the alignment process presented here takes place in the time range of a second. A similar mechanism, by which a hard contact lens centers itself over the cornea in a human eye, was discussed by Moriarty and Terrill (1996). The centrifugal spinning of volatile solutions is a convenient and efficient means of coating planar solids with thin films. This process, known as spin coating, has been widely used in many technological processes, such as deposition of dielectric layers onto silicon wafers in the microelectronic industry, for- mation of ultrathin antireflective coatings for deep UV lithography, and others. Two important stages of Physics of Thin Liquid Films 12-5 FIGURE 12.3 Deposition of water onto a patterned surface with hydrophilic microchannels with corners. The width of the channel in the corner region increases from channel (a) to channel (e). Time and therefore the volume of the condensate increase from top to bottom. When a microchannel undergoes a morphological change of its shape, the drop moves to the corner to maximize the contact area with the hydrophilic part of the substrate. (Reprinted with permission from Herminghaus et al. (2000).) © 2006 by Taylor & Francis Group, LLC 12-6 MEMS: Introduction and Fundamentals FIGURE 12.4 Infrared images of various states as seen in the experiments. The temperature increases with increas- ing brightness, so warm depression regions are white (except in (c)) and cool elevated regions are dark. Each image has its own brightness, so temperatures in different images cannot be compared. (a) A localized depression (dry spot) with a helium gas layer and d ϭ 0.025 cm. (b) A localized elevation (high spot) with an air gas layer and d ϭ 0.037 cm. (c) A dry spot with hexagons in the surrounding region and d ϭ 0.025 cm. (d) Hexagons with an air gas layer and d ϭ 0.045 cm. For more detail refer to the source. (Reprinted with permission from VanHook et al. (1997).) t = 0:00 min 10:30 min 13:50 min t = 15:30 min 17:10 min 20:15 min FIGURE 12.5 The evolution of a localized depression and formation of a dry spot in silicone oil of depth d ϭ 0.0267 Ϯ 0.0008 cm and helium in the gas layer. At t ϭ 0 (an arbitrary starting point) there is negligible defor- mation of the interface. The liquid layer begins to form a localized depression (the white circle), and in 15 minutes the interface has ruptured (h min → 0) and formed a dry spot. The dry spot continues to grow for several more min- utes before saturating. Bright (dark) regions are hot (cool) because they are closer (farther) to (from) the heater. All images have the same intensity scaling. (Reprinted with permission from VanHook et al. (1997).) © 2006 by Taylor & Francis Group, LLC the process are usually considered. The first stage occurs shortly after the liquid volume is delivered to the disk surface rotating usually at the speed of 1000–10,000r/min. At the beginning of this stage the liquid film is relatively thick (usually greater than 500 microns). The film thins mainly because of radial drainage under the influence of centrifugal forces. Inertial forces are important and can lead to the appearance of instabilities of the spinning film. The second stage occurs when the film has thinned to the point where inertia is no longer important (film thickness usually less than 100 microns), and the flow slows down considerably, but deformations of the fluid interface may still be present because of the instabilities that appeared during the first stage. The film continues to thin mainly because of solvent evaporation until the Physics of Thin Liquid Films 12-7 FIGURE 12.6 Photographs of C4 bonding based on self-alignment mechanism. (a) Layout of the chip (4 mm by 4 mm) which consists of four solder joints made of 63Sn37Pb. The upper chip is not aligned with the lower one, as can be seen from the position of the upper cross relative to four squares at the lower chip. Initial misalignment is 150 microns. (b) An enlarged picture of one of the solder joints at the initial moment. (c) An intermediate stage. (d) The final position. (e) A side view showing the cross-section of the solder joint at the final stage. (Reprinted with permis- sion from Salalha et al. (2000).) © 2006 by Taylor & Francis Group, LLC solvent becomes depleted, and the film solidifies and ceases to