betaFeedback

WorksPhysics and Astronomy

Self-similar and Universal Dynamics in Drainage of Mobile Soap Films

Abstract

Vertical soap films drain under the influence of gravity, as indicated by the downward motion of colorful horizontal interference fringes observed on their surfaces. In this study conducted with rectangular soap films, we experimentally characterize the descent dynamics of these isothickness fringes and report its self-similar nature. We also show that this result is equivalent to thickness profiles exhibiting a separation of space and time. By integrating new measurements with data from the literature across various conditions, we validate these properties and establish the universality of the dynamics, governed by a single physical scalar. Our findings provide new insights for proposing a drainage model and understanding the mechanism of marginal regeneration at the origin of the process.

Introduction

Soap films are thin liquid membranes with very particular properties. They maintain their shape at rest, like solids, forming minimal surfaces that have intrigued numerous physicists and mathematicians since the work of Plateau [1]. Simultaneously, their liquid nature enables soap films to be generated [2, 3] and to accommodate large deformations or self-heal when subject to perturbations [4, 5, 6, 7, 8, 9, 10, 11, 12]. Soap films exist across various scales, ranging from giant soap films [13, 14] to micrometer-sized soap films observed at the interfaces between bubbles within liquid foams [15]. Understanding and controlling the stability of soap films inside liquid foams is a multifactorial problem of paramount importance for industrial applications [16].

The drainage of soap films under the influence of gravity has a direct effect on their thinning dynamics and lifetime. This phenomenon captured the attention of renowned scientists in the 19th century, such as Maxwell and Brewster [17]. More recently, Mysels et al. [18] identified two limiting cases while considering vertical soap films made from different kind of surfactant solutions, highlighting that interfacial properties modify the nature of the 2D flow in the plane of the film and lead to specific, easily distinguishable flow features. Slow drainage was observed for “rigid” films that offer high resistance to any motion within the plane of the film, in contrast to the fast drainage of “mobile” films that display in-plane recirculation of film elements [19].

Mobile films are characterized by a smooth thickness profile associated with downward motion in the central part of the film, coupled with upward and rapid turbulent motion along the lateral borders of the film (Figure 1(a)). This coupling leads to typical drainage speeds scaling inversely proportional to the film width [18, 20, 21, 22]. The lateral flow has been identified by Mysels et al. [18] as being triggered by the process of marginal regeneration, which involves the formation of thin film elements at the contact between the soap film and the meniscus connecting the frame, and their rise along the lateral borders, leading to the downward motion in the central part of the film due to area conservation [22]. This scenario suggests many mechanisms at play, from the nucleation, growth, and upward transport of the thin film elements along the lateral borders to their merging with the central film. While thin film instabilities have been proposed to explain the formation of these elements [18, 19, 23, 24, 25], their complete dynamics lack characterization and modeling, which is a necessary step for understanding the drainage of vertical soap films.

Figure 1. (a) Temporal snapshots of a typical experiment: W=16W = 16 mm, H=25H = 25 mm, η=1.5\eta= 1.5 mPa.s. The central part shows horizontal interference fringes of homogeneous thickness. Thin film elements associated with the process of marginal regeneration are observed at the lateral and bottom borders (below dashed line). (b) Spatiotemporal diagram obtained along the white dashed line in (a). Colour-coded iso-thickness centers are tracked and lines interpolate points. (c) Thickness profile at different times, in colors. The red dotted line indicates z∗/H=0.3z^*/H = 0.3.

Nevertheless, the dynamics of the central part of mobile soap films have been the subject of numerous experimental studies. Two approaches coexist in characterizing the drainage dynamics. On one hand, some studies report measurements of the positions of isothickness lines (descent measurements) as functions of time [18, 26]. On the other hand, other studies involve measuring the evolution of thickness over time at specific positions (thinning measurements) [20, 21, 22, 27]. In both cases, it remains difficult to draw conclusions about the effect of control parameters such as the geometrical properties of the frame (except for the film width as mentioned above), the nature of the surfactant solution, or the film creation procedure. This challenge is partly due to the difficulty of obtaining a full spatiotemporal characterization of the dynamics. For example, in the thinning approach, the limited number of measurement positions, often just one and differing from one study to another, raises questions about the generalization of the results and the comparison between studies. In addition, there is no relation that connects the two approaches to enable comparison between them.

In this article, we present a robust procedure that allows a full spatiotemporal characterization of the drainage dynamics and an easy way to switch between the descent and thinning approaches. For this, we show the existence of a regime where the descent dynamics is self-similar, the thickness profiles exhibit a separation of space and time, and there is a power law relationship between thickness and time. For each experiment, we demonstrate that the full spatiotemporal dynamics can be described using only a few quantities: a time marking the onset of drainage, a physical scalar along with an exponent to quantify the temporal evolution of drainage, and a function providing the fine details of the self-similar dynamics and the shape of the thickness profile in the descent and thinning approaches, respectively. By compiling data from existing literature along with new experiments, we demonstrate that this regime prevails in diverse experimental conditions and that all data collapse onto a single master curve, highlighting the universality of the aforementioned function.

