In situ TEM observation of liquid-state Sn nanoparticles vanishing in a SiO2 structure: a potential synthetic tool for controllable morphology evolution from core–shell to yolk–shell and hollow structures

Precise design of hollow nanostructures can be realized via various approaches developed in the last two decades, endowing nanomaterials with unique structures and outstanding performances, showing their usefulness in a broad range of fields. Herein, we demonstrate the formation of SnO2@SiO2 hollow nanostructures, for the first time, by interaction between liquid state Sn cores and SiO2 shell structures inside Sn@SiO2 core–shell nanoparticles with real-time observation via in situ transmission electron microscopy (TEM). Based on the in situ results, designed transformation of the nanoparticle structure from core–shell Sn@SiO2 to yolk–shell Sn@SiO2 and hollow SnO2@SiO2 is demonstrated, showing the controllable structure of core–shell Sn@SiO2 nanoparticles via fixing liquid-state Sn inside a SiO2 shell which has a certain Sn containing capacity. The present approach expands the toolbox for the design and preparation of yolk–shell and hollow nanostructures, thus providing us with a new strategy for fabrication of more complicated nanostructures.


Introduction
Hollow nanoparticles have been applied in a broad range of applications, such as catalysts, 1 energy conversion and storage, 2 sensors, 3 and biomedicines, 4 due to their unique properties including large specic surface area, low density, etc., determined by their ingenious structures. Much work so far has focused on the synthesis of hollow nanoparticles, which basically can be classied into three strategies: (1) a hard templating pathway, (2) a so templating pathway, and (3) a self-templating synthetic route which includes surface-protected etching, Ostwald ripening, the Kirkendall effect, and galvanic replacement. [5][6][7][8][9][10] Generally, all these strategies consist of two basic steps, including (1) formation of a precursor for the hollow structure (e.g., core-shell nanostructures) and (2) creation of a hollow space by various methods including etching of cores, removal of templates, net outward diffusion, etc. 11 So far, the formation of a hollow structure in most of the research has been carried out and observed via ex situ techniques, in which the transient information involved during the hollowing process was missing. To gain more insight, in situ transmission electron microscopy (TEM) has been used in recent research to observe the hollowing process caused by the galvanic reaction 12,13 and the Kirkendall effect. 14,15 E. Sutter et al. used liquid-cell electron microscopy to investigate galvanic replacement reactions between Ag nanoparticle templates and aqueous Pd salt solutions. They observed the transformation from Ag nanoparticles into hollow Ag-Pd nanostructures. 12 Chee et al. used liquid-cell electron microscopy for observation of the hollow Ag nanostructure with void formation due to the nanoscale Kirkendall effect occurring in conjunction with galvanic replacement. 15 Several studies have reported the in situ observation of hollow structure formation by removing core templates from the core-shell structures. Yang et al. observed that when exposed to electron beam irradiation, ZnO@Al 2 O 3 core-shell nanowires can be transformed into ZnO@Al 2 O 3 composite nanotubes via local etching of ZnO cores. 16 Zhu et al. observed the transformation of Se-C core-shell nanoparticles into yolk-shell nanoparticles and hollow nanoparticles by electron beam irradiation, which induced the evaporation of embedded Se. 17 In this study, we report the formation of SnO 2 @SiO 2 hollow nanostructures, for the rst time, by interaction between liquid state Sn cores and SiO 2 structures that were synthesized through wet chemistry inside Sn@SiO 2 core-shell nanoparticles, with real-time observation via in situ TEM. According to the in situ observation, diffusion of liquid state Sn at 300 C contributes to the moving-out of embedded Sn cores, creating a cavity inside SiO 2 shell structures. Based on this understanding, to the best of our knowledge, we demonstrate the synthesis of yolk-shell Sn@SiO 2 and hollow SnO 2 @SiO 2 nanoparticles from core-shell Sn@SiO 2 nanoparticles for the rst time, showing nanostructure control via interaction between liquid state metal and the SiO 2 shell structure. Based on these results, this new strategy is capable of being applied in the design and preparation of yolk-shell, hollow nanostructures or more complicated nanostructures with the heating method which is expected to be achieved by using lasers or microwaves.

