betaFeedback

WorksPhysics and Astronomy

Crystal Nucleation Driven by Microjets from Hydrodynamically Interacting Thermocavitation Bubbles

Abstract

Primary nucleation predominantly dictates the physicochemical, optical and mechanical properties of crystals emerging from solution. Here, we reveal a previously unrecognized nucleation mechanism arising from bubble-bubble hydrodynamic interactions using laser-induced thermocavitation bubble pairs. Under confinement, antiphase generation of two bubbles produces a micron-scale, high-speed liquid jet. Upon impact with the surrounding supersaturated liquid, this microjet generates intense localized shear, thereby inducing crystallization of the model inorganic salt, potassium permanganate. Under identical laser energies and supersaturation levels, crystallization was not observed in single-bubble experiments, in which no microjet was present. We show that the crystallization probability increases with microjet velocity. Boundary integral simulations reveal how experimental parameters such as confinement, bubble spacing, and laser energy govern microjet formation and velocity. We quantify the relationship between jet velocity and the observed enhancement in crystal nucleation kinetics by incorporating shear into the kinetic prefactor in classical nucleation theory. By combining experiments, theory, and simulations, we establish how microjets emerging from bubble-bubble hydrodynamic interactions may influence crystallization kinetics, opening new opportunities for the manufacture of crystalline materials.

Nucleation from supersaturated solution is of fundamental interest in understanding how nature self-organizes seemingly uncorrelated molecules into complex crystalline structures [1, 2, 3]. Moreover, controlling nucleation has practical relevance for applications ranging from the manufacturing of functional materials that outperform their natural counterparts to understanding the mineral morphogenesis and the origin of homochirality [4, 5, 6, 7, 8]. However, nucleation, growth and other crystallization phenomena that often occur in dynamic multiphase environments in the presence of flows render nucleation control challenging. The use of light has emerged as a promising non-invasive approach due to its ability to spatiotemporally control the crystal nucleation and therefore the properties of emerging crystals [9, 10, 11, 12, 13]. Specific

to pulsed lasers, generally having femto- to nanosecond pulse width, the formation of crystals is associated with vapor (cavitation) bubbles formed due to the interaction of laser light with the solution [14, 15]. Despite recent progress, predictive control over crystal quality parameters such as size, purity, (pseudo)polymorphic form and morphology along with nucleation rate remains elusive [9, 15].

Although considerable effort has focused on controlling crystal quality parameters through thermodynamic variables, flow-induced hydrodynamic phenomena offer new opportunities in crystallization [16, 17, 18]. Among these hydrodynamic phenomena, cavitation bubbles are ubiquitous: as they localize high-velocity flows and interact with each other and confining surfaces through the surrounding liquid [19, 20]. However, the implications of these interactions for nucleation from solution remain largely underexplored [21] [22]. The cavitation of a single bubble, inducing strong localized flows and corresponding stresses, has been utilized in microfluidics [23, 24] and biomedical applications [25, 26]. As long as the bubble is sufficiently far from walls, free surfaces and other bubbles, the bubble will retain a spherical shape throughout its lifetime [27]. However, asymmetries in the bubble’s environment can lead to aniosotropic bubble dynamics, and even the formation of microjets [28]. These microjets of fast-moving liquid form due to the impulse generated as a result of the bubble’s interaction with its surroundings. Jetting consequently leads to transport of fluid and species away from the location of bubble inception. Common methods for induced microjets include the generation of bubbles near rigid boundaries, near free surfaces, or near other bubbles [29–31]. The jet emerging from bubble formation close to a rigid boundary can often cause surface damage [32, 33], which is therefore considered undesired in certain applications. With free interfaces hardly present in continuously functioning synthetic or physiological systems, bubble-bubble interactions present a promising approach for jet creation. Thus, inducing jets through interaction between multiple laser-induced bubbles is a versatile and non-invasive approach to controlling jet properties and has found applications in microrheology [34], single-cell poration [30, 35], mechanotransduction [36], and particle manipulation [37].

In this work, we show that hydrodynamic interactions between laser-induced thermocavitation bubbles can alter the kinetics of crystal nucleation in an aqueous supersaturated solution. Specifically, we identify a previously unexplored route to crystallization mediated by interacting bubble pairs confined within a microcapillary, see Figure 1. Under identical supersaturation conditions and laser energies, a single thermocavitation bubble does not induce crystallization within the experimental parameter range studied. In contrast, antiphase generation of two bubbles produces a micron-scale liquid jet traveling through the first bubble, driven by the expansion of the second bubble. Upon impact with the surrounding liquid, this microjet generates intense localized shear, whose magnitude scales with the velocity of the jet, leading to a substantial enhancement of the nucleation rate. In other words, the nucleation rate is observed to increase with increasing jet velocity in experiments. We first employ boundary integral simulations to elucidate how laser energy, bubble-bubble separation distance, and geometric confinement govern jet formation, morphology, and velocity. We then establish the relationship between jet velocity and crystal nucleation kinetics and interpret the observed behavior within the framework of classical nucleation theory. By integrating experiments, theory, and simulations, we demonstrate how bubble-bubble interactions under confinement can be engineered to control nucleation kinetics in solution.

Figure 1. Experimental setup and method. (A) Optical paths of nanosecond laser pulses that travel through two separate objectives triggering thermocavitation bubbles. Two cameras, one high-speed and one low-speed, provide access to different timescales, O(μs)O(\mu\mathrm{s}) and O(s)O(\mathrm{s}), respectively. (B) Illustration of thermocavitation bubbles interacting in a microcapillary. The first thermocavitation bubble (B1\mathrm{B}_1) is triggered at t=0t=0. The second bubble B2\mathrm{B}_2 is triggered when B1\mathrm{B}_1 is at its maximum size, X1,max⁡X_{1,\max}, thus making the bubbles antiphase. As B2\mathrm{B}_2 expands, a jet emerges within the B1\mathrm{B}_1 due to hydrodynamic interaction between bubbles. The distance of the center of B2\mathrm{B}_2 to nearest B1\mathrm{B}_1 interface at when X1=X1,max⁡X_1=X_{1,\max} is f−X1,max⁡/2f-X_{1,\max}/2, where ff is the distance between the laser foci. The daughter bubbles and cavity formed as a result of B1−B2\mathrm{B}_1-\mathrm{B}_2 interaction is represented with white dotted lines using circles and rectangle, respectively. Long after the bubbles disappear, KMnO4_4 solute crystals emerge from solution. The crystals then grow in size over minutes. The experimental parameters prescribed prior to each experiment, together with the quantities measured during the experiment, are summarized in a table. (C) Measured dynamic bubble sizes from high-speed microscopy images illustrated in panel B. Bubble B2\mathrm{B}_2 is formed when B1\mathrm{B}_1 reaches its maximum size at t=tB2t=t_{\mathrm{B}_2}. X1X_1 and X2X_2 correspond to the sizes of bubbles B1\mathrm{B}_1 and B2\mathrm{B}_2, respectively.

Results

Experiments

Bubble pairs are generated using two 532 nm pulsed lasers (pulse width 5-9 ns) synchronized through a delay generator. Vapor bubbles B1_1 and B2_2 are created by lasers 1 and 2, respectively, as illustrated in Figure 1. The dynamics of the bubbles and the resulting jet are governed by the absorbed laser energies, E1E_1 and E2E_2, which determine the maximum bubble sizes, X1,max⁡X_{1,\max} and X2,max⁡X_{2,\max}, the focal separation distance ff, the effective standoff distance f−X1,max⁡/2f-X_{1,\max}/2, and the time delay between bubble initiation. The formation times of bubbles B1_1 and B2_2 are denoted by tB1t_{\mathrm{B}_1} and tB2t_{\mathrm{B}_2}, respectively, with tB1=0t_{\mathrm{B}_1}=0. The effective standoff distance, defined at t=tB2t=t_{\mathrm{B}_2}, can be adjusted either by varying the focal separation ff or by modifying X1,max⁡X_{1,\max}. The latter is controlled through the laser energy supplied to B1_1 (E1E_1) and the capillary diameter dd [24]. Consequently, the experimentally prescribed parameters are the capillary diameter dd, focal separation ff, laser energies E1E_1 and E2E_2, and solution supersaturation SS. The primary observables are the jet dynamics characterized by the jet velocity, and if whether any crystal(s) are observed. Throughout this study, bubbles are generated in antiphase [38], such that B2_2 is initiated when B1_1 reaches its maximum size, X1,max⁡X_{1,\max}. Due to the high optical absorption coefficient of the solution (>1000 cm−1>1000\ \mathrm{cm}^{-1}), the incident laser energy is assumed to be fully absorbed. Further details of the optical setup, solution preparation, and experimental procedures are provided in the Methods section.

Figure 1B shows representative images from experiments, demonstrating the formation of a jet along with subsequent nucleation and growth of crystals from the supersaturated solution. The high-speed camera operating at 112,500 frames per second (fps) captures the bubble and jet dynamics, while the low-speed camera recording at 3 fps captures the crystal growth over seconds. The experiments are carried out on an aqueous KMnO4_4 solution with supersaturation S=1.5S=1.5 relative to the solubility of 0.076 g/g of water at 298 K [39]. Figure 1C highlights the formation of B2_2 at tB2=240 μst_{\mathrm{B}_2}=240\ \mathrm{\mu s} as B1_1 reaches its maximum size. Consequently, as B2_2 continues to grow, we notice the formation of a jet within B1_1 along the axis connecting the bubble centers. This fast-moving jet pierces the right side end of B1_1, creating significant shear upon impact, and even facilitating the emergence of cavity and daughter bubbles. After the collapse of B1_1 and B2_2, the residual daughter bubble(s) coalesce or dissolve over a millisecond to second time range. Crystals of observable size then emerge surrounding the residual bubble(s) over seconds. The formation of crystals was found to be independent of the daughter bubbles, discussed later in Figure 3, where even “non-piercing” jets lead to crystallization.

