Home Random quiz Cards

Chapter 11

Boundary-Value Problems for Ordinary Differential Equations

Rendered from the uploaded full LaTeX source with KaTeX. No PDF, no PDF.js, no image extraction, and no raw LaTeX displayed.

Learning goals

After completing this chapter, the reader should be able to:

  1. formulate linear and nonlinear two-point boundary-value problems;

  2. distinguish initial-value problems from boundary-value problems;

  3. implement shooting and multiple-shooting methods;

  4. derive finite-difference discretizations for linear and nonlinear BVPs;

  5. solve nonlinear finite-difference systems by Newton’s method;

  6. formulate collocation methods, including polynomial and Chebyshev collocation;

  7. derive weak formulations for self-adjoint linear BVPs;

  8. build Galerkin approximations for linear BVPs;

  9. build Newton–Galerkin methods for nonlinear BVPs;

  10. assemble finite-element stiffness matrices and load vectors;

  11. evaluate Galerkin integrals using high-order Gauss–Legendre quadrature;

  12. understand adaptive refinement, residual estimators, and practical solver choices.

Boundary-Value Problems

A second-order two-point boundary-value problem has the form y''=f(x,y,y'), \qquad a<x<b, with boundary conditions such as y(a)=\alpha, \qquad y(b)=\beta. These are Dirichlet boundary conditions. More general boundary conditions include Neumann conditions y'(a)=\gamma, \qquad y'(b)=\delta, and Robin conditions \eta_a y(a)+\mu_a y'(a)=\gamma_a, \qquad \eta_b y(b)+\mu_b y'(b)=\gamma_b.

Key point: Central difference from IVPs

An initial-value problem evolves from complete data at one point. A boundary-value problem imposes conditions at two or more points. This changes the numerical structure: BVP discretizations usually produce algebraic systems, often nonlinear.

A standard linear model problem is -y''(x)+q(x)y(x)=g(x), \qquad y(a)=\alpha,\quad y(b)=\beta. A self-adjoint variable-coefficient model is -\frac{\dd}{\dd x}\left(p(x)y'(x)\right)+q(x)y(x)=g(x), \qquad y(a)=\alpha,\quad y(b)=\beta, where p(x)\ge p_0>0, \qquad q(x)\ge0.

Well-Posedness and Maximum Principles

For linear BVPs, numerical methods are meaningful only when the continuous problem is well posed.

Theorem: Uniqueness for a coercive Dirichlet problem

Suppose p(x)\ge p_0>0, \qquad q(x)\ge0 on [a,b]. Then the homogeneous problem -\frac{\dd}{\dd x}\left(p(x)y'(x)\right)+q(x)y(x)=0, \qquad y(a)=0,\quad y(b)=0, has only the trivial solution.

Proof

Multiply the equation by y and integrate: \int_a^b \left[ -\left(p y'\right)'y+q y^2 \right]\dd x=0. Integration by parts gives \int_a^b p(x)|y'(x)|^2\,\dd x + \int_a^b q(x)|y(x)|^2\,\dd x = \left[-p(x)y'(x)y(x)\right]_a^b. The boundary term is zero because y(a)=y(b)=0. Hence \int_a^b p|y'|^2\,\dd x+\int_a^b q|y|^2\,\dd x=0. Both integrals are nonnegative. Since p\ge p_0>0, we obtain y'=0 almost everywhere. Thus y is constant. The boundary condition gives y=0.

Shooting Method

Consider y''=f(x,y,y'), \qquad y(a)=\alpha, \qquad y(b)=\beta. Introduce u_1=y, \qquad u_2=y'. Then u_1'=u_2, \qquad u_2'=f(x,u_1,u_2). The missing initial slope is s=y'(a). For each s, solve the IVP u_1(a)=\alpha, \qquad u_2(a)=s. Define the shooting residual F(s)=u_1(b;s)-\beta. The BVP is solved by finding F(s)=0.

Algorithm
Caption.

Single shooting method

  1. BVP y''=f(x,y,y'), boundary values \alpha,\beta, initial slope guess s_0

  2. For k=0,1,2,\ldots:

  3. Solve the IVP with y(a)=\alpha, y'(a)=s_k

  4. Compute F(s_k)=y(b;s_k)-\beta

  5. If |F(s_k)| is small enough:

  6. Return y(x;s_k)

  7. Update s_k using Newton, secant, or bisection

Theorem: Newton shooting sensitivity equation

For the nonlinear BVP y''=f(x,y,y'), let y(x;s) be the shooting solution with y(a;s)=\alpha, \qquad y'(a;s)=s. The derivative z(x;s)=\frac{\partial y(x;s)}{\partial s} satisfies z''=f_y(x,y,y')z+f_{y'}(x,y,y')z', \qquad z(a)=0, \qquad z'(a)=1. Therefore F'(s)=z(b;s).

Proof

Differentiate y''(x;s)=f(x,y(x;s),y'(x;s)) with respect to s. This gives z''=f_yz+f_{y'}z'. Differentiating the initial conditions y(a;s)=\alpha, \qquad y'(a;s)=s with respect to s yields z(a)=0, \qquad z'(a)=1. Since F(s)=y(b;s)-\beta, we have F'(s)=z(b;s).

Warning: Shooting can be unstable

Shooting may be ill-conditioned when the associated IVP has exponentially growing modes. Small changes in the initial slope can create huge changes at the right boundary. Multiple shooting and collocation are often more robust.

Multiple Shooting

Multiple shooting divides [a,b] into subintervals a=x_0<x_1<\cdots<x_M=b. On each subinterval, an IVP is solved with unknown initial data. Continuity conditions are imposed at the internal nodes: y_j(x_j)=Y_j, y_j(x_{j+1})=Y_{j+1}. The unknowns are the interface values and possibly slopes. The resulting nonlinear system is solved by Newton’s method.

Multiple shooting reduces sensitivity because each IVP is solved on a shorter interval.

Finite Differences for Linear BVPs

Consider -y''(x)=g(x), \qquad y(a)=\alpha, \qquad y(b)=\beta. Let x_i=a+ih, \qquad i=0,\ldots,N, \qquad h=\frac{b-a}{N}. For interior points, y''(x_i) \approx \frac{y_{i-1}-2y_i+y_{i+1}}{h^2}. Thus -\frac{y_{i-1}-2y_i+y_{i+1}}{h^2}=g(x_i), \qquad i=1,\ldots,N-1. This produces a tridiagonal linear system.

Theorem: Second-order finite-difference consistency

If y\in C^4([a,b]), then \frac{y(x_i-h)-2y(x_i)+y(x_i+h)}{h^2} = y''(x_i)+O(h^2).

