Skip to main content
Log in

A quadratic spline collocation method for the Dirichlet biharmonic problem

  • Original Paper
  • Published:
Numerical Algorithms Aims and scope Submit manuscript

Abstract

A new method based on quadratic spline collocation is formulated for the solution of the Dirichlet biharmonic problem on the unit square rewritten as a coupled system of two second-order partial differential equations. This method involves the solution of an auxiliary biharmonic problem using fast Fourier transforms and the solution of a nonsymmetric Schur complement system using preconditioned BICGSTAB, at a total cost of \(N^{2} \log N\) on an N × N uniform partition of the unit square. The results of numerical experiments demonstrate the optimality of the global accuracy of the method and also superconvergence results, in particular, third-order accuracy in the \(L^{\infty }\) norm of the solution and its fourth-order accuracy at the partition nodes and the collocation points.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Similar content being viewed by others

References

  1. Abushama, A.A.: Modified nodal cubic spline collocation for Poisson’s and biharmonic equations in the unit square. Ph.D. thesis, Colorado School of Mines, Golden (2004)

    Google Scholar 

  2. Abushama, A.A., Bialecki, B.: Modified nodal cubic spline collocation for biharmonic equations. Numer. Algorithms 43, 331–353 (2006)

    Article  MathSciNet  Google Scholar 

  3. Arad, M., Yakhot, A., Ben-Dor, G.: A highly accurate numerical solution of a biharmonic equation. Numer. Methods Partial Differential Equations 13, 375–391 (1997)

    Article  MathSciNet  Google Scholar 

  4. Ben-Artzi, M., Croisille, J., Fishelov, D.: A fast direct solver for the biharmonic problem in a rectangular grid. SIAM J. SCi. Comput. 31, 303–333 (2008)

    Article  MathSciNet  Google Scholar 

  5. Bialecki, B.: A fast solver for the orthogonal spline collocation solution of the biharmonic Dirichlet problem on rectangles. J. Comput. Phys. 191, 601–621 (2003)

    Article  MathSciNet  Google Scholar 

  6. Bialecki, B.: A fourth order finite difference method for the Dirichlet biharmonic problem. Numer. Algorithms 61, 351–375 (2012)

    Article  MathSciNet  Google Scholar 

  7. Bialecki, B., Fairweather, G.: Orthogonal spline collocation methods for partial differential equations, Numerical Analysis 2000, Vol. VII, Partial differential equations. J. Comput. Appl. Math. 128, 55–82 (2001)

    Article  MathSciNet  Google Scholar 

  8. Bialecki, B., Fairweather, G., Karageorghis, A.: Matrix decomposition algorithms for modified spline collocation for Helmholtz problems. SIAM J. Sci. Comput. 24, 1733–1753 (2003)

    Article  MathSciNet  Google Scholar 

  9. Bialecki, B., Fairweather, G., Karageorghis, A.: Optimal superconvergent one step nodal cubic spline collocation methods. SIAM J. Sci. Comput. 27, 575–598 (2005)

    Article  MathSciNet  Google Scholar 

  10. Bialecki, B., Fairweather, G., Karageorghis, A.: Matrix decomposition algorithms for elliptic boundary value problems: a survey. Numer. Algorithms 56, 253–295 (2011)

    Article  MathSciNet  Google Scholar 

  11. de Boor, C.: The method of projections as applied to the numerical solution of two point boundary value problems using cubic splines. Ph.D. Thesis, University of Michigan, Ann Arbor (1966)

    Google Scholar 

  12. de Boor, C., Swartz, B.: Collocation at Gaussian points. SIAM J. Numer. Anal. 10, 582–606 (1973)

    Article  MathSciNet  Google Scholar 

  13. Bjrstad, P.: Fast numerical solution of the biharmonic Dirichlet problem on rectangles. SIAM J. Numer. Anal. 20, 59–71 (1983)

    Article  MathSciNet  Google Scholar 

  14. Buzbee, B.L., Dorr, F.W.: The direct solution of the biharmonic equation on rectangular regions and the Poisson equation on irregular regions. SIAM J. Numer. Anal. 11, 753–763 (1974)

    Article  MathSciNet  Google Scholar 

  15. Christara, C.C.: Quadratic spline collocation methods for elliptic partial differential equations. BIT 34, 33–61 (1994)

    Article  MathSciNet  Google Scholar 

  16. Christara, C.C., Chen, T., Dang, D.M.: Quadratic spline collocation for one-dimensional linear parabolic partial differential equations. Numer. Algorithms 53, 511–553 (2010)