Figure 3. Incident jet velocities, jet morphology and cumulative nucleation probability. (A−CA-C) Incident jet velocities (vj,incdv_{\mathrm{j,incd}}) inside B1_1 at different standoff distances (f−X1,max⁡/2f-X_{1,\max}/2). The color of the markers represents the jet morphologies with distinct shapes, breakup behavior, and temporal evolution: pinch-off, straight, and piercing, as annotated using the respective colors. In addition, each jet morphology region is shaded using the respective colors with illustrative images of B1_1. The shades serve as a guide to the eye. The filled markers represent the experiments where crystals were observed inside the capillary after a fixed observation time, t=2 mint=2\ \mathrm{min} from the laser irradiation. Correspondingly, the empty markers represent experiments with no crystals observed at t=2 mint=2\ \mathrm{min}. Whether or not crystals were observed at this fixed observation time in an experiment is used to assign a binary value for probability, p=1p=1 or 00 for trials with and without crystal(s), respectively. The scalebar over B1_1 image in panel AA is 100 μm100\ \mu\mathrm{m}.

No crystallization was observed when only a single thermocavitation bubble, B1_1, was generated under otherwise identical conditions. Moreover, increasing the laser energy delivered by a factor of two likewise failed to induce crystallization under a single thermocavitation bubble event, see Supplementary section 1. This result rules out laser-induced nucleation facilitated by local evaporation around a thermocavitation bubble as a possible mechanism - a phenomenon previously shown to promote crystal nucleation [14]. The absence of nucleation with a single thermocavitation bubble points to the microjet that arises from bubble-bubble interactions as the key ingredient governing crystallization. Therefore, we investigate how the velocity of the jet, vjv_j, influences the onset of nucleation. Since the jet velocity governs both the probability and the kinetics of crystal nucleation, as discussed later, establishing a quantitative relationship between the jet velocity and the experimental parameters provided in Figure 1 is essential. Such a relationship would enable the rational design of laser excitation and confinement conditions to achieve desired nucleation outcomes. We employ a numerical model to resolve the full jet dynamics and determine the jet impingement velocity on the right side (opposite interface) of B1\mathrm{B}_1, which constitutes the key hydrodynamic parameter controlling crystallization.

Numerical simulation

We solve for the liquid momentum using an axisymmetric Boundary Integral Element Method, allowing us to simulate the interactions between two bubbles perfectly centered along the capillary axis. The rate of change in the position of the B1\mathrm{B}_1 interface is used to estimate the temporal evolution of the jet velocity. To decrease the code runtime, we reduce the computational domain by simulating the bubble-bubble interaction only using the near side of B1\mathrm{B}_1, see Figure 2A. This resembles B2\mathrm{B}_2 interacting with a meniscus, with the option to define the initial curvature of the B1\mathrm{B}_1 interface using a contact angle. In the literature, there is a numerical model for the bubble-meniscus interaction in a capillary partially filled with liquid [40], but that model fails to predict the influence of changes in capillary diameter. Moreover, bubble initialization in the model from the literature [40] employs a pressure pulse that cannot be directly correlated to experiments in this study. We address these shortcomings by introducing an initial bubble of certain pressure and volume proportional to the laser energy. Using simulations we explain the observed jet formation and measured jet velocities.

Figure 2. Temporal jet evolution and dependence of jet velocity on experimental parameters. (AA, BB) An illustration from Boundary Integral Element simulation and high-speed microscopy experiment showing jet displacement over time in a 200 μm\mu\mathrm{m} capillary and E2=19.6 μJE_2 = 19.6\,\mu\mathrm{J}. (CC) Measured temporal jet tip velocity at a fixed stand-off distance, f−X1,max⁡/2≈275 μmf-X_{1,\max}/2 \approx275\,\mu\mathrm{m}, for three different B2B_2 laser energies, E2E_2, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\,\mu\mathrm{J}. The discrete points are measurements from experiments and the continuous curves are from simulations. The incident jet velocity, vj,incdv_{j,\mathrm{incd}}, is defined as the velocity at which the jet reaches the right side end of B1B_1 (also illustrated in panel BB). The peak jet velocity, vj,pkv_{j,\mathrm{pk}}, is the maximum velocity a jet can achieve during its evolution. (DD) Peak jet velocity (vj,pkv_{j,\mathrm{pk}}) as a function of capillary diameter (dd) and absorbed laser energy for B2B_2 (E2E_2) from simulations. The three data points for E2E_2 correspond to the examples from panel CC.

In studies involving bubble-meniscus interaction in a capillary, the contact angle of the meniscus with the wall is observed to influence the jet velocity due to flow focusing. The velocity is reported to decrease with increasing contact angle [40, 41]. Thus, the initiation of the B1B_1 interface requires an estimate of its contact angle with the capillary wall. Using the angles measured from the images in our experiments, we impose an initial contact angle of 30∘30^\circ, consistent with others using borosilicate glass [42, 43]. We assume that the gas inside the bubble consists of non-condensable gas and water vapor. The partial pressure due to the non-condensable gas is assumed to undergo a polytropic process with λ=1.25\lambda= 1.25 as the polytropic index for underwater exploding bubbles [44]. Thus, the gas pressure over the liquid at the interface can be written as [45],

pl=pc+p0(V0Vb)λ−σglκ.(1)p_l = p_c + p_0\left(\frac{V_0}{V_b}\right)^\lambda- \sigma_{gl}\kappa. \tag*{(1)}

where pcp_c is the saturated vapor pressure of the condensable gas, VbV_b is the instantaneous bubble volume, p0p_0 and V0V_0 are the initial non-condensable gas pressure and bubble volume, respectively, and σgl\sigma_{gl} is the gas-liquid surface tension. κ=0.5(κ1+κ2)\kappa= 0.5(\kappa_1+\kappa_2) is the mean local curvature computed using the following equations for an axisymmetric surface [46]:

κ1=r′x′′−x′r′′(r′2+x′2)3/2,κ2=x′r.\kappa_1 = \frac{r'x''-x'r''}{(r'^2+x'^2)^{3/2}}, \qquad\kappa_2 = \frac{x'}{r}.

The superscripts ′' and ′′'' denote the first and second derivatives with respect to Γ\Gamma - the arc length of the piecewise smooth interface comprising the bubble surface, free surface and rigid boundary. We treat the liquid as inviscid and incompressible and the flow induced by the bubble motion as irrotational.

Using the theory for potential flow, the potential ϕ\phi is related to the velocity vector u\boldsymbol{u} as u=∇ϕ\boldsymbol{u} = \nabla\phi, which itself satisfies the Laplace equation,

∇2ϕ=0,(2)\nabla^2\phi= 0, \tag*{(2)}

in the flow domain Ω\Omega. For a sufficiently smooth function ϕ\phi defined over the piecewise smooth surface Γ\Gamma, Green's integral formula for a three-dimensional space can be written as [47],

c(p)ϕ(p)=∫Γ∂∂n(ϕ(q))1∣p−q∣ dΓ−∫Γϕ(q)∂∂n(1∣p−q∣) dΓ,(3)c(p)\phi(p) = \int_\Gamma\frac{\partial}{\partial n}\left(\phi(q)\right)\frac{1}{\lvert p-q\rvert}\,d\Gamma -\int_\Gamma\phi(q)\frac{\partial}{\partial n}\left(\frac{1}{\lvert p-q\rvert}\right)\,d\Gamma, \tag*{(3)}

where pp is a field point anywhere in Ω+Γ\Omega+\Gamma, q∈Γq\in\Gamma, nn is the outward unit normal, and c(p)c(p) equals 4π4\pi if p∈Ωp\in\Omega or 2π2\pi if p∈Γp\in\Gamma. Choosing pp on Γ\Gamma results in a solution for either ϕ\phi or ∂ϕ/∂n\partial\phi/\partial n on Γ\Gamma if the other is specified. Following a Lagrangian framework, the kinematic and dynamic boundary conditions on the bubble surface are written as

Δr=(∂ϕ∂r)Δt,Δx=(∂ϕ∂x)Δt,and(4)\Delta r = \left(\frac{\partial\phi}{\partial r}\right)\Delta t,\qquad\Delta x = \left(\frac{\partial\phi}{\partial x}\right)\Delta t,\qquad\text{and} \tag*{(4)}
Δϕ=[12∣∇ϕ∣2+p∞−plρL]Δt,(5)\Delta\phi= \left[\frac{1}{2}\lvert\nabla\phi\rvert^2+\frac{p_\infty-p_l}{\rho_L}\right]\Delta t, \tag*{(5)}

where p∞p_\infty is the pressure as x→∞x\to\infty and ρL\rho_L is the liquid density.