flow. Such problems are discussed in the sections on isothermal films and phase changes. Numerous applications relevant for MEMS involve the dynamics of liquid films or drops. This area is in constant progress and new exciting developments are often reported in the literature. Knight et al. (1998) describes a new method of enhancement and control of nanoscale fluid jets.They demonstrated this method with a design of a continuous-flow mixer capable of mixing flow rates of nanoliters per second within the time scale of 10 microseconds. Such a mixer can be useful in nanofabrication techniques and serve as an essen- tial part of a microreactor built on a chip. Spatially controlled changes in the chemical structure of a solid substrate can guide a deposited liquid along the substrate. Ichimura et al. (2000) reported their experimental results showing the possibility of reversible guidance of liquid motion by light irradiation of a photoresponsive solid substrate. Asymmetric irradiation of the solid surface with blue light led to movement of a 2 microliter olive oil droplet with a typical speed of 35 microns/sec. A similar irradiation with a homogeneous blue light stopped the movement of the droplet completely. The speed of the droplet and the direction of its movement were adjustable to the conditions of such irradiation. The phenomenon described has a potential applicability in design of microreactors and microchips. Schaeffer et al. (2000) proposed a new technique of creating and replicating lateral structures in films on submicron length scales. This technique is based on the fact that lateral gradients of the electric field applied in the vicinity of the film interface induce variations of surface tension and thus lead to the elec- trocapillary effect. The electrocapillary effect is similar to the thermocapillary effect previously mentioned and is addressed more thoroughly in the section on thermal effects. The electrocapillary effect triggering the electrocapillary instability of the film results in formation of ordered patterns on the film interface and focusing of the interfacial troughs and peaks in the desired locations following the master pattern of the electrodes. Schaeffer et al. (2000) reported the replication of patterns of lateral dimensions of order 140 nanometers while employing this technique. A complete investigation of the electrocapillary insta- bility of thin liquid films has not yet appeared in the literature. Lee and Kim (2000) presented a liquid micromotor and liquid–metal droplets rotating along a microchannel loop driven by continuous elec- trowetting (CEW) phenomenon based on the electrocapillary effect. They identified and developed key technologies to design, manufacture, and test the first MEMS devices employing CEW. A mathematical treatment of this and other phenomena must consider that the interface of the film lying or flowing on a solid surface is partially or entirely a free boundary whose configuration evolving both tem- porally and spatially must be determined as an integral part of the solution of the governing equations. This renders the problem too difficult and often almost intractable analytically, which might lead researchers to rely on computing only. Computing also becomes complicated because of the free-boundary character of the problem which requires a careful design of adequate numerical methods. Another property of such mathematical problems is their strong inherent nonlinearity, which is present in both governing equations and boundary conditions. This nonlinearity of the problem presents another complexity. Consideration of coupled phenomena, such as those previously mentioned, requires compact description of simultaneous instabilities that interact in intricate ways. This compact form must be tract- able and, at the same time, complex enough to retain the main features of the problem at hand. The most appropriate analytical method of dealing with the above complexities is to analyze only long scale phenomena, in which the characteristic lateral length scales are much larger than the average film thick- ness, the flow-field and temperature variations along the film are much more gradual than those normal to it, and the time variations are slow. Similar theories arise in a variety of areas of classical