Material and methods

Experiments involve creating soap films within vertical rectangular frames with varying widths WW and heights HH, ranging from 4 to 20 mm and from 5 to 30 mm, respectively. These frames are constructed from glass fibers with two diameters DD, 250 and 700 μm\mu\mathrm{m}, meticulously joined edge to edge. The surfactant solutions were made of water-glycerol mixtures, with sodium dodecyl sulfate (SDS) present at a concentration of 5.6 g.L−1^{-1}. We systematically adjusted the water-glycerol ratio to modulate the bulk viscosity η\eta within the range of 1 to 20 mPa.s. In our range of mixture compositions, the surface tension remains nearly constant, with a value of approximately 35 mN.m−1^{-1} [28]. Eighteen different experimental conditions were tested (Tab. I). In each experiment, the frame was fully immersed in the surfactant solution and then swiftly removed, typically within 1 second. Time is measured from the moment the soap film is formed and fully dissociated from the surfactant solution. The position inside the film is identified by the coordinate zz, ranging between 0 at the bottom and HH at the top. We characterized the thickness profile in the central part of the soap film using thin-film interferences, coding for the local thickness of the film [29]. White light is reflected at the film’s surface, and the evolving interference patterns were recorded with a color camera. The entire setup was placed inside a controlled humidity chamber with high relative humidity, typically at Hr≥85%H_{\mathrm{r}} \ge85\%, which allows tracking the thinnest fringe over at least half of the film height before rupture. Indeed, as shown by Champougny et al. [30], relative humidity has a negligible influence on the drainage dynamics, but plays a role on the soap film lifetime. Depending on the parameters, it took between 20 s and 30 min for the 0.2 μm\mu\mathrm{m}-thick fringe to reach z/H=0.5z/H = 0.5. In all cases, the flow pattern corresponded to that of mobile soap films.

Figure 1(a) displays three snapshots of a typical experiment, illustrating the temporal evolution of the interference pattern. The full movie can be found in Supplementary Material (Movie 1) [31]. Each colored fringe, representing an isothickness line, moves downward, thus indicating a global thinning process. Along the vertical direction, the fringes are bounded by two regions. At the top of the film, the presence of a very dark zone indicates a black film, whose thickness, typically below 50 nm [17], is smaller than the minimum thickness measurable by interferometry. At the bottom of the film, a small region (below the black dashed line in Figure 1) is disrupted by thin elements shaped like ”tadpoles” [17], created by the process of marginal regeneration. These elements remain localized to a limited area of the soap film and are expected to have a negligible impact on the thinning dynamics, unlike the thin elements generated at the lateral borders. Apart from these edge effects, the pattern remains smooth, confirming a regular thickness profile from top to bottom in the central part of the film. We initially performed descent measurements that involved tracking the center positions of isothickness lines corresponding to intensity maxima and minima while using 660 nm filtered light (spatiotemporal diagram in Figure 1(b)). In an experiment, roughly fifteen fringes are tracked corresponding to thicknesses ranging between 0.1 and 2 μ\mum. For a given film thickness hh, we denoted z(h,t)z(h,t) the position of this thickness with respect to time. We noticed that the larger hh, the faster the process and the time evolution of z(h,t)z(h,t). Then, thinning measurements were deduced from the same set of data and thickness profiles h(z,t)h(z,t) were calculated for different times as displayed in Figure 1(c). The time evolution of the thickness profile is available in a movie that can be found in Supplementary Material (Movie 2) [31].

WW (mm)HH (mm)DD (μ\mum)η\eta (mPa.s)nnmm
4257001.50.90 ±\pm0.031.19 ±\pm0.01
8257001.50.81 ±\pm0.011.29 ±\pm0.01
16257001.50.72 ±\pm0.021.30 ±\pm0.05
20257001.50.70 ±\pm0.011.49 ±\pm0.01
162570010.72 ±\pm0.011.21 ±\pm0.02
16257001.20.62 ±\pm0.011.18 ±\pm0.01
162570020.78 ±\pm0.031.02 ±\pm0.02
16257002.90.64 ±\pm0.051.23 ±\pm0.04
162570051.08 ±\pm0.041.07 ±\pm0.01
162570014.80.62 ±\pm0.011.54 ±\pm0.01
1625700200.72 ±\pm0.021.44 ±\pm0.02
1652501.50.70 ±\pm0.231.60 ±\pm0.25
167.52501.50.70 ±\pm0.011.54 ±\pm0.01
16102501.50.53 ±\pm0.021.84 ±\pm0.03
1612.52501.50.63 ±\pm0.031.40 ±\pm0.02
16152501.50.74 ±\pm0.051.70 ±\pm0.05
16202501.50.46 ±\pm0.011.54 ±\pm0.01
16302501.50.50 ±\pm0.021.86 ±\pm0.06

Table 1. Experimental conditions and exponents nn and mm obtained from the fits with Eqs. (5) and (7), respectively.

