Direct Observation of Liquid–Solid Two‐Phase Seed Particle‐Assisted Kinking in GaP Nanowire Growth

In the last decades, the metal‐assisted growth approach of semiconductor nanowires (NWs) has shown its potential in controlling crystal properties, such as crystal structure, composition, and morphology. Recently, literature reports have shown successful semiconductor NW growth with multiphase seed particles under growth conditions. Exploring alternative metal seeds and the mechanisms for growing semiconductor NWs is an exciting research field aiming to improve the control over the crystal growth process. Herein, the gallium phosphide (GaP) NW growth using Cu as seed particles inside an environmental transmission electron microscope is studied. In particular, the transformations of the Cu‐rich seed particles during the nucleation and growth of GaP NWs are observed. The supply of a relatively high amount of Ga atoms by the precursor mixture led to a solid Cu‐rich seed particle core covered by a liquid phase. Different growth dynamics within the two‐phase seed particle resulted in local competition in NW growth. As a result, the GaP NW kinked into another growth direction by forming a new interface at the NW growth front. The generated results enable insights into fundamental processes occurring in the seed particle during growth, creating leverage points for controlling the NW morphology.


Introduction
[3][4] The main difference between the addressed growth mechanisms is the state of the seed particle under growth conditions: liquid during VLS growth and solid during VSS growth of NWs.10] Therefore, looking thoroughly into the VLS and VSS growth dynamics, preferably in the same system, can be beneficial for precisely controlling the crystal structure, composition, and morphology in complex NWs and NW heterostructures.
[13] However, the high solubility of growth species in liquid Au droplets leads to the addressed reservoir effect and limits the formation of abrupt interfaces, e.g., in the Si-Ge system. [14,15]22][23][24][25][26][27][28][29][30] In the last decades, the metal-assisted growth approach of semiconductor nanowires (NWs) has shown its potential in controlling crystal properties, such as crystal structure, composition, and morphology.Recently, literature reports have shown successful semiconductor NW growth with multiphase seed particles under growth conditions.Exploring alternative metal seeds and the mechanisms for growing semiconductor NWs is an exciting research field aiming to improve the control over the crystal growth process.Herein, the gallium phosphide (GaP) NW growth using Cu as seed particles inside an environmental transmission electron microscope is studied.In particular, the transformations of the Cu-rich seed particles during the nucleation and growth of GaP NWs are observed.The supply of a relatively high amount of Ga atoms by the precursor mixture led to a solid Cu-rich seed particle core covered by a liquid phase.Different growth dynamics within the two-phase seed particle resulted in local competition in NW growth.As a result, the GaP NW kinked into another growth direction by forming a new interface at the NW growth front.The generated results enable insights into fundamental processes occurring in the seed particle during growth, creating leverage points for controlling the NW morphology.
It has been reported that Cu is a viable alternative to Au for the growth of III-V semiconductors via the VLS and VSS mechanisms. [25,31]Cu is expected to have a lower negative impact on the electronic properties of devices and is earth-abundant, in contrast to Au. [32,33] Generally, the Cu-group III element systems reveal higher eutectic temperatures than the corresponding Au-group III element systems. [34]Therefore, an extended temperature range for the growth of specific III-V semiconductor NWs via a solid Cu-based seed particle exists compared to Au.Moreover, Cu has also been reported to form more attainable polytypes of semiconductor NWs than Au, such as 4 H crystal structures. [35]ecently, a combination of the VLS and VSS mechanisms, the so-called liquid-assisted VSS mechanism, has been reported in the literature for growing semiconductor NWs. [36,37]In one of the addressed studies, seed particles consisting of solid Cu 3 Si and liquid Sn phases were used for growing Si NWs. [36]The morphology and structure of the Si NWs could be tailored by adjusting the Cu/Sn ratio.A seed particle consisting of a solid core surrounded by a liquid alloy was observed in another study about ZnTe NW growth. [37]The solid core's movement correlated with the step flow at the interface of the ZnTe NW and the liquid phase.Therefore, the liquid-assisted VSS mechanism could alter NW properties, such as the crystal structure, defect structure, and morphology.
