Dynamics of Formation of a Vapor Nanobubble Around a Heated Nanoparticle.

We study the formation of a nanobubble around a heated nanoparticle in a bulk liquid by using molecular dynamics simulations. The nanoparticle is kept at a temperature above the critical temperature of the surrounding liquid, leading to the formation of a vapor nanobubble attached to it. First, we study the role of both the temperature of the bulk liquid far away from the nanoparticle surface and the temperature of the nanoparticle itself on the formation of a stable vapor nanobubble. We determine the exact conditions under which it can be formed and compare this with the conditions that follow from a macroscopic heat balance argument. Next, we demonstrate the role of dissolved gas on the conditions required for nucleation of a nanobubble and on its growth dynamics. We find that beyond a certain threshold concentration, the dissolved gas dramatically facilitates vapor bubble nucleation due to the formation of gaseous weak spots in the surrounding liquid.

Introduction

The formation of nanobubbles around heated nanoparticles is a phenomenon that has technological relevance in applications such as cancer treatment,[1,2] catalytic reactions,[3−6] and solar energy conversion.[7,8] Nanoparticles can be heated either by exposing them to a laser pulse with a wavelength corresponding to their plasmonic resonance[1,2] or even by direct sunlight.[7,8] Exposure to high-power lasers or solar radiation raises the temperature of the nanoparticle to hundreds of kelvin,[9] leading to local heating of the liquid in their proximity to very high temperatures and eventually to the formation of a vapor nanobubble around them. These vapor nanobubbles, also known as plasmonic nanobubbles, are not only claimed to be potential candidates for efficient solar energy conversion but are also becoming a very useful tool for the therapeutic applications in cancer treatment.[1,2] In this application, the nanoparticles are engineered in such a way that they can selectively attach to the membrane of tumor cells exposing them to the high-power laser pulse and can generate nanobubbles that mechanically damage the cell membrane to destroy the tumor cells.[1,2] There have been numerous other examples where the fundamental understanding of plasmonic nanobubbles pave the way to further exploit them for a wide range of applications.[10] The formation of the nanobubble can be seen as heterogeneous nucleation where the first-order phase transition occurs at the nanoparticle present in the bulk phase. This nanoparticle facilitates the liquid–vapor phase transition. Understanding the exact mechanism of generation of plasmonic nanobubbles is also important from the fundamental point of view, since it can reveal interesting phenomena relevant to heat transfer and phase change at the nanoscale in general.[11]

The formation of a vapor nanobubble around a heated nanoparticle is a consequence of a highly out-of-equilibrium situation where the temperature gradients in the liquid can reach up to hundreds of kelvin per nanometer. The formation of a vapor nanobubble is an extremely transient process in which the bubble forms and collapses within a nanosecond. Such small time and length scales make it ideal to study the nanobubble formation by molecular dynamics (MD) simulations. Sasikumar and Keblinski[11] studied the bubble formation around a heated nanoparticle with the help of MD simulations and reported the formation of a vapor nanobubble when the temperature of the liquid in the vicinity of the nanoparticle reaches ∼90% of the critical temperature. Lombard, Biben, and co-workers[12−15] used the hydrodynamic phase field model based on free-energy density to study the threshold and kinetics of vapor bubble generation as a function of the size of nanoparticles and laser power. However, none of these studies considered the role of dissolved gas in the threshold and dynamics of nanobubble formation.

Recent experiments by Wang et al.[16] showed that dissolved gas can change the long-term growth dynamics due to diffusion of gas into the bubble. It can be argued that dissolved gas should not affect the initial explosive growth which is driven by the phase change of the liquid under extreme thermal gradients, as the energy for the latent heat of vaporization is provided by the thermal diffusion, which is orders of magnitude faster than the mass diffusion of the gas in the liquid. However, dissolved gas can play a role by changing the vapor–liquid phase diagram of the system, which will influence the threshold for vapor generation. Macroscopically, it is known that homogenous nucleation of bubbles can occur at lower temperature for increased gas concentration in the liquid.[17−19] Therefore, we expect the dissolved gas not to affect the nanobubble growth dynamics but rather to affect the threshold for the formation of a vapor nanobubble around a heated nanoparticle.