    Article  MathSciNet  Google Scholar 

  17. Christara, C., Liu, G.: Quartic spline collocation for second–order boundary value problems. In: Proceedings of the 9th HERCMA Conference on Computer Mathematics and Applications, Athens University of Economics and Business, September 23-26, 2009, Athens, Greece,pp. 1–8

  18. Christara, C.C., Ng, K.S.: Optimal quadratic and cubic spline collocation on nonuniform partitions. Computing 76, 227–257 (2006)

    Article  MathSciNet  Google Scholar 

  19. Christara, C.C., Zhu, Y., Zhang, J.: Quartic spline collocation for fourth-order boundary value problems. In: Proceedings of the 2008 Numerical Analysis conference, September 1-5 Kalamata, Greece, pp. 62–67 (2008)

  20. Daniel, J.W., Swartz, B.K.: Extrapolated collocation for two–point boundary value problems using cubic splines. J. Inst. Math. Appl. 16, 161–174 (1975)

    Article  MathSciNet  Google Scholar 

  21. Ehrlich, L.W.: Solving the biharmonic equation as coupled finite difference equations. SIAM J. Numer. Anal. 8, 278–287 (1971)

    Article  MathSciNet  Google Scholar 

  22. El-Gamel, M., Mohsen, A., El-Mohsen, A.: Sinc-Galerkin method for solving biharmonic problems. Appl. Math. Comput. 247, 386–396 (2014)

    MathSciNet  MATH  Google Scholar 

  23. Fairweather, G., Karageorghis, A., Maack, J.: Compact optimal quadratic spline collocation methods for the Helmholtz equation. J. Comput. Phys. 230, 2880–2895 (2011)

    Article  MathSciNet  Google Scholar 

  24. Gupta, M.M., Manohar, R.P.: The direct solution of the biharmonic equation using noncoupled approach. J. Comput. Phys. 33, 236–248 (1979)

    Article  MathSciNet  Google Scholar 

  25. Hackbusch, W.: Elliptic differential equations, theory and numerical treatment. Springer, New York (1992)

    Book  Google Scholar 

  26. Houstis, E.N., Christara, C.C., Rice, J.R.: Quadratic-spline collocation methods for two-point boundary value problems. Internat. J. Numer. Methods Engrg. 26, 935–952 (1988)

    Article  MathSciNet  Google Scholar 

  27. Knudson, D.B.: A piecewise Hermite bicubic finite element Galerkin method for the biharmonic Dirichlet Problem. Ph.D. Thesis, Colorado School of Mines, Golden (1997)

    Google Scholar 

  28. Layton, A.T., Christara, C.C., Jackson, K.R.: Optimal quadratic spline collocation methods for the shallow water equations. Math. Comput. Simul. 71, 187–205 (2006)

    Article  Google Scholar 

  29. Lou, Z.M., Bialecki, B., Fairweather, G.: Orthogonal sploine collocation methods for biharmonic prob;lems. Numer. Math. 80, 267–303 (1998)

    Article  MathSciNet  Google Scholar 

  30. Luo, W.-H., Huang, T.-Z., Gu, X.-M.: A Lagrange-quadratic spline compact optimal collocation method for the time tempered fractional diffusion equation, submitted

  31. Luo, W.-H., Huang, T.-Z., Wu, G.-C., Gu, X.-M.: Quadratic spline collocation for the time fractional subdiffusion equation. Appl. Math. Comput. 276, 252–265 (2016)

    MathSciNet  MATH  Google Scholar 

  32. Maack, J.: Quadratic spline collocation for Poisson’s and biharmonic equations in the unit square. M.S. Thesis, Colorado School of Mines, Golden (2009)

    Google Scholar 

  33. Meleshko, V.V.: Biharmonic problem in a rectangle. Appl. Sci. Res. 58, 217–249 (1998)

    Article  MathSciNet  Google Scholar 

  34. Peisker, P.: On the numerical solution of the first biharmonic equation. Model. Math. Anal. Numer. 22, 655–676 (1988)

    Article  MathSciNet  Google Scholar 

  35. Stephenson, J.W.: Single cell discretizations of order two and four for biharmonic problems. J. Comput. Phys. 55, 65–80 (1984)

    Article  MathSciNet  Google Scholar 

  36. Zhang, X.: Multilevel Schwarz methods for the biharmonic Dirichlet problem. SIAM J. Sci. Comput. 15, 621–644 (1994)

    Article  MathSciNet  Google Scholar 

  37. Zhang, J.: Bi-quartic spline collocation methods for fourth-order boundary value problems with an application to the biharmonic Dirichlet problem. Ph.D. Thesis, University of Toronto, Toronto, Ontario (2008)

    Google Scholar 

Download references

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Graeme Fairweather.