With Δp=p∞−pc\Delta p = p_{\infty}-p_c, we use the following characteristic scales to non-dimensionalize the associated parameters in simulation: Rmax⁡(ρL/Δp)1/2R_{\max}(\rho_L/\Delta p)^{1/2} for time, Rmax⁡(Δp/ρL)1/2R_{\max}(\Delta p/\rho_L)^{1/2} for potential and (Δp/ρL)1/2(\Delta p/\rho_L)^{1/2} for velocity. This implies that we non-dimensionalize lengths using Rmax⁡R_{\max}, which is the maximum radius of the spherical bubble when simulated in an unbounded liquid domain - with no capillary walls or B1_1 surface. Correspondingly, the change in potential on the B1_1 and B2_2 surface are given by [48, 49],

Δϕ^=[1+12∣u^∣2−ζ(V0/Vb)λ+ακ^]Δt^(6)\Delta\hat{\phi}=\left[1+\frac{1}{2}|\hat{u}|^2-\zeta(V_0/V_b)^\lambda+\alpha\hat{\kappa}\right]\Delta\hat{t} \tag*{(6)}

and

Δϕ^=[12∣u^∣2+ακ^]Δt^,(7)\Delta\hat{\phi}=\left[\frac{1}{2}|\hat{u}|^2+\alpha\hat{\kappa}\right]\Delta\hat{t}, \tag*{(7)}

respectively. Here ζ=p0/Δp\zeta=p_0/\Delta p is the strength parameter and α=σgl/(Rmax⁡Δp)\alpha=\sigma_{gl}/(R_{\max}\Delta p) is the surface tension parameter. The superscript  ^\hat{\ } denotes the dimensionless form of the parameters.

The bubble initialization requiring p0p_0 and V0V_0 is theoretically related to the initial potential energy of the bubble with p0V0∝Rmax⁡3p_0V_0\propto R_{\max}^3 [50]. Supplementary section 2 discusses the validation of the initial conditions together with the analytical equation for the estimation of ζ\zeta. It is essential to note that only two of the three dimensionless parameters ζ\zeta, R0/Rmax⁡R_0/R_{\max} and λ\lambda can be chosen independently. In all our analysis, upon fixing the non-dimensionalized initial radius R0/Rmax⁡R_0/R_{\max} as 0.1, we get ζ=313\zeta=313. The chosen value for ζ\zeta is within the 100 - 500 range for exploding bubbles [44], with the change in value within this range having negligible effect on bubble/jet dynamics, similar to others [51]. The numerical solution procedure to Eq. (3) can be found in Supplementary section 2.

Figure 2A – B shows the representative images for jet evolution at different time instances from experiments and simulations. The velocity vjv_j of the jet tip is determined based on the rate of change in the position of the jet tip. Figure 2C provides the temporal jet tip velocity for a chosen standoff distance, f−X1,max⁡/2≈275 μmf-X_{1,\max}/2\approx275\ \mu\mathrm{m}, for three different laser energies E2E_2 of B2_2. In all three cases, we first see the jet accelerate due to the momentum gain from B2_2 formation and bubble-bubble interaction. After reaching a maximum velocity, vj,pkv_{j,\mathrm{pk}} - peak jet velocity, the jet then decelerates due to surface tension counteracting the increase in surface area caused by the evolution of jet [40]. The incident velocity, vj,incdv_{j,\mathrm{incd}} is the velocity at which the jet strikes the right side of the first bubble, B1_1. It should be noted that vj,pkv_{j,\mathrm{pk}} is a characteristic velocity of the jet, while for a fixed vj,pkv_{j,\mathrm{pk}}, vj,incdv_{j,\mathrm{incd}} depends on X1,max⁡X_{1,\max} (or E1E_1 and dd). Thus, for an evolving jet vj,incdv_{j,\mathrm{incd}} increases with decreasing X1,max⁡X_{1,\max}, making it the right parameter to study jet characteristics in line with the literature [40]. However, vj,incdv_{j,\mathrm{incd}} is the necessary parameter to study crystallization characteristics, discussed later in the following section.

Discussion

In Figure 2C, the simulations show quantitatively similar jet evolution behavior when compared with high-speed microscopy experiments. However, there are some differences. The lifetime of bubble B2_2 in the simulation from Figure 2A (≈36 μs\approx36\ \mu\mathrm{s}) is shorter than in the experiments in Figure 2B ([O(100 μs)][O(100\ \mu\mathrm{s})]). We attribute the difference to neglection of viscous effects and capillary forces between the interface of B1_1 and the wall. Yet, as seen in Figure 2D, the model accurately predicts that vj,pkv_{j,\mathrm{pk}} has a power-law dependence on dd and E2E_2 with exponents of −1.18-1.18 and 0.91, respectively, closely consistent with previous observations [41]. Moreover, results presented in Figure 2D are in line with physical intuition. As dd decreases with other experimental parameters held constant, the peak jet velocity increases as capillary walls confine and direct momentum along the tube. Likewise, increasing E2E_2 provides more momentum for the jet, resulting in higher jet velocities. Furthermore, the Rmax⁡R_{\max} from simulations directly correlates to the laser energy from experiments. The determined power law fit of E2∝Rmax⁡2.603E_2\propto R_{\max}^{2.603} (see Supplementary section 2) is consistent with the reported empirical relation for energy difference above a threshold laser energy [52]. The numerical model discussed thus quantitatively captures the bubble-bubble interaction.

An early observation from high-speed microscopy recordings is the distinct jet morphologies shown as insets in Figure 3A. Based on the jet velocity, the jets show different morphologies with distinct temporal evolutions and breakup behavior. Therefore, the observed jet morphologies in experiments are categorized into three types: (i) pinch-off: jet breaks into droplet(s); (ii) straight: jet does not break up before it reaches the opposite end of B1_1, and therefore will deform the incident vapor-liquid interface but will not pierce it; and (iii) piercing: the jet pierces the opposite end of B1_1, leading to the creation of daughter bubble(s), see Figure 3A. The tendency of a jet to break up is evaluated using the Weber number We=ρLvj2rj/σglWe=\rho_Lv_j^2r_j/\sigma_{gl}, with rjr_j the jet radius. Jetting is reported to occur for We>4We>4 [53, 54]. In this work, WeWe ranges between 4 – 65, satisfying the necessary condition for jet formation and breakup. However, as the distance the jets can travel within B1_1 is limited to X1,max⁡X_{1,\max}, jets with relatively higher velocities possess sufficient momentum to reach the right side end of B1_1 before breaking up, while the jets with lower velocities break up (pinch-off) due to the Rayleigh instability [53]. The observed velocity range (4 – 8 m/s) for jet pinch-off is in agreement with the reported satellite-free drop generation range [55]. A clear observation from Figure 3A is that vjv_j increases with increasing laser energy, E2E_2, regardless of jet morphology at fixed f−X1,max⁡/2f-X_{1,\max}/2, E1E_1 and dd. This experimental observation is in line with physical intuition. The higher E2E_2, the more laser energy is available for momentum, resulting in higher jet velocities. The corresponding number of crystals observed for the experiments presented in Figure 3 is reported in Supplementary section 3.

Figure 3 shows the incident jet velocities for a range of supersaturations S={1.5,1.3,1.2}S=\{1.5,1.3,1.2\}, laser energies E1E_1 and E2E_2, and capillary diameters d={200,200,300} μmd=\{200,200,300\}\ \mu\mathrm{m}, respectively, for varying standoff distances. At the same stand-off distance we observe variation in the jet velocity between experiments in Figure 3. We attribute this variation to the experimental difficultly in maintaining the left side end curvature (contact angle) of B1_1 due to contact angle pinning and instabilities at the free interface. As Figure 3 classifies different jet morphologies, it can shed light on the question “Does jet morphology influence nucleation probability and kinetics?”. To this end, we introduce the cumulative nucleation probability at a fixed time lag, pp, as the measure of the likelihood of crystal formation due to microjet formation. Throughout this work, whether a solution did crystallize (p=1p=1, closed symbols) or not (p=0p=0, open symbols) was determined at a fixed observation time of 2 minutes. The results presented in Figure 3 show a clear dependence of pp on vj,incdv_{\mathrm{j,incd}} (incident jet velocity), SS (solution supersaturation) and dd (capillary diameter), with no influence of jet morphology. Apparent lack of a direct influence of jet morphology on nucleation probability is noteworthy, as one might intuitively expect morphologies with larger interfacial surface areas to promote heterogeneous nucleation and thereby enhance crystallization kinetics. At first glance, Figure 3A appears to indicate a morphology dependence, with most non-nucleating experiments with p=0p=0 clustered within the pinch-off regime (with blue shade). However, closer inspection of Figure 3B,C shows that this notion is misleading, as open and closed symbols are scattered across different morphologies with no distinct preference. Moreover, pinch-off type jets possess the highest surface-area-to-volume ratio with a pinched-off droplet. Therefore, this morphology should exhibit the highest nucleation probability if heterogeneous nucleation at the

vapor-liquid interfacial area were the dominant mechanism. However, the opposite trend is observed. Instead, nucleation correlates primarily with incident jet velocity, suggesting that jet-induced shear, rather than jet morphology, governs crystallization.

We see an increase in the experiments resulting in crystallization (p=1p=1, closed symbols) with increasing vj,incdv_{\mathrm{j,incd}}, SS and dd across all panels in Figure 3. Therefore, a characteristic parameter to compare experiments conducted under different experimental conditions is needed. In Figure 3, this characteristic parameter is established using a threshold jet velocity for observed crystal formation, vj,thv_{\mathrm{j,th}}, defined as the value above which the cumulative nucleation probability is greater than 0.5. For each set of SS and dd employed in this work, the pp as a function of vj,incdv_{\mathrm{j,incd}} is calculated using the logistic regression as shown in Figure 4A-C. We report an increase in vj,thv_{\mathrm{j,th}} with decreasing supersaturation - a larger characteristic jet velocity (or shear) is needed to trigger crystal nucleation at lower supersaturations. Furthermore, upon increasing dd from 200 to 300 μm\mu\mathrm{m} while maintaining S=1.3S=1.3, we observed a steep rise in the number of crystals formed (>20>20, refer Supplementary section 4). Therefore, the experiments in Figure 3C in 300 μm\mu\mathrm{m} capillary were performed with relatively lower supersaturation (S=1.2S=1.2) to facilitate easy counting of the number of crystals.

