Microfluidic jet impacts on deep pools transition from capillary-dominated cavity closure to gas pressure-dominated closure at higher Weber numbers

Abstract

Studying liquid jet impacts on a liquid pool is crucial for various engineering and environmental applications. During jet impact, the free surface of the pool deforms and a cavity is generated. Simultaneously, the free surface of the cavity extends radially outward and forms a rim. Eventually the cavity collapses by means of gas inertia and surface tension. Our numerical investigation using an axisymmetric model in Basilisk C explores cavity collapse dynamics under different impact velocities and gas densities. We validate our model against theory and experiments across a previously unexplored parameter range. Our results show two distinct regimes in the cavity collapse mechanism. By considering forces pulling along the interface, we derive scaling arguments for the time of closure and maximum radius of the cavity, based on the Weber number. For jets with uniform constant velocity from tip to tail and We ⩽ 150 the cavity closure is capillary dominated and happens below the surface (deep seal). In contrast, for We ⩾ 180 the cavity closure happens above the surface (surface seal) and is dominated by the gas entrainment and the pressure gradient that it causes. Additionally, we monitor gas velocity and pressure throughout the impact process. This analysis reveals three critical moments of maximum gas velocity: before impact, at the instant of cavity collapse, and during droplet ejection following cavity collapse. Our results provide information for understanding pollutant transport during droplet impacts on large bodies of water, and other engineering applications, like additive manufacturing, lithography and needle-free injections.

1. Introduction

The pioneering work of Worthington (1908) displaying and describing liquid impacts onto pools initiated a century-long interest into characterising such impact phenomena. Understanding the intricacies of these events is relevant for a broad spectrum of situations in nature; the noise of rain (Prosperetti et al. 1989) or the scent of earth after rain on a hot day (Joung & Buie 2015), as well as in technology such as in inkjet printing (van der Bos et al. 2014) or spray atomisation (Panão & Moreira 2005). Of particular interest has been the air entrainment, cavity formation and collapse (Lee et al. 1997; Truscott et al. 2014; Eshraghi et al. 2020; Eggers et al. 2007; Deka et al. 2018). In general, the cavity formation in pools begins when the jet impacts the free surface of the target and deflects its surface. This deflection occurs just before coalescence of the jet and the bath, as the local gas pressure builds up by the approaching liquid jet (Bouwhuis et al. 2015). Upon coalescence the inertia of the high-speed jet dominates the deflection and a hemispherical cavity is formed by the head of the jet (Speirs et al. 2018). Due to its radial expansion, the cavity has lamella shooting radially outwards, extending the walls of the cavity. The remainder of the jet impacts the base of the newly formed cavity and extends it primarily in the direction of travel, producing a slender cavity (Bouwhuis et al. 2015). In this process the kinetic energy of the jet is converted to potential surface energy and heat due to dissipation (Speirs et al. 2018).

There is, however, a stark disparity in the amount of work done on projectiles in the millimetre regime (Engel 1966; Pumphrey & Elmore 1990; Rein 1993; Oguz et al. 1995; Yarin 2006; Barolo et al. 2006; Aristoff & Bush 2009; Zhang et al. 2012; Agbaglah & Deegan 2014; Truscott et al. 2014; Fudge et al. 2021) compared to projectiles in the micrometer regime (Bouwhuis et al. 2016; Speirs et al. 2018; Quetzeri-Santiago et al. 2021; Quetzeri-Santiago & Rivas 2023). For cavities generated by projectiles in the millimetre range the collapse can be mainly attributed to the hydrostatic pressure (Oguz et al. 1995). This implies that the Bond number (Bo = ρ₀gD²/γ) is greater than one (where ρ₀ and γ are the density and surface tension of the liquid, D is a relevant length scale and g the acceleration due to gravity). The domain of interest for our work is in the micrometer regime, where collapse is driven by surface tension forces as Bo ~ O(10⁻³). This regime is relevant in emerging technologies as 3D printing (Antkowiak et al. 2011), spray painting (Herczyński et al. 2011), extreme ultraviolet lithography (Klein et al. 2015), environmental aspects (Speirs et al. 2023) and needle free injection methods (Berrospe-Rodriguez et al. 2016; Oyarte Galvez et al. 2020; van der Ven et al. 2023).