Structural characterization
A modied reverse-emulsion method was used for uniform and controllable SiO 2 encapsulation of Sn nanoparticles which were prepared though a well developed hot-injection method. 18 Sn@SiO 2 core-shell nanoparticle samples with two dimensions were produced: (a) a 20 nm core with a 20 nm shell (Fig. S1, see the ESI †) and (b) a 30 nm core with an 8 nm shell (Fig. S2, see the ESI †). The samples were loaded on a tungsten wire (as shown in Fig. 1a and b, respectively, in a vacuum at 25 C) and heated inside a TEM heating holder for in situ TEM observation (Fig. S3, see the ESI †). Several Sn@SiO 2 core-shell nanoparticles in Fig. 1a were examined by selected area electron diffraction (SAED). The diffraction spots in the corresponding pattern shown in Fig. 1c are indexed to tetragonal Sn (JCPDS no. 04-0673), indicating the existence of metallic Sn cores. Aer completing the whole heating process (from 25 C to 300 C in a vacuum) in a TEM and cooling down to 25 C, most of the Sn cores disappeared and some decreased in size, forming hollow and yolk-shell nanostructures as shown in Fig. 1d-f. Sn@SiO 2 nanoparticles shown in Fig. 1a were observed under strong electron beam conditions (37.3 pA cm À2 ). Under these conditions, signicant deformation of SiO 2 shells occurred with interconnections between them (Fig. 1d). Meanwhile, according to the SAED pattern of the hollow and yolk-shell nanostructures in Fig. 1d (Fig. S4, see the ESI †), the rings are indexed to SnO 2 (JCPDS no. 41-1445) while several dim spots are indexed to the (200) and (101) planes of tetragonal Sn (JCPDS no. 04-0673), demonstrating that the transformation of Sn to SnO 2 occurred during the heating process, probably due to the interaction with the dangling bonds in the porous SiO 2 . In order to avoid the effect of beam irradiation, weak electron beam conditions (2.7 pA cm À2 ) were applied when observing Sn@SiO 2 nanoparticles as shown in Fig. 1b. Aer heating, most of the Sn cores disappeared, leaving behind a hollow SiO 2 structure without signicant deformation (Fig. 1f), and its corresponding SAED pattern (the inset image in Fig. 1f) is indexed to SnO 2 (JCPDS no. 41-1445), indicating the existence of SnO 2 .

Dynamic motion of Sn nanodroplets inside the SiO 2 structure
We rst discuss the dynamic motion of Sn nanodroplets inside the deformed SiO 2 structure at an elevated temperature under electron beam irradiation according to the observation of detailed structural evolution and transformation that was recorded in situ during the heating process. When the temperature exceeds the melting point (T m ) of metallic Sn (bulk: 231 C, 20 nm in diameter: 223 C), 19 the SiO 2 encapsulated solid-state Sn cores are transformed to a liquid state as Sn nanodroplets. Fig. 2a shows the dynamic motion of Sn nanodroplets inside the SiO 2 structure at 300 C in a vacuum under electron beam irradiation (37.3 pA cm À2 ) captured from the movie (ESI Movie 1, see the ESI †). In the initial state, Sn cores A and B, with initial diameters of 16.9 and 21.2 nm, respectively, were located at the central positions of two adjacent Sn@SiO 2 core-shell nanoparticles with clear boundaries in between (Fig. 2a, t ¼ t 0 ). As the heating proceeded, both core A and B decreased in size (A: 11.3 nm and B: 16.6 nm, Fig. 2a, t ¼ t 0 + 314 s) and core A started to move away from the central position (Fig. 2a, t ¼ t 0 + 322 s). By t 0 + 362 s, dynamic motion in random directions inside the SiO 2 shell is observed. The instant movement directions of Sn cores are shown with blue arrows marked in the TEM images according to the movie. Aer t 0 + 368 