Proof

Taylor expansions give y(x_i+h)=y(x_i)+hy'(x_i)+\frac{h^2}{2}y''(x_i) +\frac{h^3}{6}y^{(3)}(x_i)+\frac{h^4}{24}y^{(4)}(\xi_+), and y(x_i-h)=y(x_i)-hy'(x_i)+\frac{h^2}{2}y''(x_i) -\frac{h^3}{6}y^{(3)}(x_i)+\frac{h^4}{24}y^{(4)}(\xi_-). Adding and rearranging yields \frac{y(x_i-h)-2y(x_i)+y(x_i+h)}{h^2} = y''(x_i)+O(h^2).

Algorithm
Caption.

Finite-difference method for -y''=g

  1. interval [a,b], boundary values \alpha,\beta, number of panels N

  2. h\gets(b-a)/N

  3. For i=1,\ldots,N-1:

  4. Set row i of the tridiagonal system:

  5. \item -Y_{i-1}+2Y_i-Y_{i+1}=h^2 g(x_i) \item

  6. Move known boundary terms Y_0=\alpha, Y_N=\beta to the right-hand side

  7. Solve the tridiagonal system for Y_1,\ldots,Y_{N-1}

  8. Return Y_0,\ldots,Y_N

Finite Differences for Nonlinear BVPs

For y''=f(x,y,y'), \qquad y(a)=\alpha, \qquad y(b)=\beta, use y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}, and y'(x_i)\approx \frac{Y_{i+1}-Y_{i-1}}{2h}. The nonlinear equations are R_i(Y)= \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2} - f\left( x_i,Y_i,\frac{Y_{i+1}-Y_{i-1}}{2h} \right) = 0.

Newton’s method solves J(Y^{(k)})\Delta Y^{(k)} = -R(Y^{(k)}), then updates Y^{(k+1)}=Y^{(k)}+\Delta Y^{(k)}.

Algorithm
Caption.

Newton finite-difference method for nonlinear BVPs

  1. mesh x_i, boundary values, initial guess Y^{(0)}

  2. For k=0,1,2,\ldots:

  3. Assemble nonlinear residual R(Y^{(k)})

  4. Assemble Jacobian J(Y^{(k)})

  5. Solve J(Y^{(k)})\Delta Y^{(k)}=-R(Y^{(k)})

  6. Y^{(k+1)}\gets Y^{(k)}+\Delta Y^{(k)}

  7. If \|R(Y^{(k+1)})\| and \|\Delta Y^{(k)}\| are small:

  8. Return Y^{(k+1)}

Collocation Methods

Collocation methods approximate the solution by a polynomial or piecewise polynomial u_h and enforce the differential equation at selected points: \mathcal L u_h(x_j)=g(x_j). For a nonlinear operator, \mathcal N(u_h)(x_j)=0.

For example, on [a,b], choose a polynomial u_N(x)=\sum_{k=0}^{N} c_k\phi_k(x) and impose u_N(a)=\alpha, \qquad u_N(b)=\beta, and u_N''(\xi_j)=f(\xi_j,u_N(\xi_j),u_N'(\xi_j)), \qquad j=1,\ldots,N-1.

Chebyshev Spectral Collocation

Map [a,b] to [-1,1] by x=\frac{b-a}{2}\xi+\frac{a+b}{2}. Use Chebyshev–Gauss–Lobatto nodes \xi_j=\cos\left(\frac{\pi j}{N}\right), \qquad j=0,\ldots,N. Let D_\xi be the Chebyshev differentiation matrix on [-1,1]. Then D_x=\frac{2}{b-a}D_\xi, \qquad D_x^{(2)}=\left(\frac{2}{b-a}\right)^2D_\xi^2. For the linear problem -y''+q(x)y=g(x), \qquad y(a)=\alpha, \qquad y(b)=\beta, the interior collocation equations are -\left(D_x^{(2)}Y\right)_j+q(x_j)Y_j=g(x_j), \qquad j=1,\ldots,N-1, with boundary equations Y(a)=\alpha, \qquad Y(b)=\beta.

Theorem: Spectral accuracy for analytic BVP solutions

If the solution of a linear BVP is analytic in a complex neighborhood of [a,b] and the boundary-value problem is well conditioned, then Chebyshev collocation can converge faster than any algebraic power of N^{-1}.

Proof

Analyticity implies geometric decay of Chebyshev coefficients. The Chebyshev interpolant and its differentiated approximations converge rapidly in appropriate norms. Since the differential operator is imposed at the collocation nodes and the continuous BVP is assumed well conditioned, a small collocation residual implies a small solution error. Therefore the convergence is spectral until limited by roundoff or conditioning.

Figure 11.1 Shooting method for BVPs The missing initial slope is tuned until the far boundary is satisfied.
Open visual gallery

Weak Formulation for Linear BVPs

Consider the homogeneous Dirichlet problem -\frac{\dd}{\dd x}\left(p(x)y'(x)\right)+q(x)y(x)=g(x), \qquad y(a)=0,\quad y(b)=0. Let V=H_0^1(a,b). Multiply by a test function v\in V and integrate: \int_a^b \left[ -\left(p y'\right)'v+qyv \right]\dd x = \int_a^b gv\,\dd x. Integrating by parts gives the weak formulation: \int_a^b p y'v'\,\dd x + \int_a^b q yv\,\dd x = \int_a^b gv\,\dd x \qquad \forall v\in V. Define a(y,v)= \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x, and \ell(v)=\int_a^b gv\,\dd x. The weak problem is: \text{Find }y\in V \text{ such that } a(y,v)=\ell(v) \quad \forall v\in V.

Theorem: Coercivity of the linear weak form

If p(x)\ge p_0>0 and q(x)\ge0, then a(v,v)\ge c\|v\|_{H^1(a,b)}^2 \qquad \forall v\in H_0^1(a,b), for some c>0.

Proof

We have a(v,v)=\int_a^b p|v'|^2\,\dd x+\int_a^b q|v|^2\,\dd x \ge p_0\|v'\|_{L^2}^2. By Poincare’s inequality on H_0^1(a,b), \|v\|_{L^2}\le C_P\|v'\|_{L^2}. Hence \|v\|_{H^1}^2 = \|v\|_{L^2}^2+\|v'\|_{L^2}^2 \le (C_P^2+1)\|v'\|_{L^2}^2. Therefore a(v,v)\ge \frac{p_0}{C_P^2+1}\|v\|_{H^1}^2.

Linear Galerkin Method

Let V_h\subset V be a finite-dimensional trial and test space with basis \phi_1,\ldots,\phi_m. The Galerkin approximation is: \text{Find }y_h\in V_h \text{ such that } a(y_h,v_h)=\ell(v_h) \quad \forall v_h\in V_h. Write y_h=\sum_{j=1}^{m}Y_j\phi_j. Testing with v_h=\phi_i gives \sum_{j=1}^{m} a(\phi_j,\phi_i)Y_j = \ell(\phi_i), \qquad i=1,\ldots,m. Thus KY=F, where K_{ij}=a(\phi_j,\phi_i) = \int_a^b p\phi_j'\phi_i'\,\dd x + \int_a^b q\phi_j\phi_i\,\dd x, and F_i=\int_a^b g\phi_i\,\dd x.