In this work, we looked into the dynamics of a high-speed microfluidic jet penetrating a pool. These jets are comparable in size and momentum produced in needle free applications (Schoppink & Rivas 2022). A validation process was done through both qualitative and quantitative comparisons with other numerical results, experiments and theoretical predictions. In addition we quantified the cavity profile and closure time of the cavity as a function of relevant fluid parameters. Our numerical strategy provides the opportunity to examine a broad parameter space unconstrained from experimental limitations.

2. Methodology

Experimental details

A transparent cubic bath made of acrylic with dimensions of 5 × 10 × 20 cm, was filled with water. High-speed jets were generated from a thermocavitation process and directed to impact a water pool. The setup is similar to the ones used in refs. (Quetzeri-Santiago et al. 2021; Quetzeri-Santiago & Rivas 2023). The thermocavitation process occurs inside a glass microfluidic chip filled with a Direct Red 81 solution in water at 0.5 wt. %. In thermocavitation, a expanding bubble is created at the base of the chip, due to the energy transfer to the liquid from a continuous wave laser. The expanding bubble pushes the liquid that is in front of it generating the jet (Oyarte Gálvez et al. 2020). The jet velocity U₀ and diameter R_j in these experiments ranged from 10 m/s to 40 m/s, and 50-100 µm respectively. The surface tension of water is γ = 0.072 N/m, its density ρ₀ = 1000 kg/m³ and its viscosity µ = 1 cP. Thus, the Weber number $We={\rho }_0{U}_0^2{R}_j/\gamma$, and the Reynolds numbers Re = ρ₀U₀R_j/µ range between 35-1333 and 500-4000, respectively. For all the experiments Bo ~ O(10⁻³). The processes of bubble generation, jet ejection and impact on the liquid droplet were recorded with a Photron Fastcam SAX2 coupled with a 2x Navitar microscope objective. A typical experiment duration was ~ 5 ms and the camera resolution was set to 768 × 328 pixels² at a sample rate of 30k frames per second with an exposure time of 2.5 μs.

Numerical model

We consider a liquid jet impacting a pool of identical liquid with velocity U₀. The jet is cylindrical with radius R_j and length L_j and is placed at a distance S between the free surface level of the pool and the tip of the jet. The domain is axisymmetric and filled with ambient gas. The top, right and bottom boundaries have outflow conditions imposed with the pressure as P = P_∞, zero normal velocity gradients (top and bottom, ∂u_z/∂z = 0; right, ∂u_r/∂r = 0) and zero shear stresses (top and bottom, ∂u_r/∂z = 0; right, ∂u_z/∂r = 0). Since we are studying jet impact in the micro/millimetre regime, effects of gravity are neglected (g = 0) as hydrostatic effects are small (Quetzeri-Santiago et al. 2021).

The governing equations are nondimensionalised with the initial radius of the jet R_j and the impact velocity of the jet U₀. Thus, time is non-dimensionalised as t* = tU₀/R_j. For simplicity from now on we use t* = t.

$$\frac{\partial U_i}{\partial X_i}=0$$ (2.1)

$$\frac{\partial U_i}{\partial t}+U_0\frac{U_i}{X_j}=\frac{1}{\widehat{\rho }}\left(-\frac{\partial P}{\partial X_i}+\frac{1}{Re}\frac{\partial \left(2\widehat{\mu }{D}_{ij}\right)}{\partial X_j}+\frac{1}{We}\kappa {\delta }_s{n}_i\right)$$ (2.2)

Which represent conservation of mass and momentum respectively. In here U_i is the velocity vector, P is the pressure, D_ij is the viscous stress tensor, $We={\rho }_0{U}_0^2{R}_j/\gamma$ is the Weber number and Re = ρ₀U₀R_j/μ. The last term represents capillary effects, where κ is the interface curvature. Ensuring that this term is handled at the liquid interface, the characteristic function δ_s is used. Lastly, n_i is the normal to the interface. The geometric Volume of Fluid (VoF) method is used to track the interfaces, with a VoF tracer Φ such that,

$$\Phi \left(x\right)=\{\begin{array}{ll}1,\hfill & \text{if}x\in \text{fluid}\text{phase}\hfill \\ 0,\hfill & \text{if}x\in \text{gas}\text{phase}\hfill \end{array}$$ (2.3)