physics: shallow-water theory for water waves, lubrication theory in viscous flows, slender-body theory in aerodynamics, and in dynamics of jets [e.g., Yarin, 1993]. In all of these examples, a geometrical disparity is used to practically separate the variables and to simplify the analysis. In thin viscous films, most rupture and instability phe- nomena occur on long scales, and a long-wave approach explained later is very useful. The long-wave theory approach is based on the asymptotic reduction of the governing equations and boundary conditions to a simplified system, which consists often, but not always, of a single nonlinear partial differential equation formulated in terms of the local thickness of the film varying in time and 12-8 MEMS: Introduction and Fundamentals © 2006 by Taylor & Francis Group, LLC space. The rest of the unknowns (i.e., the fluid velocity, pressure, temperature, etc.) are determined via functionals of the solution of this differential equation usually called evolution equation. The notori- ous complexity of a free-boundary problem thus is removed. The corresponding penalty is, however, the presence of the strong nonlinearity in the evolution equation(s) and the higher-order spatial derivatives (usually up to the fourth) appearing there. A simplified linear stability analysis of the problem can be carried out based on the resulting evolution equation. A weakly nonlinear analysis of the problem is also possible through that equation. However, the fully nonlinear analysis that allows one to study finite-amplitude deformations of the film interface must be performed numerically. Numerical solution of the evolution equation is incomparably less difficult than that of the original, free-boundary problem. Several encouraging verifications of the long-wave theory versus the experimental results have appeared in the literature. Burelbach et al. (1990) carried out a series of experiments in an attempt to check the long-wave theory of Tan et al. (1990) for steady thermocapillary flows induced by non-uniform heating of the solid substrate. The measured steady shapes were favorably tested against theoretical predictions for layers less than 1 mm thick under moderate heating conditions. However, the relative error was large for conditions near rupture, where the long-wave theory is formally invalid [Burelbach et al., 1988], but in all other cases the predicted and measured values of the minimal film thickness agreed within 20%. The theory (see Equation (3.6) of [Tan et al., 1990]) also predicts rupture when the parameter L exceeds a certain critical value and predicts steady patterns otherwise. Experimental results (see Figure 1 of [Burelbach et al., 1990]) show that L is an excellent qualitative indicator of whether the film ruptures. VanHook et al. (1995, 1997) performed experiments on the onset of the long-wavelength insta- bility in thin layers of silicone oil of varying thickness, aspect ratios, and transverse temperature gradients across the layer. A formation of “dry spots” at randomly varying locations was found above the critical temperature difference across the layer in qualitative agreement with corresponding numerical simulations. The experimental support for the theoretical results is discussed in various sections of this chapter. Another test for the validity of an asymptotic theory, such as the long-wave theory presented here, is the comparison between the numerical solutions for the full free-boundary problem in its original form and the solutions obtained for the corresponding long-wave evolution equations. Due to the difficulty of carrying out direct numerical simulations previously discussed, the number of such comparative studies is quite limited. Krishnamoorthy et al. (1995) performed a full-scale direct numerical simulation of the governing equations to study the rupture of thin liquid films because of thermocapillarity and found very good qualitative agreement with the results arising from the solution of the corresponding evolution equation, except for times prior to rupture. Oron (2000b) found even better agreement at rupture between his results and the direct simulations of the Navier–Stokes equations of Krishnamoorthy et al. (1995). There has been a long debate in the literature about the validity of fingered