted by the black arrow in Figure 3d reveals that the GaP growth front evolution was not completed. A stable growth front was formed at approximately 412 s after adding TMGa to the precursor flow ( Figure 3e).
The power spectrum corresponding to the HRTEM image in Figure 3e was overlaid with simulated electron diffraction patterns of the Ag (red), Cu 3 P (blue), and GaP (orange, space group: F43m) 80 phases (Figure 3f). GaP{111}/Ag{110}, GaP{111}/Cu 3 P{0001}, and Ag{110}/Cu 3 P{1120} interfaces were identified as the primary interfaces in the observed multicomponent nanostructure. Moreover, an additional Ag-Cu 3 P interface was observed directly at the triple-phase boundary between the Ag, Cu 3 P, and GaP phases. The power spectrum corresponding to the HRTEM image in (a) was overlaid with matching simulated electron diffraction patterns of the Ag (red, space group: Fm3m) 77 and Cu 3 P (blue, space group: P6 3 cm) 78 phases. The Ag (red, bottom left) and Cu 3 P (blue, top right) phases' power spectra are revealed with adapted brightness and contrast as insets in (b).
GaP Growth. As the GaP crystal growth continued along its [111] direction, the Cu 3 P facet indicated by the black arrow in Figure 3e disappeared stepwise, and the additional Ag-Cu 3 P interface's area increased, initiating phase rearrangements at the growth front ( Figure 4). The Ag phase formed a differently oriented interface with the GaP crystal, indicated by the white arrow in Figure 4a (timestamp: 482 s). GaP growth continued at the two differently oriented GaP growth fronts, and nanofacets appeared at the Ag-Cu 3 P interface (timestamp: 744 s, Figure 4b).
The zoomed-in area of the HRTEM image indicated by the white arrow in Figure 4b reveals the atomic structure at the specific heterointerfaces (Figure 4c). While the GaP-Ag interface changed, the GaP-Cu 3 P interface formed by GaP{111} and Cu 3 P{0001} facets remained unaltered. A closer look at the GaP-Ag interface change reveals that a GaP twin was formed as part of the phase rearrangements. The orange line indicates the twin's mirror plane, and the orange dotted lines highlight the mirrored GaP atomic columns in Figure 4c.
Moreover, the Ag-Cu 3 P interface changed significantly by developing local Ag{110}/Cu 3 P{1120} interfaces, exemplarily highlighted by the white line in Figure 4c. The local Ag-Cu 3 P interfaces were interrupted by crystal steps. This observation could have been the direct consequence of the GaP twin formation leading to a slight rotation of the Ag phase relative to the Cu 3 P phase. Consequently, the steps at the Ag-Cu 3 P interface could hint toward dislocations accommodating the strain originating from the phases' mismatch already observed in a previous study. 27 It is worth mentioning that a moirépattern was visible in the Ag phase in Figure 4b due to an overlap of the Ag and Cu 3 P phases. The moirépattern expanded from the Ag-Cu 3 P interface over a larger area (timestamp: 918 s, Figure 4d). This observation suggests that the phases formed additional heterointerfaces with differently oriented facets. Moreover, the GaP-Ag interface continuously changed and was no longer oriented parallel to the direction of the electron beam. Nevertheless, the addressed processes led to the accommodation of strain indicated by the continuous removal of crystal steps at the Ag-Cu 3 P interface highlighted by the white arrow in Figure 4d.
After the removal of the crystal steps at the Ag-Cu 3 P interface, again, an additional Ag-Cu 3 P interface indicated by the white arrow in Figure 4e remained beside the primary Ag{110}/Cu 3 P{1120} interface (timestamp: 1081 s). The  Figure 4f) and reacted with the supplied P atoms to form Cu 3 P layers. The area of the Cu 3 P crystal's top facet decreased during this process. Therefore, other available Cu 3 P crystal facets grew, leading to the moirépatterns observed in Figure  4e,f. The power spectrum of the HRTEM image region marked by the black rectangle in Figure 4f was overlaid with matching simulated electron diffraction patterns of the Ag and Cu 3 P phases (see inset in Figure 4f). These results confirm the growth of the Cu 3 P crystal in regions other than the addressed top facet. Moreover, additional reflections that could not be associated with a specific phase were observed. Since the Cu 3 P phase was in contact with the twinned GaP segment, the additional reflections likely corresponded to a differently oriented Cu 3 P crystal. A detailed analysis of regions with overlapping crystals causing moirépatterns can be found in Figure S2.