s, core A started to move along a xed direction toward core B and reached the edge of the SiO 2 shell at t 0 + 391 s. Meanwhile, core B moved in a similar manner toward the outer edge of the SiO 2 shell and met core A at t 0 + 421 s. During the heating process, both core A and B gradually became invisible by contrast in the SiO 2 structure (Fig. 2a, t ¼ t 0 + 556 s). The observed movement and disappearance of Sn cores may provide evidence for the vanishing of metal Sn inside the SiO 2 structure. The dynamic motion of Sn cores was probably caused by the dynamic wetting conditions which are resulted from the variation of the environment around Sn cores, such as the interface curve and pressure. It is worth mentioning that no clear cavity could be observed aer Sn cores le their previous central positions during heating, which probably can be explained by the structural deformation and viscous ow of SiO 2 under high-dose electron beam irradiation. [20][21][22][23] This structural deformation of the SiO 2 shell is clearly observed in Fig. 2b and c during heating (300 C) under electron beam irradiation (37.3 pA cm À2 ). Fig. 2b shows the breakage of two connected Sn@SiO 2 core-shell nanoparticles. The connection area existed initially between these two Sn@SiO 2 core-shell nanoparticles (Fig. 2b, t ¼ t 0 , marked by a green arrow), and then it stretched out in length shells by contrast (Fig. 2c, t ¼ t 0 ). As heating progressed, these boundaries became blurred (Fig. 2c, t ¼ t 0 + 60 s) and then disappeared (Fig. 2c, t ¼ t 0 + 108 s), indicating connection between the SiO 2 shells and merging of the two individual particles (ESI Movie 2, see the ESI †). It is reasonable to believe that the deformation and viscous ow were not induced by simply heating, since in our case, the heating temperature 300 C is far below the $0.7 T g (glass-transition temperature) of SiO 2 (T g ¼ 1373 K) and cannot cause viscous ow of SiO 2 for a signicant deformation rate. 23 In addition, the temperature increase caused by energy transfer from the electron beam under our conditions is also far below 0.7 T g according to the calculation given by Zheng et al. 20 In order to minimize the effect from electron beam irradiation, Sn@SiO 2 core-shell nanoparticles distributed in other places, which were far from the focus area for movie recording during the heating process, were also observed aer cooling down to 25 C. Reduced deformation is observed compared to the SiO 2 structures in the focus area, whereas the hollow nanoparticles still formed ( Fig. 1e), which indicated that electron irradiation is not necessary for the formation of hollow structures in our case.

Cavity formation induced by vanishing of liquid-state Sn cores
From the above discussion, the observed dynamic movements and vanishing of Sn cores show the possible occurrence of Sn diffusion in the SiO 2 structure. Fig. 3 shows the in situ observation of two different pathways, mode I and mode II, for the disappearance of Sn cores embedded in SiO 2 , providing a visual demonstration of how the cavity forms inside the silica structure induced by increasing the temperature in a vacuum. Fig. 3a shows the vanishing process in mode I of Sn cores inside the SiO 2 structure at 300 C in a vacuum, captured from the movie (ESI Movie 3, see the ESI †). This process is illustrated in Fig. 3c (8 stages marked by numbers 1-8). The Sn core with a size of 38.5 nm was embedded in the deformed continuous SiO 2 structure which was derived from the interconnected SiO 2 shell, induced by electron beam irradiation (37.3 pA cm À2 ) as discussed above. At 300 C, this Sn core rst melted to a liquid