2. Methodology

Simulations were carried out using a high-fidelity front tracking framework (Glimm et al. Reference Glimm, Isaacson, Marchesin and McBryan1981; Glimm & McBryan Reference Glimm and McBryan1985; Glimm et al. Reference Glimm, Grove, Lindquist, McBryan and Tryggvason1988, Reference Glimm, Grove, Li and Tan2000; Du et al. Reference Du, Fix, Glimm, Jia, Li, Li and Wu2006; Bo et al. Reference Bo, Liu, Glimm and Li2011), coupled with a novel grid-aligned ghost fluid method (Bempedelis & Ventikos Reference Bempedelis and Ventikos2020). This method has been validated extensively for bubble problems, with particular focus on the shock-induced collapse of bubbles, in an earlier paper by the authors (Bempedelis & Ventikos Reference Bempedelis and Ventikos2020).

The flow is modelled by the compressible Euler equations,

(2.1a–c)\begin{equation} \frac{\partial \boldsymbol{q}}{\partial t} + \boldsymbol{\nabla} \boldsymbol{\cdot} \boldsymbol{f}(\boldsymbol{q}) = 0, \quad \boldsymbol{q}\left(\boldsymbol{x},t\right) = \left( \begin{array}{c} \rho \\ \rho \boldsymbol{u} \\ E \end{array} \right), \quad \boldsymbol{f}(\boldsymbol{q}) = \left( \begin{array}{c} \rho \boldsymbol{u} \\ \rho \boldsymbol{u} \otimes \boldsymbol{u} + p \boldsymbol{\mathsf{I}} \\ \left(E+p \right) \boldsymbol{u} \end{array} \right), \end{equation}

where $\rho$ is the density, $\boldsymbol {u}$ the velocity vector, $p$ the pressure and $E$ the total energy per unit volume. The fluids are assumed immiscible, and the effects of diffusion (thermal and viscous), phase change and surface tension are neglected. For the temporal and spatial scales considered in this work, the collapse of a bubble is a phenomenon where inertial forces are dominating. Other mechanisms can thus be disregarded as they have very little effect on the flow. This is also confirmed in Betney et al. (Reference Betney, Tully, Hawker and Ventikos2015), where the relevant non-dimensional numbers (such as the Reynolds and Weber numbers) are shown to be very large. The above system of equations is closed with the stiffened gas equation of state which relates the pressure with the density and internal energy of the fluid,

(2.2)\begin{equation} p + \gamma p_\infty = \rho (\gamma-1) (e+e_\infty), \end{equation}

where $\gamma$ is the adiabatic index, $p_\infty$ the stiffened pressure constant, $e$ the specific internal energy and $e_\infty$ the energy translation factor. The values for these parameters are taken from Hawker & Ventikos (Reference Hawker and Ventikos2012), and are given in table 1. The reader is referred to Bempedelis & Ventikos (Reference Bempedelis and Ventikos2020) for a detailed description of the employed numerical framework.

Table 1.

Thermodynamic parameters for the stiffened gas equation of state.

In the simulations that follow, the air bubbles were placed in mechanical equilibrium with the surrounding liquid, which was water (for a visualization of the two-dimensional arrangement see figure 1). Both fluids were initially at rest under atmospheric pressure conditions. The density of water was set to $\rho _w = 993.89\ \textrm {kg}\ \textrm {m}^{-3}$ and that of air at $\rho _a = 1.204\ \textrm {kg}\ \textrm {m}^{-3}$. The radius of the large bubbles was set to $R_B=0.0005\ \textrm {m}$. The radius of the small bubble was set to $R_b=0.0002\ \textrm {m}$ (size ratio $R_B/R_b=2.5$). The domain was discretized with $75$ points per large bubble radius ($30$ points for the small one). As shown in Bempedelis & Ventikos (Reference Bempedelis and Ventikos2020), such a resolution is sufficient for the accurate prediction of the collapse dynamics. This is further verified in the mesh refinement study presented in § 3. In the simulations that follow, the flow field is extracted every $t=10\ \textrm {ns}$. Computations were performed without taking advantage of any symmetries, in order to be able to predict the development of three-dimensional instabilities. Periodic boundary conditions were applied to all domain boundaries other than the inlet and the outlet. Post-shock conditions ($\, p=1\ \textrm {GPa}$, corresponding to a $M \simeq 1.42$ shock wave in water) were defined at the inlet and non-reflecting conditions were set at the outlet.