Ag Phase's Surface Alterations. Now, the dynamic processes observed during the experiment and discussed above will be used to enlighten the Ag phase's surface alterations and the mechanisms of GaP nucleation and growth at an atomic scale ( Figure 5). Changes in the Ag phase's surface were observed after adding TMGa to the PH 3 supply. Based on our observations, we conclude that Ga accumulation in the Ag phase was the most probable initial step causing the formation of surface layers. However, since we could not directly prove the enrichment of the Ag phase with Ga atoms by complementary methods, alternative scenarios leading to this observation are added to the discussion (Figure 5a).
In the first scenario, the Ag phase's oversaturation with Ga atoms could have led to the diffusion of Ga atoms toward the Ag phase's surface and a subsequent chemical reaction with P atoms supplied by the gas phase (mechanism I, Figure 5a). In the second scenario, the Ag phase could have mediated the Cu 3 P phase's partial dissolution at the Ag-Cu 3 P interface. Subsequently, the Cu 3 P compound could have been redeposited at the Ag phase's surface (mechanism II, Figure  5a). In the third scenario, Cu and P atoms could have diffused along the surface across the heterointerface and covered the Ag phase's surface (mechanism III, Figure 5a).
In the literature, ordered layers were observed at liquid− solid, 81 liquid−vapor, 82 and solid−vapor 40 interfaces. The formation mechanisms included the supersaturation of a phase by specific atomic species segregating at the phase's surface 82 and the surface diffusion of atomic species from one phase to another. 40 However, it is worth mentioning that the methods used in this study did not allow the clarification of the exact formation mechanism.
Consequently, complementary methods are essential for studying dynamic processes at an atomic scale. One possible (but challenging) way to gain a deeper understanding of the altered Ag phase's surface might be the determination of its chemical composition using high-resolution scanning trans-mission electron microscopy (STEM)-electron energy loss spectroscopy (EELS). Moreover, complementary theoretical modeling could be used to compare the suggested formation mechanisms and determine the likelihood of their occurrences. 83 Nucleation Mechanism. In contrast to the above case, a more detailed analysis of the diffusion processes was possible for the GaP nucleation. The combination of observations, including the GaP nucleus formation at the Cu 3 P phase's surface, its growth toward the Ag-Cu 3 P interface by replacing the Cu 3 P phase, and alterations in the Ag phase upon adding TMGa to the PH 3 supply, suggests that counter-diffusion of Cu + and Ga 3+ cations and redox reactions were involved in the GaP nucleation and growth (Figure 5b). , were observed at the Ag-Cu 3 P interface. Moreover, GaP twinning is highlighted by the mirror plane (orange line) and the mirrored GaP atomic columns (orange dotted lines). (d) The Cu 3 P crystal was removed layer-by-layer at the Ag-Cu 3 P interface, highlighted by the white arrow showing a step. (e) At the Ag-Cu 3 P-GaP triple-phase boundary, the orientation of the Cu 3 P facet deviated from the Cu 3 P facet forming the primary interface with the Ag phase. (f) The alteration of the Ag-Cu 3 P interface continued with the removal of Cu 3 P layers along the Ag-Cu 3 P interface. The step at the Ag-Cu 3 P interface highlighted by the white arrow confirms the discussed process. Since the removed Cu atoms could not attach to the shrinking top facet of the Cu 3 P crystal indicated by the black arrow, Cu 3 P growth continued in other regions. The Cu 3 P phase's expansion already observed in (b) became more pronounced, leading to overlaps of the Ag and GaP phases with the Cu 3 P phase. The power spectrum of the area marked by the black rectangle in the HRTEM image (adapted brightness and contrast) was overlaid with matching simulated electron diffraction patterns of the Ag and Cu 3 P phases and is shown as an inset in (f). Moreover, additional reflections were observed but could not be allocated to a specific phase.
The supply of atomic species from specific regions within the multicomponent nanoparticle (e.g., P atoms from the Cu 3 P phase) and the propagation of the GaP growth front indicate the following chemical reaction as the driving force for the observed diffusion processes: Cu P Ga 3Cu GaP 3 + + Therefore, Cu + ions must have diffused from the GaP-Cu 3 P interface toward the Ag-Cu 3 P interface. At the Ag-Cu 3 P interface, the Ga atoms supplied directly by the Ag phase or through diffusion along the Ag-Cu 3 P interface reduced the Cu + cations by being oxidized to Ga 3+ cations. Simultaneously, the resulting Ga 3+ cations diffused toward the GaP-Cu 3 P interface, reacting with the P 3− anions of the Cu 3 P phase's anion sublattice. The Cu 3 P phase's crystal structure was essential to facilitate this process and matched well with the GaP phase's crystal structure (see Figure S3). Consequently, the discussed process can be described as a topotactic reaction.
