A Singularly Perturbed Convection Diffusion Turning Point Problem with an Interior Layer

Size: px
Start display at page:

Download "A Singularly Perturbed Convection Diffusion Turning Point Problem with an Interior Layer"

Transcription

1 A Singularly Perturbed Convection Diffusion Turning Point Problem with an Interior Layer E. O Riordan, J. Quinn School of Mathematical Sciences, Dublin City University, Glasnevin, Dublin 9, Ireland. Abstract A linear singularly perturbed interior turning point problem with a continuous convection coefficient is examined in this paper. Parameter uniform numerical methods composed of monotone finite difference operators and piecewise-uniform Shishkin meshes, are constructed and analysed for this class of problems. A refined Shishkin mesh is placed around the location of the interior layer and we consider disrupting the centre point of this fine mesh away from the point where the convection coefficient is zero. Numerical results are presented to illustrate the theoretical parameter-uniform error bounds established. Keywords: Singularly Perturbed; Shishkin mesh; Interior Turning Point 1 Introduction Interior layers exhibiting a hyperbolic tangent profile can arise in solutions of singularly perturbed quasilinear problems of the form εu uu b(x)u = f(x), x (0, 1), b 0, u(0) > 0, u(1) < 0; (1.1) when the singular perturbation parameter ε > 0 can be arbitrarily small. Asymptotic expansions can be used to locate the interior point p, where u(p) = 0, to within an O(ε) neighbourhood of some known point d. That is, p (d Cε, d + Cε) (see Howes [5]). In this paper, we examine numerical methods for a class of linear problems associated with the above nonlinear problem. For example, problems of the form εy 2 tanh( d x 2ε )y y = f, x (0, 1), y(0) > 0 > y(1), d (0, 1) (1.2) will be studied. For such a problem, an interior layer forms about the interior point d. A parameter-uniform numerical method [1], based on an upwind finite difference operator and a piecewise-uniform Shishkin mesh ([1]), will be constructed and analysed in this paper for these type of problems. We will also consider the effect of centring the piecewise-uniform mesh at some point d N (d Cε ln N, d + Cε ln N). The resulting numerical analysis may prove useful in any future examination of the nonlinear problem (1.1), where the point p is only known to lie within some interval (d Cε, d + Cε). Singularly perturbed turning point problems of the form εy a(x)y + b(x)y = f(x), x (0, 1), a(d) = 0, d (0, 1), b > 0 This research was supported by the Irish Research Council for Science, Engineering and Technology 1

2 have been studied by several authors (Farrell [3], Berger et al.[2]). In this case, where the convective coefficient a is independent of ε, then the nature of any interior layer is different to the problem being considered in the current paper. Depending on the quantity b(d)/a (d), there may be no interior layer or a power-law layer in the vicinity of the point d. Also, the restriction a (x) 1 2 a (d), x (0, 1) is placed on this problem, whereas in (1.2), the convective coefficient ã := 2 tanh( d x 2ε ) satisfies ã (x) ã (d), x (0, 1). Exponential interior layers can be generated by considering linear problems with discontinuous coefficients of the form: Find y C 1 (0, 1) such that εy a(x)y b(x)y = f(x), x (0, d) (d, 1) a(x) α > 0, x < d, a(x) α < 0, a(d + ) a(d ). Interior layers of exponential type form in the vicinity of the point of discontinuity in the coefficient a (Farrell et al. []). The numerical analysis associated with such problems relies heavily on the fact that the coefficient a is strictly bounded away from zero in the interval (0, d) (d, 1). Finally, in [6], we examined a boundary turning point problem of the form: Find y C 1 (0, 1) such that εy + a(x)y = f(x), x (0, 1), y(0), y(1) given. The presence of the coefficient a ε (x) C(1 e αx/ε ), a ε (0) = 0, generated a boundary layer of exponential type in the vicinity of the boundary point x = 0. For this problem, the turning point at x = 0. In the current paper, the point where the convection coefficient vanishes is in the interior of the domain. As a consequence, the analysis for the current problem is significantly different to the analysis presented in [6]. In particular, the discrete error analysis is more intricate. In 2 we state the class of problems examined in this paper and derive a priori bounds on the derivatives of the solutions. In 3, we construct and analyse a set of numerical methods for this class of problems. The numerical methods consist of an upwind finite difference operator on piecewise uniform meshes, which are fine in the vicinity of the interior layer. In the final section, some numerical results are presented to illustrate the theoretical error bounds. Note that throughout the paper, the notation f (k) denotes the k-th derivative of f, C denotes a generic constant that is independent of ε and N, and. denotes the maximum pointwise norm. 2 Continuous Problem Consider the following problem class on the unit interval Ω = (0, 1). Find y ε such that L ε y ε (x) := (εy ε a ε y ε by ε )(x) = f(x), x Ω, y ε (0) = y 0 > 0, y ε (1) = y 1 < 0, a ε, b, f C 2 ((0, 1) \ {d}) C 0 [0, 1], b(x) 0, x Ω, (P ε ) a ε (x) > 0 for x [0, d), a ε (d) = 0, a ε (x) < 0 for x (d, 1]. We will show that the solution to (P ε ) exhibits an interior layer in a neighbourhood of the point d. Additional restrictions on the function a ε are listed in (2.1) below. Assumptions on the coefficient a ε in (P ε ) 2

3 Denote Ω := (0, d) and Ω + := (d, 1). Define the limiting functions a 0 and a+ 0 as a 0 (x) := lim ε 0 a ε (x), x [0, d) and a + 0 (x) := lim ε 0 a ε (x), x (d, 1] and a ± 0 (d) := lim x d ± a± 0 (x). Assume the following conditions on a ε ; a ε (x) > α ε (x), x d, α ε (x) := θ tanh (r(d x)/ε), θ > 2r > 0, x Ω, (2.1a) x a ε (t) dt C, x Ω, (2.1b) t=0 ϕ ± ε (x) := (a ± 0 a ε)(x) satisfies ϕ ± ε (x) ϕ ± ε (d) e ± θ 2ε (d x), x Ω ±. (2.1c) Note that (2.1a) implies a 0 (x) θ, x Ω and a + 0 (x) θ, x Ω +. The differential operator L ε defined in problem (P ε ) satisfies the following minimum principle. Theorem 2.1. Let L ε be the differential operator defined in (P ε ) and z C 2 (Ω) C 0 ( Ω). If min {z(0), z(1)} 0 and L ε z(x) 0 for x Ω, then z(x) 0 for all x Ω. Proof. The proof is by contradiction. Assume that there exist a point p Ω such that z(p) < 0. d t=x αε(t)dt It follows from the hypotheses that p / {0, 1}. Define the auxiliary function u = ze 1 2ε and note that u(p) < 0. Choose q Ω such that u(q) = min u(x) < 0. Therefore, from the definition of q, we have u (q) = 0 and u (q) 0. But then, using αε 2 > 2εα ε, we have L ε z(q) = [εu + 1 ε (α ε(2a ε α ε ) 2εα ε + εb)( u)](q)e 1 2ε which is a contradiction. d t=x αε(t)dt > 0 Lemma 2.1. Assuming (2.1), there exists a unique solution, y ε, of (P ε ) such that y (k) ε (x) Cε k, k = 0, 1, 2, x Ω. Proof. Define the two barrier functions ψ ± (x) := f 2θr [(x d)α ε(x) + θ] + max { y 0, y 1 } ± y ε (x), which are nonnegative at x = 0, 1. From (2.1) we have and εα ε a ε α ε { εα ε α ε α ε = ( 2r θ 1)α εα ε { 0, x d, 0, x d,, (2.2a) 2εα ε a ε α ε 2εα ε α 2 ε = ( 2r θ 1)α2 ε 2θr 2θr, x Ω. (2.2b) Using the definition of the problem (P ε ) and the inequalities (2.2), we can deduce that L ε ψ ± f [(x d)(εα ε a ε α 2θr ε) + (2εα ε a ε α ε )] + f 0. Hence, using Theorem 2.1, we have y ε C on Ω. We now bound the derivatives of y ε. Using (P ε ) and the assumptions in (2.1) we have for any x Ω; x x εy ε (t) dt = εy ε(x) εy ε(0) = (f + by ε + a ε y ε)(t) dt, (2.3a) t=0 x t=0 (a ε y ε)(t) dt [a εy ε ] x 0 + y ε x t=0 t=0 x t=0 (by ε + f)(t) dt b y ε + f. a ε(t) dt C, (2.3b) (2.3c) 3

4 By the Mean Value Theorem z (0, ε) s.t. ε y ε(z) 2 y ε. (2.3d) Using (2.3) with x = z we obtain ( ε y ε(0) f b + 2 a ε + Then for any x [0, 1] we have ε y ε(x) 2 f + z t=0 ) a ε(t) dt y ε C. ( ( x )) b + 2 a ε + a ε(t) dt y ε. t=0 Hence y ε Cε 1 and use the differential equation in (P ε ) to bound y ε. We split the problem (P ε ) into left and right problems around x = d as follows. Define y L (x), x d, y ε (x) = y R (x), x d, where the left and right problems are defined by L ε y L (x) = f(x), x Ω, y L (0) = y 0, y L (d) = y ε (d), (P L) ; L ε y R (x) = f(x), x Ω + y R (d) = y ε (d), y R (1) = y 1, (P R). From [6], we have the existence of a unique y L and y R s.t. y (k) L\R Cε k, k = 0, 1, 2. We decompose each y L\R into the sum of a regular component, v L\R, and a layer component, w L\R. If a ε satisfied the bound a ε C > 0 for all x (0, d), then, as in [], we would simply define the left regular component as the solution of L ε v L = f with suitable boundary conditions and the left layer component as the solution of L ε w L = 0, w L (0) = 0, w L (1) = (y ε v L )(d). Since a ε does not satisfy such a lower bound, we study the problem L v L (x) := εv L + a 0 v L bv L = f(x), x Ω, v L (0) = y 0, v L (d) = (v 0 + εv 1 )(d), (2.a) where v L = v 0 + εv 1 + ε 2 v 2 and the subcomponents v 0, v 1, v 2 satisfy (a 0 v 0 + bv 0 )(x) = f(x), x (0, d], v 0 (0) = y 0, (2.b) { (a v 0 v 1 + bv 1 )(x) = 0 } (x), x (0, d) lim t d v 0 (t), x = d, v 1 (0) = 0, (2.c) L v 2 (x) = v 1(x), x Ω, v 2 (0) = v 2 (d) = 0. (2.d) Note that in problem (2.a), the coefficient a ε, of the first derivative term, has been replaced by the strictly positive a 0 defined in (2.1). We incorporate the error (L ε L )v L into the layer component w ε, which is, noting (2.1c), defined as the solution of L ε w L (x) = ϕ ε v L(x), x Ω, w L (0) = 0, w L (d) = (y ε v L )(d). (2.e) Similarly v R and w R satisfy L + v R (x) := εv R + a + 0 v R bv R = f(x), x Ω +, v R (d) = v, v R (1) = y 1, (2.5a)