Figure 4. Cumulative nucleation probability and calculation of solute-solution interfacial tension. (AA–CC) The logistic regression function plot for the panels AA–CC in Figure 3, respectively. vj,thv_{j,\mathrm{th}} is the threshold velocity for crystal formation above which the cumulative nucleation probability, p≥0.5p \ge0.5. The errorbars are the standard errors. The representation of the markers is similar to Figure 3. (DD) Schematic of the cluster formation via particle addition and removal processes in the presence of shear flow. The arrows represent the linear velocity streamlines. (EE) The plot shows the estimates from Eq. (10). Appropriate values for the variables are supplied for the parameters discussed in this work, S={1.5,1.3,1.2}S = \{1.5, 1.3, 1.2\}, d={200,200,300} μmd = \{200, 200, 300\}\ \mu\mathrm{m} and vj,thv_{j,\mathrm{th}} from panels AA–CC, respectively. The slope corresponds to the value of BB in Eq. (10), with which we estimate the solute-solution interfacial tension σsl\sigma_{\mathrm{sl}}. The error associated with the vj,thv_{j,\mathrm{th}} is estimated from the shaded region, which corresponds to the standard errors of the logistic function in panels AA–CC.

The impingement of a jet onto the liquid surrounding the bubble can create shear [35], thereby altering the nucleation kinetics [17, 56], as illustrated in Figure 4D. However, the exact mechanism of how shear facilitates nucleation is still under debate, shrouded in mechanisms involving enhanced mass transfer due to flow, agitation-enhanced cluster aggregation and shear-induced molecular alignment of clusters [22, 57, 58]. From classical nucleation theory, the nucleation rate, JJ, which is the number of nuclei formed per unit volume per unit time, is expressed as [59],

J=ASexp⁡[−B/log⁡2S],(8)J = AS \exp\left[-B/\log^{2} S\right], \tag*{(8)}

where BB is a dimensionless thermodynamic parameter given by B=16πv02σsl3/(3kB3T3)B = 16\pi v_{0}^{2}\sigma_{\mathrm{sl}}^{3}/(3k_{\mathrm{B}}^{3}T^{3}). Here v0v_{0} is the molecular volume of the solute, σsl\sigma_{\mathrm{sl}} is the solute-solution interfacial tension, kBk_{\mathrm{B}} is Boltzmann’s constant and TT is the absolute temperature. AA is the supersaturation-independent kinetic parameter for homogeneous nucleation. The influence of liquid shear rate γ˙\dot{\gamma} on the nucleation rate is accounted for in AA with empirical evidence showing a power law dependence, A∝γ˙nA \propto\dot{\gamma}^{n} [60–64]. In addition, for aqueous glycine, Forsyth et al. [65] demonstrated AA to positively scale with the solution container’s surface area Π\Pi as A∝γ˙ΠA \propto\dot{\gamma}\Pi. Combining the relations, we get the empirical correlation A∝(vj/d)ndA \propto(v_j/d)^{n}d, with γ˙∼vj/d\dot{\gamma} \sim v_j/d (see Figure 4D) and Π∼πd\Pi\sim\pi d over unit length of the capillary. This thus establishes the relationship to the governing parameters accounted for in the pre-exponential kinetic factor AA, while keeping the thermodynamic parameters BB and SS constant.

By definition, JJ can be expressed as N/(tobsV)N/(t_{\mathrm{obs}}V), where NN is crystal count, tobst_{\mathrm{obs}} is observation time and VV the capillary volume. We establish the crystallization threshold criterion as N≈1N \approx1, with a fixed observation time of tobs=2 mint_{\mathrm{obs}} = 2\ \mathrm{min}. Therefore, near the crystallization threshold, this makes JJ only a function of capillary volume, J∝1/(πd2L)J \propto1/(\pi d^{2}L). Since we maintain a constant capillary length (LL) throughout this work, we simplify,

J/A=Sexp⁡[−B/log⁡2S]∝d−3(vj,th/d)−n/C.(9)J/A = S\exp\left[-B/\log^{2} S\right] \propto d^{-3}(v_{j,\mathrm{th}}/d)^{-n}/C. \tag*{(9)}

CC is a constant whose value is based on NN, tobst_{\mathrm{obs}}, LL, temperature, and solute-solvent properties [59]. The thermal penetration depth in the liquid surrounding the bubble is calculated using the lifetime of the bubbles, which is O(10 μm)O(10\ \mu\mathrm{m}). However, the momentum penetration depth due to jet impingement, calculated using Newton’s approximation of impact depth [66], is O(100 μm)O(100\ \mu\mathrm{m}). We therefore neglect the effect of temperature variations in the liquid surrounding the bubble in Eq. (9) and take T=298 KT = 298\ \mathrm{K} throughout our analysis. Correspondingly, we neglect the change in SS due to TT. Moreover, the solute penetration depth that would exist due to exchange of solvent at the B1_1 ends (gas-liquid interface) [14] is calculated to be O(1 μm)O(1\ \mu\mathrm{m}). This, however, is much less than the momentum penetration depth due to microjets, and therefore allows us to ignore the solute boundary layer surrounding B1_1 in our calculations. As a result, we take SS as the supersaturation in the bulk. By rearranging Eq. (9), we get

−log⁡[d3(vj,th/d)nS]=−B/log⁡2S+log⁡C.(10)-\log\left[d^{3}\left(v_{\mathrm{j,th}}/d\right)^{n}S\right]=-B/\log^{2}S+\log C. \tag*{(10)}

With three unknowns: nn, BB, and CC, in the above equation, at least three datasets of the other variables are required to compute the unknowns. To this end, we condense all the data presented in Figure 4A–C into three datasets. Thus for S={1.5,1.3,1.2}S=\{1.5,1.3,1.2\} and d={200,200,300}d=\{200,200,300\} μ\mum, the corresponding vj,thv_{\mathrm{j,th}} discussed in Figure 4A–C are respectively used. This results in n=0.71n=0.71 and B=0.049B=0.049. The estimates after substituting the appropriate values of variables in Eq. (10) are shown in Figure 4E, in which BB is the slope of the fit line. This estimated value for nn is in close agreement with the empirically observed 1–3 range from literature that employ various solutes and shearing techniques [60–65], see Supplementary section 5. Furthermore, using the estimated value for BB from the fit, we can calculate σsl=2.8\sigma_{\mathrm{sl}}=2.8 mJ/m2^{2} at 298 K, which is in close agreement with 3.16 mJ/m2^{2} calculated using the Mersmann relation [67]. See Supplementary section 6 for the calculations of σsl\sigma_{\mathrm{sl}}. The fact that the extracted values of nn and BB are both in agreement with literature values thus strengthens the validity of the proposed scaling of the kinetic parameter with microjet-induced shear rate in Eq. (10) in the context of classical nucleation theory [68, 69]. The proposed scaling suggests that a threshold jet velocity and, by extension, a critical shear rate above which nucleation is more likely. Moreover, the threshold jet velocity above which nucleation is more likely depends also on the solution supersaturation and capillary diameter. This dependence can be rationalized within the classical nucleation theory framework, as increasing SS and dd would increase the amount of solute available to form nuclei within the volume of interest.

The model solute studied, KMnO4\mathrm{KMnO_4} has an increasing solubility with temperature and therefore is expected not to crystallize due to the local heating of solvent by laser. This study thus distinguishes itself from the light-driven crystal nucleation of retrograde (inverse) soluble compounds via heating [70]. Moreover, a single cavitation bubble triggered under identical conditions did not induce crystallization. Therefore, previously proposed solute accumulation at the vapor-liquid interface due to thermocavitation[14] cannot explain observations in this study. Thus, the observed correlation of higher cumulative crystallization probabilities with increasing jet velocity in Figure 4A–C along with the CNT analysis presented in Figure 4E supports the hypothesis of shear-induced nucleation emerging from hydrodynamic bubble-bubble interactions. This presents a novel route to nucleation from solution. However, a critical look into these results, as well as a careful comparison to the literature, is needed.

The first question we should ask is: Can shear influence the nucleation in this work? Shear induced nucleation is reported for colloids, proteins, polymers and other soft matter systems, as well as for molecular solutes [62, 71–73]. Unlike colloids, imaging of individual molecules in solution under shear is experimentally challenging. Therefore the precise shear-induced nucleation mechanism for molecular solutes has so far remained elusive. In this work, we estimate shear rates O(105 s−1)O(10^{5}\ \mathrm{s^{-1}}) – among the largest reported in literature [22], applied over the lifetime of the microjet O(10−6 s)O(10^{-6}\ \mathrm{s}) – the shortest reported in literature (see Supplementary section 5). To answer the question above, we employ the Péclet number, Pe=γ˙l2/DPe=\dot{\gamma}l^{2}/D, as the critical threshold where flow-driven transport outpaces quiescent relaxation due to thermal fluctuations. ll is the characteristic length scale and DD is the diffusion coefficient. For KMnO4\mathrm{KMnO_4}, DD is taken as 1×10−91\times10^{-9} m2^{2}/s. For the calculation of ll, we use the critical cluster diameter of KMnO4\mathrm{KMnO_4} from CNT in the absence of any external field for the lowest supersaturation employed in this work, S=1.2S=1.2. The critical cluster diameter is calculated as 1.64 nm, see Supplementary section 7 for the calculation [59]. Using the above values, the estimated Pe≈O(10−4)Pe\approx O(10^{-4}), which is significantly lower than unity – not sufficient to argue that shear is a prominent mechanism in the enhanced nucleation observed in this work. Alternatively, we can calculate to what length scale the pre-nucleation clusters need to grow so that Pe≈1Pe\approx1. This calculation suggests a characteristic cluster size in the range of 100 to 300 nm. From a thermodynamic point of view, it is problematic to have such large clusters in (metastable) equilibrium. Therefore, it is more natural to assume that the effect of shear impacts the crystal growth stage rather than the nucleation stage. In the crystal growth stage, the length scales are obviously much larger than the relevant scales in the nucleation stage.