Self-similar descent dynamics

In Figure 1(b), we observe that all curves display the same behavior whatever hh within one time dilation factor decreasing with hh: the evolution of zz for a given hh starts slowly, accelerates, and then slows down at a later stage. This evolution marks the presence of an inflection point, highlighting a maximum in speed that occurs at a fixed position zmax⁡z_{\max}, regardless of the thickness. These elements suggest that the descent dynamics exhibits self-similarity such as

z(h,t)=Hf(t−t0T(h)),(1)z(h,t) = H f\left(\frac{t-t_0}{T(h)}\right), \tag*{(1)}

with ff being an empirical function, independent of thickness hh, that monotonically decreases from 1 to 0, T(h)T(h) a function of hh that quantifies scale invariance in time, and t0t_0 a parameter that marks the onset of the self-similar regime.

We have tested this expression based on a drainage speed analysis in experiments. For each tracked thickness, we calculated the drainage speed V(h,t)V(h,t) by differentiating z(h,t)z(h,t) with respect to time. The drainage speed is represented as a function of time in Figure 2(a), while it is shown as a function of zz (on the vertical axis) in Figure 2(b). We then measured the maximum speed Vmax⁡(h)V_{\max}(h) and its corresponding time tmax⁡(h)t_{\max}(h) and position zmax⁡(h)z_{\max}(h), marked with circles and diamonds in Figure 2. As clearly visible in the inset in Figure 2(b), the maximum speed occurs at the same position, zmax⁡(h)≃0.8Hz_{\max}(h)\simeq0.8H, independent of hh, while Vmax⁡(h)V_{\max}(h) and tmax⁡(h)t_{\max}(h) are increasing and decreasing functions of hh, respectively. These measurements are confronted with Eq. (1). Taking the time derivative of this equation gives the drainage speed

Figure 2. (a) V(h,t)=∣∂z(h,t)/∂t∣V(h,t)=|\partial z(h,t)/\partial t| as a function of tt for various thicknesses hh, indicated by different colors, using the same data set as in Figure 1. For each thickness, the maximum speed Vmax⁡(h)V_{\max}(h) and its corresponding time tmax⁡(h)t_{\max}(h) are marked with open circles. (b) z(h,t)z(h,t) as a function of V(h,t)V(h,t) for the same data set, using the same colors as in (a). For each thickness hh, the trajectory, parameterized by time, progresses downward, and the maximum speed Vmax⁡(h)V_{\max}(h) and its corresponding position zmax⁡(h)z_{\max}(h) are marked with open diamonds. Inset: zmax⁡/Hz_{\max}/H as a function of hh. The dashed line represents zmax⁡=0.8Hz_{\max}=0.8H.

V(h,t)=∣∂z(h,t)∂t∣=HT(h)∣f′(t−t0T(h))∣.(2)V(h,t)=\left|\frac{\partial z(h,t)}{\partial t}\right|=\frac{H}{T(h)}\left|f'\left(\frac{t-t_0}{T(h)}\right)\right|. \tag*{(2)}

At t=tmax⁡(h)t=t_{\max}(h), the speed is maximum. Given that isothickness lines are moving downward, f′f' is negative, and the maximum speed is written as

Vmax⁡(h)=HT(h)∣f′(tmax⁡(h)−t0T(h))∣.(3)V_{\max}(h)=\frac{H}{T(h)}\left|f'\left(\frac{t_{\max}(h)-t_0}{T(h)}\right)\right|. \tag*{(3)}

Defining u=(t−t0)/T(h)u=(t-t_0)/T(h), ∣f′∣|f'| is maximum for umax⁡=(tmax⁡(h)−t0)/T(h)u_{\max}=(t_{\max}(h)-t_0)/T(h). With this notation, Eq. (3) yields

tmax⁡(h)=C1Vmax⁡(h)+t0,(4)t_{\max}(h)=C\frac{1}{V_{\max}(h)}+t_0, \tag*{(4)}

with C=Humax⁡∣f′(umax⁡)∣C=H u_{\max}|f'(u_{\max})|. Given that ff is assumed independent of hh, CC is constant, thus identifying an affine relationship between tmax⁡t_{\max} and 1/Vmax⁡1/V_{\max} with CC and t0t_0 being the slope and the intercept, respectively. This relation is validated experimentally in Figure 3(a), allowing a measurement of CC and t0t_0 for each experiment. Without loss of generality, we fixed that ∣f′∣|f'| reaches its maximum in u=1u=1, setting umax⁡=1u_{\max}=1 and the definition of T(h)=tmax⁡(h)−t0T(h)=t_{\max}(h)-t_0. Experimentally, T(h)T(h) is a decreasing function of the thickness (inset of Figure 3(b)), with a dependency that is consistent with a power law

Figure 3. (a) tmax⁡(h)t_{\max}(h) as a function of 1/Vmax⁡(h)1/V_{\max}(h). The dashed line is a fit with an affine function, and its intercept defines t0t_0. Inset: T(h)T(h) as a function of hh in log-log scale ; the best fit with a power law, as proposed in Eq. (5), gives n=0.72n = 0.72 and κ=2.4⋅10−4 m0.72.s\kappa= 2.4 \cdot10^{-4}\ \mathrm{m}^{0.72}\mathrm{.s}. (b) z(h,t)z(h,t) as a function of (t−t0)/T(h)(t-t_0)/T(h) rescaling all the descent positions from Figure 1(b) with color coding for thickness. Same data set as in Figure 1.