structures of the film interface often arising from the solution of the evolution equations and whether they are artifacts of the asymptotic reduction applied. Direct solution of the Navier–Stokes equations [Krishnamoorthy et al., 1995] provides convincing evidence supporting the validity of the evolution equations even in the domain where some assumptions leading to their derivation are violated. The analysis of thin liquid films has progressed significantly in recent years. In the review article by Oron et al. (1997) such analyses were unified into a simple framework in which the special cases naturally emerged. In this chapter the physics of thin liquid films is reviewed with emphasis on the phenomena of considerable interest for MEMS. The theory of drop spreading, despite its importance, is not included here. Refer to other reviews [de Gennes, 1985; Leger and Joanny, 1992; Oron et al., 1997] for more detailed information. The general evolution equation describing the general dynamics of thin liquid films is derived following Oron et al. (1997) and is discussed in the next section. The topic addressed in the second section is isothermal films, where the physical effects discussed are viscous, surface tension, gravity, and centrifugal forces along with van der Waals interactions. The third section examines the influence of thermal effects on the dynamics of liquid films. The fourth section considers the dynamics of liquid films undergoing phase changes, such as evaporation and condensation. Physics of Thin Liquid Films 12-9 © 2006 by Taylor & Francis Group, LLC 12.2 The Evolution Equation for a Liquid Film on a Solid Surface We now describe the long-wave approach and apply it to a flow of a viscous liquid in a film. The film is supported below by a solid horizontal plate and is bounded above by an interface separating the liquid and a passive gas and slowly evolving in space and time, as given by its equation z ϭ h (x, y, t). Assume the possibility of external interfacial forces Π with the components {Π 3 , Π 1 , Π 2 } in the normal and tangential to the film surface directions, respectively, determined by the vectors n ϭ , t 1 ϭ , t 2 ϭ . (12.1) The components of the vectors n, t 1 , t 2 in Equation (12.1) are specified in the order of x-, y-, and z- direc- tions, where x and y are the spatial coordinates in the given solid plane and z is normal to the latter and directed across the film. The presence of a conservative body force determined by the potential φ acting on the liquid phase, such as gravity, centrifugal, or van der Waals force, is accounted for as well. We note that the vectors t 1 , t 2 are not orthogonal, but it is sufficient for our later application that (n, t 1 ) and (n, t 2 ) con- stitute pairs of orthogonal vectors. The letter subscripts denote the partial derivatives with respect to the corresponding variable. The liquid considered in this work is assumed to be a simple Newtonian incompressible viscous fluid whose dynamics are well described by the Navier–Stokes and mass conservation equations, provided that the length scales characteristic for the flow domain are within the continuum range exceeding several molec- ular spacings. The mass conservation and Navier–Stokes equations for such a liquid in three dimensions have the form u x ϩ v y ϩ w z ϭ 0, ρ (u t ϩ uu x ϩ vu y ϩ wu z ) ϭ Ϫp x ϩ µ (u xx ϩ u yy ϩ u zz ) Ϫ φ x , ρ (v t ϩ uv x ϩ vv y ϩ wv z ) ϭ Ϫp y ϩ µ (v xx ϩ v yy ϩ v zz ) Ϫ φ y , (12.2) ρ (w t ϩ uw x ϩ vw y ϩ ww z ) ϭ Ϫp z ϩ µ (w xx ϩ w yy ϩ w zz ) Ϫ φ z , where ρ , µ are, respectively, the density and kinematic viscosity of the liquid; u, v, w are the respective components of the fluid velocity vector v in the directions x, y, z; t is time; and p is pressure. The classical boundary conditions between the liquid and the solid surface supporting it are those of no-penetration w ϭ 0 and no-slip u ϭ 0, v ϭ 0. These conditions are appropriate for the continuous films to be considered. Problems with a contact line, where the liquid on a solid surface spreads or recedes will not be examined in this chapter. The reader interested in this topic is referred to the review papers by de Gennes (1985), Leger and Joanny (1992), and Oron et al. (1997). The boundary conditions at the solid surface are therefore w ϭ 0, u ϭ 0, v ϭ 