To investigate the conditions for bubble nucleation and the role of dissolved gases, we perform simulations of a pure liquid around a heated nanoparticle and determine the conditions for the nucleation of a vapor nanobubble and its growth dynamics. The temperature of the nanoparticle (TNP) is kept at a constant value that is much higher than the critical temperature (Tc) of the liquid. The temperature of the liquid “far away” from the nanoparticle surface is also kept constant by having an isothermal wall, the temperature of which is much lower than Tc (see Figure ). Our results are compared to theoretical predictions based on a macroscopic heat balance.

Figure 1

Schematic of a vapor nanobubble formed around a heated nanoparticle. Initially, the temperature of the whole system is constant and equal to a value TW. At time t = 0, the temperature of nanoparticle TNP is suddenly raised to a temperature far above Tc while keeping the wall temperature fixed at TW.

Simulation Methodology

Molecular dynamics (MD) simulations were performed with the aid of the open source code GROMACS[20] to simulate the formation of a nanobubble around a heated nanoparticle. We used three types of molecules in our simulations: first acts as liquid (L), second as solid (S) and third as gas (G). The nanoparticle and wall are modeled by a collection of solid particles (S) arranged in an face-centered cubic lattice and connected with the neighboring particles by nonlinear elastic springs that act as chemical bonds. These particles can vibrate around their equilibrium positions while interacting with liquid and gas particles. They also interact with other solid particles by an interaction potential. For convenience, we refer to the type of particles that are predominantly in the liquid phase as “liquid particles” and in a similar way, we use the terms “solid particles” and “gas particles”. The nonbonded interaction between the particles is described by a Lennard-Jones potentialwhere ϵ is the interaction strength between particles i and j and σ is the characteristic size of the particles. The potential is truncated at a relatively large cutoff radius (rc) of 5σLL, where σLL is the size of the liquid particles. The particles in the nanoparticle and the solid wall are connected by the finitely extensible nonlinear elastic (FENE) potential[21] as given bywhere for the value of the spring constant ks, we used ks = 30ϵSS/σ2 and for rk, we used 1.5σSS, which are consistent with the previous MD studies on nanobubble generation around a heated nanoparticle.[11,22,23] The reason for using the FENE bond potential (which is normally used for coarse-grained polymer simulations) to connect the solid particles is that this allows the nanoparticle to be heated to arbitrarily high temperatures without melting. There are around 135 000 moving LJ particles (liquid and gas) in the simulation box, whereas the nanoparticle consists of around 1400 LJ particles and the wall consists of 50 000 LJ particles. The time step for updating the particle velocities and positions was set at , where m is mass of the solid particles. The mass of all Lennard-Jones particles is set as 20 Da or atomic mass unit. The time step was chosen such that its value is sufficiently smaller than the shortest time scale in the system.[24]

We now explain the choice of the boundary conditions; see Figure . First, in z-direction, we put in walls of constant temperature TW to thermally equilibrate the system, as otherwise the mean temperature would keep on increasing. The input of thermal energy of the hot nanoparticle with TNP > TW must be balanced. Both lower and upper wall are kept at constant temperature TW. The wall thickness is large enough so that the LJ particles are equilibrated and no artefacts from too thin walls arise. Next, on the choice of boundary conditions in x, y-direction (“lateral”-direction, see Figure : we chose periodic boundary conditions for computational efficiency). This is possible and reasonable, provided that the particles at the edge of the box are so far from the hot particle in the center that they do not feel (or at least hardly feel) any thermal or density gradient caused by the heating in the center. As seen from the (latter) Figures and 4, this is indeed the case. Then, the particles leaving the box on the right-hand side and entering it on the left-hand side (or vice versa) are sufficiently equilibrated and can be seen as being “at infinite distance”.

Figure 3

Radial density profile of liquid molecules around a heated nanoparticle fitted to eq . The black line indicates the radius of the nanobubble Rb and the shaded region depicts the width of the interface, w.