5 L ε w R (x) = ϕ + ε v R(x), x Ω +, w R (d) = (y ε v R )(d), w R (1) = 0, (2.5b) where v is chosen in an analogous fashion to v L (d) so that we may bound the derivatives of v R appropriately. Lemma 2.2. If v L and w L are the solutions of (2.) and v R and w R are the solutions of (2.5) then for k = 0, 1, 2, 3, we have the following bounds on the derivatives of v L\R and w L\R : v (k) L (x) C(1 + ε2 k ) and w (k) L (x) Cε k e θ 2ε (d x), x Ω, v (k) R (x) C(1 + ε2 k ) and w (k) R (x) Cε k e θ 2ε (x d), x Ω +, where θ is given in (2.1a). Proof. Analysis of problems (2.a) can be carried out effectively in the same manner as [1, 3.3] to establish the bounds v (k) 0, v(k) 1 Consider the barrier functions ψ ± (x) = me 1 2ε C and v(k) 2 Cε k, k = 0, 1, 2, 3. (2.6) d t=x αε(t)dt ± w L (x), x Ω, m := max { w L (d), 2 r v L ε }, which are nonnegative at x = 0, 1. Using (2.1), (2.6) we have L ε ψ ± (x) mθr 2ε e θ 2ε (d x) + δ ε (0) v L e θ 2ε (d x) 0. Note that d t=x α ε(t) dt= εθ r ln cosh ( r ε (d x)). Thus we can use (2.1) and cosh t 1 2 et, t > 0, to show that e θ 2ε (d x) e 1 2ε d t=x αε(t)dt 2 θ 2r e θ 2ε (d x). (2.7) Hence using the minimum principle in Theorem 2.1 we obtain w L (x) Ce θ 2ε (d x). The bounds on the derivatives of w L are established as in [6]. Analysis of v R and w R is performed in the same manner. 3 The Discrete Problem and Error Analysis Given the bounds in Lemma 2.2 on the solution, it is natural to refine the mesh in the vicinity of the point d. We examine such a mesh below. In addition, we consider the effect of centring the mesh at some other point d N near the point d. Consider the following finite difference method. Find Y ε such that L N ε Y ε (x i ) := (εδ 2 a ε D b)y ε (x i ) = f(x i ), x i Ω N ε, Y ε (0) = y 0, Y ε (1) = y 1, { D D := if a ε (x i ) 0 D + if a ε (x i ) < 0, δ2 Z(x i ) := D+ (Z(x i ) Z(x i 1 )), h i := x i x i 1, (h i+1 + h i )/2 (P N ε ) where D ± are the standard forward and backward finite difference operators. We define the 5

6 piecewise uniform mesh, Ω N ε, with a refined mesh centred at d N by d N d < pσ, p 1 2, µ = 1 2 min {dn, 1 d N }, σ := min {µ, ε r ln N}, (3.1a) Ω N ε := x i H 0 := N (dn σ), h := N σ, H 1 := N (1 dn σ), (3.1b) x i = H 0 i, 0 i N, x i = x N + h(i N ), N < i 3N,, Ω N ε := Ω N ε \ {x 0, x N }. (3.1c) x i = x 3N + H 1 (i 3N ), 3N < i N. Note, if d N = d then set p = 0. If d N d, then the mesh is not aligned to the point d. To prove the existence of a discrete solution to (P N ε ), we construct discrete analogues of the barrier functions used in Lemma 2.1. However, in place of α ε in these barrier functions, we will construct and use a discrete analogue A ε. We identify the nearest mesh point to the left of d in this non-aligned mesh as x Q. That is Define the following mesh functions x Q := max {x i x i d}. (3.2) S(x i ) := (1 + ρ) Q i (1 ρ) Q i, C(x i ) := (1 + ρ) Q i + (1 ρ) Q i, (3.3a) T (x i ) := S(x i) C(x i ), ρ = rhε 1, 0 i N. (3.3b) The mesh functions S, C and T can be thought of as discrete analogues of the 2 sinh, 2 cosh and tanh functions centred at x Q respectfully. Note that from (3.1), we have ρ CN 1 ln N 0 as N ε. Some properties of these mesh functions are given in Lemma A.1 in Appendix A. Define the operators D h Z(x i) := 1 h (Z(x i) Z(x i 1 )), D + h := D h Z(x i+1), δ 2 h := 1 h [(D+ h D h )Z(x i)]. (3.) Observe that in the fine mesh; D h = D and in the coarse mesh; D h D. The purpose of defining such operators is for convenience. We present identities and inequalities that result when these operators are applied to the mesh function T in Lemma A.2 in Appendix A. We now present the discrete function A ε which will replace α ε in discrete analogues of the barrier functions in Lemma 2.1. We distinguish between the cases σ < µ (σ Cε ln N) and σ = µ (σ C). When σ < µ, we define A ε as follows A ε (x i ) := A ε (x N ) + (x N A ε (x 3N x i ) 5r 2µ N (1 2p), 0 i < N, θt (x i ), ) (x i x 3N ) 5r 2µ N (1 2p) 3N, N i 3N, < i N. (3.5a) When σ = µ, we consider the case µ = 1 2 (1 dn ) (where H 0 h = H 1 ) and define A ε as follows A ε (x N ) + H 0 h θ(t (x i) T (x N )), 0 i < N, A ε (x i ) := (3.5b) N θt (x i ), i N. 6

7 When µ = dn 2 (where H 0 = h H 1 ) we can define A ε in the same manner. We prove convergence of A ε to α ε in the following lemma, the proof of which is given in Appendix B. Lemma 3.1. For sufficiently large N, the function α ε defined in (2.1) and the mesh function A ε defined in (3.5) satisfies: (α ε A ε )(x i ) CN 1 (ln N) 2 + CN (1 2p). By the assumption (2.1a), and assuming N is sufficiently large, we have The finite difference operator L N ε a ε (x i ) A ε (x i ), x i Ω N ε. (3.6) satisfies the following discrete minimum principle. Theorem 3.1. Let L N ε be the difference operator defined in (Pε N ) and Z be a mesh function on Ω N ε. If min {Z(x 0 ), Z(x N )} 0 and L N ε Z(x i ) 0 for x i Ω N ε, then Z(x i ) 0 for all x i Ω N ε. Proof. Suppose that Z(x q ) = min i Z(x i ) < 0. If follows from the hypotheses that q / {0, N} and x q Ω N ε. Since Z(x q ) is the minimum value we have D Z(x q ) 0, D + Z(x q ) 0 and δ 2 Z(x q ) 0. If x q < d, to avoid a contradiction we must have L N ε Z(x q ) 0 but if b(x q ) > 0 then L N ε Z(x q ) > 0 or if b(x q ) = 0 then Z(x q 1 ) = Z(x q ) = Z(x q+1 ) < 0. Repeat this argument will eventually lead us to conclude that either L N ε Z(x q ) > 0 or Z(x 0 ) < 0 for x q < d which is a contradiction. If x q > d the same argument leads to a contradiction and finally a contradiction can also be established for x q = d. Lemma 3.2. There exists a unique solution, Y ε, of (P N ε ) such that Y ε (x i ) C, x i Ω N ε. Proof. We first establish the following list of inequalities; εδ 2 A ε (x i ) A ε (x i )D A ε (x i ) 0, 0 < i < Q, (3.7a) A ε (x i )A ε (x i 1 ) A ε (x i+1 )A ε (x i ) εδ 2 A ε (x i ) A ε (x i )D + A ε (x i ) 0, Q < i < N, (3.7b) 2ε h i+1 + h i (h i D + h i+1 D + )A ε (x i ) θr, 0 < i Q, (3.7c) 2ε h i+1 + h i (h i D + h i+1 D + )A ε (x i ) θr, Q < i < N, (3.7d) which are discrete counterparts to the inequalities (2.2). For convenience in this proof, we write A ε (x i ) = A i and h i+1 + h i = Σh. Recall from the assumption (2.1) that θ > 2r. We first consider the case where σ < µ. Using (3.5a) we can show that D A ε (x i ) = { 5r 2µ N (1 2p), 0 < i N, 3N < i N, θd h T i, N < i 3N } (3.8a) and δ 2 A ε (x i ) = 0, max {i, N i} < N, 2(D + D )(A i ) h+h 0,1, i = N, 3N, θδ 2 h T i, 3N < i < N. (3.8b) For (3.7a), the result is trivial for 0 < i < N. For i = N, using Lemma A.2 and (3.8), for sufficiently large N, we have εδ 2 A i A i D A i = ( Nε 2d N + A i )( D A i ) + N 2d N εd + A i 5rθ µ N (1 2p) N 2d N (5rθN 2(1 p) ) 0. 7

8 For N < i < Q, use (A.6g). Bound (3.7b) in the same manner. For (3.7c), for 0 < i N, using (A.6), for sufficiently large N, we have A i A i 1 2ε Σh (h id + h i+1 D + )A i A i A i 1 A 2 N (1 Cρ) 2 θ θ2 > θr. For N < i Q, use (A.6h). The bound (3.7d) is established in the same manner. Next, consider the case where σ = µ = 1 2 (1 dn ). Using (3.5b) we can show that D A ε (x i ) = θd h T i, δ 2 A ε (x i ) = { h H 0 θδ 2 h T i, i < N, 2h H 0 +h θδ2 h T i, i = N, θδ2 h T i, i > N }. (3.9) Using Lemma A.2 and H 0 h = H 1, we can prove (3.7a)-(3.7d), in the same manner as the case σ < µ. The case σ = µ = dn 2 follows suit. Define the discrete barrier functions ψ ± (x i ) := f θr [(x i x Q )A ε (x i ) + max x i Ω N ε (x i x Q )A ε (x i )] + max { A, B } ± Y ε (x i ), (3.10) which are nonnegative at x = 0, 1. Using (3.5a) and Lemma A.2 we can find { D ± [(x i x Q )A ε (x i )] = (x i x Q )D ± 0, i Q A ε (x i ) + A ε (x i±1 ) 0, i Q. (3.11) Using this, we can show that L N ε ψ ± (x i ) { f θr [(x i x Q )(εδ 2 A i A i D A i ) (A i A i 1 2ε Σh (h id + h i+1 D + )A i )] + f, 0 < i Q, f θr [(x i x Q )(εδ 2 A i A i D + A i ) (A i+1 A i 2ε Σh (h id + h i+1 D + )A i )] + f, Q < i < N. Using the bounds in (3.7), we have L N ε ψ ± (x i ) 0 for x i Ω N ε. Thus using Theorem 3.1, we have Y ε C on Ω N ε. As with the continuous problem (P ε ), we split the discrete problem (Pε N ) into left and right discrete problems centred around x = x Q as follows. Define Y L (x i ), x i x Q, Y ε (x i ) = Y R (x i ), x i x Q, where the discrete left and right problems are defined by L N ε Y L (x i ) = f(x i ), x i Ω N ε (0, x Q ), Y L (0) = y 0, Y L (x Q ) = Y ε (x Q ), (P N L ) L N ε Y R (x i ) = f(x i ), x i Ω N ε (0, x Q ), Y R (x Q ) = Y ε (x Q ), Y R (1) = y 1. (P N R ) We decompose each Y L\R into the sum of a regular component, V L\R, and a layer component, W L\R. We define V L as the solution of L N, V L (x i ) := (εδ 2 + a 0 D b)v L (x i ) = f(x i ), x i Ω N ε (0, x Q ) V L (0) = y 0, V L (x Q ) = (V 0 + εv 1 )(x Q ), (3.12a) 8

