Void nucleation, growth, and coalescence
In the experiments we observed the evolution of voids that was categorized into the nucleation, growth, and coalescence processes. It has been documented that the voids would be formed through the thermal heating28, local stress concentrators caused by extended defects (e.g., dislocations and grain boundaries)29, Kirkendall effect30 and atomic displacement induced vacancy zones31. The e-beam-induced temperature rise was normally in the range of several degrees32,33, indicating that heating did not dominate the formation of voids. The local stress concentrators and the Kirkendall effect might not mainly account for the void formation either, because no grain boundaries and/or diffusion couple with different rates were involved during the process. In the absence of material discontinuities, on the other hand, the direct transfer of e-beam energy to atoms during collision may be large enough to knock them out of their lattice sites, creating the vacancies so that the condensation of vacancy clusters induces the voids. On the basis of both energy and momentum conservation, the maximum kinetic energy EA transferred from an electron to an atom through collision is estimated using the following equation,
$$\:{\text{E}}_{\text{A}}\text{}\text{=}\text{}\text{561}\text{ε}\text{}\left(\text{ε}\text{}\text{+}\text{}\text{2}\right)\text{}\text{/}\text{}\text{A}$$
1
where \(\:\text{A}\) is the weight of atom and ε = E/(mc2), E is the original energy of electron in TEM with 200 kV accelerating voltage, m is the mass of electron, and c is velocity of light. For the typical 200 keV e-beam in TEM, the EA values were calculated to be 2.62 and 16.37 eV for Hg and S, respectively. Upon comparison to the vacancy formation energies Ev of 4.5 and 7.8 eV for Hg and S, it indicates that it was easier for S atoms to be knocked out than that for Hg atoms. Surprisingly, according to the EDS analysis the loss of Hg was more than that of S element (Supplementary Fig. 6), which was owing to the occurrence, movement, and evaporation of Hg droplets (Fig. 3). Such evolution of Hg droplets would further offer localized reduction in displacement energy for S, energetically favoring the vacancy clustering to facilitate the evolution of voids.
In the next stage, the dynamical evolution includes the simultaneous growth and coalescence of existing voids as well as the formation of few new ones. Compared to the nucleation of new ones, the small voids tended to add into the existing voids and to merge to form the large ones, which was probably driven by the reduction of the total energy of the system. Interestingly, unlike the direct coalescence of voids, e.g., the behavior reported in the Bi nanoparticles induced by heating, the coalescence between two neighboring voids in the HgS NCs was achieved by the bridge with ~ 18 nm (Fig. 3). Ultimately, the large voids turned into a long crack-like feature to further lower the total energy. Such feature preferentially propagated along the < 001 > direction of the crystal due to the fewer number of bonds bridging unit cells perpendicular to this direction. In this work, the structural evolution of the crystal tended to form the voids instead of the volume contraction so that the framework was maintained. This was probably due to the thermodynamic preference for reduction of the total surface energy of the voids greater than the crystal surface energy.
Formation, movement and coalescence of Hg nanodroplets
During the e-beam excitation of HgS semiconductor, a number of electron-hole (e–h) pairs are generated within an excitation volume. For a 200 kV e-beam with the typical current of 1.2 nA used here, the local rate of carrier pair generation was calculated to be approximately 3.10 × 105 pairs per second in the HgS NC (Supplementary Fig. 7, Supplementary Tables 1–2 and Supplementary Notes). The utilization of such generated electrons was then roughly estimated. As presented in Fig. 2, about three discernible Hg nanodroplets with a diameter of ~ 2 nm (~ 7.26 × 103 atoms) formed at 2 s while ten bigger nanodroplets (diameter of ~ 10 nm) appeared at 400 s which contained about 3.02 × 106 Hg atoms. Within these e-beam excitation periods, the number of generated electrons were 6.20 × 105 and 1.24 × 108 for 2 s and 400 s, respectively. Despite the rough estimation, it provided a pictorial understanding for the formation of Hg nanodroplets (Hg2+ + 2e \(\:\to\:\) Hg), namely, the formation of the 2 nm (~ 2.3%) and 10 nm (~ 4.9%) nanodroplets only consumed several percent of the reductive electrons, whereas most of the carriers were recombined in the material.