Therefore, the one-fluid approximation is used in the momentum equation (2.2) by means of the following arithmetic equations:

$$\begin{array}{cc}\widehat{A}\left(\Phi \right)=\Phi +\left(1-\Phi \right)\frac{A_g}{A_l}& \forall A\in \left\{\rho ,\mu \right\}\end{array}$$ (2.4)

The incompressible Navier-Stokes equations are solved using the finite volume partial differential equation solver Basilisk C Popinet (2009, 2018). With Basilisk, a variety of partial differential equation can be solved with parallelization capabilities on an adaptive mesh refinement (AMR) grid. An example of the mesh refinement used in this work can be seen in figure S1 in the supplementary materials. This solver employs The Bell-Colella-Glaz (BCG) scheme (Bell et al. 1989), which is a robust second order upwind scheme. In this scheme a projection method is used similar to Chorin (1967) where the pressure and velocity solutions for equations 2.1 and 2.2 are decoupled. In this work, we use an improvement on Chorin’s method, where we couple the projection and diffusion-convection steps by the BCG scheme. The projection method is also known as a fractional step method, where intermediate iterative steps are used to uncouple the pressure solution while maintaining a divergence-free velocity field.

Validation

Validation of the code was performed first qualitatively comparing simulations of a microfluidic jet impacting a liquid droplet, with the experiments extracted from one of our previous works (Quetzeri-Santiago et al. 2021). Figures 1 b) and c), illustrate the capabilities of the numerical technique to reproduce the traversing and embedding phenomena observed in the experiments. The numerical setup is similar to the one in figure 1, but instead of a deep pool we initialise a droplet with radius R_d. Next, we tested the ability of the code to reproduce the traversing and embedding threshold obtained experimentally and reported in previous works (Quetzeri-Santiago et al. 2021). After the impact of a microfluidic jet onto a droplet a cavity is created and if the impact velocity is enough to overcome the surface tension of the droplet will traverse it completely. The critical Weber number for traversing the droplet $W{e}_{crit}\approx 64{\left(\frac{D_{drop}}{2D_{jet}}\right)}^{1/2}$ was found by comparing the Young-Laplace and dynamic pressures in the cavity. To assess the validity of We_crit was experimentally compared with the Weber number based on the jet inertia and the droplet surface tension γ_d, i.e., $W{e}_{jet}={\rho }_0{U}_0^2{R}_j/{\gamma }_d$, for a given Ohnesorge number $Oh=\frac{\sqrt{We}}{Re}$. In our simulations, we maintain a constant Oh while varying Re, an unattainable condition to the experiments due to the inherent properties of liquids. Figure 1 b) shows excellent agreement between the experiments and simulations. Furthermore, simulations show that at Re < 200 the threshold increases and deviates from the experimental threshold which is lower than the prediction We_jet/We_crit = 1. This indicates that viscous dissipation can influence the traversing process for more viscous liquids than the ones used on the experiments.

Figure 1.

a) Numerical set-up for the study of a jet impact on a droplet. A liquid jet with radius R_j impacts with a velocity U₀, viscosity µ₀, and density ρ₀ a pool with height H of the same liquid. b) Phase diagram displaying the outcome of droplet penetration based on Re and We_jet/We_crit. With the embedding cases as closed markers and traversing cases as open markers. The experimental data is curved in We-Re space, as it is probed for constant Ohnesorge numbers $Oh=\frac{\sqrt{We}}{Re}$. c) Simulation results of a microfluidic jet impacting a droplet. When the jet has enough inertia to go through the droplet we name it traversing. In contrast if the inertia is not enough we call it embedding. d) Experimental results showing the traversing and embedding of a microfluidic jet on a water droplet (Quetzeri-Santiago et al. 2021).Times are made non-dimensional by diving by R_j/U₀.