Theorem: Galerkin orthogonality

Let y\in V solve the weak problem and y_h\in V_h solve the Galerkin problem. Then a(y-y_h,v_h)=0 \qquad \forall v_h\in V_h.

Proof

The exact solution satisfies a(y,v_h)=\ell(v_h) \qquad \forall v_h\in V_h. The Galerkin solution satisfies a(y_h,v_h)=\ell(v_h) \qquad \forall v_h\in V_h. Subtracting gives a(y-y_h,v_h)=0.

Theorem: Cea-type best approximation estimate

Assume a(\cdot,\cdot) is continuous and coercive on V: |a(u,v)|\le M\|u\|_V\|v\|_V, \qquad a(v,v)\ge \alpha\|v\|_V^2. Then \|y-y_h\|_V \le \frac{M}{\alpha} \inf_{v_h\in V_h}\|y-v_h\|_V.

Proof

For any v_h\in V_h, coercivity gives \alpha\|y-y_h\|_V^2 \le a(y-y_h,y-y_h). Using Galerkin orthogonality with y_h-v_h\in V_h, a(y-y_h,y_h-v_h)=0. Therefore a(y-y_h,y-y_h)=a(y-y_h,y-v_h). By continuity, a(y-y_h,y-v_h) \le M\|y-y_h\|_V\|y-v_h\|_V. Canceling \|y-y_h\|_V and taking the infimum over v_h gives the result.

Finite Elements with Piecewise Linear Basis Functions

Let a=x_0<x_1<\cdots<x_N=b. The piecewise linear hat basis functions satisfy \phi_i(x_j)=\delta_{ij}. For homogeneous Dirichlet conditions, the unknown basis functions are \phi_1,\ldots,\phi_{N-1}.

On an element e=[x_i,x_{i+1}], with length h_e=x_{i+1}-x_i, the local basis functions are \varphi_1(x)=\frac{x_{i+1}-x}{h_e}, \qquad \varphi_2(x)=\frac{x-x_i}{h_e}. Their derivatives are \varphi_1'(x)=-\frac1{h_e}, \qquad \varphi_2'(x)=\frac1{h_e}. For constant p, the local stiffness matrix is K^{(e)} = \frac{p}{h_e} \begin{pmatrix} 1&-1\\ -1&1 \end{pmatrix}. For constant q, the local mass matrix is M^{(e)} = \frac{q h_e}{6} \begin{pmatrix} 2&1\\ 1&2 \end{pmatrix}.

Algorithm
Caption.

Finite-element assembly for a linear BVP

  1. mesh x_0,\ldots,x_N, coefficients p,q, source g

  2. Initialize global matrix K and vector F to zero

  3. For each element e=[x_i,x_{i+1}]:

  4. Compute local stiffness entries

  5. \item K_{mn}^{(e)}=\int_{x_i}^{x_{i+1}} \item \left(p\varphi_n'\varphi_m'+q\varphi_n\varphi_m\right)\dd x \item

  6. Compute local load entries

  7. \item F_m^{(e)}=\int_{x_i}^{x_{i+1}}g\varphi_m\,\dd x \item

  8. Add local entries to global matrix and vector

  9. Apply boundary conditions

  10. Solve the resulting linear system

High-Precision Quadrature for Galerkin Integrals

Galerkin and finite-element methods require integrals such as \int_{x_i}^{x_{i+1}} p(x)\varphi_m'(x)\varphi_n'(x)\,\dd x, \int_{x_i}^{x_{i+1}} q(x)\varphi_m(x)\varphi_n(x)\,\dd x, and \int_{x_i}^{x_{i+1}} g(x)\varphi_m(x)\,\dd x. If the coefficients are not constant, these integrals must be approximated accurately.

Use the reference element \xi\in[-1,1], \qquad x=x_c+\frac{h_e}{2}\xi, \qquad x_c=\frac{x_i+x_{i+1}}{2}. Then \int_{x_i}^{x_{i+1}} F(x)\,\dd x = \frac{h_e}{2} \int_{-1}^{1} F\left(x_c+\frac{h_e}{2}\xi\right)\dd \xi. An r-point Gauss–Legendre rule is \int_{-1}^{1}F(\xi)\,\dd \xi \approx \sum_{\ell=1}^{r}w_\ell F(\xi_\ell), ] and is exact for all polynomials of degree at most \[ 2r-1.

Explicit low-order Gauss–Legendre rules are: r=1: \qquad \xi_1=0, \qquad w_1=2. r=2: \qquad \xi_{1,2}=\pm\frac1{\sqrt3}, \qquad w_1=w_2=1. r=3: \qquad \xi_1=0,\quad \xi_{2,3}=\pm\sqrt{\frac35}, w_1=\frac89, \qquad w_2=w_3=\frac59. r=4: \qquad \xi_{1,2}=\pm \sqrt{\frac{3}{7}-\frac{2}{7}\sqrt{\frac65}}, \qquad \xi_{3,4}=\pm \sqrt{\frac{3}{7}+\frac{2}{7}\sqrt{\frac65}}, w_{1,2}= \frac{18+\sqrt{30}}{36}, \qquad w_{3,4}= \frac{18-\sqrt{30}}{36}. r=5: \qquad \xi_1=0, \xi_{2,3}=\pm\frac13\sqrt{5-2\sqrt{\frac{10}{7}}}, \qquad \xi_{4,5}=\pm\frac13\sqrt{5+2\sqrt{\frac{10}{7}}}, w_1=\frac{128}{225}, w_{2,3}=\frac{322+13\sqrt{70}}{900}, \qquad w_{4,5}=\frac{322-13\sqrt{70}}{900}.

Theorem: Polynomial exactness of Gauss quadrature

An r-point Gauss–Legendre rule on [-1,1] integrates exactly every polynomial of degree at most 2r-1.

Proof

The Gauss nodes are the roots of the Legendre polynomial P_r, which is orthogonal to all polynomials of degree at most r-1. For any polynomial p of degree at most 2r-1, divide by P_r: p=P_r q+s, \qquad \deg q\le r-1, \qquad \deg s\le r-1. The quadrature rule is interpolatory on r nodes, so it integrates s exactly. At the nodes, P_r=0, so the quadrature of P_rq is zero. The exact integral of P_rq is also zero by orthogonality. Therefore the rule integrates p exactly.

Chapter summary: Quadrature choice for Galerkin integrals

For piecewise polynomials of degree k: \varphi_i\varphi_j \quad \text{has degree }2k, and \varphi_i'\varphi_j' \quad \text{has degree }2k-2. If coefficients p,q,g are polynomial, choose Gauss order r so that 2r-1 is at least the degree of the full integrand. If coefficients are smooth but not polynomial, increase r until the quadrature error is negligible compared with the discretization error.