0 at z ϭ 0. (12.3) At the film surface z ϭ h(x, y, t) the boundary conditions are formulated in the vector form [e.g., Wehausen and Laitone, 1960]: h t ϩ v и ∇ * h ϳ w ϭ 0, (12.4a) T и n ϭ Ϫ2H ~ σ n ϩ ∇ s σ ϩ Π, (12.4b) where T is the stress tensor of the liquid, Π is the prescribed forcing at the interface, H ~ is the mean curvature of the interface determined from 2H ~ ϭ ∇ * и n ϭ Ϫ , (12.5) h xx (1 ϩ h 2 y ) ϩ h yy (1 ϩ h 2 x ) Ϫ 2h x h y h xy ᎏᎏᎏᎏ (1 ϩ h 2 x ϩ h 2 y ) 3/2 {0, 1, h y } ᎏ ͙ 1 ෆ ϩ ෆ h ෆ 2 y ෆ {1, 0, h x } ᎏ ͙ 1 ෆ ϩ ෆ h ෆ 2 x ෆ {Ϫh x , Ϫh y , 1} ᎏᎏ ͙ 1 ෆ ϩ ෆ h ෆ 2 x ෆ ϩ ෆ h ෆ 2 y ෆ 12-10 MEMS: Introduction and Fundamentals © 2006 by Taylor & Francis Group, LLC ∇* ϭ (∂/∂x, ∂/∂y, ∂/∂z) is the gradient operator and ∇ s is the surface gradient with respect to the inter- face z ϭ h(x, y, t). Note that in Equation (12.4) the “dot” represents both the inner product of two vectors and the product of a tensor and a vector, respectively. Equation (12.4a) is the kinematic boundary condition formulated in the absence of interfacial mass transfer and represents the balance between the normal component of the liquid velocity at the interface and the velocity of the interface itself. An appropriate change should be made in Equation (12.4a) to accommodate the phenomena of evaporation or condensation (see the section on phase changes). Equation (12.4b), which constitutes the balance of interfacial stresses in the absence of interfacial mass transfer, has three components. The physical meaning of its two tangential components is that the shear stress at the interface is balanced by the sum of the respective Π i , i ϭ 1, 2 and the surface gradient of surface tension σ . The normal component of Equation (12.4b) states that the difference between the normal interfacial stress and Π 3 exhibits a jump equal to the product of twice the mean curvature of the film interface and surface ten- sion. This jump is known in the literature as the capillary pressure. When the external force Π is zero, and the fluid has zero viscosity or the fluid is static v ϭ 0, then T и n и n ϭ Ϫp, and Equation (12.4b) reduces to the well-known Young–Laplace equation. This equation describes, for instance, the excess pressure in an air bubble gauged to the external pressure, as twice the surface tension divided by the bubble radius (see e.g., [Landau and Lifshitz, 1987]). The subsequent derivations closely follow those made by Oron et al. (1997) when explicitly extended into three dimensions. Projecting Equation (12.4b) onto the directions n, t 1 , t 2 , respectively, yields Ϫp ϩ ϭ 2 ~ H σ ϩ Π 3 , µ [(u z ϩ w x )(1 Ϫ h 2 x ) Ϫ (v z ϩ w y )h x h y Ϫ (u y ϩ v x )h y Ϫ 2(u x Ϫ w z )h x ] ϭ ΂ Π 1 ϩ ΃ (1 ϩ h 2 x ϩ h 2 y ) 1/2 , (12.6) µ [Ϫ(u z ϩ w x )h x h y ϩ (v z ϩ w y )(1 Ϫ h 2 y ) Ϫ (u y ϩ v x )h x Ϫ 2(v y Ϫ w z )h y ] ϭ ΂ Π 2 ϩ ᎏ ∂ ∂ σ y ᎏ ΃ (1 ϩ h 2 x ϩ h 2 y ) 1/2 . Let us now introduce scales appropriate for thin films where the transverse length scale is much smaller than the lateral ones. Assume length scales in the lateral directions, x and y, to be defined by wavelength λ of the interfacial disturbance on a film of mean thickness d. The film is referred to as thin film if the interfacial distortions are much longer than the mean film thickness, that is, ε ϭ ᎏ λ d ᎏ ϽϽ 1. (12.7) The z-coordinate (normal to the solid substrate) is normalized with respect to d, while the coordinates x, y are scaled with λ or equivalently d/ ε . Thus the dimensionless z-coordinate is defined as ς ϭ , (12.8a) while the dimensionless x- and y-coordinates are given by ξ ϭ , η ϭ . (12.8b) It is assumed that in the new spatial variables no rapid variations occur as ε → 0, then , , ϭ O(1). (12.8c) ∂ ᎏ ∂ ς ∂ ᎏ ∂ η ∂ ᎏ ∂ ξ ε y ᎏ d ε x ᎏ d z ᎏ d ∂ σ ᎏ ∂ x 2 µ [u x (h 2 x Ϫ 1) ϩ v y (h 2 y Ϫ 1) ϩ h x h y (u y ϩ v x ) Ϫ h x (u z ϩ w x ) Ϫ h y (v z ϩ w y )] ᎏᎏᎏᎏᎏᎏᎏᎏ 1 ϩ h 2 x ϩ h 2 y Physics of Thin Liquid Films 12-11 © 2006 by Taylor & Francis Group, LLC If the lateral components of the velocity field u, v are assumed to be of order one and U 0 denotes the charac- teristic velocity of the problem, the dimensionless fluid velocities in the x- and y- directions are defined as U ϭ , V ϭ . (12.8d) Then the continuity Equation (12.2) requires that the z-component of the velocity field w is small, and the