A similar observation was reported in the literature showing the (partial) replacement of the Cu 3 P phase by GaP via a cation exchange reaction in a solution-based process. 84 The high concentration of Cu vacancies in Cu 3 P nanocrystals was identified as essential for their capability to perform cation exchange reactions. 85 The same approach, using Cu 3 P nanocrystals as starting materials, was applied to synthesize InP 84−87 and In 1−x Ga x P 84 nanoparticles. Here, instead of a solution containing Ga 3+ cations, the Ag phase mediated the supply of Ga atoms.
The above-mentioned studies reveal the stabilization of III− V compounds, including GaP, 84 in their metastable wurtzite polytype. In this study, GaP nuclei with the zincblende crystal structure were synthesized. The formation of stacking faults in the GaP nucleus (see Figure 3b) could have been caused by the slightly different formation mechanism and a potential minor mismatch between the phases at the heterointerface (see Figure S3). After the Ag and Cu 3 P phases were rearranged at the GaP growth front, changing the growth mechanism (see Figure 3d), no stacking faults were observed in the associated GaP segment.
Consequently, the choice of synthesis parameters, such as the temperature, could impact the stabilization of specific III− V semiconductor polytypes. Moreover, it is worth mentioning that literature reported the presence of interfaces consisting of Cu 3 P{1120} and InP{1100} facets when Cu 3 P crystals were partly replaced by wurtzite InP segments in cation exchange reactions. 85 Although, in the here-presented study, the observed GaP-Cu 3 P growth front also involved a Cu 3 P{1120} facet during the GaP nucleation, another GaP-Cu 3 P interface involving a Cu 3 P{0001} facet formed (see Figure 3b). This additional heterointerface could have prevented stabilizing the metastable wurtzite polytype. Therefore, controlling the heterointerfaces evolving during the nucleation step by synthesizing defined Ag-Cu 3 P seed particles could be essential for crystal structure engineering via the here-presented approach.
Finally, it is worth mentioning that the observed dynamic processes and formation mechanisms correspond to a specific set of synthesis parameters. Therefore, temperature, partial pressure, and carrier gas variations will likely result in alternative synthesis mechanisms. In particular, under certain conditions, the Ag phase could initiate GaP nucleation at the Ag-Cu 3 P-gas triple-phase boundary, similar to the reported Agassisted nucleation of GaP nanowires on a substrate. 88 Growth Mechanism. The growth mechanism with associated diffusion pathways could also be clarified after the initial GaP nucleation step (Figure 5c). In particular, similar to the GaP nucleation discussed above, the chemical reaction 1 was indirectly observed at the Ag-Cu 3 P-GaP triple-phase boundary.
The stepwise replacement of Cu 3 P by GaP evolved from the Ag-Cu 3 P-GaP triple-phase boundary. It led to crystal steps at the GaP-Cu 3 P interface since Cu + and Ga 3+ cations had to be interchanged along the GaP-Cu 3 P interface, requiring more time for Cu + cations farther away from the triple-phase boundary (see Figure 3e). Simultaneously, the Cu + cations that were reduced to Cu atoms at the triple-phase boundary diffused through/along the Ag phase to a thermodynamically favorable Cu 3 P facet at the Ag-Cu 3 P-gas triple-phase boundary.  Figure 3b, and (c) Figure 3e are shown to illustrate the nucleation and growth mechanisms as well as the underlying potential diffusion pathways. (a) The reaction of Ga atoms with supplied P atoms (mechanism I), the Ag-mediated dissolution and redeposition of Cu 3 P (mechanism II), and the surface diffusion of Cu and P atoms across the Ag-Cu 3 P interface (mechanism III) could have caused the observed Ag phase's surface alterations. (b) The GaP nucleation and growth required exchanging Cu + and Ga 3+ cations through the Cu 3 P phase and a redox reaction at the Ag-Cu 3 P interface. (c) After forming a stable GaP growth front, a similar process was observed involving the exchange of Cu + and Ga 3+ cations along the GaP-Cu 3 P interface. Moreover, Cu 3 P was redeposited at a thermodynamically favorable Cu 3 P facet in contact with the Ag and gas phases. The Ag phase acted as a diffusion medium for the Cu atoms, and the P atoms supplied via the gas phase led to a chemical reaction at the Ag-Cu 3 P-gas triple-phase boundary.