Additional information

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix

Appendix

We shall prove that the matrix P defined in (4.34) is symmetric and positive definite. It follows from (3.15) and (2.10) that

$$ Z^{T} B^{D} Z=\frac{1}{8}{\Lambda} +I_{N} \equiv {\Lambda}_{1}. $$
(A.1)

Using (2.8), we have

$$ -4 \leq \lambda_{i} < 0, \quad i = 1,\ldots,N, $$
(A.2)

which yields

$$ \frac{1}{2} \leq \frac{1}{8}\lambda_{i}+ 1, \quad i = 1,\ldots,N. $$
(A.3)

It follows from (A.1), (2.8), and (A.3) that BD is symmetric and that the diagonal entries of Λ1 are positive. Consequently, BD is positive definite. Since BD is symmetric and positive definite, it follows from (4.37) and (4.27) that − P22, appearing on the right-hand side of (4.34), is symmetric and positive definite. It thus remains to show that \(P_{21}P_{11}^{-1}P_{12}\), appearing on the right-hand side of (4.34), is symmetric and positive definite. To this end, we first derive a formula for \(P_{21}P_{11}^{-1}P_{12}\). It follows from (4.35)–(4.37) that

$$ P_{11}= \left[ \begin{array}{cc} P_{11}^{1} & P_{11}^{2} \\ 0 & P_{11}^{1} \end{array} \right], \quad P_{12}= \left[ \begin{array}{c} P_{12}^{1} \\ P_{12}^{2} \end{array} \right], \quad P_{21}= \left[ \begin{array}{cc} P_{21}^{1} & P_{21}^{2} \end{array} \right], $$
(A.4)

where

$$ P_{11}^{1}=B^{D} \otimes A^{D} + A^{D}\otimes \left( B^{D} - \displaystyle\frac{h^{2}}{12}A^{D}\right), \quad P_{11}^{2}=-B^{D} \otimes B^{D}, $$
(A.5)
$$ P_{12}^{1}=-E \otimes B^{D}, \quad P_{12}^{2}=E \otimes A^{D} + F\otimes \left( B^{D} - \displaystyle\frac{h^{2}}{12}A^{D}\right), $$
(A.6)
$$ P_{21}^{1}=C^{D}\otimes B^{D}, \quad P_{21}^{2}=\frac{h^{2}}{12} P_{21}^{1}. $$
(A.7)

Using (A.5), (3.15), (2.10), we have

$$ (Z^{T}\otimes Z^{T})P_{11}^{1}(Z\otimes Z) =h^{-2}\left[\left( \frac{1}{8}{\Lambda} +I_{N}\right)\otimes {\Lambda} +{\Lambda} \otimes \left( \frac{1}{24}{\Lambda} +I_{N}\right)\right]\equiv {\Lambda}_{2}. $$
(A.8)

Using (A.2), we have that

$$ \frac{5}{6} \leq \frac{1}{24}\lambda_{i}+ 1, \quad i = 1,\ldots,N. $$
(A.9)

It follows from (A.8), (2.8), (A.2), (A.3), and (A.9) that P11 is symmetric and that diagonal entries of Λ2 are negative. Thus, \(P_{11}^{1}\) is negative definite and hence nonsingular. With

$$ R=(P_{11}^{1})^{-1}P_{11}^{2} (P_{11}^{1})^{-1}, $$
(A.10)

we have

$$P_{11}^{-1}= \left[ \begin{array}{cc} (P_{11}^{1})^{-1} & -R \\ 0 & (P_{11}^{1})^{-1} \end{array} \right]. $$

The last equation and (A.4) yield

$$ P_{21} P_{11}^{-1} P_{12} = P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1} - P_{21}^{1} R P_{12}^{2} + P_{21}^{2}(P_{11}^{1})^{-1} P_{12}^{2}. $$
(A.11)

We next study the terms on the right-hand side in (A.11). Using (A.7), (A.6), (4.26), and (4.27), we obtain

$$ P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1}= -\frac{1}{4h} \left( [\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes B^{D}\right) (P_{11}^{1})^{-1} \left( [\textbf{e}_{1}|\textbf{e}_{N}]\otimes B^{D}\right). $$
(A.12)