On the theoretical side, Mura and Zaccone [17] developed a shear-modified classical nucleation theory (CNT) framework where nucleation is governed by two competing effects: enhanced molecular transport and nucleus straining. This framework predicted an optimal shear rate at which the nucleation rate is enhanced to an order of magnitude compared to the case in the absence of flow. Debuysscher et al. [22] explored the proposed framework experimentally with glycine reaching shear rates as high as O(104 s−1)O(10^{4}\ \mathrm{s^{-1}}) in a millimeter-sized capillary. These authors observed a maximum enhanced nucleation at shear rates O(103 s−1)O(10^{3}\ \mathrm{s^{-1}}) with nucleation rate decreasing at higher shear rates as qualitatively predicted by Mura and Zaccone [17]. Yet, in line with our conclusion, Debuysscher et al. [22] also concluded that their experiments cannot be described by shear-modified CNT.

To explain the observed shear-induced nucleation enhancement in their experiments, Debuysscher et al. [22] and other groups refer to the existence of non-classic prenucleation clusters that are larger than what CNT predicts [74, 75]. As prenucleation clusters are presumed to be both larger and diffuse more slowly, PePe numbers of order one can be reached, emphasizing the role of flow over diffusion. Also, due to growing indirect experimental evidence supporting the existence of non-classical clusters as large as 100 nm [65] for a broad range of inorganic aqueous solutions [76–78], we acknowledge our results could be interpreted in this context. However, the exact physicochemical composition of these clusters is experimentally challenging to pinpoint. We also acknowledge that such large prenucleation clusters are problematic conceptually. For our system, this would mean equilibrium cluster sizes involving 108−10910^{8} - 10^{9} ions, which is theoretically very difficult to accept without having to invoke some unknown stabilization mechanism. Therefore, we would like to also propose alternative explanations for our observations.

One factor that may play a role in the enhanced nucleation is the supersonic airflow due to the cavity collapse (see Figure 1B and Supplementary section 8). The supersonic airflow generated at the impacted B1B_1 interface would enhance the γ˙\dot{\gamma} in PePe [79, 80]. The supersonic airflow exiting the cavity formed upon microjet impact closely resembles that of compressible flow through a nozzle - where here the “nozzle” is a liquid cavity shrinking rapidly in time [81], see Supplementary section 8. While the presence of the cavity is evident for piercing jets, the straight jets however due to its relatively low velocities would create cavities with extremely short length and time scale - which are experimentally challenging to capture. Overall, this supersonic airflow, extensively studied in literature [82, 83], with velocities O(1000−10000 m/s)O(1000 - 10000\ \mathrm{m/s}) over the liquid nozzle can create larger shear rates [O(108 s−1)][O(10^{8}\ \mathrm{s}^{-1})] (see Supplementary section 8 for calculation). These transient shear rates are three orders of magnitude larger than what we predict on the basis of the incident jet velocity, and may create localized hydrodynamic environments in which advective transport and mechanical straining of much smaller prenucleation clusters become significant [17]. At the same time, the timeframe in which the shear rate is so high is also very small. Whether it is long enough to allow for formation of embryonic nuclei is unclear, but the potential for enhanced nucleation is obviously there during the cavity collapse.

Another mechanism that may contribute to the enhanced nucleation rate involves the coupled concentration and temperature fields surrounding the bubble. The expanding bubble is accompanied by both a solute-rich accretion layer, approximately O(100 nm)O(100\ \mathrm{nm}) thick, and a thermal boundary layer extending several micrometers into the liquid [14]. Because the jet originates from liquid adjacent to the bubble interface, the strong extensional flow associated with jet formation is expected to stretch and thin both layers. Upon impact and coalescence with the opposite meniscus, the reduced thickness of the thermal boundary layer may enhance heat transfer away from the concentrated accretion zone, resulting in local cooling. Since the accretion layer already possesses an elevated solute concentration, even a modest temperature decrease could transiently increase the local supersaturation, thereby providing an additional thermodynamic driving force for nucleation alongside the hydrodynamic effects discussed above. Just like the supersonic airflow created due to the transient cavity collapse explanation, confirming thermal effects is experimentally challenging due to the time and length scales involved in microjet evolution. Yet these two explanations could work in concert with the shear-induced nucleation hypothesis.

An observation worth mentioning is the different induction times reported in the literature compared to this study. The induction time for crystallization is defined as the time elapsed from the moment of shear until the first crystal is detected. In this work, we observe an induction time in the order of milliseconds to seconds - restricted within the fixed observation of 2 minutes. However, the induction times reported for different solutes and shearing methods in the literature is in orders of tens of seconds to minutes as tabulated in Supplementary section 5 [60–65]. It should also be noted that the maximum shear rate reported in literature is two orders of magnitude lower than this work, O(105 s−1)O(10^{5}\ \mathrm{s}^{-1}). While, in literature, the applied shear spans the whole induction time when compared to a very short time O(10−6 s)O(10^{-6}\ \mathrm{s}) in this work - the lifetime of the microjet. This characteristic of microjet-induced shear described in this work thus offers a unique opportunity to separate the nucleation stage from the crystal growth stage. As the controlled shear experiments in the literature apply the shear for a prolonged time O(10 s)O(10\ \mathrm{s}), it is very likely that the growth stage is affected by the shear.

In summary, we demonstrated a crystal nucleation mechanism driven by microjets generated through bubble-bubble hydrodynamic interactions in a supersaturated aqueous solution of a model salt. We identified a threshold jet velocity above which nucleation becomes significant and showed that this threshold increases with decreasing supersaturation and capillary diameter in experiments. By combining experiments, simulations, and theory, we propose microjet-induced shear as the mechanism promoting crystal nucleation. More broadly, this work reveals bubble-bubble hydrodynamic interactions as a viable route for flow-assisted control of nucleation kinetics and phase transitions - a previously overlooked phenomenon. The prospective applications range from geothermal magma flows [84] to next-generation manufacturing technologies for pharmaceuticals[85] and high-performance perovskite solar cells [86].

Methods

Optical setup

For sample illumination, we used white LED light (KL1500, Zeiss). An infinity-corrected 10x objective (Plan Fluor, Nikon) with 0.3 numerical aperture was used for laser 1. An 8x objective (Planapochromat, Carl Zeiss) with 0.1 numerical aperture was used for laser 2. The 10x objective was also used for imaging with resolution ≈2 μm/px\approx2\,\mu\mathrm{m}/\mathrm{px} for a high-speed camera (FASTCAM NOVA S16, Photron) and ≈0.74 μm/px\approx0.74\,\mu\mathrm{m}/\mathrm{px} for a low-speed camera (Imager Pro LX 16M, LaVision). The basis on how to build the setup with bill of materials is reported elsewhere [87]. Both lasers 1 and 2 are pulsed Nd:YAG lasers with the configurations listed in Table. I.

ParameterLaser 1Laser 2
ModelNano L 90-100Powerlite DLS 8000, Continuum
Wavelength532 nm532 nm
Pulse duration5 ns7 − 9 ns
Diameter at source4 mm9 mm
Diameter at objective8 mm3 mm
TEM₀₀yesno

Table I. Laser parameters.

Solution preparation

Ultrapure water (18.2 MΩ cm, ELGA Purelab) was used to prepare the supersaturated KMnO₄ (223468, Sigma-Aldrich) solutions in an 8 ml vial (SF8, BGB). A solubility of 7.6 g/100g-H₂O at 298 K for KMnO₄ was used [39, 88]. After preparation, the solution was stirred at 1000 rpm over a hot plate maintained at 333 K for one hour to accelerate the dissolution. Then the solution was placed in the oven maintained at 323 K overnight (> 12 hours) for complete dissolution. During the experiments, the vials were placed over a hot plate maintained at 333 K and constantly stirred at 1000 rpm.

Experimental protocol

For each experiment, the solution was freshly loaded into a new capillary. The warm solution from the hot plate was transferred using a micro-pipette. The warmth prevents spontaneous nucleation while handling. Once the supersaturated solution was transferred, the capillary ends were covered using droplets of silicone oil (378321, Sigma-Aldrich) to prevent evaporation. The step-by-step experimental procedure during each trial is detailed in Table. II.

Data availability

All data and code supporting the findings of this study are available from the corresponding author upon request.

Acknowledgments

This work was funded by LightX project under NWO Open Technology Programme (project number 16714). We thank Ing. E.F.J. Overmars for his support with the Nano L 90-100 laser procurement. We thank the members of LightX user committee: Dr. Jana Sonnenschein (Bayer AG), Dr. Bart Zwijnenburg (Nobian B.V.), Dr. Rob Geertman (Janssen Pharmaceutica), Dr. Andreas Sieber (Lonza Group) and Prof. dr. Antoine E. D. M. van der Heijden (TU Delft).

