Vapor compression and energy dissipation in a collapsing laser-induced bubble Open Access

The collapse of cavitation bubbles often leads to shock waves emission, light radiation, and rebound bubbles. The occurrence of these phenomena suggests the presence of a gaseous phase within the bubble, which is highly compressed during the collapse. However, its nature and influence on the bubble dynamics is to date still a subject of debate. Furthermore, although it is widely accepted that the non-condensable gas within the bubble undergoes adiabatic compression, the role of condensable vapors remains vague as equilibrium conditions are at best only satisfied at the bubble interface but not in the bubble interior.^1–3 This problem was already introduced in the last century by Plesset,⁴ who speculated about the inability of vapor to change phase at the same rate of the bubble shrinkage. Successively, several sophisticated numerical models have been proposed to capture the influence of phase change on the bubble dynamics. Fujikawa and Akamatsu,⁵ and later Akhatov et al.,⁶ developed numerical models that include liquid compressibility, and heat and mass transfer. They investigated the incidence of non-equilibrium processes at the bubble wall due to thermal inertia of condensing vapor, concluding that this could lead to the occurrence of supercritical conditions at the final stage of the collapse. They therefore highlighted the possibility of the vapor to behave as a non-condensable gas in the case where the volume reduction rate of the bubble was much higher than the condensation rate. In addition, Akhatov et al.⁶ introduced a sticking coefficient of water vapor, which played as a tuning parameter to fit experimental data and predicted the condensed vapor at the bubble–liquid interface. Their work was further endorsed by Szeri et al.,⁷ who studied the heat and mass transfer during cavitation bubbles collapse, concluding that the latter occurs so fast that thermal diffusion and phase change effects are nearly obviated, as the vapor condensation rate is much slower than the bubble volume reduction rate. More recently, Magaletti et al.⁸ conducted a similar numerical investigation considering phase change, occurrence of supercritical conditions, thermal conduction, and liquid compressibility effects, reporting the disappearance and reappearance of the liquid–vapor interface during the final stage of the collapse because of a transition to supercritical conditions of the vapor. They also concluded that, in agreement with Fujikawa and Akamatsu,⁵ purely vapor bubble may be able to emit shock waves at collapse. Lately, Liang et al.⁹ studied the transition from nonlinear to linear oscillations of collapsing cavitation bubbles. They developed a novel approach based on the Gilmore¹⁰ model, with which they could fit the progressive condensation of water vapor during nonlinear bubble oscillations from experimental data by means of a tuning parameter, obtaining in turn the partial pressure of condensable vapor and non-condensable gas within the bubble. The results were in good agreement with the ones of Akhatov et al.⁶ Following this approach, Wen et al.¹¹ were able to track the bubble dynamics of millimeter-sized spherical cavitation bubbles up to the fourth oscillation. The relevance of phase change on the bubble motion has been also discussed theoretically. Already for linear oscillations, it is possible to distinguish regimes where the vapor is trapped inside the bubble while keeping equilibrium conditions at the interface.^12,13 For strongly nonlinear oscillations, Fuster et al.³ have shown that this asymptotic limit is reached for large values of the accommodation coefficient, where the net flux across the interface is eventually dominated by diffusion effects.⁷ Thus, the relevance of non-equilibrium conditions at the interface and the consequences of it on the bubble motion remain an open problem that needs of careful experimental investigations.