Since BD and \(P_{11}^{1}\) are symmetric, (A.12) implies that \(P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1}\) is also symmetric. From (A.7), (A.6), (4.26), (4.27), and (3.15) it follows that

$$ P_{21}^{1} =\frac{2}{h}[\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes \left( \frac{1}{8}T+I_{N}\right) =\frac{1}{4h}[\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T +\frac{2}{h}[\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}, $$
(A.13)
$$ P_{12}^{2} = \frac{1}{8h^{2}} [\textbf{e}_{1}|\textbf{e}_{N}]\otimes T+ \frac{1}{h^{2}}[\textbf{e}_{1}|\textbf{e}_{N}]\otimes \left( \!\frac{1}{24}T + I_{N}\!\right) \! = \frac{1}{6h^{2}} [\textbf{e}_{1}|\textbf{e}_{N}]\otimes T+ \frac{1}{h^{2}}[\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}. $$
(A.14)

Using the last two equations, we have

$$\begin{array}{@{}rcl@{}} P_{21}^{1} R P_{12}^{2} &=& \frac{1}{24h^{3}}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&+\frac{1}{4h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &&+ \frac{1}{3h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&+ \frac{2}{h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}). \end{array} $$
(A.15)

Since BD is symmetric, it follows from (A.5) that \(P_{11}^{2}\) is symmetric. This and the symmetry of \(P_{11}^{1}\) imply that R of (A.10) is symmetric since

$$R^{T}=((P_{11}^{1})^{T})^{-1}(P_{11}^{2})^{T} ((P_{11}^{1})^{T})^{-1} = (P_{11}^{1})^{-1}P_{11}^{2}(P_{11}^{1})^{-1} = R. $$

Since T of (2.7) and R are symmetric so are the matrices

$$([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T), \quad ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}), $$

appearing on the right-hand side of (A.15). It follows from (3.15) that ADT = TAD, BDT = TBD. Hence (A.5) implies that

$$ P_{11}^{2}(I_{N}\otimes T)=(I_{N}\otimes T) P_{11}^{2} $$
(A.16)

and

$$P_{11}^{1} (I_{N}\otimes T)=(I_{N}\otimes T) P_{11}^{1}. $$

The last equation yields

$$ (P_{11}^{1})^{-1}(I_{N}\otimes T)=(I_{N}\otimes T) (P_{11}^{1})^{-1}. $$
(A.17)

Using (A.10), (A.16), and (A.17), we have

$$ R(I_{N}\!\otimes T) = (P_{11}^{1})^{-1} P_{11}^{2} (P_{11}^{1})^{-1} (I_{N}\!\otimes T) = (I_{N}\!\otimes T) (P_{11}^{1})^{-1} P_{11}^{2} (P_{11}^{1})^{-1} = (I_{N}\!\otimes T) R. $$
(A.18)

For \(Q=(P_{11}^{1})^{-1}\) or R, (A.17) and (A.18) yield

$$\begin{array}{@{}rcl@{}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) Q ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) &=& ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (I_{N} \otimes T) Q ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &=& ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) Q (I_{N} \otimes T) ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &=& ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) Q ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T). \end{array} $$
(A.19)

The matrices

$$([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}), \quad ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T), $$

appearing on the right-hand side of (A.15) are symmetric since the symmetry of T, R, and (A.19) with Q = R yield

$$\begin{array}{@{}rcl@{}} [([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N})]^{T} &=&([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &=&([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}), \end{array} $$
$$\begin{array}{@{}rcl@{}} [([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T)]^{T} &=& ([\textbf{e}_{1}|\textbf{e}_{N}]^{T}\otimes T) R ([\textbf{e}_{1}|\textbf{e}_{N}] \otimes I_{N}) \\ &=&([\textbf{e}_{1}|\textbf{e}_{N}]^{T}\otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}] \otimes T). \end{array} $$

Since all matrices on the right-hand side of (A.15) are symmetric so is \(P_{21}^{1} R P_{12}^{2}\). In a similar way we show that \(P_{21}^{2} (P_{11}^{1})^{-1} P_{12}^{2}\) is also symmetric. Since all matrices on the right-hand side of (A.11) are symmetric so is \(P_{21}P_{11}^{-1} P_{12}\).

We next show that \(P_{21}P_{11}^{-1} P_{12}\) is positive definite. Using (A.12), (3.15), and (A.19) with \(Q=(P_{11}^{1})^{-1}\), we have that

$$\begin{array}{@{}rcl@{}} P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1} &=& -\frac{1}{4h} \left( [\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes \left( \frac{1}{8}T+I_{N}\right)\right)\\ &&\times(P_{11}^{1})^{-1} \left( [\textbf{e}_{1}|\textbf{e}_{N}]\otimes \left( \frac{1}{8}T+I_{N}\right)\right)\\ &=& -\frac{1}{256h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T)\\ &&-\frac{1}{32h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N})\\ &&-\frac{1}{32h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&-\frac{1}{4h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N})\\ &=& -\frac{1}{256h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T)\\ &&-\frac{1}{16h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\&&- \frac{1}{4h}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}). \end{array} $$

Using (A.7), (4.26), (3.15), (A.14), and (A.19) with Q = (P11)− 1, we have

$$\begin{array}{@{}rcl@{}} P_{21}^{2} (P_{11}^{1})^{-1} P_{12}^{2} &=&\frac{h}{6} \left( [\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes \left( \frac{1}{8}T+I_{N}\right)\right)\\ &&\times(P_{11}^{1})^{-1} \left( \frac{1}{6h^{2}} [\textbf{e}_{1}|\textbf{e}_{N}]\otimes T+ \frac{1}{h^{2}}[\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}\right)\\ &=& \frac{1}{288h}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\&&+ \frac{1}{48h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N})\\ &&+ \frac{1}{36h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\&&+ \frac{1}{6h}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N})\\ &=& \frac{1}{288h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\&&+\frac{7}{144h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T)\\ &&+ \frac{1}{6h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}). \end{array} $$

The last two equations yield

$$\begin{array}{@{}rcl@{}} P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1}+P_{21}^{2} (P_{11}^{1})^{-1} P_{12}^{2} \!&=&\! -\frac{1}{2304h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&\!-\frac{1}{72h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&\!- \frac{1}{12h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ \!&=&\! -\frac{1}{72h} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (P_{11}^{1})^{-1} \\ &&\!\times(I_{N}\otimes (T + 6 I_{N}))([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &&\!- \frac{1}{2304h}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes T) (P_{11}^{1})^{-1} \\ &&\times ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T). \end{array} $$
(A.20)

Using (2.10) and (A.8), we have

$$\begin{array}{@{}rcl@{}} &&-(Z^{T}\otimes Z^{T}) \left[(P_{11}^{1})^{-1} (I_{N}\otimes \left( T + 6 I_{N}\right)\right] (Z\otimes Z)\\ &=&-[(Z^{T}\otimes Z^{T}) P_{11}^{1} (Z\otimes Z)]^{-1} (Z^{T}\otimes Z^{T})\\ &&\times\left[ (I_{N}\otimes (T + 6 I_{N}) \right] (Z\otimes Z) \\ &=&-{\Lambda}_{2}^{-1} \left[I_{N}\otimes ({\Lambda} + 6I_{N})\right]. \end{array} $$
(A.21)

It follows from (2.8) and (A.2) that the diagonal entries of Λ + 6IN are positive. Recall that the diagonal entries of Λ2 are negative. Thus it follows from (A.21) that \(-(P_{11}^{1})^{-1} \left [ (I_{N}\otimes (T + 6I_{N}) \right ]\) is symmetric and its eigenvalues are positive. Consequently \(-(P_{11}^{1})^{-1} \left [ (I_{N}\otimes (T + 6I_{N}) \right ]\) is positive definite. Since \(P_{11}^{1}\) is negative definite, \(-(P_{11}^{1})^{-1}\) is positive definite. It follows from (2.10) and (2.8) that the eigenvalues of T are nonzero which makes T nonsingular. Hence, for nonzero v in R2N, w = ([e1|eN] ⊗ T)v, z = ([e1|eN] ⊗ IN)v are nonzero. Consequently, the positive definiteness of \(-(P_{11}^{1})^{-1} \left [ (I_{N}\otimes (T + 6I_{N}) \right ]\) and \(-(P_{11}^{1})^{-1}\) yield

$$\begin{array}{@{}rcl@{}} &&-\textbf{v}^{T} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T}\otimes T) (P_{11}^{1})^{-1} \left[ (I_{N}\otimes (T + 6I_{N}) \right] ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \textbf{v}\\ &=& -\textbf{w}^{T} (P_{11}^{1})^{-1}\left[ (I_{N}\otimes (T + 6I_{N})\right] \textbf{w}>0,\\ &&-\textbf{v}^{T} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T}\otimes T) (P_{11}^{1})^{-1} ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \textbf{v}\\ &=& -\textbf{z}^{T} (P_{11}^{1})^{-1} \textbf{z}>0. \end{array} $$

