Practical Asymptotics Practical Asymptotics Edited by H.K. KUlKEN Eindhoven University of Technology, Eindhoven, The Netherlands Reprinted from Journal of Engineering Mathematics, Volume 39, Nas. 1-4 (March 2001) SPRINGER-SCIENCE+ BUSINESS MEDIA, B.V. ISBN 978-94-010-3827-0 ISBN 978-94-010-0698-9 (eBook) DOI 10.1007/978-94-010-0698-9 TABLE OF CONTENTS Practical asymptotics byH. K. Kuiken 1-2 Shearflow over aparticulate or fibrousplate by C. Pozrikidis 3-24 Current-voltage characteristics from an asymptotic analysis of the MOSFETequa- tions by E. Cumberbatch, H.Abebe and H. Morris 25-46 Separating shearflowpast asurface-mounted blunt obstacle by S. Bhattacharyya, S.C.R. Dennis and F.T. Smith 47-62 Microwavejoining oftwo long hollow tubes: an asymptotic theory and numerical simulations by G.A. Kriegsmann and J.Luke 63-78 Fastcomputation of limit cycles in an industrial application byS.Gueron and N.Liron 79-86 Asymptotic analysis of the steady-state and time-dependent Bermanproblem by1.R. King and S.M.Cox 87- 130 Generation ofwater wavesand boresby impulsive bottom flux byM. Landrini and P.A. Tyvand 131-170 On the asymptotic analysis of surface-stress-driven thin-layer flow byL. W.Schwartz 171-188 Matched asymptotic expansions and the numerical treatment of viscous-inviscid interaction by A. E.P.Veldman 189-206 Stokes flow around an asymmetric channel divider; a computational approach us- ing matlab by J.D.Fehribach and A.M.J. Davis 207-220 The frozen-field approximation and the Ginzburg-Landau equations ofsupercon- ductivity by H.G. Kaper and H. Nordborg 221-240 Analytical approximations tothe viscous glass-flow problem in the mould-plunger pressing process, including an investigaton ofboundary conditions by S.W. Rienstra and T.D.Chandra 241-259 Asymptotic adaptive methods for multi-scale problems in fluid mechanics by R. Klein, N. Botta, T. Schneider, C. D. Munz, S. Roller, A. Meister, L. Hoffmann and T. Sonar 261-343 Asymptotic analysis of the flow of shear-thinning foodstuffs in annular scraped heat exchangers by A.D.Fitt and C. P.Please 345-366 The evolution oftravelling waves from chemical-clock reactions by S.J. Preece,J.Billingham and A. C. King 367-385 Journal ofEngineering Mathematics 39: 1-2,200I. ©2001 KluwerAcademicPublishers. Practical asymptotics H. K. KUIKEN DepartmentofMathematics andComputing Science,Eindhoven UniversityofTechnology,P.O.Box513, 5600MB Eindhoven,TheNetherlands Received25July 1998,accepted inrevisedform 26September 2000 Abstract. The termpractical asymptotics isexplained.Itisargued that many practical problems are amenable toand will benefit from the asymptotic approach. Fifteen papers by authors who are active in a wide range of disciplines demonstratethis.Itisarguedthattheteachingofasymptoticmethods shouldremainanintegralpartof anysophisticatednumericscurriculum, raising itfarabovethelevelofmerenumbercrunching. Keywords: asymptotics, asymptotics teaching,non-dimensionalisation,numerics,real-world problems. A few years ago this journal published a special issue entitled Large-Scale Numerical Mod elling ofProblemsInvolving theNavier-Stokes Equations (32(1997) 101-280).That appeared to be a very timely publication. Indeed, these days applied mathematics seems, at first sight anyway, to become more and more dominated by direct numerical simulation. Admittedly, this leads to new insights which, itwould seem, could not havebeen attained by other means. The purpose of that issue was tohighlight that particular aspect. The interest of large sections ofthe community of applied mathematicians being focussed intheabove direction nowadays, theolder andbroad subject ofasymptotics seems tobelosing the popularity it once enjoyed and the impact it had. This is partly because many believe that asymptotics deals with exceptional cases which are usually outside the practical domain. However,this isamisconception! Tomake this clear, we offerthe following argument: Having non-dimensionalised your problem, consideratypical dimensionless group ND = (XIk,(Xkz2(Xk33•••, where the k, are real-valued exponents that can beboth positive and negative. Weknow that, within agiven setofunits, eachofthephysical parameters can assume valuesthat mayvary (Xi widely depending upon the problem in hand, i.e.