Detailed Characterization of the Surface and Growth Mechanism of Monodisperse Ni3Sn4 Nanoparticles.

Synthesis of most tin-based bimetallic nanoparticles is a challenging task because of the differences in the redox potential and the melting point between both components. This article presents a co-reduction synthesis of monoclinic Ni3Sn4 nanoparticles. Varying time and temperature gives the possibility to control the size of the nanoparticles in the range of 4-12 nm. The products were characterized by X-ray diffraction, high-resolution transmission electron microscopy, X-ray photoelectron spectroscopy, and energy-dispersive X-ray spectroscopy measurements. Although the synthesis was conducted entirely oxygen free, the postsynthetic treatment undertaken under air leads to the formation of an amorphous oxide shell. The oxide shell consists of an outer tin-rich region and a nickel-rich region at the interface to the metallic Ni3Sn4 core. On the basis of the investigation of the particles at different stages of the synthesis, we propose a growth mechanism for the Ni3Sn4 nanocrystals. These results can be a guidepost for the synthesis of other tin-based bimetallic nanoparticles.

Introduction

Metallic and especially bimetallic nanomaterials exhibit properties that are interesting for application in optical, catalytic, or electronic devices.[1−5] The main challenge in the synthesis of these materials arises from differences in such properties as the melting point or the redox potentials of both metals. It is problematic to find reaction conditions under which two metals with different properties can be simultaneously reduced to form bimetallic nuclei or to ensure the further uniform growth of the particles. Tin-based bimetallic nanoparticles always deal with high differences regarding the melting point caused by the extraordinary low melting point of tin (232 °C).[6−13] One promising intermetallic compound is the combination of tin and nickel, which leads to three stable Ni–Sn phases (Ni3Sn, Ni3Sn2, and Ni3Sn4) in accordance with the phase diagram.[14] The nickel–tin compounds are highly interesting for the application as anode material in lithium-ion batteries, where they could replace the currently used graphite anode, yielding higher specific capacities or as catalysts in acetylene hydrogenation.[15] The controlled production of one of the crystalline Ni–Sn phases involves many challenges. High-energy ball milling often leads to additional phases like tin oxide and mixtures of Ni–Sn phases.[16−20] However, the ball-milling technique lacks the control of size, shape, and crystal structure of the produced material, needed for most of the applications of such materials.[21] Deposition methods such as electrochemical or sputtering deposition deal with similar difficulties.[22−26] Colloidal synthesis starting with metal salts is an interesting approach to overcome these issues. Reduction of metal salts with sodium borohydride leads to polydisperse crystalline nanoparticles in a size range from 5 to 50 nm but with mixed phases of tin oxides and Ni–Sn phases like Ni3Sn2 and Ni3Sn.[27−29] Dong et al. used KBH4 as a reducing agent and obtained amorphous micrometer particles containing Ni and Sn in a ratio of 3:4.[30] Depending on the initial weights of the metal salts, Antonin et al. synthesized Ni–Sn samples containing all three Ni–Sn phase compositions (Ni3Sn, Ni3Sn2, and Ni3Sn4) with side products of NiO and SnO2 in a size range <10 nm.[31] Therefore, the wet chemical synthesis is a useful approach to gain more control of the size of the particles but still synthesizing single-crystalline phases is not easy to achieve. Applying tin templates as a starting material is a promising approach because it allows high control of size, shape, and crystal structure; however, it is always limited by the possibility to synthesize the appropriate templates. Chou and Schaak synthesized hollow single-crystalline Ni3Sn4 nanorods with tin nanorods as templates and cube-shaped NiSn3 nanocrystals, starting with spherical tin nanoparticles.[32,33] Watanabe et al. obtained Ni3Sn2 nanocubes with an averaged edge length of 15 nm using almost spherical tin nanoparticles.[34] Wang et al. synthesized spherical Ni3Sn4 nanoparticles with an average size of 40 nm; however, there is no information about the templates in this study.[6]