Irrespective from the fact that equilibrium conditions are sustained at the interface or not, the influence of phase change on the dynamics of bubbles has been experimentally confirmed in several works. In relation to single bubble sonoluminescence, Vazquez and Putterman¹⁴ observed an increased collapse cushioning and decreased light emission at increasing water temperature. The latter observation was further confirmed by Toegel et al.¹⁵ and Hopkins et al.,¹⁶ who highlighted the importance of the partial pressure of vapor trapped within the bubble during the collapse on the intensity of light emission. Later, Tinguely¹⁷ and Phan et al.¹⁸ investigated the dynamics of laser-induced cavitation bubbles in water at different temperatures, showing that the higher the water temperature, hence the vapor pressure, the larger the rebound bubble. To explain these effects, numerical models based on the slightly compressible versions of the Rayleigh–Plesset equation proposed by Keller and Miksis¹⁹ and Gilmore¹⁰ use the bubble internal pressure to fit the bubble radius evolution.^9,20,21 However, due to technological challenges involved in measuring the bubble contents at the sub-millimeter and sub-millisecond scale within the bubbles, direct probing of the inner bubble pressure remains uncertain²² and may hinder some limitations of the model when using experimental data to fit the evolution of the bubble radius. Among others, some limitations of these models are that numerically describe the dynamics of perfectly spherical cavitation bubbles neglecting effects such as chemical reactions, phase change, or strongly nonlinear effects related to liquid and gas compressibility. In addition to the aforementioned uncertainties regarding the modeling of phase change processes, the influence of non-spherical deformations deserves particular attention as its effect has been indeed observed in various studies. Brennen²³ investigated the bubble fission process due to bubble shape instabilities, concluding that the energy dissipated by the mixing and turbulence due to bubble fission may be preponderant compared to the conventional viscous and acoustic damping. Delale and Tunç²⁴ developed a numerical model also accounting for deviations from sphericity, confirming the results of Brennen.²³ Moreover, Supponen et al.²⁵ experimentally investigated deformed cavitation bubbles and reported that bubbles experience weak jetting phenomena even with reduced anisotropy. Bubble shape perturbations have been also largely investigated in relation to single bubble sonoluminescence, where the development of hydrodynamic instabilities at the bubble–liquid interface has been shown to determine the stability diagrams in which light emission is observed.²⁶ 

In this work, we investigate the influence of the vapor content during the collapse of laser-generated single cavitation bubbles in aqueous ammonia by systematically varying the ammonia mass fraction $w NH 3$ in solution. The latter two-component solutions have similar densities but different $p v *$ values, which we use to investigate the influence of the bubble internal composition. Compared to single bubble sonoluminescence experiments, the main difference is that it is possible to investigate the influence of the vapor content in regimes in conditions where the bubble oscillation is not stable gaining further insight about the role of phase change on extremely violent transient collapses. In addition, because the system's temperature is kept constant, we avoid some problems related to thermal expansion-related misalignment of optical components, sensitivity variation of measuring instruments, and change of laser energy absorbed by the liquid at bubble generation. These experiments allow us to investigate the influence of vapor content on various variables, including (i) the rebound size and collapse time, (ii) microscopic shape of rebound bubble, (iii) luminescence, (iv) radiated shock at collapse, and (v) liquid pressure build up prior to final collapse. Our findings show that all five vary significantly with the NH₃ concentration in solution, evidencing the significant role of the latter on the bubble collapse and supporting the notion of vapor compression. Furthermore, our measurements provide evidence of the effect of the bubble contents on the acoustic emission at bubble collapse.

The experimental apparatus relies on a laser-based technique for the generation of single cavitation bubbles^27–30 schematically shown in Fig. 1. A bubble arises from the plasma generated by a 9-ns Nd:YAG laser pulse (Quantel CFR 400, 532 nm) focused into a point at the center of an extended volume of liquid (aqueous ammonia in this work). The laser beam, redirected with a set of high-intensity mirrors, is enlarged tenfold with a beam expander and focused with an off-axis parabolic mirror, immersed into the liquid in the test chamber. The anisotropy parameter ζ for the generated bubbles was kept below the topological limit between spherical and toroidal collapse (⁠ $ζ < 4 × 10 − 4$⁠), such that any reentrant jet did not pierce the bubble at collapse (a detailed description of ζ is given by Obreschkow et al.³¹ and Supponen et al.²⁵). To this end, we generated bubbles with a maximum radius $R 0 ≈ 1.5$ mm. This size was achieved by adjusting the laser beam energy with a neutral-density filter. Aqueous ammonia was contained in a transparent gas-tight box (cubic shape, 18 cm edge length), which prevented ammonia leaks and hence concentration variations. A total of six ammonia mass fractions $w NH 3$ were exploited in the experiments, ranging from 0 (pure water) to 0.05. The pressure $p ∞$ and the temperature $T ∞$ of the liquid at rest were kept constant and equal to the atmospheric pressure (⁠ $≈ 97$ kPa) and ambient temperature (⁠ $≈ 20 ° C$⁠), respectively.

