Accelerated finite elements schemes for parabolic stochastic partial differential equations

Abstract

For a class of finite elements approximations for linear stochastic parabolic PDEs it is proved that one can accelerate the rate of convergence by Richardson extrapolation. More precisely, by taking appropriate mixtures of finite elements approximations one can accelerate the convergence to any given speed provided the coefficients, the initial and free data are sufficiently smooth.

This is a preview of subscription content, access via your institution.

References

  1. 1.

    Asadzadeh, M., Schatz, A., Wendland, W.: A new approach to Richardson extrapolation in the finite element method for second order elliptic problems. Math. Comput. 78(268), 1951–1973 (2009)

    MathSciNet  Article  Google Scholar 

  2. 2.

    Blum, H., Lin, Q., Rannacher, R.: Asymptotic error expansion and Richardson extrapolation for linear finite elements. Numer. Math. 49, 11–37 (1986)

    MathSciNet  Article  Google Scholar 

  3. 3.

    Brzeźniak, Z., Carelli, E., Prohl, A.: Finite-element-based discretizations of the incompressible Navier–Stokes equations with multiplicative random forcing. IMA J. Numer. Anal. 33–3, 771–824 (2013)

    MathSciNet  Article  Google Scholar 

  4. 4.

    Carelli, E., Prohl, A.: Rates of convergence for discretizations of the stochastic incompressible Navier–Stokes equations. SIAM J. Numer. Anal. 50–5, 2467–2496 (2012)

    MathSciNet  Article  Google Scholar 

  5. 5.

    Davie, A.J., Gaines, J.G.: Convergence of numerical schemes for the solution of parabolic stochastic partial differential equations. Math. Comput. 70, 121–134 (2001)

    MathSciNet  Article  Google Scholar 

  6. 6.

    Gerencsér, M., Gyöngy, I.: Localization errors in solving stochastic partial differential equations in the whole space. Math. Comput. 86(307), 2373–2397 (2017)

    MathSciNet  Article  Google Scholar 

  7. 7.

    Gyöngy, I., Krylov, N.V.: Accelerated finite difference schemes for linear stochastic partial differential equations in the whole space. SIAM J. Math. Anal. 43–5, 2275–2296 (2010)

    MathSciNet  Article  Google Scholar 

  8. 8.

    Hall, E.J.: Accelerated spatial approximations for time discretized stochastic partial differential equations. SIAM J. Math. Anal. 44, 3162–3185 (2012)

    MathSciNet  Article  Google Scholar 

  9. 9.

    Hall, E.J.: Higher order spatial approximations for degenerate parabolic stochastic partial differential equations. SIAM J. Math. Anal. 45(4), 2071–2098 (2013)

    MathSciNet  Article  Google Scholar 

  10. 10.

    Hausenblas, E.: Finite element approximation of stochastic partial differential equations driven by poisson random measures of jump type. SIAM J. Numer. Anal. 46–1, 437–471 (2008)

    MathSciNet  Article  Google Scholar 

  11. 11.

    Kovács, M., Larsson, S., Lindgren, F.: Strong convergence of the finite element method with truncated noise for semilinear parabolic stochastic equations with additive noise. Numer. Algorithms 53, 309–320 (2010)

    MathSciNet  Article  Google Scholar 

  12. 12.

    Kovács, M., Printems, J.: Strong order of convergence of a fully discrete approximation of a linear stochastic Volterra type evolution equation. Math. Comput. 83, 2325–2346 (2014)

    MathSciNet  Article  Google Scholar 

  13. 13.

    Krylov, N.V., Rozovskii, B.L.: On Cauchy problem for linear stochastic partial differential equations. Math. USSR Izvestija 11–4, 1267–1284 (1977)

    Article  Google Scholar 

  14. 14.

    Lemaire, V., Pagès, G.: Multi-level Richardson–Romberg extrapolation. Bernoulli 23–4A, 2643–2692 (2017)

    Article  Google Scholar 

  15. 15.

    Malliavin, P., Thalmaier, A.: Numerical error for SDE: Asymptotic expansion and hyperdistributions. C. R. Math. Acad. Sci. Paris 336, 851–856 (2003)

    MathSciNet  Article  Google Scholar 

  16. 16.

    Rannacher, R.: Richardson extrapolation for a mixed finite element approximation of a plain blending problem (Gesellschaft für angwandte Mathematik and Mechanik, Wissenschaftliche Jahrestagung, West Germany, Apr. 1–4, 1986) Zeitschrift für andgewandte Mathematik und Mechanik, vol. 67–5 (1987)

  17. 17.

    Rannacher, R.: Richardson Extrapolation with Finite Elements, Numerical Techniques in Continuous Mechanics, pp. 90–101. Springer, Berlin (1987)

    Google Scholar 

  18. 18.

    Richardson, L.F.: The approximate arithmetical solution by finite differences of physical problems including differential equations, with an application to the stresses in a masonry dam. Philos. Trans. R. Soc. A 210, 307–357 (1911)

    MATH  Google Scholar 

  19. 19.

    Rozovskii, B.L.: Stochastic Evolution Systems. Linear Theory and Applications to Nonlinear Filtering. Kluwer, Dordrecht (1990)

    Book  Google Scholar 

  20. 20.

    Sidi, A.: Practical Extrapolation Methods. Cambridge University Press, Cambridge (2003)

    Book  Google Scholar 

  21. 21.