9 where V L = V 0 + εv 1 + ε 2 V 2 and V 0, V 1, V 2 satisfy (a 0 D + b)v 0 (x i ) = f(x i ), x i Ω N ε (0, x Q ], V 0 (0) = y 0, (3.12b) { (a δ 0 D + b)v 1 (x i ) = 2 V 0 (x i ), x i Ω N } ε (0, x Q ) δ 2, V V 0 (x i 1 ), x i = x 1 (0) = 0, (3.12c) Q L N, V 2 (x i ) = δ 2 V 1 (x i ), x i Ω N ε (0, x Q ), V 2 (0) = V 2 (x Q ) = 0. (3.12d) We incorporate the error (L N ε L N, )V L into the discrete layer component, W L, which is, noting (2.1c), defined as the solution of the following discrete problem Similarly V R and W R satisfy L N ε W L (x i ) = ϕ ε D V L (x i ), x i Ω N ε Ω, (3.12e) W L (0) = 0, W L (x Q ) = (Y L V L )(x Q ). (3.12f) L N,+ V R (x i ) := (εδ 2 + a + 0 D+ b)v R (x i ) = f(x i ), x i Ω N ε (x Q, 1), (3.13a) V R (x Q ) = v, V R (1) = y 1, (3.13b) L N ε W R (x i ) = ϕ + ε D + V R (x i ), x i Ω N ε (x Q, 1), (3.13c) W L (x Q ) = (Y R V R )(x Q ), W R (1) = 0. (3.13d) where v is defined analogously to V L (x Q ) in (3.12). We now determine bounds on V L\R, W L\R, D ± V L\R and on the error V L\R v L\R. But first we present a mesh function that we will use in the barrier functions for the layer component, along with some of its properties. The proof of this Lemma is given in Appendix C. Lemma 3.3. For sufficiently large N, the mesh function defined as satisfies Ŵ (x i ) := Q j=i+1 ( Ŵ (x i ) Ce θ 2ε (d x i), 1 + α ) ε(x j ) 1 h j, 0 i < Q, Ŵ (x Q ) := 1 2ε Ŵ (x i ) (1 Cρ)e θ 2ε (d x i), 0 i Q, (3.1a) N i Q (σ < µ), 0 i Q (σ = µ). (3.1b) Ŵ (x i) CN (1 p), i N (σ < µ). (3.1c) Lemma 3.. If v L, w L are the solutions of (2.), v R, w R are the solutions of (2.5), V L, W L are the solutions of (3.12) and V R, W R are the solutions of (3.13) then we have the following bounds D V L (x i ) C, x i Ω N ε (0, x Q ]; (V L v L )(x i ) CN 1 x i, x i Ω N ε [0, x Q ], D + V R (x i ) C, x i Ω N ε [x Q, 1); (V R v R )(x i ) CN 1 (1 x i ), x i Ω N ε [x Q, 1], { Ce W L (x i ) θ 2ε (d xi) + CN (1 p) N, i Q (σ < µ), 0 i Q (σ = µ), CN (1 p), 0 i N (σ < µ), { Ce W R (x i ) θ 2ε (xi d) + CN (1 p), Q i 3N (σ < µ), Q i N (σ = µ), CN (1 p), i N (σ < µ). 3N 9

10 Proof. If Z is a mesh function on Ω N ε [0, x Q ] then we can easily prove that: if Z x0 0 and (a 0 D + b)z ΩN ε [0,x Q ]\x 0 0 then Z ΩN ε [0,x Q ] 0. (3.15) Consider the functions ψ ± (x i ) = y 0 + f θ x i ± V 0 (x i ). Using (2.1), (3.12b) and (3.15) we can show that V 0, D V 0, δ 2 V 0 C. Similarly we can show that V 1, D V 1, δ 2 V 1 C. For (3.12d), using the results in [1, 3.3] we have V 2 (x i ) Cx i, x i Ω N ε [0, x Q ]. Repeat [1, 3.5,Lemma 3.1] and obtain ε D V 2 C. Thus it follows that D V L C. Using (2.) with (3.13) we can define the errors E 0 := V 0 v 0, E 1 := V 1 v 1 and E := V ε v ε as the solutions of (a 0 D + b)e 0 (x i ) = a 0 (v 0 D v 0 )(x i ), x Ω N ε (0, x Q ], E 0 (0) = 0, (3.16a) (a 0 D + b)e 1 (x i ) = (a 0 (v 1 D v 1 ) (v 0 δ2 V 0 ))(x i ), x i Ω N ε (0, x Q ], E 1 (0) = 0, L N, E(x i ) = (L L N, )v L (x i ), x i Ω N ε (0, x Q ), E(0) = 0, E(x Q ) = (E 0 + εe 1 )(x Q ). (3.16b) (3.16c) Using Lemma 2.2, we can use standard truncation error estimates to show that v 0 D v 0 C v 0 max h i CN 1, (3.17a) v 0 δ 2 v 0 C v 0 max h i CN 1, (3.17b) v 1 D v 1 C v 1 max h i CN 1, (3.17c) (L L N, )v L C(ε v L + v L ) max h i CN 1. (3.17d) Using suitable barrier functions, we can easily show that E 0 (x i ) CN 1 x i. D v 0 )(x i ) = 1 xi h i t=x i 1 v 0 (x i) v 0 (t) dt we can show that Using (v 0 { (v 0 δ 2 V 0 )(x i ) (v 0 δ 2 v 0 )(x i ) + δ 2 E 0 (x i ) CN 1 CN + 1, i N, C, i = N. Rewrite (3.16b) as E 1 (x i ) = [E 1 (x i 1 ) + h i = i i [ j=1 k=j Using (1 + h i b(x i ) a 0 (x i) ) 1 < 1 and using (3.17) we have a 0 (a 0 (v 1 D v 1 ) (v 0 δ 2 V 0 ))(x i )](1 + bh i a 0 (x i )) 1 (1 + bh i a (x k )) 1 ] h j 0 a (a 0 (v 1 D v 1 ) (v 0 δ 2 V 0 ))(x j ). 0 E 1 (x i ) i h i CN 1 + C i j=1 h j (v 0 δ 2 V 0 )(x j ). For i < N, we clearly have E 1(x i ) CN 1 + C h i icn 1 CN 1 and for i N we have E 1 (x i ) CN 1 + Ch N + C h i (i N )CN 1 CN 1. Use suitable barrier functions with (3.16c) to find that E(x i ) CN 1 x i for x i Ω N ε [0, x Q ]. The proof for the bounds on V R can be established in an analogous manner. We now bound the layer component W L. We first establish a few inequalities. Using 1 2 et 10

11 cosh t e t, t > 0 and using (2.1), (A.5) and Lemma A.2, bound ϕ ε N as follows and a ε for sufficiently large Ŵ (x i 1 ) e θ 2ε (d x i+h) 1 2 e θ 2ε (d x i), N < i < Q (σ < µ), 0 < i < Q (σ = µ) (3.18a) α ε (x i ) θ 2, 0 i N, (σ < µ) (3.18b) ϕ ε (x i ) N (1 p), 0 i N, (σ < µ). (3.18c) Consider the functions ψ ± (x i ) = m 1 Ŵ (x i ) + m 2 N (1 p) x i ± W L (x i ), m 2 := 2 θ ϕ ε (d) D V L, m 1 := max { W L (d), 2ε r m 2} which are nonnegative at x = 0, 1. Consider the σ < µ case. Using Lemma 3.3, (3.18) and (3.13) and noting for any x i, we have α ε(x i+1 ) D + α ε (x i ) α ε(x i ). Then for sufficiently large N we have L N ε ψ ± (x i ) m 1 (εδ 2 Ŵ α ε D Ŵ )(x i ) α ε (x i )m 2 N (1 p) + D V L ϕ ε (x) m 1 h i+1 2ε(h i+1 + h i ) (α2 ε 2εα ε)ŵ (x i 1) α ε (x i )m 2 N (1 p) + D V L ϕ ε (x) m 2 θ 2 N 1 + ϕ ε (d) D V L N (1 p) 0, 0 < i N, m 1 θr ε e θ 2ε (d x i) + ϕ ε (d) D V L e θ 2ε (d x i) 0, N < i < Q. Hence using Theorem 3.1, we have W L (x i ) C(Ŵ (x i) + N (1 p) ). The case σ = µ is proved in the same manner as for N < i < Q above. The proof of the bound on W R can be performed in a similar way. A bound on the error Y ε y ε is given in the following lemma. Lemma 3.5. If y ε is the solution of (P ε ) and Y ε is the solution of (P N ε ) then where d N d < pσ and p 1 2. (Y ε y ε )(x i ) CN 1 (ln N) 2 + CN (1 p), x i Ω N ε, Proof. First, if σ < µ then for i N, using Lemma 2.2, Lemma 3. and (3.1), (A.5) we have (Y ε y ε )(x i ) W L (x i ) + w L (x i ) + V L v L CN (1 p). Similarly, if σ < µ then for i 3N, we have (Y ε y ε )(x i ) CN (1 p). We now need to examine the error over all mesh points x i [x L, x R ] where x L = 0 if σ = µ or x L = x N if σ < µ and x R = 1 if σ = µ or x R = x 3N if σ < µ. Note the implication that for any x i (x L, x R ), we have h i /ε CN 1 ln N. The error E(x i ) := (Y ε y ε )(x i ) is defined as a solution of the following L N ε E(x i ) = (L ε L N ε )y ε, x i Ω N ε (x L, x R ), E(x L\R ) CN (1 p). (3.19) Note the following truncation error estimate ε(δ 2 y ε y ε )(x i ) 2ε h i+1 + h i i+1 1 h j=i j xj t=x j 1 y ε (s) y ε (x i ) ds dt s=x i. (3.20) t 11