One aspect relevant to this study is the kinking of NWs. [38][52] However, only very few studies have involved a detailed investigation of NW kinking going beyond simulations or conventional microscopy. [49]In particular, the relationship between particle phase and kinking behavior is not yet understood.Therefore, exploring alternative seed materials, growth modes, and ways to manipulate the NW growth is essential to progress in the formation of complex nanostructures with advanced functionalities. [39,41,43]n this work, we used Cu as seed material to promote gallium phosphide (GaP) NW growth and investigated kinking processes occurring at specific growth conditions.The study was performed in an aberration-corrected environmental transmission electron microscope (TEM) with an integrated metal-organic chemical vapor deposition system (MOCVD). [53]The nucleation of GaP NWs was observed and captured by high-resolution TEM (HRTEM) movies.The phases involved in the nucleation and NW growth process were determined via energy-dispersive X-ray spectroscopy (EDS) and power spectra obtained from the acquired HRTEM images.Unexpectedly, growth via the liquid-assisted VSS mode achieved by increasing the Ga supply led, in many cases, to NW kinking, which could be attributed to the different growth dynamics in the two phases of the growth promoter.This study reveals the potential of alternative seed materials to form semiconductor NWs via unique growth modes.A fundamental understanding of NW kinking is crucial for gaining control over the semiconductor NW morphology that might be relevant for thermal, optoelectronic, and biological applications. [49]

Results and Discussion
The GaP NW nucleation and growth experiments were performed in an environmental TEM with an integrated MOCVD system, allowing the supply of precursors such as phosphine (PH 3 ) and trimethylgallium (TMGa).Cu nanoparticles (NPs) acting as seed particles for the GaP NWs were generated by a homebuilt spark ablation system. [54]A differential mobility analyzer (DMA) enabled the selection of NPs with a specific mobility diameter for the deposition on microelectromechanical systems (MEMS) TEM heating chips via an electric precipitator.The TEM heating chips with deposited Cu NPs were then transferred to the environmental TEM.Subsequently, the precursors were supplied via separated lines connected to the microscope by a side port injector (see Section 4 for details).
The deposited Cu NPs were annealed at 600 °C in an H 2 atmosphere to reduce their defect concentration and remove surface contamination, such as oxides and C-based residues.The spherical Cu NPs obtained after the annealing step had a diameter of %30 nm, and surface contamination and defects were drastically reduced (see Figure S1, Supporting Information).After the temperature was decreased to %420 °C and the H 2 flow was switched off, PH 3 was supplied.Straight after supplying PH 3 to the Cu NPs, phase transformations were observed.The phase transformations yielded hexagonal-shaped NPs, as exemplarily shown in the HRTEM image in Figure 1a.The addressed NP was observed at the edge of the TEM heating chip's amorphous SiN x membrane and tilted close to one of its zone axes.The reflections in the power spectrum corresponding to the HRTEM image in Figure 1a matched well with the simulated electron diffraction pattern of the Cu 3 P phase (blue, space group: P6 3 cm) oriented in its ½001 zone axis (Figure 1b).This observation agrees with other experiments presented in the literature where the supply of a P precursor to Cu NPs yielded Cu 3 P NPs with hexagonal shapes. [55]

GaP NW Nucleation
The nucleation of GaP was initiated by adding TMGa to the PH 3 supply (see Section 4 for details).An HRTEM movie (see Movie S1, Supporting Information) was acquired to capture the dynamic process occurring straight after starting the TMGa supply, using a synthesized Cu 3 P NP that was tilted close to its ½102 zone axis (see Figure S2, Supporting Information).The Cu 3 P NP reacted with the supplied Ga atoms, yielding an NP heterostructure (Figure 1c).The reflections of the power spectrum corresponding to the HRTEM image in Figure 1c matched the simulated electron diffraction pattern of the zincblende (ZB) GaP phase (orange, space group: F43m) and reflections associated with specific planes of the Cu 9 Ga 4 phase (green, space group: P43m).