droplet (stage 1 in Fig. 3c, outlined with an orange dashed line) and aer heating, showed lower contrast compared with the asformed ultra-small particles ($2 nm) distributed around it, indicating that the Sn core did not exist as a lled spherical droplet but a balloon-like spherical liquid layer that adhered on the inner surface of the cavity in the SiO 2 structure (Fig. 3a, t ¼ t 0 ) (stage 2 and 3 in Fig. 3c; the layer is shown by using a green dashed line). It is reasonable to consider that uniform diffusion of Sn may occur throughout the outward directions perpendicular to the Sn-SiO 2 interface (shown as white arrows in stage 2 and 3 in Fig. 3c). As the diffusion proceeded, the liquid Sn layer decreased in thickness (stage 3 in Fig. 3c) and then broke with the appearance of a hole (Fig. 3a, t ¼ t 0 + 5 s) (stage 4 in Fig. 3c). The balloon-like spherical liquid layer turned to a bowllike structure and then vanished quickly with the decrease in the area of the Sn-SiO 2 interface (Fig. 3a, t 0 + 8 s-t 0 + 11 s) (stage 4-7 in Fig. 3c), nally forming a cavity in the SiO 2 structure (Fig. 3a, t 0 + 12 s) (stage 8 in Fig. 3c). In addition, the edge between the single and overlapped liquid Sn layer makes it easy to distinguish by contrast (Fig. 3a, t 0 + 5 s-t 0 + 8 s). Moreover, the as-formed cavity also shrunk with the further deformation of the SiO 2 structure under electron beam irradiation (Fig. 3a, Under weaker electron beam irradiation (2.7 pA cm À2 ), another type of vanishing process of the Sn core was observed at 300 C in a vacuum, named mode II, as shown in Fig. 3b, which is captured from the movie (ESI Movie 4, see the ESI †). Fig. S5a † shows that a Sn@SiO 2 core-shell nanoparticle (core diameter: ca. 90 nm and shell thickness: ca. 8 nm) (blue arrow) with a small inner cavity and numerous hollow-structured SiO 2 nanoparticles surrounding it was selected. The whole hollowing process is illustrated in Fig. 3d, in which Sn cores are outlined with orange dashed lines. In the beginning, a part of the liquid Sn core was already diffused out, forming a cavity beneath the SiO 2 shell (Fig. 3b, t ¼ t 0 ) (stage 2 in Fig. 3d), and then the remaining Sn core decreased in size with time (Fig. 3b, t 0 + 60 st 0 + 479 s) (stage 2-4 in Fig. 3d) until the Sn core disappeared completely (Fig. S5b, see the ESI †) (stage 6 in Fig. 3d). In this process, the free surface of the Sn core, which was initially concave, turned nearly at and nally became convex. According to the observation, Sn may diffuse out through the interconnected SiO 2 shell structure (Fig. S5a, see the ESI †) since no leakage was observed and no Sn particles formed outside during our observation. Moreover, the wetting angle that is shown in Fig. 3b at t ¼ t 0 + 60 s was less than 90 and kept changing as heating proceeded, displaying dynamic wetting conditions on the curved inner surface of the shell. Combined with the results from Fig. 3a, this indicates that the embedded Sn droplets can completely wet the inner surface of the SiO 2 shell in the asprepared Sn@SiO 2 core-shell structure.
This cavity formation realized by vanishing of Sn cores through simply raising the temperature could be a facile method to create a hollow structure. Several previous studies showed in situ observation of core-removal hollowing processes including using the Kirkendall effect, 14,15,24 local etching of cores, 16 evaporation of cores, 17 draining away of liquid cores, 25 etc. Our observation showed a core-removal hollowing process by diffusion of liquid-state metal cores inside the shell structure, which has not been reported to the best of our knowledge.