12 For any s [x i 1, x i+1 ] [x L, x R ] Ω N ε, using (P ε ), (2.1) and Lemma 2.1, we can bound ε y ε (s) y ε (x i ) as follows ε y ε (s) y ε (x i ) ( a ε y ε + ( b + Thus using (3.21) with (3.20) we have 1 0 a ε dt) y ε + b y ε + f ) s x i C h i ε 2. (3.21) (L ε L N ε )y ε (x i ) Ch i ε 2 Cε 1 N 1 ln N, x i Ω N ε (x L, x R ). (3.22) Hence, using the barrier functions in (3.10) we can show that E(x i ) Cε 1 N 1 ln N[(x i x Q )A ε (x i ) + max x i Ω N ε [x L,x R ] (x i x Q )A ε (x i )] + CN (1 p) (3.23) Cσ ε N 1 ln N + CN (1 p) CN 1 (ln N) 2 + CN (1 p), σ < µ, C ε N 1 ln N + CN (1 p) CN 1 (ln N) 2 + CN (1 p), σ = µ. Complete the proof using Theorem 3.1. The nodal bound on the error is easily extended to a global error bound. Lemma 3.6. If y ε is the solution of (P ε ) and Y ε is the solution of (P N ε ) then (Y ε y ε )(t) CN 1 (ln N) 2 + CN (1 p), t [0, 1]. where Y ε is the piecewise linear interpolant of Y ε on [0, 1]. Proof. Proof follows the corresponding proof in [1, Th m 3.12]. Numerical Examples Example 1 (3.2) (3.25) In this example from the class (P 1 ) we consider a ε (x) = ( x 2 ) tanh ( 1.1 ε (0.6 x)), b(x) = e 5x, f(x) = cos(3x), A = 2, B = 3, d = 0.6. We choose r = 1 and θ = 2.1. We consider various values of p. To generate a value of d N, we start with a scale factor of κ = 0.99 and reduce by 0.01 until d N = d ± κp min { 1 2 min (d, 1 d), ε r ln N} satisfies d dn < pσ. (.1) We consider p = 0, p = 0.5 and p = 1 and using (.1) we centre the mesh at d N = d, d N d 0.5σ and d N d σ. We compute the approximate errors E N ε = max Ω N ε Ω8192 ε U N ε U 8192 ε (.2) where U M ε is Uε M, the numerical solution of (Pε N ) using N = M mesh points, interpolated onto the mesh Ω N ε Ω 8192 ε. Table.1 displays the approximate errors Eε N and the uniform errors E N = max ε Eε N, using (.2). Shifting the mesh off-centre within the limit of p 1 2 has little to no effect on the differences. We further test for an effect on the value of p by producing the computed rates of convergence p N ε and the uniform rates of convergence p N, computed using the double mesh principle (see [1]), as shown in Table.2. The N (1 p) factor established in Lemma 3.5 is not evident for p 1 2. However, for p = 1 we see a collapse in the computed rates of convergence. 12

13 Eε N d N = d d N d + 0.5σ d N d + σ N ε E N Table.1: Approximate errors for Example 1. p N ε d N = d d N d + 0.5σ d N d + σ N ε p N Table.2: Computed orders of convergence for Example 1. 13

14 In Figure.1, we present comparisons between numerical solutions using N = 32 and the mesh centred at d N = d and ε = 2 0, ε = 2 2, ε = 2 5 and ε = Fig..1: Solution of (P N ε ) with N = 32, d N = d for ε = 2 0, ε = 2 2, ε = 2 5 and ε = In Figure.2, we present comparisons between numerical solutions using N = 32, ε = 2 8 and the mesh centred at d N = d, d N d 0.5σ and d N d σ. The location of the meshes for each case are superimposed on the figure below the graphs. Fig..2: Solution of (P N ε ) with N = 32, ε = 2 8 and d N d pσ for p = 0, p = 1 2 and p = 1. 1

15 References [1] P.A. Farrell, A.F. Hegarty, J.J.H. Miller, E. O Riordan, G.I. Shishkin, Robust Computational Techniques for Boundary Layers, Chapman and Hall/CRC, Boca Raton, FL. (2000) [2] A.E. Berger, H. Han, R.B. Kellogg, A Priori Estimates and Analysis of a Numerical Method for a Turning Point Problem, Mathematics of Computation, 2(166), 65-92, (198) [3] P.A. Farrell, Sufficient Conditions for the Uniform convergence of a difference scheme for a Singularly Perturbed Turning Point Problem, SIAM J. Numer. Anal., 25(3), , (1988) [] P.A. Farrell, A.F. Hegarty, J.J.H. Miller, E. O Riordan, G.I. Shishkin, Global Maximum Norm Parameter-Uniform Numerical Method for a Singularly Perturbed Convection- Diffusion Problem with Discontinuous Convection Coefficient, Math. and Computer Modelling, 0, , (2000) [5] F.A. Howes, Boundary-interior layer interactions in nonlinear singular perturbation theory, Memoirs of the American Mathematical Society, 15(203), (1978) [6] E. O Riordan, J. Quinn, Parameter-uniform numerical methods for some linear and nonlinear singularly perturbed convection diffusion boundary turning point problems, BIT Numer. Math., 51(2), , (2011) A Properties of the discrete functions S, C and T. Lemma A.1. For sufficiently large N, the mesh functions S and C defined in (3.3) satisfy: S(x i 1 ) > S(x i ), i; S(x i ) > 0, i < Q; S(x Q ) = 0; S(x i ) < 0, i > Q; (A.1a) C(x i 1 ) > C(x i ), i < Q; C(x i 1 ) < C(x i ), i > Q; C(x i ) C(x Q ) = 2, i; (A.1b) } { C(x i Q 2, i ) C(xi ), i < Q 1+ρ C(x i+1 ) C(x Q 1 i, C(x i ) i ) 1 ρ, Q i ; (A.1c) } { i Q, C(x i ) (1 + ρ)c(xi ), i < Q C(x Q < i, (1 ρ)c(x i ) i 1 ) ; (A.1d) C(x i ), Q i T (x i ) > 0, i < Q; T (x Q ) = 0; T (x i ) < 0, i > Q; (A.1e) T (x i ) < 1, T (x i ) tanh( r ε (Q i)h) CNρ2, i; T (x N ), T (x 3N ) 1 Cρ > 1 2. (A.1f) Proof. For convenience in this proof, we write any mesh function Z(x i ) as Z i. Clearly C i 2, i and S i 0, i Q, S i 0, i Q. Use S i S i 1 = ρc i and C i C i 1 = ρs i to establish the bounds in (A.1a) and (A.1b). For (A.1c), we can show C i 1 = C i + ρs i. We can use C i < S i < C i (A.2) to show that (1 ρ)c i < C i 1 < (1 + ρ)c i. Combine these with (A.1b) to establish (A.1c). For (A.1d), we can show that C i+1 = 1 (C 1 ρ 2 i ρs i ). Thus from (A.2), we have C i+1 < 1+ρ C 1 ρ 2 i = 1 1 ρ C i. Similarly C i+1 > 1 1+ρ C i. Combine these with (A.1b) to establish (A.1d). The expressions in (A.1e) follow from (A.1a) and (A.1b). The first bound in (A.1f) follows from (A.2). Using the inequalities e (1 t/2)t 1 + t e t 1 + 2t, t [0, 0.5], (A.3a) e (1+t)t 1 t e t 1 t/2, t [0, 0.5], (A.3b) 15

16 we can show that for any sufficiently small ρ > 0 and 0 j N, we have (1 CN ρ 2 ) sinh ρj CN ρ [(1 + ρ)j (1 ρ) j ] (1 + CN ρ 2 ) sinh ρj + CN ρ 2 (A.a) (1 CN ρ 2 ) cosh ρj 1 2 [(1 + ρ)i + (1 ρ) i ] cosh ρj. (A.b) With a change of variables, we can prove the second bound in (A.1f) using (A.3), (A.) with (3.3) for sufficiently large N. Note that using (3.1) and (3.2), for sufficiently large N we have x 3N x Q x N (1 p)σ Q N (1 p) N, (A.5a) x Q (1 p)σ 3N Q (1 p) N. (A.5b) For i = N and for sufficiently large N, using (A.5), we can write C i (1 + Cρ)(1 + ρ) Q N and S i (1 Cρ)(1 + ρ) Q N. Hence S i /C i 1 Cρ > 1 2. We can write a similar statement for i = 3N. Lemma A.2. Assuming (2.1), then for sufficiently large N, the operators in (3.) and the mesh function T defined in (3.3) satisfy: D h T (x i) < 0, 0 < i N, D + h T (x i) < 0, 0 i < N, (A.6a) 1 + ρ 1 ρ D h T (x i) D + h T (x i) D h T (x i), 0 < i < Q, (A.6b) D + h T (x Q) = D h T (x Q), (A.6c) 1 + ρ 1 ρ D+ h T (x i) D h T (x i) D + h T (x i), Q < i < N, (A.6d) T (x i ) T (x N ) Cρ, i < N, T (x i) T (x 3N ) Cρ, i > 3N, (A.6e) ε D + h T (x N ), ε D h T (x 3N ) 5rθN 2(1 p), when σ = ε r ln N, (A.6f) εδh 2 T (x i) θt (x i )D h T (x i) 0, 0 < i < Q, (A.6g) εδ 2 h T (x i) θt (x i )D + h T (x i) 0, Q < i < N, (A.6h) T (x i )T (x i 1 ) ε θ (D h + D+ h )T (x i) 2r, 0 < i Q, (A.6i) T (x i+1 )T (x i ) ε θ (D h + D+ h )T (x i) 2r, Q < i < N. (A.6j) Proof. To ease notation, we write U(x i ) as U i for any mesh function U. Using (3.3) with (3.), we can show that D h T (x i) = r ε (T it i 1 1) = r/ε(1 ρ2 ) Q i δ 2 h T (x i) = r ε T i(d + + D )T i. C i C i 1, (A.7a) (A.7b) Use Lemma A and (A.7) to establish (A.6a), (A.6c) and (A.6e). For (A.6b), using Lemma A we have C i+1 C i 1 (1+ρ) 2 C i C i 1. Thus D + h T (x i) = r/ε(1 ρ2 ) Q (i+1) (1 + ρ)2 r/ε(1 ρ 2 ) Q i C i+1 C i 1 ρ 2 C i C i 1 = 1 + ρ 1 ρ D h T (x i). Also, from Lemma A.1 and (A.7) we have δh 2T (x i) 0, i Q. Thus using (3.), we have D + h T (x i) D h T (x i), i Q. Verify (A.6d) in the same manner. For (A.6f), when i = N then using Lemma A.1 and (3.3) we have C i+1 C i C 2 i+1 = [(1 + ρ) Q (i+1) + (1 ρ) Q (i+1) ] 2 (1 + ρ) 2(Q (i+1)). (A.8) 16

