Arxiv:2105.12365V2 [Physics.Flu-Dyn] 3 Aug 2021

Flow proﬁles near receding three–phase contact lines: Inﬂuence of surfactants

Benedikt B. Straub,1 Henrik Schmidt,1 Peyman Rostami,1, 2 Franziska Henrich,1 Massimiliano Rossi,3, 4 Christian J. Kähler,4 Hans–JürgenButt,1 and Günter K. Auernhammer1, 2 1Max Planck Institute for Polymer Research, Ackermannweg 10, D–55128 Mainz, Germany 2Leibniz-Institut fürPolymerforschung, Hohe Straße 6, D–01069 Dresden, Germany 3Department of Physics, Technical University of Denmark, DTU Physics Building 309, DK–2800 Kongens Lyngby, Denmark 4Institute of Fluid Mechanics and Aerodynamics, Bundeswehr University Munich, D–85577 Neubiberg, Germany (Dated: August 5, 2021) The dynamics of wetting and dewetting is largely determined by the velocity field near the contact lines. For water drops it has been observed that adding surfactant decreases the dynamic receding contact angle even at a concentration much lower than the critical micelle concentration (CMC). To better understand why surfactants have such a drastic effect on drop dynamics, we constructed a dedicated a setup on an inverted microscope, in which an aqueous drop is held stationary while the transparent substrate is moved horizontally. Using astigmatism particle tracking velocimetry, we track the 3D displacement of the tracer particles in the flow. We study how surfactants alter the flow dynamics near the receding contact line of a moving drop for capillary numbers in the order of 6 10− . Even for surfactant concentrations c far below the critical micelle concentration (c CMC) Marangoni stresses change the flow drastically. We discuss our results first in a 2D model that considers advective and diffusive surfactant transport and deduce estimates of the magnitude and scaling of the Marangoni stress from this. Modeling and experiment agree that a tiny gradient in 1 surface tension of a few µN m− is enough to alter the flow profile significantly. The variation of the Marangoni stress with the distance from the contact line suggests that the 2D advection-diffusion model has to be extended to a full 3D model. The effect is ubiquitous, since surfactant is present in many technical and natural dewetting processes either deliberately or as contamination.

I. INTRODUCTION contact line diverges [4]. To circumvent this kind of un- physical singularities various ways of introducing cutoff lengths or slip boundary conditions on the solid substrate The wetting and dewetting of a solid surface by a liq- are discussed [4–8]. In typical substrates the length scales uid are fundamental processes of many natural phenom- where these effects become relevant are nanoscopic, i.e., ena and technical applications. Examples include drops inaccessible with optical measurement techniques [9]. gliding down a window glass or being blown across a windscreen as well as drop and bubble motion in coating, As soon as the strong assumption of simple liquids painting, printing and flotation. Drops on a substrate are (only one component, Newtonian viscosity, constant sur- surrounded by a contact line, i. e., a line where substrate, face tension, etc.) is left, the modeling is much more liquid and gas meet. The Young relation predicts a spe- complex, if existing [10]. In quasi-static situations sur- cific contact angle θ as a function of the surface tensions factants change the surface tensions and consequently at the liquid-gas, liquid-substrate and substrate-gas in- the contact angles of sitting drops. Under constant flow terface [1]. For all practical purposes, the contact angle conditions, the presence of surfactants can change the at non-moving contact lines is limited between the static boundary conditions on free liquid-gas interfaces. This advancing and receding contact angle due to imperfec- is essentially due to the the fact that surface dilation or tions of the substrate. This contact angle hysteresis give contraction due to hydrodynamic flow induces gradients rise to a lateral adhesion force that has to be overcome in the surfactant surface concentration and consequently to make a drop [2]. When the contact line is moving, the the surface tension. These gradients in surface tension, or contact angle is velocity dependent. For simple liquids, Marangoni stresses, counteract the delating or contract- arXiv:2105.12365v2 [physics.flu-dyn] 3 Aug 2021 the dynamics of wetting, i.e., the velocity dependent ad- ing surface flows, i.e., modify the stress-free boundary vancing and receding contact angle, are well described condition at free surfaces. This is nicely illustrates in by hydrodynamic modeling and molecular kinetic theory rising bubbles in surfactant solutions, which form stag- [3–8]. Essentially, the hydrodynamic energy dissipation nant caps even at very low concentrations [11–13], i.e., is concentrated near the moving contact line, because of the boundary condition becomes effectively non-moving. the apparently diverging shear rate at the moving con- These stagnant caps change the rise velocity and flow tact line. The connection between the dissipation at the field in and outside the bubble. A similar process is ob- contact line and the macroscopic wetting dynamics near served in the coalescence dynamics of drops [14–17] with contact lines has been described using the Navier–Stokes liquid pools. In evaporating drops, the natural convec- equation. As Huh and Scriven pointed out, without fur- tion due to thermal Marangoni effects [18] is modified ther approximations, the dissipation near the receding for drops containing more than one component [19], e.g., 2 additional surfactants [20, 21], particles [22], or salt [23]. experimental data. Our results show that (i) minute con- Landau-Levich films form above a critical speed, when centrations can influence the dynamics surfactant laden plates are being pulled out of a liquid pool. The prop- drops significantly and (ii) the dynamics of surfactant erties of these films and the hydrodynamic flow field in laden drops is a non-local process in which, most proba- the pool depend on the surfactant concentration in the bly, the flow in the entire drop has to be considered. liquid [24–26]. When a continuous motion of the contact line is enforced, i.e., when at speeds below the critical speed for film formation, different scenarios have been II. RELEVANT MECHANISMS observed. The presence of insoluble surfactants can in- duce an unsteady motion of the contact line [27] and A. 2D advection-diffusion model structured deposits on the substrate [28]. The situation can change qualitatively for soluble sur- 1. General idea factants. Studies with a rotating drum setup show that even small amounts of soluble surfactants strongly re- In a simple model that takes into account various duce the dynamic receding contact angle [29–32]. By previous studies [14, 15, 19, 26, 29–32, 36–40], a two– adding surfactant at a concentration between 5–30% of dimensional model can be used as a working hypothe- the critical micelle concentration (CMC), the critical ve- sis (Figure 1): At the receding contact line new liquid- locity for film formation reduces by one order of magni- gas interface is created and thus the liquid surface is tude. This strong decrease in the receding contact angle expanded. This expanded surface is less covered with was attributed to surface tension gradients (Marangoni surfactant than the liquid surface in equilibrium at the effect) near the receding contact line, where continuously given surfactant concentration. Surfactants reach this in- a new surface was assumed to be created. terface in two ways: (1) Surfactants, that were adsorbed Measurements of the three–dimensional flow pattern at the solid-liquid interface, can be directly transferred of surfactant solutions near receding contact lines would to the new liquid surface at the contact line. Other- shed light on the true mechanism behind this strong influ- wise, they remain on the solid interface. (2) Surfactants ence of small surfactant concentrations on the dynamic diffuse from the bulk solution towards the new surface, receding contact angle. For pure water, such measure- supported by advection due to the bulk flow. The equi- ments had been carried out by Kim et al. using a To- librium surfactant concentration in the vicinity of the mographic PIV setup [33, 34]. Qian et al. measured the solid is transported with the bulk flow towards the con- liquid velocities near the substrate in the vicinity of the tact line. This equilibration process leads to a gradient contact lines using particle image velocimetry and flood in surface tension and thus a Marangoni stress along the illumination [35]. Since surfactants change the hydrody- interface. The Marangoni stress is directed towards the namic boundary condition at the liquid-gas interface of three–phase contact line and counteracts the flow along liquids [14, 15, 36], surface velocimetry has been carried the liquid-gas interface. Higher surfactant concentrations out. Studies using plates withdrawn from a liquid pool should lead to higher Marangoni stress. The surface ac- with surfactant solutions reveal a strong impact of the tivity of the surfactant influences the Marangoni effect. contact line velocity on the surface flow [37, 38]. A recir- Based on previous results by some of us [29, 31, 40], we culation flow on the surface was observed in dynamic wet- proposed that the Marangoni effect scales with the rela- ting experiments of drops with insoluble surfactants [39]. tive concentration, as measured in percent of the CMC. Yet, the internal dynamics can only be approximated in- directly with these methods. Other studies measured the internal flow during dewetting macroscopically [26, 31]. 2. Typical diffusion distance Due to resolution limits, these studies were not able to resolve the area near the contact line and liquid-gas in- The wedge geometry has no intrinsic length scale. So terface. the characteristic length scales of the system have to be In this work we compare quantitatively measured flow deduced by the from the sample properties. profile near receding contact lines in surfactant solutions To derive a characteristic diffusion distance, we com- to modeling considerations of 2D models. In the fol- pare the surface excess of the surfactants to their bulk lowing section, we describe a advection-diffusion model concentration [41]. Using a linear dependence of the for receding contact lines and explore some consequences equilibrium surface excess Γ on the bulk concentration thereof on the surface-tension gradients. The advection- c [29, 40] diffusion model gives important predictions on the magnitude and scaling of the surface-tension gradients. We Γ = αc. (1) then report measurements of the flow profile and deduce the surface tension gradient close to receding contact line The parameter α has the dimension of a length. We note from these measurements. The comparison of the mod- that Equation 1 is only valid for small concentrations. eling part to the experiments leads us to the conclusion In equilibrium, the length α is the thickness of a liquid that a purely 2D approach is not compatible with the layer underneath the liquid-gas interface that contains 3