T(h)=κh−n,(5)T(h)=\kappa h^{-n}, \tag*{(5)}

where nn is a positive exponent and κ\kappa a constant.

In Figure 3(b), z(h,t)/Hz(h,t)/H is plotted as a function of (t−t0)/T(h)(t-t_0)/T(h) for all thicknesses. All data collapse onto the same master curve, which corresponds to the empirical function ff. With this procedure, we validate that for a given experiment, the descent dynamics exhibits a self-similar process fully described by two physical scalars, κ\kappa and t0t_0, a dimensionless function ff, and a scaling exponent nn.

Separation of space and time in the thickness profile

Given that T(h)T(h) is well approximated by a power law of hh (Eq. (5)), it is possible to switch from the descent approach to the thinning approach, additionally revealing a separation of space and time variables in the latter. In fact, Eqs. (1) and (5) result in

h(z,t)=(κt−t0)m(f−1(zH))m(6)h(z,t) = \left(\frac{\kappa}{t-t_0}\right)^m \left(f^{-1}\left(\frac{z}{H}\right)\right)^m \tag*{(6)}

with m=1/nm = 1/n and f−1f^{-1} the inverse function of ff. For any fixed position in space between 00 and HH, noted z⋆z^\star, this equation predicts a thickness evolution as

h(z⋆,t)∝(κt−t0)m(7)h(z^\star,t) \propto\left(\frac{\kappa}{t-t_0}\right)^m \tag*{(7)}

and Eq. (6) can thus be expressed as

h(z,t)=h(z⋆,t)Sz⋆(zH)(8)h(z,t) = h(z^\star,t) S_{z^\star}\left(\frac{z}{H}\right) \tag*{(8)}

where Sz⋆S_{z^\star} is a shape function that is proportional to (f−1)m(f^{-1})^m with a constant that depends on the chosen measurement position z⋆z^\star to ensure Sz⋆(z/H)=1S_{z^\star}(z/H)=1 in z=z⋆z=z^\star. The predictions of Eqs. (7) and (8) are tested with the same dataset used in Figure 1. In Figure 4, we observe that all thickness profiles at successive times collapse onto a single curve when plotting h(z,t)/h(z⋆,t)h(z,t)/h(z^\star,t) as a function of z/Hz/H, with z⋆=0.3 Hz^\star=0.3\,H marked by the red dotted line in Figure 1(c) and Figure 4. The time evolution of h(z⋆,t)h(z^\star,t) is consistent with Eq. (7), as evidenced by the power law behavior when plotting h(z⋆,t)h(z^\star,t) as a function of t−t0t-t_0 in a log-log scale (inset of Figure 4).

Figure 4. z/Hz/H as a function of h(z,t)/h(z⋆,t)h(z,t)/h(z^\star,t) with z⋆=0.3Hz^\star= 0.3H rescaling all the thickness profiles from Figure 1(c) with color coding for time. Inset: h(z⋆,t)h(z^\star,t) as a function of t−t0t-t_0 in log-log scale ; the best fit with a power law, as proposed in Eq. (7), gives m=1.30m = 1.30. Same data set as in Figure 1.

To conclude, for a given experiment, the thinning dynamics is characterized by a unique thickness profile, up to a multiplicative factor that decreases with a power law of (t−t0)(t-t_0). This validates the possibility to switch easily between the descent and thinning approaches based on power laws to describe the relationship between thickness and time (Eqs. (5) and (7)).

Universality of the approach and of the functions

We systematically studied the validity of our approach by performing on all experiments (see Tab. I) the same procedures described previously to characterize the descent and thinning approaches. This allowed us to quantify how the parameters and functions described above vary with frame dimensions (width WW and height HH), and with the bulk viscosity η\eta of the solution. With the eighteen different experimental conditions explored in this study, there is typically a factor 90 between the slowest and fastest time scales, representing a significant test for our approach.

In the descent approach, the function ff describes the time evolution of the positions of isothickness lines up to multiplicative scaling factors in hh (Eq. (1)). In Figure 5(a), the function ff is plotted for several values of the parameters WW, HH and η\eta. Very remarkably, all experiments aggregate onto the same curve, suggesting the function ff

Figure 5. (a) Rescaled descent positions for all experiments. Marker types represent widths, marker sizes correspond to heights, and marker colors indicate viscosity. For clarity, only two isothickness lines are plotted for each experiment — one with a small thickness and one with a large thickness. Consequently, 36 isothickness lines are displayed out of a set of 300 available. (b) Coefficient CC, defined in Eq. (4), as a function of film height HH. The dashed line is a fit highlighting the proportionality between the two quantities.

is universal for this set of parameters. In particular, the inflection point, which indicates the presence of a maximum speed, is consistently located at 0.8H0.8H for all values of hh whatever the experimental conditions. The universality can also be tested by examining the coefficient CC in Eq. (4). In experiments, we found that CC is proportional to HH (Figure 5(b)) and independent of other parameters such as the width WW and viscosity η\eta (see Appendix A), validating that ff is identical for all these experiments. Finally, the exponent nn in T(h)=κh−nT(h)=\kappa h^{-n} (Eq. (5)), was measured systematically. It varies roughly between 0.5 and 1, and statistics over the eighteen experiments lead to a mean value nˉ=0.70±0.14\bar{n}=0.70\pm0.14, with no discernible correlation with the control parameters (Tab. I).

