Axisymmetric mixed free boundary value problem

Mixed Free Boundary Value Problem for Laplace equation in axisymmetrical case is considered. We take into consideration mean and Gauss curvatures of the free boundary. The problem of this type arise on investigating thermal equillibrium of two phases. We take into account capillary forces acting in...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Datum:2007
1. Verfasser: Shcherbakov, E.
Format: Artikel
Sprache:Englisch
Veröffentlicht: Інститут прикладної математики і механіки НАН України 2007
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/10114
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Axisymmetric mixed free boundary value problem / E. Shcherbakov // Нелинейные граничные задачи. — 2007. — Т. 17. — С. 130-150. — Бібліогр.: 12 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
_version_ 1859758357118713856
author Shcherbakov, E.
author_facet Shcherbakov, E.
citation_txt Axisymmetric mixed free boundary value problem / E. Shcherbakov // Нелинейные граничные задачи. — 2007. — Т. 17. — С. 130-150. — Бібліогр.: 12 назв. — англ.
collection DSpace DC
description Mixed Free Boundary Value Problem for Laplace equation in axisymmetrical case is considered. We take into consideration mean and Gauss curvatures of the free boundary. The problem of this type arise on investigating thermal equillibrium of two phases. We take into account capillary forces acting in intermediate layer separating different phases. Plane model of the equillibrium without capillary forces was considered in the paper([1]). We consider variational problem whose solution is generalized solution of the boundary value problem.We prove regularity of solution, analyticity of free boundary and investigate its properties.
first_indexed 2025-12-02T02:07:46Z
format Article
fulltext 130 Нелинейные граничные задачи 17, 130-150 (2007) c©2007. E. Shcherbakov AXISYMMETRICAL MIXED FREE BOUNDARY VALUE PROBLEM Mixed Free Boundary Value Problem for Laplace equation in axisymmetrical case is considered. We take into consideration mean and Gauss curvatures of the free boundary. The problem of this type arise on investigating thermal equillibrium of two phases. We take into account capillary forces acting in intermediate layer separating different phases. Plane model of the equillibrium without capillary forces was considered in the paper([1]). We consider variational problem whose solution is generalized solution of the boundary value problem. We prove regularity of solution, analyticity of free boundary and investigate its properties. Keywords and phrases: Mixed boundary value problem, mean curvature, Gausss curvature, variational problem, free boundary MSC (2000): 35R35 1. Boundary Value Problem. All the domains considered in the paper are meridional sections of axisymmetrical domains.We are searching for a domain Ω, lying in the half stripΠ = {(x, y) |x ∈ (−1, 1) , y ∈ (−∞, 0)}which is sym- metrical with regard to the axis y. The boundary of Ω consists of the set Γ = {(±1, y) , y ∈ (−∞, 0)} and unknown Σ with the endponts (±1, 0). We suppose the curve Σto be twice differentiable everywhwere except of the set ΛΣ of the points lying on the y-axis where it possibly posesses singularity. We are searching also for a function u = u (x, y), u (x, y) = u (−x, y), a solution of the following equation 1 x · ∂ ∂x ( x · ∂u (x, y) ∂x ) + ∂2u (x, y) ∂y2 = 0 (x, y) ∈ Ω, x > 0 (1.1) We suppose that it satisfies the following boundary conditions u (x, y) = 1, (x, y) ∈ Σr, (1.2) ∂u ∂ν (±1, y) + α · u (±1, y) = 0, y ∈ (−∞, 0) , (1.3) Axisymetric mixed free boundary value problem 131 u (∞) = 0 (1.4) l 1 2 · |∇u|2 + κ · H(x, y) + θ · K(x, y) = λ, (x, y) ∈ Σr ⋂ (Π − ΛΣ) , x 6= ±1. (1.5) Here Σr is the set of regular points of the free arcΣ for the Dirichlet boundary value problem (see[2]). The functions H (x, y), K (x, y) re- present values of mean and Gauss curvature of the axisymmetrical surface S at a point (x, y) of its meridional section Ω. The letter νdenotes external normal to the set Γ. The numbers α, κ, θ are nonnegative real numbers. When function u satisfies Laplace equation, κ = θ = 0, we have plane boundary value problem studied in (see[1]). In the sequel we shall call to the problem (1.1)-(1.5) as the main boundary value problem. 