After the formation, those Hg nanodroplets exhibited relatively large movement speed of ~ 56 nm/s at the dose rate of 4340 e/Å2·s, in comparison to that of ~ 5–10 nm/s observed for the Bi nanodroplets34. Although the movement of metal nanodroplets has been rarely reported, several mechanisms about the driving forces originated from the gradient fields including temperature21, surface energy35, electric field36 and light field37 have been proposed for other common liquid droplets (e.g., water). Those factors might not be applicable or play the essential role for the phenomena observed in this work. It has also been shown that the ratchet surface leaded to the propulsion of the liquids, which was driven by the viscous force between the solid/liquid interface due to the Leidenfrost effect38. The HgS nanocrystals in our work adopted a saw-tooth morphological feature so that the Hg nanodroplets preferred to curve concavely near the tops of the ridges while presenting the convex characteristics elsewhere (Supplementary Fig. 8). In this scenario, we speculated that such ratchet surface may contribute to the fast movement of the Hg nanodroplets. During the e-beam-induced reduction of Hg, the nucleation of the nanodroplets occurred once the vapor pressure of Hg exceeded the saturated one. Such variation in the surface curvature would generate a pressure differential Δp between the concave ridge and the neighboring convex position, forming a net force to drive the motion of the nanodroplet (Supplementary Fig. 8). Such a pictorial understanding was again verified by the fact that the glide velocity of the Hg nanodroplets increased dramatically at the high e-beam dose rates since more Hg vapor was formed upon the irradiation (Fig. 3).
During the movement, the Hg nanodroplets would coalesce with a relatively large distance of ~ 5 nm through the bridges (Fig. 3), which was different from that observed in other systems. For example, the Bi nanodroplets underwent a sudden coalescence on the SrBi2Ta2O9 platelet39, while the fusion or coalescence of Au nanocrystals was realized by the nanochannel with a critical spacing of less than 1 nm40–42. Moreover, the coalescence of nanobubbles in liquids was observed to occur within a distance of ~ 2 nm between two bubbles43. In this work, the e-beam irradiation induced the reduction of Hg2+ to Hg on the HgS NCs. As the time elapsed, the accumulation of Hg atoms might create the atomic chains accordingly, resulting in the formation of relatively long bridges that finally leaded to the coalescence between the nanodroplets.
Mechanism for continuous ink-jetting behavior
It is known that the irradiation of water by the electron beam generally generates a variety of radiolysis products, including the common oxidative •OH, H2O2 and reductive solvated electron eh− species44. The steady-state concentrations of the radiolysis species were calculated and summarized (Supplementary Tables 3, 4, 5), presenting that the concentration of the oxidative species was higher than that of the reductive ones. Meanwhile, the oxidative species possess very high reduction potentials (Supplementary Table 6), which have been reported to play the vital role for etching the nanostructures (especially metals) in liquids31,45–48. Despite the inevitable existence of slight etching, the HgS semiconductor studied here demonstrated the distinct behavior from that was shown in the metal nanostructures (almost pure etching).
Different types of scavengers were designed to suppress a certain number of specific species so that the dominated ones for the ink-jetting phenomenon would be clarified. Figure 5a shows the sequential TEM images in the electrospun liquid cells with the addition of H2O2 (scavenger for reductive species). In this case, the e-beam-induced oxidative species would predominate. The HgS NCs were gradually etched and more cavity feature was formed with the increment of time. No ink-jetting phenomenon was observed even during the long irradiation for ~ 300 s. The above behavior was further validated in the thin carbon liquid cells with the same addition of H2O2 (Fig. 5b). Obvious contrast from the bubbles was observed, manifesting the liquid environment inside the carbon liquid cell (Supplementary Fig. 9). Similar etching phenomenon was also observed as the time elapsed to several hundreds of seconds. Additional ex-situ experiments were designed to distinguish the etching effect from either H2O2 or •OH (see Methods for details). Figure 5c depicts the TEM images of the HgS NCs treated mainly by •OH. The morphological change was seen at 10 min and it became obvious after 30 min. Most parts of the NCs were etched off after 180 min, resulting in the complete collapse of the initial bipyramid structure. In contrast, the HgS NCs with the addition of pure Co2+ or H2O2 presented no obvious morphological change at the same time scale (Supplementary Fig. 10). Upon comparison of the above results, it indicates that the •OH plays the dominant role in etching the HgS NCs.
Figure 5d shows the time-dependent TEM images of the ink-jetting phenomenon in the electrospun liquid cells with the addition of methanol (scavenger for oxidative species, Supplementary Movie 5). Similar to the structural evolution of HgS NCs in water, the voids arose rapidly and the ink-jetting phenomenon occurred consecutively. Distinctly, the first ink-jetting occurred at ~ 66 s, which was significantly shorter than that in the water condition (~ 314 s). Meanwhile, the interval time\(\text{∆}\text{t}{\prime}\text{}\text{=}{\text{}\text{t}}_{\text{n}}\text{-}\text{}{\text{t}}_{\text{n-1}}\) between each ink-jetting event also became shorter, e.g., the fifth jetting behavior was finished within only ~ 4 s. Given the nature of oxidative scavenger in the case of methanol, it suggests that the reductive electrons might play the key role for the continuous ink-jetting phenomenon. This was further evidenced by the fact that the Hg nanodroplets immediately nucleated under the reductive environment after the e-beam excitation for 2 s in the gas cells with the absence of •OH (Fig. 2). For the case of the liquid cells, the longer time needed to trigger the ink-jetting (Fig. 4, 338s) indicates that part of the reductive electrons was consumed by the oxidative species and the long-time accumulation of the sufficient electrons therefore allowed the subsequent ink-jetting.