The same holds for the thinning approach. The universality of the function Sz∗S_{z^*} with z∗=0.3Hz^*=0.3H is demonstrated in Figure 6(a) using the same dataset, where one sees a good collapse of the points on the left-hand side. Note that a qualitative illustration of the universality of the thickness profile is available in the Supplementary Material (Movie 3) [31], for soap films of different widths. The exponent mm in h(z∗,t)∝(κ/(t−t0))mh(z^*,t)\propto\left(\kappa/(t-t_0)\right)^m (Eq. (7)), takes values between 1 and 2, with a mean value mˉ=1.41±0.25\bar{m}=1.41\pm0.25. The universality of both ff and Sz∗S_{z^*} is consistent given that both functions are linked through Sz∗∝(f−1)mS_{z^*}\propto\left(f^{-1}\right)^m and that mm does not vary much between experiments. Nevertheless, the small variation between experiments was used to check that m=1/nm=1/n for a given experiment (Eq. (6)), as validated in Figure 6(b).

Figure 6. (a) Rescaled thickness profiles: the left side displays our data with the same markers as in Figure 5(a), while the right side shows data from the literature. Again, only two temporal profiles are retained for each experiment. The grey zone highlights the values z<0.2Hz<0.2H, which often correspond to an area of thickness perturbation associated with the presence of thin film elements at the bottom border of the soap film. (b) mm as a function of 1/n1/n for the eighteen different experimental conditions. The dashed line represents the unity line.

Furthermore, this assertion is tested using available data from the literature that provide thickness profile measurements at different times under various experimental conditions [18, 20, 32, 33]. Details on data extraction and processing are given in Appendix B. In the aforementioned studies, we reviewed images and/or text descriptions to ensure that the experiments displayed flow pattern characteristics of mobile soap films. Sz∗S_{z^*} with z∗=0.3Hz^*=0.3H obtained from these profiles is displayed on the right-hand side of Figure 6(a) for comparison with our data shown on the left-hand side of the same figure. In all cases, Sz∗S_{z^*} is remarkably similar for z/H>0.2z/H>0.2. For z/H<0.2z/H<0.2 (grey zone in Figure 6(a)), the profiles no longer superimpose, a fact that we attribute to the presence of the thin film elements generated at the bottom border by marginal regeneration, perturbing the thickness profiles.

For the same data from the literature, we did not have access to time t0t_0 to test the prediction of Eq. (7), h(z∗,t)∝(κ/(t−t0))mh(z^*,t)\propto\left(\kappa/(t-t_0)\right)^m. Assuming m=mˉ=1.41m=\bar{m}=1.41 for these data, we could check that h(z∗,t)−1/mˉh(z^*,t)^{-1/\bar{m}} is an affine function of time, validating the prediction (see Appendix B). This time dependence is also validated with studies providing measurements of the thickness at a given position only [20, 21, 22, 27]. Hudales et al. [20] found that the local thinning rate, ∂h/∂t\partial h/\partial t, scales as −h2-h^2, which is fully consistent with h(z∗,t)h(z^*,t) evolving as ∝1/(t−t0)m\propto1/(t-t_0)^m with m=1m=1. Berg et al. [21] characterized the evolution of the thickness with an empirical law that is consistent with our prediction in the long time limit with measured exponents mm between 0.9 and 1.8. We have also verified that such a time evolution is consistent with the data of Tan et al. [27] and Seiwert et al. [22].

Discussion

Our study demonstrates that the drainage of mobile soap films can be characterized using either of two approaches. In the descent approach, the positions of isothickness lines exhibit self-similarity, whereas in the thinning approach, thickness profiles exhibit a separation of space and time. In both approaches there is a power law relationship between thickness and time and the dynamics is characterized by a few quantities only: a time t0t_0 marking the onset of drainage, a physical scalar κ\kappa, an exponent (nn or mm) quantifying the temporal evolution, and a dimensionless function (ff or Sz∗S_z^*) providing the fine details of the self-similar dynamics and the shape of the thickness profile in the descent and thinning approaches, respectively.

Time t0t_0 marks the onset of drainage with respect to the time the soap film is formed. In practice, this time is shorter than the typical values of the scaling times T(h)T(h) measured for a given experiment, and t−t0t-t_0 could be replaced by tt in the long-term evolution of the thickness at a given position. Nevertheless, the values of t0t_0 are systematically negative, a fact that we attribute to the finite initial thickness of the film expected from Frankel’s law [3, 18, 21]. If we note hih_i the initial thickness, we expect hi∝(κ/−t0)mh_i \propto\left(\kappa/-t_0\right)^m from Eq. (7). In experiments, we found typical values around a few microns, which is consistent with the order of magnitude expected from Frankel’s law given a pulling velocity around 1 cm s−11\ \mathrm{cm\,s^{-1}}.