To quantify the numerical convergence, the energy distribution over time is calculated. The supplementary materials provides further details of the energy calculation. We show the energy allocation for different resolutions over the penetration time frame in a bar plot presented in Figure S2 in the supplementary materials. The energy is normalised by the total energy initially present at highest refinement (r₀/Δ = 1024). From this bar plot we draw multiple conclusions. First we note that over time the total energy is not fully conserved, albeit that increasing the refinement does mitigate the losses. Therefore, we attribute this energy loss to be inherent to the numerical method. Regarding the distribution of energy the fractions are comparable, especially for the three highest refinements. This makes evident that the numerical process converges at resolution (r₀/Δ = 512).

3. Experimental results

Similarly to the case of the impact of a microfluidic jet onto a droplet, air is entrained when the jet impacts a deep liquid pool, and a cavity is formed (figure 2). The cavity continues expanding in both the radial and the z direction, until it collapses. Previous research shows that for We ≫ 1 during the cavity expansion the process is inertial and the cavity adopts a slender shape (Bouwhuis et al. 2016). Upon reaching the maximum cavity size, interfacial tension starts to influence the cavity dynamics, as kinetic energy is converted into surface energy of the newly formed cavity. The time it takes to reach this regime is approximated by relating the dynamic pressure and the Young-Laplace pressure of the cavity (Quetzeri-Santiago et al. 2021). However, depending on the Weber number the cavity can collapse collapse below the original position of the surface (deep seal, figure 2 a) or from the top of surface (surface seal, figure 2 b). The shape of the cavities and bubbles entrapped are similar to those of impacts on capillary bridges (Quetzeri-Santiago & Rivas 2023). In these experiments we observe deep seal from We ≈ 35 − 200, while surface seal is observed from We ≈ 300 − 400 (see figure 2). We note that this transition is observed for a singular Bond number Bo = 1.4×10⁻³, as in the experiments the only parameter that was varied was the impact speed. These findings align with the regime map described in van der Ven et al. (2023), categorising them within the “splashing substrate” region, specifically located at its leftmost boundary, considering a shear modulus G of water equal to 0.

Figure 2.

Snapshots of a liquid jet impacting a liquid pool. a) Deep seal, We = 200, the cavity collapses below the original position of the pool surface. b) Close up of the surface at the onset the surface seal, We = 400. A crown is formed before it collapses on itself. c) Cavity evolution during surface seal. The cavity forms in a similar way to the deep seal case for t < 0.6 ms, but afterwards the whole cavity volume remains trapped as a bubble inside the liquid pool. Times are made non-dimensional by diving by R_j/U₀.

In figure 3 a) we show simulation results of the cavity profile evolution of the impact of a cylindrical jet with uniform velocity onto a pool for We = [50 − 400], for constant Re and Bo. Furthermore, R_j = U₀ = ρ₀ = 1, and are kept constant throughout the paper. Thus, Weber number was varied by changing the surface tension, with We = 1/γ. For all the Weber numbers at t = 30 the cavity evolution is similar, as is inertia dominated. However, at t > 62 a deviation from the profiles is observed. Similar to the experiments, for We ≈ 180 a rim forms and propels the pool surface upward from its equilibrium surface level. The rim is thinner and shoots higher up as the Weber number increases. In contrast for We < 150, surface tension prohibits a slender rim to develop and to advance above the equilibrium surface level. Now the rim of the cavity is flattened and develops into a spherical blob of liquid. Consequently, the seal mechanism differs in both cases, while for We > 180 the cavity closure is above the pool equilibrium surface level, the opposite is true for We < 150. Although the qualitative phenomena are similar between the experiments and simulations, the critical Weber number for transitioning from surface seal to deep seal is ≈ 30% lower in the simulations compared to the experiments. We also noticed that in our simulations, the point at which the deep seal pinches off is closer to the surface compared to what was observed in the experiments. Additionally, the upward jet resulting from the cavity collapse, i.e., a Worthington jet observed in experiments for We < 200 is not reproduced in the simulations. We attribute these discrepancies to the uniform velocity across the jet in the simulations, which is not be the case for the experiments. Furthermore, we would like to note that the characteristics of cavity collapse and Worthington jet formation are influenced by small disturbances of the cavity and the shape of the jet plays a major role (Michon et al. 2017; Quetzeri-Santiago & Rivas 2023).