    Talay, D., Tubaro, L.: Expansion of the global error for numerical schemes solving stochastic differential equations. Stoch. Anal. Appl. 8, 94–120 (1990)

    MathSciNet  Article  Google Scholar 

  22. 22.

    Walsh, J.B.: Finite element methods for parabolic stochastic PDE’s. Potential Anal. 23(1), 1–43 (2005)

    MathSciNet  Article  Google Scholar 

  23. 23.

    Yan, Y.: Galerkin finite element methods for stochastic parabolic partial differential equations. SIAM J. Numer. Anal. 43(4), 1363–1384 (2005)

    MathSciNet  Article  Google Scholar 

Download references

Acknowledgements

This work started while István Gyöngy was an invited professor at the University Paris 1 Panthéon Sorbonne. It was completed when Annie Millet was invited by the University of Edinburgh. Both authors want to thank the University Paris 1, the Edinburgh Mathematical Society and the Royal Society of Edinburgh for their financial support. The authors want to thank the anonymous referees for their careful reading and helpful remarks.

Author information

Affiliations

Authors

Corresponding author

Correspondence to Annie Millet.

Additional information

Publisher's Note

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

Appendix

Appendix

The aim of this section is to prove that the example described in Sect. 6.2 satisfies Assumptions 2.4 and 2.5.

For \(k=1,\ldots , d\), let \(e_k\in {\mathbb {Z}}^d\) denote the k-th unit vector of \({\mathbb {R}}^d\); then \({\mathbb {G}}={\mathbb {Z}}^d\) and

$$\begin{aligned} \Gamma =\left\{ \sum _{k=1}^d \epsilon _k e_k \; : \; \epsilon _k\in \{-1,0,1\} \, \text{ for } k=1, \ldots , d \right\} . \end{aligned}$$

For fixed \(k=1, \ldots , d\) (resp. \(k\ne l \in \{1, \ldots , d\}\)) let

$$\begin{aligned} {{\mathcal {I}}}(k)=\{1, \ldots , d\} \setminus \{k\}, \quad \text{ resp. } {{\mathcal {I}}}(k,l)=\{1, \ldots , d\} \setminus \{k,l\}. \end{aligned}$$
(7.1)

Note that in the particular case \(d=1\), the functions \(\psi \) gives rise to the classical linear finite elements. Then for \(\mathbf{i} \in {{\mathbb {Z}}}^d\), we have for \(k=0, 1, \ldots , d\):

$$\begin{aligned} R_\mathbf{i}:= (\psi _\mathbf{i},\psi )= {d\atopwithdelims ()k} \Big (\frac{1}{6}\Big )^k \Big ( \frac{2}{3}\Big )^{d-k} \quad \text{ if } \; \sum _{l=1}^d |i_l|=k. \end{aligned}$$
(7.2)

Furthermore, given \(k=0, 1, \ldots , d\), there are \(2^k\) elements \(\mathbf{i}\in {\mathbb {Z}}^d\) such that \(\sum _{l=1}^d |i_l|=k\). Therefore, we deduce

$$\begin{aligned} \sum _{\mathbf{i}\in {\mathbb {Z}}^d} (\psi _\mathbf{i},\psi )= \sum _{k=0}^d 2^k {d\atopwithdelims ()k} \Big (\frac{1}{6}\Big )^k \Big ( \frac{2}{3}\Big )^{d-k} = \sum _{k=0}^d {d\atopwithdelims ()k} \Big (\frac{2}{6}\Big )^k \Big ( \frac{2}{3}\Big )^{d-k}=1, \end{aligned}$$

which yields the first compatibility conditon in (2.20).

We at first check that Assumption 2.4 holds true, that is

$$\begin{aligned} \delta \sum _{\mathbf{i}\in {\mathbb {Z}}^d} U_\mathbf{i}^2 = \delta |U|^2_{\ell _2({\mathbb {Z}}^d)} \le \Big |\sum _{\mathbf{i}\in {\mathbb {Z}}^d} U_\mathbf{i} \psi _\mathbf{i}\Big |_{L^2}^2 =\sum _{\mathbf{i}, \mathbf{j}\in {\mathbb {Z}}^d} R_{\mathbf{i}-\mathbf{j}} U_\mathbf{i} U_\mathbf{j}, \quad U\in \ell _2({\mathbb {Z}}^d). \end{aligned}$$

for some \(\delta >0\). For \(U\in \ell _2({{\mathbb {Z}}}^d)\) and \(k=1, \ldots , d,\) let \(T_k U=U_{e_k}\), where \(e_k\) denotes the k-th vector of the canonical basis.

For \(U\in \ell _2({\mathbb {Z}}^d)\) we have

$$\begin{aligned} \Big | \sum _\mathbf{i} U_\mathbf{i} \psi _\mathbf{i}\Big |_{L^2}^2&= \sum _\mathbf{i\in {\mathbb {Z}}^d} \int _{[0,1]^d} \left[ U_\mathbf{i} \prod _{j=1}^d ( 1-{x_j}) + \sum _{k=1}^d (T_k U)_\mathbf{i}\; x_k \prod _{j\in {{\mathcal {I}}}(k)}(1-x_j) \right. \\&\left. \quad + \sum _{1\le k_1<k_2\le d} (T_{k_1} \circ T_{k_2} U)_\mathbf{i}\; x_{k_1} x_{k_2}\right. \\&\left. \quad \times \prod _{j\in {{\mathcal {I}}}(k_1,k_2)}(1-x_j) +\cdots + (T_1 \circ T_2 \cdots \circ T_d U)_\mathbf{i} \prod _{k=1}^d x_k\right] ^{2} dx, \end{aligned}$$