It is worth mentioning that, during the heating process, ultra-small liquid-state Sn nanodroplets (diameter < 5 nm) appeared and were distributed in the SiO 2 structure accompanied by the disappearance of Sn cores (Fig. 3a). To collect more information on this phenomenon, a SiO 2 sphere distributed with ultra-small nanodroplets and attached on a tungsten wire was observed during heating (ESI Movie 5, see the ESI †). The captured images of morphology evolution are shown in Fig. 4a. During the heating process, the SiO 2 structure kept shrinking with shape transformation from a sphere to a spherical cap (t 0t 0 + 107 s). Meanwhile, no change in the area of the contact interface (red double arrow, t 0 + 4 s and t 0 + 104 s) was observed.  To show this shrinkage more clearly, a blue dashed line triangle was drawn with vertices A, B and C located at the positions of three characteristic Sn nanodroplets (t ¼ t 0 ). The change in size and shape of these triangles shows the shrinkage of SiO 2 , and the relative positions of these triangles indicate that the SiO 2 structure did not move but clung to the tungsten wire during the observation. The mean diameter of distributed Sn nanodroplets inside SiO 2 in each image is measured to show the growth of the as-formed nanodroplets (Fig. 4b), which increased from 1.8 nm to 2.9 nm. The corresponding size histograms are shown in Fig. S6. † Notably, during our observation, bigger Sn nanodroplets grew with time while small ones gradually disappeared, resulting in a decrease in the total number of Sn nanodroplets (Fig. 4c). This phenomenon indicates the occurrence of Ostwald ripening which is the result of the reduction in interface energy by the diffusion process inside the SiO 2 structure. 26,27 Controllable nanostructure of Sn@SiO 2 To further investigate the formation of the hollow structure via Sn diffusion in the SiO 2 structure, Sn@SiO 2 core-shell nanoparticles with various shell thicknesses (3.0 nm, 6.8 nm, and 29.0 nm) are synthesized ( Fig. 5b-d). Fig. 5a shows that the mean diameter of the original Sn nanoparticles used for the cores of Sn@SiO 2 is 16.9 nm. Fig. 5e shows the narrow size distribution of Sn nanoparticles and Sn@SiO 2 nanoparticles. To form the hollow SiO 2 structure, these three Sn@SiO 2 samples were transferred to a sealed aluminum pan and heated in a differential scanning calorimeter (DSC) under xed heating conditions (from 25 C to 300 C, ramping rate: 10 C min À1 , hold for 30 min at 300 C, naturally cool down to 25 C). The corresponding DSC results of the heating experiments are found in Fig. S7 (see the ESI). † In all samples, the endothermic peaks located at around 232 C indicate the melting of Sn cores embedded inside the SiO 2 shell. The melting temperature is determined from the intersection point of the initial base line and auxiliary line drawn through the linear section of the peak slope as shown by the black dashed line in Fig. S7. † 28 The melting temperatures of Sn cores embedded in the SiO 2 shell with shell thicknesses of 3.0, 6.8 and 29.0 nm are 228.0, 228.5 and 228.5 C, respectively, which are lower than the melting temperature of bulk Sn (231.9 C) due to the size-dependent melting effect. According to the model provided by Lai et al., the melting temperature of Sn nanoparticles with a diameter of 16.9 nm is 220.9 C, which is lower than our measured value. 19 It is believed that the SiO 2 matrix environment may inuence the surface atom mean-square displacement (msd) of the Sn core particle, thus affecting the melting temperature of Sn. 29 The endothermic peaks located at 199.6, 191.6 and 169.4 C were found during the initial heating of the as-prepared Sn@SiO 2 core-shell nanoparticles shown in Fig. S7a, † which may come from the desorption of residual organics and water. In comparison, the DSC curves of Sn@SiO 2 aer the heating experiment only show the peaks from the melting of Sn cores (Fig. S7b †). Aer heating, Sn@SiO 2 with a 3.0 nm SiO 2 shell was transformed into the yolk-shell structure as shown in Fig. 5f and i, in which most of the Sn cores remain inside the SiO 2 shell but decrease in size. With a thicker shell (6.8 nm), the Sn@SiO 2 shown in