Description of the event sequenceTime sequence (minutes)
1. Fill the capillary with supersaturated solutionat 0
2. Place the filled capillary over the sample stage in the microscope and allow it to cool down. The LED light is switched off during this process.0 − 2
3. Scan the capillary end to end to check if any crystals have formed. If there are no crystals, then align the capillary with its mid along the length over the objective.2 − 6
4. Trigger a few laser pulses with the shutter closed to warm up the lasers.6 − 6.5
5. Prepare the camera and oscilloscope to wait for the trigger. The laser is shot at ≈ 7 min.6.5 − 7
6. The high-speed camera records for 3 seconds, while the low-speed camera continues recording until 2 min. After the 5th second, the intensity of the LED is kept at a minimum, as the large exposure time of the low-speed camera allows in more light.7 − 9

Table II. Experimental protocol

Author contributions

N.N., A.V. and H.B.E. designed and performed the experiments; N.N., R.H. and J.T.P. analyzed the boundary integral element model; N.N. and H.J.M.K. analyzed the classical nucleation theory; N.N. and P.J.T.V. analyzed the logistic regression method; N.N. and V.K. developed the experimental setup; N.N., R.H., J.T.P. and H.B.E. wrote the manuscript.

Competing interests

The authors declare no conflict of interest.

Supplementary Information: Crystal Nucleation Driven by Microjets from Hydrodynamically Interacting Thermocavitation Bubbles

Nagaraj Nagalingam; Adithya Vignesh; Vikram Korede; Herman J.M. Kramer; Peter J.T. Verheijen; Jan Groenewold; Remco Hartkamp; Johan T. Padding; Huseyin Burak Eral

SI.pdf

Single bubble

Figure S1 illustrates B₁ formation (with no B₂) in a 200 μm capillary. We observed no crystal formation post the bubble collapse within the standard observation time of 2 minutes. The laser energy supplied is E1=304.75 μJE_1 = 304.75\,\mu\mathrm{J}, which is approximately twice the laser energy supplied for B₁ in 200 μm capillary in the main text.

Figure S1. B₁ formation with no B₂. No crystals were recorded after a fixed observation time of 2 minutes from laser irradiation. The laser energy supplied is E1=304.75 μJE_1 = 304.75\,\mu\mathrm{J}.

Boundary Integral Element Method (BIEM) solution procedure.

For an axisymmetric problem with cylindrical coordinates (r,θ,z)(r,\theta,z), all the variables are independent of the rotation angle and therefore can be expressed as [1, 2],

1∣p−q∣ dS=4r(ξ)D[(dzdξ)2+(drdξ)2]12K(k) dξ(11)\frac{1}{\lvert p-q\rvert}\,\mathrm{d}S = \frac{4r(\xi)}{D} \left[ \left(\frac{\mathrm{d}z}{\mathrm{d}\xi}\right)^2+ \left(\frac{\mathrm{d}r}{\mathrm{d}\xi}\right)^2 \right]^{\frac{1}{2}} K(k)\,\mathrm{d}\xi \tag*{(11)}

and,

∂∂n(1∣p−q∣)dS=4r(ξ)D3{[dzdξ(r(ξ)+ri)−drdξ(z(ξ)−zi)]2k2(ξ)dzdξriE(k)1−k2(ξ)+2k2(ξ)dzdξriK(k)}dξ(12)\frac{\partial}{\partial n} \left( \frac{1}{\lvert p-q\rvert} \right)\mathrm{d}S = \frac{4r(\xi)}{D^3} \left\{ \left[ \frac{\mathrm{d}z}{\mathrm{d}\xi}(r(\xi)+r_i)- \frac{\mathrm{d}r}{\mathrm{d}\xi}(z(\xi)-z_i) \right] \frac{2}{k^2(\xi)} \frac{\mathrm{d}z}{\mathrm{d}\xi}r_i \frac{E(k)}{1-k^2(\xi)} + \frac{2}{k^2(\xi)} \frac{\mathrm{d}z}{\mathrm{d}\xi}r_iK(k) \right\}\mathrm{d}\xi \tag*{(12)}

With each elements of the boundary parameterized by ξ∈[0,1]\xi\in[0,1], p=(ri,zi)p=(r_i,z_i) and q=(r(ξ),z(ξ))q=(r(\xi),z(\xi)). D=[(r(ξ)+ri)2+(z(ξ)−zi)2]12D=[(r(\xi)+r_i)^2+(z(\xi)-z_i)^2]^{\frac{1}{2}} and k=2r(ξ)ri/Dk=2\sqrt{r(\xi)r_i}/D. K(k)K(k) and E(k)E(k) are the complete elliptic integrals of the first and second kind respectively, approximated by the polynomial expansion [3],

K(k)=P(η)−Q(η)ln⁡(η)(13)K(k)=P(\eta)-Q(\eta)\ln(\eta) \tag*{(13)}
E(k)=R(η)−S(η)ln⁡(η).(14)E(k)=R(\eta)-S(\eta)\ln(\eta). \tag*{(14)}

PP, QQ, RR and SS are tabulated polynomials of fourth-order, and η=1−k2\eta=1-k^2.

When the point under consideration (pp) lies within the interval of integration, logarithmic singularity in K(k)K(k) and E(k)E(k) exists at η=0\eta=0. In addition, a special case exists when this point lies over the axisymmetric axis. The detailed calculations required for these cases are reported in thesis by [2].

Numerical integration

The integral equation is solved numerically along the surface Γ\Gamma as [1, 4]$,

2πϕi=∑j=1N∫Sj∂∂n(ϕj)1∣pi−qj∣ dΓ−∑j=1N∫Sjϕj∂∂n(1∣pi−qj∣) dΓ.(15)2\pi\phi_i=\sum_{j=1}^{N}\int_{S_j}\frac{\partial}{\partial n}(\phi_j)\frac{1}{\lvert p_i-q_j\rvert}\,\mathrm{d}\Gamma-\sum_{j=1}^{N}\int_{S_j}\phi_j\frac{\partial}{\partial n}\left(\frac{1}{\lvert p_i-q_j\rvert}\right)\,\mathrm{d}\Gamma. \tag*{(15)}

Since we have a deforming bubble, it is necessary to have a continuously differentiable surface at each node points. For this purpose, the surface contour is approximated by a finite number of cubic elements as [5],

r(ξ)=ajr+bjrξ+cjrξ2+djrξ3,(16)r(\xi)=a_j^r+b_j^r\xi+c_j^r\xi^2+d_j^r\xi^3, \tag*{(16)}
z(ξ)=ajz+bjzξ+cjzξ2+djzξ3.(17)z(\xi)=a_j^z+b_j^z\xi+c_j^z\xi^2+d_j^z\xi^3. \tag*{(17)}

As either ϕ\phi or ∂ϕ/∂n\partial\phi/\partial n is unknown over the surface, both the parameters are chosen to linearly vary within each surface element as,

ϕ(ξ)=[(1−ξ) ξ][ϕjϕj+1](18)\phi(\xi)=[(1-\xi)\ \xi]\begin{bmatrix}\phi_j\\\phi_{j+1}\end{bmatrix} \tag*{(18)}
∂ϕ(ξ)∂n=[(1−ξ) ξ][(∂ϕ/∂n)j(∂ϕ/∂n)j+1].(19)\frac{\partial\phi(\xi)}{\partial n}=[(1-\xi)\ \xi]\begin{bmatrix}(\partial\phi/\partial n)_j\\(\partial\phi/\partial n)_{j+1}\end{bmatrix}. \tag*{(19)}

For each element, the integrals on the hight-hand side of the Eq. (15) can be evaluated as [6],

Ie=[k1][ϕjϕj+1]−[k2][(∂ϕ/∂n)j(∂ϕ/∂n)j+1](20)I_e=[k_1]\begin{bmatrix}\phi_j\\\phi_{j+1}\end{bmatrix}-[k_2]\begin{bmatrix}(\partial\phi/\partial n)_j\\(\partial\phi/\partial n)_{j+1}\end{bmatrix} \tag*{(20)}

with

[k1]=∫01[(1−ξ) ξ]∂∂n(1∣pi−qj∣) dξ(21)[k_1]=\int_0^1[(1-\xi)\ \xi]\frac{\partial}{\partial n}\left(\frac{1}{\lvert p_i-q_j\rvert}\right)\,\mathrm{d}\xi \tag*{(21)}
[k2]=∫01[(1−ξ) ξ]1∣pi−qj∣ dξ.(22)[k_2]=\int_0^1[(1-\xi)\ \xi]\frac{1}{\lvert p_i-q_j\rvert}\,\mathrm{d}\xi. \tag*{(22)}

The global equations are obtained by summing the contributions from all elements. Thus

Hi,jϕj=Gi,j(∂ϕ∂n)j,i,j=1,2,3,…,N(23)H_{i,j}\phi_j=G_{i,j}\left(\frac{\partial\phi}{\partial n}\right)_j,\quad i,j=1,2,3,\ldots,N \tag*{(23)}

where

Hi,j=∑[(k1)i,j+δi,j c(p)], where(24)H_{i,j}=\sum\left[(k_1)_{i,j}+\delta_{i,j}\,c(p)\right],\ \text{where} \tag*{(24)}
δi,j=1 ∀ i=j and δi,j=0 ∀ i≠j,(25)\delta_{i,j}=1\ \forall\ i=j\ \text{and}\ \delta_{i,j}=0\ \forall\ i\ne j, \tag*{(25)}
Gi,j=∑(k2)i,j.(26)G_{i,j}=\sum(k_2)_{i,j}. \tag*{(26)}

Eq. S5 is composed of Fredholm integral equation of the first and second kind. During the time marching process, sawtooth instability over the moving boundary profile may appear due to the numerical solution to Fredholm integral equation of the first kind [4]. Therefore a five-point smoothing technique introduced by Longuet-Higgins and Cokelet [7] for equally spaced points is used,

f^i=fi−ω(fi−2−4fi−1+6fi−4fi+1+fi+2).(27)\widehat{f}_i=f_i-\omega(f_{i-2}-4f_{i-1}+6f_i-4f_{i+1}+f_{i+2}). \tag*{(27)}

Where ω=1/16\omega=1/16, fif_i is the quantity prior to smoothing and f^i\widehat{f}_i is the smoothed quantity. Besides, the nodes of the moving boundaries after updation can become too close to (or too far from) each other, which may also cause numerical instabilities [8]. Therefore, the boundary is re-discretized after each time step.