dimensionless fluid velocity in the z-direction is defined as W ϭ (12.8e) We stress that the characteristic velocity U 0 is not specified here for the sake of generality. The freedom of choosing this value is thus given to the user. We just note one of the possible choices but not the unique one U 0 ϭ µ / ρ d, which is known in the literature as a “viscous velocity.” Time is scaled in the units of λ /U 0 , so that the asymptotically long-time behavior of the film is con- sidered. The dimensionless time is therefore defined via τ ϭ (12.8f) Finally, because of the assumed slow lateral variation of the film interface, one expects locally parallel flow in the liquid, so that the pressure gradient is balanced with the viscous stress p x ϰ µ u zz , and the dimen- sionless interfacial stresses, body-force potential and pressure are defined, respectively, as (Π 1 , Π 2 , Π 3 ) ϭ (Π ˆ 1 , Π ˆ 2 , ε Π ˆ 3 ), (Φ, P) ϭ ( φ , p). (12.8g) Notice that pressure is asymptotically large similar to the situation arising in the lubrication effect [Schlichting, 1968]. If all these dimensionless variables are substituted into the governing system of Equations (12.2)–(12.5), the following scaled system is obtained: U ξ ϩ V η ϩ W ς ϭ 0, (12.9a) ε R(U τ ϩ UU ξ ϩ VU η ϩ WU ς ) ϭ ϪP ξ ϩ U ς ς ϩ ε 2 (U ξξ ϩ U ηη ) Ϫ Φ ξ , (12.9b) ε R(V τ ϩ UV ξ ϩ VV η ϩ WV ς ) ϭ ϪP η ϩ V ςς ϩ ε 2 (V ξξ ϩ V ηη ) Ϫ Φ η , (12.9c) ε 3 R(W τ ϩ UW ξ ϩ VW η ϩ WW ς ) ϭ ϪP ς ϩ ε 2 W ςς ϩ ε 4 (W ξξ ϩ W ηη ) Ϫ Φ ς . (12.9d) At ς ϭ 0: W ϭ 0, U ϭ 0, V ϭ 0. (12.10) At ς ϭ H: W ϭ H τ ϩ UH ξ ϩ VH η , (12.11a) ϭ P ϩ Π ˆ 3 ϩ , (12.11b) (U ς ϩ ε 2 W ξ )(1 Ϫ ε 2 H 2 ξ ) Ϫ ε 2 (V ς ϩ ε 2 W η )H ξ H η Ϫ ε 2 (U η ϩ V ξ )H η Ϫ 2 ε 2 (U ξ Ϫ W ς ) H ξ ϭ (Π ˆ 1 ϩ Σ ξ )[1 ϩ ε 2 (H 2 ξ ϩ H 2 η )] 1/2 , (12.11c) S ෆ ε 3 [H ξξ (1 ϩ ε 2 H 2 η ) ϩ H ηη (1 ϩ ε 2 H 2 ξ ) Ϫ 2 ε 2 H ξ H η H ξη ] ᎏᎏᎏᎏᎏᎏ [1 ϩ ε 2 (H 2 ξ ϩ H 2 η )] 3/2 2 ε 2 [U ξ ( ε 2 H 2 ξ Ϫ 1) ϩ V η ( ε 2 H 2 η Ϫ 1) ϩ ε 2 H ξ H η (U η ϩ V ξ ) Ϫ H ξ (U ς ϩ W ξ ) Ϫ H η (V ς ϩ W η )] ᎏᎏᎏᎏᎏᎏᎏᎏᎏᎏ 1 ϩ ε 2 (H 2 ξ ϩ H 2 η ) ε d ᎏ µ U 0 d ᎏ µ U 0 ε U 0 t ᎏ d w ᎏ ε U 0 v ᎏ U 0 u ᎏ U 0 12-12 MEMS: Introduction and Fundamentals © 2006 by Taylor & Francis Group, LLC (V ς ϩ ε 2 W η )(1 Ϫ ε 2 H 2 η ) Ϫ ε 2 (U ς ϩ ε 2 W ξ )H ξ H η Ϫ ε 2 (U η ϩ V ξ )H ξ Ϫ 2 ε 2 (V η Ϫ W ς )H η ϭ (Π ˆ 2 ϩ Σ η )[1 ϩ ε 2 (H 2 ξ ϩ H 2 η )] 1/2 . (12.11d) Here H ϭ h/d is the dimensionless thickness of the film and Σ ϭ εσ / µ U 0 is the dimensionless surface ten- sion normalized with respect to its characteristic value. The Reynolds number R and the inverse capillary number S ෆ are defined by R ϭ , S ෆ ϭ . (12.12) The continuity Equation (12.9a) is now integrated in ς across the film from 0 to H ( ξ , η , τ ), and Equations (12.10) and (12.11a) are used along with integration by parts to obtain H τ ϩ ͵ H 0 U dς ϩ ͵ H 0 V d ς ϭ 0. (12.13) Equation (12.13) is a more convenient form of the kinematic condition because only two of three com- ponents of the fluid velocity field appear explicitly. It also warrants conservation of mass in a domain with a deflecting upper boundary. The solution of the governing Equations (12.2)–(12.5) is sought in the form of expansion of the dependent variables into asymptotic series in powers of the small parameter ε : U ϭ U (0) ϩ ε U (1) ϩ ε 2 U (2) ϩ …, V ϭ V (0) ϩ ε V (1) ϩ ε 2 V (2) ϩ …, W ϭ W (0) ϩ ε W (1) ϩ ε 2 W (2) ϩ …, P ϭ P (0) ϩ ε P (1) ϩ ε 2 P (2) ϩ …. (12.14) One way to approximate the solution of the governing system is to assume that R, S ෆ ϭ O(1) as ε → 0. Under this assumption the inertial terms, measured by ε R, are one order of magnitude smaller than the dominant viscous terms, consistent with the local-parallel-flow assumption.The surface tension terms,meas- ured by S ෆ ε 3 , are two orders of magnitude smaller and would be lost. It is essential to retain surface-tension effects at leading order, so it is assumed that capillary effects are strong relative to those of viscosity and S ෆ ϭ S ε Ϫ3 . (12.15) It is then assumed that R, S ϭ O(1), as ε → 0. Equations (12.14) and (12.15) are substituted into Equations (12.9)–(12.11) and (12.13), and the resulting equations are sorted with respect to the powers of ε . At leading order in ε the governing system becomes, after omitting the superscript “zero” in U (0) , V (0) , W (0) , and P (0) , U ς ς ϭ (P ϩ Φ) ξ , (12.16a) V