Figure 3.

a) Superposition of cavity profiles by cylindrical jets at Re = 2·10⁴, for different Weber numbers, indicating inverse relation between rim thickness and Weber number. b) Simulation of jets with We = 200, all starting with identical tip velocities but varying tail velocities. At times t < 43, the cavity is similar for all cases. Yet, at time t = 73 the cavity collapsed for the cases of u_t/U₀ > 0.8. At a time t = 492, all cavities collapsed with a surface seal, but the cases of u_t/U₀ ⩽ 0.2. In general, jets with higher tail velocities exhibit earlier cavity collapse.

Indeed, in the experiments, due to the decelerating nature of the thermocavitation bubble expansion, the impacting jet exhibited a difference between the jet tail velocity u_t and the jet tip velocity U₀. The result of the difference in velocity between the tip and the tail, results in a Matryoshka effect, i.e., droplets reopening the cavity before it closed, leading to a deeper pinch-off point (Hurd et al. 2015). Conversely, in our simulations, the velocity remained uniform throughout the entire jet. To bridge this gap, we conducted simulations in which, for simplicity, we implemented a linearly decreasing velocity gradient from the jet’s tip to its tail. In this approach, we established the velocity at the tip of the jet as our reference point and systematically adjusted the tail’s velocity to 10%, 20%, and so on, up to 100% of the tip’s velocity.

The outcome of these simulations at We = 200 imposing the aforementioned velocity profiles, are depicted in figure 3 b). We note that higher tail velocities correlated with an earlier cavity collapse, as well as with a pinch-off point closer to the surface. Notably, when tail velocities ranged from 0.9 to 0.4 relative to the tip, the impact result manifested as a surface seal. In contrast, tail velocities in the range of 0.3 to 0.1 yielded a deep seal. Moreover, as observed in figure 4 we recover the deep seal phenomena, and particularly the Worthington jet observed in the experiments. Consequently, by incorporating a falling linear velocity gradient within our jet simulations, we not only achieved qualitative alignment with our experimental data on deep pools but also quantitatively replicated the transition from deep to surface seal (where We was calculated using the jet tip velocity as the characteristic velocity). However, for simplicity in the remaining of the discussion we keep the jets with uniform velocity distribution.

Figure 4.

Jet impacting on a pool with We = 200 and u_t/U₀ = 0.3. In this case the jet breaks into droplets before the cavity collapses, creating a Matryoshka effect. Furthermore, a Worthington jet similar to that of the experiments is recovered. The simulation shows the gas velocity field on the right hand side of the snapshots.

4. Cavity dynamics model

When the cavity is formed, part of the kinetic energy of the jet transforms into surface energy by the creation of new surface. Thus, the free surface of the cavity has more surface energy than a pool in equilibrium. It is therefore energetically favourable for the interface to restore its rest state. In this way surface tension forces (F_γ) counteract the radial expansion of a cavity induced by inertia. Nevertheless, liquid jet inertia drags gas inside the cavity. By using Bernoulli’s principle, along a streamline extending from the outside of the surface to the inside, one notes that the gas density and velocity contribute to a pressure difference. Following, the derivation from Eshraghi et al. (2020),

$${\left[\frac{1}{2}{\rho }_g{u}_g^2+P\right]}_c={\left[\frac{1}{2}{\rho }_g{u}_g^2+P\right]}_{\infty }$$ (4.1)

$$\Delta P=P_{\infty }-P_c\approx \frac{1}{2}{\rho }_g{u}_g^2$$ (4.2)

where ρ_g and u_g are the ambient gas density and velocity, P is the air pressure at a point close to the rim P_c and a point far away near the undisturbed surface of the pool P_∞.

This gradient in pressure induces a force that pulls towards the centre where the pressure is lower. In the remainder of this work, we call this the Bernoulli suction force (F_ΔP).

To characterise the collapse time we model the trajectory of the rim of the cavity by considering the radial component of the surface tension force and Bernoulli suction force (see figure 5). We assume that the forces only act radially. This enables us to to find analytical expressions for the pinch-off time. However, in reality this is a simplification as it does not consider the rim to translate vertically.

Figure 5.

Force diagram on the rim of the cavity and cavity parameters. The rim has a diameter 2a and the forces acting to close the surface are the pressure gradient F_ΔP and the surface tension force F_γ.