Bulk flow Marangoni stress Diffusion

Transfer Advection Stick Contact line (CL) Moving substrate

FIG. 1. Simple 2D model to explain the Marangoni stress in the case of a receding contact line for aqueous surfactant FIG. 2. Sketch of the area close to the moving three–phase solutions. A new liquid-gas interface is created at the contact contact line. line. Hence, the liquid surface is expanded. This expanded interface is initially less covered with surfactants (small dots with tail) than the surface in equilibrium. Below the dashed line, advection transports surfactant towards the contact line. with the diffusion constant of the surfactant D. Surfactants reach the expanded surface either by advection To estimate the advected distance during this diffusion and diffusion away from the contact line or by direct transfer time, we use the velocity of the liquid-gas interface U from the solid-liquid interface to the liquid-gas interface at int the contact line. Part of the surfactants remain absorbed on in the surfactant-free case as derived by Moffatt [3] the solid interface after the contact line passed. This leads to sin θ θ cos θ a concentration gradient, i.e., a Marangoni stress along the U = U − . (5) int θ sin θ cos θ liquid-gas interface, which opposes the surface flow. − Consequently, the advected distance is given by as many surfactant molecules as the liquid-gas interface, x = t U . (6) Figure 2. The length α can be calculated by using the adv D int Gibbs adsorption isotherm As the estimates given in Table I indicate, different sur- c dγ factants can have huge differences in these characteristic Γ = . (2) −RT dc length scales. We will check these dependences experimentally in the later sections of this work. Here, R is the ideal gas constant and T is the absolute temperature. Figure 2 defines the orientation of the length α. By integrating the Gibbs adsorption isotherm B. Gradients along the liquid-gas interface in a 2D between c = 0 and CMC of the surfactant, one gets the flow length α γ γ A first hint that this simple 2D advection-diffusion α = 0 − . (3) 2RT c model might have to be extended came from a set of experiments in the rotating drum setup [30]. The absolute Here, γ0 is the surface tension of pure water 1 magnitude of the changes in the dynamic contact angle (72.4 mN m− ) and γ is the equilibrium surface tension of the surfactant solution at CMC. Since the liquid depends strongly on the possible transport mechanism of layer with the thickness α contains as many surfactant the surfactant between the receding and the advancing molecules as the surface, α is a good measure of a typi- contact line. For example, blocking the transport along cal diffusion length in the simple 2D advection-diffusion the liquid-gas interface between the advancing and the re- model above. For surfactants that give a similar surface ceding contact line changes significantly the magnitude tension γ at very different absolute concentrations, i.e. of the change in the dynamic contact angles [30]. for surfactants with strongly different CMC, the length Further limitation of this simple model becomes ap- scale α can vary significantly. parent, when considering the stress balance at the free surface close to the receding contact line. The surface tension gradient γ due to the adsorption of the sur- ∇k 3. Typical advected distance factants on the freshly formed liquid-gas interface has to be balanced by the viscous stress of the hydrodynamic flow τ = µ ~v , with v the flow component parallel to In the 2D advection-diffusion system described above, p p the liquid-gas∇⊥ interface µ the viscosity of the liquid and the transport across α is purely diffusive. This allows us and the gradient parallel and perpendicular to to estimate first a diffusion time t . D the∇k liquid-gas∇⊥ interface α2 t = , (4) D τ = µ ~vp = γ. (7) 2D ∇⊥ ∇k 4

TABLE I. Critical micelle concentration (CMC), surface tension at CMC γ, diffusion coefficient D at 25 °C [31], the characteristic diffusion (α) and advection distances (xadv) of two typical non-ionic surfactants. −1 2 −1 Surfactant Formula CMC [mM] γ [mN m ] D [m s ] α [µm] xadv [µm] −10 Octyl triglycole C8E3 7.5 27.3 4.6 10 2.5 1.2 × −10 Dodecyl pentaglycole C12E5 0.07 30.7 2.9 10 244 6900 ×