A high-speed camera (Shimadzu HPV-X2) filming at up to 10 million frames per second, backlighted by a collimated LED, recorded shadowgrams of the bubble. The camera is equipped with a 105 mm objective (Nikon AF-S Micro 1:2.8 GED), and a 2× teleconverter (Nikon AF-S TC-20E). A safety filter (532 nm high-pass filter) is mounted in front of the objective to prevent reflected laser beams to accidentally reach the camera sensor. The camera records 256 frames per film at a fixed resolution of 400 × 250 pixels at all frame rates below 10 × 10⁶. At the latter, the resolution is halved. The results of bubble dynamics presented in this work were obtained from high-speed recordings at 500 000 fps, whereas luminescence was recorded at 10 × 10⁶ fps. The instantaneous radius R of the bubble was obtained with automated image processing from high-speed recordings by retrieving the equivalent radius $R eq$ of the bubble axial cross sectional area A assuming spherical symmetry: $R = R eq = A / π$⁠. The bubble cross sectional area is represented by the black area in the bubble shadowgram recorded by the camera.

Vapor pressure data of aqueous ammonia solutions are reported in Fig. 2(a). The density ρ of aqueous ammonia with a mass fraction from 0 to 0.05 is taken from Green and Perry³² as 998.2, 993.9, 989.5, 985.3, 981.1, and 977.0 kg m⁻³, respectively. The speed of sound c of aqueous ammonia is retrieved from high-speed movies and is 1479, 1487, 1494, 1510, 1516, and 1535 m s⁻¹ for the same solutions cited before, respectively. The critical point of ammonia, as reported in Fig. 2(b), is defined at 132.5 °C and 11.28 MPa, whereas the critical point of water is defined at 374 °C and 22.06 MPa.³² At experimental conditions, ammonia and water coexist with the liquid phase in vapor form and can therefore be liquefied by compression only. As reported by Narita et al.,⁴⁰ who investigated the dissolved oxygen in aqueous ammonia, the concentration of dissolved gas decreases with increasing ammonia concentration.

A detailed procedure for fitting the bubble dynamics from experimental data is described in the  Appendix.

Figure 3(a) shows the radial evolution in time in normalized coordinates, $R / R 0$ vs $t / t R$⁠, of single cavitation bubbles in aqueous ammonia, with R₀ the radius of the bubble at its maximum expansion, and $t R = 0.915 R 0 ρ / p ∞$ the Rayleigh⁴³ collapse time for an empty bubble collapsing without dissipation effects. The plot distinctly illustrates the dependence of the maximum radius of the rebound bubble R₁ on the concentration of the ammonia solution, which increases from 19% of R₀ for the least concentrated solution to 57% of R₀ for the most concentrated solution. On the other hand, R₀ remained nearly constant at 1.5 mm (±6% due to laser power oscillation, see Table I) independently of the ammonia concentration. This pointed out that the influence of the ammonia content on the first bubble oscillation is negligible, suggesting that the physics of the plasma at generation is not altered by the presence of ammonia in solution and that mass transfer effects do not affect the growth of the bubble. Nevertheless, as shown in the nested plot in Fig. 3(a), and in Fig. 3(b), the bubble lifetime was gradually prolonged up to approximately 6% with increasing ammonia concentration.

Furthermore, we noticed additional details strictly related to ammonia content in solution. For instance, high-speed shadowgraphs, displayed in Fig. 4(a), show that the rebound bubble in pure water has a rather irregular shape with various defects at its interface. The development of these deformations seems to be prevented by an augmentation of the ammonia content in solution as schematically represented in Fig. 4(b). The bubble sketched on the left-hand side has a lower vapor content, leading to a larger growth of reentrant jets and surface perturbations. On the contrary, greater vapor pressure in the bubble on the right-hand side opposes to the liquid contraction, leading to a more spherical collapse.