3. The collapse of a focusing triangular bubble array

We commence the discussion by describing in brief the shock-induced collapse of a focusing triangular bubble array. For simplicity, the case considered here refers to the collapse of an array of cylindrical bubbles (i.e. a two-dimensional configuration, figure 1), with parameters corresponding to case I in table 2.

Table 2.

Bubble arrangement and domain parameters.

Collapse begins when the incident shock wave impacts the large bubbles. Following the impact, the shockwave gets transmitted in the gas and reflected back in the liquid as a strong rarefaction. The bubble interface is accelerated, and the pressure relaxation leads to the formation of a high-speed jet on the upstream side of each bubble (figure 2a). The collapse of the large bubbles is not symmetric: the inner lobes are less compressed compared with their outer counterparts. This occurs as the incident shockwave weakens in the region between the bubbles, due to its interaction with the reflected rarefaction waves. A strong water hammer shockwave is generated upon impact of the liquid jet on the far wall of each bubble (figure 2b).

Figure 2.

Bubble interfaces and pressure contours at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of cylindrical air bubbles in water. $(a)$$t=490\ \textrm {ns}$, $(b)$$t=790\ \textrm {ns}$, $(c)$$t=990\ \textrm {ns}$, $(d)$$t=1070\ \textrm {ns}$.

The presence of the large bubbles and the associated flow patterns (such as the reflected rarefaction waves) has a significant effect on the collapse of the central one. Contrary to the shock-induced collapse of a single bubble, several waves contribute to the collapse of the central bubble: the incident shock wave, the water hammer shock waves and the waves generated at the collapse of the inner lobes (figure 2c). The way these waves interact (and eventually impact the bubble) is highly complex (figure 2d). The parameters that govern the dynamics of the collapse of the central bubble are two: the geometric properties of the arrangement and the strength of the incident shock.

At this point, a grid convergence study was carried out in order to demonstrate the validity of our simulations. Three simulations were performed: one with the resolution employed throughout the paper ($75$ points per initial large bubble radius, in consideration of the highly demanding three-dimensional simulations and the available computational resources), and two with increased resolutions ($150$ and $300$ points per initial large bubble radius). Although convergence is a concern when considering inviscid equations, previous works on the collapse of single bubbles have shown that such convergence can be achieved if specific metrics, that are important to the processes studied, are considered. These metrics include collapse times, velocity and shock pressure at main jet impact (Hawker & Ventikos Reference Hawker and Ventikos2012; Bempedelis & Ventikos Reference Bempedelis and Ventikos2020). Figure 3 shows the time history of the maximum pressure in the domain, a metric of particular interest in our study. The predicted pressure profiles are in good agreement; though differences in pointwise values exist (and are to a degree attributed to the flow field sampling frequency), all key features in the collapse process are well captured by all considered resolutions.

Figure 3.

Time history (following shock impact) of the maximum pressure in the domain for different grid resolutions, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of cylindrical air bubbles in water.

4. The effects of interbubble distance

Assuming that the bubbles lie on the same plane, a three-bubble focusing configuration (see figure 1) may be fully described by three parameters: the size ratio between the large and small bubbles $R_B/R_b$, and the distance between the bubble centres in the two other directions $S_x/R_B$, $S_y/R_B$ (or $S_z/R_B$ in the case of three-dimensional simulations). In the present section, we investigate the effects of different bubble spacing (in the $x$ direction) for both cylindrical (two-dimensional) and spherical (three-dimensional) bubbles. Two arrangements are simulated: in the first scenario (case I) the bubbles are placed farther apart than the second one (case II). The parameters of these configurations can be seen in table 2. The total number of elements in the three-dimensional simulations is 47 250 000 and 40 500 000 for cases I and II, respectively.

Other than describing the collapse process in detail, the aim of this study is to investigate the possibility of localized energy focusing in such configurations. Hence, the analysis commences from figure 4, where we plot the time history of the maximum pressure in the domain for both cases in two and three dimensions.

Figure 4.

Time history (following shock impact) of the maximum pressure in the domain, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of $(a)$ cylindrical and $(b)$ spherical air bubbles in water.