17 Using (A.5), (A.8) and the inequality (1 + t)/(1 t) e 2t, 0 < t 0.5, assuming σ = ε r ln N, then we have D + h T (x i) = rε 1 (1 ρ 2 ) Q (i+1) C i+1 C i rε 1 (1 + ρ)q (i+1) (1 ρ) Q (i+1) (1 + ρ) 2(Q (i+1)) = 1+ρ 1 ρ rε 1 ( 1 ρ 1 + ρ ) Q N 5rε 1 e 2ρ(Q N ) = 5rε 1 N 2(1 p). Use (A.6a) to bound D + h T (x i) from above. Bound D T (x 3N ) in the same manner. For (A.6g), using (2.1), (A.7) and (A.6b) with Lemma A.1 and (A.6a) for i < Q, we have εδ 2 h T i θt i D h T i = T i (r(d + + D )A i θd h T i) T i ( 2r (1 ρ) θ)d h T i 0. The bound (A.6h) is established in the same manner. For (A.6i) on 0 < i Q, using (2.1), Lemma A.1, (A.6a) and (A.7) we have T i T i 1 ε θ (D h + D+ h )T i T i T i 1 2ε θ D h T i = 2r 2r θ + (1 θ )T it i 1 2r θ. The bound (A.6j) is established in the same manner. B Proof of Lemma 3.1 Proof. First, using (3.1), (3.2) and the identity tanh (X + Y ) = tanh X+tanh Y 1+tanh X tanh Y we have tanh( r ε (x Q x i )) = tanh[ r ε (d x i) + r ε (x Q d)] = tanh( r ε (d x i)) + tanh( r ε (x Q d)) 1 + tanh( r ε (d x i)) tanh( r ε (x Q d)) tanh( r ε (d x i)) + ρ 1 ρ tanh( r ε (d x i)) + Cρ. Construct a corresponding lower bound to give tanh( r ε (d x i)) tanh( r ε (x Q x i )) Cρ CN 1 ln N, i. (B.1) Consider i S := { [ N, 3N ] if σ < µ, [ N, N] if σ = 1 2 (1 dn ), [0, 3N ] if σ = dn 2 } Z. Using (3.1), we have x Q x i = x N + h(q 3N ) x N h(i 3N ) = (Q i)h. Now using Lemma A.2 we have A ε (x i ) θ tanh( r ε (x Q d)) CNρ 2. For all other i / S, using Lemma A.2 and (3.5) we have A ε (x i ) θ tanh( r ε (x Q x i )) θ T N, 3N + CN (1 2p) + C T i T N, 3N θ tanh( r ε (x Q x i )) C T N, 3N 1 + CN (1 2p) + Cρ + C 1 θ tanh( r ε (x Q x i )) Cρ + CN (1 2p) + Cρ + C 1 θ tanh( r ε (x Q x N, 3N )) Cρ + CN (1 2p) + C 1 θ tanh( r ε (Q ( N, 3N ))h) Cρ + CN (1 2p) + C 1 T N, 3N + C T N, 3N θ tanh( r ε (Q ( N, 3N ))h) Cρ + CN (1 2p) + Cρ + CNρ 2. Thus, i, we have A ε (x i ) θ tanh( r ε (x Q x i )) Cρ + CN (1 2p) + CNρ 2 CN (1 2p) + CN 1 (ln N) 2. (B.2) Combine (B.1) and (B.2) to complete the proof. 17

18 C Proof of Lemma 3.3 Proof. We can easily bound Ŵ from below using 1 + t et, t > 0 with (2.1), (3.1) and (A.5) as follows Ŵ (x i ) Q j=i+1 e αε(x j ) 2ε h j = e 1 2ε Q j=i+1 αε(x j)h j e θ 2ε Q j=i+1 h j = e θ 2ε (x Q x i ) 1 2 e θ 2ε (d x i). Note the Rectangle Rule for Numerical Integration: Partition the interval [a, b] into N subintervals with width h = 1/N, where t j = a + jh. Then b a N f(s) ds = h f(t j ) + O((b a)h 2 f ). j=1 Using (A.5) with the rectangle rule on N we have i Q when (σ < µ) or on 0 i Q when (σ = µ), 1 2ε Q j=i+1 α ε (x j )h j 1 2ε xq x i α ε (t) dt Cσ ε (h/ε)2 1 xq α ε (t) dt CN 2 (ln N) 3. (C.1) 2ε x i Using the inequality 1 + t e t t2 /2, t > 0 with (2.1), (2.7), (3.1), (A.5) and (C.1) we have Ŵ (x i ) Q j=i+1 e αε(x j ) 2ε h j e 1 2 ( αε(x j ) 2ε h j ) 2 e CN(N 1 ln N) 2 e 1 2ε Q j=i+1 αε(x j)h j Ce 1 2ε x Q x i α ε(t) dt Ce 1 d 2ε x α i ε(t) dt Ce θ 2ε (d xi). Finally, we have Ŵ (x i) Ŵ (x i 1) = αε(x i)h i 2ε Ŵ (x i ) > 0. Thus for i N (2.1) and (A.5) we have when (σ < µ), using Ŵ (x i ) Ŵ (x N ) Ce θ 2ε (d xn/) CN (1 p). 18

Interior Layers in Singularly Perturbed Problems

Interior Layers in Singularly Perturbed Problems Interior Layers in Singularly Perturbed Problems Eugene O Riordan Abstract To construct layer adapted meshes for a class of singularly perturbed problems, whose solutions contain boundary layers, it is

More information

Parameter-Uniform Numerical Methods for a Class of Singularly Perturbed Problems with a Neumann Boundary Condition

Parameter-Uniform Numerical Methods for a Class of Singularly Perturbed Problems with a Neumann Boundary Condition Parameter-Uniform Numerical Methods for a Class of Singularly Perturbed Problems with a Neumann Boundary Condition P. A. Farrell 1,A.F.Hegarty 2, J. J. H. Miller 3, E. O Riordan 4,andG.I. Shishkin 5 1

More information

A PARAMETER ROBUST NUMERICAL METHOD FOR A TWO DIMENSIONAL REACTION-DIFFUSION PROBLEM

A PARAMETER ROBUST NUMERICAL METHOD FOR A TWO DIMENSIONAL REACTION-DIFFUSION PROBLEM A PARAMETER ROBUST NUMERICAL METHOD FOR A TWO DIMENSIONAL REACTION-DIFFUSION PROBLEM C. CLAVERO, J.L. GRACIA, AND E. O RIORDAN Abstract. In this paper a singularly perturbed reaction diffusion partial

More information

A Linearised Singularly Perturbed Convection- Diffusion Problem with an Interior Layer

A Linearised Singularly Perturbed Convection- Diffusion Problem with an Interior Layer Dublin Institute of Technology ARROW@DIT Articles School of Mathematics 201 A Linearised Singularly Perturbed Convection- Diffusion Problem with an Interior Layer Eugene O'Riordan Dublin City University,

More information

Iterative Domain Decomposition Methods for Singularly Perturbed Nonlinear Convection-Diffusion Equations

Iterative Domain Decomposition Methods for Singularly Perturbed Nonlinear Convection-Diffusion Equations Iterative Domain Decomposition Methods for Singularly Perturbed Nonlinear Convection-Diffusion Equations P.A. Farrell 1, P.W. Hemker 2, G.I. Shishkin 3 and L.P. Shishkina 3 1 Department of Computer Science,

More information

Parameter robust methods for second-order complex-valued reaction-diffusion equations

Parameter robust methods for second-order complex-valued reaction-diffusion equations Postgraduate Modelling Research Group, 10 November 2017 Parameter robust methods for second-order complex-valued reaction-diffusion equations Faiza Alssaedi Supervisor: Niall Madden School of Mathematics,

More information

A uniformly convergent numerical method for a coupled system of two singularly perturbed linear reaction diffusion problems

A uniformly convergent numerical method for a coupled system of two singularly perturbed linear reaction diffusion problems IMA Journal of Numerical Analysis (2003 23, 627 644 A uniformly convergent numerical method for a coupled system of two singularly perturbed linear reaction diffusion problems NIALL MADDEN Department of

More information

Quintic B-spline method for numerical solution of fourth order singular perturbation boundary value problems

Quintic B-spline method for numerical solution of fourth order singular perturbation boundary value problems Stud. Univ. Babeş-Bolyai Math. 63(2018, No. 1, 141 151 DOI: 10.24193/subbmath.2018.1.09 Quintic B-spline method for numerical solution of fourth order singular perturbation boundary value problems Ram

More information

A coupled system of singularly perturbed semilinear reaction-diffusion equations

A coupled system of singularly perturbed semilinear reaction-diffusion equations A coupled system of singularly perturbed semilinear reaction-diffusion equations J.L. Gracia, F.J. Lisbona, M. Madaune-Tort E. O Riordan September 9, 008 Abstract In this paper singularly perturbed semilinear

More information

Available online at ScienceDirect. Procedia Engineering 127 (2015 )

Available online at   ScienceDirect. Procedia Engineering 127 (2015 ) Available online at www.sciencedirect.com ScienceDirect Procedia Engineering 127 (2015 ) 424 431 International Conference on Computational Heat and Mass Transfer-2015 Exponentially Fitted Initial Value

More information

Remarks on the analysis of finite element methods on a Shishkin mesh: are Scott-Zhang interpolants applicable?

Remarks on the analysis of finite element methods on a Shishkin mesh: are Scott-Zhang interpolants applicable? Remarks on the analysis of finite element methods on a Shishkin mesh: are Scott-Zhang interpolants applicable? Thomas Apel, Hans-G. Roos 22.7.2008 Abstract In the first part of the paper we discuss minimal

More information

ACCURATE NUMERICAL METHOD FOR BLASIUS PROBLEM FOR FLOW PAST A FLAT PLATE WITH MASS TRANSFER

ACCURATE NUMERICAL METHOD FOR BLASIUS PROBLEM FOR FLOW PAST A FLAT PLATE WITH MASS TRANSFER FDS-2000 Conference, pp. 1 10 M. Sapagovas et al. (Eds) c 2000 MII ACCURATE UMERICAL METHOD FOR BLASIUS PROBLEM FOR FLOW PAST A FLAT PLATE WITH MASS TRASFER B. GAHA 1, J. J. H. MILLER 2, and G. I. SHISHKI

More information

SPLINE COLLOCATION METHOD FOR SINGULAR PERTURBATION PROBLEM. Mirjana Stojanović University of Novi Sad, Yugoslavia

SPLINE COLLOCATION METHOD FOR SINGULAR PERTURBATION PROBLEM. Mirjana Stojanović University of Novi Sad, Yugoslavia GLASNIK MATEMATIČKI Vol. 37(57(2002, 393 403 SPLINE COLLOCATION METHOD FOR SINGULAR PERTURBATION PROBLEM Mirjana Stojanović University of Novi Sad, Yugoslavia Abstract. We introduce piecewise interpolating

More information

An exponentially graded mesh for singularly perturbed problems

An exponentially graded mesh for singularly perturbed problems An exponentially graded mesh for singularly perturbed problems Christos Xenophontos Department of Mathematics and Statistics University of Cyprus joint work with P. Constantinou (UCY), S. Franz (TU Dresden)

More information

Superconvergence analysis of the SDFEM for elliptic problems with characteristic layers

Superconvergence analysis of the SDFEM for elliptic problems with characteristic layers Superconvergence analysis of the SDFEM for elliptic problems with characteristic layers S. Franz, T. Linß, H.-G. Roos Institut für Numerische Mathematik, Technische Universität Dresden, D-01062, Germany

More information

A COMPUTATIONAL TECHNIQUE FOR SOLVING BOUNDARY VALUE PROBLEM WITH TWO SMALL PARAMETERS

A COMPUTATIONAL TECHNIQUE FOR SOLVING BOUNDARY VALUE PROBLEM WITH TWO SMALL PARAMETERS Electronic Journal of Differential Equations, Vol. 2013 (2013), No. 30, pp. 1 10. ISSN: 1072-6691. URL: http://ejde.math.txstate.edu or http://ejde.math.unt.edu ftp ejde.math.txstate.edu A COMPUTATIONAL