Figure 4

Temperature variation along the axis parallel and perpendicular to the wall originating from the center of the nanoparticle. The temperature profile is calculated by solving the heat equation numerically and compared with MD simulations and the spherically symmetric result. The black curve represent the analytical result obtained for perfect spherical symmetry; the red and green curves are the numerical result obtained from COMSOL for the geometry that has been used in MD simulations.

Initially, the system is equilibrated at constant temperature by coupling the whole system to a constant temperature bath, equal to TW. At time t = 0, liquid and gas particles are disconnected from the temperature coupling while the wall remains connected to the temperature coupling at TW while the temperature of the nanoparticle is set to TNP by coupling it to a separate thermostat. So the gas/liquid is free to set its “own” temperature, constrained by the fixed temperatures of the wall and the nanoparticle. Two separate velocity-rescale thermostats have been used to maintain both constant TW and constant TNP with a time constant of 1 ps. TW is chosen in such a way that it should be less than the critical temperature of the fluid, whereas TNP is varied in the range such that its minimum value is always much higher than the critical temperature of the fluid. The pressure is kept constant at p/pc = 0.308 (where pc is the critical pressure of the Lennard-Jones particles) by semi-isotropic pressure coupling, which means that the simulation box can expand or contract only in the z-direction to keep the pressure constant. Berendsen pressure coupling has been used to maintain the constant pressure with compressibility equal to 4.5 × 10–5 bars–1 and time constant as 1 ps. The complete set of Lennard-Jones parameters that we used in our simulations are given in Table . The parameters for L and G particles are chosen in such a way that the critical temperature of the L particles should be much higher than the highest wall temperature used in the system and for G particles the critical temperature should be much lower than the lowest wall temperature used in our simulations. The typical system size is 20 × 20 × 22 nm3 in x-, y-, and z-direction, respectively, with the z dimension changing during the simulation to maintain constant pressure.

Table 1

Value of Various LJ Parameters Used in the MD Simulations

i–j	σij (nm)	ϵij (kJ/mol)
L–L	0.34	3.0
G–G	0.5	1.0
S–S	0.30	3.0
L–G	0.42	1.73
S–L	0.32	3.0
S–G	0.42	1.0

In Figure , we show a typical profile of a vapor nanobubble around a heated nanoparticle for both a single-component liquid and a liquid with dissolved gas in it. The average density field of liquid particles in radial direction around the nanoparticle was calculated as a function of time to investigate the formation of a nanobubble. A nanobubble is considered to form if the density of liquid particles in the vicinity of the nanoparticle is less than the critical density of the liquid.[15] The radius Rb of the nanobubble was obtained by fitting the relationto the radial density profile, where ρL is the liquid density, ρV the vapor density, and w the width of the liquid–vapor interface. Figure shows the typical radial density profile of liquid particles fitted to eq . The density in both cases was obtained by averaging over 100 simulation snapshots with a time gap of 1000 steps between each snapshot. Note that the density of the molecules in radial direction has been calculated in two different ways: first, by assuming the center of the nanoparticle at the origin and second, by assuming the center of the nanobubble at the origin. The centers of the nanoparticle and the nanobubble are not exactly located at the same position as the nanoparticle can move a little bit inside the nanobubble. However, the position of the nanoparticle always fluctuates around the center of the nanobubble. So for better accuracy, the center of the nanobubble is used to calculate the radius while the center of the nanoparticle is used to calculate the density around the nanoparticle to examine the formation of the nanobubble. By “center of nanoparticle” we mean the center of mass of the nanoparticle. The center of the nanobubble is calculated by calculating the center of mass of voids present inside the nanobubble. In doing so, the whole simulation box is divided into a fine three-dimensional grid and the grid points that do not contain any LJ particle are referred to as “voids”.

Figure 2

Typical snapshot of a vapor nanobubble formed around a heated nanoparticle for a single-component liquid (left) and with a gas dissolved in the liquid (right). In this case, kBTNP/ϵLL is equal to 5.54 and kBTW/ϵLL is equal to 0.97 and the mole fraction of gas molecules xg for the snapshot on the right is set as 0.011. These snapshots are taken at 400 ps where the systems were at the steady state or “quasi equilibrium”.