Fig. 5g and j contains a much smaller Sn core aer the heat treatment. A big proportion of the SiO 2 structure became hollow, and part of it had small Sn cores retained in it. As for the Sn@SiO 2 with the thickest SiO 2 shell (29.0 nm), no Sn cores are found aer heating, only the hollow SiO 2 structure remains. Based on these results, it is found that a larger proportion of the metal Sn core in core-shell Sn@SiO 2 nanoparticles with a thicker SiO 2 shell moves out under the same heating conditions. One possible reason for this phenomenon seems to be that the as-prepared SiO 2 shell has the capacity for xing a certain amount of the Sn element which comes from the melted Sn cores by heating above the melting point of Sn. To collect more evidence for this hypothesis, Sn@SiO 2 core-shell nanoparticles with various core sizes but uniform SiO 2 shells with a thickness of 29.0 nm are synthesized. Aer the heating process under the same conditions, we found that the morphologies of the as-obtained nanoparticles show an evolution from the solid Sn@SiO 2 core-shell structure to the yolk- shell structure and hollow structure (Fig. 6a). These three structures can be distinguished by using the volume ratio (R re ) between the residual Sn core and the original Sn core: R re ¼ 1 for core-shell nanoparticles, 0 < R re < 1 for yolk-shell nanoparticles and R re ¼ 0 for hollow nanoparticles. Fig. 6b shows that when the shell thickness is xed, nanoparticles with higher R re will be obtained with a bigger size of the original Sn core aer the heating process. That is to say, Sn@SiO 2 core-shell nanoparticles with relatively larger Sn cores may retain a greater proportion of the Sn core aer the heating process. The volume of the residual Sn core was calculated by assuming it to be a spherical cap with the method given in the ESI. † 30 Combined with the results from Fig. 5, we found that the amorphous SiO 2 shell that was fabricated through the wet chemistry method in this case provides space with a certain capacity for xing the Sn element aer heating above the melting temperature to the liquid state. Silica that was made from wet chemistry methods generally contains physically adsorbed water molecules which are connected to the hydroxyl groups on the surface of the silica structure by hydrogen bonding, based on Zhuravlev's models. 31 With this consideration, as discussed in Fig. 1f, SnO 2 is observed aer heating in a vacuum at 300 C. This formation of SnO 2 may be contributed by the interactions between liquid state Sn and the physically absorbed water molecules and hydroxyl groups on the surface of silica, 31-33 as shown in the following equation: The as-formed SnO 2 can easily get attached on the silica surface with hydroxyl groups. This phenomenon is different from that in our previous study, in which Sn nanoparticles were conned in porous silica spheres. 32 In this case, the SiO 2 matrix structure was pre-calcined, in which physically adsorbed water was removed, and aer melt-freeze cycles of Sn nanoparticles, no morphology change occurred. During the heating process, as the physically adsorbed water is gradually consumed, the oxidation reaction of Sn tends to stop. Moreover, the as-formed SnO 2 may ll and block the channels for the movement of liquid state Sn, thus keeping the rest of the Sn core unmoved so that the half-core or yolk-shell Sn@SiO 2 structure can be formed. This yolk-shell structure formation indicates that the amount of Sn exceeds the Sn containing capacity of the shell, making the shell "saturated". To quantify this capacity, the volume ratio (R vs ) between the as-formed void (regarded as equal to the volume of moved-out Sn) and SiO 2 shell aer the heating process is used here, which is determined from the amount of the Sn element distributed in a unit volume of the SiO 2 shell. Fig. 6c shows the distribution of R vs , which is calculated using the measured data from more than 200 Sn@SiO 2 yolk-shell nanoparticles. The mean value of R vs is calculated to be 2.74, based on which the maximum diameter of the original Sn core for the formation of the hollow structure with a SiO 2 shell thickness of 29.0 nm is calculated to be 17.9 nm.