(Xi '" lO'",where the integers n, can range from large negative to large positive values.Thus, ND '" lOP, and we can ask ourselves: what is the probability that p '" O? Ofcourse, p is most likely to beless than -lor larger than 1,which means that the problem can be simplified through the application of asymptotic techniques, without seriously affecting the practical usefulness of the reduced model.The technique, orrather the craft, by which the reduction isaccomplished iscalled PracticalAsymptotics. It is a highly intuitive process and is based, to a large extent, onphysical reasoning. Papers which demonstrate the power of practical asymptotics should involve both large scale computations on the full model and computations based on the asymptotically reduced 2 H. K. Kuiken model. Acomparison between thetwoshould then illustrate aconsiderable reduction incom puting effort in terms of cpu time.Furtheremphasis should be put on the explicit representa tionoftrends,physical relationships, rules, laws, etc.,asexpressed by mathematical formulae resulting from the asymptotics with numerically generated coefficients. This, then, should contrast with the practice of large-scale numerics which usually results in the presentation of numbers, (colour) plates orcontour plots relating to specific examples. In the present issue fifteen papers have been collected, each seeking to demonstrate the usefulness and practical nature of the asymptotic approach for the discipline from which it arose. These papers address subjects as varied as: the production of glass bottles, semicon ductors, surface-tension-driven flows, microwave-assisted joining of tubes, viscous-inviscid interaction, industrial limit cycles, heat exchangers used in the production of foodstuffs, wa ter waves generated by bottom topography, channel flows, chemical-clock reactions, Stokes flows, separated flows,superconductivity, flowsoverfibrous plates, multiscale problems aris ing inlow-Mach-number flows;awide range of subjects indeed! We hope that this issue will demonstrate the continuing usefulness and validity of the asymptotic approach in reducing the complexity of mathematical models to apractical mini mum, i.e., without unduly sacrificing their accuracy, thus achieving large reductions in com puting effort and increased physical understanding. Webelieve thattheteaching ofasymptotic methods should remain an integral part of the academic curriculum as an indispensable tool for the numericist wishing to tackle problems ofever-increasing complexity. JournalofEngineeringMathematics 39: 3-24,200I. ©2001KluwerAcademicPublishers. Shear flow over a particulate or fibrous plate C.POZRIKIDIS DepartmentofMechanicalandAerospaceEngineering UniversityofCalifornia,SailDiego,LaJolla, California92093-0411, U.S.A.(e-mail:[email protected], InternetURL:http://stokes.ucsd.edu/c..,pozrikidis) Received: 15September 1999,acceptedinrevisedform 10April2000 Abstract. Simple shearflowoveraporousplateconsistingofaplanararrayofparticlesisstudiedasamodelof flowovera membrane.The mainobjective istocompute theslipvelocity defined with reference tothe velocity profilefarabove the plate,andthedrift velocity induced by theshear flowunderneath the plate.Thedifference betweenthesetwovelocitiesisshowntobeproportional tothethicknessoftheplate.Whenthegeometryofthe particlearrayisanisotropic,thedirectionsoftheslipanddrift velocityaregenerally different fromthedirection ofthe overpassing shear flow. An integral formulation is developed todescribe flowover a plate consisting of a periodic latticeofparticles witharbitrary shape,and integralrepresentations for the velocity and pressure are developedintermsofthedoubly-periodicGreen'sfunctionofthree-dimensionalStokesflow.Basedontheintegral representation,asymptotic expressions for the slipand drift velocityarederived todescribe the limitwhere the particlesize issmallcomparedtothe inter-particleseparation, andnumerical resultsare presented for spherical andspheroidalparticlesofarbitrarysize.Theasymptoticresultsarefoundtobeaccurateoveranextendedrangeof particlesizes.Tostudythelimitofsmallplateporosity,theavailablesolutionforshearflowoveraplanewallwith acircularorificeisusedtodescribeflowoveraplatewithahomogeneousdistributionofcircularperforations,and expressionsfortheslipanddriftvelocityarederived.Correspondingresultsarepresentedforaxialandtransverse shearnowoveraperiodicarrayofcylindersarrangeddistributedinaplane.Streamlinepatternillustrationsconfirm thatanegativedriftvelocityisduetotheonsetofeddiesbetweenclosely-spacedparticles. Keywords: Stokesflow,integralequations,shear flow,porousplate,membrane. 