The overlay with simulations reveals that the ZB GaP phase was tilted close to its ½110 zone axis, while the Cu 9 Ga 4 phase was tilted off its zone axis, and only reflections corresponding to its ð110Þ planes were visible.Considering the largest observed interplanar spacing in Figure 1d is %6.203Å, which matches with the one for Cu 9 Ga 4 (110) planes (6.1358 Å) and not with those for the Cu 3 P(100) (6.0270 Å) or Cu(100) (3.6149 Å) planes, [56][57][58] it is likely that the Cu 9 Ga 4 phase formed in this process.According to the Cu-Ga phase diagram, [59] the Cu 9 Ga 4 phase can exist in three modifications (γ 1 -, γ 2 -, and γ 3 -modifications), which slightly vary in their composition. [57]ased on the chosen process temperature (420 °C) and neglecting the presence of P, the γ 3 -modification was considered the most likely phase to be present during the performed experiments. [59]Consequently, P atoms initially accumulated in the Cu-based phase in the form of Cu 3 P must have reacted with Ga atoms, enabling the nucleation of the GaP phase.It is worth mentioning that the electron beam continuously located at the NP could have led to the observed contamination of its surface, highlighted by a white arrow in Figure 1c.

Seed Particle Transformation
In the next step, we focused on investigating stable GaP NW growth after the nucleation stage.[62] For that purpose, a different strategy was applied to perform the addressed experiment.After the nucleation, the TMGa flow was reduced to decrease the GaP NW growth rate.Subsequently, the TMGa flow was switched off while keeping the PH 3 flow and the temperature constant.The Ga atoms in the seed particle acted as a reservoir to maintain the growth of the GaP phase.The GaP NW growth stopped when the Ga reservoir was depleted, and instead, P accumulated in the seed particle (see below).Subsequently, a different GaP NW was selected, which was oriented such that it could be tilted into one of its zone axes while maintaining its GaP-seed particle interface as parallel as possible to the electron beam direction.
The GaP NW shown in the HRTEM image in Figure 2a had a diameter of %40 nm.Its power spectrum, revealed in the inset of Figure 2a (orange, bottom right), highlights the orientation of the GaP phase in its ½110 zone axis.The reflections observed in the power spectrum of the HRTEM image's seed particle area shown in the inset of Figure 2a (blue, top right) do not allow a detailed crystallographic analysis.However, since the TMGa supply was stopped and the Cu-based phase depleted from Ga, we expect a subsequent formation of the Cu 3 P phase in the seed particle by  the chemical reaction of Cu with available P. It is worth mentioning that the interface between the GaP phase and the seed particle was not oriented parallel to the direction of the electron beam at this stage.
The TMGa flow was again added to the precursor supply to reinitiate GaP NW growth.The HRTEM images in Figure 2a-c are selected averaged frames extracted from an HRTEM movie (see Movie S2, Supporting Information) and reveal the phase transformation occurring in the seed particle due to the Ga supply.Before the phase transformation of the initial Cu-rich phase started, no GaP growth was observed.Within %50 s after TMGa was added to the precursor supply, the seed particle rearranged at the growth front leading to an expansion of the NW diameter indicated by a white arrow in Figure 2b.The chemical reaction proceeded over the whole seed particle resulting in an interface between the new phase and the GaP NW that was oriented parallel to the direction of the electron beam.
The phase transformation was completed %102 s after adding the TMGa supply.Clear lattice fringes could be observed in the seed particle and the NW (Figure 2c).The power spectrum in Figure 2d corresponds to the framed region in the HRTEM image in Figure 2c, and simulated electron diffraction patterns of the ZB GaP (orange, oriented in its ½110 zone axis) and Cu 9 Ga 4 (green, oriented in its ½110 zone axis) phases matched the observed patterns.Therefore, the phase transformation led to the presence of the Cu 9 Ga 4 phase, which was arranged at the ZB GaP½111 growth front with specific crystal orientation.The analysis of the epitaxial relationship between the two phases revealed that the ZB GaP(111) planes were oriented parallel to the Cu 9 Ga 4 ð111Þ planes.It is also worth mentioning that the lattice spacings and orientations of the ZB GaPð002Þ and Cu 9 Ga 4 ð003Þ planes, as well as the ZB GaPð220Þ and Cu 9 Ga 4 ð330Þ, are similar, which could explain the preferred epitaxial relation of the two phases (also observed in another NW; see Figure S3, Supporting Information).
As part of the phase transformation, pronounced faceting of the Cu 9 Ga 4 phase could be observed.Considering that the sharp boundaries of the Cu 9 Ga 4 phase in the projection are not edges formed by two differently oriented facets but are surface facets oriented parallel to the direction of the electron beam, a more detailed analysis of the faceting can be performed.Since we did not observe drastic contrast changes in the HRTEM image close to those boundaries, which would be expected if two differently oriented Cu 9 Ga 4 facets formed the boundaries resulting in crystal thickness variations, a surface facet oriented parallel to the direction of the electron beam is the more plausible scenario.The different Cu 9 Ga 4 surface facets are indicated by green labels (F1-4) in Figure 2c, and their crystallographic orientations are highlighted by green circles and labels in the corresponding power spectrum in Figure 2d.