The differential mass of the rim m = ρπa²r(t)dϕ, where we assume the rim to be circular in cross section with a the rim radius. This rim is subjected to two forces the Bernoulli suction force F_ΔP and a surface tension force F_γ, thus the rim equation of motion is (Eshraghi et al. 2020),

$$m\ddot{r}=F_{\Delta P}+F_{\gamma }=-2ar\left(t\right)d\varphi \Delta P\left(t\right)-4\gamma ad\varphi$$ (4.3)

$$\ddot{r}=-\frac{2\Delta P}{{\rho }_g\pi a}-\frac{4\gamma }{{\rho }_{g}ar\left(t\right)}.$$ (4.4)

Where $\ddot{r}$ is the rim acceleration, r(t) is the rim position from the centre of impact at any time, ΔP = P_∞ − P_c is the pressure gradient, and ϕ is the radial coordinate. To get an analytical solution for the radial coordinate of the rim, we look into the limits where one force is negligible, which we explore in the next sections.

Radial surface tension regime

To explain the pinch-off time for We ~ 1, we neglect the Bernoulli suction force F_ΔP and assume the only force driving the collapse is the surface tension acting in the horizontal coordinate. Therefore, this problem reduces to that of the collapse of a liquid ring and follow the ideas of Texier et al. (2013). Since the Reynolds number Re ≫ 1, viscous dissipation can be neglected and we can use potential flow to describe the dynamics. Using mass conservation and that the pressure is governed by the Laplace law, we arrive to the following equation for the evolution of r (Texier et al. 2013),

$$\frac{d\left(r\dot{r}\right)}{dt}ln\left(\frac{r+a}{r}\right)+\frac{{\dot{r}}^2}{2}\left(\frac{r^2}{{\left(r+a\right)}^2}-1\right)=-\frac{\gamma }{{\rho }_{0}r}\left(1+\frac{r}{r+a}\right)$$ (4.5)

Assuming that the thickness of the ring is constant, due to volume conservation ra is also constant (Texier et al. 2013). Therefore, we can linearise equation 4.5, considering that r ≫ a during most time of the closure, and that we take time for the cavity collapse from the moment it reached its maximum radius r_max we obtain,

$$\ddot{r}=-\frac{2\gamma }{{\rho }_0{r}_{max}{a}_0}.$$ (4.6)

Where a₀ is the rim radius at r_max. By integrating 4.6 and using t = 0 as the time from which the cavity collapse starts, then the initial expansion velocity $\dot{r}=0$ and r = r_max, thus,

$$r\left(t\right)=r_{max}-\frac{\gamma }{{\rho }_0{r}_{max}{a}_0}{t}^2.$$ (4.7)

Therefore, the closure time is

$$t_c=\sqrt{\frac{{\rho }_0{r}_{max}^2{a}_0}{\gamma }}$$ (4.8)

We note that r_max is reached at a time $t_{\gamma }\approx {\rho }_0^2{R}_j^3{U}_0^3/\gamma$ when the surface tension starts to make the rim move towards the centre of the cavity (Bouwhuis et al. 2015; Quetzeri-Santiago et al. 2021). Furthermore, the cavity radius evolves as R(z, t) ≈ R_j(U₀t + z))^1/2 (Bouwhuis et al. 2015). Substituting t_γ in the last equation we get,

$$r_{max}\sim \frac{{\rho }_0{R}_j^2{U}_0^2}{\gamma }=R_{j}We.$$ (4.9)

We confirm this linear relation for We < 180 by performing simulations as shown in figure 6 a). Noting that a₀ ~ γ, we obtain,

$$t_c\sim r_{max}\sim We.$$ (4.10)

Figure 6.

a) Maximum radius of the cavity r_max scaled by the radius of the jet R_j as a function of the Weber number. r_max ~ We for the deep seal regime. b) The relation between the radius of the rim a and the Weber number, where a ~ We⁻¹. The radii are taken just before pinch-off. These simulations were performed for Re = 20000