2. Variational Problem. We use the variational method for procuring solution of described free boundary value problem. Let I(u, Ω) := ∫ ∫ Ω |∇u(x, y)|2 |x| dx dy − λ ∫ ∫ B |x| dx dy + κ ∫ Σ |x| ds + 2θ ∫ Σ f(ẋ) ds + α ∫ Γ u2 ds (2.1) Here B is complement Π \ Ω̄ to closed domain Ω̄ in the half strip Π, z = z (s) = y (s) + ix (s) is the natural parametric representation of curve Σ, ẋ (s)- derivative dx ds , and f is the function of the following type: 2f(t) = − √ 1 − t2 · t∫ 0 (arcsin σ + σ √ 1 − σ2 − π 2 )× ×(1 − σ2)− 3 2 dσ + E0 √ 1 − t2. (2.2) Let us consider the functional I over the setDof the admissible pairs (u, Ω) defined as follows. The boundary of the domain Ω consists 132 E. Shcherbakov of rectifiable curves Σ = δΩ ⋂ Π, symmetrical with regard toy axis and of the set Γ of the above mentioned type. It is clear that the curve Σ connecting the points (±1, 0) define admissible domains. We denote as Ξthe set of admissible curves Σ.We denote as W the class of admissible functions u,u (x, y) = u (−x, y). These functions are continuous in the domainΩ̄ \ΛΣ. They assume boundary value equal to one on the curve Σ, and vanish at the infinite point. We will suppose that the functions u (x, y) posess on the setΩ generalized derivatives of the first order such that weighted Dirichlet integral is bounded ∫ ∫ Ω |∇u(x, y)|2 · |x| dx dy < ∞ Variational problem: Find minimum of the functionalI = I (u, Ω) on the set D. 3. Symmetrization of functions and domains. Let Ω be an admissible domain, Ωh =Ω ⋂ [(x, y) , −1 ≤ x ≤ 1, y > −h, h > 0] and I0 (u, Ωh) = I0 ′ (u, Ωh) + I0 ′′ (u, Ωh) (3.1) I0 ′ (u, Ωh) = ∫ ∫ Ωh |∇u(x, y)|2 |x| dx dy (3.2) I0 ′′ (u, Ωh) = α · ∫ ∫ Γh |u(x, y)|2 ds (3.3) h ≤ ∞. In this section we are going to study the behaviour of the functional I0 under symmetrization of the solution of the boundary value problem (1.1)-(1.4). We start with the following principle of maximum (see[1]). Lemma 3.1. Let Σ be analytic curve symmetrical with regard to axis y and u = u (x, y) - a solution of the boundary value problem (1.1)-(1.4). Then the functionuattains its maximum value on the curveΣand this value is equal to one. Axisymetric mixed free boundary value problem 133 Proof. For the function u given we can choose a number h sufficiently large and such that the following inequality takes place |u(x,−h)| < 1 2 · sup {u (z) , z ∈ Ω}. The function u cannot assume its maximum value inside of the domain Ω. It cannot also have the maximum value on the line Γ. It means that the maximum value can be achieved only on the curve Σ. It is clear that this value is equal to one. The lemma is proved. Let us now consider symmetrization of the function u ∈ W with regard to the axis x. We can extend the function under consideration as unity into the complement to the domain Ω to the domain Π. Let us put t = x2 , y = y ν (t, y) = u (√ t, y ) , if ≥ 0,and u (√ |t|, y ) , ift < 0. Besides we put ν equal to one at the points where the function uassumes the same value. Thus we we have that the function ν is defined everywhere in the half-strip Π. Let G = {(t, y, z) : (t, y) ∈ Π, z < ν (t, y)} We have from the lemma (3.1) that G ⊂ {(t, y, z) : (t, y) ⊂ Π, z < 1} We denote as G (y0) a section of the domain G by the plane {(t, y, z) : y = y0}. It is clear that δG (y0) coincides with the graph of the functionν0 (t) := ν (t, y0). Let us now put ρ1 (y0) = inf {ν0 (t) : |t| ≤ 1}, ρ2 (y0) = sup {ν0 (t) : |t| ≤ 1}. We see that the intersection of δG (y0)with the line l (y0, ρ) = {(t, y0, ρ) : |t| ≤ 1, ρ1 ≤ ρ ≤ ρ2} consists of the even number of the points tk (ρ), 1 ≤ k ≤ 2n. Let us denote as 2 · T (ρ) the measure of the intersection of G (y0) with the line l0 (y0, ρ), 2 · T (ρ) = t2 (ρ) − t1 (ρ) + · · · + t2n (ρ) − t2n−1 (ρ) It is clear that this measure is equal to 2 when ρ < ρ1 (y0) The function T (ρ)thus defined is an increasing function. Definition 3.1. Let ν• = ν• (t, y) be the function corresponding to the function ν in the following way 134 E. Shcherbakov • ν• (−t, y) = ν• (t, y) , t ∈ [0, 1], • ν• (T (ρ)) = ρ, ρ1 (y) ≤ ρ ≤ ρ2 (y) , y ≤ y0. We call the function u• = u• (x, y) = ν• (x2, y) , |x| ≤ 1 the sym- metrization of the function u with regard to the axis x. The line Σ is the level line for the function u which means that symmetrization of the function leads to the Steiner symmetrization of the domain Ω with regard to the axis y. We denote as Ω• the symmetrization of the domain Ωand we select the notation u• for the symmetrization of the function u. Now we are going to define the symmetrization of the function u with regard to the axis x. Once again we extend this function into the half-strip Π by the rule u (x, y) = u (x,−y) ,−∞ < y ≤ 0. Let us define as G ′ the domain G with its reflection in the plane (x, y). We denote as G ′ (x0)the intersection of the domain G with the plane {x = x0}. It is clear that the boundary δG ′ (x0)is the graph of the function