The addressed increase in the GaP NW diameter during the phase transformation could be due to the interface change from Cu 3 P-GaP to Cu 9 Ga 4 -GaP.The matching of the Cu 9 Ga 4 and GaP phases at the interface might hint toward a lower energy interface promoting the increase in diameter (interface area).A similar process in VSS NW growth has been observed in the Au-seeded GaAs NW growth, [60] where the change in diameter was attributed to the formation of low-energy facets.Moreover, defects in the GaP NW segment that grew during the seed particle's phase transformation were observed.A white arrow indicates the defects in Figure 2c that likely originated in the seed particle's crystal structure change affecting the processes at the GaP NW growth front.
After a sharp interface formed between the ZB GaP NW and the Cu 9 Ga 4 phases, oriented parallel to the direction of the electron beam, a physical transformation of the seed particle was observed (see Movie S3, Supporting Information).The faceted seed particle (Figure 2c) was reshaped by corner truncation (Figure 3a).Corner truncation in faceted nanocrystals is a wellknown process driven by minimizing the total surface free energy. [61,63]It is worth mentioning that during the physical transformation of the seed particle, the GaP NW-seed particle interface changed significantly.White arrows in Figure 3a indicate the presence of unfinished GaP bilayers evolving from the GaP-Cu 9 Ga 4gas triple-phase boundary.After the physical transformation, the seed particle revealed a hemispherical shape (Figure 3b).The power spectrum of the HRTEM image in Figure 3b (see Figure S4, Supporting Information) did not show significant alterations from the power spectrum in Figure 2d, indicating no further phase transformations or crystal rotations accompanying the reshaping processes.The elevated temperatures might have enabled a high diffusivity of the involved atoms leading to the thermodynamically more stable shape.Furthermore, the interface area between the ZB GaP NW and the Cu 9 Ga 4 phase decreased slightly during the reshaping of the seed particle resulting in a new ZB GaP surface facet indicated by a white arrow in Figure 3b.

GaP NW Growth and Kinking
After the seed particle was reshaped at the ZB GaP NW tip, multilayer growth (nucleation of new bilayers before the bilayer closest to the growth front could fully evolve) [64] promoted by the Cu 9 Ga 4 phase proceeded for approximately 10 GaP bilayers (see Movie S4, Supporting Information).Multilayer GaP NW growth via the VSS mechanism yielded a ZB GaP segment without the formation of twins/stacking faults (Figure 4a).After the addressed GaP NW segment grew from a solid seed particle, a liquid phase was observed covering the Cu 9 Ga 4 phase (Figure 4b).Therefore, we conclude that the growth from a solid seed was a transition regime, and the liquid phase formed due to the accumulation of Ga atoms under the chosen synthesis conditions.This observation agrees with the Cu-Ga binary phase diagram when neglecting the presence of P. [59] It is worth mentioning that it is difficult to exclude a thin liquid layer already covering the solid seed particle during the VSS growth in the transition regime.
Interestingly, the Ga-rich liquid phase was more pronounced at the GaP-liquid-gas triple-phase boundary involving the ZB GaP surface facet with a different orientation indicated by the white arrow in Figure 3b.Likely, the contact angle between the seed particle and the ZB GaP phase played a crucial role in the favored accumulation of the liquid phase in the addressed region.As a result of the selective accumulation of the liquid phase, local ZB GaP growth via the VLS mechanism with a significantly increased growth rate was observed (see Movie S4, Supporting Information).The growth rate at the solid-solid interface was lower (only a few bilayers formed from %235 to %435 s after the supply of TMGa), leading to a gap between the two growth fronts (Figure 4a-c).Therefore, different VLS and VSS growth dynamics within the same GaP NW caused this observation, and the diffusion pathways of the growth species likely played a crucial role.