Observation of snapshots of the last instant of the collapsing bubble displayed in Fig. 4(c) reveals a brighter hot-spot for bubbles collapsing in pure water than in aqueous ammonia at $w NH 3 = 0.01$⁠, indicating a larger energy density achieved due to a smaller minimum radius at collapse. Although light emission is no longer detected at larger ammonia concentrations, the displayed snapshots of collapsing bubbles in the most diluted solutions show multiple irregular hot-spots.

The hydrophone signals U over time, displayed in Fig. 4(d), also pointed out the dependence of the emitted shock waves on the ammonia content in solution. The signal of the shock wave emitted upon bubble collapse in pure water (⁠ $w NH 3 = 0$⁠) exhibits multiple peaks, while it smooths out at larger ammonia concentrations (⁠ $w NH 3 = 0.05$⁠). In addition, we surprisingly observed that the pressure build up recorded by the hydrophone before the main shock is more pronounced at larger ammonia concentration.

We noticed several details about cavitation bubbles in aqueous ammonia at increasing concentration, namely, (i) longer collapse time, followed by larger rebound bubbles, (ii) increased regularity of the spherical microscopic shape of rebound bubble, (iii) reduced light emission at collapse, (iv) enhanced regularity of the radiated shock's shape at collapse, and (v) larger liquid pressure build up prior to final collapse. These observations, strictly related to the bubble contents, evidence an increased bubble internal pressure during the collapse, likely due to an increased vapor pressure of the two-component aqueous ammonia solution. In particular, the longer collapse time, followed by larger rebounds, is likely due to the cushioning effect of the compressed gaseous phase, whose pressure increases with increasing ammonia content.^14,16,17

Bubble shape's irregularities, probably due to a loss of sphericity during the last moments of its collapse, decrease when the ammonia concentration is increased. Although we are unable to observe the very final stage of the collapse, a faster increase in the bubble pressure at collapse reduces the retraction of the liquid, increasing the minimum radius at collapse, and inhibits the crumpling of the bubble, thus the growth of deviations from sphericity, as illustrated in Fig. 4(b). Such deviations from sphericity are likely brought on by the development of reentrant non-piercing jets and/or hydrodynamic instabilities at the bubble–liquid interface such as Rayleigh–Taylor instabilities.^{20,31,44–47} Optical systems, as the one employed for this work, are well known to generate bubbles with surface perturbations.^20,48 Our reasoning about an augmented bubble pressure at collapse is also backed by visualizations of the light emitted at bubble collapse.

Observations of light emission are consistent with the works of Toegel et al.,¹⁵ Vazquez and Putterman,¹⁴ Moss et al.,⁴⁹ and Hopkins et al.,¹⁶ who also observed higher light intensity with lower water vapor pressure within the bubble.

Finally, the analysis of the shock waves emitted at bubble collapse support the idea of a faster increase in the bubble internal pressure at collapse due to the increased vapor pressure in more concentrated ammonia solutions. The indented shock wave signal recorded at bubble collapse in water, displayed in Fig. 4(d), probably arises due to successive collapses of different bubble fragments, whereas the smoother signal for the bubble in concentrated aqueous ammonia is synonym with more uniform bubble collapse. Moreover, the more pronounced pressure build up in the liquid surrounding the bubbles in more concentrated aqueous ammonia is likely attributed to a greater deceleration of the liquid surrounding the bubble as the internal pressure of the latter increases.^33,43