Given \(\alpha \in (0,1)\) if we let

$$\begin{aligned} I(\alpha )= & {} \int _0^\alpha (1-x)^2 dx = \alpha - \alpha ^2 + \frac{\alpha ^3}{3}, \\ J(\alpha )= & {} \int _0^\alpha x(1-x) dx = \frac{\alpha ^2}{2}-\frac{\alpha ^3}{3}, \; K(\alpha )= \int _0^\alpha x^2 dx = \frac{\alpha ^3}{3} , \end{aligned}$$

restricting the above integral on the set \([0,\alpha ]^d\), expanding the square and using the Cauchy Schwarz inequality we deduce the existence of some constants \(C(\gamma _1,\gamma _2, \gamma _3)\) defined for \(\gamma _i\in \{0,1, \ldots , d\}\) such that

$$\begin{aligned} \Big | \sum _\mathbf{i} U_\mathbf{i} \psi _\mathbf{i}\Big |_{L^2}^2 \ge&\sum _\mathbf{i} |U_\mathbf{i}|^2 \Big [ I(\alpha )^d + {d\atopwithdelims ()1} K(\alpha )\, I(\alpha )^{d-1} \\&+ {d\atopwithdelims ()2} K(\alpha )^2\, I(\alpha )^{d-2} + \cdots + K(\alpha )^d\Big ] \\&-\, 2 \Big ( \sum _\mathbf{i} |U_\mathbf{i}|^2 \Big ) \sum _{\gamma _1+\gamma _2+\gamma _3=d, \gamma _2+\gamma _3\ge 1} C(\gamma _1,\gamma _2, \gamma _3)\; I(\alpha )^{\gamma _1}\; J(\alpha )^{\gamma _2} \; K(\alpha )^{\gamma _3}\\ \ge&\, |U|_{\ell _2({{\mathbb {Z}}}^d)}^2 \Big ( \alpha ^d - \sum _{l=d+1}^{3d} C_l \alpha ^l \Big ), \end{aligned}$$

where \(C_l\) are some positive constants. Choosing \(\alpha \) small enough, we have \( \big | \sum _\mathbf{i} U_\mathbf{i} \psi _\mathbf{i}\big |_{L^2}^2 \ge \frac{\alpha ^d}{2} |U|_{\ell _2({{\mathbb {Z}}}^d)}^2\), which implies the invertibility Assumption 2.4.

We now prove that the compatibility Assumption 2.5 holds true. For \(l=1, \ldots , d\), \(n=0, \ldots , d-1\):

$$\begin{aligned} (D_l \psi _\mathbf{i} , D_l\psi )&= - \, 2^{d-1-n} \, \Big ( \frac{1}{6}\Big )^n \, \Big ( \frac{1}{3}\Big )^{d-1-n} \; \text{ for } |i_l|=1, \; \sum _{r \ne l, 1\le n\le d} |i_r|=n, \end{aligned}$$
(7.3)
$$\begin{aligned} (D_l \psi _\mathbf{i} , D_l \psi )&= + \, 2^{d-n} \, \Big (\frac{1}{6}\Big )^n \, \Big ( \frac{1}{3}\Big )^{d-1-n} \; \text{ for } |i_l|=0, \; \sum _{r\ne l, 1\le r\le d} |i_r|=n. \end{aligned}$$
(7.4)

For \(n=1, \ldots , d-1\) and \(k_1<k_2<\cdots < k_n\) with \(k_r\in {{\mathcal {I}}}(l)\), where \({{\mathcal {I}}}(l)\) is defined in (7.1), let

$$\begin{aligned} \Gamma _l(k_1, \ldots , k_n)&=\Big \{ \sum _{r=1}^n \epsilon _{k_r} e_{k_r} \; : \; \epsilon _{k_r}\in \{ -1, 1\}, r=1, \ldots , n \Big \} ,\\ \Gamma _l(l;k_1, \ldots , k_n)&=\Big \{ \epsilon _l e_l + \sum _{r=1}^n \epsilon _{k_r} e_{k_r} \; : \; \epsilon _l \in \{-1,1\} \; \text{ and } \epsilon _{k_r}\in \{ -1, 1\}, r=1, \ldots , n\Big \} . \end{aligned}$$

Then \(| \Gamma _l (k_1, \ldots , k_n)|=2^n\) while \(| \Gamma _l (l;k_1, \ldots , k_n)|=2^{n+1}\). For \(l=1, \ldots , d\), the identities (7.3) and (7.4) imply

$$\begin{aligned} \sum _{\lambda \in \Gamma }&(D_l \psi , D_l \psi _\lambda ) = \left[ (D_l\psi , D_l\psi ) + \sum _{\epsilon _l\in \{-1,+1\}} (D_l \psi , D_l \psi _{\epsilon _l e_l})\right] \\&\quad + \sum _{n=1}^{d-1} \sum _{k_1<k_2< \cdots <k_n, k_j\in {{\mathcal {I}}}(l)} \left[ \sum _{\lambda \in \Gamma _l(k_1,\ldots , k_n)} (D_l\psi , D_l\psi _\lambda ) + \sum _{\lambda \in \Gamma _l(l; k_1, \ldots , k_n)} (D_l\psi , D_l\psi _\lambda ) \right] \\&= \Big [ 2^d \Big ( \frac{1}{3}\Big )^{d-1} -2\times 2^{d-1} \Big ( \frac{1}{3}\Big )^{d-1}\Big ] \\&\quad +\sum _{n=1}^{d-1} 2^n {d-1 \atopwithdelims ()n} \Big [ 2^{d-n} \Big ( \frac{1}{6} \Big )^n \Big ( \frac{1}{3}\Big )^{d-1-n} -2\times 2^{d-1-n} \Big (\frac{1}{6}\Big )^n \Big ( \frac{1}{3}\Big )^{d-1-n}\Big ] =0 . \end{aligned}$$