Not only the size and phase control of nanocrystalline Ni–Sn phases require further investigation but also information about their growth mechanism and detailed structure is still limited. Therefore, in the current study, we developed a co-reduction synthesis, leading to the formation of single-crystalline, colloidal Ni3Sn4 nanoparticles, with size controllable in a range between 4 and 12 nm. We investigated the size and shape of the nanoparticles by transmission electron microscopy (TEM), the crystal structure by X-ray diffraction (XRD), and the composition within the nanoparticle with high-resolution TEM (HRTEM), X-ray photoelectron spectroscopy (XPS), and energy-dispersive X-ray spectroscopy (EDX) measurements. On the basis of this information we could propose a detailed model of the synthesized nanoparticles. In addition to that, we analyzed the influence of reaction time and reaction temperature on the growth of the nanoparticles, which gave us insight into the formation and growth mechanism of Ni3Sn4 particles in colloidal solution.

Results and Discussion

Colloidal single-crystalline Ni3Sn4 nanoparticles were synthesized via a hot injection of the reducing agent at 230 °C. Metal chlorides were used as a metal source, whereas oleylamine acted as a solvent and at the same time as a ligand for the emerging nanoparticles, keeping them from agglomeration. To prevent any side reactions before the reduction, the solution must be free of water and oxygen, which is achieved by evacuating the setup. Before reducing the metal ions, addition of lithium-bis(trimethylsilyl)amide induces the formation of metal–oleylamine precursors.[21,35] To obtain Ni3Sn4 nanoparticles with a narrow size distribution, the solution was held at 230 °C for 21 h. Traces of agglomerates were sorted out via centrifugation.

The size and shape of the synthesized nanoparticles were determined by TEM, and their crystallographic structure and phase composition by XRD, HRTEM, and XPS.

In Figure , a representative TEM image of the particles is shown together with the corresponding size distribution. The nanoparticles are all uniform and quasispherical, with a mean diameter of 3.7 ± 0.6 nm.

Figure 1

(a) TEM image of quasispherical nanoparticles synthesized at 230 °C for 21 h with (b) the corresponding size distribution. The mean and standard deviation of the nanoparticles’ diameter is 3.7 ± 0.6 nm.

In Figure , the resulting X-ray diffraction pattern is plotted. Because of the substantial broadening of the peaks, a Rietveld analysis was conducted to get information about the crystal structure parameters and the size of the particles. We started the refinement with the monoclinic Ni3Sn4 structure (ICSD, PDF: 04-007-1116). After refinement of the particle size, strain, and the lattice parameters, as well as the atomic positions (see Table ), we achieved a good agreement between the experimental and computed pattern (blue and green line in Figure , respectively).

Figure 2

Rietveld analysis of the diffraction pattern with the Ni3Sn4 crystal structure (ICSD, PDF: 04-007-1116). No additional phase is observed.

Table 1

Fit Parameter of the Rietveld Analysis of the Ni3Sn4 Nanoparticle Sample

particle diameter	2.55(5) nm
strain	1.27(1)%
lattice parameters	β = 101.87(7)°
	a = 12.21(1) Å
	b = 4.07(2) Å
	c = 5.278(3) Å
atom positions	atom	x	y	z
	Ni1	0	0	0
	Ni2	0.234(1)	0	0.309(1)
	Sn1	0.521(1)	0	0.263(1)
	Sn2	0.203(1)	0	0.760(1)

Thus, we can rule out the presence of additional crystalline phases in this sample. To compare the particle size obtained by TEM measurements with the results from Rietveld analysis, a volume-weighted mean diameter D4,3(36,37) has to be calculated from TEM data (4.2 nm for this sample). In comparison with 2.5 nm, the particle diameter received from Rietveld analysis, there is a distinct difference in size not explainable by the fit quality. In fact, the assumption of an amorphous oxide shell surrounding the nanoparticles would explain the low diameter obtained from the Rietveld analysis. Regarding the galvanic series, it is expected that both metals are easy to oxidize. Thus, the formation of an oxidized shell is likely in contact with the atmosphere, whose exact composition needs to be investigated.