Macroscopic Modeling

Nanobubble Formation

In this section, we describe a framework to understand the conditions that lead to the formation of a vapor nanobubble around a heated nanoparticle from a macroscopic viewpoint. As a criterion for the appearance of the nanobubble, we use the condition that the liquid temperature in the neighborhood of the nanoparticle equals the spinodal temperature Tspin.[11][11] (A calculation of spinodal temperature Tspin for the single and binary mixtures of Lennard-Jones molecules is shown in the Supporting Information.)

In our MD simulations, it is shown that in the case of a vapor bubble, a layer of liquid is always formed around the nanoparticle due to high attractive force from the densely packed molecules in the nanoparticle. To be consistent with our simulations, we therefore choose the criterion that the liquid temperature at a distance 2σLL from the nanoparticle surface becomes equal to the spinodal temperature Tspin. We will next show that for our parameter setting, the temperature Ts′ of the fluid at radius Rp′ = Rp + 2σLL is very close to the temperature Ts at the surface of the nanoparticle. Conservation of energy gives the followingwhere κL′ is the thermal conductivity of the liquid. The nanoparticle temperature TNP and the liquid temperature at the nanoparticle surface Ts are related in terms of GSL, the interfacial thermal conductance of the solid–liquid interface, also known as Kapitza conductance.[25] A similar statement relating the energy flow between the surface of radius Rp′ and the cold wall of radius RW at temperature TW leads to the followingFrom the previous two equations, it followsWith this result we findWith the present parameter values, the term on the right-hand side is found to be around 0.01–0.02 so that Ts and Ts′ are essentially equal.

In the previous considerations, we have assumed spherical symmetry which is clearly not fulfilled, as shown in Figure . Moreover the cold walls are on the top and bottom but not on the sides. To test the error associated with the spherical symmetry assumption, we numerically solved the heat conduction equation for exactly the same geometry that we used for the MD simulations, with appropriate boundary conditions, using an FEM-based commercial solver, COMSOL.[26]Figure shows the temperature profile along lines originating from the center of the nanoparticle, one perpendicular to the wall along the z axis and another parallel to the wall along the y axis. The black line shows the 1/r behavior from the analytical solution for a spherically symmetric system. The data points in the figure show the temperature from the MD calculation averaged over spherical shells concentric with the nanoparticle. The scatter close to the nanoparticle is due to statistical fluctuations, as indicated by the error bars. There are some obvious differences between the various results, yet small enough to justify the use of the spherical symmetry assumption.

A second assumption implicit in eq is that the thermal conductivity of the liquid κL does not vary with radial position, despite the large temperature gradient of hundreds of kelvin over a few nanometers distance. This assumption looks quite radical at the first glance due to the well-known dependence of the thermal conductivity of the Lennard-Jones liquid on the temperature and density.[27] A systematic validation of this assumption is shown in the Supporting Information.