Based on the above discussion, a competition exists for the reduction of Hg2+ to the formation of Hg droplets, i.e., the reductive electrons promote such behavior whereas the oxidative species inhibit it. Once the accumulated electrons reach the threshold value during such competition process, the ink-jetting behavior occurs. Since the reductive electrons are consumed in each reduction-involved ink-jetting process, a certain amount of time would be thereby required to accumulate enough electrons for the next jetting behavior. Figure 5e shows the relationship between the e-beam irradiation time and the ink-jetting frequency under different liquid conditions. At each frequency, the irradiation time needed to trigger the ink-jetting under the addition of methanol was shorter than that in the water. The interval time between two sequential ink-jetting processes is quantitatively displayed in Fig. 5f. Compared to the interval time of ~ 11–22 s in water, the time span under the addition of methanol was significantly shortened (~ 3–6 s). The phenomenon was ascribed to the consumption of a certain number of oxidative species through the addition of methanol. In this scenario, more electrons had the chance of being involved in the reduction process, leading to the shortened time for triggering the ink-jetting behavior.
Schematic diagram of nanodroplet dynamics in gas and liquid cells
The overall evolution pictures of Hg Nanodroplets in the gas and liquid cells are schematically summarized in Fig. 6. In the gas cells (Fig. 6a), the direct transfer of e-beam energy to the Hg and S atoms knocks them out of the lattice sites, creating a large number of the vacancies. The condensation of the vacancy clusters induces the voids, which is also accompanied by the appearance of Hg nanodroplets due to the reduction of Hg2+ to Hg. Subsequently, the voids gradually grow up after coalescence through the bridges, which further turn into the crack-like structure preferentially along the < 001 > long-axis direction. Meanwhile, the Hg nanodroplets move rapidly, and come into coalescence through the bridges when they are close each other, forming the bigger ones accordingly. With the continuous coalescence process, the large droplets are finally formed and stay relatively stable on the substrate.
In the case of liquid cells (Fig. 6b), the voids also arise rapidly in the preliminary stage upon the e-beam excitation, which are evolved into the large ones as the time elapses. Compared to that in the gas cells, distinctly, some smaller Hg nanodroplets appear in the early stage, but without the apparent growth. After the long-time excitation so that the accumulated electrons achieve the threshold, the ink-jetting behavior begins to occur. Such ink-like feature spread, and then becomes less obvious at the later stage. After a certain time of additional e-beam excitation, the second ink-jetting appears once the reaccumulation of the reductive electrons becomes sufficient for triggering such behavior. In this regard, the continuous the ink-jetting behaviors become feasible, with the interval time of two neighboring ones depending on the competition between the reductive electrons (accumulation) and the oxidative species (consumption of electrons). Ultimately, a stable liquid Hg layer is formed after the several ink-jetting processes at a certain dose rate.
In summary, the distinct evolution dynamics of the Hg nanodroplets mediated at the solid-gas and solid-liquid interfaces were directly visualized and statistically investigated by in-situ TEM. Upon exposure to e-beam in the gas cells, the voids and Hg nanodroplets nucleated and grew up, which were then coalesced into the large ones through the nanobridges. The voids could be further evolved into the crack-like structure preferentially along the < 001 > direction of the HgS solid substrate. The Hg droplets moved fast at the solid-gas interface and finally became relatively stable at each dose rate, which was distinct from the evaporation behavior observed at the solid-vacuum interface. In contrast to the typical behavior of the voids and Hg nanodroplets observed at either solid-gas or solid-vacuum interface, statistical results revealed the occurrence of the ink-jetting phenomenon in the liquid cells. Two competitive factors governed the interval time between the neighboring ink-jetting events, namely, the reductive electron species facilitated the shortening of such period while the oxidative species like •OH acted in the opposite manner. These phenomena were further verified by the ex-situ experiments with the addition of additional H2O2 or methanol. Our results contribute to advance a fundamental understanding of the full and unique evolution picture of liquid metal nanodroplets at nanoscale solid-gas and solid-liquid interfaces, and provide the experimental approach of potentially modulating droplets behavior with controllable functionality through interface engineering.