This proves the second identity in (2.20) when \(i=j\). Furthermore, (7.3) implies

$$\begin{aligned} \sum _{\lambda \in \Gamma } R^{ll}_\lambda \lambda _l \lambda _l&= - \sum _{\lambda \in \Gamma } (D_l \psi , D_l \psi _\lambda ) \lambda _l \lambda _l = - \sum _{\epsilon _l\in \{-1,1\}} \big (D_l \psi , D_l \psi _{\epsilon _l e_l} \big ) \\&\quad - \sum _{n=1}^{d-1} \sum _{k_1<k_2< \cdots <k_n, k_j\in {{\mathcal {I}}}(l)} \sum _{\lambda \in \Gamma _l(l;k_1, \ldots , k_n)} (D_l \psi , D_l \psi _\lambda )\\&= 2\times 2^{d-1} \Big (\frac{1}{3}\Big )^{d-1} + \sum _{d=1}^n {d-1 \atopwithdelims ()n} 2^{n+1} \times 2^{d-1-n} \Big ( \frac{1}{6}\Big )^n \Big ( \frac{1}{3}\Big )^{d-1-n} \\&= 2 \sum _{n=0}^{d-1} {d-1 \atopwithdelims ()n} \Big (\frac{2}{6}\Big )^n \Big (\frac{2}{3}\Big )^{d-1-n} =2,\quad l=1,\ldots ,d. \end{aligned}$$

Furthermore, given \(k\ne l \in \{1, \ldots , d\}\),

$$\begin{aligned} \sum _{\lambda \in \Gamma } R^{ll}_\lambda \lambda _k\lambda _k&= - \sum _{\lambda \in \Gamma } (D_l \psi , D_l \psi _\lambda ) \lambda _k \lambda _l=0. \end{aligned}$$

Indeed, for \(n=1, \ldots , d-1\), \(k_1<\cdots < k_n \) where \(k_r\in {{\mathcal {I}}}(l)\) and at least one of the indices \(k_r\) is equal to k for \(r=1, \ldots ,n\), given \(\lambda \in \Gamma _l(k_1, \ldots , k_n)\) we have using (7.3) and (7.4)

$$\begin{aligned} \sum _{\epsilon _l\in \{-1,1\}} (D_l\psi ,D_l\psi _{\epsilon _le_l+\lambda }) \lambda _k \lambda _l = - 2^{d-1-n} \Big (\frac{1}{6}\Big )^n \Big (\frac{1}{3}\Big )^{d-1-n} \times (-1+1)=0. \end{aligned}$$

This proves the second identity in (2.22) when both derivatives agree.

Also note that for \(k\ne l \in \{1, \ldots , d\}\) we have \(\sum _{\lambda \in \Gamma } R^{kl}_\lambda =0\). Indeed, for \(\lambda \) as above

$$\begin{aligned} (D_k \psi , D_l \psi _\lambda )&+ \sum _{\epsilon _l\in \{-1,1\}} (D_k\psi , D_l\psi _{\epsilon _le_l+\lambda }) \\&= 2^{d-n} \Big (\frac{1}{6}\Big )^n \Big (\frac{1}{3}\Big )^{d-1-n} -2\times 2^{d-1-n} \Big (\frac{1}{6}\Big )^n \Big (\frac{1}{3}\Big )^{d-1-n} =0, \end{aligned}$$

while \(R^{kl}_\lambda =0\) for other choices of \(\lambda \in \Gamma \).

We now study the case of mixed derivatives. Given \(k\ne l \in \{1, \ldots , d\}\) recall that \({{\mathcal {I}}}(k,l)=\{ 1, \ldots , d\} \setminus \{k,l\}\). Then for \(k\ne l \in \{1, \ldots , d\}\) and \(\mathbf{i} \in {\mathbb {Z}}^d\) we have for \(n=0, \ldots , d-2\)

$$\begin{aligned} (D_k \psi _\mathbf{i}, D_l \psi )&=\,0 \; \text{ if } |i_k \, i_l|\ne 1, \end{aligned}$$
(7.5)
$$\begin{aligned} (D_k \psi _\mathbf{i}, D_l \psi )&= - \,\Big (\frac{1}{2}\Big )^2 \; \Big ( \frac{1}{6}\Big )^n \; \Big (\frac{2}{3}\Big )^{d-n-2} \; \text{ if } \; i_k\, i_l=1, \quad \sum _{r\in {{\mathcal {I}}}(k,l)} |i_r|=n, \end{aligned}$$
(7.6)
$$\begin{aligned} (D_k \psi _\mathbf{i}, D_l \psi )&= +\, \Big (\frac{1}{2}\Big )^2 \; \Big ( \frac{1}{6}\Big )^n \; \Big (\frac{2}{3}\Big )^{d-n-2} \; \text{ if } \; i_k \, i_l=- 1, \quad \sum _{r\in {{\mathcal {I}}}(k,l)} |i_r|=n. \end{aligned}$$
(7.7)