More information

On the relationship of local projection stabilization to other stabilized methods for one-dimensional advection-diffusion equations

On the relationship of local projection stabilization to other stabilized methods for one-dimensional advection-diffusion equations On the relationship of local projection stabilization to other stabilized methods for one-dimensional advection-diffusion equations Lutz Tobiska Institut für Analysis und Numerik Otto-von-Guericke-Universität

More information

An Exponentially Fitted Non Symmetric Finite Difference Method for Singular Perturbation Problems

An Exponentially Fitted Non Symmetric Finite Difference Method for Singular Perturbation Problems An Exponentially Fitted Non Symmetric Finite Difference Method for Singular Perturbation Problems GBSL SOUJANYA National Institute of Technology Department of Mathematics Warangal INDIA gbslsoujanya@gmail.com

More information

Numerical integration of singularly perturbed delay differential equations using exponential integrating factor

Numerical integration of singularly perturbed delay differential equations using exponential integrating factor MATHEMATICAL COMMUNICATIONS 51 Math. Commun. 017), 51 64 Numerical integration of singularly perturbed delay differential equations using exponential integrating factor Lakshmi Sirisha Challa and Yanala

More information

1. Nonlinear Equations. This lecture note excerpted parts from Michael Heath and Max Gunzburger. f(x) = 0

1. Nonlinear Equations. This lecture note excerpted parts from Michael Heath and Max Gunzburger. f(x) = 0 Numerical Analysis 1 1. Nonlinear Equations This lecture note excerpted parts from Michael Heath and Max Gunzburger. Given function f, we seek value x for which where f : D R n R n is nonlinear. f(x) =

More information

On a Two-Point Boundary-Value Problem with Spurious Solutions

On a Two-Point Boundary-Value Problem with Spurious Solutions On a Two-Point Boundary-Value Problem with Spurious Solutions By C. H. Ou and R. Wong The Carrier Pearson equation εü + u = with boundary conditions u ) = u) = 0 is studied from a rigorous point of view.

More information

Fitted operator finite difference scheme for a class of singularly perturbed differential difference turning point problems exhibiting interior layer

Fitted operator finite difference scheme for a class of singularly perturbed differential difference turning point problems exhibiting interior layer Fitted operator finite difference scheme for a class of singularly perturbed differential difference turning point problems exhibiting interior layer Dr. Pratima Rai Department of mathematics, University

More information

Chapter 2 First Order Differential Equations

Chapter 2 First Order Differential Equations Chapter 2 First Order Differential Equations 2.1 Problem A The plan for this book is to apply the theory developed in Chapter 1 to a sequence of more or less increasingly complex differential equation

More information

ODE Final exam - Solutions

ODE Final exam - Solutions ODE Final exam - Solutions May 3, 018 1 Computational questions (30 For all the following ODE s with given initial condition, find the expression of the solution as a function of the time variable t You

More information

Applications of the periodic unfolding method to multi-scale problems

Applications of the periodic unfolding method to multi-scale problems Applications of the periodic unfolding method to multi-scale problems Doina Cioranescu Université Paris VI Santiago de Compostela December 14, 2009 Periodic unfolding method and multi-scale problems 1/56

More information

Abstract. 1. Introduction

Abstract. 1. Introduction Journal of Computational Mathematics Vol.28, No.2, 2010, 273 288. http://www.global-sci.org/jcm doi:10.4208/jcm.2009.10-m2870 UNIFORM SUPERCONVERGENCE OF GALERKIN METHODS FOR SINGULARLY PERTURBED PROBLEMS

More information

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation

1. Introduction Boundary estimates for the second derivatives of the solution to the Dirichlet problem for the Monge-Ampere equation POINTWISE C 2,α ESTIMATES AT THE BOUNDARY FOR THE MONGE-AMPERE EQUATION O. SAVIN Abstract. We prove a localization property of boundary sections for solutions to the Monge-Ampere equation. As a consequence

More information

Finite difference method for elliptic problems: I

Finite difference method for elliptic problems: I Finite difference method for elliptic problems: I Praveen. C praveen@math.tifrbng.res.in Tata Institute of Fundamental Research Center for Applicable Mathematics Bangalore 560065 http://math.tifrbng.res.in/~praveen

More information

LECTURE 3: DISCRETE GRADIENT FLOWS

LECTURE 3: DISCRETE GRADIENT FLOWS LECTURE 3: DISCRETE GRADIENT FLOWS Department of Mathematics and Institute for Physical Science and Technology University of Maryland, USA Tutorial: Numerical Methods for FBPs Free Boundary Problems and

More information

Finite Element Superconvergence Approximation for One-Dimensional Singularly Perturbed Problems

Finite Element Superconvergence Approximation for One-Dimensional Singularly Perturbed Problems Finite Element Superconvergence Approximation for One-Dimensional Singularly Perturbed Problems Zhimin Zhang Department of Mathematics Wayne State University Detroit, MI 48202 Received 19 July 2000; accepted

More information

A uniformly convergent difference scheme on a modified Shishkin mesh for the singularly perturbed reaction-diffusion boundary value problem

A uniformly convergent difference scheme on a modified Shishkin mesh for the singularly perturbed reaction-diffusion boundary value problem Journal of Modern Methods in Numerical Mathematics 6:1 (2015, 28 43 A uniformly convergent difference scheme on a modified Shishkin mesh for the singularly perturbed reaction-diffusion boundary value problem

More information

A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION

A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION A LOCALIZATION PROPERTY AT THE BOUNDARY FOR MONGE-AMPERE EQUATION O. SAVIN. Introduction In this paper we study the geometry of the sections for solutions to the Monge- Ampere equation det D 2 u = f, u

More information

Goal. Robust A Posteriori Error Estimates for Stabilized Finite Element Discretizations of Non-Stationary Convection-Diffusion Problems.

Goal. Robust A Posteriori Error Estimates for Stabilized Finite Element Discretizations of Non-Stationary Convection-Diffusion Problems. Robust A Posteriori Error Estimates for Stabilized Finite Element s of Non-Stationary Convection-Diffusion Problems L. Tobiska and R. Verfürth Universität Magdeburg Ruhr-Universität Bochum www.ruhr-uni-bochum.de/num

More information

A New Improvement of Shishkin Fitted Mesh Technique on a First Order Finite Difference Method

A New Improvement of Shishkin Fitted Mesh Technique on a First Order Finite Difference Method A New Improvement of Shishkin Fitted Mesh Technique on a First Order Finite Difference Method With applications on singularly perturbed boundary value problem ODE Rafiq S. Muhammad College of Computers

More information

Nonlinear stabilization via a linear observability

Nonlinear stabilization via a linear observability via a linear observability Kaïs Ammari Department of Mathematics University of Monastir Joint work with Fathia Alabau-Boussouira Collocated feedback stabilization Outline 1 Introduction and main result

More information

Approximation of interior-layer. solutions. in a singularly perturbed semilinear. reaction-diffusion problem

Approximation of interior-layer. solutions. in a singularly perturbed semilinear. reaction-diffusion problem Numerische Mathematik manuscript No. (will be inserted by the editor) Natalia Kopteva Martin Stynes Approximation of interior-layer solutions in a singularly perturbed semilinear reaction-diffusion problem

More information

Boundary Value Problems and Iterative Methods for Linear Systems

Boundary Value Problems and Iterative Methods for Linear Systems Boundary Value Problems and Iterative Methods for Linear Systems 1. Equilibrium Problems 1.1. Abstract setting We want to find a displacement u V. Here V is a complete vector space with a norm v V. In

More information

Measure and Integration: Solutions of CW2

Measure and Integration: Solutions of CW2 Measure and Integration: s of CW2 Fall 206 [G. Holzegel] December 9, 206 Problem of Sheet 5 a) Left (f n ) and (g n ) be sequences of integrable functions with f n (x) f (x) and g n (x) g (x) for almost

More information

ONLINE APPENDIX TO: NONPARAMETRIC IDENTIFICATION OF THE MIXED HAZARD MODEL USING MARTINGALE-BASED MOMENTS

ONLINE APPENDIX TO: NONPARAMETRIC IDENTIFICATION OF THE MIXED HAZARD MODEL USING MARTINGALE-BASED MOMENTS ONLINE APPENDIX TO: NONPARAMETRIC IDENTIFICATION OF THE MIXED HAZARD MODEL USING MARTINGALE-BASED MOMENTS JOHANNES RUF AND JAMES LEWIS WOLTER Appendix B. The Proofs of Theorem. and Proposition.3 The proof

More information

Robust exponential convergence of hp-fem for singularly perturbed systems of reaction-diffusion equations

Robust exponential convergence of hp-fem for singularly perturbed systems of reaction-diffusion equations Robust exponential convergence of hp-fem for singularly perturbed systems of reaction-diffusion equations Christos Xenophontos Department of Mathematics and Statistics University of Cyprus joint work with

More information

Global Solutions for a Nonlinear Wave Equation with the p-laplacian Operator

Global Solutions for a Nonlinear Wave Equation with the p-laplacian Operator Global Solutions for a Nonlinear Wave Equation with the p-laplacian Operator Hongjun Gao Institute of Applied Physics and Computational Mathematics 188 Beijing, China To Fu Ma Departamento de Matemática

More information

Cholesky factorisations of linear systems coming from a finite difference method applied to singularly perturbed problems

Cholesky factorisations of linear systems coming from a finite difference method applied to singularly perturbed problems Cholesky factorisations of linear systems coming from a finite difference method applied to singularly perturbed problems Thái Anh Nhan and Niall Madden The Boundary and Interior Layers - Computational

More information

A posteriori error estimation for elliptic problems

A posteriori error estimation for elliptic problems A posteriori error estimation for elliptic problems Praveen. C praveen@math.tifrbng.res.in Tata Institute of Fundamental Research Center for Applicable Mathematics Bangalore 560065 http://math.tifrbng.res.in

More information

Lecture 18 Oct. 30, 2014

Lecture 18 Oct. 30, 2014 CS 224: Advanced Algorithms Fall 214 Lecture 18 Oct. 3, 214 Prof. Jelani Nelson Scribe: Xiaoyu He 1 Overview In this lecture we will describe a path-following implementation of the Interior Point Method

More information

Universal asymptotics in hyperbolicity breakdown

Universal asymptotics in hyperbolicity breakdown IOP PUBLISHING Nonlinearity 2 (2008) 557 586 NONLINEARITY doi:0.088/095-775/2/3/00 Universal asymptotics in hyperbolicity breakdown Kristian Bjerklöv and Maria Saprykina Department of Mathematics and Statistics,

More information

THE INVERSE FUNCTION THEOREM FOR LIPSCHITZ MAPS

THE INVERSE FUNCTION THEOREM FOR LIPSCHITZ MAPS THE INVERSE FUNCTION THEOREM FOR LIPSCHITZ MAPS RALPH HOWARD DEPARTMENT OF MATHEMATICS UNIVERSITY OF SOUTH CAROLINA COLUMBIA, S.C. 29208, USA HOWARD@MATH.SC.EDU Abstract. This is an edited version of a