uρL To get an estimate of the upper bound of the gradients Re = (11) in surface tension, we assume that a gradient is strong µ eneough to stop the surface flow at the liquid-gas interface completely. Such situation is solved analytically by µu Ca = (12) Taylor in what is typically called a scraping flow [42, 43]. γ The tangental stress along the liquid-gas interface needed to stop the flow is given by [44] In these definitions ρ is the mass density, L a characteristic length scale, and u a characteristic velocity. 2µU sin θ θ cos θ Additionally, one further dimensionless number could µ ~v = − , (8) p 2 2 ∇⊥ xint θ sin θ play a role: the Damköhlernumber [19]. This is given by − the ratio between the diffusion time to the interface and where θ is the contact angle in radiants and xint the the adsorption time at the interface. However, previous distance to the contact line (Figure 2). Integrating this studies have shown that the overall effect on the dynamic from infinity to the a finite distance from the contact receding contact angle is independent on the charge and line gives an upper bound for the expected differences in molar mass of the surfactant [31, 32]. These studies con- the surface tension needed to completely stop the surface cluded that the rate limiting step is not the adsorption to flow away from the contact line the liquid-gas interface but the transport to the interface. We follow this result and assume in this work that the sin θ θ cos θ ∆γ = 2µU ln xint − . (9) adsorption process of the surfactant from the subsurface − θ2 sin2 θ − to the surface is fast compared transport process to the Importantly, in this 2D case the change in surface ten- (sub)surface. sion has only a weak logarithmic dependence on the dis- The typical number given in Table II show important tance from the contact line. To get an estimate of the characteristics of the flow of surfactant solutions close to receding contact lines. The flow is laminar (Re 1) and expected changes, we enter typical values for the exper- 1 dominated by capillary effects (La 1 and Ca 1). imental parameters into Equation 9 (U = 200 µm s− , These relations stay valid, if the length α ist taken as a µ = 1 mPa s, θ = π/4 , and xint = 50 µm), see also the experimental part below. With these numbers, we get characteristic length. 1 a surface tension difference in the order of 5 µN m− . ≈ This extremely small value is essentially due to the low TABLE II. Estimates of typical dimensionless numbers: viscosity of water. Laplace number La, Reynolds number Re, and Capillary number Ca, with a characteristic length L of 50 µm, the viscosity of water µ = 1 mPa s, the density of water ρ = 3 1 C. Dimensionless numbers 1000 kg/m , and a characteristic velocity of 200 µm s− . Surface tension La Re Ca 1 6 Besides the estimates of characteristic length, time and 72 mN m− 3600 0.01 2.8 10− × surface tension scales as given above, we use dimension- 1 6 40 mN m− 2000 0.01 5.0 10− less numbers to choose a good set of experimental set- × tings. In the present case, several dimensionless numbers are connected to the problem. The Laplace number La measures the ratio of the capillary forces to the iner- D. Questions to the experiments tial forces in a fluid flow. The Laplace number is also the ratio of the Reynolds Re and the capillary number Ca. Both numbers give dimensionless velocities. The The above considerations give a couple of non-trivial Reynolds number relates the fluid inertia to the viscous predictions on the characteristics of the flow field in sur- dissipation. It is often used to characterize the crossover factant solutions close to receding contact lines that can from laminar to turbulent flow. The capillary number be checked experimentally: (i) Close to the receding con- give the ratio of viscous forces to surface tension forces: tact line, a region with a slowed down or stagnated surface flow should be observed. (ii) The size of this region γρL Re should depend strongly on the type of the surfactant, es- La = = (10) µ2 Ca pecially, its values of the characteristic length α. (iii) The 5 absolute value of the change in surface tension needed to the measurement area was at least a factor of 30 larger produce these effects is very small, in the range of several than the measurement area. Therefore, we assumed that 1 µN m− . (iv) The surface tension should vary slowly with the measured flow fields are independent of any influence a logarithmic dependence on the distance to the contact of the prism. In this work, we considered receding con- 1 line. tact lines with a constant velocity of 200 µm s− , i.e., the To test these modeling observations experimentally, we drop was pulled away from the prism (Figure 3). A stable varied two important parameters: (i) the characteristic steady–flow condition was achieved 5 s after starting the length α by choosing two surfactants with different CMC substrate motion and could be held for more than 60 s. (or the absolute surfactant concentration at compara- ble surface tensions) and (ii) by changing the concentration of these surfactant to test for a possible dependence C. Astigmatism particle tracking velocimetry on the Laplace number (or the concentration relative to CMC). Three–dimensional particle trajectories were measured by astigmatism particle tracking velocimetry (APTV). APTV is a single–camera method that uses a deliberate III. MATERIAL AND METHODS astigmatic aberration of the optical system, obtained by introducing a cylindrical lens in front of the camera sen- A. Sample preparation sor [46–48]. Figure 4 shows a sketch of the used APTV setup. The working principle of APTV is shown in Fig- We study solutions of nonionic surfactants, to avoid ure 4 (b). The optical configuration of an APTV sys- charge effects. The surfactants dodecyl pentaglycole tem has not one, but two principal focal planes (Fxz and (C12E5) and octyl triglycole (C8E3), purchased from Fyz). The reason for that is the cylindrical lens in front of Sigma–Aldrich, were used without further purification. the camera chip, which introduces a controlled astigmatic The properties of the surfactants are listed in Table I. aberration. The particle image deformation depends on To track the flow motion we use fluorescent polystyrene the relative distance of the particle to the focal axes. In particles (diameter 2 µm or 4 µm) dispersed in the the example in Figure 4 (b), particle 1 is on the Fxz plane aqueous surfactant solutions. The particles were ei- and therefore it is focused in x–direction (x–diameter ther synthesized in house (Rhodamine B: Ex./Em. small) and defocused in y–direction (y–diameter large). 550 nm/600 nm) or bought from microParticles GmbH On the other hand, particle 3 is on the Fyz plane and (PS–FluoRed: Ex./Em. 530 nm/607 nm). To avoid sedi- therefore it is focused in y–direction and defocused in x– mentation of the particles a 1:1 mixture of water (H2O, direction. If the x– and y–diameters of particle images Arium®611 or Arium®pro VF/UF & DI/UV ultra– at different axial positions are plotted together, they col- pure water systems, Sartorius, resistivity of 18.2 MΩ) lapse on a single curve as a function of the axial particle and deuterated water (D2O purchased from Sigma– position (Figure 4 (c)). This curve can be obtained by Aldrich, purity 99 %) was used. Previous experiments a calibration procedure performed on reference particle have already demonstrated that these particles can faith- images at known axial positions. The axial position of a fully follow the internal and interfacial flow of droplets measured particle image is then determined by compar- [21, 45]. Precision cover glasses (No. 1.5H, 24 60 mm2) ing its x and y diameters with the calibration curve. In × were mechanically cleaned with acetone, isopropanol, our analysis we follow the APTV procedure as described ethanol using cleanroom wipes. Afterwards, the cover in [47, 48]. glasses were hydrophobized with trichloro(1H,1H,2H,2H– For the APTV measurements, we used a 40 X micro- perfluorooctyl)silane (Sigma–Aldrich) via the gas phase. scope objective in combination with a 150 mm cylindri- The cover glasses were placed side by side with two 10 µL cal lens. Images were acquired with a rate of 50 Hz. The drops of the silane in a desiccator for 2 min. A mag- maximum measurement volume of the setups was around netic ventilator ensured a homogeneous distribution of 400 450 60 µm3 with uncertainty in the particle po- × × the silane in the gas phase. sition determination of less than 0.25 µm in the lateral ± direction and 1 µm in the axial direction. Detailed information about± the setups and data processing can be B. Experimental setup found in the Supplemental Material [49]. The velocity fields were obtained from an average of 3 To image the moving contact line for a long time and experiments for each fluid solution. In each experiment, with high resolution, the drop was pinned to a prism an average of 168 tracer particles were tracked and the while the substrate was moved smoothly back and forth respective positions and velocities measured with APTV. (Figure 3 (a)) by a piezoelectric motor (LPS–45, Physik The position and inclination of the substrate was deter- Instrumente, Germany). We placed this setup above mined experimentally from the trajectories of particles the objective lens of an inverted fluorescence microscope. stuck on the substrate or trapped at the contact line. Thus, the moving contact line was stationary as seen by All data are shown in a coordinate system in which the the microscope. The distance between the prism and plane at z = 0 corresponds to the substrate. The posi- 6

Prism

Fluid

Moving substrate

Objective

FIG. 3. Sketch of the experimental setup. The shown flow field for pure water was measured close to the receding three–phase contact line and is shown in the co–moving frame of the contact line. The measured particle velocities within a volume around each grid point were averaged. The liquid close to the substrate has the highest velocity, directed parallel to the moving substrate and decreasing with increasing distance from the substrate. When getting closer to the contact line the flow is 1 directed upwards. Contact line velocity: 200 µm s− ; receding contact angle: 78° 3°. ± (a) (b) Sample 1 Fxz 1

Objec ve 2 Field z lens 2 3 Filter F cube Camera yz y 3 Cylindrical lens x

Illumina¢on

Objec¡ ve

(c)

1 z

FIG. 4. (a) Sketch of an APTV setup with a cylindrical lens in front of the camera chip. (b) Working principle of APTV. The image of a fluorescent labelled particle changes depending on its distance to the objective lens, i.e., the z position. (c) The scatter plot of x– and y–extensions of particle images at different z positions collapses on a single curve that is dependent on the axial variable z. This curve is obtained by a calibration procedure on reference particle images. The calibration curves are used to calculate the z position of measured particle images. tion of the contact line was also identified in each image a volume around each grid point were averaged with an and used to correct the stream–wise coordinate x, so that an average number of 13 particle velocities for each grid the final x–coordinate measures the distance of particles point. from the contact line (with x = 0 corresponding to the contact line). Finally, all data points were projected on the xz–plane. The measured particle velocities within 7

IV. MEASURED CHARACTERISTIC OF THE For both surfactants, the region near the liquid-gas in- FLOW FIELD terface has a reduced velocity; the deviation field points towards the contact line (Figure 5 (h), (i)). The surface A. Velocity and deviation fields flow is slower than in pure water. The surface velocity is reduced in the entire observation volume near the free To compare experimental flow profiles (e.g. Figure 5 surface. As will be shown in the next section, this reduc- (a–c)) to hydrodynamic calculations, we assume a simpli- tion in surface velocity results from a Marangoni stress fied geometry near the contact line: Rather than treating caused by gradients in surfactant concentration, which the full three–dimensional dewetting problem, we take opposes the surface flow. a two-dimensional wedge geometry, because we do not If we consider the shown fluid volume of Figure 5 (h) observe significant transverse flow within the resolution as a control volume, we can apply the conservation of of our measurements. For this geometry, hydrodynamic mass. Fluid enters the volume at x = 100 µm from the flow is known for simple liquids, like water [3–6]. These right and leaves the volume at z = 60 µm parallel to the theories provide an analytical solution for the region very liquid-gas interface. Since the outflow of the fluid near near the three–phase contact line (inner solution) and the the free surface is reduced due to the Marangoni stress, bulk flow (outer solution). Due to the size of the tracer the inflow of the fluid into the volume is also reduced particles, we could not resolve experimentally the flow in to fulfil mass conservation; the deviation field points to the inner region. Therefore, we compare our measure- the right at x = 100 µm. The same general qualitative ments only to the outer theoretical solution, which is the behavior is observed for surfactants with very difference same for all these theories. We use the theory of Moffatt α, showing this does not depend on the characteristic [3] for stress–free liquid-gas interface and constant con- length scale α but is a generic feature. tact line velocity. Figure 5 (g) depicts the deviation field between measurement (a) and theory (d) for pure water. This deviation field does not show a flow field but the B. Deduced changes in surface tension normalized velocity difference between measurement and theory as calculated by: To quantify the gradient in surface tension and surface excess, i.e., the surface concentration of surfactant ud = um ut, wd = wm wt, (13) molecules, along the free surface, we use the already men- − − tioned Equation (7). We first calculate the shear stress from the measured velocity profiles. Since any tangental 2 2 ~vd ud + wd stress occurring at a liquid-gas interface originates from a Sd = | | = . (14) ~v 2 2 surface tension gradient, we calculate the surface tension t put + wt | | gradient along the free surface γ from the shear stress ∇k Here, u and w are the mean localp velocity components in τ = µ ~vp in the flow profile just under the free surface ∇⊥ x and z direction at a given (x, z) position. The indices [50], Equation (7): τ = µ ~vp = γ. The velocity ~vp ∇⊥ ∇k indicate deviation d, measurement m and theory t. The is the velocity component parallel to the free surface. relative strength of the deviation Sd is represented by the To calculate the velocity derivative, we fit the veloc- color code and the length of the arrows, the angular de- ity in the direction normal to the free surface by a cubic viation by the orientation of the arrows. The deviation polynomial. We differentiate the polynomial function to field of water (Figure 5 (g)) has a randomly distributed calculate the shear stress τ. We use the viscosity of water angular deviation and its magnitude is <30 %. Due to µ = 1 mPa s to calculate the stress for the different solu- this randomness and the small absolute deviation, mea- tions. Due to the scattering of the experimental data the surement and theory agree very well. The random noise uncertainty of the calculated stress is around 20 %. Since shows that no bias errors are present. At the liquid-gas the statistical error becomes too large closer than 20 µm interface, little deviation occurs. As expected, the liquid- to the contact line, we exclude this region from further gas interface is stress–free. analysis. The surface tension gradient γ is integrated In the presence of surfactants (30 %CMC), the receding from right to left to obtain the change∇ ink surface tension 1 contact angle at a velocity of 200 µm s− decreased from ∆γ (Figure 6). 78° (pure water) to 54° (C12E5, 0.38 ppm) and 25° (C8E3, For all measured concentrations, a clear surface ten- 40 ppm). Because of this change in contact angle, it is sion gradient was measurable. Far away from the contact impossible to directly compare the measured flow fields of line (>70 µm) the surface–tension gradient vanishes. At surfactant solutions to those measured in pure water. No this distance the gradient in surface concentration is too hydrodynamic theory is available to describe the flow of small to produce measurable effects, i.e., we consider the surfactant solutions near moving contact lines. As shown surface tension to be close to constant. When approach- above, the water measurements match the hydrodynamic ing the contact line, the surface tension deviates from its theory for water. So, we can compare the surfactant equilibrium value. For both surfactants the magnitude measurements with the hydrodynamic theory for pure in surface tension difference ∆γ increases with increasing water [3] at the respective contact angle. surfactant concentration, Figure 6. Although the change 8

(a) (b) (c)

(d) (e) (f)

(g) (h) (i)

FIG. 5. Direct comparison between the measured velocity fields (a–c), the analytical solution of Moffatt [3] (d–f) and the corresponding deviation fields (g–i). Each column represents one measurement solution and the contact line velocity is always 1 200 µm s− . The first column is the measured flow field of pure water (a), (d) is the corresponding analytical solution and (g) is the calculated deviation field between (a) and (d). The second column shows the measured flow field for a 30 %CMC C12E5 solution (b) and the calculated deviation field (h). The third column shows the results for a 30 %CMC C8E3 solution. The comparison between the measured flow fields and the analytical solution shows that in the case of surfactant solutions, the flow at the interface between liquid and air is reduced. This is shown clearly by the deviation fields (h, i). in surface tension for the 5 %CMC solutions is small, it is sponds to a decreasing surface concentration, more pre- significantly larger than our resolution limit that is given cisely surface excess, of surfactant near the contact line. by the apparent surface tension gradient for water, see Supplemental Material [49]. We quantify this range of the increased surface tension by a decay length LD over V. COMPARING EXPERIMENTAL DATA TO which ∆γ reduces to a fraction of 1/e of its initial value. THE MODELING IDEAS The decay lengths of the surface tension gradient differ by less than a factor of two between the used surfactants A. Gradients in surface tension and concentrations, Table III. Additionally, there is no clear tendency of how the decay length depends on the characteristic length α, i.e., on the type of the surfac- The measured flow fields and absolute values match to tant, or on the Laplace number La, i.e., on the balance some extend the expected values. As discussed in Section of the surface forces to the viscous forces in the liquid. II B, the magnitude of the change in surface tension is in 1 So neither molecular nor local dynamic properties seem the order of µN m− . This is actually expected due to to play a dominating role in defining the hydrodynamic the close relation between the magnitude of this gradient flow close to the receding contact line in these surfactant and the viscosity of the liquid. For the same reason, solutions. the surface excess near the contact line is only around 0.01 % lower than the equilibrium surface excess. These Although the surface tension near the contact line small changes of surface excess strongly influence the flow 1 is only increased by 1–2 µN m− , the gradient in sur- profile and the apparent receding contact angle (Figure 5) face tension, i.e. the Marangoni stress, is around at very small absolute concentrations of the surfactants 2 70–130 µN m− . This increase in surface tension corre- at 30 %CMC of only 0.38 ppm and 40 ppm for C12E5 and 9

TABLE III. Solution parameters: relative concentration in %CMC, absolute concentration in the bulk c in parts per million, equilibrium surface tension γ, equilibrium surface excess Γ, surface area per molecule Amol, length α and decay length LD for diﬀerent surfactant concentrations. 1 2 2 Name [%CMC] c [ppm] γ [mN m− ] Γ [mol m− ] Amol [nm ] α [µm] LD [µm] 6 C12E5 5 0.06 58.2 0.9 10− 1.9 244 17.5 3.9 × ± 6 C12E5 15 0.2 47.5 2.6 10− 0.7 244 12.4 1.6 × ± 6 C12E5 30 0.4 40.2 5.1 10− 0.3 244 15.8 2.1 × ± 6 C8E3 5 6.8 53.2 0.9 10− 1.8 2.5 17.4 4.8 × ± 6 C8E3 15 20.3 43.4 2.8 10− 0.6 2.5 15.9 5.5 × ± 6 C8E3 30 40.7 35.4 5.6 10− 0.3 2.5 23.3 6.0 × ±

lengths LD of the surface tension gradients are almost the same (Table III). (ii) This measured decay length matches none of the expected advected lengths xadv. For C8E3 the decay length is much larger than the advected length x L , for C E the inequality is inverted adv D 12 5 xadv LD. For a precise comparison one would have to take the actual surface velocities in calculating the advected length. However, this does not resolve the discrepancies because changing the surface velocity by maximum a factor of three over a few tens of micrometers does not change the advected distance to the amount needed to identify it with the decay length LD. These discrepancies clearly show the limitations of the 2D modeling considerations in Section II. Especially, neither the properties summarized in the characteristic FIG. 6. (a) and (b) show the calculated change in surface length α (molecular properties of the surfactant) nor the tension ∆γ for different concentrations of C12E5 and C8E3 Laplace number La (characteristic of the dynamics close along the free surface xInt. See Fig. 2 for the axis direction. to the receding contact line) seem to have major effect. The data was fitted with an exponential function and the From this we conclude that one of the essential ap- dotted lines are the corresponding prediction bounds. proximation of the simple model has to be dropped. The advection-diffusion approach is a simple consequence of mass conservation and has a strong basis. In contrast C8E3, respectively. This observation indicates that even the 2D approximations has a much weaker basis and has minor contamination or small amounts of additives may already be shown to have some limitations [30]. In the strongly influence the flow near contact lines. following we explorer possibilities beyond this 2D approx- Our experimental results give clear data on the flow imation. field close to the contact line and consequently on the gradients in surface tension and surface excess. We can, however, draw no conclusions on the amount of surfac- B. Beyond the 2D approximation tant transferred from the solid-liquid interface to the liquid-gas interface directly at the moving contact line, As shown in a previous study [30], the surface trans- because we lack data in the last 20 µm. Our data port of the surfactant has an important influence on the ≈ shows that the surface is also expanded in our observa- dynamic receding contact angles. When taking the entire tion region. The surface velocity is increasing with in- drop into account, also a fully 3D contribution becomes creasing distance from the contact line over a distance apparent. Close to the receding side of the drop, the gen- that is roughly given by the decay length LD, Figure 5. eration rate for new liquid-gas interface is highest and the Since we lack data very close to the receding contact line, liquid surface is measurably expanded, Figure 5. At the we are unable to quantify the true surface generation di- side of the drop, between the advancing and receding end rectly at the receding contact line. of the drop, there is a region in which the surface flow in There are some substantial differences of the exper- the drop is parallel to the contact line, as sketched in Fig- imental data to the modeling expectations. (i) The ure 7. The further the liquid-gas interface is away from decay in surface tension is much faster than the loga- the receding contact line, or the ”older” this surface is, rithmic decay predicted by the simple 2D model. (ii) the closer the surface concentration is to the equilibrium For all concentrations and both surfactants, the decay surface concentration. Thus, at the sides of the drop, 10 where no fresh liquid-gas interface is generated, the sur- the receding contact line the concentration of surfactant face concentration of surfactant is higher than close to is reduced leading to a Marangoni stress, i.e., gradient in the receding contact line. Correspondingly, there is a surface tension. surface tension gradient, a large scale Marangoni stress, Two system parameters are important in a 2D model parallel to the contact line that can drive surface trans- considering advective and diffusive transport of surfac- port of surfactant from the drop sides to the receding tant to the free surface. (i) The ratio between surface end of the drop. This Marangoni stress can modify the excess and volume concentration of the surfactant give over all flow profile in the drop, leading to an additional a characteristic diffusion length α. (ii) The dimension- transport of surfactant to the region close to the receding less Laplace number La gives the ratio between capillary contact line. In this 3D model, the spatial distribution forces and inertial forces. The characteristic numbers α of surface excess of surfactant is also determined by the and La are varied by using different surfactants and dif- entire flow profile in the drop and not by the local pro- ferent concentrations. The 2D advection-diffusion model cesses close to the receding contact line, i.e., thus a fully gives some strong predictions on the characteristics of the 3D flow allows to resolve the discrepancies described in flow field, namely the order of magnitude of the surface the previous section. tension gradient and its dependence on the distance to the contact line. The flow velocity and viscosity of the aqueous drops studied only leads to a very small gradient in surface tension and surface excess in the observed volume. The experimentally measured flow fields allow to check these predictions. The magnitude of the measured surface tension gradient fits nicely in the predicted range 1 of a few µN m− . In contrast, the spatial dependence does not match the 2D advection-diffusion model, but decays much faster and does not depend on α and La. To resolve this, we propose a 3D flow field that is beyond the 2D approximation initially assumed. This conclusion is supported by the fact that the range of the surface tension gradients seems to be independent of the type of the surfactant and its concentration, i.e., of α and La. FIG. 7. Sketch of the three–dimensional Marangoni stress The dynamics of surfactant laden drops shows a strong around the drop’s surface (purple arrows). The black arrows coupling of the internal hydrodynamic flow and the sur- in the drop indicate the surface flow. At the sides of the drop, between the areas of the advancing and receding contact factant dynamics at the liquid-gas interface. This cou- lines, no liquid-gas interface is generated and thus the surface pling can limit the applicability of 2D approximations sig- tension is closer to equilibrium and higher than close to the nificantly. Our results imply that even small pollutions receding contact line. We measure that the surface tension or additives may play a major role in dynamic dewetting close the receding side is reduced (Figure 6). Hence, a surface processes. tension gradient along the drop’s surface exists that can drive surface transport of surfactant from the sides of the drop to the receding end of the drop. Note the different length scales. ACKNOWLEDGMENTS The width of the drop is several mm and the region with lower surface tension close to the receding contact line has an Funded by the Deutsche Forschungsgemeinschaft extension in flow direction of several tens of µm. (DFG, German Research Foundation) — Project–ID 265191195 — SFB 1194, A02 (P. R.), A06 (B. B. S., G. K. A.) and C03 (H.J. B.). H. S. acknowledges financial support from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) through Project VI. SUMMARY AND CONCLUSION No. AU321/3 within the Priority Program 1681. C. J. K. and M. R. acknowledge financial support through Project We have analyzed the flow field in the vicinity of the No. KA1808/22–1. The research leading to these re- receding contact line of a water drop with and without sults has received funding from the European Research surfactant. The measured flow profiles support the idea Council under the European Union’s Seventh Frame- that receding contact angles decrease even at low surfac- work Programme (FP7/2007-2013)/ERC grant agree- tant concentration because of a Marangoni effect. Near ment n◦883632 (H.J. B., B. B. S.).

[1] T. Young, An essay on the cohesion of ﬂuids, Philosoph- (1805). ical Transactions of the Royal Society of London 95, 65 11

[2] N. Gao, F. Geyer, D. W. Pilat, S. Wooh, D. Vollmer, ticle/surfactant mixtures: the importance of particle– H.-J. Butt, and R. Berger, How drops start sliding over interface interactions, Langmuir 31, 4113 (2015). solid surfaces, Nature Physics 14, 191 (2017). [21] A. Marin, R. Liepelt, M. Rossi, and C. J. Kähler, [3] H. K. Moffatt, Viscous and resistive eddies near a sharp Surfactant-driven flow transitions in evaporating corner, Journal of Fluid Mechanics 18, 1 (1964). droplets, Soft Matter 12, 1593 (2016). [4] C. Huh and L. E. Scriven, Hydrodynamic model of steady [22] S. Manukyan, H. M. Sauer, I. V. Roisman, K. A. Bald- movement of a solid/liquid/fluid contact line, Journal of win, D. J. Fairhurst, H. Liang, J. Venzmer, and C. Tro- Colloid and Interface Science 35, 85 (1971). pea, Imaging internal flows in a drying sessile polymer [5] O. V. Voinov, Hydrodynamic of wetting, Journal of Flu- dispersion drop using spectral radar optical coherence to- idmechanics 11, 714 (1976). mography (sr-oct), Journal of Colloid and Interface Sci- [6] R. G. Cox, The dynamics of the spreading of liquids on ence 395, 287 (2013). a solid surface. part 1. viscous flow, Journal of Fluid Me- [23] A. Marin, S. Karpitschka, D. Noguera-Mar´ın, M. A. chanics 168, 169 (1986). Cabrerizo-V´ılchez, M. Rossi, C. J. Kähler,and M. A. [7] D. Bonn, J. Eggers, J. Indekeu, J. Meunier, and E. Rolley, Rodr´ıguez Valverde, Solutal marangoni flow as the cause Wetting and spreading, Reviews of Modern Physics 81, of ring stains from drying salty colloidal drops, Phys. 739 (2009). Rev. Fluids 4, 041601 (2019). [8] J. H. Snoeijer and B. Andreotti, Moving contact lines: [24] C.-W. Park, Effects of insoluble surfactants on dip coat- Scales, regimes, and dynamical transitions, Annual Re- ing, Journal of Colloid and Interface Science 146, 382 view of Fluid Mechanics 45, 269 (2013). (1991). [9] E. Bonaccurso, M. Kappl, and H.-J. Butt, Hydrody- [25] R. Krechetnikov and G. M. Homsy, Surfactant effects namic force measurements: Boundary slip of water on in the landau–levich problem, J. Fluid Mech. 559, 429 hydrophilic surfaces and electrokinetic effects, Physical (2006). Review Letters 88, 076103 (2002). [26] H. C. Mayer and R. Krechetnikov, Landau-levich flow [10] G. Lu, X. D. Wang, and Y. Y. Duan, A critical review visualization: Revealing the flow topology responsible of dynamic wetting by complex fluids: From newtonian for the film thickening phenomena, Physics of Fluids 24, fluids to non-newtonian fluids and nanofluids, Advances 052103 (2012). in Colloid and Interface Science 236, 43 (2016). [27] K. S. Varanasi and S. Garoff, Unsteady motion of re- [11] R. Davis and A. Acrivos, The influence of surfactants on ceding contact lines of surfactant solutions: the role of the creeping motion of bubbles, Chemical Engineering surfactant re-self-assembly, Langmuir 21, 9932 (2005). Science 21, 681 (1966). [28] B. K. Beppler, K. S. Varanasi, S. Garoff, G. Evmenenko, [12] S. S. Sadhal and R. E. Johnson, Stokes flow past bub- and K. Woods, Influence of fluid flow on the deposition bles and drops partially coated with thin films. part 1. of soluble surfactants through receding contact lines of stagnant cap of surfactant film –exact solution, Journal volatile solvents, Langmuir 24, 6705 (2008). of Fluid Mechanics, 126, 237 (1983). [29] D. Fell, G. Auernhammer, E. Bonaccurso, C. J. Liu, [13] S. S. Dukhin, V. I. Kovalchuk, G. G. Gochev, M. Lotfi, R. Sokuler, and H.-J. Butt, Influence of surfactant con- M. Krzan, K. Malysa, and R. Miller, Dynamics of rear centration and background salt on forced dynamic wet- stagnant cap formation at the surface of spherical bub- ting and dewetting, Langmuir 27, 2112 (2011). bles rising in surfactant solutions at large reynolds num- [30] D. Fell, N. Pawanrat, E. Bonaccurso, H.-J. Butt, and bers under conditions of small marangoni number and G. K. Auernhammer, Influence of surfactant transport slow sorption kinetics, Adv Colloid Interface Sci 222, 260 suppression on dynamic contact angle hysteresis, Colloid (2015). and Polymer Science 291, 361 (2013). [14] Y. Amarouchene, G. Cristobal, and H. Kellay, Noncoa- [31] F. Henrich, D. Fell, D. Truszkowska, M. Weirich, M. Any- lescing drops, Phys. Rev. Lett. 87, 206104 (2001). fantakis, T. H. Nguyen, M. Wagner, G. K. Auernham- [15] O. Manor, I. U. Vakarelski, X. S. Tang, S. J. O’Shea, mer, and H.-J. Butt, Influence of surfactants in forced G. W. Stevens, F. Grieser, R. R. Dagastine, and D. Y. C. dynamic dewetting, Soft Matter 12, 7782 (2016). Chan, Hydrodynamic boundary conditions and dynamic [32] D. Truszkowska, F. Henrich, J. Schultze, K. Koynov, forces between bubbles and surfaces, Phys. Rev. Lett. H. J. Rader, H.-J. Butt, and G. K. Auernhammer, Forced 101, 024501 (2008). dewetting dynamics of high molecular weight surfactant [16] S. V. Iasella, N. Sun, X. Zhang, T. E. Corcoran, S. Garoff, solutions, Colloids and Surfaces A: Physicochemical and T. M. Przybycien, and R. D. Tilton, Flow regime transi- Engineering Aspects 521, 30 (2017). tions and effects on solute transport in surfactant-driven [33] H. Kim, S. Große, G. E. Elsinga, and J. Westerweel, Full marangoni flows, Journal of Colloid and Interface Science 3d–3c velocity measurement inside a liquid immersion 553, 136 (2019). droplet, Experiments in Fluids 51, 395 (2011). [17] M. L. Sauleda, H. C. W. Chu, R. D. Tilton, and S. Garoff, [34] H. Kim, C. Poelma, G. Ooms, and J. Westerweel, Exper- Surfactant driven marangoni spreading in the presence of imental and theoretical study of dewetting corner flow, predeposited insoluble surfactant monolayers, Langmuir, Journal of Fluid Mechanics 762, 393 (2015). Langmuir 37, 3309 (2021). [35] B. Qian, J. Park, and K. S. Breuer, Large apparent slip [18] F. Giorgiutti-Dauphinéand L. Pauchard, Drying drops, at a moving contact line, Physics of Fluids 27, 091703 The European Physical Journal E 41, 32 (2018). (2015). [19] H. Manikantan and T. M. Squires, Surfactant dynamics: [36] D. Y. C. Chan, E. Klaseboer, and R. Manica, Dynamic hidden variables controlling fluid flows, Journal of Fluid deformations and forces in soft matter, Soft Matter 5, Mechanics 892, P1 (2020). 2858 (2009). [20] M. Anyfantakis, Z. Geng, M. Morel, S. Rudiuk, and [37] B. B. Luokkala, S. Garoff, R. D. Tilton, and R. M. Suter, D. Baigl, Modulation of the coffee-ring effect in par- Interfacial structure and rearrangement of nonionic sur- 12

factants near a moving contact line, Langmuir 17, 5917 checked: 11 June 2021. (2001). [45] M. Rossi, A. Marin, and C. J. Kähler,Interfacial flows [38] B. K. Beppler, K. S. Varanasi, S. Garoff, G. Evmenenko, in sessile evaporating droplets of mineral water, Physical and K. Woods, Influence of fluid flow on the deposition Review E 100, 033103 (2019). of soluble surfactants through receding contact lines of [46] H. P. Kao and A. S. Verkman, Tracking of single fluo- volatile solvents, Langmuir 24, 6705 (2008). rescent particles in three dimensions: use of cylindrical [39] D. L. Leiske, C. Monteux, M. Senchyna, H. A. Ketel- optics to encode particle position, Biophysical Journal son, and G. G. Fuller, Influence of surface rheology on 67, 1291 (1994). dynamic wetting of droplets coated with insoluble sur- [47] C. Cierpka, M. Rossi, R. Segura, and C. J. Kähler,On the factants, Soft Matter 7, 7747 (2011). calibration of astigmatism particle tracking velocimetry [40] F. Henrich, D. Linke, H. M. Sauer, E. Dörsam,S. Hardt, for microflows, Measurement Science and Technology 22, H.-J. Butt, and G. K. Auernhammer, Forced dynamic 015401 (2011). dewetting of structured surfaces: Influence of surfactants, [48] M. Rossi and C. J. Kähler,Optimization of astigmatic Phys. Rev. Fluids 4, 124202 (2019). particle tracking velocimeters, Experiments in Fluids 55, [41] J. K. Ferri and K. J. Stebe, Which surfactants reduce 1809 (2014). surface tension faster? a scaling argument for diffusion- [49] See Supplemental Material at [URL will be inserted by controlled adsorption, Advances in Colloid and Interface publisher] for a detailed description of the astigmatism Science 85, 61 (2000). particle tracking velocimetry setups, data processing and [42] G. I. Taylor, Similarity solutions of hydrodynamic prob- additional data. lems, Aeronautics and Astronautics 4, 214 (1960). [50] L. D. Landau and E. M. Lifshitz, Fluid Mechanics, 2nd [43] G. K. Batchelor, An Introduction to fluid dynamics ed., Course of Theoretical Physics, Vol. 6 (Butterworth- (Cambridge Univ Press, 2000). Heinemann, 1987). [44] https://en.wikipedia.org/wiki/Taylor scraping flow, Last Flow profiles near receding three–phase contact lines: Influence of surfactants Supplemental Material

Benedikt B. Straub,1 Henrik Schmidt,1 Peyman Rostami,1, 2 Franziska Henrich,1 Massimiliano Rossi,3, 4 Christian J. Kähler,4 Hans–JürgenButt,1 and Günter K. Auernhammer1, 2

1Max Planck Institute for Polymer Research, Ackermannweg 10, D–55128 Mainz, Germany 2Leibniz-Institut f¨urPolymerforschung, Hohe Straße 6, D-01069 Dresden, Germany 3Department of Physics, Technical University of Denmark, DTU Physics Building 309, DK–2800 Kongens Lyngby, Denmark 4Institute of Fluid Mechanics and Aerodynamics, Bundeswehr University Munich, D–85577 Neubiberg, Germany (Dated: August 5, 2021) arXiv:2105.12365v2 [physics.flu-dyn] 3 Aug 2021

CONTENTS

I. Astigmatism Particle Tracking Velocimetry ii

II. Velocity and deviation ﬁelds iv

References v

i I. ASTIGMATISM PARTICLE TRACKING VELOCIMETRY

We performed the experiments in two facilities, setup 1 at the Max Planck Institute for Polymer Research in Mainz, and setup 2 at the Institute of Fluid Mechanics and Aerodynam- ics at the Bundeswehr University Munich. We obtained similar results in both experiments.

The ﬁnal plots for pure water, C12E5 and 5 %CMC C8E3 were obtained from setup 1 and the remanining C8E3 plots from setup 2.

FIG. S1. (a) Sketch of the experimental setups. (b, c) Examples of astigmatic particle images obtained with setup 1 and 2, respectively. The contact line of the drop is highlighted in red or green.

The setup 1 consisted of a Leica DMI 6000B inverted microscope in combination with a Photron Fastcam 1.1. Images were recorded with a recording speed of 50 Hz which provided good spatial and temporal resolution. The optical elements were a 150 µm achromatic cylindrical lens (Thorlabs) directly in front of the camera sensor and a 40X microscope objective (LUCPLFLN, Olympus, Japan). Illumination was provided by a mercury lamp in combination with an appropriate ﬁlter cube to adjust the used wavelength. As tracer particles, red–ﬂuorescent polystyrene spheres with a diameter of 2 µm were used (microParticles

GmbH, PS–FluoRed). Small polystyrene beads are slightly denser than water (ρPS = 1050 kg/m3). Therefore, to avoid sedimentation o he particles a 1:1 mixture of water and deuterated water was. Furthermore, the sedimentation time of the particles (even in pure water) is much larger than the time scales of the phenomena investigated, therefore it can safely be assumed that they follow faithfully the fluid flow [S1]. Figure S1 (a) shows a schematic sketch of the setup. The maximum measurement volume of this setup configuration is around 400 450 60 µm. × × The detection of the particle position and shape is made in two steps. The first step is a

ii pre–detection of the particles using a binarized image. In the second step the pre–detection information is used as the start parameter for an iterative two–dimensional Gaussian fitting procedure to the full intensity profile of the particle image. See [S2] for the used Gaussian fit function. The fit parameters of the Gaussian function are the extensions in both direction and middle point of the tracer particles. To minimize errors the fitting is done iteratively with varying segmentation sizes. To calibrate the system several particles are placed on top of a glass cover slip. A drop of the measurement fluid is placed above the particles without disturbing them. The focus of the microscope objective is moved through the probe with known step sizes. The result is a calibration curve like shown in FIG. 3 (c) of the main article. The fit function for the principal axes with the dependent axial variable z can be found in [S2]. The particle coordinates are combined to particle trajectories by using the “track” al- gorithm of Crocker and Grier [S3]. Due to the cylindrical lens the image is also globally deformed. Therefore, the lateral particle coordinates have to be transformed back to an undeformed image. We took pictures of a standard calibration grid (PS20, Pyser Optics, Kent, UK) before and after the introduction of the cylindrical lens into the beam path. We used the ”bUnwarpJ” [S4, S5] Plugin of ImageJ [S6–S8] to generate a transformation matrix. We used this matrix to transform the lateral coordinates of the particles back into the undeformed system. The error estimation was carried out using test data. These test data were measured the same way as the calibration curves. The axial position of the particles is therefore known. The calibration is used on this data and the deviation between the known axial position and the calculated axial position with the calibration is used as the axial error. The axial standard deviation is 0.98 µm and the estimated lateral uncertainty is 0.25 µm. ± ±

The setup 2 was similar to setup 1 but used different components. Specifically, it consisted of an inverted microscope Axio Observer Z1 (Carl Zeiss AG) in combination with an Imager sCMOS camera (LaVision GmbH), with images also recorded at 50 Hz. The optical elements were a 150 µm cylindrical lens place also in front of the camera sensor and a 40 X microscope objective (LD Plan–Neofluar, Carl Zeiss AG). The illumination was provided by a high- power green light–emitting diode (LED). As tracer particles, red–fluorescent (Rhodamine B) polystyrene spheres with a diameter of 4 µm were used (synthesized at the Max Planck

iii Institute for Polymer Research). To obtain the calibration images for setup 2, a similar procedure as the one described for setup 1 was used. Instead, a diﬀerent approach was used to obtain the axial coordinate. Rather than comparing the x and y extensions, the shape of the measured particle images was directly compared with the shape of the reference particle images by means of the normalized cross–correlation function. More details on this calibration approach are given in [S9]. Also in this case, a standard calibration grid (Thorlabs) was used to map the image distortion and get the coordinates in real units. The measurement volume obtained with this conﬁguration was about 380 430 70 µm3, × × with an estimated uncertainty of 0.7 µm in the axial direction, and 0.1 µm in the lateral ± ± direction.

II. VELOCITY AND DEVIATION FIELDS

The deviation field of a 5 %CMC C12E5 also shows the discussed deviation close the the free surface and in the bulk flow (Fig. S2 (a)). Although the change in surface tension is very small, it is still a factor 3 higher than the noise level we measured for pure water. We attribute the small systematic variation of the calculated surface tension difference for water to effect of the finite size of our tracer particles. Note that this systematic error has a different dependence on the distance from the contact line and falls off faster.

FIG. S2. (a) Deviation ﬁeld of 5 %CMC C12E5 solution. (b) Change in surface tension of water,

5 %CMC C12E5 and 5 %CMC C8E3.

iv [S1] M. Rossi, A. Marin, and C. J. Kähler,Interfacial flows in sessile evaporating droplets of mineral water, Physical Review E 100, 033103 (2019). [S2] C. Cierpka, M. Rossi, R. Segura, and C. J. Kähler, On the calibration of astigmatism particle tracking velocimetry for microflows, Measurement Science and Technology 22, 015401 (2011). [S3] J. C. Crocker and D. G. Grier, Methods of digital video microscopy for colloidal studies, Journal of Colloid and Interface Science 179, 298 (1996). [S4] C. O. S. Sorzano, P. Thevenaz, and M. Unser, Elastic registration of biological images using vector-spline regularization, IEEE Transactions on Biomedical Engineering 52, 652 (2005). [S5] I. Arganda-Carreras, C. O. S. Sorzano, R. Marabini, J. M. Carazo, C. Ortiz-de Solorzano, and J. Kybic, Consistent and elastic registration of histological sections using vector-spline regularization, in Computer Vision Approaches to Medical Image Analysis, Vol. 4241, edited by R. R. Beichel and M. Sonka (Springer Berlin Heidelberg, 2006) pp. 85–95. [S6] J. Schindelin, I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J. Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona, Fiji: an open-source platform for biological-image analysis, Nature Methods 9, 676 (2012). [S7] C. A. Schneider, W. S. Rasband, and K. W. Eliceiri, Nih image to imagej: 25 years of image analysis, Nature Methods 9, 671 (2012). [S8] C. T. Rueden, J. Schindelin, M. C. Hiner, B. E. DeZonia, A. E. Walter, E. T. Arena, and K. W. Eliceiri, Imagej2: Imagej for the next generation of scientific image data, BMC Bioinformatics 18, 529 (2017). [S9] R. Barnkob, C. J. Kähler,and M. Rossi, General defocusing particle tracking, Lab on a Chip 15, 3556 (2015).