For further information about the crystal structure and composition, HRTEM combined with EDX measurements was performed. In Figure , an ensemble of nanoparticles is shown and most of them show lattice fringes, indicating their crystalline character. To identify the lattice plane distances and the related crystal structure, a fast Fourier transformation (FFT) of the red marked particle was performed. The nanoparticle is oriented in [13̅1] direction, and all diffraction spots can be assigned to the same monoclinic Ni3Sn4 crystal structure used for the Rietveld analysis, substantiating the single-crystalline nature of the nanoparticles. A distinct amorphous shell surrounding the nanoparticle is not clearly visible, which can be explained by the lower contrast of an oxidized species against the crystalline metallic core. Hence, further investigation is needed.

Figure 3

(a) HRTEM image of crystalline nanoparticles after 21 h of reaction time. The corresponding (b) FFT image of the red marked area can be assigned to the lattice plane distances and the corresponding angles of the Ni3Sn4 phase.

Spatially resolved EDX measurements give the first hint at the detailed composition of the nanoparticles. In Figure , element-specific mappings over an ensemble of nanoparticles are shown. The elemental distribution of tin is more spread out than that of nickel, indicating a tin-rich oxide shell, which is unexpected due to the lower standard electrode potential of nickel.

Figure 4

EDX mapping of Ni3Sn4 nanoparticles; (a) scanning-TEM image of an ensemble of nanoparticles with mapping of (b) the nickel signal, (c) the tin signal, and (d) the combination of both signals.

To examine the shell in more detail, XPS measurements combined with low-energy sputter etching using argon clusters were conducted. XPS measurements are particularly suitable to characterize the surface of the nanoparticles and give information about the composition of the interior of the particle. The XP survey spectra and XP spectra of the Ni 2p and Sn 3d signals before and after the etching procedure are shown in Figure . All signals in the survey spectra can be explained by the presence of Ni, Sn, O, and C in the sample. Small Si-related features in the initial survey spectrum can be explained by the Si-containing lithium base used in the synthesis. The Sn 3d signal was fitted using two doublet signals with a spin–orbit splitting of 8.4 ± 0.1 eV and a ratio of both doublet signal intensities of 0.67 (Figure S1a). The best fit was achieved using three species, a metallic one and two oxidized species, which can be explained by the presence of Sn(II) and Sn(IV) in the nanoparticles’ oxide shell.[38] The asymmetry of the metallic signal was considered. The binding energies for the three species are EB(Sn0) = 484.4 ± 0.1 eV, EB(SnII) = 485.9 ± 0.2 eV, and EB(SnIV) = 486.5 ± 0.1 eV. The values for the oxidized species are in good agreement with the literature. The binding energy of the metallic species is around 0.4 eV lower than the literature value for metallic Sn. This can be explained by the fact that the Sn0 is present in the form of a Ni3Sn4 compound. From the Ni 2p signal, only the Ni 2p3/2 part was fitted (Figure S1b). The metallic species dominates the Ni signal at the end of the sputtering procedure. The additional features that are observed at the beginning of the etching are explained by oxidized Ni species present in the nanoparticles’ oxide shell. The metallic species was fitted with an asymmetric signal. The binding energy EB(Ni0) = 852.5 ± 0.1 eV is in good agreement with the literature. No significant shift of the binding energy is observed due to the alloying with Sn according to the literature. Two satellites at energies 3.7 ± 0.2 and 6.0 ± 0.2 eV above the metal main peak and with a signal intensity between 5 and 15% of the main metallic signal were observed according to the literature. This information was used to better fit the spectra with superimposing features of the metallic and the oxidized Ni species. It is known that the XPS signals of oxidized Ni species are shaped by (often unresolved) multiplet splitting. Often this results in a rather broad enveloping peak.[39,40] To estimate the signal intensity and thus the atomic fraction of oxidized Ni, we used one broad signal and one satellite, yielding a good fitting of the data. It has to be pointed out that Ni is known to be reduced by argon sputtering.[41,42] Therefore, we used the total amount of Ni species for further analysis.