1. Introduction Animportant areaofresearch inbiomechanics concerns theeffect ofashear flowon theequi librium shapes of, and mass transport through, membranes enclosing vesicles and biological cells [1-3].The membranestypically consistoflipid bilayers,sometimes resting onanetwork of proteins comprising the cytoskeleton, separating the vesicle or cytoplasmic fluid from the ambient plasmaorbufferfluid.From thepointofviewofhydrodynamics,amembrane maybe regarded asascreenorsieve,andthe problem offlowthrough orover itmay bestudied within the more general context of flowover aporousor irregular plate separating two semi-infinite regionsoccupied bythesameordifferent fluids.This moregeneral problem ispertinent tosev eral other areas of biofluiddynamics and mainstream engineering fluid mechanics involving, forexample,flowoverarrangements of cilia orbundled tubes. Previous studies of the thermodynamics and hydrodynamics of polymeric and biological membraneshavemodeled thenetwork ofthe fundamental molecularunitscomprising amem brane as a planar lattice of particles over which the smaller molecules of the solute slide [4, 5].The theory seekstopredict thespecies diffusivityand toestablish the relationship between the pressure drop and the geometrical properties of the membrane expressed in terms of the hydrodynamic resistance for flow normal to the membrane. Ishii [6] studied flow normal to 4 C:. lJozrikitiis a planar lattice ofspherical particles in the asymptotic limit where the particle size is small compared tothe particle separation.Tothe author'sknowledge, shearflowover aporousplate or particulate plane has not been addressed explicitly by previous authors, although three related classes ofproblems have been discussed in detail. The firstclassofproblemsinvolvesflowoveranirregularsurface with smallorlargeampli tude protuberancespossiblyoffractal nature [7,8]. Issuesofinterestincludetheestablishment ofthe physical origin oftheno-slip boundary condition,thecomputation ofthe slipvelocityas afunction ofthe morphologyofthe surface irregularities,and the study oftransport properties in terms ofeffective heat or masstransfer coefficients.The second classofproblems involves shear flow over a semi-infinite particulate matrix modeling a porous medium. Reviews and numerical simulationsfortwo- andthree-dimensional arrangements were presentedbyLarson and Higdon [9,10] and Sangani and Behl [11]. Theirresults illustrated the dependence of the slip velocity on the geometry ofthe porous medium microstructure, and helped to establish a theoretical foundation for Brinkman's equation governing the structure ofthe flow near the surface of a porous medium, also pointing out its limitations. The third class of problems involves shear flow over a wall with a circular hole or side pore, possibly in the presence of suction that drives the fluid through the hole or into the pore, with applications to particle entrainment and particle screening [12-15]. Inthispaper,westudyshearflowoveraplanarlatticeofparticlesformingaporoussurface. When the thicknessoftheplate iscomparableto,or larger than,thegapsbetween the particles, regions ofrecirculating flow develop in the intervening spacing, and the flow decays rapidly underneath the array to give a virtually quiescent lower fluid. When, however, the particle size is smaller than the inter-particle separation,the shear flow penetrates the lower fluid,and a uniform drift velocity is established underneath the array. If the particle shape or lattice geometry is anisotropic, the direction of the drift velocity is generally different from that of the overpassing shear flow.Far above the lattice, the flowreduces to simple shear flow with a macroscopic slip velocity similarto that established inshear flowovera porous material. The main goal of the present work is to illustrate the dependence ofthe slip and drift velocity on the particle shape and lattice geometry. The asymptotic and numerical studies are based on integral representations of periodic Stokes flow using an appropriate Green's function. In implementing the numerical procedure for solving an integral equation of the first kind for the traction over a particle surface, a general method isdeveloped for removing theeigenfunction of the single-layerStokeshydro dynamic potential. Tocomplement the results for three-dimensional flow over a planar array, tworelatedconfigurationsarealsoconsidered.Thefirstconfigurationinvolves shearflowover aflatplate of zero thickness containing a homogeneous distribution ofcircularperforations of smallsize, which isrelevant to shear flowover aporous plate with a large solidarealfraction. The second configuration involves longitudinal and transverse shear flow over an array of cylinders. Comparisons of the results for three-and two-dimensional configurations illustrate once again the fundamental differences in the nature ofthe corresponding flows. 2. Shearflowover a planarparticle lattice Consider infinite simple shear flow over a two-dimensional lattice of identical rigid particles positioned at the vertices of a regular lattice that is parallel to the xy plane, as shown in Shearflowover aparticulate orfibrousplate 5 z y / x Figure I. Shearflowoveraplanarlatticeofparticles.Farbelowthelattice,thevelocitytendstotheuniformdrift velocitywhosedirection and magnitude aredetermined bythe particle shape and size, the latticegeometry,and thedirectionoftheshearflow. Figure 1.Thelattice isdescribed by two base vectors 31and 32, so that the designated centers oftwo particles labelled n and m are related by (I) where i1and ii are two integers. Far above the lattice, as z tends to + 00, the flow reduces to simple shear flow with shear rate y in the direction ofthe unit vector e that is parallel to the xy plane. Thus, the slope of the velocity isrequired toexhibit the asymptotic behavior . du hm - -+ ye (2) z--->+oodz while the pressure tends to aconstant value denoted by p"?". Farbelow the lattice, asztends to - 00, the shear stress isrequired to vanish, du lim - -+ 0 (3) z--->-oodz and the pressure tends to a constant value denoted by p Integration ofthe preceding two - 00. equations with respect to z produces the asymptotic forms lim u -+ yze +U+oo +edt (4) z--->+oo and lim u -+ U-oo +edt, (5) z~-oo whereU+oo and U-oo aretwoconstant velocities parallel to the xy plane,bothtobecomputed aspart of the solution, and 'edt' stands for exponentially decaying terms. The magnitude and direction ofthe slip velocity U+oo depend on the definition ofthe origin ofthe z axis; in the present case, the origin coincides with the designated particle centers [9, 16]. The magnitude and directionofthe uniformdriftvelocity induced bythe shearflow underthe lattice,U-oo,on 6 C.Pozrikidis the otherhand, isindependent ofthe origin ofthe zaxis. Neither the slip velocity nor the drift velocity is necessarily oriented in the direction ofthe shear flow. In practice, the shear flow may be generated by the translation ofa flat plate located at z = zp with velocity Upparallel + to the particle lattice. Using the asymptotic form (4), we find Up = Yzpe U"?", which provides us with an expression for evaluating ye and U+oo, when another linear relationship between them has been established. Returning to the problem ofinfinite shear flow, we perform a force balance overacontrol volume that is confined between (a) four planes that are perpendicular to the x)' plane and enclose one lattice cell and thus one particle, (b) two planes that are parallel to the .ry plane located far above or below the particle lattice, and (c) the surface ofthe enclosed particle, to obtain F == [ f(x) dS(x) = flYAe+A(p-OO - p+OO)ez, (6) }Particle where F is the force exerted on one particle, f = a .n is the hydrodynamic traction, a is the stress tensor, n is the unit vector normal to the particle pointing into the fluid, /l- is the fluid viscosity, A is the area of one lattice cell, and e, is the unit vector pointing along the z axis. The motion ofthe fluid isgoverned by the equations ofStokesflow [17] '\7·u=o (7) which are to be solved subject to (a) the far-field conditions expressed by Equations (4) and (5), (b) the no-slip and no-penetration condition u = 0 on the particle surface, and (c) the periodicity condition (8) where iI and i: are two integers.Ourmain objective is to compute the slip and drift velocities asfunctions ofthe particle shape and size and ofthe lattice geometry, in the limit ofvanishing Reynolds number. 2.1. INTEGRAL FORMULATION To prepare the ground for the integral formulation, we introduce the Green's function ofthe equations ofStokes flow describing the doubly-periodic flow due toa two-dimensional lattice of point forces that is identical to the particle lattice shown in Figure I. The velocity and pressure field induced by the point forces at the point x are expressed by I III(x) = --Gij(x,xo) bj, (9) 8Jr/l- where G is the velocity Green's function tensor, P is the pressure Green's function vector, and b is the strength of a point force; one arbitrarily selected point force is located at XQ. The Green's functions for the velocity and pressure satisfy the periodicity condition shown in Equation (8).Moreover, as z - Zotends to +00, we require the asymptotic behavior 8Jr 8Jr lim G(x,xo) -+ - - (z - zo)J +edt, lim P(x, xo) -+ -e +edt, (10) z->+oo A z->+oo A z where J is the identity matrix but with the third diagonal component corresponding to the z axis set equal to zero. As z - Zo tends to - 00, all components ofthe Green's functions are