Algorithm
Caption.

High-order Gauss quadrature on an element

  1. element [x_i,x_{i+1}], integrand F(x), Gauss nodes \xi_\ell, weights w_\ell

  2. h_e\gets x_{i+1}-x_i, x_c\gets(x_i+x_{i+1})/2

  3. I\gets0

  4. For \ell=1,\ldots,r:

  5. x_\ell\gets x_c+(h_e/2)\xi_\ell

  6. I\gets I+w_\ell F(x_\ell)

  7. Return (h_e/2)I

Nonlinear Galerkin Method

Consider a nonlinear BVP written abstractly as \mathcal N(y)=0 with boundary conditions. A typical example is -y''+q(x)y+\gamma y^3=g(x), \qquad y(a)=0,\quad y(b)=0. Multiplying by v\in V and integrating gives: \int_a^b y'v'\,\dd x + \int_a^b q yv\,\dd x + \int_a^b \gamma y^3v\,\dd x - \int_a^b gv\,\dd x = 0. Define the nonlinear residual R(y;v) = \int_a^b y'v'\,\dd x + \int_a^b q yv\,\dd x + \int_a^b \gamma y^3v\,\dd x - \int_a^b gv\,\dd x. The nonlinear Galerkin method is: \text{Find }y_h\in V_h \text{ such that } R(y_h;v_h)=0 \quad \forall v_h\in V_h. With y_h=\sum_{j=1}^{m}Y_j\phi_j, we obtain nonlinear algebraic equations R_i(Y)=R(y_h;\phi_i)=0, \qquad i=1,\ldots,m.

Newton–Galerkin Linearization

Newton’s method requires the Frechet derivative of the residual. For R(y;v)= \int_a^b y'v'\,\dd x + \int_a^b q yv\,\dd x + \int_a^b \gamma y^3v\,\dd x - \int_a^b gv\,\dd x, the derivative in direction w is R'(y)[w;v] = \int_a^b w'v'\,\dd x + \int_a^b qwv\,\dd x + \int_a^b 3\gamma y^2wv\,\dd x. Newton’s correction \delta_h\in V_h solves R'(y_h^{(k)})[\delta_h;v_h] = -R(y_h^{(k)};v_h) \qquad \forall v_h\in V_h. Then y_h^{(k+1)}=y_h^{(k)}+\delta_h.

Algorithm
Caption.

Newton–Galerkin method for nonlinear BVPs

  1. finite-dimensional space V_h, initial guess y_h^{(0)}

  2. For k=0,1,2,\ldots:

  3. Assemble nonlinear residual vector

  4. \item R_i=R(y_h^{(k)};\phi_i) \item

  5. Assemble Jacobian matrix

  6. \item J_{ij}=R'(y_h^{(k)})[\phi_j;\phi_i] \item

  7. Solve J\Delta Y=-R

  8. Y^{(k+1)}\gets Y^{(k)}+\Delta Y

  9. If \|R\| and \|\Delta Y\| are small enough:

  10. Return y_h^{(k+1)}

Theorem: Newton–Galerkin Jacobian for cubic reaction

For the nonlinear weak residual R(y;v)= \int_a^b y'v'\,\dd x + \int_a^b q yv\,\dd x + \int_a^b \gamma y^3v\,\dd x - \int_a^b gv\,\dd x, the Newton Jacobian matrix in the basis \{\phi_i\} is J_{ij} = \int_a^b \phi_j'\phi_i'\,\dd x + \int_a^b q\phi_j\phi_i\,\dd x + \int_a^b 3\gamma (y_h^{(k)})^2\phi_j\phi_i\,\dd x.

Proof

Differentiate R(y+\epsilon w;v) with respect to \epsilon at \epsilon=0. The derivative of y' is w', the derivative of qy is qw, and the derivative of \gamma y^3 is 3\gamma y^2w. Taking w=\phi_j and v=\phi_i gives the matrix entry.

Quadrature for Nonlinear Galerkin Terms

Nonlinear Galerkin terms such as \int_a^b \gamma (y_h)^3\phi_i\,\dd x and Jacobian terms such as \int_a^b 3\gamma (y_h)^2\phi_j\phi_i\,\dd x must be evaluated accurately.

If y_h is a polynomial of degree k on each element, then (y_h)^3\phi_i has degree at most 4k, and (y_h)^2\phi_j\phi_i has degree at most 4k. Therefore, for polynomial coefficients, an r-point Gauss rule with 2r-1\ge4k integrates these nonlinear cubic terms exactly.

For non-polynomial nonlinearities such as \sin(y_h), \qquad e^{y_h}, \qquad \frac{1}{1+y_h^2}, there is no finite exact Gauss rule. In this case one uses high-order quadrature and checks convergence by increasing r.

Warning: Quadrature error can destroy Galerkin accuracy

If the quadrature rule is too low order, the numerical method is no longer the intended Galerkin method. Underintegration may introduce consistency errors, spurious modes, or incorrect Newton Jacobians.

Petrov–Galerkin and Least-Squares Galerkin Methods

In the standard Galerkin method, the trial and test spaces are the same. In a Petrov–Galerkin method, y_h\in V_h, \qquad v_h\in W_h, where V_h and W_h may differ.

A least-squares Galerkin method minimizes a residual norm: y_h=\arg\min_{v_h\in V_h}\|\mathcal L v_h-g\|_{L^2(a,b)}^2. For a linear operator \mathcal L, the normal equations are (\mathcal L y_h,\mathcal L v_h)=(g,\mathcal L v_h) \qquad \forall v_h\in V_h. For nonlinear problems, y_h=\arg\min_{v_h\in V_h}\|\mathcal N(v_h)\|^2 leads to nonlinear normal equations.

A Posteriori Residual Estimation

Let y_h be an approximate solution of -y''+q y=g. The strong residual on an element is R_e(x)=g(x)+y_h''(x)-q(x)y_h(x). For piecewise linear finite elements, y_h''=0 inside each element, so R_e(x)=g(x)-q(x)y_h(x). Jump residuals measure derivative discontinuities at nodes: J_i= y_h'(x_i^-)-y_h'(x_i^+). A typical residual indicator is \eta_e^2 = h_e^2\|R_e\|_{L^2(e)}^2 + \frac12 h_e \sum_{x_i\in\partial e\cap(a,b)} |J_i|^2.

Adaptive Mesh Refinement

Adaptive finite-element methods follow the loop: \text{solve} \longrightarrow \text{estimate} \longrightarrow \text{mark} \longrightarrow \text{refine}. Elements with large \eta_e are refined. This concentrates degrees of freedom where boundary layers, sharp gradients, or singularities occur.

Algorithm
Caption.