Figure 5

Comparison of XP spectra before and after the etching procedure of (a) the survey spectra, (b) the Sn 3d signals, and (c) the Ni 2p signals.

In Figure , the intensities, normalized on each species’ maximum value are plotted as a function of the sputtering time for the XPS signals of C, Ni, and Sn. The carbon signal decreases with a high slope at the beginning of the etching procedure and is almost steady after 3000 s. This behavior indicates a fast degradation of the organic layer surrounding the nanoparticles. Comparing the gradient of different signals can give a hint of the location of the elements within the nanoparticles. On the basis of the fitted XPS signal for Sn, the evolution of the oxidized and metallic species are plotted separately. Because of the reduction of metallic Ni by argon sputtering, a separation of the metallic and oxidized Ni species does not yield reliable results and was, consequently, not conducted.[42,43]

Figure 6

Overview of normalized intensities during the sputtering procedure for selected species. The Sn signal is divided into a metallic (Sn0) and oxidized species (Sn). The Ni signal is divided into the summed signal and a nickel signal (Niotc “outside the core”) resulting from the assumption of a stoichiometric Ni3Sn4 core. Lines are a guide for the eye.

Assuming a perfect Ni3Sn4 crystal, the gradients of the Sn and Ni signal would be identical, which is not the case here. The increase of the Sn signal is significantly higher than that of the Ni signal, indicating an inhomogeneous distribution of both elements in the nanoparticles, with a Sn-rich surface. The difference between the slope of the oxidized and metallic Sn species is in line with an oxidic shell and a metallic core of the nanoparticles. This result is in good agreement with the HRTEM and EDX measurements in Figures and 4. Furthermore, the intensity of oxidized Sn is slowly decreasing after 1000 s of sputtering time, implying a slight degradation of the oxide shell.

As mentioned before, the Ni signal cannot be divided into oxidized and metallic species without significant errors. Instead, we assume that the metallic core consists of stoichiometric Ni3Sn4 and all metallic Sn signal results from that core. On the basis of that idea and the atomic fractions of the species (Figure S2), the metallic Ni signal belonging to the core of the nanoparticle can be calculated and subtracted from the overall Ni signal (see the Supporting Information for details). The evolution of the remaining Ni signal is plotted as Niotc (outside the core). Although the oxidation state of this signal is not distinguishable, its evolution differs from that of all tin signals, indicating a different distribution of Ni compared to that of Sn outside the core with Ni predominantly in the inner region of the shell.

Summing up all information received by TEM, XRD, HRTEM, EDX, and XPS, a detailed model of the synthesized nanoparticles is sketched in Figure . The nanoparticles have core–shell geometry with a monoclinic Ni3Sn4 core and an amorphous shell. The shell itself contains an irregular distribution of both elements Sn and Ni, presumably oxidized. Regarding the galvanic series, nickel atoms are easier to oxidize than tin atoms. Therefore, it is surprising that the fraction of Sn is higher near the surface and that Ni is located near the interface of the core and the shell. Within the Ni3Sn4 crystal structure, tin atoms are more weakly bound than the nickel atoms, resulting in a higher diffusion rate reflected by the big difference of the melting temperature between tin (232 °C) and nickel (1455 °C).[44] Consequently, using tin in intermetallic nanoparticles can prevent the less noble metal from oxidation.

Figure 7

Schema of the synthesized nanoparticles regarding the element distribution. For visualization reasons, the thickness of the oxide shell is overestimated.