Initialization

The computational domain has the B1_1 interface on one end with the other end open. Gravity is neglected as the tube is horizontal and the vertical dimension (capillary diameter) is O(100 μm)O(100\ \mu\mathrm{m}). The initial B1_1 interface curvature radius is defined using the contact angle of B1_1 with the capillary wall, θ\theta. It is to be noted that this contact angle is dynamic during the simulations run and changes based on the fluid flow. The bubble B2_2 with radius R0R_0 is positioned at distance hh away from the B1_1 interface.

The node that connects the B1_1 interface to the capillary wall is considered a part of both the interface and the capillary wall. While solving the boundary integral equation, we treat this common node as a part of wall and impose ∂ϕ/∂t=0\partial\phi/\partial t=0 on it. The velocity of this common node along the wall (∂z/∂t\partial z/\partial t) is computed using the gradient Δϕ/Δz\Delta\phi/\Delta z at this node with respect to the neighbouring node on wall.

Validation

Below we illustrate the validity of model with respect to the initial contact angle (θ\theta) of B1_1. vj,pkv_{j,\mathrm{pk}} is the maximum velocity a jet can achieve during its development evolution - a numerical model is necessary to estimate the temporal change in jet velocity. The analytical expression for vj,pkv_{j,\mathrm{pk}} is proposed by Peters et al. [9] as,

vj,pk∝cos⁡θ(1+βcos⁡θ)blog⁡(sin⁡θ−b1−b+sin⁡θ−1)(28)v_{j,\mathrm{pk}}\propto\frac{\cos\theta(1+\beta\cos\theta)}{b\log\left(\frac{\sin\theta-b}{1-b}+\sin\theta-1\right)} \tag*{(28)}

where b=1+h0cos⁡θb=1+h_0\cos\theta. h0h_0 is a fitting parameter and β\beta is a proportionality factor.

Fig. S2A shows the relation between the peak jet velocity (vj,pkv_{j,\mathrm{pk}}) and the cosine of contact angle. The fit line is derived from Eq. (28) with 11.75 as the constant of proportionality and by substituting h0=0.26h_0=0.26 and β=0.84\beta=0.84 for d=200 μmd=200\ \mu\mathrm{m} [9]. The method thus predicts the correlation same as the analytical expression from literature. In addition, the change in vj,pkv_{j,\mathrm{pk}} for ζ=100−500\zeta=100-500 was less than 5% relative to ζ=313\zeta=313.

In Figure S2B the model predicts a power law dependence on dd for vj,pkv_{j,\mathrm{pk}} with −1.18-1.18 index, consistent with the previous observations [10]. Figure S2C shows that the

Figure S2. Validation of the Boundary Integral Element Method code. (A) Influence of the jet velocity due to change in the initial contact angle. The simulations are for d=200 μmd = 200\,\mu\mathrm{m}, h=275 μmh = 275\,\mu\mathrm{m} and Rmax⁡=88.8 μmR_{\max} = 88.8\,\mu\mathrm{m}. The fit line correspond to Eq. (28) with 11.75 as the constant of proportionality. (B) Peak jet velocity (vj,pkv_{\mathrm{j,pk}}) as a function of capillary diameter (dd) from simulations. (C) Simulated Rmax⁡R_{\max} as a function of absorbed laser energy in experiments, E2E_2. The average standard error over mean for E2E_2 is ±0.3 μJ\pm0.3\,\mu\mathrm{J}. Rmax⁡R_{\max} is the maximum radius of the bubble when simulated in an unbound liquid domain - with no capillary walls or B1 surface. The three data points correspond to the examples from Figure 2C in the main text.

Rmax⁡R_{\max} from simulations to directly correlate to the laser energy from experiments. The determined power law fit index E2∝Rmax⁡2.603E_2 \propto R_{\max}^{2.603} is consistent with the reported empirical relation for energy difference above a threshold laser energy [11].

Analytical equation for ζ\zeta calculation

With the dimensionless initial radius of the bubble R^0=R0/Rmax⁡\hat{R}_0 = R_0/R_{\max}, we calculate [4]

ζ=[R^03−1−3α(1−R^02)](λ−1)R^03λ(1−R^03(1−λ)).(29)\zeta= \left[\hat{R}_0^3 - 1 - \frac{3}{\alpha}\left(1-\hat{R}_0^2\right)\right]\frac{\left(\lambda-1\right)}{\hat{R}_0^{3\lambda}\left(1-\hat{R}_0^{3(1-\lambda)}\right)}. \tag*{(29)}

The above relation is derived from the Rayleigh–Plesset equation, using the boundary conditions dR/dt=0dR/dt = 0 at R=R0R = R_0 and R=Rmax⁡R = R_{\max}. Thus Eq. (29) can be used to calculate ζ\zeta, which is used as one of the initial conditions provided we establish R0/Rmax⁡R_0/R_{\max} and λ\lambda.

The crystal count for experiments reported in Figure 3 of main text

The crystal count recorded in experiments are reported in Figure S3. The results are for the experiments reported in Figure 3 of the main text.

Figure S3. Crystal count. The recorded crystal count for experiments from Figure 3 of main text. The markers are adapted from Figure 3 in main text. The crystals were recorded after a fixed observation time of 2 minutes from laser irradiation.

Bubble pair experiments within 300m capillary

Figure S4 illustrates the crystal formation in experiments for S=1.3S = 1.3 within 300 μm\mu\mathrm{m} capillary. The experiments involved formation of bubble pairs with E1=219 μJE_1 = 219\,\mu\mathrm{J} and E2=16.9 μJE_2 = 16.9\,\mu\mathrm{J}, similar to the experiments from the main text. In total the experiment was repeated 8 times out of which 5 trials resulted in >20> 20 crystals and the other 3 trials resulted in 2 − 3 crystals.

Illustration of crystal formation

Figure S4. Illustration of crystal formation. The crystals were recorded after a fixed observation time of 2 minutes from laser irradiation.

Index n values from literature

Table. SI contains the index value nn from the relation J∝γ˙nJ \propto\dot{\gamma}^{n} (in main text) referenced from literature.

Solute-solution interfacial tension calculation

Table. SII provides the values required for calculation of solute-solution interfacial tension σsl\sigma_{\mathrm{sl}} with B=16πvsl2/(3kBT3)B = 16\pi v_{\mathrm{sl}}^2/(3k_{\mathrm{B}}T^3).

Mersmann proposed a simple relation for the solute-solution interfacial tension calculation from thermodynamics as [18],

σ∝T(ρM)2/3log⁡(ρC∗ρL).(30)\sigma\propto T\left(\frac{\rho}{M}\right)^{2/3}\log\left(\frac{\rho}{C^*\rho_{\mathrm{L}}}\right). \tag*{(30)}

Where TT is the temperature, ρ\rho is the solute density, MM molar mass of solute, ρL\rho_{\mathrm{L}} is the solution density, and C∗C^* is the weight fraction in g/g of solution. By substituting the respective values from Table. SII and Table. SIII in the above relation, we calculate the value of σsl\sigma_{\mathrm{sl}} for KMnO4_4 relative to σsl=2.2 mJ/m2\sigma_{\mathrm{sl}} = 2.2\,\mathrm{mJ/m^2} for KCl [19, 20]. The calculated value of σsl\sigma_{\mathrm{sl}} for KMnO4_4 is 3.16 mJ/m2^2 at 298 K.

nnCompoundγ˙\dot{\gamma} (1/s)Induction time (s)Experimental setup design
1.443lysozyme0 − 10-Lid driven flow employing a rotating chip [12]
1.226butyl paraben0 − 2001200Taylor-Couette flow with rotating inner cylinder [13]
3.029isotactic polypropylene0 − 8-Linkam CSS-450 shear stage [14]
2.08paracetamol0 − 20050Rotating parallel plate geometry [15]
1glycine10 − 1000100Rotating Couette flow and capillary flow devices [16, 17]

Table SI. Empirically observed nn values from literature. γ˙\dot{\gamma} is the shear rate. The induction time is for the highest shear rate reported in the respective literature.