For \(n=1, \ldots , d-2\) and \(k_1<\cdots <k_n\) with \(k_r\in {{\mathcal {I}}}(k,l)\) for \(r=1,\ldots , n\), set

$$\begin{aligned} \Gamma _{k,l}(k_1, \ldots , k_n)=\Big \{ \sum _{r=1}^n \epsilon _{k_r} e_{k_r} \; : \; \epsilon _r\in \{-1,1\} \Big \}. \end{aligned}$$

For \(n=0\) there is no such family of indices \(k_1<\cdots <k_n\) and we let \(\Gamma _{k,l}(\emptyset )=\{0\}\). Thus for \(n=0, \ldots , d-2\), \(| \Lambda _{k,l}(k_1,\ldots , k_n)|=2^n\). Using the identities (7.5)–(7.7) we deduce

$$\begin{aligned} \sum _{\lambda \in \Gamma }&(D_k \psi , D_l \psi _\lambda ) = \sum _{n=0}^{d-2} \; \sum _{k_1<k_2<\cdots <k_n, k_r\in {{\mathcal {I}}}(k,l)} \sum _{\lambda \in \Gamma _{k,l}(k_1,\ldots , k_n)} \big [ (D_k\psi , D_l\psi _{e_k+e_l+\lambda }) \nonumber \\&\quad + (D_k\psi , D_l\psi _{e_k-e_l+\lambda })+ (D_k\psi , D_l\psi _{-e_k+e_l+ \lambda }) + (D_k\psi , D_l\psi _{-e_k-e_l+\lambda }) \big ] \nonumber \\ =&\sum _{n=0}^{d-2} {d-2 \atopwithdelims ()n} \, 2^n \, \Big [ - \Big ( \frac{1}{2}\Big )^2 \, \Big (\frac{1}{6}\Big )^n \, \Big (\frac{2}{3}\Big )^{d-2-n} + \Big ( \frac{1}{2}\Big )^2 \, \Big (\frac{1}{6}\Big )^n \, \Big (\frac{2}{3}\Big )^{d-2-n} \nonumber \\&\quad + \Big ( \frac{1}{2}\Big )^2 \, \Big (\frac{1}{6}\Big )^n \, \Big (\frac{2}{3}\Big )^{d-2-n} - \Big ( \frac{1}{2}\Big )^2 \, \Big (\frac{1}{6}\Big )^n \, \Big (\frac{2}{3}\Big )^{d-2-n}\Big ] = 0,\quad k\ne l. \end{aligned}$$
(7.8)

This completes the proof of the second identity in (2.20) when \(i\ne j\), and hence (2.20) holds true. Furthermore, the identities (7.6) and (7.7) imply for \(i\ne j \in \{1, \ldots , d\}\) and \(\{i,j\}=\{k,l\}\)

$$\begin{aligned} \sum _{\lambda \in \Gamma }&(D_k \psi , D_l \psi _\lambda )\lambda _k \lambda _l = \sum _{n=0}^{d-2} \; \sum _{k_1<k_2<\cdots <k_n, k_r\in {{\mathcal {I}}}(k,l)} \sum _{\lambda \in \Gamma _{k,l}(k_1,\ldots , k_n)} \big [ (D_k\psi , D_l\psi _{e_k+e_l+\lambda }) \\&\qquad -\, (D_k\psi , D_l\psi _{e_k-e_l+\lambda }) - (D_k\psi , D_l\psi _{-e_k+e_l+ \lambda }) + (D_k\psi , D_l\psi _{-e_k-e_l+\lambda }) \big ] \\&\quad = -\, 4 \, \Big ( \frac{1}{2}\Big )^2 \sum _{n=0}^{d-2} {d-2 \atopwithdelims ()n} \, 2^n \, \Big (\frac{1}{6}\Big )^n \, \Big (\frac{2}{3}\Big )^{d-2-n}\\&\quad =- \sum _{n=0}^{d-2} {d-2 \atopwithdelims ()n} \, \Big (\frac{2}{6}\Big )^n \Big ( \frac{2}{3}\Big )^{d-2-n}= -1. \end{aligned}$$

Equation (7.5) proves that \((D_k \psi , D_l \psi _\lambda )=0\) if \(|\lambda _k \lambda _l|\ne 1\). Hence using (7.8) we deduce that for any \(r=1, \ldots , d\),

$$\begin{aligned} \sum _{\lambda \in \Gamma } (D_k \psi , D_l \psi _\lambda )\lambda _r \lambda _r =0. \end{aligned}$$

Let \(r\in {{\mathcal {I}}}(k,l)\) and for \(n=1, \ldots , d-3\), let \(k_1<\cdots < k_n\) be such that \(k_j\in \{1, \ldots , d\} \setminus \{k,l,r\}\) and \(\lambda = \sum _{j=1}^n \epsilon _{k_j} e_{k_j}\) for \(\epsilon _{k_j}\in \{-1,1\}\), \(j=1, \ldots , n\). Then for any choice of \(\epsilon _k\) and \(\epsilon _l\) in \(\{-1,1\}\) the equalities (7.6) and (7.7) imply that

$$\begin{aligned} (D_k\psi , D_l \psi _{\lambda + \epsilon _k e_k + \epsilon _l e_l + e_r} ) = (D_k\psi , D_l \psi _{\lambda + \epsilon _k e_k + \epsilon _l e_l - e_r} ) . \end{aligned}$$