Prediction of Maximum Bubble Radius

When a vapor nanobubble is formed around the heated nanoparticle, it grows for a while and reaches a steady state due to the finite size of the system. We can predict the steady-state radius again from the heat balance by assuming that at the liquid–vapor boundary, the temperature equals Tspin. At steady state, the heat coming out of the nanoparticle will get transferred across the vapor layer and exactly the same amount will be conducted through the liquid toward the cold wall. This heat balance can be written aswhere RSS is the steady-state radius of the nanobubble and q, the heat flux through the vapor phase, consists of two contributions: a conductive heat flux and a ballistic heat flux. The conductive heat flux qc is dominated by the solid-vapor conductance qc = GSV(TNP – Ts), where GSV is the solid-vapor interfacial conductance and Ts is the temperature of the vapor near the nanoparticle surface. The expression for the ballistic heat flux qb for a “Knudsen gas” is given by[14]where α is the thermal accommodation coefficient, ρs is the density of the liquid on the surface of the nanoparticle, and m is the mass of one liquid particle. Expression for the ballistic heat flux is the difference between the energy fluxes associated with the incoming and outgoing molecules from the nanoparticle surface. The thermal accommodation coefficient α is a dimensionless parameter that characterizes the probability with which the molecules stick or leave the nanoparticle surface. From our simulations, α is calculated as follows[28,29]where Tr and Ti are the temperatures of incident and reflected vapor molecules, respectively. The conductive flux in the vapor phase qc is found to be at least an order of magnitude less than qb, which is primarily due to the value of GSV, which is typically 20 times smaller than GSL.[14] Neglecting the contribution from qc, we can write eq asNote that in this expression, we have assumed radial symmetry in the temperature profile of the liquid, which is less justified here as the radius of the nanobubble is comparable to the distance between the liquid–vapor interface and the wall. Nevertheless, we still used eq to get a rough estimate of the maximum nanobubble radius and replaced the wall temperature TW with the temperature averaged over a sphere of radius TW, which is at most 10% larger than TW. Eq is solved using RSS for appropriate values for the parameters and compared with the RSS, as obtained from MD simulations.

Results and Discussion

Formation of a Nanobubble

Figure shows steady-state profiles of the liquid density and temperature around the nanoparticle obtained from MD simulations, all averaged over spherical shells. In this figure, r = 0 corresponds to the center of the nanoparticle, the surface of which is at r/σLL ∼ 5. The various lines correspond to different cold wall temperatures for a fixed nanoparticle temperature. The horizontal black lines indicate the critical density and critical temperature of the liquid. Note that initially, the liquid around the nanoparticle has uniform temperature and density while the temperature is equal to the wall temperature TW and the density varies according to TW. For example, when kBTW/ϵLL = 0.69, the initial density of the liquid is 0.78 and when kBTW/ϵLL = 0.97, the liquid density is 0.62. The large density values at the nanoparticle surface are due to the strong attraction that the very closely spaced nanoparticle molecules exert on the liquid molecules. We consider a nanobubble to have formed when the density of the liquid molecules falls below the critical density. In the case of Figure a, a nanobubble is considered to have formed for the conditions corresponding to the lowest line kBTW/ϵLL = 0.97. In this example, the minimum density is found at a distance of about 2σLL from the particle surface, a behavior that we have encountered in all examples we have studied. This is the reason why, in the macroscopic model, to test bubble formation, we look at the temperature at a distance of 2σLL from the nanoparticle surface. Analysis of the sensitivity of the results to this particular choice is presented in the Supporting Information. The sharp drop in temperature at the nanoparticle surface visible in Figure b is the effect of the Kapitza resistance.

Figure 5

Variation of (a) density and (b) temperature as a function of radial distance from the center of the nanoparticle for various wall temperatures TW. The black line indicates the critical density and critical temperature of the liquid molecules. In this case, the temperature of the nanoparticle kBTNP/ϵLL is kept at a constant value of 5.54 and the concentration of gas molecules is 0.

As explained before, in this work, we use two different criteria for nanobubble formation for the MD simulation and macroscopic theory. In the former one, the criterion is that the liquid density falls below the critical density of the Lennard-Jones molecules. For the macroscopic theory prediction, we use the criterion that the temperature near the nanoparticle surface exceeds spinodal temperature Tspin. Consistency of the two criteria requires that at the liquid–vapor interface, the density of liquid molecules should fall below the critical density and the temperature in the interfacial region should cross spinodal temperature. We tested this consistency from the measurement of density and temperature around nanoparticle when a nanobubble has been judged to form. Some typical results are shown in Figure where it can be observed that critical density and spinodal temperature coincide at the liquid–vapor interface.

Figure 6

Variation of density and temperature around a heated nanoparticle as a function of the radial distance from the nanoparticle center. In this case, kBTNP/ϵLL is set to 5.54 and kBTW/ϵLL to 0.97. The shaded region indicates the liquid–vapor interface where the temperature crosses the spinodal temperature and the density of the liquid molecules goes past the critical density. It shows that both criteria for nanobubble nucleation (T = Tspin and ρ = ρc at the liquid–vapor interface) are consistent.