Adaptive Galerkin loop for a BVP

  1. initial mesh \mathcal T_h, tolerance \tau

  2. For k=0,1,2,\ldots:

  3. Solve the Galerkin problem on \mathcal T_h

  4. Compute element indicators \eta_e

  5. If (\sum_e\eta_e^2)^{1/2}\le\tau:

  6. Return y_h

  7. Mark elements with large \eta_e

  8. Refine marked elements

Practical Comparison of BVP Methods

Chapter summary: Choosing a BVP method
Shooting.

Simple and uses IVP solvers. Good for mild scalar problems. Can be unstable for sensitive or stiff BVPs.

Multiple shooting.

More robust than single shooting. Produces larger nonlinear systems but improves conditioning.

Finite differences.

Simple, sparse, and easy for one-dimensional problems. Boundary conditions and nonlinearities are straightforward.

Collocation.

High accuracy and flexible polynomial representation. Nonlinear systems are natural.

Chebyshev collocation.

Spectral accuracy for smooth solutions on simple intervals. Dense matrices and conditioning require care.

Linear Galerkin.

Natural for self-adjoint linear BVPs. Gives symmetric positive definite systems under coercivity assumptions.

Nonlinear Galerkin.

Powerful for nonlinear BVPs. Requires residual and Jacobian assembly, usually with high-order quadrature.

Finite elements.

Excellent for variable coefficients, weak solutions, adaptive refinement, and extension to PDEs.

High-order quadrature.

Essential when coefficients, sources, or nonlinear terms are not exactly integrable by low-order formulas.

Figure 11.2 Finite-difference BVP grid Boundary values enter the algebraic system directly.
Open visual gallery

Exercises

The following exercise bank is intentionally large. Basic problems test definitions, computations, and essential formulas. Starred exercises require proofs, careful discretization, weak formulations, quadrature analysis, or nonlinear solver design. Problems marked \star, \star\star, and \star\star\star are progressively harder.

Basic problems

Exercise 11.1 Basic BVP definition

Define a two-point boundary-value problem and give one example.

Exercise 11.2 Basic Dirichlet conditions

What are Dirichlet boundary conditions?

Exercise 11.3 Basic Neumann conditions

What are Neumann boundary conditions?

Exercise 11.4 Basic Robin conditions

Write a Robin boundary condition at x=a.

Exercise 11.5 Basic Shooting residual

For y''=f(x,y,y'),\qquad y(a)=\alpha,\qquad y(b)=\beta, define the shooting residual.

Exercise 11.6 Basic Convert to first-order system

Convert y''=f(x,y,y') to a first-order system.

Exercise 11.7 Basic Finite-difference second derivative

Derive y''(x_i)\approx \frac{y_{i-1}-2y_i+y_{i+1}}{h^2}.

Exercise 11.8 Basic Finite-difference system

Build the finite-difference system for -y''=g,\qquad y(0)=0,\qquad y(1)=0 on a grid with N=4 subintervals.

Exercise 11.9 Basic Boundary terms

Explain how nonzero Dirichlet boundary values enter the finite-difference right-hand side.

Exercise 11.10 Basic Nonlinear residual

Write the nonlinear finite-difference residual for y''=y^2+x.

Exercise 11.11 Basic Newton correction

State the Newton correction equation for a nonlinear finite-difference BVP system.

Exercise 11.12 Basic Collocation

Explain the idea of collocation for BVPs.

Exercise 11.13 Basic Chebyshev nodes

Write the Chebyshev–Gauss–Lobatto nodes on [-1,1].

Exercise 11.14 Basic Weak form

Derive the weak form of -y''=g,\qquad y(a)=0,\quad y(b)=0.

Exercise 11.15 Basic Galerkin matrix

For -y''+qy=g, write the Galerkin stiffness matrix entries.

Exercise 11.16 Basic Hat functions

Define the piecewise linear hat function \phi_i on a uniform mesh.

Exercise 11.17 Basic Local stiffness matrix

Derive the local stiffness matrix for -y'' using piecewise linear finite elements.

Exercise 11.18 Basic Local mass matrix

Derive the local mass matrix for piecewise linear finite elements on one element.

Exercise 11.19 Basic Gauss two-point rule

State the two-point Gauss–Legendre rule on [-1,1].

Exercise 11.20 Basic Quadrature mapping

Map a Gauss rule from [-1,1] to an element [x_i,x_{i+1}].

Exercise 11.21 Basic Nonlinear Galerkin residual

Write the Galerkin residual for -y''+y^3=g,\qquad y(a)=y(b)=0.

Exercise 11.22 Basic Nonlinear Jacobian

For the residual in the referenced result, write the Newton Jacobian bilinear form.

Exercise 11.23 Basic Residual estimator

Define an element residual for -y''+qy=g.

Exercise 11.24 Basic Adaptive loop

List the four main steps of an adaptive Galerkin method.

Exercise 11.25 Basic Method comparison

Give one advantage and one disadvantage of shooting, finite differences, and finite elements.

Intermediate problems \star

Exercise 11.26 Intermediate Uniqueness

Prove the referenced result.

Exercise 11.27 Intermediate Shooting sensitivity

Prove the referenced result.

Exercise 11.28 Intermediate Finite-difference consistency

Prove the referenced result.

Exercise 11.29 Intermediate Tridiagonal system

Derive the tridiagonal matrix for -y''=g,\qquad y(0)=\alpha,\qquad y(1)=\beta.

Exercise 11.30 Intermediate Nonlinear finite-difference Jacobian

Derive the Jacobian entries for R_i(Y)= \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2} - f\left(x_i,Y_i,\frac{Y_{i+1}-Y_{i-1}}{2h}\right).

Exercise 11.31 Intermediate Multiple shooting equations

Write the nonlinear algebraic system for multiple shooting with three subintervals.

Exercise 11.32 Intermediate Chebyshev scaling

Derive D_x=\frac{2}{b-a}D_\xi, \qquad D_x^{(2)}=\left(\frac{2}{b-a}\right)^2D_\xi^2.

Exercise 11.33 Intermediate Chebyshev collocation system

Write the Chebyshev collocation system for -y''+y=g,\qquad y(-1)=0,\qquad y(1)=0.

Exercise 11.34 Intermediate Weak formulation

Derive the weak formulation for -\frac{\dd}{\dd x}(p y')+qy=g, \qquad y(a)=y(b)=0.

Exercise 11.35 Intermediate Coercivity

Prove the referenced result.

Exercise 11.36 Intermediate Galerkin orthogonality

Prove the referenced result.

Exercise 11.37 Intermediate Cea estimate

Prove the referenced result.

Exercise 11.38 Intermediate Element stiffness

Derive the local stiffness matrix for variable coefficient p(x) using quadrature.

Exercise 11.39 Intermediate Element load vector

Derive the local load vector F_m^{(e)}=\int_e g\varphi_m\,\dd x.

Exercise 11.40 Intermediate Gauss exactness

Prove the referenced result.