High-angle annular dark-eld scanning transmission electron microscopy (HAADF-STEM), energy dispersive spectroscopy (EDS) elemental mapping and X-ray photoelectron spectroscopy (XPS) were used for investigating the element distribution of the as-formed hollow SiO 2 structure. The HAADF-STEM image in Fig. 7a shows a high Z (atomic number) contrast of the inner surface in the hollow SiO 2 structures. Fig. 6 (a) After heating in a DSC (loaded in a sealed Al pan and heated from 25 C to 300 C at a ramping rate of 10 C min À1 ), the Sn@SiO 2 core-shell nanoparticles are transformed into yolk-shell (1-15) and hollow nanoparticles (16), showing an evolution from core-shell to yolk-shell and hollow nanoparticles. Sn@SiO 2 core-shell nanoparticles with the same shell thickness (29.0 nm) but various core diameters (20-59 nm), shown in a-2 as "d", are used as starting materials. The scale bar is 50 nm. (b) Volume ratio between the residual Sn core and the original Sn core (R re ) obtained by measuring more than 200 particles with various diameters from 13.3 to 59.7 nm. The increasing ratio of the remaining Sn core with a bigger original core diameter indicates that Sn@SiO 2 nanoparticles with bigger Sn cores will retain more Sn after heating, showing controllable Sn diffusion based on size difference. (c) Histogram of volume ratios between the as-formed void and shell (R vs ) which is close to Gaussian distribution. The mean value of R vs is 2.74 AE 0.58, indicating a possible Sn containing capacity of the SiO 2 structure. The volume of the residual Sn core was calculated with the method given in the ESI. Meanwhile, many bright spots in the shell area are found, indicating the existence of the embedded Sn element which has a higher Z contrast than SiO 2 ( Fig. 7a and S8d †). The elemental map in Fig. 7b shows that the Sn element is uniformly distributed inside the SiO 2 hollow structure. During ex situ observation, similarly, the deformation of SiO 2 is found: shrinkage occurred, and the surface became rough aer electron beam focusing ( Fig. S8a and b †). Fig. S8e † and the inset image show the HRTEM image and its corresponding fast Fourier transform (FFT) pattern, respectively. The interplanar distance is calculated to be 0.343 nm, corresponding to the (110) crystal planes of tetragonal SnO 2 (JCPDS no. 41-1445) which is in good agreement with the SAED results shown in Fig. 1f and S4. † To check the Sn distribution on the surface and near surface area (3-10 nm), an XPS wide scan was performed on Sn@SiO 2 coreshell nanoparticles before and aer heat treatment in a DSC (Fig. 7c). Before heat treatment, no signal from Sn was detected, while a strong signal from Sn appeared when testing the Sn@SiO 2 core-shell nanoparticles aer heat treatment, showing strong evidence for outward diffusion of Sn. In addition, the XPS narrow scan (Fig. S9 †) shows that most of the Sn in the hollow structure existed as SnO 2 , 34 which is in good agreement with the previous discussion. It must also be mentioned that In-Sn@SiO 2 (In-Sn alloy nanoparticle coated with SiO 2 ) nanoparticles were prepared by a similar method (Fig. 8a-c). The In-Sn alloy has an even lower melting temperature. We applied the same conditions of the heating process as for Sn@SiO 2 on In-Sn@SiO 2 . The hollow structure also appeared aer being heated above the melting temperature (Fig. 8d), indicating that the diffusion of both In and Sn occurred inside the SiO 2 structure, showing a potentially facile method for nanostructure control of low melting alloy (LMA) nanoparticles embedded inside SiO 2 (LMA@SiO 2 ).

Conclusions
In summary, controllable nanostructure evolution of Sn@SiO 2 from core-shell to yolk-shell and hollow has been achieved by the interaction between the liquid metal Sn and SiO 2 shell structure synthesized through a wet chemistry method. The formation of a cavity inside the SiO 2 structure and transformation of Sn@SiO 2 core-shell nanoparticles into hollow SiO 2 nanostructures by vanishing of liquid Sn cores inside Sn@SiO 2 was observed in real-time via in situ TEM. The xing of liquid state Sn inside the SiO 2 shell structure takes the role of removing the cores in core-shell nanoparticles thus creating the cavity for the hollow nanostructure. Importantly, based on our ndings, we demonstrated control over the morphologies of hollow/yolk-shell Sn@SiO 2 nanostructures by adjusting the volume ratio between the core and shell of starting Sn@SiO 2 core-shell nanoparticle precursors. Furthermore, our ndings provide important knowledge of cavity formation through heating low melting point metal nanoparticles inside a SiO 2 structure, as well as a potential tool for designing complicated hollow nanostructures in future studies.