To focus more on the growth mechanism of the nanoparticles, we stopped the synthesis after 1 and 5 h at 230 °C, obtaining nanoparticles shown in Figure . In comparison to the first sample held at 230 °C for 21 h, there are some larger nanoparticles surrounded by many smaller ones. The corresponding size distribution is broadened, and due to the single large particles, there is a not negligible fraction centered at 10.5 nm, which decreases for the sample heated for 5 h at 230 °C (Figure b). Thus, more extended heating at 230 °C leads to a narrower size distribution combined with a smaller mean diameter of the particles. Therefore, no Ostwald ripening is observed at 230 °C but there is a disappearance of the larger nanoparticles. This disappearance is unexpected since smaller particles tend to dissolve more readily in a colloidal solution because of the high surface tension. In addition to that, after 5 h, at 230 °C, we observed the formation of a small fraction of aggregates, which was removed before TEM measurements by centrifugation. The precipitate obtained from the synthesis conducted for 21 h contains an additional β-Sn phase (see Figure S3). To clarify the origin and following disappearance of these larger nanoparticles, we studied the differences in their structure and composition by HRTEM and EDX.

Figure 8

TEM images and the corresponding size distribution of nanoparticles heated for (a) 1 h and (b) 5 h at 230 °C.

In Figure , an HRTEM image of the sample with 1 h reaction time is shown. All nanoparticles, regardless of their size, are crystalline and exhibit clearly visible lattice fringes. The smaller nanoparticles are single crystalline in contrast to the larger ones, which consist of more than one crystalline domain. The single-crystalline nanoparticle in Figure b can be identified as Ni3Sn4 oriented in [13̅1] direction. The composition of the polycrystalline nanoparticles was compared to the monocrystalline ones by HRTEM–EDX measurements (Figure S4). The Sn content of the polycrystalline particles is twice as high as the Ni content. Therefore, the larger nanoparticles differ in the crystallinity and composition. We assume two different growth paths for both kinds of particles (see below for the details of the suggested growth mechanism).

Figure 9

(a) HRTEM image of nanoparticles heated for 1 h at 230 °C. The large nanoparticle exhibits a polycrystalline structure, (b) HRTEM image of a single-crystalline nanoparticle with (c) the corresponding FFT. The lattice planes can be assigned to the Ni3Sn4 crystal structure.

Thus, the synthesis of Ni3Sn4 nanoparticles can be channeled from a bimodal size distribution with a mixture of single-crystalline nanoparticles and polycrystalline nanoparticles to a narrow size distribution of only single-crystalline nanoparticles by extending the duration of the reaction at 230 °C. A slight effect of the nanoparticle growth can be observed but is still relatively small taking the long reaction time of over 20 h into account. To speed up the nanoparticle growth, we added a second temperature step after nucleation at 230 °C. The samples were heated to 300 °C for different durations between 1 and 24 h. All samples were characterized regarding their size and crystallographic structure. The TEM images are shown in Figure for three selected reaction times.

Figure 10

TEM images of the resulting nanoparticles after heating to 300 °C for (a) 1 h, (b) 4 h, and (c) 12 h and (d) the corresponding size distribution. (e) HRTEM image after 12 h.

The nanoparticles remain quasispherical, but their mean diameter changes after 1 h of heating at 300 °C to 5.2 ± 1.0 nm, after 4 h to 8.7 ± 1.7 nm, and after 12 h to 12.2 ± 2.3 nm. Further heating (24 h) leads to the formation of large agglomerates, not stable in colloidal solution (Figure S5e). The results of three more samples, obtained after 0, 2.5, and 8 h are summarized in the Supporting Information (Figure S5). The increasing diameter of the particles and the standard deviation combined with an increase of time indicates Ostwald ripening as the growth mechanism. To rule out the formation of additional phases while heating the reaction solution to 300 °C, XRD measurements were done. The normalized diffraction patterns are plotted in Figure together with the reference data for Ni3Sn4 and β-Sn (AMCSD: 0011248). As expected, the broadening of the diffraction peaks decreases with longer reaction times, indicating the presence of larger crystallites within the samples. The sample heated for 24 h at 300 °C (Figure ) consists of large agglomerates with a diameter of several hundred nanometers and a negligible volume fraction of small nanoparticles. In addition to that, the sample exhibits diffraction peaks, which can be assigned to β-Sn and Ni2PSn (ICSD, PDF: 04-010-2577). Apparently, high temperatures and long reaction times lead to a release of phosphorus out of the solvent tributylphosphine.