There are two important points of note in these plots. The first is that in the case of cylindrical bubbles (figure 4a), the predicted peaks in pressure are of similar level for both arrangements, whereas in the case of spherical bubbles (figure 4b) much larger pressures are predicted for the more compact arrangement. The second point is that in the case of the compact arrangement we predict larger pressures when moving from two to three dimensions, whereas this does not happen in the case of the loose arrangement.

The above two findings are indications that the dynamics of cylindrical and spherical bubble arrays differ significantly. Hence, conclusions drawn from two-dimensional studies are not always directly translatable to spherical configurations. We continue with a detailed analysis of the collapse process for each case, highlighting the origins of the differences mentioned. It is noted that the discussion is not restricted to the particular (symmetric) arrangement. The involved mechanisms (i.e. the likely contribution of waves emitted at the collapse of the upstream bubbles to the collapse of the central one) are present and influence – to a different extent – the levels of pressure amplification in cases of arbitrary (in terms of bubble placement and sizes) configurations as well (see appendix A.1).

4.1. Cylindrical bubbles

The first peak in the pressure (at $t=690\ \textrm {ns}$ following shock impact) occurs at the generation of the water hammer shock wave in the large bubbles. Water hammer pressure in the loose arrangement is slightly higher due to the larger distance between the bubbles. Subsequent peaks (around $t=890\ \textrm {ns}$) occur at the generation of waves during the collapse of the bubble lobes.

The collapse process is visualized in figures 2 (for case I) and 5 (for cases I and II). The way the central bubble collapses is different for the two cases. In case I, due to the larger spacing, the incident shock wave has enough space to advance through the large bubbles and impact the central one directly (figure 2a). As a result, the bubble is already under compression when the water hammer waves from the large bubbles arrive (figure 2b,c). At $t=1070\ \textrm {ns}$, a small rise in pressure is observed as the waves from the collapse of the lobes interact at the centreline (figure 2d). Finally, at $t=1090\,\textrm {ns}$, a large pressure peak ($p \simeq 9.54\ \textrm {GPa}$) is spotted as the central bubble collapses under the combined effect of all aforementioned waves.

Figure 5.

Bubble interfaces and pressure contours at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of cylindrical air bubbles in water. Left half, case I; right half, case II. $(a)$$t=790\ \textrm {ns}$, $(b)$$t=990\ \textrm {ns}$.

In the case of the compact arrangement, the central bubble is almost completely shielded by the incident wave by the time the waves from the large bubble arrive and start compressing it laterally (figure 5a). The collapse of the bubble is realized in two distinct steps: the water hammer shock waves are the first to arrive and compress the top part of the bubble. The waves generated at the collapse of the lobes follow at a later stage, but their point of origin is very close to the bottom part of the bubble (figure 5b). As a result, the bottom part of the bubble collapses laterally first (at $t=1030\ \textrm {ns}$), creating a strong wave ($p \simeq 5.13\ \textrm {GPa}$). The peak pressure ($p \simeq 8.68\ \textrm {GPa}$) is generated at the collapse of the top part of the bubble which follows shortly after (at $t=1070\ \textrm {ns}$), under the combined effect of all the above waves.

Despite the described differences, both arrangements result in an amplification of the pressure compared with the water hammer shock of a single bubble (predicted to be $p \simeq 3.25\ \textrm {GPa}$ in a simulation with similar parameters). The means of amplification are common in both cases: the waves emitted during the collapse of the large bubbles contribute directly and positively to the collapse of the central one.

4.2. Spherical bubbles

As already discussed, the arrangements under study display a different behaviour when initially spherical three-dimensional bubbles are considered. The time history of the maximum pressure in the domain is presented in figure 4(b). A first pressure peak occurs at the impact of the liquid jets on the far bubble walls, at $t=560\ \textrm {ns}$ (collapse is a faster process in three dimensions). The peaks that follow until approximately $t=730\ \textrm {ns}$ are related to the collapse of the lobes, but mainly to the interaction of the waves upstream of the large bubbles.