Exercise 11.41 Intermediate Quadrature order selection

For quadratic finite elements and polynomial coefficient q of degree 3, find a Gauss order that exactly integrates q\phi_i\phi_j.

Exercise 11.42 Intermediate Nonlinear Galerkin residual

Derive the nonlinear Galerkin residual for -y''+q(x)y+\gamma y^3=g.

Exercise 11.43 Intermediate Newton–Galerkin Jacobian

Prove the referenced result.

Exercise 11.44 Intermediate Quadrature for cubic terms

If y_h is piecewise quadratic, determine a Gauss order that integrates (y_h)^3\phi_i exactly.

Exercise 11.45 Intermediate Least-squares Galerkin

Derive the least-squares Galerkin equations for \mathcal L y=g.

Exercise 11.46 Intermediate Jump residuals

Explain why derivative jumps appear in finite-element residual estimators.

Exercise 11.47 Intermediate Adaptive marking

Describe maximum marking and Dörfler marking for adaptive refinement.

Exercise 11.48 Intermediate Boundary layer

Explain why adaptive meshes are useful for boundary-layer problems.

Exercise 11.49 Intermediate Shooting instability

Construct a linear BVP for which shooting is numerically ill-conditioned.

Exercise 11.50 Intermediate Collocation versus Galerkin

Compare collocation and Galerkin methods in terms of residual enforcement.

Advanced problems \star\star

Exercise 11.51 Advanced Maximum principle

Prove a maximum principle for -y''+q(x)y=g with q\ge0.

Exercise 11.52 Advanced Discrete maximum principle

Prove a discrete maximum principle for the standard finite-difference discretization of -y''+qy=g.

Exercise 11.53 Advanced Fourth-order finite differences

Derive a fourth-order finite-difference method for -y''=g.

Exercise 11.54 Advanced Nonuniform finite differences

Derive a second-derivative formula on a nonuniform mesh.

Exercise 11.55 Advanced Newton convergence

State local convergence conditions for Newton’s method applied to a nonlinear BVP discretization.

Exercise 11.56 Advanced Damped Newton

Develop a damped Newton method for nonlinear BVP systems.

Exercise 11.57 Advanced Continuation method

Use parameter continuation to solve a nonlinear BVP with multiple solution branches.

Exercise 11.58 Advanced Bratu problem

Study the Bratu problem -y''=\lambda e^y,\qquad y(0)=y(1)=0. Discuss multiplicity and Newton initialization.

Exercise 11.59 Advanced Collocation order

Analyze the order of polynomial collocation for a smooth second-order BVP.

Exercise 11.60 Advanced Chebyshev spectral proof

Prove the referenced result more rigorously for a linear constant-coefficient problem.

Exercise 11.61 Advanced Chebyshev conditioning

Investigate conditioning of Chebyshev second-derivative matrices with Dirichlet boundary conditions.

Exercise 11.62 Advanced Tau method

Compare Chebyshev collocation with the Chebyshev tau method.

Exercise 11.63 Advanced Spectral Galerkin

Formulate a spectral Galerkin method for -y''+y=g using basis functions satisfying homogeneous boundary conditions.

Exercise 11.64 Advanced Quadrature aliasing

Explain quadrature aliasing in nonlinear Galerkin methods and propose a de-aliasing strategy.

Exercise 11.65 Advanced Mass lumping

Study mass lumping for one-dimensional finite elements.

Exercise 11.66 Advanced Higher-order finite elements

Derive local matrices for quadratic finite elements on a reference interval.

Exercise 11.67 Advanced Reference element

Derive the reference-element transformation for stiffness and mass integrals.

Exercise 11.68 Advanced Variable coefficients

Analyze quadrature requirements for variable p(x) and q(x) in finite elements.

Exercise 11.69 Advanced Nonlinear Jacobian symmetry

Determine conditions under which the Newton–Galerkin Jacobian is symmetric.

Exercise 11.70 Advanced Energy minimization

Show that the linear Galerkin solution minimizes an energy functional.

Exercise 11.71 Advanced A posteriori reliability

Prove a reliability bound for a residual estimator in a simple one-dimensional Poisson problem.

Exercise 11.72 Advanced A posteriori efficiency

Prove a local efficiency estimate for a residual estimator.

Exercise 11.73 Advanced hp refinement

Compare h-refinement, p-refinement, and hp-refinement for smooth and singular BVP solutions.

Exercise 11.74 Advanced Mixed formulation

Introduce u=y' and formulate a mixed finite-element method for a second-order BVP.

Exercise 11.75 Advanced Petrov–Galerkin stability

State an inf-sup condition for a Petrov–Galerkin BVP method.

Research-level problems \star\star\star

Exercise 11.76 Research-level Nonlinear bifurcation

Study bifurcation diagrams for nonlinear BVPs using continuation and Newton– Galerkin discretization.

Exercise 11.77 Research-level Pseudo-arclength continuation

Derive pseudo-arclength continuation for nonlinear BVPs with turning points.

Exercise 11.78 Research-level Goal-oriented adaptivity

Develop a dual-weighted residual estimator for a quantity of interest.

Exercise 11.79 Research-level Certified BVP enclosures

Study interval or radii-polynomial methods for certifying solutions of nonlinear BVPs.

Exercise 11.80 Research-level Boundary layers

Analyze singularly perturbed BVPs with boundary layers and compare fitted meshes, adaptive FEM, and spectral collocation.

Exercise 11.81 Research-level Shock-like internal layers

Study nonlinear BVPs with sharp internal layers and design adaptive discretizations.

Exercise 11.82 Research-level hp spectral element BVP

Develop an hp-spectral element method for a one-dimensional BVP.

Exercise 11.83 Research-level Discontinuous Galerkin BVP

Formulate a discontinuous Galerkin method for a second-order BVP using interior penalty.

Exercise 11.84 Research-level Isogeometric Galerkin

Study isogeometric Galerkin discretization for BVPs using spline basis functions.

Exercise 11.85 Research-level Quadrature in nonlinear FEM

Analyze how quadrature error affects Newton convergence in nonlinear finite-element BVPs.

Exercise 11.86 Research-level Adaptive quadrature inside Newton

Design an adaptive quadrature strategy for evaluating nonlinear Galerkin residuals and Jacobians.

Exercise 11.87 Research-level Matrix-free Newton–Krylov

Develop a matrix-free Newton–Krylov method for a nonlinear Galerkin BVP system.

Exercise 11.88 Research-level Preconditioning Galerkin systems

Design preconditioners for linearized Galerkin systems arising from nonlinear BVPs.

Exercise 11.89 Research-level Spectral integration matrices

Compare Chebyshev differentiation-matrix methods with spectral integration methods for BVPs.

Exercise 11.90 Research-level Green’s functions

Construct a Green’s function for a linear second-order BVP and compare with finite element approximation.