For better visualization of the liquid and solid phases in the seed particle, masks on the Cu 9 Ga 4 ð110Þ and ð110Þ reflections and, subsequently, an inverse fast Fourier transform (FFT) was applied to the power spectrum corresponding to Figure 4c.The image in Figure 4d obtained by the described procedure was overlaid with the outlines of the seed particle (black) and the ZB GaP NW (orange).The bright region indicates the solid Cu 9 Ga 4 phase.The projection shows that the solid was covered by a liquid phase located between the bright area and the outline of the seed particle (black) in Figure 4c.
The zoomed-in HRTEM image of the region indicated by a black rectangle in Figure 4c reveals the liquid phase forming a sharp interface with the ZB GaP phase.Moreover, the white arrows in Figure 4a-c show the same position in the GaP NW.The indicated ZB GaP region was in close contact with the liquid-solid interface in the seed particle during growth.The zoomed-in HRTEM image of the area highlighted by a red rectangle in Figure 4c shows that the image contrast varies significantly (Figure 4f ).However, no defects, including dislocations, could be observed.A possible explanation for the nonuniform image contrast could be a variation in crystal thickness due to the presence of the liquid-solid interface in the seed particle during GaP NW growth.
The gap between the GaP NW segments grown via the VLS and VSS mechanisms became more significant with time (see Movie S5, Supporting Information).While %54 GaP bilayers grew at %635 s after the supply of TMGa in the segment where the liquid phase accumulated at the GaP growth front, only a few GaP bilayers formed in the other part (Figure 5a).The imbalance in growth rates led to a new interface between the Cu 9 Ga 4 and ZB GaP phases, indicated by a white arrow in Figure 5a.The new interface fully evolved at %685 s after adding the TMGa supply (Figure 5b).
The power spectrum in Figure 5c corresponds to the area in the white rectangle of the HRTEM image in Figure 5b, which was overlaid with the simulated electron diffraction patterns of the same phases and for the same zone axes as used in Figure 2c and S4b, Supporting Information.Interestingly, the formation of the new Cu 9 Ga 4 -ZB GaP interface was not accompanied by alterations in the epitaxial relationship between the two phases.The new interface between the seed particle and the GaP NW was again oriented parallel to the direction of the electron beam, and growth proceeded in the ZB GaP½111 direction.The new growth front appeared rough at %810 s after adding the TMGa supply (Figure 5d).
A closer look at the ZB GaP NW segment grown in the final stage of the new interface formation (see Movie S6, Supporting Information) shows the presence of ZB GaP twins (black rectangle, Figure 5d).One of those ZB GaP twins already formed earlier, and a white arrow indicates its location in Figure 5b.The zoomedin HRTEM image of the addressed region (Figure 5e) shows the ZB GaP twin by highlighting the mirror plane (solid orange line) and the mirrored GaP atomic columns (dashed orange lines).
As a result of the kinking, the ZB GaP crystal revealed a region, indicated by a red rectangle in Figure 5d, where inhomogeneities could be observed.The zoomed-in HRTEM image of the addressed area suggests that the contrast variations were likely caused by different crystal thicknesses or faceting in the region where the kinking event was initiated (Figure 5f ).The straight dotted orange lines overlaid over selected atomic columns of the GaP phase reveal no significant bending of the planes around the addressed inhomogeneities, which could have indicated the presence of defects.
Further GaP NW growth led to the formation of a GaP NW-seed particle interface that was not oriented parallel to the direction of the electron beam (Figure 6a).This observation might have been caused by or initiated the liquid phase wetting the sidewall of the GaP NW, indicated by a white arrow in Figure 6a.As a further consequence of the liquid phase in contact with the GaP NW sidewall, the Ga-rich liquid shell was consumed to promote GaP growth antiparallel to the initial GaP½111 growth direction (Figure 6b).Interestingly, the power spectrum, as an inset in Figure 6b, reveals that the Cu 9 Ga 4 crystal did not change its orientation during the whole process (see Figure 2d and 5c).
The same behavior was observed in another experiment for a GaP NW tilted close to its ½110 zone axis (see Figure S3 and Movie S7, Supporting Information).Analogous to the main experiment, the kinking led to the change in the GaP NW growth from one to another <111> direction.Moreover, the epitaxial relationship between the GaP NW and the Cu 9 Ga 4 phases was the same as highlighted for the GaP NW in Figure 5.An additional experiment performed at higher V/III ratios (see Figure S5 and Table S3, Supporting Information) revealed the possibility of achieving straight GaP NW growth facilitated by a Cu 9 Ga 4 seed particle not covered by a liquid Ga-rich phase.Consequently, Cu-seeded GaP NW growth using relatively low V/III ratios for the precursor supply can initiate the formation of kinks caused by the locally different growth dynamics observed in the liquid-assisted VSS growth mode.