More information

Numerical Solution of Second Order Singularly Perturbed Differential Difference Equation with Negative Shift. 1 Introduction

Numerical Solution of Second Order Singularly Perturbed Differential Difference Equation with Negative Shift. 1 Introduction ISSN 749-3889 print), 749-3897 online) International Journal of Nonlinear Science Vol.8204) No.3,pp.200-209 Numerical Solution of Second Order Singularly Perturbed Differential Difference Equation with

More information

EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE. Leszek Gasiński

EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE. Leszek Gasiński DISCRETE AND CONTINUOUS Website: www.aimsciences.org DYNAMICAL SYSTEMS SUPPLEMENT 2007 pp. 409 418 EXISTENCE RESULTS FOR QUASILINEAR HEMIVARIATIONAL INEQUALITIES AT RESONANCE Leszek Gasiński Jagiellonian

More information

CONVERGENCE OF FINITE DIFFERENCE METHOD FOR THE GENERALIZED SOLUTIONS OF SOBOLEV EQUATIONS

CONVERGENCE OF FINITE DIFFERENCE METHOD FOR THE GENERALIZED SOLUTIONS OF SOBOLEV EQUATIONS J. Korean Math. Soc. 34 (1997), No. 3, pp. 515 531 CONVERGENCE OF FINITE DIFFERENCE METHOD FOR THE GENERALIZED SOLUTIONS OF SOBOLEV EQUATIONS S. K. CHUNG, A.K.PANI AND M. G. PARK ABSTRACT. In this paper,

More information

Boundary layers in a two-point boundary value problem with fractional derivatives

Boundary layers in a two-point boundary value problem with fractional derivatives Boundary layers in a two-point boundary value problem with fractional derivatives J.L. Gracia and M. Stynes Institute of Mathematics and Applications (IUMA) and Department of Applied Mathematics, University

More information

Everywhere differentiability of infinity harmonic functions

Everywhere differentiability of infinity harmonic functions Everywhere differentiability of infinity harmonic functions Lawrence C. Evans and Charles K. Smart Department of Mathematics University of California, Berkeley Abstract We show that an infinity harmonic

More information

On solving linear systems arising from Shishkin mesh discretizations

On solving linear systems arising from Shishkin mesh discretizations On solving linear systems arising from Shishkin mesh discretizations Petr Tichý Faculty of Mathematics and Physics, Charles University joint work with Carlos Echeverría, Jörg Liesen, and Daniel Szyld October

More information

in Bounded Domains Ariane Trescases CMLA, ENS Cachan

in Bounded Domains Ariane Trescases CMLA, ENS Cachan CMLA, ENS Cachan Joint work with Yan GUO, Chanwoo KIM and Daniela TONON International Conference on Nonlinear Analysis: Boundary Phenomena for Evolutionnary PDE Academia Sinica December 21, 214 Outline

More information

Piecewise Smooth Solutions to the Burgers-Hilbert Equation

Piecewise Smooth Solutions to the Burgers-Hilbert Equation Piecewise Smooth Solutions to the Burgers-Hilbert Equation Alberto Bressan and Tianyou Zhang Department of Mathematics, Penn State University, University Park, Pa 68, USA e-mails: bressan@mathpsuedu, zhang

More information

Centrum voor Wiskunde en Informatica

Centrum voor Wiskunde en Informatica Centrum voor Wiskunde en Informatica Modelling, Analysis and Simulation Modelling, Analysis and Simulation High-order time-accurate schemes for singularly perturbed parabolic convection-diffusion problems

More information

A first-order system Petrov-Galerkin discretisation for a reaction-diffusion problem on a fitted mesh

A first-order system Petrov-Galerkin discretisation for a reaction-diffusion problem on a fitted mesh A first-order system Petrov-Galerkin discretisation for a reaction-diffusion problem on a fitted mesh James Adler, Department of Mathematics Tufts University Medford, MA 02155 Scott MacLachlan Department

More information

UNIFORMLY CONVERGENT 3-TGFEM VS LSFEM FOR SINGULARLY PERTURBED CONVECTION-DIFFUSION PROBLEMS ON A SHISHKIN BASED LOGARITHMIC MESH

UNIFORMLY CONVERGENT 3-TGFEM VS LSFEM FOR SINGULARLY PERTURBED CONVECTION-DIFFUSION PROBLEMS ON A SHISHKIN BASED LOGARITHMIC MESH INTERNATIONAL JOURNAL OF NUMERICAL ANALYSIS AND MODELING, SERIES B Volume 4, Number 4, Pages 35 334 c 23 Institute for Scientific Computing and Information UNIFORMLY CONVERGENT 3-TGFEM VS LSFEM FOR SINGULARLY

More information

MATH 220: PROBLEM SET 1, SOLUTIONS DUE FRIDAY, OCTOBER 2, 2015

MATH 220: PROBLEM SET 1, SOLUTIONS DUE FRIDAY, OCTOBER 2, 2015 MATH 220: PROBLEM SET 1, SOLUTIONS DUE FRIDAY, OCTOBER 2, 2015 Problem 1 Classify the following PDEs by degree of non-linearity (linear, semilinear, quasilinear, fully nonlinear: (1 (cos x u x + u y =

More information

Math 564 Homework 1. Solutions.

Math 564 Homework 1. Solutions. Math 564 Homework 1. Solutions. Problem 1. Prove Proposition 0.2.2. A guide to this problem: start with the open set S = (a, b), for example. First assume that a >, and show that the number a has the properties

More information

ASYMPTOTIC METHODS IN MECHANICS

ASYMPTOTIC METHODS IN MECHANICS ASYMPTOTIC METHODS IN MECHANICS by I. Argatov and G. Mishuris Aberystwyth University 2011 Contents Introduction............................... 4 1 Asymptotic expansions 5 1.1 Introduction to asymptotic

More information

NUMERICAL ANALYSIS OF THE CAHN-HILLIARD EQUATION AND APPROXIMATION FOR THE HELE-SHAW PROBLEM, PART II: ERROR ANALYSIS AND CONVERGENCE OF THE INTERFACE

NUMERICAL ANALYSIS OF THE CAHN-HILLIARD EQUATION AND APPROXIMATION FOR THE HELE-SHAW PROBLEM, PART II: ERROR ANALYSIS AND CONVERGENCE OF THE INTERFACE NUMERICAL ANALYSIS OF THE CAHN-HILLIARD EQUATION AND APPROXIMATION FOR THE HELE-SHAW PROBLEM, PART II: ERROR ANALYSIS AND CONVERGENCE OF THE INTERFACE XIAOBING FENG AND ANDREAS PROHL Abstract. In this

More information

Perturbation Methods

Perturbation Methods Perturbation Methods GM01 Dr. Helen J. Wilson Autumn Term 2008 2 1 Introduction 1.1 What are perturbation methods? Many physical processes are described by equations which cannot be solved analytically.

More information

arxiv: v1 [nlin.ps] 18 Sep 2008

arxiv: v1 [nlin.ps] 18 Sep 2008 Asymptotic two-soliton solutions solutions in the Fermi-Pasta-Ulam model arxiv:0809.3231v1 [nlin.ps] 18 Sep 2008 Aaron Hoffman and C.E. Wayne Boston University Department of Mathematics and Statistics

More information

for all subintervals I J. If the same is true for the dyadic subintervals I D J only, we will write ϕ BMO d (J). In fact, the following is true

for all subintervals I J. If the same is true for the dyadic subintervals I D J only, we will write ϕ BMO d (J). In fact, the following is true 3 ohn Nirenberg inequality, Part I A function ϕ L () belongs to the space BMO() if sup ϕ(s) ϕ I I I < for all subintervals I If the same is true for the dyadic subintervals I D only, we will write ϕ BMO

More information

A uniformly convergent numerical method for singularly perturbed nonlinear eigenvalue problems

A uniformly convergent numerical method for singularly perturbed nonlinear eigenvalue problems A uniformly convergent numerical method for singularly perturbed nonlinear eigenvalue problems Weizhu Bao and Ming-Huang Chai Abstract In this paper we propose a uniformly convergent numerical method for

More information

e x = 1 + x + x2 2! + x3 If the function f(x) can be written as a power series on an interval I, then the power series is of the form

e x = 1 + x + x2 2! + x3 If the function f(x) can be written as a power series on an interval I, then the power series is of the form Taylor Series Given a function f(x), we would like to be able to find a power series that represents the function. For example, in the last section we noted that we can represent e x by the power series

More information

A posteriori error estimates for the adaptivity technique for the Tikhonov functional and global convergence for a coefficient inverse problem

A posteriori error estimates for the adaptivity technique for the Tikhonov functional and global convergence for a coefficient inverse problem A posteriori error estimates for the adaptivity technique for the Tikhonov functional and global convergence for a coefficient inverse problem Larisa Beilina Michael V. Klibanov December 18, 29 Abstract

More information

High-Gain Observers in Nonlinear Feedback Control

High-Gain Observers in Nonlinear Feedback Control High-Gain Observers in Nonlinear Feedback Control Lecture # 1 Introduction & Stabilization High-Gain ObserversinNonlinear Feedback ControlLecture # 1Introduction & Stabilization p. 1/4 Brief History Linear

More information

Sample Final Questions: Solutions Math 21B, Winter y ( y 1)(1 + y)) = A y + B

Sample Final Questions: Solutions Math 21B, Winter y ( y 1)(1 + y)) = A y + B Sample Final Questions: Solutions Math 2B, Winter 23. Evaluate the following integrals: tan a) y y dy; b) x dx; c) 3 x 2 + x dx. a) We use partial fractions: y y 3 = y y ) + y)) = A y + B y + C y +. Putting

More information

A FINITE DIFFERENCE DOMAIN DECOMPOSITION ALGORITHM FOR NUMERICAL SOLUTION OF THE HEAT EQUATION

A FINITE DIFFERENCE DOMAIN DECOMPOSITION ALGORITHM FOR NUMERICAL SOLUTION OF THE HEAT EQUATION mathematics of computation volume 57, number 195 july 1991, pages 63-71 A FINITE DIFFERENCE DOMAIN DECOMPOSITION ALGORITHM FOR NUMERICAL SOLUTION OF THE HEAT EQUATION CLINT N. DAWSON, QIANG DU, AND TODD

More information

Laplace s Equation. Chapter Mean Value Formulas

Laplace s Equation. Chapter Mean Value Formulas Chapter 1 Laplace s Equation Let be an open set in R n. A function u C 2 () is called harmonic in if it satisfies Laplace s equation n (1.1) u := D ii u = 0 in. i=1 A function u C 2 () is called subharmonic

More information

1 Lyapunov theory of stability

1 Lyapunov theory of stability M.Kawski, APM 581 Diff Equns Intro to Lyapunov theory. November 15, 29 1 1 Lyapunov theory of stability Introduction. Lyapunov s second (or direct) method provides tools for studying (asymptotic) stability