u0 (x0, y) , |x0| < 1.We can transform this graph in the same way as it was done with the graph of G (y0) of the function ν0.As a result we get the function u•• (t, y) whose graph will be symmetric with regard to the plane (x, z). As in preceding case the symmetrization of the function of this type leads us to the symmetrization of the domain Ω ⋃ {(x, y) |x| ≤ 1, y > 0, (x,−y) ∈ Ω} - in this case with regard to the axisx. Let us study now the behaviour of the functionalI0 under symmetrizations of the solutions of the boundary value(1.1)-(1.4) just defined. Theorem 3.1. Let u (x, y)be a solution of the problem (1.1)-(1.4)in the domain Ω and u• = u• (x, y)-its symmetrization with regard to some of its axis. Then I0 (u•, Ω•) ≤ I0 (u, Ω) (3.4) Proof. For the beginning we consider symmetrization with regard to the axis y. After substitution of the variables u, ν by the variables u•, ν• under the sign of integral I0 we arrive at the following expression I ′ 0 (u•, Ω•) = 4 · 1∫ 0 1∫ 0 t ( ν• t 2 ) · dt · dy + 1∫ 0 1∫ 0 ( ν• y 2 ) · dt · dy (3.5) Axisymetric mixed free boundary value problem 135 Now the function T = T (ρ) is monotone one for each y0 ∈ (−∞, 0). Hence the following representation takes place 1∫ 0 ( ν• y )2 (t, y0) · dt = ρ2(y0)∫ ρ1(y0)   ( δT δy )2 ∣∣∣ δT δρ ∣∣∣   (ρ, y0) · dρ (3.6) This representation is a plane consequence of the second property from the defintion 3.1(see[4]). The function u is a real analytic func- tion. This means that the graph of the function ν0 (t) is the union of the monotone curves. For each of this curves we can do calculations which leads us to the formula (3.6). Thus we get as a result the following formula 1∫ 0 (νy) 2 (t, y0) dt = ρ2(y0)∫ ρ1(y0) 2n(ρ)∑ i=1   ( δti δy )2 ∣∣∣ δti δρ ∣∣∣   (ρ, y0) dρ (3.7) Now we use the Shwartz inequality to get the following result   2·n(ρ)∑ i=1 (−1)i · δti δy   2 ≤ 2n(ρ)∑ i=1 ( δti δy )2 ∣∣∣ δti δρ ∣∣∣ ·   2n(ρ)∑ ı=1 (−1)i · δti δρ   = = 2n(ρ)∑ i=1   ( δti δy )2 ∣∣∣ δti δρ ∣∣∣   ∣∣∣∣ δT δρ ∣∣∣∣ From this inequality we easily get ρ2(y0)∫ ρ1(y0)   ( δT δy )2 ∣∣∣ δT δρ ∣∣∣   (ρ, y0) dρ = ρ2(y0)∫ ρ1(y0)   ( δti δy )2 ∣∣∣ δti δρ ∣∣∣   (ρ, y0) dρ (3.8) Comparing the expressions (3.6)-(3.8) we get the following result 0∫ −∞ 1∫ 0 ( νy •2 ) (t, y) dt · dy ≤ 0∫ −∞ 1∫ 0 ( νy •2 ) (t, y) dt · dy (3.9) 136 E. Shcherbakov We prove now that there also takes place the following inequality 0∫ −∞ 1∫ 0 ( νt •2 ) (t, y) · t dt · dy ≤ 0∫ −∞ 1∫ 0 ( νt •2 ) (t, y) · t dt · dy (3.10) To this end let us consider the following quiet evident inequality   2n(ρ)∑ i=1 (−1)i ti   2 ≤   2n(ρ)∑ i=1 ti∣∣∣ δti δρ ∣∣∣   ·   2n(ρ)∑ i=1 (−1)i δti δρ   = =   2n(ρ)∑ i=1 ti∣∣∣ δti δρ ∣∣∣   · ∣∣∣∣ δT δρ ∣∣∣∣ . We get from it the result T∣∣∣ δT δρ ∣∣∣ ≤   2n(ρ)∑ i=1 ti∣∣∣ δti δρ ∣∣∣   . It leads us to the inequality 0∫ −∞ 1∫ 0 t (νt •)2 (t, y) dt · dy ≤ 0∫ −∞ ρ2(y0)∫ ρ1(y0) T∣∣∣ δT δρ ∣∣∣ dρ dy ≤ 0∫ −∞ ρ2(y0)∫ ρ1(y0) 2n(ρ)∑ i=1 ti∣∣∣ δti δρ ∣∣∣ dρ dy = 0∫ −∞ 1∫ 0 t ( νt 2 ) (t, y) dt · dy. It is in fact the inequality (3.10). Now uniting the inequalities (3.9)- (3.10)we arrive at the result Í0 (u•, Ω•) ≤ Í0 (u, Ω) (3.11) It is now left to prove that ∫ Γ u•2 ds ≤ ∫ Γ u2 ds (3.12) Axisymetric mixed free boundary value problem 137 The function T (ρ) is a monototone one. Hence for each fixed y0,−∞ ≤ y0 ≤ 0, the function u• achieves its minimal value on on the line {x = ±1}. It means that on the setΓ the values of the function u• do not exceed the values of the function u. These arguments lead us to the conclusion that the inequality(3.12) is valid. From the inequalities(3.11)-(3.12)it now follows that under the symmetrization of the function uwith regard to the axis y the inequality (3.4)takes place. The same result also takes place for the symmetrization with regard to the axis x. This time it will be even more easy to prove the assertion as we will have now in the expression (3.12) equality instead of the inequality. The theorem is proved. 