This clearly yields that for \(r\in {{\mathcal {I}}}(k,l)\) we have

$$\begin{aligned} \sum _{\lambda \in \Gamma } (D_k \psi , D_l \psi _\lambda )\lambda _k \lambda _r = \sum _{\lambda \in \Gamma } (D_k \psi , D_l \psi _\lambda )\lambda _l \lambda _r =0. \end{aligned}$$

Finally, given \(n=2, \ldots , d\) and \(k_1< \cdots <k_n\) where the terms \(k_j\in {{\mathcal {I}}}(k,l)\), then given any choice of \(\epsilon _k\) and \(\epsilon _l\) in \(\{-1,1\}\), the value of \((D_k \psi , D_l \psi _{\epsilon _k e_k + \epsilon _l e_l + \lambda })\) does not depend on the value of \(\lambda \in \Gamma _{k,l}(k_1,\ldots , k_n)\). Therefore, if we fix \(r_1\ne r_2\) in the set \({{\mathcal {I}}}(k,l)\), for fixed n there are as many choices of indices \(k_1<\cdots <k_n\) such that \(\epsilon _{r_1} \epsilon _{r_2}=1\) that of such indices such that \(\epsilon _{r_1} \epsilon _{r_2}=-1\). This proves

$$\begin{aligned} \sum _{\lambda \in \Gamma } (D_k\psi , D_l\psi _\lambda ) \lambda _{r_1} \lambda _{r_2}=0, \end{aligned}$$

which completes the proof of the first identity in (2.22) for mixed derivatives; hence (2.22) holds true.

We now check the compatibility condition (2.21). Fix \(j\in \{1, \ldots , d\}\); then

$$\begin{aligned} (D_j\psi , \psi )= & {} 2^{d-1} \Big ( \prod _{k\ne j} \int _0^1 (1-x_k)^2 dx_k\Big ) \nonumber \\&\quad \times \Big [ \int _0^1 (-1) (1-x_j) dx_j + \int _{-1}^0 (1+x_j) dx_j\Big ] =0, \end{aligned}$$
(7.9)

while

$$\begin{aligned} (D_j \psi , \psi _{e_j})&=2^{d-1} \Big ( \prod _{k\ne j} \int _0^1 (1-x_k)^2 dx_k\Big ) \int _0^1 (-1) \big (1+(x_j-1)\big ) dx_j = -\, \frac{1}{2} \Big (\frac{2}{3}\Big )^{d-1}, \nonumber \\ (D_j \psi , \psi _{-e_j})&=2^{d-1} \Big ( \prod _{k\ne j} \int _0^1 (1-x_k)^2 dx_k\Big ) \int _{-1}^0 \big (1- (x_j+1)\big ) dx_j = \frac{1}{2} \Big (\frac{2}{3}\Big )^{d-1}. \end{aligned}$$
(7.10)

For \(n=1, \ldots , d-1\) and \(k_1<\cdots <k_n\) where the indexes \(k_r\), \(r=1, \ldots , n\) are different from j we have for any \(\lambda \in \Gamma _j(k_1,\ldots , k_n)\)

$$\begin{aligned} (D_j\psi , \psi _{\lambda }) =&\, 2^{d-(n+1)} \Big (\prod _{k\in \Gamma \setminus \{j,k_1, \ldots , k_n\}} \int _0^1 (1-x_k)^2 dx_k\Big ) \times \Big ( \prod _{r=1}^n \int _0^1 x_{k_r} (1-x_{k_r}) dx_{k_r} \Big ) \nonumber \\&\times \, \Big [ \int _0^1 (-1) (1-x_j) dx_j + \int _{-1}^0 (1+x_j) dx_j\Big ] = 0, \end{aligned}$$
(7.11)

while

$$\begin{aligned} (D_j\psi , \psi _{e_j+\lambda }) =&\, 2^{d-(n+1)} \Big (\prod _{k\in \Gamma \setminus \{j,k_1, \ldots , k_n\}} \int _0^1 (1-x_k)^2 dx_k\Big )\nonumber \\&\times \Big ( \prod _{r=1}^n \int _0^1 x_{k_r} (1-x_{k_r}) dx_{k_r} \Big ) \nonumber \\&\times \int _0^1 (-1) \big ( 1+ (x_j-1)\big ) dx_j = -\frac{1}{2} \; \Big ( \frac{2}{3}\Big )^{d-(n+1)} \; \Big ( \frac{1}{6}\Big )^n , \end{aligned}$$
(7.12)

and

$$\begin{aligned} (D_j\psi , \psi _{-e_l+\lambda }) =&\, 2^{d-(n+1)} \Big (\prod _{k\in \Gamma \setminus \{j,k_1,\ldots , k_n\}} \int _0^1 (1-x_k)^2 dx_k\Big )\nonumber \\&\times \Big ( \prod _{r=1}^n \int _0^1 x_{k_r} (1-x_{k_r}) dx_{k_r} \Big ) \nonumber \\&\times \int _{-1}^0 \big ( 1- (x_j+1)\big ) dx_j = \frac{1}{2} \; \Big ( \frac{2}{3}\Big )^{d-(n+1)} \; \Big ( \frac{1}{6}\Big )^n . \end{aligned}$$
(7.13)