The collapse process for both cases is visualized in figures 6 and 7, displaying the pressure contours along the $y=0$ and $x=0$ planes, respectively. In the loose arrangement (case I), collapse of the small bubble is not carried out by waves from the large bubbles. Instead, the collapse is driven by the incident shock wave (figure 6). There is, however, an indirect effect from the presence of the large bubbles: the reflected rarefaction waves relax the pressure of the incident wave. As a result, the collapse of the central bubble is weaker. The collapse takes place at $t=730\ \textrm {ns}$ and no peaks associated with water hammer waves are observed (figure 7b). Instead, the peak in the pressure ($p \simeq 7.85\ \textrm {GPa}$) occurring at $t=770\ \textrm {ns}$ is found upstream of the central bubble, at the interaction of the waves produced during its collapse along the $y=0$ plane (figure 7c). Interestingly, the maximum peak is smaller than the one predicted in the case of cylindrical bubbles. Moreover, the collapse of the central bubble is of similar intensity to that of the larger bubbles, or of an isolated spherical bubble (where pressure was predicted (via a two-dimensional axisymmetric simulation) to rise up to $p \simeq 8.24\ \textrm {GPa}$). This means that there is a range of interbubble distances where the configuration displays a complete change of character: from strongly focusing energy through chain reactions to lessening the magnitude of the produced waves.

Figure 6.

Bubble interfaces and pressure contours on the $y=0$ plane at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of spherical air bubbles in water. Left half, case I; right half, case II. $(a)$$t=590\ \textrm {ns}$, $(b)$$t=710\ \textrm {ns}$.

Figure 7.

Bubble interfaces and pressure contours on the $x=0$ plane at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of spherical air bubbles in water. Left half, case I; right half, case II. $(a)$$t=590\ \textrm {ns}$, $(b)$$t=740\ \textrm {ns}$, $(c)$$t=770\ \textrm {ns}$.

In the case of the compact arrangement (case II), the collapse of the small bubble is carried out in a two-step manner similar to the two-dimensional case. The first large peak in pressure (at $t=760\ \textrm {ns}$) is associated with the lateral collapse of the bottom part of the bubble. The second peak (at $t=800\ \textrm {ns}$) is the largest one (reaching values up to $p \simeq 21.20\ \textrm {GPa}$) and is associated with the collapse of the top part of the bubble.

The relaxation (induced by the large bubbles) in the incident shock pressure is more pronounced on the $y=0$ plane, compared with the $x=0$ one. This leads to the stronger compression of the bubbles along the latter. However, in the case of the compact arrangement the dominant direction of compression changes once the waves from the large bubbles arrive. This results in different patterns for the splitting of the central bubble, as will be shown shortly.

Figure 8 shows a three-dimensional view of the interface of the spherical bubbles at different time instants. In the first frame (figure 8a), the bubble interfaces are depicted at a moment following jet impact. As discussed, in the case of the loose arrangement the central bubble is already significantly compressed. In the case of the compact arrangement (case II), the large bubbles have split and have assumed a horseshoe-shaped form. In the second frame (figure 8b), we observe the lateral compression of the central bubble (more pronounced on its bottom part) for case II, whereas the central bubble for case I has split into two parts. In the following frame (figure 8c), the central bubble for case II has also split, but along a different direction to the one observed for case I. At this point the large bubble rings have also split. At later times (figure 8d), the large bubble rings are shown to have remerged, though only in the case of the loose arrangement.

Figure 8.

Bubble interfaces for cases I (coloured in bronze) and II (in light green) at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a triangular array of spherical air bubbles in water. $(a)$$t=690\ \textrm {ns}$, $(b)$$t=740\ \textrm {ns}$, $(c)$$t=790\ \textrm {ns}$, $(d)$$t=990\ \textrm {ns}$.

5. Energy focusing in other configurations

In the previous section we considered a triangular arrangement of three bubbles in both two and three dimensions (figure 9a). In the present section, our investigation is extended to alternative configurations of focusing arrangements in three dimensions, which share the same cross-section in the $y=0$ plane, allowing thus for an appreciation of the effects of different extrusions in the third dimension. In three-dimensional space, a focusing configuration may also involve five bubbles in a pyramid-like arrangement (figure 9b). Another possible arrangement involves two bubbles, with the large one forming a torus (figure 9c). A configuration such as this is of course very difficult to achieve in water. Experiments of this kind are, however, conducted in either gels (with very high water content, 99.5 %–99.9 % by weight) or even in plastics/polymers, where the visualization of the phenomena takes advantage of the very high intensities emerging and utilizes non-visible light modalities. For the range of shockwaves of interest, such materials behave like liquids, to a degree that makes them indistinguishable from pure water. Casting and arranging bubble configurations of practically arbitrary complexity is fully feasible when such materials are used. The simulations for the new arrangements were carried out with 82 687 500 elements, as the domain was extended along the $y$ direction (such that $D_y/R_B=D_x/R_B=7$), in order to accommodate the additional bubbles. All other parameters were set as reported in § 2 and table 2 (case I).