Finally, additional experiments were performed to evaluate the impact of exposing the sample to the electron beam on the described observations.For that purpose, the whole synthesis procedure was reproduced without exposing the sample to the electron beam by closing the gun valve.Subsequently, the gun valve was opened, and TEM images were acquired without tilting the NWs in one of their zone axes or improving the image quality by advanced alignment procedures.The whole procedure focused on minimizing the sample's exposure to electrons.The additional experiments suggest that the chosen synthesis parameters, including the relatively high TMGa partial pressure, were crucial for observing the liquid-solid two-phase seed particle-assisted GaP NW kinking (see Figure S6 and S7, Supporting Information).Any electron beam effects played minor roles in the observed processes.

Conclusion
In summary, we have investigated the in situ growth of GaP NWs using Cu as an alternative growth promoter.The Cu NPs were first transformed into Cu 3 P NPs by supplying PH 3 .After adding TMGa to the precursor supply, nucleation of GaP NWs was observed, involving the transformation of the Cu 3 P NPs into GaP-Cu 9 Ga 4 NP heterostructures.GaP NW growth experiments were performed on a pregrown Cu-seeded GaP NW by adding the Ga precursor to the PH 3 supply.The Cu-rich phase transformed into a faceted Cu 9 Ga 4 phase.The phase transformation was followed by a physical transformation yielding a Cu 9 Ga 4 seed particle with a hemispherical shape.After a short period of GaP VSS growth, a liquid phase covering the Cu 9 Ga 4 seed particle phase was observed.The local accumulation of the liquid phase led to GaP VLS growth with a significantly higher growth rate than observed for the segment in contact with the Cu 9 Ga 4 phase.The locally different growth dynamics led to the formation of a new GaP-Cu 9 Ga 4 interface.Consequently, the GaP NW kinked, leading to growth along another GaP<111> direction.As part of this process, the liquid phase started wetting the GaP NW sidewall, resulting in GaP NW growth along the NW sidewall antiparallel to the initial GaP½111 growth direction.The kinking, particularly the presence of a liquid-solid interface at the growth front, led to variations in the crystal thickness indicated by abrupt HRTEM image contrast changes.However, there were no signs of increased defect concentrations in the kink region.Additional experiments supported those findings and excluded a significant impact of the electron beam on the observations affecting our conclusions.This study gives fundamental insights into the kinking of GaP NWs using liquid-solid two-phase seed particles and highlights potential ways to control their morphology.

Experimental Section
Synthesis and Deposition of Cu NPs: Cu NPs were generated in a homebuilt spark ablation system.Cu rods were used as the anode and cathode in the particle generator.An H 2 /N 2 mixture was used as a carrier gas to transport the generated agglomerates through a furnace kept at 850 °C.The Cu NPs had different diameters after passing the furnace.A DMA was used to select NPs with a mobility diameter of %30 nm.The selected Cu NPs were then deposited on a MEMS-based chip by an electrostatic precipitator.A detailed description of the generator can be found elsewhere. [54]EMS-Based Heating Chips for in situ TEM Experiments: MEMS-based heating chips from Norcada Inc. were used to perform the TEM experiments.The chips consist of SiN x membranes where W coils are embedded to enable Joule heating.There are 19 holes in the middle of the chips with a few nm thick SiN x film at the edges.The chips could be heated up to 1100 °C with a homogeneous temperature profile along the central region of the chip.The temperature was controlled in constant resistance mode by the Blaze software supplied by Hitachi High-Technologies.
TEM Sample Holder and Gas/Precursor Supply: A double-tilt holder manufactured by Hitachi High-Technologies was used for the in situ TEM experiments.The TEM was connected to a gas handling system, which enabled the supply of precursors.While PH 3 was supplied directly to the sample, the TMGa precursor was transported by H 2 from a bubbler to the gas handling system, where the gas mixture was further diluted.Both precursor flows were supplied directly to the sample via separated lines of a side port injector integrated into the microscope.