All else equal, i.e., $p ∞$⁠, R₀, and ζ, a larger bubble internal pressure would only occur if a larger amount of matter in the gaseous form is compressed during the collapse. In this regard, although we cannot exclude that the non-condensable gas within the bubbles may have an effect on their dynamics,⁶ we argue that our observations are not a consequence of the varying partial pressure of non-condensable components. Specifically, non-condensable gases within laser-induced cavitation bubbles may result from vaporization of dissolved non-condensable gas at bubble generation, chemical reactions from plasma recombination, and diffusion from the liquid. For bubbles similar to those presented in this work, diffusion effects were shown to be negligible with respect to the other two gas sources.^4,6,26,48 Furthermore, laser-generated gas due to chemical reactions is assumed to be proportional to the initial plasma energy density, hence to the bubble potential energy $E p 0$⁠,^9,20 which was nearly constant. We could also neglect any laser-generated gas due to the different liquid composition, as the dissociation energy of water and ammonia is similar (498 kJ/mol for the first O–H bond in water and 435 kJ/mol for the NH₂–H bond of ammonia).^50,51 In turn, the constant bubble volume at maximum expansion highlights the independence of the plasma nature on the ammonia content. Conversely, this would likely reflect on the bubble size, absorbed energy, and/or energy partitioning at bubble generation. This actually corroborates the effectiveness of employing aqueous ammonia as a successful tool for the investigation of cavitation bubbles. Finally, vaporization due to plasma generation, proportional to the dissolved gas saturation, may give rise to variations in the partial pressure of non-condensable gas. As mentioned before, Narita et al.⁴⁰ reported a decreased dissolved oxygen in aqueous ammonia with increasing ammonia concentration. However, an increased ammonia content would translate into a decreased vaporized non-condensable gas, which contravene our observations of increased bubble pressure.

Altogether, despite we are unable to directly probe the bubble contents and their behavior during the bubble collapse, our observations provide evidence that the vapor contained within a collapsing bubble is compressed. As claimed in several numerical works present in the literature,^{1,3,5–8,12,13,52} irrespective if equilibrium conditions are meet at the interface or not, condensation kinetics of vapors within the bubble are slower than the volume reduction rate. Similar works have also been published recently.^53–56 Vapors are then unable to fully condense during the collapse due to the thermal lag at the bubble interface, likely transitioning to the non-condensable gaseous state and providing the bubble with stronger means to resist and delay its collapse. The development of a thermal boundary layer near the bubble wall was introduced earlier by Fujikawa and Akamatsu,⁵ Yasui,⁵⁷ Storey and Szeri,¹ and Akhatov et al.⁶ in their numerical models. In light of these theories, as the vapor pressure promptly rises with increasing ammonia concentration in the liquid [Fig. 2(a)], a greater amount of vapor remains trapped into the bubble at collapse due to the higher heat dissipation required for condensations at the bubble wall.

It is to be noted that in our experiments, the compressed gaseous phase is a two-component mixture of ammonia vapor and water vapor. As shown in Fig. 2(a), the water vapor partial pressure decreases in more concentrated solutions, whereas the ammonia vapor partial pressure quickly increases. Therefore, during the bubble collapse at increasingly concentrated aqueous ammonia solutions, the ratio of ammonia to water vapor that gets compressed rapidly changes. Moreover, as the binodal curve of ammonia locates well above that of water [Fig. 2(b)], ammonia condenses more slowly and under higher pressure than water vapor. It is however interesting to note that supercritical conditions may be reached more easily for ammonia. Furthermore, in our multicomponent system, a mass-transfer boundary layer may also concurrently limit the condensation process along with heat transfer, adding an additional level of complexity to the problem.

In addition to supporting compression of vapors, it is to be noted that our results seem to be in contrast with theories focused on the idea that luminescence is due to shock-initiated thermal emission within the collapsing bubbles.^49,57 In particular, Evans⁵⁸ found that small deformations of a nearly spherical converging shock wave increase as the shock converges, eventually limiting the process of light emission. This would be the case of our bubbles in more diluted solutions, which were observed with a more deformed surface. Nevertheless, they emitted more light at collapse than the most spherical bubbles in more concentrated aqueous ammonia. Our results are thus more in line with the theory of luminescence phenomena due to adiabatic compression of the bubble contents, also supported by Brenner et al.²⁶ 

Finally, it is interesting and surprising to observe that the presence of ammonia did not influence the maximum radius of the bubbles. This highlights the purely inertial nature of the bubble growth. Furthermore, as the rebound ensuing the collapse is largely affected by the ammonia content, our results evidence that a higher concentration of ammonia in the water build the bubble contents up, likely increasing the vapor pressure within the bubble, highlighting the effectiveness of this technique to study the effect of the vapor pressure on cavitation bubbles.