Figure 9.

Schematic of different focusing configurations. $(a)$ Triangular arrangement, $(b)$ pyramidal arrangement, $(c)$ toroidal arrangement.

Animations depicting the collapse of all three configurations shown in figure 9 are available as supplementary movies 1, 2 and 3 at https://doi.org/10.1017/jfm.2020.535 for the cases of the triangular, pyramidal and toroidal arrangements, respectively. Note that in the animations, time is measured relative to the beginning of the simulations rather than following initial shock impact. It is interesting to mention that no three-dimensional instabilities are observed, as also reported for the collapse of a single bubble (Hawker & Ventikos Reference Hawker and Ventikos2012; Bempedelis & Ventikos Reference Bempedelis and Ventikos2020). Similar to the previous section, we begin the analysis from the time history of the maximum pressure in the domain (figure 10). It is instantly evident that the pyramidal and toroidal arrangements offer significantly higher levels of pressure amplification.

Figure 10.

Time history (following shock impact) of the maximum pressure in the domain, $p=1\ \textrm {GPa}$ shock-induced collapse of different arrangements of air bubbles in water.

5.1. Pyramidal arrangement

The first pressure peak in the case of the five-bubble array is related to the water hammer shock waves of the large bubbles, which occur at the same time as in the triangular array configuration (figures 11a and 12a). Subsequent rises (until $t \simeq 840\ \textrm {ns}$) correspond to the collapse of the bubble lobes and the interaction of these waves upstream of the bubble (figures 11b and 12b). At $t=860\ \textrm {ns}$, a large pressure peak occurs as these waves interact at the centreline, upstream of the central bubble, which has assumed a mushroom-like shape (figures 11c and 12c). The bubble collapses shortly after, generating strong waves reaching up to $p \simeq 55.51\ \textrm {GPa}$ (figures 11d and 12d).

Figure 11.

Volume rendering of the density gradient magnitude at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a pyramidal array of spherical air bubbles in water. $(a)$$t=640\ \textrm {ns}$, $(b)$$t=740\ \textrm {ns}$, $(c)$$t=840\ \textrm {ns}$, $(d)$$t=940\ \textrm {ns}$.

Figure 12.

Bubble interfaces and pressure contours on the $y=0$ plane at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a pyramidal array of spherical air bubbles in water. $(a)$$t=640\ \textrm {ns}$, $(b)$$t=740\ \textrm {ns}$, $(c)$$t=840\ \textrm {ns}$, $(d)$$t=940\ \textrm {ns}$.

5.2. Toroidal arrangement

The collapse of the toroidal bubble occurs much later compared with the large spherical bubbles (at $t=690\ \textrm {ns}$). In fact, the time required for the jet to impact the far bubble wall is similar to the (two-dimensional) case of cylindrical bubbles. Following the main jet impact, the torus is split in two rings (figure 13a), and later in three rings, following the collapse of the inner one (figure 13b). The waves from the collapse of the inner ring strengthen as they converge towards the centreline, where they interact (upstream of the small bubble) at $t=950\ \textrm {ns}$ (figure 13c). Shortly after, the central bubble collapses producing very large pressures, reaching $p \simeq 140.76\ \textrm {GPa}$ (figure 13d).

Figure 13.

Volume rendering of the density gradient magnitude at different time instants following shock impact, $p=1\ \textrm {GPa}$ shock-induced collapse of a pair of toroidal and spherical air bubbles in water. $(a)$$t=760\ \textrm {ns}$, $(b)$$t=890\ \textrm {ns}$, $(c)$$t=950\ \textrm {ns}$, $(d)$$t=1040\ \textrm {ns}$.