For a given thickness hh in the descent approach, drainage can be characterized either by a typical descent time T(h)T(h) scaling as κh−n\kappa h^{-n} or equivalently by a typical descent speed H/T(h)H/T(h) scaling as Hhn/κHh^n/\kappa. We found that nn varies slightly between experiments but we could not identify any correlation with the control parameters. The variations are small with a mean value nˉ=0.70±0.14\bar{n}=0.70\pm0.14. The equivalence in the thinning approach is a thickness evolving as (κ/(t−t0))m\left(\kappa/(t-t_0)\right)^m for a given position, with m=1/nm=1/n, in agreement with our data and empirical laws proposed by other studies.

Our analysis suggests that the functions ff and Sz∗S_z^* are universal within the parameter range explored in this study and in the data taken from the literature. While these functions remain empirical at this stage, this universality highlights that the mechanisms at play are the same, and only the time or speed scales differ between different setups or experimental conditions. However, certain conditions must be met for this self-similar regime to be observed. First, marginal regeneration along the lateral borders needs to be the dominant mechanism, and the scaling time for a given thickness, T(h)T(h), has to be shorter than the typical time scale τP\tau_P expected for the Poiseuille flow that would characterize rigid films. For the latter, this time scale is the characteristic time for a thickness hh to travel through the entire height and scales as τP∼(ηH)/(h2ρg)\tau_P\sim(\eta H)/(h^2\rho g) [26, 34]. The power −2-2 in thickness is different from the power −nˉ=−0.70±0.14-\bar{n}=-0.70\pm0.14 measured here when marginal regeneration is active. Quantitatively, taking H=25 mmH=25\ \mathrm{mm}, W=16 mmW=16\ \mathrm{mm}, η=1.5 mPa s\eta=1.5\ \mathrm{mPa\,s} would lead to a time scale around 4000 s4000\ \mathrm{s} for the 1 μm1\ \mathrm{\mu m}-thick fringe to be advected, in comparison to the 10 s10\ \mathrm{s} measured in the present study. In some situations, marginal regeneration seems to be prevented when thickness is not homogeneous inside the soap films, as highlighted in the case of giant soap films where the dynamics are then dominated by evaporation [35, 36]. Second, even if marginal regeneration is present, edge effects at both the top and bottom should be minimized to recover the functions found in the present study. On the top, the extent of the black film should remain small, which is not always the case with certain surfactants or over extended drainage periods [37]. The region affected by thin film elements generated at the bottom edge should also be of limited spatial extent. These considerations support our choice of fixing z∗=0.3Hz^*=0.3H as a suitable compromise, as this value is small enough to provide a good signal-to-noise ratio but not so small that it avoids the region perturbed by the rising thin film elements.

Conclusion

Our study provides a general framework to compare studies with each other and gain insights into how marginal regeneration drives the drainage of vertical soap films. The universality of the dimensionless functions suggests a robust characterization of the dynamics through the measurements of κ\kappa and nn (or mm), and investigating how these parameters depend on the control parameters. Given that the exponent does not vary much in experiments, fixing nn (or mm) to its mean value would allow for a quantitative comparison between experiments through the parameter κ\kappa. Another step forward would be to relate the occurrence of the self-similar regime to interfacial properties and other experimental conditions in order to probe the transition between the mobile and rigid soap film limits.

Acknowledgments

This work was supported by the National Research Agency (ANR-20-CE30-0019). The authors are grateful to Emma Simon for her participation at the early stage of the project, and to Isabelle Cantat and Emmanuelle Rio for fruitful discussions.

Self-similar coefficient CC

Figure 7 presents the self-similar coefficient CC as a function of the bulk viscosity η\eta (a) and the frame width WW (b). It shows that, in the range explored in this study, the self-similar coefficient CC is independent of η\eta and WW. This supports the idea that self-similarity holds and ff is universal, consistent with the fact that CC depends only on HH in Eq. (4). The two panels of the figure have been obtained for H=25H = 25 mm, and show that in average C≈11C \approx11 mm. This value is represented by a black horizontal dashed line in Figure 7 (a) and (Figure 7b).

Figure 7. Self-similar coefficient CC as a function of (a) bulk viscosity η\eta and (b) frame width WW, with frame height H=25H = 25 mm. The average value for CC on both panels is around C≈11C \approx11 mm, and is represented by a black horizontal dashed line.