4. Dirichlet Principle. We will show now that solutions of mixed boundary value prob- lem(1.1)-(1.4) possess extremal property usually called as Dirichlet Principle. Let Ω be an admisssible domain with the curve Σ monotone in each quadrant from lower half-plane. We supppose also that this curve consists of the regular points and is symmetric with regard to the axis y. The aim of this section consists of the proof of the Dirichlet Principle for the solutions of the problem (1.1)–(1.4). This problem is singular because of infinity of the domain. In the paper [1] the principle was proved in the plane case using the method of exhausting the infinite domain with finite ones. Dealing with our case in the same way we prove the folowing result. Theorem 4.1. Let Ω be an admissible domain whose boundary arc Σ was just described.Let u be admissible function so that (u, Ω) ∈ D. Let u0 = u0 (x, y) be a solution of the problem (1.1)–(1.4) in the domainΩ. Then I0 (u0, Ω) ≤ I0 (u, Ω) . (4.1) Proof. Let Eh be the class of functions defined in Ωh with finite integral I0 assuming in the mean the boundary value equal to unity on the arc Σ and equal to zero on the line y = −h. Let us consider the extremal sequence νn (h) for the functional I0(u, Ωh) defined by the condition lim n→∞ I0 (νn (h) , Ωh) 138 E. Shcherbakov inf {I0 (u, Ωh) , u ∈ Eh} = d From parallelogram equality we get I0 (νn (h) , Ωh) + I0 (νm (h) , Ωh) 2 = I0 ( νn (h) + νm (h) 2 , Ωh ) + +I0 ( νn (h) − νm (h) 2 , Ωh ) ≥ d + I0 ( νn (h) − νm (h) 2 , Ωh ) (4.2) In accordance with the definition of the constantdwe get from (4.1) that the sequences Í0 (νn (h) , Ωh) , ´́ I0 (νn (h) , Ωh) are fundamental ones. As for the domains of the considered type the inequality of Friedrichs takes place (see [6]) than the sequence νn (h)is fundamental in the functional space W 1,2 (Ωh). We denote by νh = νh (x, y) the limit of this sequence. On the dislocation of the boundary of the domain the functions of the bounded set from the space W 1,2 (Ωh) behave themselves in equicontinuous way. It is clear that this function satisfies the equation (1.3). Passing to the limit under the sign of the integral we get also that the following condition takes place lim n→∞ 0∫ −h (νn)2 (±1, y) dy = 0∫ −h (νh) 2 (±1, y) dy. Let hn := −n, and Ωn,νn - sequences corresponding to hn. Now we are going to construct a solution ν = ν (x, y) of the problem (1.1)– (1.4)considering the sequence νnin the domain Ω = ⋃ Ωn. Let y0 be the maximal distance from the points of the curve Σ to the axis x. Let us consider the function ν∞ = ν∞ (x, y) defined as follows v∞ (x, y) = = − ∑∞ m=0 2α λm 2·J0(λm)·[λm 2+α2] · eλm·(y+y0) · J0 (λm · x) (4.3) Here J0 = J0 (t)-Bessel function of zero order and λm-solution of the equation λm · J ′ 0 (λm) + α · J0 (λm) = 0 In the half-strip Π the function v∞ (x, y)is the unique solution of the problem (1.1)–(1.4) The difference between the functions v∞ (x, y)− Axisymetric mixed free boundary value problem 139 vn (x, y) ≥ 0, |x| ≤ 1, y = y0 Taking the condiion (1.3) into account we get that the functionvn (x, y) satisfies the following condition νn (x, y) ≤ ν∞ (x, y) , (x, y) ∈ Π (y0) (4.4) Let us consider the difference νn−νm, m < n in the domain Ωh. From the condition (4.4)for the points (x, y) ∈ (y = −h) we get that the limit of this difference for mtending to infinity is equal to zero. The difference under consideration is equal to zero on the arc Σ and on the set Γ satisfies to the condition(1.3). It means that the sequenceνnis fundamental in the sense of the uniform convergence in each domain Ω (n0) = Ω ⋂ {y > −n0} , n0 ∈ N Let ν = ν (x, y) be the limit of the sequence {νn}. It is clear that it satisfies the equation (1.1). The class of the functions admissible for the domain Ωh does not diminish when nincreases. It means that the functionν = ν (x, y) satisfies the condition (1.2)at the points of the curve Σ. The inequality (4.4) means that this function satisfies also the condition (1.4)at the infinity. It is easy to prove that it also satisfies the condition (1.3)on the set Γ. The above said means that the function ν = ν (x, y) satisfies (1.1)–(1.4). It is an admissible function for the the variational problem for the functional I0 (u, Ω). In the usual way we prove that this functional achieves its minimal value on this function [2]. The theorem is proved. 5. Solution of boundary value problem. We are going to prove here that using the solutions of the problem (1.1)–(1.4) we are able to construct a solution of the main problem. To begin with we recall that the numberE0 can be selected arbitrarily large. This permits us to prove the following result. Lemma 5.1. For arbitrary non negative values of λ,κ, θthere exists a numberE0such that the values of the functional I (u, Ω) are non negative on the set of admissible pairs (u, Ω) ∈ D. Proof. Let Q (u, Ω) be the following expression Q (u, Ω) := −λ · ∫ ∫ B |x| · dx · dy + θ · ∫ Σ f (ẋ) · ds (5.1) 140 E. Shcherbakov In accordance with the formula (2.2) we get for the values of E0 sufficiently large the following inequality Q ≥ −λ · θ − θ · ρ · 1∫ 0 ( arcsin σ + σ · √ 1 − σ2 ) × × (1 − σ2) − 3 2 dσ + θ · E0 · ρ ≥ 0 (5.2) Here ρ is the maximal distance from the points of the arc Σto the axis x. It is clear that the functional I (u, Ω)−Q (u, Ω) assumes non- negative values over the set of the admissible functions. Now we are ready to prove the main result. Theorem 5.1. Let the number c0 > 0, satisfies condition κ−c0θ > 0 Then for all nonnegative α, λ there exists an admissible pair (ue, Ωe) ∈ D, such that I (ue, Ωe) = inf {I (u, Ω) , (u, Ω) ∈ D} The curve Σe corresponding to the domain Ω is piece-wise analytic and it is nondecreasing for the points with x > 0 The function ue is a solution of the main problem. Proof. Let{en} be a minmizing sequence en = (un, Ωn) for the variati- onal problem from the section 2. Without restriction we may assume that the curves Σn = δΩn ⋂ Π are piece-wise analytic ones consisting of the regular points(see [7]). The functional Q behaves monotonically under symmetrization of the functions and domains defined before. The functional I behaves itself in the same way. It means that we may assume monotone behaviour of the curves Σnin each quadrant. Using Dirichlet principle we can also assume that the functions un are the solutions of the problem (1.1)–(1.4) in the domain Ωn. In accordance with the lemma 5.1 we can asssume that for the numbers E0 sufficiently large the functional I is non negative over the set of admissible functions. It means that the set of ordinates of the points of intersection of the curves Σn with y - axis is bounded. It follows from the Helly theorem that the sequence of the curves Σn converges to the limit curve Σe. Besides we get convergence of the domains Ωn to the domain Ωe as to the kernel(see [9]). The Axisymetric mixed free boundary value problem 141 functions un are the solutions of the equation(1.4). It means that we can assume that they are traces of the harmonic functions defined over the domainsΩ• n obtained by their rotation about y-axis. From this it follows that the sequence of the functions unis compact in the sense of the uniform convergence inside of the domain Ωe. We leave for the convergent subsequence the notation of the proper sequence. Hence the sequence{un}is convergent inside of the domain Ωe. Let us denoteue := lim n→∞ un (x, y). The functions unare limited by the function ν∞. It means that the limit function ueassumes at the infinity the value equal to zero. Let us now consider the behaviour of the function ue on the set Σe ⋃ Γe. It was said earlier that the functions from the limited set of the Sobolev space are equicontinuous in the mean with regard to the shift of the boundary. The arcs of the set Σe−{x = 0} satisfy Lipshitz condition. It means that the function ue assumes in the mean the value equal to one on the arc Σ(see[6],[10]). The points of the set Σe\{x = 0} are regular ones. Hence the function ue assumes on Σe the boundary value equal to one with possibl eexeption of the points lying on the y-axis. We can assume that the functions un are extended across analytic set Γ. Passing to the limit on Γ we get that the limit function is a solution of equations(1.1)– (1.4). We will show now that the function ue satisfies almost every- where on Σe the boundary condition (1.5). The functional I (u, Ω) is semicontinuous from below on the set D (see [4],[7]), whence it follows that it attains its minimal value on the pair (ue, Ωe) ∈ D. Let z0 = y0 + i · x0, 0 < x0 < 1-be any point on the curveΣesuch that tangential line exists at this point. Let z• = z + ǫ · F be a transformation defined by the infinitely differentiable function F with support lying in the disk B (z0, δ). For the values ǫ > 0 sufficiently small the mapping z• (z) is the topological one. The Bjorke theorem (see [10]) permits us to consider the function F as the extension on B (z0, δ) of the finite infinitely differentiable function given on Σδ e = Σe ∩B (z0, δ). The necessary condition for the element (ue, Ωe) 142 E. Shcherbakov to be extremal can be written in the form 0 = |Σ(ǫ)|∫ 0 [( 2 · i · λ · x · ¯̇z + 4 · i · u2 ezx − κ i ) F ] ds − |Σ(ǫ)|∫ 0 [ (κ · x · ¯̇z − 2 · i · θ · fẋ · +ẏ · ¯̇z + 2 · θ · f · ¯̇z) dF ds ] ds (5.3) Here z = y + i · x, uez = 2−1 · ( δue δy − i δue δx ) and |σ (ǫ)| is the length of the curve Σǫ e = Σe ⋂ {|z − z0| < ǫ} . The condition (see(5.2)), is the condition of existence of generalized deri- vative for the function κ ·x · ¯̇z−2 · i ·θ · fẋ · ẏ · ¯̇z+2 ·θ · f · ¯̇z ∈ L2 ([0, |Σǫ|]). Let Φ1 = Φ1 (s) , Φ2 = Φ2 (s) are the following functions Φ1 (s) = κ · x · ẏ − 2θ · fẋ · ẋ · ẏ + 2θ · f · ẏ, (5.4) Φ2 (s) = −κ · x · ẋ − 2θ · fẋ · ẏ 2 − 2θ · f · ẋ (5.5) From above said it follows that they are absolutely continuous. The following equations are the consequences of the conditions G (ẏ, Φ1, Φ2, y) = 0, G = ẏ ·Φ1 − √ 1 − ẏ2 ·Φ2 − 2 · θ · f − κ · y (5.6) H (ẋ, Φ1, Φ2) = 0, H = ẋ ·Φ1+ √ 1 − ẋ2 ·Φ2+2 ·θ · fẋ · √ 1 − ẋ2, (5.7) It is easy to show that these equations are solvable in ẋ, ẏ for the values of the parameterE0 sufficiently large. We get also that ẋ = Ψ1 (Φ1, Φ2), ẏ = Ψ (Φ1Φ2) for some differentiable functions Ψ1, Ψ2. The functions Φ1, Φ2, y are absolutely continuous as the functions of the parameter s of natural parametrization of the curve Σe. It means that the functions ẋ, ẏ are differentiable almost everywhere on Σǫ e. Taking into account the above-mentioned results we get from (5.3) the equality −i · κ · ż · x · k (z) + κ · ẋ · ¯̇z + 2 · ı θ · ẍ · ¯̇z = (5.8) −2 · i · λ · x · ¯̇z − 4 · i · ( δue δz )2 · ż + κ i (5.9) Axisymetric mixed free boundary value problem 143 Here k (z)-curvature of the curve Σǫ e. In the axially symmetric case we have 2 · H (z) = k (z) + ( δue δz ) · (x · |∇ue|) −1 (5.10) K (z) = − ẍ x . (5.11) Using (5.6)-(5.8) we get that the function ue satisfies almost every- where on the curve Σe boundary condition (1.5). Using (5.6)–(5.8) we get that the function ue satisfies almost everywhere on the curve Σe boundary condition (1.5). Using a‘priory estimates for |∇ue| from the paper [2] we get that the curve Σe ⋂ {x > 0} is a Liapunov curve. From Shauder estimates it follows that this curve is infinitely differentiable one. In the usual way (see [4], [7]) we prove that the curve Σ consists of the analytic arcs. The theorem is proved. 6. Free boundary. In this part we investigate the contact of free boundary with the lines {x = ±1} and {y = 0}. Let us begin with the set { Σ̄ ∩ {x = 1} } . We shall prove for the first the following lemma which is the simple generalization of the result proved in ([1]). Lemma 6.1 Let b be the length of the segment Σ̄ ∩ {x = 1} and {un}-the sequence of the functions from extremal sequence {un, Ωn}. Then lim n→∞ 0∫ −b un (1, y)dy = b (6.1) Proof. Let us consider the parts Σn = δΩn ∩Π lying in the half space {y > −b + ǫ0}. Then from assumption of the lemma it follows that they are contained in the rectangular Πǫ = (−b + ǫ0 < y < 0 ) × (1 − ǫ, 1) , 0 < ǫ < 1, 0 < ǫ0 < b 144 E. Shcherbakov Let us compare the functions un, n > n (ǫ, ǫ0), in the rectangular Πǫwith the functionuǫwhich is a solution of (1.1) and satisfies the following conditions uǫ (x, b − ǫ) = 0 = uǫ (x, 0) , 1 − ǫ < x < 1 (6.2) uǫ (1 − ǫ, y) = 1,−b + ǫ0 < y < 0 (6.3) duε dx + α · uε (1, y) = 0,−b + ε0 < y < 0 (6.4) The function uεcan be represented in the following form uε (x, y) = ∑ m≥0 Am · Vm (x, ε) · sinm π −b + ε0 · y (6.5) Vm (x, y) = − α m · Y 1 m (x, ε) + Y2 m (x, ε) (6.6) y′′ + x−1 · y′−m2 · y = 0 (6.7) Y 1 m (1, ε) = 0, ( Y 1 m )′ (1, ε) = 1 (6.8) Y 2 m (1, ε) = 1, ( Y 2 m )′ (1, ε) = 0 (6.9) The functions Y k m, k = 1, 2, are constructed as linear combinations of the functions l0 (mx) , K0 (mx), constituting the fundamental system for equation (6.6) (see [11]). It is can be easily verified that Y 1 m (x) = −K0 (m) · l0 (m · x) + l0 (m) · K0 (m · x) m · l0 (m) · K ′ 0 (m) − l′0 (m) · K0 (m) (6.10) Y 2 m (x) = −K ′ 0 (m) · l0 (m · x) − l0 (m) · K0 (m · x) l0 (m) · K ′ 0 (m) − l′0 (m) · K0 (m) (6.11) Using integral representations for the functions l0 (mx) , K0 (mx) (see [11]) we obtain the following estimate for the functions Vm Vm (1 − ε, ε0) > m 4 · eε·m 2 (6.12) The functions uε from (6.5) evidently satisfy the condition (6.2). Let us take coefficients Am in the form Am = 4 π · m · Vm (1 − ε, ε0) , Axisymetric mixed free boundary value problem 145 Than we get that the function uεsatisfies also the condition (6.3). The estimate (6.12) is not exact but it permits us to differentiate the series (6.5)by terms. This means that the function uε satisfies the condition (1.1). Thus we have that the functionuε is a solution of the problem (1.1),(6.2)–(6.4). Now we evidently have 0∫ −b+ǫ0 uǫ (1, y) dy = m0∫ −b+ǫ0 0 ∑ m=0 1 4 · π · Vm (1 − ǫ, ǫ0) · sin π −b + ǫ0 y · dy+ + ∑ m≥m0 4 (−b + ǫ0) π · m2 · Vm (1 − ǫ, ǫ0) · [1 − (−1)m] For a given positive number ǫ0 we can choose the numbers m0, n (ǫ0) sufficiently large and a positive number ǫ sufficiently small such that for n > n (ǫ0) we get the inequality 0 ≤ − 0∫ −b+ǫ0 un (1, y) · dy + b ≤ 2 · ǫ0 Tending n to infinity we get the result we need. The lemma is proved. Theorem 6.1. Let (u, Ω) be a solution of variational problem from the point 2. Let us assume that the inequalityλ < α takes place.Then for the arc Σeof the boundary Ωeconnecting the points (−1, 0) , (1, 0)we have Σ̄e ∩ {x = ±1} = {(−1, 0) , (1, 0)} (6.13) Proof. Let us suppose that the theorem is not correct. In this