Bernoulli suction regime

For We ≫ 1, the contribution of surface tension to the collapse can be neglected, and is dominated by the Bernoulli suction force, using equations 4.2 and 4.4 (Eshraghi et al. 2020),

$$\ddot{r}=\frac{2\Delta P}{{\rho }_g\pi a}=-\frac{u_g^2}{\pi a}=-c_1$$ (4.11)

$$r\left(t=t_c\right)=-\frac{1}{2}{c}_1{t}_c^2+{\dot{r}}_0{t}_c+r_0=0,$$ (4.12)

Solving equation 4.12 we get the time for cavity pinch-off,

$$t_c=\frac{{\dot{r}}_0\pm \sqrt{{{\dot{r}}_0}^2+2c_1{r}_0}}{c_1}\sim \frac{{\dot{r}}_{0}a}{u_g^2}.$$ (4.13)

Where we considered that the initial cavity radius r₀ ≈ 0. In this regime as shown in figure 3, the cavity radius evolves similarly for all Weber numbers up to times t ≈ 60, thus ${\dot{r}}_0\nsim We$. In the simulations we tracked the gas velocity and we observed that u_g remains relatively constant within a 10% margin before impact for We > 200 (see figure S3 in the supplementary materials). Furthermore, as shown in figure 6, by tracking a in terms of the Weber number we show that a ~ We⁻¹. Therefore,

$$t_c\sim W{e}^{-1}.$$ (4.14)

Model comparison with simulations

Figure 7 a) shows the collapse time in terms of the Weber number for three different Reynolds numbers (or pool viscosities), where the Weber number was varied by changing the surface tension with R_j = U₀ = ρ₀ = 1. Here we observe that in all the cases the closure time reaches a maximum and then it decreases. This can be explained by a collapse regime transition from capillary dominated to the air suction dominated. Indeed, in figure 7 a) we show that the scalings obtained in equations 4.10 and 4.14, match very well with the simulations. Here we also notice that for a Reynolds number Re = 2 × 10³, the maximum time of closure is ≈ 40 % shorter than for the potential flow case. This can be attributed to a smaller radial extension due to viscous dissipation. In contrast, for We = 600, and Re = 2000, the time of collapse is ≈ 50 % larger than for the potential flow. Here we argue that viscosity delays the rim and upward sheet formation, slowing down overall the closure dynamics. Since equations 4.14 and 4.10 do not depend on Re, we expect that the transition from a capillary deep seal to the pressure driven surface seal do not depend on Re. Figure 7 shows that the transition occurs at We ≈ 180 for all the explored Re, confirming the closure time independence from Re. This transition is within the same order of magnitude as that predicted by Aristoff & Bush (2009) for solid spheres for Bo ⩽ 3. We note that the difference between the predicted value of Aristoff & Bush (2009) and the present results might be due to the fact that the transition depends on the sphere density and size, with the surface seal transition proportional to the sphere size (Eshraghi et al. 2020).

Figure 7.

a)Dependency of the time of closure t_c on the Weber number We, for different Reynolds numbers Re. The black diamonds correspond to the simulation data points. The dotted black line correspond to a spline through the simulation data. The red and blue dotted lines correspond to the approximations given by equations 4.10 and 4.14. The general trend is similar for each Reynolds number, we observe a global maximum for the time of closure where the two regimes meet. b) Superposition of cavity profiles by cylindrical jets at Re = 5 · 10³ and We = 50, varying the ambient gas density ρ_g in the range of four times and a quarter to that of atmospheric air. Here we observe that the time of cavity collapse, decreases with increasing air density.

Effect on ambient gas density

From our discussion in section 4.2, we would expect that the variation in pressure gradient would determine the time of collapse. Given that the pressure gradient is dominated by the ambient gas density, we expect that an increase in gas density decreases the time of cavity collapse. For We = 50, an increase in four times the gas density with respect to the ambient pressure ρ_g = 4, results in a cavity collapse at t ≈ 131, as shown by figure 7 b). In contrast, for gas densities at ambient pressure ρ_g = 1, and under ambient pressure ρ_g = 0.5, the cavity collapses at t ≈ 236 and t ≈ 266 respectively. These results are in line with research on water entry of a sphere at reduced ambient pressures (Gilbarg & Anderson 1948; Abelson 1970; Yakimov 1973). Furthermore, in the recent work of (Williams et al. 2022), it was found that the most important gas parameter influencing the lamella ejection is the gas density. Although, in the latter, air density prevents the cavity collapse by resisting the contact line movement (Williams et al. 2022).