Rescaling the thickness profiles from literature data

We present here the rescaling of experimental data from the literature. For each experiment found in the literature, we first display the original thickness profiles obtained at different times. Then, we set z⋆=0.3 Hz^\star= 0.3\,H and plot the rescaled profiles h(z,t)/h(z⋆,t)h(z,t)/h(z^\star,t) on one hand, and we verify the power law between h(z⋆,t)h(z^\star,t) and tt on the other hand (Eq. (7)). Since the time t0t_0 cannot be obtained for the literature data, the power law (7) is verified by plotting in lin-lin scales the thickness h(z⋆,t)h(z^\star,t) to the power −1/mˉ-1/\bar{m} as a function of time tt, where mˉ=1.41\bar{m} = 1.41 is the mean value for the exponent mm obtained in our study.

Figure 8 is obtained from Hudales et al. [20]. Figure 9 and Figure 10 are obtained from Yu et al. [33]: each figure corresponds to a specific surfactant solution, either comprising several SDS concentrations without FC1157, or the same SDS concentration but different FC1157 concentrations.

Figure 8. Data adapted from figure 1 of Hudales et al. [20]. (a) Thickness profiles at different times, coded in colors. (b) Rescaled profiles obtained after normalizing by h(z⋆,t)h(z^\star,t), where z⋆=0.3 Hz^\star= 0.3\,H. (c) h−1/mˉ(z⋆,t)h^{-1/\bar{m}}(z^\star,t) as a function of tt to validate the power law (7) between these two quantities.

Figure 9. Data adapted from figure 7 of Yu et al. [33], where the FC1157 concentration is varied. The left column shows thickness profiles at different times, coded in colors. The middle column represents the rescaled profiles obtained after normalizing by h(z⋆,t)h(z^{\star},t), where z⋆=0.3 Hz^{\star}=0.3\,H. The right column contains the plot of h−1/mˉ(z⋆,t)h^{-1/\bar{m}}(z^{\star},t) as a function of tt to validate the power law (7) between these two quantities. The FC1157 concentration of the solution is 0.25 CMC for the first line, 1 CMC for the second line and 4 CMC for the third line, while the SDS concentration is fixed to 1 CMC.

Figure 10. Data adapted from figure 6 of Yu et al. [33], where the SDS concentration is varied. The left column shows thickness profiles at different times, coded in colors. The middle column represents the rescaled profiles obtained after normalizing by h(z⋆,t)h(z^{\star},t), where z⋆=0.3Hz^{\star}=0.3H. The right column contains the plot of h−1/mˉ(z⋆,t)h^{-1/\bar{m}}(z^{\star},t) as a function of tt to validate the power law (7) between these two quantities. The SDS concentration of the solution is 0.25 CMC for the first line, 1 CMC for the second line, 6 CMC for the third line, and 10 CMC for the last line.