case the number b from the precedent lemma is positive. Using Green theorem we get ∫ ∫ Ωe |x| · |∇u|2 dx dy + α · ∫ Γ u2 e dy = α · ∫ Γ ue dy (6.14) Let us consider the pair (u•, Ω•) for which u• (x, y) = u (x, y − b)and Ω•is the domain obtained by the dislocation by the number b of the 146 E. Shcherbakov domain Ω in the positive direction of the axis y. The pair (u•, Ω•), evidently, belongs to the set D. We denote as Σ• the part of the boundary of Ω• lying in the closure of the half-strip Π and connecting the points (−1, 0) , (1, 0) . Taking (6.13) into account we get I0 (u•, Ω•) = I0 (u, Ω) + 2 · b · (λ − α) − b2 2 (6.15) ∫ Σ• |x| · ds < ∫ Σ |x| · ds (6.16) |l•|∫ 0 f (ẋ) · ds = |l|∫ 0 f (ẋ) · ds (6.17) Here l,l•are the lengths of the curves Σ,Σ•respectively. From the conditions (6.13)–(6.16) it follows that I (u•, Ω•) < I (u, Ω) . The theorem is proved. We will show now that the curve Σe = ∂Ω ⋂ Π̄ for extremal pair (ue, Ωe) does not have points on the line {y = 0} for the exception of its end points. Lemma 6.2. Let u0 = u0 (x, y) be a solution of the problem (1.1)- (1.4) in the half strip Π. Then there exist a number М such that for a sufficiently small neighbourhood of the point (1, 0) the following inequality takes place u0 y (x, y) ≥ M · ln [ 1 + 1 α · (1 − x) ] (6.18) Proof. Without any difficulties we get for the function u0 the following representation u0 (x, y) = 1 + 2 · α π · 0∫ −∞ I0 (µ · x) · sin µ · y µ · [µ · I1 (µ) + α · I0 (µ)] rmdµ (6.19) Axisymetric mixed free boundary value problem 147 Here I0, I1-Bessel functions of purely imaginary argument. The pos- sibility of such representation follows from asymptotic estimate (see [12]) In (µ) = eµ √ 2π · µ · [ 1 + O ( µ −1 2 )] , µ → ∞ (6.20) It follows from (6.19) that |u0 y(x, 0)| > 2α π · N1∫ 0 I0 (µ · x) [µ · I ′ 0 (µ) + α · I0 (µ)] dµ + 3−1 · (1−x)−1)∫ N1 eµ·(x−1) µ + α dµ (6.21) The number N1 in the condition (6.21) is selected in such way that for the function O ( µ −1 2 ) from (6.20) takes place inequality ∣∣∣O ( µ −1 2 )∣∣∣ < 1 2 . From inequality (6.21) we now get |u0 y(x, 0)| > 2α π · N1∫ 0 I0(µx) [µI1(µ) + αI0(µ)] dµ +3−1e−1 · ln [ 1 1 − x + α ] − 3−1e−1 · ln [ 1 N1 + α ] (6.22) The inequality (6.22) implies the result we need. The lemma is proved Lemma 6.3. Let (u, Ω) be a solution of variational problem from section 2 such that Σ ∩ (−1, 1) 6= ∅. Then the function uy tends to infinity logarithmically when x tends to unit. Proof. Let us extend the function u as unity to the half strip Π. Let us consider the function w := u− u0 in the half strip Π. We can extend the function w as an odd function across the axis x. From assumption of the lemma it follows that there exists a positive number such that 148 E. Shcherbakov Σ ∩ (0, 1) = [c, 1) , c < 1. We consider now the reduction of the extended function w to the set Πc = {(x, y) |0 < c < 1 ,−∞ < y < ∞} . Let us consider sine transform ws (c, µ) := 0∫ −∞ w (c, y) · sin µ · ydy of the function w (c, y). We can easily verify that for the function w in the strip Πc takes place the following representation. w (x, y) = 2 π · 0∫ −∞ ws (c, µ)× ×µ·I0[2µ·(1−x)]+α·I1[2µ·(1−x)] µ·I0[2µ·(1−c)]+α·I1[2µ·(1−c)] · sin µ · ydµ (6.23) From the representation (6.23) it follows that that the function wy (x, 0) is bounded in the neighbourhood of the point (1, 0). It means that the function uy (x, o) jointly with the function u0 y (x, 0) logarithmically tends to the infinity. The lemma is proved. Theorem 6.2. Let (ue, Ωe) be a solution of variational problem from section 2. Let us suppose that for the numbers λ, α takes place inequali- ty λ2 < α. Then the number c, [c, 1) = Σe ∩ (0, 1), is equal to unit. Proof. From the theorem 5.1 it follows that the curve Σe consists of the analytic curves. Let us consider variation δI under local variation δx̃ of the boundary δI (Ωe, ue, δ~x, δu) = ∫ Σe [ λ − |∇ue| 2] · δ~x · ~ν · ds+ +κ · δ ∫ Σe |x| · ds + θ · δ ∫ Σe f (ẋ) · ds (6.24) We can select as δx̃ the function whose graph for x > 0 represents a step with its basis of the length ∣∣∣ln −1 2 (1 − x0) ∣∣∣ Axisymetric mixed free boundary value problem 149 centered at the point x0 and of the same height. In this case we haveh δ ∫ Σe |x| ds + δ ∫ Σe f (ẋ) ds = O (∣∣∣ln −1 2 (1 − x0) ∣∣∣ ) , x0 → 1 (6.25) From lemma 6.3 we get ∫ Σe [ λ2 − |∇ue| 2]·δx̃·~ν · ds x +κ·δ ∫ Σe |x|· ds+θ·δ ∫ Σe f (ẋ)· ds < 0 (6.26) It means that δI (Ωe, ue,δx̃,δu) < 0 for the points x0 sufficiently near to one. We denote by the letter Ω̃ the domain obtained from Ωe with the help of displacement δx̃. Let ũ be a solution of boundary problem (1.1)-(1.4) in the domain Ω̃. On the