The Cu atoms then reacted with the supplied P atoms to form Cu 3 P, which was redeposited at the addressed Cu 3 P facet.
The growth mechanism and diffusion pathways became more complex after the Cu 3 P facet, indicated by the black arrow in Figure 3e, disappeared, and a GaP twin was observed. Likely, the twinning of the GaP phase led to the formation of a differently oriented Cu 3 P phase segment, which made the analysis nontrivial. Nevertheless, it can be expected that after the GaP twinning, the chemical reaction 1 and the redeposition of Cu 3 P mediated by the Ag phase were essential steps to facilitate the GaP growth.

CONCLUSIONS
The results herein enlighten the fundamental processes involved in synthesizing Ag-Cu 3 P-GaP nanoparticles by supplying Ga-and P-containing precursors to Ag-Cu 3 P seed particles. Straight after supplying both precursors, dynamic processes were observed solely in the Ag phase, including the formation of ordered layers at the Ag phase's surface.
Subsequently, GaP nucleated at the Cu 3 P phase's surface and grew toward the Ag-Cu 3 P interface. We identified the counter-diffusion of Cu + and Ga 3+ cations through the Cu 3 P phase, facilitated by the high Cu vacancy concentration in the Cu 3 P phase, as an essential process enabling GaP nucleation. After the Ag and Cu 3 P phases were rearranged at the GaP growth front, similar mechanisms were involved in promoting GaP growth at the GaP-Cu 3 P interface. Cu atoms accumulated in/at the Ag phase and diffused to the Ag-Cu 3 P-gas triplephase boundary where the Cu atoms reacted with supplied P atoms forming Cu 3 P at a specific Cu 3 P crystal facet, not in contact with the GaP phase.
After the addressed Cu 3 P crystal facet's area decreased significantly, Cu 3 P growth continued in other regions, indicated by moirépatterns, and additional interfaces between the phases were formed. Consequently, further GaP growth became complex and involved significant Ag-Cu 3 P interface reconstructions. However, despite the slightly different mechanisms for GaP nucleation and growth, the Cu 3 P acting as a P reservoir and the Ag phase mediating the diffusion of Ga and Cu atoms were essential for synthesizing Ag-Cu 3 P-GaP multicomponent nanoparticles.
This study enlightens the fundamental formation mechanisms of multicomponent nanoparticles. Epitaxial relationships, diffusion of cations/atoms, and redox reactions at interfaces have been identified as essential aspects of the synthesis that need to be considered when aiming to control multicomponent nanoparticles' properties. The findings of this study will help create interface facet-engineered multicomponent nanoparticles with potential in catalysis.

Ag-Cu Nanoparticle Synthesis and Deposition on MEMS-Based Heating Chips.
Ag-Cu agglomerates were generated in a home-built spark ablation system. 89 In the reactor, a spark between a Ag anode and a Cu cathode ablated material from both electrodes. A N 2 /H 2 carrier gas was used to transport the formed agglomerates from the reactor through a furnace kept at 850°C to sinter the agglomerates into spherical, phase-segregated nanoparticles. A differential mobility analyzer system allowed sintered bimetallic nanoparticles with a selected mobility diameter of 30 nm to remain in the gas stream. The selected bimetallic nanoparticles were deposited on a MEMS-based heating chip by an electrostatic precipitator. A detailed process description, including relevant parameters, can be found in a recently published study. 26

MEMS-Based Heating Chips for in Situ TEM Investigations.
MEMS-based chips with a thin SiN x membrane and an embedded W coil for Joule heating from Norcada Inc. were used for the in situ TEM experiments. The SiN x membrane contained 19 regions with a few nm thick SiN x layer and holes in their center. The chips allowed sample heating up to 1100°C with a homogeneous temperature profile along the central region of the chip. The temperature variation was performed in constant resistance mode and controlled by the Blaze software supplied by Hitachi High-Technologies.