ς ς ϭ (P ϩ Φ) η , (12.16b) (P ϩ Φ) ς ϭ 0, (12.16c) H τ ϩ UH ξ ϩ VH η Ϫ W ϭ 0, (12.16d) U ξ ϩ V η ϩ W ς ϭ 0 (12.16e) with the boundary conditions at ς ϭ 0: W ϭ 0, U ϭ 0, V ϭ 0, (12.17) and at ς ϭ H: P ϭ ϪΠ ˆ 3 Ϫ S(H ξξ ϩ H ηη ), U ς ϭ Π ˆ 1 ϩ Σ ξ , (12.18) V ς ϭ Π ˆ 2 ϩ Σ η . ∂ ᎏ ∂η ∂ ᎏ ∂ξ σ ᎏ U 0 µ U 0 d ρ ᎏ µ Physics of Thin Liquid Films 12-13 © 2006 by Taylor & Francis Group, LLC We note here that Equations (12.16)–(12.18) are linear with respect to the variables U, V, W, P. The only nonlinearity of this problem is associated, as seen from Equation (12.19) in conjunction with the kinematic condition Equation (12.16d), with the local film thickness H( ξ , η , τ ). Solving Equations (12.16)–(12.18) yields U ϭ ΄ ς 2 Ϫ H ς ΅ (Φ Ϫ Π ˆ 3 | ς ϭH Ϫ S∇ 2 H) ξ ϩ ς (Π ˆ 1 ϩ Σ ξ ), V ϭ ΄ ς 2 Ϫ H ς ΅ (Φ Ϫ Π ˆ 3 | ς ϭH Ϫ S∇ 2 H) η ϩ ς (Π ˆ 2 ϩ Σ η ), (12.19) W ϭ Ϫ ͵ ς 0 (U ξ ϩ V η )d ς , P ϭ ϪΠ ˆ 3 | ς ϭH ϪS∇ 2 H. If Equation (12.19) is substituted into the mass conservation Equation (12.13), one obtains the appro- priate evolution equation for the interface, H τ ϩ ∇ и [H 2 (Π ˆ * ϩ ∇Σ)] ϩ ∇ и {H 3 [∇(Π ˆ 3 Ϫ Φ| ς ϭH ) ϩ S∇∇ 2 H]} ϭ 0, (12.20) where Π ˆ * ϭ (Π ˆ 1 , Π ˆ 2 ) is the tangential projection of the dimensionless vector Π ˆ , ∇ ≡ (∂/∂ ξ , ∂/∂ η ) and ∇ 2 ϵ ∂ 2 /∂ ξ 2 ϩ ∂ 2 /∂ η 2 . In two dimensions (∂/∂ η ϭ 0) this evolution equation reduces to H τ ϩ [H 2 (Π ˆ 1 ϩ Σ ξ )] ξ ϩ {H 3 [(Π ˆ 3 Ϫ Φ| ς ϭH ) ξ ϩ SH ξξξ ]} ξ ϭ 0. (12.21) In these equations the location of the film interface H ϭ H( ξ , η , τ ) is unknown and is determined from the solution of the corresponding partial differential equation. When such a solution is obtained, the components of the velocity and the pressure fields can be determined from Equation (12.19). The physical significance of the terms becomes apparent when Equations (12.20) and (12.21) are writ- ten in the original dimensional variables: µ h t ϩ ∇ ෆ и [h 2 (Π * ϩ ∇ ෆ σ )] ϩ ∇ ෆ и {h 3 [∇ ෆ (Π 3 Ϫ φ | zϭh ) ϩ σ ∇ ෆ ∇ ෆ 2 h]} ϭ 0, (12.22) with ∇ ෆ ϵ (∂/∂x, ∂/∂y), ∇ ෆ 2 ϵ (∂ 2 /∂x 2 ϩ ∂ 2 /∂y 2 ) and µ h t ϩ [h 2 (Π 1 ϩ σ x )] x ϩ {h 3 [(Π 3 Ϫ φ | zϭh ) x ϩ σ h xxx ]} x ϭ 0. (12.23) The first term in Equations (12.22) and (12.23) represents the effect of viscous damping, while the next ones account, respectively, for the effects of the imposed tangential interfacial stress, non-uniformity of surface tension, the imposed normal interfacial stress, body forces, and surface tension on the dynamics of the film. In the following examples, two- and three-dimensional cases are examined. Unless specified, only dis- turbances periodic in x and y are discussed. Thus, λ is the wavelength of these disturbances, and 2 π d/ λ is the corresponding dimensionless wavenumber. In accordance with this, Equations (12.20)–(12.23) are normally solved with periodic boundary conditions. These equations whether in two or three dimensions are of fourth order in each of the spatial variables, and therefore four boundary conditions are needed to define a well-posed mathematical problem. These four boundary conditions imply periodicity of the solution H and its first, second, and third derivatives with respect to the corresponding spatial variable. At the same time, Equations (12.20)–(12.23) are of first order in time, thus one initial condition is needed to complete the well-posed statement of the problem. This initial condition representing the location of the film interface at t ϭ 0 or τ ϭ 0 is usually taken as a small-amplitude random or sinusoidal distur- bance on top of the uniform state given by H ϭ 1. In two dimensions it can be written by H( τ ϭ 0, ξ ) ϭ 1 ϩ δ sin( ξ ϩ ϕ ) or H( τ ϭ 0, ξ ) ϭ 1 ϩ δ rand( ξ ), (12.24) 1 ᎏ 3 1 ᎏ 2 1 ᎏ 3 1 ᎏ 2 1 ᎏ 3 1 ᎏ 2 1 ᎏ 3 1 ᎏ 2 1 ᎏ 2 1 ᎏ 2 12-14 MEMS: Introduction and Fundamentals © 2006 by Taylor & Francis Group, LLC [...]