More information

An extended complete Chebyshev system of 3 Abelian integrals related to a non-algebraic Hamiltonian system

An extended complete Chebyshev system of 3 Abelian integrals related to a non-algebraic Hamiltonian system Computational Methods for Differential Equations http://cmde.tabrizu.ac.ir Vol. 6, No. 4, 2018, pp. 438-447 An extended complete Chebyshev system of 3 Abelian integrals related to a non-algebraic Hamiltonian

More information

or E ( U(X) ) e zx = e ux e ivx = e ux( cos(vx) + i sin(vx) ), B X := { u R : M X (u) < } (4)

or E ( U(X) ) e zx = e ux e ivx = e ux( cos(vx) + i sin(vx) ), B X := { u R : M X (u) < } (4) :23 /4/2000 TOPIC Characteristic functions This lecture begins our study of the characteristic function φ X (t) := Ee itx = E cos(tx)+ie sin(tx) (t R) of a real random variable X Characteristic functions

More information

High order geometric smoothness for conservation laws

High order geometric smoothness for conservation laws High order geometric smoothness for conservation laws Martin Campos-Pinto, Albert Cohen, and Pencho Petrushev Abstract The smoothness of the solutions of 1D scalar conservation laws is investigated and

More information

Equilibria with a nontrivial nodal set and the dynamics of parabolic equations on symmetric domains

Equilibria with a nontrivial nodal set and the dynamics of parabolic equations on symmetric domains Equilibria with a nontrivial nodal set and the dynamics of parabolic equations on symmetric domains J. Földes Department of Mathematics, Univerité Libre de Bruxelles 1050 Brussels, Belgium P. Poláčik School

More information

MATH 205C: STATIONARY PHASE LEMMA

MATH 205C: STATIONARY PHASE LEMMA MATH 205C: STATIONARY PHASE LEMMA For ω, consider an integral of the form I(ω) = e iωf(x) u(x) dx, where u Cc (R n ) complex valued, with support in a compact set K, and f C (R n ) real valued. Thus, I(ω)

More information

I. The space C(K) Let K be a compact metric space, with metric d K. Let B(K) be the space of real valued bounded functions on K with the sup-norm

I. The space C(K) Let K be a compact metric space, with metric d K. Let B(K) be the space of real valued bounded functions on K with the sup-norm I. The space C(K) Let K be a compact metric space, with metric d K. Let B(K) be the space of real valued bounded functions on K with the sup-norm Proposition : B(K) is complete. f = sup f(x) x K Proof.

More information

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games

Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games Stability of Feedback Solutions for Infinite Horizon Noncooperative Differential Games Alberto Bressan ) and Khai T. Nguyen ) *) Department of Mathematics, Penn State University **) Department of Mathematics,

More information

OPTIMALITY CONDITIONS AND ERROR ANALYSIS OF SEMILINEAR ELLIPTIC CONTROL PROBLEMS WITH L 1 COST FUNCTIONAL

OPTIMALITY CONDITIONS AND ERROR ANALYSIS OF SEMILINEAR ELLIPTIC CONTROL PROBLEMS WITH L 1 COST FUNCTIONAL OPTIMALITY CONDITIONS AND ERROR ANALYSIS OF SEMILINEAR ELLIPTIC CONTROL PROBLEMS WITH L 1 COST FUNCTIONAL EDUARDO CASAS, ROLAND HERZOG, AND GERD WACHSMUTH Abstract. Semilinear elliptic optimal control

More information

A Parameter-uniform Method for Two Parameters Singularly Perturbed Boundary Value Problems via Asymptotic Expansion

A Parameter-uniform Method for Two Parameters Singularly Perturbed Boundary Value Problems via Asymptotic Expansion Appl. Math. Inf. Sci. 7, No. 4, 1525-1532 (2013) 1525 Applied Mathematics & Information Sciences An International Journal http://dx.doi.org/10.12785/amis/070436 A Parameter-uniform Method for Two Parameters

More information

Stochastic homogenization 1

Stochastic homogenization 1 Stochastic homogenization 1 Tuomo Kuusi University of Oulu August 13-17, 2018 Jyväskylä Summer School 1 Course material: S. Armstrong & T. Kuusi & J.-C. Mourrat : Quantitative stochastic homogenization

More information

Hyperbolic Systems of Conservation Laws

Hyperbolic Systems of Conservation Laws Hyperbolic Systems of Conservation Laws III - Uniqueness and continuous dependence and viscous approximations Alberto Bressan Mathematics Department, Penn State University http://www.math.psu.edu/bressan/

More information

Iterative Methods for Linear Systems

Iterative Methods for Linear Systems Iterative Methods for Linear Systems 1. Introduction: Direct solvers versus iterative solvers In many applications we have to solve a linear system Ax = b with A R n n and b R n given. If n is large the

More information

On Smoothness of Suitable Weak Solutions to the Navier-Stokes Equations

On Smoothness of Suitable Weak Solutions to the Navier-Stokes Equations On Smoothness of Suitable Weak Solutions to the Navier-Stokes Equations G. Seregin, V. Šverák Dedicated to Vsevolod Alexeevich Solonnikov Abstract We prove two sufficient conditions for local regularity

More information

THEOREM OF OSELEDETS. We recall some basic facts and terminology relative to linear cocycles and the multiplicative ergodic theorem of Oseledets [1].

THEOREM OF OSELEDETS. We recall some basic facts and terminology relative to linear cocycles and the multiplicative ergodic theorem of Oseledets [1]. THEOREM OF OSELEDETS We recall some basic facts and terminology relative to linear cocycles and the multiplicative ergodic theorem of Oseledets []. 0.. Cocycles over maps. Let µ be a probability measure

More information

Course 224: Geometry - Continuity and Differentiability

Course 224: Geometry - Continuity and Differentiability Course 224: Geometry - Continuity and Differentiability Lecture Notes by Prof. David Simms L A TEXed by Chris Blair 1 Continuity Consider M, N finite dimensional real or complex vector spaces. What does

More information

On finite time BV blow-up for the p-system

On finite time BV blow-up for the p-system On finite time BV blow-up for the p-system Alberto Bressan ( ), Geng Chen ( ), and Qingtian Zhang ( ) (*) Department of Mathematics, Penn State University, (**) Department of Mathematics, University of

More information

First, let me recall the formula I want to prove. Again, ψ is the function. ψ(x) = n<x

First, let me recall the formula I want to prove. Again, ψ is the function. ψ(x) = n<x 8.785: Analytic Number heory, MI, spring 007 (K.S. Kedlaya) von Mangoldt s formula In this unit, we derive von Mangoldt s formula estimating ψ(x) x in terms of the critical zeroes of the Riemann zeta function.

More information

Some recent results on controllability of coupled parabolic systems: Towards a Kalman condition

Some recent results on controllability of coupled parabolic systems: Towards a Kalman condition Some recent results on controllability of coupled parabolic systems: Towards a Kalman condition F. Ammar Khodja Clermont-Ferrand, June 2011 GOAL: 1 Show the important differences between scalar and non

More information

MAS153/MAS159. MAS153/MAS159 1 Turn Over SCHOOL OF MATHEMATICS AND STATISTICS hours. Mathematics (Materials) Mathematics For Chemists

MAS153/MAS159. MAS153/MAS159 1 Turn Over SCHOOL OF MATHEMATICS AND STATISTICS hours. Mathematics (Materials) Mathematics For Chemists Data provided: Formula sheet MAS53/MAS59 SCHOOL OF MATHEMATICS AND STATISTICS Mathematics (Materials Mathematics For Chemists Spring Semester 203 204 3 hours All questions are compulsory. The marks awarded

More information

Analysis of a Streamline-Diffusion Finite Element Method on Bakhvalov-Shishkin Mesh for Singularly Perturbed Problem

Analysis of a Streamline-Diffusion Finite Element Method on Bakhvalov-Shishkin Mesh for Singularly Perturbed Problem Numer. Math. Theor. Meth. Appl. Vol. 10, No. 1, pp. 44-64 doi: 10.408/nmtma.017.y1306 February 017 Analysis of a Streamline-Diffusion Finite Element Method on Bakhvalov-Shishkin Mesh for Singularly Perturbed

More information

An interface problem with singular perturbation on a subinterval

An interface problem with singular perturbation on a subinterval Xie Boundary Value Problems 2014, 2014:201 R E S E A R C H Open Access An interface problem with singular perturbation on a subinterval Feng Xie * * Correspondence: fxie@dhu.edu.cn Department of Applied

More information

Lipschitz Metrics for a Class of Nonlinear Wave Equations

Lipschitz Metrics for a Class of Nonlinear Wave Equations Lipschitz Metrics for a Class of Nonlinear Wave Equations Alberto Bressan and Geng Chen * Department of Mathematics, Penn State University, University Park, Pa 1680, USA ** School of Mathematics Georgia

More information

Applied/Numerical Analysis Qualifying Exam

Applied/Numerical Analysis Qualifying Exam Applied/Numerical Analysis Qualifying Exam August 9, 212 Cover Sheet Applied Analysis Part Policy on misprints: The qualifying exam committee tries to proofread exams as carefully as possible. Nevertheless,

More information

On the Stability of the Best Reply Map for Noncooperative Differential Games

On the Stability of the Best Reply Map for Noncooperative Differential Games On the Stability of the Best Reply Map for Noncooperative Differential Games Alberto Bressan and Zipeng Wang Department of Mathematics, Penn State University, University Park, PA, 68, USA DPMMS, University

More information

GALERKIN TIME STEPPING METHODS FOR NONLINEAR PARABOLIC EQUATIONS

GALERKIN TIME STEPPING METHODS FOR NONLINEAR PARABOLIC EQUATIONS GALERKIN TIME STEPPING METHODS FOR NONLINEAR PARABOLIC EQUATIONS GEORGIOS AKRIVIS AND CHARALAMBOS MAKRIDAKIS Abstract. We consider discontinuous as well as continuous Galerkin methods for the time discretization

More information

Solutions Preliminary Examination in Numerical Analysis January, 2017

Solutions Preliminary Examination in Numerical Analysis January, 2017 Solutions Preliminary Examination in Numerical Analysis January, 07 Root Finding The roots are -,0, a) First consider x 0 > Let x n+ = + ε and x n = + δ with δ > 0 The iteration gives 0 < ε δ < 3, which

More information

Green s function estimates for a singularly perturbed convection-diffusion problem

Green s function estimates for a singularly perturbed convection-diffusion problem Green s function estimates for a singularly perturbed convection-diffusion problem Sebastian Franz Natalia Kopteva Abstract We consider a singularly perturbed convection-diffusion problem posed in the

More information

Structurally Stable Singularities for a Nonlinear Wave Equation

Structurally Stable Singularities for a Nonlinear Wave Equation Structurally Stable Singularities for a Nonlinear Wave Equation Alberto Bressan, Tao Huang, and Fang Yu Department of Mathematics, Penn State University University Park, Pa. 1682, U.S.A. e-mails: bressan@math.psu.edu,

More information