Synthesis of Sn nanoparticles
The Sn nanoparticles were synthesized using a well-developed method. 18 25 mL of oleylamine was vacuum dried in a 50 mL three-necked ask at 140 C for 1.5 h and then naturally cooled down to 30 C. Anhydrous SnCl 2 (0.5 mmol) (min. 98%, Junsei) was added into the ask before heating to 140 C and holding for an extra 30 minutes to remove the moisture. Aerwards, the ask was heated to 190 C under an atmosphere of argon with constant stirring, followed by injection of 3.6 mmol lithium bis(trimethylsilyl)amide (LiN(SiMe 3 ) 2 ) (1.0 M in toluene, Aldrich) and a second injection of 0.6 mmol diisobutylaluminium hydride (DIBAH, 1.0 M in toluene, Kanto) within 10 s. The reaction solution became brown and then gradually turned black. The solution was stirred at 180 C for 1 h to ensure the near-complete reaction of the metal precursor and then cooled down to 50 C in an ice-water bath. 500 mL of dried oleic acid (OA, Junsei) were added before cooling down to 30 C. The obtained solution was washed with ethanol and centrifuged (8000 rpm, 5 min). The precipitate was redispersed in a solvent mixture (6 mL, volume ratio: OA : hexane ¼ 1 : 50) and then washed with ethanol. Finally, the puried precipitate was redispersed in 25 mL of hexane to form a stock solution of Sn nanoparticles. In addition, InSn 4 nanoparticles were prepared in a similar way by using InCl 3 mixed with SnCl 2 in a molar ratio of 1 : 4 as the precursor.

Synthesis of Sn@SiO 2 core-shell nanoparticles
To encapsulate Sn nanoparticles with a uniform SiO 2 shell, a modied reverse micro-emulsion method was used. A typical  4 nanoparticles after heating in a DSC (from 25 C to 300 C, 10 C min À1 , loaded in a sealed Al pan). 0.61 g (0.594 mL) of polyethylene glycol mono-4-nonylphenyl ether (n ¼ approx. 5, TCI, Tokyo, Japan) and 62 mL of deionized water (puried using a PureLabo system, Organo/ELGA, >18 MU). This emulsion was sonicated for 15 min to produce a good dispersion of the nanoparticles without aggregation. Aerward, 50 mL of TEOS (tetraethyl orthosilicate) were added under vigorous stirring (800 rpm). Aer 30 min hydrolysis and condensation of TEOS were triggered upon injection of 30 mL of an ammonia solution (min. 28.0%, Junsei, Tokyo, Japan). The mixture was stirred at 400 rpm for 10 h. Aerward, the obtained samples (Sn@SiO 2 core-shell nanoparticles) were washed with ethanol three times and nally dried in a vacuum at room temperature. In particular, the thickness of the SiO 2 shell can be easily tuned by changing the concentration of the Sn nanoparticle-dispersed stock solution.

In situ TEM observation
In situ TEM observations were carried out using a JEM-1000K RS TEM (JEOL) operated at 1000 kV. A wire-type heating holder (JEOL, EM-Z081834SWHH) was used. The wire lament was heated using a DC generated by dry cells. The lament temperature was xed by the calibrated current value. [35][36][37] The powder samples (Sn@SiO 2 ) were loaded on a tungsten wire and heated up to 300 C under vacuum. TEM images were recorded with a charge-coupled device with an exposure time of typically 0.5 s and 5 s. The electron beam current ux for strong electron beam conditions was 37.3 pA cm À2 , and that for weak electron beam conditions was 2.7 pA cm À2 , as measured with a faraday gauge.

Ex situ TEM observation
The ex situ TEM observation of Sn@SiO 2 was carried out using a conventional TEM (JEM-2000FX microscope, JEOL) operated at 200 kV and a high-resolution eld-emission gun scanning TEM (STEM, JEM-ARM200F Cold, JEOL) operated at 200 kV at room temperature and under a pressure of 10 À6 Pa.

Conflicts of interest
There are no conicts to declare.