ParameterDescriptionValue
BBGradient(slope) [Figure 4 from main text]0.048
v0v_0Solute molecular volume9.71×10−29 m39.71 \times10^{-29}\ \mathrm{m^3}
kBk_BBoltzmann constant1.381×10−23 m2kg/(s2K)1.381 \times10^{-23}\ \mathrm{m^2kg/(s^2K)}
TTTemperature298 K
ρKMnO4\rho_{\mathrm{KMnO_4}}Solute density2700 kg/m32700\ \mathrm{kg/m^3}
ρL−KMnO4\rho_{\mathrm{L-KMnO_4}}Solution density1060 kg/m31060\ \mathrm{kg/m^3} (at 25 °C and S=1S = 1)
CKMnO4∗C^{*}_{\mathrm{KMnO_4}}Solute weight fraction0.0706 g/g of solution (at 25 °C and S=1S = 1)
MKMnO4M_{\mathrm{KMnO_4}}Molar mass of solute158.03 g/mol
NAN_AAvogadro constant6.023×1023 mol−16.023 \times10^{23}\ \mathrm{mol^{-1}}

Table SII. Physical parameter values for aqueous KMnO4_4 solution.

ParameterDescriptionValue
v0v_0Solute molecular volume6.238×10−29 m36.238 \times10^{-29}\ \mathrm{m^3}
ρKCl\rho_{\mathrm{KCl}}Solute density1989 kg/m31989\ \mathrm{kg/m^3}
ρL−KCl\rho_{\mathrm{L-KCl}}Solution density1175 kg/m31175\ \mathrm{kg/m^3} (at 25 °C and S=1S = 1)
CKCl∗C^{*}_{\mathrm{KCl}}Solute weight fraction0.2645 g/g of solution (at 25 °C and S=1S = 1)
MKClM_{\mathrm{KCl}}Molar mass of solute74.55 g/mol

Table SIII. Physical parameter values for aqueous KCl solution.

Critical cluster size

Using the free-energy barrier relation for a nucleus to form, the critical cluster radius rcr_c is given by [21],

rc=2σslM8.314ρTln⁡S.(31)r_c = \frac{2\sigma_{\mathrm{sl}}M}{8.314\rho T \ln S}. \tag*{(31)}

In the above equation the σsl\sigma_{\mathrm{sl}} calculated in Section SVI is used. For KMnO4_4 the values reported in Table SII, such as MM, ρ\rho and TT are used for calculation. The minimum value of S=1.2S = 1.2 used in this work is used for calculation. The calculation results in rc=0.82r_c = 0.82 nm.

Cavity collapse

Figure S5 shows the collapse of the cavity at the impacted B1B_1 interface by microjet. The collapse of the cavity creates a supersonic airflow. Since velocity is proportional to shear generated, while velocities O(101 m/s)O(10^1\ \mathrm{m/s}) create shear rate O(105 1/s)O(10^5\ 1/\mathrm{s}), velocities O(10000 m/s)O(10000\ \mathrm{m/s}) can create shear rate O(108 1/s)O(10^8\ 1/\mathrm{s}).

Figure S5. Collapse of the cavity. The red lines mark the boundary of the cavity. The white arrows represent the air exiting the cavity during its collapse.

Videos

Movie S1: Pinch-off jet. High-speed microscopy experiment with 112,500 fps showing the first thermocavitation bubble (B1B_1) is triggered at t=0t = 0. The second bubble B2B_2 is triggered when B1B_1 is at its maximum size thus making the bubbles antiphase. As B2B_2 expands, a jet emerges within the B1B_1 due to hydrodynamic interaction between bubbles (Figure 1B in main text). Each frame contains a timestamp and scale bar. The confining walls of the microfluidic channel are above and below the viewing plane. The experiment is for S=1.5S = 1.5 and d=200 μmd = 200\ \mu\mathrm{m}, with B2B_2 laser energies, E2=8.6 μJE_2 = 8.6\ \mu\mathrm{J}, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\ \mu\mathrm{J}.

Movie S2: Straight jet. High-speed microscopy experiment with 112,500 fps showing the first thermocavitation bubble (B1B_1) is triggered at t=0t = 0. The second bubble B2B_2 is triggered when B1B_1 is at its maximum size thus making the bubbles antiphase. As B2B_2 expands, a jet emerges within the B1B_1 due to hydrodynamic interaction between bubbles (Figure 1B in main text). Each frame contains a timestamp and scale bar. The confining walls of the microfluidic channel are above and below the viewing plane. The experiment is for S=1.5S = 1.5 and d=200 μmd = 200\ \mu\mathrm{m}, with B2B_2 laser energies, E2=13.3 μJE_2 = 13.3\ \mu\mathrm{J}, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\ \mu\mathrm{J}.

Movie S3: Crystal emergence. Low-speed microscopy experiment with 3 fps showing the emergence of KMnO4_4 crystals corresponding to the high-speed microscopy experiment from Movie S2: Straight jet. Each frame contains a timestamp and scale bar. The confining walls of the microfluidic channel are above and below the viewing plane. The experiment is for S=1.5S = 1.5 and d=200 μmd = 200\ \mu\mathrm{m}, with B2B_2 laser energies, E2=13.3 μJE_2 = 13.3\ \mu\mathrm{J}, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\ \mu\mathrm{J}.

Movie S4: Piercing jet. High-speed microscopy experiment with 112,500 fps showing the first thermocavitation bubble (B1B_1) is triggered at t=0t = 0. The second bubble B2B_2 is triggered when B1B_1 is at its maximum size thus making the bubbles antiphase. As B2B_2 expands, a jet emerges within the B1B_1 due to hydrodynamic interaction between bubbles (Figure 1B in main text). Each frame contains a timestamp and scale bar. The confining walls of the microfluidic channel are above and below the viewing plane. The experiment is for S=1.5S = 1.5 and d=200 μmd = 200\ \mu\mathrm{m}, with B2B_2 laser energies, E2=19.6 μJE_2 = 19.6\ \mu\mathrm{J}, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\ \mu\mathrm{J}.

Movie S5: Crystal emergence. Low-speed microscopy experiment with 3fps showing the emergence of KMnO4_4 crystals corresponding to the high-speed microscopy experiment from Movie S4: Piercing jet. Each frame contains a timestamp and scale bar. The confining walls of the microfluidic channel are above and below the viewing plane. The experiment is for S=1.5S = 1.5 and d=200 μmd = 200\ \mu\mathrm{m}, with B2B_2 laser energies, E2=19.6 μJE_2 = 19.6\ \mu\mathrm{J}, and for B1B_1 laser energy, E1=157.6 μJE_1 = 157.6\ \mu\mathrm{J}.

References

  1. [1]B. Taib, G. Doherty, and J. Blake, in Proceedings of 8th Australasian Fluid Mechanics Conference, Vol. 52 (1983).
  2. [2]B. B. Taib, Boundary integral method applied to cavitation bubble dynamics, Ph.D. thesis, University of Wollongong (1985).
  3. [3]C. Hastings Jr, J. T. Wayward, and J. P. Wong Jr, Approximations for digital computers, Vol. 2040 (Princeton University Press, 2015).
  4. [4]A. Dadvand, B. C. Khoo, M. T. Shervani-Tabar, and S. Khalilpourazary, Engineering Analysis with Boundary Elements 36, 1595 (2012).
  5. [5]R. H. Bartels, J. C. Beatty, and B. A. Barsky, An introduction to splines for use in computer graphics and geometric modeling (Morgan Kaufmann, 1995).
  6. [6]G. P. Lennon, P. L.-F. Liu, and J. A. Liggett, Water Resources Research 15, 1102 (1979).
  7. [7]M. S. Longuet-Higgins and E. Cokelet, Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences 350, 1 (1976).
  8. [8]M. T. Shervani-Tabar, A. Dadvand, B. C. Khoo, and M. R. H. Nobari, Theoretical and Computational Fluid Dynamics 23, 297 (2009).
  9. [9]I. R. Peters, Y. Tagawa, N. Oudalov, C. Sun, A. Pros- peretti, D. Lohse, and D. van der Meer, Journal of Fluid Mechanics 719, 587–605 (2013).
  10. [10]Y. Tagawa, N. Oudalov, C. W. Visser, I. R. Peters, D. van der Meer, C. Sun, A. Prosperetti, and D. Lohse, Phys. Rev. X 2, 031002 (2012).
  11. [11]V. Venugopalan, A. Guerra, K. Nahen, and A. Vogel, Phys. Rev. Lett. 88, 078103 (2002).
  12. [12]S. Stroobants, M. Callewaert, M. Krzek, S. Chinnu, P. Gelin, I. Ziemecka, J. F. Lutsko, W. De Malsche, and D. Maes, Crystal Growth & Design 20, 1876 (2020).
  13. [13]J. Liu and Å. C. Rasmuson, Crystal Growth & Design 13, 4385 (2013).
  14. [14]F. De Santis, R. Pantani, and G. Titomanlio, Polymer 90, 102 (2016).
  15. [15]C. Devos, A. Vananroye, R. Cardinaels, C. Xiouras, T. Van Gerven, and S. Kuhn, Soft Matter 19, 5896 (2023).
  16. [16]C. Forsyth, P. A. Mulheran, C. Forsyth, M. D. Haw, I. S. Burns, and J. Sefcik, Crystal Growth & Design 15, 94 (2015).
  17. [17]C. Forsyth, I. S. Burns, P. A. Mulheran, and J. Sefcik, Crystal Growth & Design 16, 136 (2016).
  18. [18]A. Mersmann, Journal of Crystal Growth 102, 841 (1990).
  19. [19]A. Faravar and M. Manteghian, International Journal of New Chemistry 6, 43 (2019).
  20. [20]A. J. Alexan­der and P. J. Camp, Crystal Growth & Design 9, 958 (2009).
  21. [21]D. Kashchiev and G. M. van Rosmalen, Crystal Research and Technology 38, 555 (2003).

Paper details

vlnc:W_mcekup5v
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 11

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents