Lecture 9 Notes: Surface Tension

18.354J Nonlinear Dynamics II: Continuum Systems 9 Lecture 9 Surface Tension Moving on, we shall now use our knowledge of variational principles to investigate the shape of water droplets and soap films. The surface S of a droplet has energy Z E[S] = γ dS (204) with the coefficient of proportionality γ being the surface tension. The questions we now want to answer are: What are the equilibrium shapes predicted by this energy? How can we incorporate constraints, such as fixed volume, into the problem? Is the solution stable? First we will look at the problem in one dimension, and then extend our analysis to many dimensions. 9.1 Two-dimensional bubble Consider a two-dimensional bubble, corresponding to an area B confined by a non-intersecting closed path ∂B in the plane. Let’s assume we can describe parameterize path by y(x). The surface energy is Z Z p E[y] = γ ds = γ dx 1 + y 02 . (205) ∂B ∂B We have already solved the problem to determine the curve that connects two points while satisfying δE/δy = 0, the solution being a straight line. Now we need to introduce the constraint of a fixed volume, and this is done by adding a Lagrange multiplier. Thus we consider the functional Z Z p 02 E[y] = γ dx 1 + y − λ dx y (206) ∂B B the second integral being the volume of the bubble. We combine this into Z h p i E[y] = dx γ 1 + y 02 − λy Inserting the integrand into the Euler-Lagrange equations gives !0 y0 γ p = λ. 1 + y 02 (207) (208) The term on the left hand side is called the mean curvature of the surface, and the solution is readily shown to be a circle (as we expect). You can test this yourself since you know the equation of a circle is just x2 + y 2 = r2 . 9.2 Soap film between two hoops Now lets be a little more ambitious and consider a cylindrically symmetric two dimensional surface, confined by two parallel rings located at x = −a and x = a. Given the axi-symmetry of the problem, we may express the surface energy is Z Z E=γ dA = γ 2πr ds. (209) A A √ The arclength ds is simply 1 + r02 dx, and we see that through symmetry we have reduced this to a one dimensional problem. The problem to be solved is now Z a p ! E[r] = 2πγ dx r 1 + r02 = min. (210) −a Putting the integrand into the Euler-Lagrange equation gives   p d rr0 02 √ 1+r − =0 dx 1 + r02 (211) which can be simplified to rr00 − r02 = 1 (212) We can see immediately that a particular solution of this equation is given by the catenoid 8 r = a cosh(x/a), (213) which corresponds to the special case R/a = cosh(1). For other ratios, we need to solve (212) numerically. 9.3 Rayleigh-Plateau Instability In the previous section, we considered the shape of a soap film, stretched between two hoops. We know, however, that the film breaks after a certain extension, but it is not clear that there is anything in our problem to account for this. What is wrong? To answer this question we must analyze the stability of the solution we have found. This is a topic that we will address in more detail later on in the course, but shall touch upon now. In one-dimensional calculus we check if our extremum is a maximum or a minimum, and if it is the former it is unstable. We now adopt the same approach in variational calculus, and consider the instability of a thin cylinder of fluid. This was studied by Plateau in the 1870’s, who was interested in the formation of droplets from a jet. In cylindrical coordinates y(x) = r0 for the perfect cylinder. The cylinder has surface energy and we look for an extremum by considering the functional Z Z p E[r] = γ dx 2πr 1 + r02 − λ dx πr2 , (214) where the second part is just the volume constraint. We shall perturb the shape and show that if the wavelength of disturbance is greater than 2πr0 , the energy of the system decreases. Using the analysis from above, we know that an extremum is required to satisfy the Euler-Lagrange equations   1 r r00 γ √ − = λr. (215) 1 + r02 (1 + r02 )3/2 8 You can find more soap film solutions at http://www.susqu.edu/brakke/evolver/evolver.html. There, you will also download a surface evolver code to investigate minimal surfaces in more detail. For the perfect cylinder y(x) = r0 is a solution with λ = γ/r0 . Note that λ has dimensions of force/area, and is in fact the pressure. We now consider a perturbation such that r(x) = r0 +  cos kx. (216) This perturbation is in some sense arbitrary, because any perturbation can be decomposed into sines and cosines. Thus  Z  p π 2 2 2 2 E[r] = γ 2π(r0 +  cos kx) 1 +  k sin kx − (r0 +  cos kx) dx. (217) r0 Assuming that  is small, we can expand the square root, yielding    Z  1 2 2 2 π 2 E[r] = γ 2π(r0 +  cos kx) 1 +  k sin kx − (r0 +  cos kx) dx. r0 2 (218) Expanding everything out, we find that terms linear in  cancel out, and we are left with  Z Z  cos2 kx 2 2 2 E[r] = γ πr0 dx +  γπ r0 k sin kx − dx. (219) r0 The first term corresponds to the energy E0 of the undeformed problem, so we are really interested in the sign of the second integral ∆E. If it is positive then the energy is increased as a result of the perturbation, and the system will be stable. However, if the integral is negative, then by perturbing the system the energy is decreased, and the system must be unstable. Since the disturbance is periodic, we integrate over one wavelength, 2π/k, to determine the energy per wavelength. We know that Z 2π Z 2π π dx sin2 kx = dx cos2 kx = , (220) k 0 0 so that the energy change, ∆E, is   1 ∆E =  γπ r0 k − . r0 k 2 2 (221) So now we see that if r0 k > 1 the system is stable, because ∆E is positive. However, if r0 k < 1 then ∆E is negative and the system is unstable. Since k = 2π/λ this can be rearranged to say that the system is unstable to wavelengths λ greater 2πr0 , the circumference of the√cylinder. This was Plateau’s results, which Rayleigh found to be √ wrong by a factor of 2, due to hydrodynamic effects within the cylinder (i.e., λ = 2 2πr0 is the correct answer). The physical reasoning arises from the fact that by perturbing the cylinder you may be increasing the length of the surface along the cylinder, but you are reducing the cross sectional area. It is the trade off between these two that determines the stability. So does this answer our question of why the soap film breaks? Perhaps. You see, if the distance between the two hoops is less than the 2π/r0 , then the system is not open to long wavelength perturbations, and is therefore stable. As soon as the separation exceeds the critical wavelength, instability sets in and the bubble breaks. Note, however, that a critical part of our stability analysis relied on the constraint of constant volume to determine instability. With the soap film this is not present, so perhaps an alternative mechanism is responsible (e.g., possibly the shape of the soap film between the two hoops becomes so eccentric that surfaces in the middle touch, and there is a pinch off?). What would happen if the ends of the two hoops were closed, so that the volume constraint did then apply? Our stability analysis would now seem more relevant. The only way to check is to now go do some experiments (a good course project!).