TEM Characterization Techniques: The in situ TEM experiments were carried out in an environmental TEM (HF-3300s, Hitachi High-Technologies, Japan) at Lund. [65] The microscope is equipped with a cold field emission gun, which was operated at 300 kV in our experiment, an imaging aberration corrector (CEOS BCOR), and a complementary metal-oxide-semiconductor (CMOS) camera (Oneview IS, Gatan, USA).The HRTEM images and movies were recorded with an electron dose rate of %600-6800 e Å À2 s. (for details, see Table S1 and S2 Nucleation of GaP NWs using Cu NPs: After the Cu NPs were transformed into Cu 3 P NPs by the supply of PH 3 , the nucleation of GaP NWs was initiated by adding TMGa, using H 2 as a carrier gas, at 420 °C.Detailed parameters used for the Cu-Cu 3 P transformation and GaP NW nucleation presented in Figure 1 and S2, Supporting Information are summarized in Table S3, Supporting Information.
Liquid-Assisted VSS Growth Mode: Detailed information about the parameters used for the liquid-assisted VSS growth of GaP NWs presented in Figure 2-6, S3-S4, Supporting Information, and Movie S1-S7, Supporting Information is shown in Table S3, Supporting Information.
Additional Experiments: Detailed information about the parameters used for the additional experiments presented in Figure S5-S7, Supporting Information is shown in Table S3, Supporting Information.
Simulation of Electron Diffraction Patterns: The crystallographic information files of the Cu 3 P, Cu 9 Ga 4 , and GaP phases were downloaded from the Inorganic Crystal Structure Database (ICSD, NO. 15056, NO. 2359, and NO.77087).The lattice parameters were not temperature-corrected due to a lack of thermal expansion coefficients associated with the specific phases.SingleCrystal (Version 4.1.9)software was used to simulate electron diffraction patterns for different phases with various crystal orientations (see Table S4, Supporting Information for details).
Data Processing: For Figure 1-6 and S1-S7, Supporting Information, DigitalMicrograph (Version 3.42.3048.0)from Gatan was used to analyze HRTEM images and extract HRTEM images from the HRTEM movies.Moreover, power spectra of the HRTEM images were obtained by applying an FFT to the highlighted areas of the HRTEM images.The processing parameters for extracting HRTEM images from the raw HRTEM movie data are summarized in Table S5, Supporting Information.Movie S1-S7, Supporting Information correspond to three acquired HRTEM movies processed with the in situ player integrated into the DigitalMicrograph software (see Table S6, Supporting Information for details).The time stamps and scale bars were added by ImageJ (Version 1.52a).ImageJ was also used to compress the HRTEM movies extracted from the DigitalMicrograph software.The figures in the main text and the Supporting Information were prepared using Photoshop from Adobe (Version 23.1.0).

Figure 2 .
Figure 2. a-c) Selected averaged frames of an HRTEM movie (see Movie S2, Supporting Information) showing the phase transformation of a GaP NW seed particle: (a) %37 s, (b) %71.5 s, and (c) %102 s after adding TMGa to the PH 3 supply.a) A Cu-rich phase is arranged at the GaP NW growth front.The power spectra of the ZB GaP (orange, bottom right) and the seed particle (blue, top right) phases are shown as insets in (a).(b) Shows the phase transformation of the seed particle initiated by the Ga supply.(c) The phase transformation is completed %102 s after TMGa is added to the PH 3 flow.The green labels (F1-4) highlight the surface facets of the seed particle.d) The power spectrum corresponding to the white rectangle in (c) is overlaid with simulated electron diffraction patterns of the two involved phases.The power spectra of the Cu 9 Ga 4 (green, top right) and ZB GaP (orange, bottom right) phases are shown as insets in (d).Green circles and the corresponding labels indicate the orientations of the surface facets highlighted in (c).

Figure 1 .
Figure 1.a) HRTEM image of a Cu 3 P NP acquired at 420 °C after the PH 3 supply is started.b) Corresponding power spectrum with an overlaid simulated electron diffraction pattern of the Cu 3 P phase (blue).c) HRTEM image of a different NP acquired after adding TMGa to the PH 3 supply.The white arrow in (c) indicates C-based contamination of the NP surface.d) Corresponding power spectrum with an overlaid simulated electron diffraction pattern of the ZB GaP phase (orange) and simulated reflections of specific Cu 9 Ga 4 planes (green).The power spectra of the Cu 9 Ga 4 (green,

Figure 3 .