Exercise 11.91 Research-level Sturm–Liouville eigenvalue BVP

Extend the Galerkin method to Sturm–Liouville eigenvalue problems.

Exercise 11.92 Research-level Fractional BVPs

Discuss how weak forms and quadrature change for fractional boundary-value problems.

Exercise 11.93 Research-level Neural Galerkin BVP

Formulate a neural Galerkin or residual-minimization method for a nonlinear BVP and compare it with finite elements.

Exercise 11.94 Research-level Reproducible BVP benchmarks

Design reproducible benchmarks comparing shooting, finite differences, collocation, linear Galerkin, nonlinear Galerkin, and Chebyshev spectral methods.

Exercise 11.95 Research-level Automatic method selection

Design a decision system that selects a BVP method based on stiffness, nonlinearity, smoothness, boundary layers, and desired accuracy.

Exercise solutions

The solutions below are converted from the uploaded LaTeX solution file. Open an accordion to view the full solution.

This chapter gives detailed worked solutions for the exercises in Chapter 11. Each solution includes the problem formulation, the method, the mathematical derivation, the conclusion, and a diagnostic comment.

Exercise 11.1

Problem formulation.

Define a two-point boundary-value problem and give one example.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.2

Problem formulation.

What are Dirichlet boundary conditions?

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.3

Problem formulation.

What are Neumann boundary conditions?

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.4

Problem formulation.

Write a Robin boundary condition at x=a.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.5

Problem formulation.

For y''=f(x,y,y'),\qquad y(a)=\alpha,\qquad y(b)=\beta, define the shooting residual.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.6

Problem formulation.

Convert y''=f(x,y,y') to a first-order system.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.7

Problem formulation.

Derive y''(x_i)\approx \frac{y_{i-1}-2y_i+y_{i+1}}{h^2}.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.8

Problem formulation.

Build the finite-difference system for -y''=g,\qquad y(0)=0,\qquad y(1)=0 on a grid with N=4 subintervals.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.9

Problem formulation.

Explain how nonzero Dirichlet boundary values enter the finite-difference right-hand side.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.10

Problem formulation.

Write the nonlinear finite-difference residual for y''=y^2+x.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.11

Problem formulation.

State the Newton correction equation for a nonlinear finite-difference BVP system.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.12

Problem formulation.

Explain the idea of collocation for BVPs.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.13

Problem formulation.

Write the Chebyshev–Gauss–Lobatto nodes on [-1,1].

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.14

Problem formulation.

Derive the weak form of -y''=g,\qquad y(a)=0,\quad y(b)=0.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.15

Problem formulation.

For -y''+qy=g, write the Galerkin stiffness matrix entries.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.16

Problem formulation.

Define the piecewise linear hat function \phi_i on a uniform mesh.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.17

Problem formulation.

Derive the local stiffness matrix for -y'' using piecewise linear finite elements.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.18

Problem formulation.

Derive the local mass matrix for piecewise linear finite elements on one element.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.19

Problem formulation.

State the two-point Gauss–Legendre rule on [-1,1].

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.20

Problem formulation.

Map a Gauss rule from [-1,1] to an element [x_i,x_{i+1}].

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.21

Problem formulation.

Write the Galerkin residual for -y''+y^3=g,\qquad y(a)=y(b)=0.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.22

Problem formulation.

For the residual in the referenced result, write the Newton Jacobian bilinear form.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.23

Problem formulation.

Define an element residual for -y''+qy=g.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.24

Problem formulation.

List the four main steps of an adaptive Galerkin method.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.25

Problem formulation.

Give one advantage and one disadvantage of shooting, finite differences, and finite elements.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.26

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.27

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.28

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.29

Problem formulation.

Derive the tridiagonal matrix for -y''=g,\qquad y(0)=\alpha,\qquad y(1)=\beta.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.30

Problem formulation.

Derive the Jacobian entries for R_i(Y)= \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2} - f\left(x_i,Y_i,\frac{Y_{i+1}-Y_{i-1}}{2h}\right).

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.31

Problem formulation.

Write the nonlinear algebraic system for multiple shooting with three subintervals.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.32

Problem formulation.

Derive D_x=\frac{2}{b-a}D_\xi, \qquad D_x^{(2)}=\left(\frac{2}{b-a}\right)^2D_\xi^2.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.33

Problem formulation.

Write the Chebyshev collocation system for -y''+y=g,\qquad y(-1)=0,\qquad y(1)=0.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.34

Problem formulation.

Derive the weak formulation for -\frac{\dd}{\dd x}(p y')+qy=g, \qquad y(a)=y(b)=0.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.35

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.36

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.37

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.38

Problem formulation.

Derive the local stiffness matrix for variable coefficient p(x) using quadrature.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.39

Problem formulation.

Derive the local load vector F_m^{(e)}=\int_e g\varphi_m\,\dd x.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.40

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.41

Problem formulation.

For quadratic finite elements and polynomial coefficient q of degree 3, find a Gauss order that exactly integrates q\phi_i\phi_j.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.42

Problem formulation.

Derive the nonlinear Galerkin residual for -y''+q(x)y+\gamma y^3=g.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.43

Problem formulation.

Prove the referenced result.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.44

Problem formulation.

If y_h is piecewise quadratic, determine a Gauss order that integrates (y_h)^3\phi_i exactly.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.45

Problem formulation.

Derive the least-squares Galerkin equations for \mathcal L y=g.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.46

Problem formulation.

Explain why derivative jumps appear in finite-element residual estimators.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.47

Problem formulation.

Describe maximum marking and Dörfler marking for adaptive refinement.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.48

Problem formulation.

Explain why adaptive meshes are useful for boundary-layer problems.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.49

Problem formulation.

Construct a linear BVP for which shooting is numerically ill-conditioned.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.50

Problem formulation.

Compare collocation and Galerkin methods in terms of residual enforcement.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.51

Problem formulation.

Prove a maximum principle for -y''+q(x)y=g with q\ge0.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.52

Problem formulation.

Prove a discrete maximum principle for the standard finite-difference discretization of -y''+qy=g.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.53

Problem formulation.

Derive a fourth-order finite-difference method for -y''=g.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.54

Problem formulation.

Derive a second-derivative formula on a nonuniform mesh.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.55

Problem formulation.

State local convergence conditions for Newton’s method applied to a nonlinear BVP discretization.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.56

Problem formulation.

Develop a damped Newton method for nonlinear BVP systems.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.57

Problem formulation.

Use parameter continuation to solve a nonlinear BVP with multiple solution branches.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.58

Problem formulation.

Study the Bratu problem -y''=\lambda e^y,\qquad y(0)=y(1)=0. Discuss multiplicity and Newton initialization.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.59

Problem formulation.

Analyze the order of polynomial collocation for a smooth second-order BVP.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.60

Problem formulation.

Prove the referenced result more rigorously for a linear constant-coefficient problem.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.61

Problem formulation.