Figure 11

XRD data of Ni3Sn4 nanoparticles for different heating times at 300 °C.

Figure 12

Rietveld analysis of a sample heated for 24 h at 300 °C. Next to the usual Ni3Sn4 phase, additional phases of β-Sn and Ni2PSn are present and plotted separately from the total Rietveld refinement.

To learn more about those phases and the phase composition, Rietveld analysis (Figure ) was conducted. The ratio of the phases Ni3Sn4/β-Sn/Ni2PSn is about 4:1:1. The refined parameters of the analysis can be found in the Supporting Information (Table S1). In contrast to the nanoparticles not heated to 300 °C but held at 230 °C for 21 h, the unit cell is barely distorted. The crystallite size of the β-Sn phase is in the region of micrometers, followed by crystallites of more than hundred nanometers for the Ni3Sn4 phase. Only the Ni2PSn phase shows crystallite sizes in the nanometer range of approximately 9 nm.

Taking all of the above information into account, a growth mechanism can be postulated and is sketched in Figure . We assume the formation of liquid tin-rich nuclei owing to the higher standard electrode potential of tin.[45,46]The fact that nickel nanoparticles did not form in reactions conducted without a tin source supports this theory (results not shown here for brevity). The liquid nuclei can easily aggregate during collisions in the reaction solution, which leads to the formation of a small fraction of larger particles (Figure ). Both the nuclei and the larger particles react with Ni monomers. However, assuming a diffusion-controlled reaction, in which each particle reacts with the same amount of Ni monomers, regardless of its size, the fraction of Ni increases faster in the small particles. Thus, smaller nanoparticles more quickly reach the stoichiometric ratio of Ni3Sn4. Meanwhile, their melting point changes from 232 °C (Sn) to higher values up to 795 °C (Ni3Sn4), resulting in a liquid–solid phase transition. Thus, further collisions of these particles are not likely to result in new aggregation. Heating to higher temperatures leads to further growth by Ostwald ripening for the Ni3Sn4 nanoparticles and agglomeration of the tin-rich particles, which still might contain some liquid regions. Hence, the pure Ni3Sn4 nanoparticles can be separated from the byproducts of the reaction by centrifugation. During purification, the nanoparticle dispersion is exposed to the atmosphere, resulting in the formation of an oxidized shell, which could be avoided by working under an inert gas atmosphere.

Figure 13

Suggested growth mechanism for Ni3Sn4 nanoparticles.

Conclusions

In this work, we developed a colloidal synthesis of Ni3Sn4 nanoparticles with high control of the phase composition and size. We used TEM, HRTEM, XRD, EDX, and XPS to characterize the monoclinic crystal structure, the different sizes of the nanoparticles, and the formation of the oxide shell. We showed that the nanoparticles with almost 4 nm size reveal a distorted unit cell, which recovers after treatment at 300 °C, forming larger nanoparticles up to 12 nm. An additional Ni2PSn phase can be observed after 24 h caused by the decomposition of one of the solvents. At lower temperatures, the synthesis can easily be modified to eliminate nanoparticles with a different phase composition than that of Ni3Sn4. Exposing the nanoparticles to the air causes the formation of an oxide shell with a high Sn concentration at the surface, protecting nickel from oxidation. Using all given information, we proposed a growth mechanism for this synthesis, demonstrating the essential steps from nucleation over Ostwald ripening to the stable nanoparticle. In general, the simultaneous reduction of nickel and tin can be performed successfully although the redox potentials differ. The huge melting point difference and the resulting faster diffusion of tin atoms lead to the formation of a tin-rich oxide shell. This synthetic method could be transferred to other tin-based bimetallic nanoparticles containing, for example, cobalt or iron.