Note that the number of terms \( (D_j\psi , \psi _{\epsilon _l e_l+\lambda }) \) with \(\epsilon _l=-1\) or \(\epsilon _l =+1\) is equal to \({d-1 \atopwithdelims ()n} 2^n\). Therefore, the identities (7.9)–(7.13) imply that for any \(j=1, \ldots , d\) we have

$$\begin{aligned}&\sum _{\lambda \in \Gamma } \lambda _j R^j_\lambda =- \sum _{\lambda \in \Gamma } (D_j \psi , \psi _\lambda )\; \lambda _j = \frac{1}{2} \Big (\frac{2}{3}\Big )^{d-1} - \frac{1}{2} \Big (\frac{2}{3}\Big )^{d-1} (-1) \end{aligned}$$
(7.14)
$$\begin{aligned}&\qquad + \frac{1}{2} \sum _{n=1}^{d-1} {d-1 \atopwithdelims ()n} 2^n \Big ( \frac{2}{3}\Big )^{d-1-n} \Big ( \frac{1}{6}\Big )^n - \frac{1}{2}\sum _{n=0}^{d-1} {d-1 \atopwithdelims ()n} 2^n \Big ( \frac{2}{3}\Big )^{d-1-n} \Big ( \frac{1}{6}\Big )^n \times (-1) \nonumber \\&\quad = \sum _{n=0}^{d-1} {d-1 \atopwithdelims ()n} \Big ( \frac{2}{6}\Big )^n \Big (\frac{2}{3}\Big )^{d-1-n} = 1. \end{aligned}$$
(7.15)

This proves (2.21) when \(i=k\).

Let \(k\ne j \in \{1,\ldots , d\}\) and given \(n=1,\ldots , d-1\) let \(k_1<\cdots < k_n\) be indices that belong to \({{\mathcal {I}}}(j)\) such that one of the indices \(k_r, r=1,\ldots , n\) is equal to k. Given any \(\lambda \in \Gamma _j(k_1,\ldots , k_n)\) we deduce that

$$\begin{aligned} (D_j \psi , \psi _{e_l+\lambda }) \lambda _k + (D_j \psi , \psi _{-e_l+\lambda }) \lambda _k =0. \end{aligned}$$

This completes the proof of the identity (2.21).

In order to complete the proof of the validity of Assumption 2.5, it remains to check that the identities in (2.23) hold true. Recall that for \(\lambda \in \Gamma \) and \(i,j,k,l\in \{1,\ldots ,d\}\) we have

$$\begin{aligned} Q_\lambda ^{ij,kl}=\int _{{\mathbb {R}}^d} z_k z_l D_j\psi _\lambda (z) D_i^*\psi (z) dz =-\int _{{\mathbb {R}}^d} z_k z_l D_j\psi _\lambda (z) D_i \psi (z) dz. \end{aligned}$$

For \(p=1,\ldots , 4\), \(n=1,\ldots , d-p\) and \(i_1,\ldots , i_p \in \{1, \ldots ,d\}\) with \(i_1,\ldots , i_p\) pairwise different let

$$\begin{aligned}&{\mathcal {I}}_n(i_1,\ldots , i_p):=\Big \{ \sum _{\alpha =1}^n \epsilon _\alpha e_{k_\alpha } ; \epsilon _\alpha \in \{-1,+1\}, \; 1\le k_1< \cdots <k_n\le d, \\&\quad k_\alpha \not \in \{i_1,\ldots , i_p\} \, \text{ for } \alpha =1,\ldots , n\Big \}, \end{aligned}$$

and \( {\mathcal {I}}_0(i_1,\ldots , i_p)=\{0\}\).

First suppose that \(i=j\).

First let \(k=l=i\); then for \(n=0,\ldots , d-1\) and \( \mu \in {\mathcal {I}}_n(i)\) we have

$$\begin{aligned} Q_{\mu }^{ii,ii} + Q_{\mu + e_i}^{ii,ii} + Q_{\mu - e_i}^{ii,ii}=0. \end{aligned}$$

Let \(k=l\) with \(k\ne i\); then then for \(n=0,\ldots , d-1\) and \(\mu \in {\mathcal {I}}_n(i)\) we have

$$\begin{aligned} Q_{\mu }^{ii,kk} + Q_{\mu + e_i}^{ii,kk} + Q_{\mu -e_i}^{ii,kk}=0. \end{aligned}$$

Let \(l=i\) and \(k\ne i\); then for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1,+1\} \) and \( \mu \in {\mathcal {I}}_n(i,k)\) we have

$$\begin{aligned} Q_{\mu + \epsilon e_i + e_k}^{ii,ki} + Q_{\mu + \epsilon e_i - e_k}^{ii,ki}=0. \end{aligned}$$

A similar result holds for \(k=i\) and \(l\ne i\). Furthermore, \(Q_\lambda ^{ii,ki}=0\) if \(\lambda \) is not equal to \(\mu + \epsilon e_i + e_k\) or \(\mu + \epsilon e_i - e_k\) with \(\mu \in {\mathcal {I}}_n(i,k)\) for some n.

Let \(k\ne l\) with \(k\ne i\) and \(l\ne i\); then for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1,+1\}\) and \(\mu \in {\mathcal {I}}_n(k,l)\) we have

$$\begin{aligned} Q_{\mu + \epsilon e_k + e_l}^{ii,kl} + Q_{\mu + \epsilon e_k - e_l}^{ii,kl}=0, \end{aligned}$$

while \(Q_\lambda ^{ii,kl}=0\) if \(\lambda \) is not equal to \(\mu + \epsilon e_i + e_k\) or \(\mu + \epsilon e_i - e_k\) with \(\mu \in {\mathcal {I}}_n(i,k)\) for some n.

We now suppose that \(i\ne j\).