Investigate conditioning of Chebyshev second-derivative matrices with Dirichlet boundary conditions.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.62

Problem formulation.

Compare Chebyshev collocation with the Chebyshev tau method.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.63

Problem formulation.

Formulate a spectral Galerkin method for -y''+y=g using basis functions satisfying homogeneous boundary conditions.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.64

Problem formulation.

Explain quadrature aliasing in nonlinear Galerkin methods and propose a de-aliasing strategy.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.65

Problem formulation.

Study mass lumping for one-dimensional finite elements.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.66

Problem formulation.

Derive local matrices for quadratic finite elements on a reference interval.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.67

Problem formulation.

Derive the reference-element transformation for stiffness and mass integrals.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.68

Problem formulation.

Analyze quadrature requirements for variable p(x) and q(x) in finite elements.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.69

Problem formulation.

Determine conditions under which the Newton–Galerkin Jacobian is symmetric.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.70

Problem formulation.

Show that the linear Galerkin solution minimizes an energy functional.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.71

Problem formulation.

Prove a reliability bound for a residual estimator in a simple one-dimensional Poisson problem.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.72

Problem formulation.

Prove a local efficiency estimate for a residual estimator.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.73

Problem formulation.

Compare h-refinement, p-refinement, and hp-refinement for smooth and singular BVP solutions.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.74

Problem formulation.

Introduce u=y' and formulate a mixed finite-element method for a second-order BVP.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.75

Problem formulation.

State an inf-sup condition for a Petrov–Galerkin BVP method.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Exercise 11.76

Problem formulation.

Study bifurcation diagrams for nonlinear BVPs using continuation and Newton– Galerkin discretization.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.77

Problem formulation.

Derive pseudo-arclength continuation for nonlinear BVPs with turning points.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.78

Problem formulation.

Develop a dual-weighted residual estimator for a quantity of interest.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.79

Problem formulation.

Study interval or radii-polynomial methods for certifying solutions of nonlinear BVPs.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.80

Problem formulation.

Analyze singularly perturbed BVPs with boundary layers and compare fitted meshes, adaptive FEM, and spectral collocation.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.81

Problem formulation.

Study nonlinear BVPs with sharp internal layers and design adaptive discretizations.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.82

Problem formulation.

Develop an hp-spectral element method for a one-dimensional BVP.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.83

Problem formulation.

Formulate a discontinuous Galerkin method for a second-order BVP using interior penalty.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.84

Problem formulation.

Study isogeometric Galerkin discretization for BVPs using spline basis functions.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.85

Problem formulation.

Analyze how quadrature error affects Newton convergence in nonlinear finite-element BVPs.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.86

Problem formulation.

Design an adaptive quadrature strategy for evaluating nonlinear Galerkin residuals and Jacobians.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.87

Problem formulation.

Develop a matrix-free Newton–Krylov method for a nonlinear Galerkin BVP system.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Verify the stationarity residual, feasibility residual, and descent or curvature condition, not only the final numerical value.

Exercise 11.88

Problem formulation.

Design preconditioners for linearized Galerkin systems arising from nonlinear BVPs.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Distinguish forward error from backward error; a small residual does not imply a small forward error when the problem is ill-conditioned.

Exercise 11.89

Problem formulation.

Compare Chebyshev differentiation-matrix methods with spectral integration methods for BVPs.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.90

Problem formulation.

Construct a Green’s function for a linear second-order BVP and compare with finite element approximation.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The order claim should be confirmed by a Taylor expansion or a log-log refinement table using a problem with a known exact solution.

Exercise 11.91

Problem formulation.

Extend the Galerkin method to Sturm–Liouville eigenvalue problems.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.92

Problem formulation.

Discuss how weak forms and quadrature change for fractional boundary-value problems.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Check both the singular kernel evaluation and the long-history summation; fractional errors often come from the memory term rather than the algebraic solver.

Exercise 11.93

Problem formulation.

Formulate a neural Galerkin or residual-minimization method for a nonlinear BVP and compare it with finite elements.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For -(p y')'+qy=g,\qquad y(a)=y(b)=0, multiply by a test function v\in H_0^1(a,b) and integrate: \int_a^b -(p y')'v\,\dd x+\int_a^b qyv\,\dd x = \int_a^b gv\,\dd x. Integrating by parts and using v(a)=v(b)=0, \int_a^b p y'v'\,\dd x+\int_a^b q yv\,\dd x = \int_a^b gv\,\dd x. With y_h=\sum_jY_j\phi_j, the linear system is \sum_j \left[ \int_a^b p\phi_j'\phi_i'\,\dd x+ \int_a^b q\phi_j\phi_i\,\dd x \right]Y_j = \int_a^b g\phi_i\,\dd x.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.94

Problem formulation.

Design reproducible benchmarks comparing shooting, finite differences, collocation, linear Galerkin, nonlinear Galerkin, and Chebyshev spectral methods.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

For a second-order BVP y''=f(x,y,y'),\qquad y(a)=\alpha,\quad y(b)=\beta, introduce the unknown initial slope y'(a)=s. Solve the IVP and define the shooting residual F(s)=y(b;s)-\beta. The BVP is solved when F(s)=0. Newton’s method for the shooting parameter is s_{k+1}=s_k-\frac{F(s_k)}{F'(s_k)}. The derivative F'(s) can be obtained from the variational equation.

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

The formula should be verified by substitution into the defining equation and, when possible, by a small numerical test.

Exercise 11.95

Problem formulation.

Design a decision system that selects a BVP method based on stiffness, nonlinearity, smoothness, boundary layers, and desired accuracy.

Method.

Use shooting residuals, finite-difference discretization, weak forms, Galerkin assembly, quadrature, and nonlinear Newton linearization.

Detailed solution and justification.

A BVP is discretized either by finite differences or by a weak formulation. For finite differences, y''(x_i)\approx \frac{Y_{i-1}-2Y_i+Y_{i+1}}{h^2}. For a Galerkin method, write y_h=\sum_jY_j\phi_j and impose a(y_h,\phi_i)=\ell(\phi_i). For nonlinear problems, define residuals R_i(Y) and solve R(Y)=0 by Newton’s method: J(Y_k)s_k=-R(Y_k).

Conclusion.

The displayed derivation gives the requested formula, proof, or computation. The final expression should be checked against the assumptions and notation of the corresponding chapter.

Diagnostic comment.

Always test the result on the scalar model equation or Fourier mode and verify the amplification factor satisfies the stated stability bound.

Interactive tools

The theory and solutions come first. These tools are for review, laboratories, games, memory cards, and randomized assessment.

Professional visualsAnimated diagrams and conceptual graphics. Programming practiceDebug JavaScript/PHP method snippets with corrected code.

Interactive PHP laboratories for this chapter

These experiments run inside the same web page. The student changes the input data and presses Run PHP. Results are returned by the PHP server and displayed immediately.