Gas velocity during the impact process

To conclude our analysis of the closure dynamics upon the impact of a liquid jet, we examined the maxim gas velocity u_g,max at each instant for We = 50, 150 and 400. These Weber numbers, show representative cases of deep and surface seals. Our results indicate that there are three peaks in u_g,max. These peaks occur in three instants: I) just before impact, II) during cavity closure and III) when droplets are shed after the cavity collapse, regardless of the Weber number, see figure 8.

Figure 8.

Maximum simulated gas velocity for different Weber numbers, at Re = 10000. Roman numbers I, II and III, highlight velocity peaks and correspond to the times just before impact, cavity collapse and droplet ejection after cavity collapse. I) Simulation snapshot of the jet just before impacting the liquid pool. This peak is observed for all We at t ≈ 4. II) Simulation snapshots of the moment of the cavity collapse for We = 150 and We = 400. III) Simulation snapshot of We = 400 at the time of droplet ejection after cavity collapse. The simulation snapshots show the gas pressure (left hand side) and the magnitude of the gas velocity field (right hand side).

The highest value of u_g,max is observed at the moment of cavity closure, which is the largest for the highest Weber number. Additionally, we found that u_g,max remains relatively constant within a 10 % margin before impact for We > 200 (see figure S3 in supplementary materials). The latter observation indicates that the gas is displaced by the jet at a constant velocity prior to the impact, for the cases where a surface seal is formed.

Figure 8 (I), (II), and (III) illustrate simulation snapshots. In these images, the gas pressure field is depicted on the left, while the velocity field is shown on the right. Notably, in figure 8 (II) for We = 400, a low-pressure region forms within the cavity, whereas this phenomenon is not observed for We = 50 and 150. In contrast, the low pressure region is not observed for We = 50 and 150. This observation supports the hypothesis that the pressure gradient, rather than capillary forces, drives cavity closure in surface seals, while capillary forces dominate in deep seals.

5. Conclusions

We investigated the dynamics of high-speed microfluidic jet impacts on a liquid pool, focusing on the formation and collapse of cavities in the micrometer regime. Previous research primarily explored millimetre-scale projectiles at high Bond numbers, where cavity dynamics is determined by the interplay between hydrostatic pressure and gas inertia. In contrast, in this paper, we delved into the range of micrometer-sized projectiles, where the Bond number is low and the interplay between surface tension and gas-inertia govern the cavity closure dynamics.

Our experimental setup involved generating high-speed jets from thermocavitation. The resulting jets impacted a water pool, enabling us to qualitatively and quantitatively examine the cavity formation and closure in a parameter range that has not been previously explored. The numerical simulations we employed provided us with the freedom to explore a wide range of parameter combinations, unconstrained by experimental limitations.

Comparing our experimental and numerical results, we observed two distinctive regimes of cavity closure: capillary and air suction-driven. In the capillary-dominated regime We ⩽ 150, surface tension played a predominant role, and we obtained analytical expressions that indicated a closure time scaling with We (equation 4.10). In the air suction regime We ⩾ 180, the initial velocity of expansion was the crucial factor, and the closure time scaled inversely with We (equation 4.14).

Our findings revealed that the transition from capillary-driven to air suction-driven closure occurred around We ≈ 180 for the range of Reynolds numbers considered. This transition point was independent of Reynolds number, indicating an inertia dominated phenomena, where viscous dissipation is negligible. However, it would be interesting to test these ideas numerically and experimentally on a wider range of values for both the Reynolds and Bond numbers.

Our results also shed light on the intricate interplay between gas density and cavity collapse dynamics. Furthermore, we tracked the maximum gas velocity and pressures for the whole impact process. We observed that the largest maximum gas velocity was attained at the moment of cavity collapse. We found other peaks in maximum velocity values at instants just before jet impact and when droplets are shed after the cavity collapse. These insights into micrometer-scale cavity formation and closure offer valuable knowledge for applications like 3D printing and needle-free injections, and pollutant distribution transport.