First suppose that \(k=i\) and \(l=j\); then for \(n=0,\ldots , d-1\) and \(\mu \in {\mathcal {I}}_n(i)\) we have

$$\begin{aligned} Q_{\mu }^{ij,ij} + Q_{\mu + e_j}^{ij,ij} + Q_{\mu -e_j}^{ij,ij}=0. \end{aligned}$$

Let \(k=l=i\); then for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1+1\}\) and \(\mu \in {\mathcal {I}}_n(i,j)\) we have

$$\begin{aligned} Q_{\mu + \epsilon e_i +e_j}^{ij,ii} + Q_{\mu + \epsilon e_i -e_j}^{ij,ii}=0, \end{aligned}$$

while \(Q_\lambda ^{ij,ii}=0\) is \(\lambda \) is not equal to \(\mu +\epsilon e_i +e_j\) or \(\mu + \epsilon e_i-e_j\) where \(\mu \in {\mathcal {I}}_n(i,j)\) for some n. A similar result holds exchanging i and j for \(k=l=j\).

Let \(k=l\) with \(k\not \in \{i,j\}\) and \(l\not \in \{i,j\}\); then for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1,+1\}\) and \(\mu \in {\mathcal {I}}_n(i,j)\) we have

$$\begin{aligned} Q_{\mu + \epsilon e_i +e_j}^{ij,kk} + Q_{\mu + \epsilon e_i -e_j}^{ij,kk}=0, \end{aligned}$$

while \(Q_\lambda ^{ij,kk}=0\) if \(\lambda \) is not equal to \(\mu +\epsilon e_i +e_j\) or \(\mu + \epsilon e_i-e_j\) with \(\mu \in {\mathcal {I}}_n(i,j)\) for some n.

Let \(l=i\) and \(k\not \in \{i,j\}\); then for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1+1\}\) and \(\mu \in {\mathcal {I}}_n(i,k)\) we have

$$\begin{aligned} Q_{\mu + \epsilon e_i +e_k}^{ij,ki} + Q_{\mu + \epsilon e_i -e_k}^{ij,ki}=0, \end{aligned}$$

while \(Q_\lambda ^{ij,ki}=0\) is \(\lambda \) is not equal to \(\mu +\epsilon e_i +e_k\) or \(\mu + \epsilon e_i-e_k\) where \(\mu \in {\mathcal {I}}_n(i,k)\) for some n. A similar result holds exchanging i and j for \(k=l=j\).

Finally, let \(k\ne l\) with \(k\not \in \{i,j\}\) and \(l\not \in \{i,j\}\); then for \(n=0,\ldots , d-4\), \(\epsilon _i, \epsilon _j, \epsilon _k \in \{-1+1\}\) and \(\mu \in {\mathcal {I}}_n(i,j,k,l)\) we have

$$\begin{aligned} Q_{\mu + \epsilon _i e_i +\epsilon _j e_j + \epsilon _k e_k +e_l}^{ij,kl} + Q_{\mu + \epsilon _i e_i +\epsilon _j e_j + \epsilon _k e_k -e_l}^{ij,kl}=0, \end{aligned}$$

while \(Q_\lambda ^{ij,kl}=0\) is \(\lambda \) is not equal to \(\mu + \epsilon _i e_i +\epsilon _j e_j + \epsilon _k e_k +e_l\) or \(\mu + \epsilon _i e_i +\epsilon _j e_j + \epsilon _k e_k -e_l\) where \(\mu \in {\mathcal {I}}_n(i,j,k,l)\) for some n. These computations complete the proof of the first identity in (2.23). Recall that for \(i,k\in \{1,\ldots , d\}\) and \(\lambda \in \Gamma \) we let

$$\begin{aligned} {\tilde{Q}}_\lambda ^{i,k}:=\int _{{\mathbb {R}}^d} z_k D_i\psi _\lambda (z) \psi (z) dz. \end{aligned}$$

Let \(k=i\); for \(n=0,\ldots , d-1\) and \(\mu \in {\mathcal {I}}_n(i)\) we have

$$\begin{aligned} {\tilde{Q}}^{i,i}_\mu + {\tilde{Q}}^{i,i}_{\mu + e_i} + {\tilde{Q}}^{i,i}_{\mu -e_i}=0. \end{aligned}$$

Let \(k\ne i\); for \(n=0,\ldots , d-2\), \(\epsilon \in \{-1,0, +1\}\) and \(\mu \in {\mathcal {I}}_n(i,k)\) we have

$$\begin{aligned} {\tilde{Q}}_{\mu + \epsilon e_i + e_k}^{i,k} + {\tilde{Q}}_{\mu + \epsilon e_i - e_k}^{i,k}=0 \end{aligned}$$

while \({\tilde{Q}}_\lambda ^{i,k}=0\) if \(\lambda \) is not equal to \(\mu +\epsilon e_i +e_k\) or \(\mu + \epsilon e_i-e_k\) where \(\mu \in {\mathcal {I}}_n(i,k)\) for some n. This completes the proof of the second identity in (2.23); therefore Assumption 2.5 is satisfied for these finite elements. This completes the verification of the validity of Assumptions 2.42.5 for the function \(\psi \) defined by (6.2).

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Gyöngy, I., Millet, A. Accelerated finite elements schemes for parabolic stochastic partial differential equations. Stoch PDE: Anal Comp 8, 580–624 (2020). https://doi.org/10.1007/s40072-019-00154-6

Download citation

Keywords

  • Stochastic parabolic equations
  • Richardson extrapolation
  • Finite elements

Mathematics Subject Classification

  • Primary 60H15
  • 65M60
  • Secondary 65M15
  • 65B05