Methods

Oleylamine (technical grade, 70%), tin chloride (SnCl2, 98%), nickel chloride (NiCl2, 98%), tributylphosphine (93.5%), lithium-bis(trimethylsilyl)amide (LiN(SiMe3)2, 97%), toluene (99.8%), diisobutylaluminium hydride in tetrahydrofuran (1 M DIBAH in tetrahydrofuran), and oleic acid (technical grade, 90%) were purchased from Sigma-Aldrich.

The synthesis of Ni3Sn4 nanoparticles was performed on the basis of the synthesis of tin nanoparticles established by Kravchyk et al.[21] In the first step, 130 mL of (0.4 mol) oleylamine was dried at 140 °C under argon/vacuum using a Schlenk line. Then, 95 mg of (0.5 mmol) tin chloride crystals was added to the solution and again dried for 30 min. Thereafter, the solution was heated up to 235 °C under argon atmosphere. A freshly prepared solution of 1054 mg of (6.3 mmol) lithium-bis(trimethylsilyl)amide in 3 mL of toluene was injected. The color of the solution turned from slightly yellow to orange immediately, accompanied by gas evolution. Directly thereafter, a freshly prepared solution of 49 mg of (0.375 mmol) nickel chloride in 5 mL of tributylphosphine was injected as well, followed by an injection of 2 mL (2 mmol) of the reducing agent DIBAH, turning the solution brown. The temperature had dropped during the injections to around 230 °C at which it stayed for 21 h. Afterward, the solution was cooled down slowly by removing the heating mantle, centrifuged with methanol, and the black precipitate was redispersed in an oleic acid/toluene solution, washed with methanol, and redispersed in toluene. In the last step, the nanoparticle dispersion was centrifuged and the obtained supernatant containing the colloidal dispersed Ni3Sn4 nanoparticles was used for further investigations.

X-ray Powder Diffraction

Powder X-ray diffraction (XRD) data were recorded using a PANalytical X’Pert Pro diffractometer with Cu Kα radiation and a Bragg Brentano θ–θ setup. The samples were prepared by dropping the nanoparticle dispersion onto low-background silicon sample holders. The diffraction patterns were analyzed by Rietveld refinement with Maud software version 2.80.[47]

Transmission Electron Microscopy

Transmission electron microscopy (TEM) measurements were performed using a Zeiss EM 900N microscope with an acceleration voltage of 80 kV. High-resolution TEM (HRTEM) images were collected on JEOL 2100F with an electron acceleration voltage of 200 kV. Carbon-coated copper grids were used as sample holders covered with diluted nanoparticle solution. For energy-dispersive X-ray spectroscopy (EDX), Oxford INCA Energy TEM250 with SDD detector X-Max80 was used.

X-ray Photoelectron Spectroscopy

XPS measurements were performed on an ESCALAB 250 Xi system (Thermo Fisher, East Grinstead, U.K.) with monochromatized Al Kα radiation (hν = 1486.6 eV), with a spot diameter of 650 μm. Charge compensation was applied if necessary. For sample preparation, an aliquot of the nanoparticle dispersion was dropped on an Au-coated substrate and dried. Data analysis was conducted with Avantage software version 5.977. All signals were charge-referenced to the C 1s signal (EB = 284.8 eV). The sputtering procedure was undertaken carefully with large argon clusters with a kinetic energy of 4 keV while the sample was rotated. Between every etching step and the subsequent measurement, a delay time was introduced.