References

References

  1. [1]J. Plateau, Statique expérimentale et théorique des liquides soumis aux seules forces moléculaires (Gauthier-Villars, Paris, France, 1873).
  2. [2]Nierop, E. a V., Scheid, B., & Stone, H. A. (2008). On the thickness of soap films: an alternative to Frankel’s law. Journal of Fluid Mechanics, 602, 119.
  3. [3]L. Saulnier, F. Restagno, J. Delacotte, D. Langevin, and E. Rio, What Is the Mechanism of Soap Film Entrainment?, Langmuir 27, 13406 (2011).DOI
  4. [4]Y.-J. Chen and P. H. Steen, Dynamics of inviscid capillary breakup: collapse and pinchoff of a film bridge, Journal of Fluid Mechanics 341, 245 (1997).DOI
  5. [5]L. Courbin and H. A. Stone, Impact, puncturing, and the self-healing of soap films, Physics of Fluids 18, 091105 (2006).DOI
  6. [6]T. Gilet and J. W. M. Bush, The fluid trampoline: droplets bouncing on a soap film, Journal of Fluid Mechanics 625, 167 (2009).DOI
  7. [7]R. E. Goldstein, H. K. Moffatt, A. I. Pesci, and R. L. Ricca, Soap-film Möbius strip changes topology with a twist singularity, PNAS 107, 21979 (2010).DOI
  8. [8]G. Kirstetter, C. Raufaste, and F. Celestini, Jet impact on a soap film, Phys. Rev. E 86, 036303 (2012).DOI
  9. [9]R. E. Goldstein, J. McTavish, H. K. Moffatt, and A. I. Pesci, Boundary singularities produced by the motion of soap films, PNAS 111, 8339 (2014).DOI
  10. [10]L. Salkin, A. Schmit, P. Panizza, and L. Courbin, Generating Soap Bubbles by Blowing on Soap Films, Phys. Rev. Lett. 116, 077801 (2016).DOI
  11. [11]B. B. Stogin, L. Gockowski, H. Feldstein, H. Claure, J. Wang, and T.-S. Wong, Free-standing liquid membranes as unusual particle separators, Science Advances 4, eaat3276 (2018).DOI
  12. [12]R. E. Goldstein, A. I. Pesci, C. Raufaste, and J. D. Shemilt, Geometry of catenoidal soap film collapse induced by boundary deformation, Phys. Rev. E 104, 035105 (2021).DOI
  13. [13]P. Ballet and F. Graner, Giant soap curtains for public presentations, Eur. J. Phys. 27, 951 (2006).DOI
  14. [14]S. Mariot, M. Pasquet, V. Klein, F. Restagno, and E. Rio, A new setup for giant soap films characterization, The European Physical Journal E 44, 1 (2021).DOI
  15. [15]I. Cantat, S. Cohen-Addad, F. Elias, F. Graner, R. Höhler, O. Pitois, F. Rouyer, and A. Saint-Jalmes, Foams: structure and dynamics (OUP Oxford, 2013).DOI
  16. [16]P. Stevenson, Foam Engineering, in Foam Engineering (John Wiley & Sons, Ltd, 2012).
  17. [17]G. Gochev, D. Platikanov, and R. Miller, Chronicles of foam films, Advances in colloid and interface science 233, 115 (2016).DOI
  18. [18]K. J. Mysels, S. Frankel, and K. Shinoda, Soap films: studies of their thinning and a bibliography (Pergamon press, 1959).
  19. [19]C. Trégouët and I. Cantat, Instability of the one-dimensional thickness profile at the edge of a horizontal foam film and its plateau border, Physical Review Fluids 6, 114005 (2021).DOI
  20. [20]J. Hudales and H. Stein, Marginal regeneration of a mobile vertical free liquid film, Journal of colloid and interface science 138, 354 (1990).DOI
  21. [21]S. Berg, E. A. Adelizzi, and S. M. Troian, Experimental study of entrainment and drainage flows in microscale soap films, Langmuir 21, 3867 (2005).pubmed.ncbi.nlm.nih.gov/15835948
  22. [22]J. Seiwert, R. Kervil, S. Nou, and I. Cantat, Velocity field in a vertical foam film, Physical Review Letters 118, 048001 (2017).DOI
  23. [23]A. Aradian, E. Raphael, and P.-G. De Gennes, “marginal pinching” in soap films, EPL (Europhysics Letters) 55, 834 (2001).DOI
  24. [24]H. Lhuissier and E. Villermaux, Bursting bubble aerosols, Journal of Fluid Mechanics 696, 5 (2012).DOI
  25. [25]A. Gros, A. Bussonnière, S. Nath, and I. Cantat, Marginal regeneration in a horizontal film: Instability growth law in the nonlinear regime, Phys. Rev. Fluids 6, 024004 (2021).DOI
  26. [26]F. Elias, J.-C. Bacri, C. Flament, E. Janiaud, D. Talbot, W. Drenckhan, S. Hutzler, and D. Weaire, Magnetic soap films and magnetic soap foams, Colloids and Surfaces A: Physicochemical and Engineering Aspects A collection of papers presented at the 5th European Conference on Foams, Emulsions, and Applications, EUFOAM 2004, University of Marne-la-Vallee, Champs sur Marne (France), 5-8 July, 2004, 263, 65 (2005).DOI
  27. [27]S. N. Tan, Y. Yang, and R. G. Horn, Thinning of a vertical free-draining aqueous film incorporating colloidal particles, Langmuir 26, 63 (2010).DOI
  28. [28]H. Khan, M. J. Seddon, R. V. Law, N. J. Brooks, E. Robles, J. T. Cabral, and O. Ces, Effect of glycerol with sodium chloride on the krafft point of sodium dodecyl sulfate using surface tension, Journal of colloid and interface science 538, 75 (2019).DOI
  29. [29]Atkins, L. J., & Elliott, R. C. (2010). Investigating thin film interference with a digital camera. American Journal of Physics, 78, 1248.
  30. [30]L. Champougny, J. Miguet, R. Henaff, F. Restagno, F. Boulogne, and E. Rio, Influence of evaporation on soap film rupture, Langmuir 34, 3221 (2018).DOI
  31. [31]See Supplemental Material at [URL will be inserted by the publisher] for movies.
  32. [32]T. Rosi, L. M. Gratton, P. Onorato, and S. Oss, Light interference from a soap film: a revisited quasi-monochromatic experiment, Physics Education 54, 018018 (2019).DOI
  33. [33]X. Yu, K. Qiu, X. Yu, Q. Li, R. Zong, and S. Lu, Stability and thinning behaviour of aqueous foam films containing fluorocarbon and hydrocarbon surfactant mixtures, Journal of Molecular Liquids 359, 119225 (2022).DOI
  34. [34]L. W. Schwartz and R. V. Roy, Modeling Draining Flow in Mobile and Immobile Soap Films, Journal of Colloid and Interface Science 218, 309 (1999).pubmed.ncbi.nlm.nih.gov/10489306
  35. [35]M. Pasquet, F. Restagno, I. Cantat, and E. Rio, Thickness profiles of giant soap films, Phys. Rev. Fluids 8, 034001 (2023).arxiv.org/abs/2209.03468
  36. [36]M. Pasquet, F. Boulogne, F. Restagno, and E. Rio, Lifetime of vertical giant soap films: role of the relative humidity and film dimensions, Soft Matter 20, 2374 (2024).DOI
  37. [37]S. Sett, S. Sinha-Ray, and A. Yarin, Gravitational drainage of foam films, Langmuir 29, 4934 (2013).DOI

Paper details

vlnc:W_b237ms8d
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

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