basis of inequality (6.26) and Dirichlet principle we now get I ( ũ, Ω̃ ) < I (ue, Ωe) (6.27) Let us symmetrize the function ũ and domain Ω̃ in order to axis y. We denote as Ũ the symmetrization of the function ũ and through Ω• the symmetrization of the domain Ω̃ in regard to the axis y. Then x0∫ −∞ U2 (1, y) · dy ≤ 0∫ −∞ ũ2 (1, y) · dy (6.28) Consider now the function Ú (x, y) = U (x, y + h) and domain Ώ obtained from domain Ω• by dislocation in the negative direction of the axis ydefined by the numberh. The pair ( ú, Ώ ) is admissible for the extremal problem from the part 2. Using inequalities (6.27)- (6.28)we arrive at the condition I ( Ú , Ώ ) < I (ue, Ωe) + 2 · (α − λ) · ln −1 2 |(1 − x0)| (6.29) The function [ λ − |∇ue| 2] · δx̃ is negative and is of the order ln− 1 2 |(1 − x0)| when x0 tends to 1. The length of the supporter of δx̃ 150 E. Shcherbakov is equal to ln− 1 2 |(1 − x0)|. It means that for the points x0 sufficiently near the unity we get on the basis (6.29) the inequality I ( Ú , Ώ ) < I (ue, Ωe) (6.30) As it was already said the pair ( Ú Ώ ) is admissible. Thus the inequali- ty (6.30)means that the assumption c < 1 leads us to the contradiction. The theorem is proved. 1. Bazalii B.V. and Shelepov V.Yu. Variational Methods in a Mixed Problem of Thermal Equilibrium with a Free Boundary // Amer. Math. Soc. Transl., (2)126, (1985), 77-92 . 2. Gilbarg D. and Trudinger N.S. Elliptic Partial Differential Equations of Second Order, Springer Verlag, Berlin, Heidelberg, 1983. 3. Polya G.and Szego G. Isoperimetric Inequalities in Mathematical Physics, Princeton, Univ.Press,1951. 4. Garabedian P.R., Lewy H. and Shiffer M. Axially Symmetric Сavitational Flow // Annals of Mathematics, 56(1952), 560-602. 5. Garabedian P.R. Partial Differential Equations, New York-London-Sydney, J.Willey, Inc., 1964. 6. Michlin S.G. Linear Partial Differential Equations, Moscow, Vischaya Shkola, 1977(in Russian). 7. Chthterbakov E. Free Boundary Value Problem for Axisymmetric Fluid‘s Flow with Surface Tension and Wedging Forces, // Zeitschrift fur Analysis und ihre Anwendungen,17(4), (1998), 937-961. 8. Kolmogorov A.N. and Fomin S.V. Elements of the Functional Analysis and Theory of Functions, Moscow, Izdatelstvo Fiziko-matematicheskoi Literatury, 1972 (in Russian). 9. Suvorov G.D. Families of Topological Mappings, Novosibirsk,1965 (in Russian). 10. Kufner J., John O. and Fucik S. Function Spaces, Prague, Academia, 1977. 11. Tichonov A.N. and Samarskii A.A. Equations of Mathematical Physics, Mos- cow, Nauka, 1966 (in Russian). 12. Zorich V.A. Mathematical Analysis, v. 2, Moscow, 2002 (in Russian). Kuban State University 149 Stavropolskaya st., 350040, Krasnodar, Russian Federation Received 1.03.07
id nasplib_isofts_kiev_ua-123456789-10114
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
issn 0236-0497
language English
last_indexed 2025-12-02T02:07:46Z
publishDate 2007
publisher Інститут прикладної математики і механіки НАН України
record_format dspace
spelling Shcherbakov, E.
2010-07-23T14:25:06Z
2010-07-23T14:25:06Z
2007
Axisymmetric mixed free boundary value problem / E. Shcherbakov // Нелинейные граничные задачи. — 2007. — Т. 17. — С. 130-150. — Бібліогр.: 12 назв. — англ.
0236-0497
https://nasplib.isofts.kiev.ua/handle/123456789/10114
Mixed Free Boundary Value Problem for Laplace equation in axisymmetrical case is considered. We take into consideration mean and Gauss curvatures of the free boundary. The problem of this type arise on investigating thermal equillibrium of two phases. We take into account capillary forces acting in intermediate layer separating different phases. Plane model of the equillibrium without capillary forces was considered in the paper([1]). We consider variational problem whose solution is generalized solution of the boundary value problem.We prove regularity of solution, analyticity of free boundary and investigate its properties.
en
Інститут прикладної математики і механіки НАН України
Axisymmetric mixed free boundary value problem
Article
published earlier
spellingShingle Axisymmetric mixed free boundary value problem
Shcherbakov, E.
title Axisymmetric mixed free boundary value problem
title_full Axisymmetric mixed free boundary value problem
title_fullStr Axisymmetric mixed free boundary value problem
title_full_unstemmed Axisymmetric mixed free boundary value problem
title_short Axisymmetric mixed free boundary value problem
title_sort axisymmetric mixed free boundary value problem
url https://nasplib.isofts.kiev.ua/handle/123456789/10114
work_keys_str_mv AT shcherbakove axisymmetricmixedfreeboundaryvalueproblem