We performed simulations for various combinations of TNP and TW and identified the region in this parameter space for which a nanobubble nucleates. The results corresponding to three different gas mole fractions xg are shown in Figure . Simulations were performed for three different values of the gas fraction xg, 0, 0.011, and 0.022. The background color in Figure indicates the prediction of nanobubble formation (brown = no bubble, blue = stable bubble), as obtained from the approximation Ts ≈ Ts′ together with eqs and 5The boundary between the two colors is set by the criterion Ts = Tspin. The small circles in Figure correspond to the MD simulations; colors blue and red indicate the formation or absence of a nanobubble, respectively. As expected, the results show that the nucleation of a nanobubble is more likely for higher wall and nanoparticle temperatures. Figure a,b shows a reasonable agreement between the macroscopic theory predictions and MD simulations. However, there are clear deviations for xg = 0.022 (Figure c), on which we will comment later.

Figure 7

Values of wall temperature TW and nanoparticle temperature TNP for which a vapor nanobubble is nucleating around a heated nanoparticle when (a) xg = 0, (b) xg = 0.011, and (c) xg = 0.022.

It can be observed from Figure a–c that the minimum values of TNP and TW required to nucleate a nanobubble decrease significantly with an increase in gas concentration. The primary reason for this reduction is the reduction in the critical point of the binary mixture of Lennard-Jones particles. A binary mixture of Lennard-Jones particles can be approximately described as a one-component fluid with an effective interaction parameter ϵM. Application of the van der Waals one-fluid conformal solution mixing rules gives[30]where x is the molar fraction of component i. In the present case, n = 2 and i = 1, 2 and ϵ and σ are the Lennard-Jones parameters of the mixture components. Critical temperature of the mixture is given by Tc,M = 1.313ϵM/kB.[30] For the Lennard-Jones parameters used in this study (see Table ), the critical temperature of the single-component liquid is 474 K and that for the binary mixture of Lennard-Jones particles is 466 and 458 K when xg = 0.011 and 0.022, respectively. There have been some experimental studies on the homogenous nucleation of bubbles in the presence of a noncondensable gas that showed similar behavior, i.e., the increase in the gas concentration decreases the saturation temperature of the liquid, which results in the nucleation of bubbles at lower temperatures compared with the pure liquid.[17,19] Although the change in the critical temperature of the mixture due to the presence of dissolved gas is relatively small, it has significant effect on the nucleation conditions, as can be observed from the shift in the boundary of nucleation boundaries in Figure . Figure shows the density profile of liquid molecules around a heated nanoparticle for different gas concentrations at the steady state. The density near the nanoparticle decreases with increase in the gas concentration, which further demonstrates that the dissolved gas enhances the nucleation of nanobubbles.

Figure 8

Density of liquid around a heated nanoparticle for various values of the mole fraction of gas molecules dissolved in liquid. TNP = 2000 K and TW = 350 K for this figure.

Figure clearly demonstrates that the gas molecules dissolved in the bulk liquid enhance the formation of a vapor nanobubble. However, for high gas concentration, the enhancement is more than that predicted by theory (see Figure c). At the highest gas mole fraction, eq is clearly not able to predict the nucleation of a nanobubble at high TNP and low TW because the gas solubility decreases with increase in temperature.[31] As a consequence, the solution becomes oversaturated in the high-temperature region, which facilitates the nucleation of a nanobubble. The criterion Ts = Tspin used before therefore fails. We can determine the appropriate threshold value Ts = Tth from eq in conditions for which the MD simulations prove the nucleation of a bubble. That is, we fit eq to the boundary, as set by the MD data points of Figure c to obtain the values of Tth as a function of TNP. Figure a shows the result of the fit, which now by construction fulfills the nanobubble nucleation conditions for xg = 0.022, whereas Figure b shows the Tth normalized by the mixture critical temperature Tc,M as a function of TNP obtained from the fit. For kBTNP/ϵLL < 4.5, the value of Tth is constant and equal to Tspin, suggesting that the nucleation of the vapor nanobubble can be explained by the change of phase due to the crossing of spinodal temperature near the nanoparticle surface. Beyond kBTNP/ϵLL = 4.5, Tth decreases monotonically, which shows that the nanobubble nucleation is dictated by the high oversaturation of gas molecules near the hot nanoparticle.