The energy balance in Eq. (2) revealed that a mere 1% of the initial potential energy of the bubble $E p 0$ was reconverted into potential energy at the rebound $E p 1$ in pure water, resulting in a substantial normalized damping $D ̂ tot$ of the bubble oscillation of approximately 99%, where $D ̂ tot = D tot / E p 0$⁠. Conversely, in aqueous solutions with increasing ammonia content, $D ̂ tot$ gradually decreases down to approximately 80% at $w NH 3 = 0.05$⁠.

This result seems to contradict the predictions of analytical models based on the Keller and Miksis¹⁹ and Gilmore¹⁰ models, as shown in Fig. 5(b) (right y-axis), where the value of the energy emitted in the shock wave at bubble collapse $E ̂ sw model$⁠, where $E ̂ sw model = E sw model / E p 0$⁠, for different values of initial non-condensable gas $p g 0$ is per construction comparable to $D ̂ tot$ (see the  Appendix). Interestingly, the trend of $E ̂ sw model$ is not reproduced by the reported experimental results.

The trend of $E ̃ sw coll / E p 0$ shown in Fig. 5 indicates the presence of two regimes associated with bubble collapse in aqueous ammonia at various concentrations. At low ammonia contents (⁠ $w NH 3 < 0.03$⁠), results indicate that the controlling mechanism of oscillation damping is different from that of liquid compressibility, as $E ̃ sw coll / E p 0$ monotonically decreases as a function of $D ̂ tot$⁠, contrary to what we expected. Conversely, at large ammonia concentrations (⁠ $w NH 3 ≥ 0.03$⁠), the liquid compressibility becomes the controlling mechanism of oscillation damping as $E ̃ sw coll / E p 0$ is monotonically increasing with $D ̂ tot$⁠. This experimental observation seems contrary to the predictions provided by simple models based on the Keller and Miksis¹⁹ and Gilmore¹⁰ equation, where the main damping mechanism is controlled by liquid compressibility implying that the energy of the emitted wave and the damping of the bubble should be directly correlated. The reasons why simplified models based on Keller and Miksis¹⁹ and Gilmore¹⁰ equation are not able to explain experimental observations should be topic of more careful future investigations. One possible explanation in light of diverse theories developed in the early 2000s is that deviations from sphericity are energy-consuming, which eventually dissipate energy into the liquid as heat.^23,24,59 The bubble fragmentation and later coalescence phenomena have related extremely small length scales at which viscous dissipation may be magnified by several orders of magnitude compared to the spherical case. Although other mechanisms related to strongly nonlinear effects in the liquid or the presence of chemical reactions may also contribute to the effective damping of the bubble and will need detailed investigations, the results are in line with previous observations, as deviations from sphericity are the largest in the most diluted solutions, where we expect a smaller minimum radius at bubble collapse. In this case, the damping effect due to the liquid compressibility seems to be overtaken by this second mechanisms of energy dissipation. Our reasoning is further endorsed by the work of Baghdassarian et al.⁴⁸ and Supponen et al.,²⁰ who reported that laser-induced bubbles are naturally formed with surface perturbations.

Finally, although this could be a mere coincidence, it is also interesting to note that the $E sw coll / E p 0$ curve peaks at approximately the same ammonia concentration whereby the ammonia vapor pressure overtakes the one of water (Fig. 2). In practice, our results suggest that the perfectly spherical bubble collapse hypothesized in the Keller and Miksis¹⁹ and the Gilmore¹⁰ models is inaccurate under certain conditions. As these models account for dissipation mainly due to liquid compressibility, the use of these models may hinder the relevance of other dissipation effects during the collapse of the bubble which are not taken into account. This may in turn compromise the accuracy of the bubble dynamics fitting from experimental data.

In this study, we provide experimental evidence about the effect of vapor content on transient cavitation bubbles:

• Consistently with the predictions for single bubble sonoluminescence experiments, the vapor is trapped inside the bubble and compressed during the violent collapse. Through an increase in the vapor pressure within the bubble by adding ammonia to water, we observe a striking role of the vapor on the amplitude of the rebound bubble. We infer that the progressively decreasing bubble deformation as a function of the ammonia concentration is due to a larger resistance of the bubble content to compression, which limits the growth of deformations and/or hydrodynamic instabilities as a consequence of the smaller values of the acceleration during the last stages of the collapse.