Environmental TEM and HRTEM Image/Movie Acquisition. A Hitachi HF-3300S environmental TEM operated at 300 kV was used to perform the experiments. The microscope was equipped with a cold field emission gun as the electron source, an imaging aberration corrector (CEOS BCOR), and a complementary metal−oxide− semiconductor (CMOS) camera (Gatan OneView IS camera). Electron dose rates of ∼1500 e/Å 2 s to 4400 e/Å 2 s were used to acquire HRTEM images and movies (see Tables S1 and S2). The background pressure next to the sample was ∼9.20 × 10 −4 Pa. Detailed information about the used environmental TEM and its capabilities are available elsewhere. 90 Sample Holder and Precursor Supply. A custom-built double tilt holder from Hitachi was used to perform the experiments. The gas handling system enabled the controlled supply of PH 3 and TMGa via a setup similar to the ones used in commercially available MOCVD systems. It is worth mentioning that PH 3 was supplied in its pure form. In contrast, TMGa was stored in a bubbler kept at −20°C. Its condensed phase was in equilibrium with the gas phase, which was transported from the bubbler via the H 2 carrier gas to the gas handling system. The TMGa/H 2 mixture was further diluted with H 2 before entering the microscope. The precursor flows were supplied directly to the sample via separate lines of a side port injector integrated into the microscope column. The chemicals used in this study, including PH 3 (flammable, toxic) and TMGa (pyrophoric), must be handled with care and appropriate equipment.
Process Parameters for the Experiment. The first step of the approach to nucleate GaP using bimetallic seed particles was forming Ag-Cu 3 P nanoparticles. The Ag-Cu nanoparticles were not pretreated under an H 2 atmosphere at elevated temperatures, as previously reported by our group. 27 The adapted procedure is summarized in Table S3. The GaP nucleation was initiated at a temperature of 420°C by adding TMGa to the PH 3 supply. The TMGa supply was increased stepwise to achieve GaP nucleation. After the nucleation event, the TMGa supply was decreased significantly to reduce the GaP growth rate. The PH 3 supply was kept constant during the whole experiment. The process parameters used for the GaP nucleation and growth by Ag-Cu 3 P seed particles are summarized in Tables S4 and S5. The total and partial pressures were estimated based on a recently published study. 90 The graph in Figure S4 reveals the variation of process conditions over time and specific experiment events.
Simulation of Electron Diffraction Patterns. Power spectra were loaded in the SingleCrystal software, and the corresponding scale bars were used to set the camera lengths for the electron diffraction pattern simulations. The cif files were downloaded from the Inorganic Crystal Structure Database. The parameters for the SingleCrystal simulations are summarized in Table S6. The simulations did not consider the phases' thermal expansions due to the lack of the Cu 3 P phase's thermal expansion coefficient in the literature. It is worth mentioning that electron diffraction patterns do not contain exactly the same information as HRTEM images' power spectra. Still, overlaying them is an adequate method to confirm the presence of specific phases.
Data Processing/Software. DigitalMicrograph from Gatan (version 3.50.3584.0) with the integrated in situ player was used to acquire and process HRTEM images and movies. Tables S7 and S8 summarize the processing parameters for extracting HRTEM images/ movies from the HRTEM movie raw data. HRTEM movies in the Supporting Information were compressed and labeled (scale bars and timestamps) using ImageJ (version 1.53p). The power spectra were calculated using DigitalMicrograph and loaded in SingleCrystal from CrystalMaker Software Ltd. (version 4.1.6) to overlay them with simulated electron diffraction patterns. The figures were prepared using Adobe Photoshop from Adobe (version 22.1.0).

* sı Supporting Information
The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsnano.3c00140. Figures S1−S4 show a comparison between the Ag phases' power spectra before and after adding TMGa to the PH 3 supply, an analysis of the observed moireṕ atterns, the crystal structure matching between the GaP and Cu 3 P phases at their interface, and a graph that reveals the variation of process conditions over time and specific experiment events; Tables S1−S8 contain information about the acquisition parameters for the HRTEM images/movies, the process parameters for the formation of Ag-Cu 3 P nanoparticles by Ag-Cu seed particles, the process parameters for the GaP nucleation and growth by Ag-Cu 3 P seed particles, the parameters for the electron diffraction pattern simulations, and the processing parameters for extracting HRTEM images/ movies from the HRTEM movie raw data (PDF) Movies S1: Ag phase's surface alterations after adding TMGa to the PH 3 supply (AVI) Movie S2: GaP nucleation and growth after increasing the TMGa supply (AVI)