Figure 3. Selected averaged frames of an HRTEM movie (see Movie S3, Supporting Information) showing the shape of the seed particle at a) %120 s and b) %125 s after adding TMGa to the precursor supply.The white arrows in (a) indicate the partial growth of GaP bilayers during the reshaping process.As a result of the physical transformation, the seed particle's shape became hemispherical, and the GaP NW-seed particle interface decreased, resulting in a new GaP surface facet indicated by a white arrow in a zoomed-in region of the HRTEM image shown as an inset in (b) highlights the new GaP surface facet.

Figure 4 .
Figure 4. a-c) Selected averaged frames of an HRTEM movie (see Movie S4, Supporting Information) showing the competition between the VLS and VSS growth mechanisms within the same GaP NW.(a) shows the state of the GaP NW at %262 s after the supply of TMGa, which is soon after the physical transformation of the seed particle.(b,c) A liquid phase accumulated close to the GaP NW surface facet indicated in Figure 3b and covered the whole Cu 9 Ga 4 phase.d) The image obtained after applying masks on the Cu 9 Ga 4 ð110Þ and ð110Þ reflections and, subsequently, an inverse FFT to the power spectrum corresponding to Figure 4c is overlaid with the outlines of the seed particle (black) and the ZB GaP NW (orange).The bright area highlights the Cu 9 Ga 4 phase, while the remaining area inside the seed particle represents the liquid phase.e) The zoomed-in HRTEM image of the region indicated by a black rectangle in (c) reveals the liquid-solid interface in the seed particle and the arrangement of the phases at the ZB GaP NW growth front.f ) Another zoomed-in HRTEM image of the area highlighted by a red rectangle in (c) shows the image contrast variation due to crystal thickness changes caused by the presence of the liquid-solid interface in the seed particle during the growth of the GaP crystal.White arrows in (a-c) indicate the position of the associated GaP NW region.

Figure 5 .
Figure 5. Selected averaged frames of an HRTEM movie (see Movie S5-S6, Supporting Information) show the kinking of the GaP NW. a) A new Cu 9 Ga 4 -ZB GaP interface indicated by a white arrow formed due to the different growth dynamics of the VLS and VSS mechanisms.b) The new interface, parallel to the direction of the electron beam, fully evolved, and a twin, highlighted by a white arrow, formed in the ZB GaP phase.c) The power spectrum corresponding to the HRTEM image in (b) is overlaid with the same simulated electron diffraction patterns of the phases used in Figure 2d.No significant alterations of the patterns corresponding to the Cu 9 Ga 4 (green, top right inset) and ZB GaP (orange, bottom right inset) phases are observed.d) After the new interface fully evolved, the NW growth proceeded in the ZB GaP½111 direction and the growth front became rough over time.e) The zoomed-in HRTEM image from the area in the black rectangle in (d) shows a ZB GaP twin that had already formed at an earlier stage (see white arrow in (b)).The solid orange line highlights the mirror plane, while the dashed orange lines indicate the mirrored GaP atomic columns.f ) The zoomed-in HRTEM image of the area in the red rectangle in (d) shows contrast inhomogeneities in the region where kinking occurred.The straight dotted orange lines suggest that the lattice planes of the GaP phase did not reveal deviations in the addressed region, as can be expected in the case of a high defect concentration.Therefore, thickness variations are the most likely reason for this observation.

Figure 6 .
Figure 6.a) Selected averaged frames of an HRTEM movie and b) an HRTEM image acquired after this HRTEM movie showing the processes occurring after the kinking of the GaP NW.(a) The GaP NW-seed particle interface is no longer oriented parallel to the direction of the electron beam at %945 s after adding TMGa to the precursor supply.This observation is caused by or initiated by the liquid phase wetting the GaP NW sidewall, indicated by the white arrow in (a).(b) As a consequence of the liquid phase wetting the GaP NW sidewall, the growth proceeded antiparallel to the initial ZB GaP½111 growth direction.The inset in (b) reveals the power spectrum corresponding to the Cu 9 Ga 4 phase, which still shows no alterations or crystal rotations compared to the Cu 9 Ga 4 phase's power spectra insets in Figure 2d and 5c.
, Supporting Information) The background pressure next to the sample was %1.10 Â 10 À4 Pa for all experiments.