The last two equations show, on using (A.20), that \(P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1}+P_{21}^{2} (P_{11}^{1})^{-1} P_{12}^{2}\) is positive definite. Using (A.15), (A.19) with Q = R, and (A.18), we have

$$\begin{array}{@{}rcl@{}} -P_{21}^{1} R P_{12}^{2} &=& - \frac{1}{24h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) (I_{N} \otimes T) R (I_{N}\otimes T) ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &&-\frac{7}{12h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes T) \\ &&- \frac{2}{h^{3}}([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &=& -\frac{1}{24h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) \left[ R (I_{N} \otimes T) (I_{N}\otimes T) \right.\\ &&\left.+ 14 R I_{N}\otimes T + 48R \right] ([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}) \\ &=& -\frac{1}{24h^{3}} ([\textbf{e}_{1}|\textbf{e}_{N}]^{T} \otimes I_{N}) R \left[ I_{N} \otimes \left( T^{2}+ 14 T + 48 I_{N} \right) \right]\\ &&\times([\textbf{e}_{1}|\textbf{e}_{N}]\otimes I_{N}). \end{array} $$
(A.22)

Using (A.10), (A.5), (2.10), (A.8), and (A.1), we have

$$\begin{array}{@{}rcl@{}} &&-(Z^{T}\otimes Z^{T}) R \left[ I_{N}\otimes (T^{2}+ 14 T + 48I_{N}) \right] (Z\otimes Z) \\ &=& [(Z^{T}\otimes Z^{T}) P_{11}^{1} (Z\otimes Z)]^{-1} (Z^{T}\otimes Z^{T}) (B^{D}\otimes B^{D}) (Z\otimes Z) \\ &&((Z^{T}\otimes Z^{T}) P_{11}^{1} (Z\otimes Z))^{-1} \\ &&(Z^{T}\otimes Z^{T}) \left[ I_{N}\otimes (T^{2}+ 14 T + 48I_{N}) \right] (Z\otimes Z) \\ &=& {\Lambda}_{2}^{-2} ({\Lambda}_{1}\otimes {\Lambda}_{1}) \left[ I_{N}\otimes ({\Lambda}^{2}+ 14 {\Lambda} + 48 I_{N}) \right]. \end{array} $$
(A.23)

Since the diagonal entries of Λ1 are positive and the diagonal entries of Λ2 are negative, the diagonal entries of \({\Lambda }_{2}^{-2} ({\Lambda }_{1} \otimes {\Lambda }_{1})\) are positive. The diagonal entries of Λ2 + 14Λ + 48IN are positive since

$$x^{2}+ 14 x + 48=(x + 6)(x + 8)> 0, \quad -4\leq x<0. $$

It follows from (A.23) that − R [(IN ⊗ (T2 + 14T + 48IN)] is symmetric and that its eigenvalues are positive. Consequently, the matrix − R [(IN ⊗ (T2 + 14T + 48IN)] is positive definite. For nonzero v in R2N, w = ([e1|eN] ⊗ IN)v is nonzero, and hence, (A.22) and the positive definiteness of − R [(IN ⊗ (T2 + 14T + 48IN)] yield

$$-\textbf{v}^{T} P_{21}^{1} R P_{12}^{2} \textbf{v}=-\frac{1}{24h^{3}} \textbf{w}^{T} R \left[ (I_{N}\otimes (T^{2}+ 14 T + 48I_{N})\right] \textbf{w}>0. $$

This shows that \(-P_{21}^{1} R P_{12}^{2}\) is positive definite. Since \(P_{21}^{1} (P_{11}^{1})^{-1} P_{12}^{1}+P_{21}^{2} (P_{11}^{1})^{-1} P_{12}^{2}\) and \(-P_{21}^{1} R P_{12}^{2}\) are positive definite, it follows from (A.11) that \(P_{21}P_{11}^{-1}P_{12}\) is also positive definite. This completes the proof that P of (4.34) is symmetric and positive definite.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Bialecki, B., Fairweather, G., Karageorghis, A. et al. A quadratic spline collocation method for the Dirichlet biharmonic problem. Numer Algor 83, 165–199 (2020). https://doi.org/10.1007/s11075-019-00676-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11075-019-00676-z

Keywords

Navigation