Figure 9

(a) Fitting of eq with MD data of nucleation conditions for xg = 0.022 by considering Tth as a fitting parameter. (b) Values of Tth obtained from the fit as a function of TNP, which can explain the nucleation of a vapor nanobubble for xg = 0.022. Tth = Tspin shows the regime where nanobubble nucleation is controlled by latent heat required to change the phase of the mixture. Tth < Tspin indicates that the nucleation of a nanobubble is controlled by the oversaturation of gas. TNP at which this transition occurs should decrease with increasing xg.

Growth Dynamics

By calculating the density as a function of time after nucleation, we can follow the bubble growth. At every instant of time, we use the fit of eq to determine the bubble radius. The results are shown in Figure . The bubble growth follows the t1/6 behavior observed in the experiments by Wang et al.[16] This time dependence can be explained by the balancing of plasmonic heating with the latent heat of vaporization of the liquid.[16] Wang et al.[16] argued that the efficiency of heat transfer from the nanoparticle surface during the initial growth of the nanobubble is dependent on its volume. The efficiency of the heat transfer is directly proportional to the ratio of the volume of the nanoparticle to the volume of the nanobubble, which leads to the Rb(t) ∝ t1/6 behavior.[16] Although the growth dynamics of the nanobubble follows t1/6 behavior independent of the gas concentration, the prefactor increases with the gas concentration. It can also be observed that after an initial explosive growth, the radius of the nanobubble reaches a steady value that is due to the finite size of the system. The steady radius of the nanobubble is calculated by using eq and compared with MD results in Figure . Note that the thermal accommodation coefficient α and the density of molecules at the nanoparticle surface ρs, which are input parameters to eq are calculated from the MD simulation at every xg. So in a way, the comparison of the maximum radius in Figure serves as a nice consistency check between the MD simulations and macroscopic theory.

Figure 10

Radius of nanobubble as a function of time for different concentrations of gas molecules in liquid. The inset shows the same data on a log–log scale, which demonstrates that the radius of nanobubble is consistent with that for a t1/6 behavior.

Figure 11

Steady-state radius of the nanobubble as a function of the gas mole fraction in the liquid. kBTNP/ϵLL is equal to 5.54, and kBTW/ϵLL is equal to 0.97 for all data points. Black data points are calculated from eq , which seems to be consistent with the data points obtained from MD simulations.

Summary

Molecular dynamics (MD) simulations were performed to study the formation and growth dynamics of a nanobubble around a heated nanoparticle. The system consists of a nanoparticle dispersed in the bulk liquid that is in contact with an isothermal wall far away from the nanoparticle surface. Combinations of the nanoparticle temperature and the wall temperature that lead to the formation of a nanobubble were determined from MD simulations and were found to be in good agreement with theoretical predictions based on heat balance argument. The role of dissolved gas in the bulk liquid on the formation of nanobubble was analyzed. We found that dissolved gas enhances the nucleation of a nanobubble because of the decrease in the critical temperature of the mixture. As long as the conditions are such that the gas solution is not supersaturated, the lowering of the critical temperature is sufficient to explain the nucleation conditions. For a given gas concentration, depending on the nanoparticle and cold wall temperature, conditions can be reach to cause gas solution to become locally supersaturated. When this happens, the gas oversaturation dictates the nucleation rather than critical temperature. It would be interesting to predict this transition theoretically.

The time dependence of the radius of the nanobubble is calculated and found to follow t1/6 behavior, in agreement with the experimental observations.[16] After the initial explosive growth, the size of the nanobubble reaches a steady state due to the finite size of the system. The steady-state radius of the nanobubble was also calculated from the heat balance arguments and was found to be consistent with the MD simulations.