... two interfaces of the film are mutually j attracting when the separation distance is relatively large This drives the instability of the flat state of the film surface On the other hand, the two interfaces of the film are mutually repelling when the separation distance is relatively short This leads to a final saturation of the amplitude of the interfacial undulation © 2006 by Taylor & Francis Group,... where τR is the time of rupture and b is the constant which should be determined from the matching with the far-from-rupture solution The minimal thickness of the film close to the rupture point is therefore © 2006 by Taylor & Francis Group, LLC 1 2-2 2 MEMS: Introduction and Fundamentals expected to decrease linearly with time This allows the long-wave analysis to be extrapolated closer to the point where... along with the availability and affordability of fast computers helped to advance the study of the pertinent phenomena The main interest is centered about the pattern formation and the quest for the dominant mechanisms driving the film evolution In the context of the latter issue the polemics are ongoing between the two candidates, namely thin film instability arising from the interaction between the intermolecular... film is one-dimensional across it, and the energy input from absorption of radiation energy in the thicker part of the film is greater than in its thinner part Thus, the interfacial temperature at the depression is lower than at the crest of the interface, and the thermocapillary stress drives the liquid into the depression promoting stabilization of the interface All this is different from the standard... attributed the type of film dewetting to the relative position of the average thickness of the film d and the location of the minimum of the function ∂φ/∂h When the film is thicker than the thickness corresponding to the minimum of ∂φ/∂h, the film dewets by formation of holes In the opposite case, dewetting sets in by formation of liquid ridges which break up further into droplets In either case, ripening... the frequencydependent dielectric properties of the materials in the layered system The potential φ of the van der Waals forces is frequently specified in terms of the excess intermolecular free energy ∆G These two values are related each to other via ∂∆G φ ϭ ᎏ ∂h (12.29) It follows in this case from Equation (12.22) in the 3-D case and Equation (12.23) in the 2-D case that the film is unstable to infinitesimal... (12.48b) Here c is the specific heat of the fluid, kth is its thermal conductivity, ϑ0 is the temperature of the rigid sub strate assumed to be uniform, and q is the rate of internal heat generation The boundary condition Equation (12.48b) is Newton`s cooling law, and αth is the heat-transfer coefficient describing the rate of heat transfer from the liquid to the ambient gas phase held at the constant temperature... Second, the linear stability analysis is carried out here in the two-dimensional case The same can be done in the three-dimensional case with respect to the normal modes hЈ ϭ hЈ exp(ikxx ϩ ikyy ϩ ωt), hЈ ϭ const, 0 0 where kx, ky are, respectively, the wavenumbers in the x and y directions As in the physical problem at hand, the symmetry is such that the spatial variables x and y are interchangeable and the. .. d/Lm, where Lm is the mean penetration length of the incident radiation by the wave of the wavelength λ Assuming that the film is non-scattering, the solid surface underneath is non-reflecting and the intensity of absorbed radiation is equal to the intensity of internal heat sources, the latter is expressed [Oron, 2004a] by q(z) ϭ q exp[Ϫaλ(h Ϫ z)], where h represents the location of the interface and... is only an estimate based on the linear stability analysis, and the effect of nonlinearities on the rate of film leveling can be found only from the solution of Equation (12.25) Equations (12.25a, b) with the obvious change in the sign of the gravity term in each of these also apply to the case of a film on the underside of a plate This case is known in the literature as the Rayleigh–Taylor instability . of film surface deformations is well-correlated with the wavelength of the most amplified linear mode pro- portional to d 2 . Similar conclusions about the dominance of the nucleation mechanism were. τ R is the time of rupture and b is the constant which should be determined from the matching with the far-from-rupture solution. The minimal thickness of the film close to the rupture point is therefore b ξ ᎏ 2 b 2 ᎏ 2 d ᎏ hЈ 0 48 π 2 d 5 σ ᎏ AЈ σ a 2 ᎏ 3d AЈ ᎏ 2 πσ 1 ᎏ d Physics. compact form must be tract- able and, at the same time, complex enough to retain the main features of the problem at hand. The most appropriate analytical method of dealing with the above complexities