• The results are endorsed by the luminescence phenomena, whose brighter emission in cavitation bubbles generated in pure water, thus with a smaller minimum radius, is synonym of larger energy density.

• The liquid pressure build up prior to the final collapse and the shape of the shock wave emitted successively also indicate that the bubble internal pressure increases faster during the bubble collapse at the instant of collapse for increasing ammonia in solutions. Our reasoning behind the compression of vapors is that vapor–liquid phase transition are restrained, likely by mass and heat transfer.

Moreover, we demonstrate that:

• While the initial growth and subsequent compression of a laser-induced bubble is nearly insensitive to the partial vapor pressure of the host fluid, the damping mechanisms controlling the amplitude of the rebound are greatly influenced by the vapor content. Liquid compressibility effects alone cannot explain the oscillation damping of collapsing bubbles. Some phenomena that could explain the experimental observations include phase change effects during the last stages of the bubble collapse, chemical dissociation reactions, and bubble fragmentation due to Rayleigh–Taylor instabilities, which may eventually lead to the appearance of extremely small scale agitation that would ultimately increase the viscous dissipation.

• These other dissipation phenomena may play a preponderant role particularly in bubbles with low internal pressure at collapse, hence diluted solutions, prevalent in most of real-life applications.

It would be however interesting to obtain absolute pressure measurement obtaining the calibration constant G as a function of the ammonia content in solution. Furthermore, it would be of great relevance to study the shock waves at bubble collapse in more details, ideally investigating experimentally the shock waves behavior in the near-field.

Our findings may help improving the development of more sophisticated numerical models for improved prediction of the cavitation process. We therefore point out the urge to include more precise phase transition prediction in numerical models and call into question the disregard of aspherical perturbed collapsing bubbles. These findings also contribute to our general understanding of cavitation bubbles, with several opportunities for practical applications, including understanding and controlling cavitation erosion processes, regulating the collapse intensity for controlled sonochemical reactions, and improving cavitation-based biomedical technologies. Moreover, they also provide novel information of cavitation in aqueous ammonia, whose occurrence is to be reckoned. Alongside an already-extensive use in chemical plants, ammonia as carbon-free fuel and green hydrogen carrier has gained significant attention over the last decade.^60,61

This work was supported by the MSCA-ITN of the EU Horizon 2020 Research and Innovation program (Grant Agreement No. 813766), and the Swiss National Science Foundation (Grant No. 179018).

The authors have no conflicts to disclose.

Davide Bernardo Preso: Conceptualization (equal); Formal analysis (equal); Investigation (lead); Methodology (equal); Validation (equal); Visualization (equal); Writing – original draft (lead). Daniel Fuster: Conceptualization (equal); Formal analysis (equal); Methodology (equal); Supervision (equal); Writing – review & editing (equal). Armand Baptiste Sieber: Investigation (supporting); Methodology (equal); Visualization (equal); Writing – review & editing (equal). Danail Obreschkow: Formal analysis (equal); Methodology (equal); Validation (equal); Writing – review & editing (equal). Mohamed Farhat: Conceptualization (equal); Formal analysis (equal); Funding acquisition (lead); Project administration (equal); Supervision (equal); Writing – review & editing (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The Rayleigh–Plesset,⁴ the Keller and Miksis,¹⁹ and the Gilmore¹⁰ models are numerically solved with a fourth-order Runge–Kutta method with adaptive time stepping. With the last two, the internal pressure of a collapsing bubble is estimated by fitting the experimental data up to the second oscillation. The model is constrained to find the value of $p g 0$⁠, the free parameter, with which the model best fits the maximum amplitude of the rebound bubble after collapse. The parameters used for simulations with the Keller and Miksis¹⁹ and the Gilmore¹⁰ model in this work are reported in Table II. An example of fitting with the two models is reported in Fig. 6. In the Keller and Miksis¹⁹ and the Gilmore¹⁰ models, we neglected viscous dissipation effects for simplicity.