a survey of numerical methods in fractional calculus...in fractional calculus fractional derivatives...
TRANSCRIPT
![Page 1: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/1.jpg)
A Survey of Numerical Methodsin Fractional Calculus
Fractional Derivatives in Mechanics: State of the ArtCNAM Paris, 17 November 2006
Kai [email protected], [email protected]
Gesellschaft fur numerische Simulation mbH, BraunschweigInstitut Computational Mathematics, TU Braunschweig
K. Diethelm, Numerical Methods in Fractional Calculus – p. 1/21
![Page 2: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/2.jpg)
Outline
The basic problem
Grunwald-Letnikov methods
Lubich’s fractional linear multistep methods
Methods based on quadrature theory
Examples
Extensions
K. Diethelm, Numerical Methods in Fractional Calculus – p. 2/21
![Page 3: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/3.jpg)
The Basic Problem
Find a numerical solution for the fractional differential equation
Dαy(x) = f(x, y(x))
with appropriate initial condition(s), where Dα is
either the Riemann-Liouville derivative
or the Caputo derivative.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 3/21
![Page 4: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/4.jpg)
Fundamental Definitions
Riemann-Liouville integral: Jαf(x) =1
Γ(α)
∫ x
0
(x−t)α−1f(t)dt
Riemann-Liouville derivative: Dαf(x) = DnJn−αf(x), n=dαe
natural initial condition for DE: Dα−ky(0) = y(k)0 , k=1, 2, . . . , n
Caputo derivative: Dα∗ f(x) = Jn−αDnf(x), n=dαe
natural initial condition for DE: Dky(0) = y(k)0 , k=0, 1, . . . , n−1
K. Diethelm, Numerical Methods in Fractional Calculus – p. 4/21
![Page 5: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/5.jpg)
Fundamental Definitions
Riemann-Liouville integral: Jαf(x) =1
Γ(α)
∫ x
0
(x−t)α−1f(t)dt
Riemann-Liouville derivative: Dαf(x) = DnJn−αf(x), n=dαe
natural initial condition for DE: Dα−ky(0) = y(k)0 , k=1, 2, . . . , n
Caputo derivative: Dα∗ f(x) = Jn−αDnf(x), n=dαe
natural initial condition for DE: Dky(0) = y(k)0 , k=0, 1, . . . , n−1
K. Diethelm, Numerical Methods in Fractional Calculus – p. 4/21
![Page 6: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/6.jpg)
Fundamental Definitions
Riemann-Liouville integral: Jαf(x) =1
Γ(α)
∫ x
0
(x−t)α−1f(t)dt
Riemann-Liouville derivative: Dαf(x) = DnJn−αf(x), n=dαe
natural initial condition for DE: Dα−ky(0) = y(k)0 , k=1, 2, . . . , n
Caputo derivative: Dα∗ f(x) = Jn−αDnf(x), n=dαe
natural initial condition for DE: Dky(0) = y(k)0 , k=0, 1, . . . , n−1
K. Diethelm, Numerical Methods in Fractional Calculus – p. 4/21
![Page 7: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/7.jpg)
Fundamental Definitions
Riemann-Liouville integral: Jαf(x) =1
Γ(α)
∫ x
0
(x−t)α−1f(t)dt
Riemann-Liouville derivative: Dαf(x) = DnJn−αf(x), n=dαe
natural initial condition for DE: Dα−ky(0) = y(k)0 , k=1, 2, . . . , n
Caputo derivative: Dα∗ f(x) = Jn−αDnf(x), n=dαe
natural initial condition for DE: Dky(0) = y(k)0 , k=0, 1, . . . , n−1
K. Diethelm, Numerical Methods in Fractional Calculus – p. 4/21
![Page 8: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/8.jpg)
Fundamental Definitions
Riemann-Liouville integral: Jαf(x) =1
Γ(α)
∫ x
0
(x−t)α−1f(t)dt
Riemann-Liouville derivative: Dαf(x) = DnJn−αf(x), n=dαe
natural initial condition for DE: Dα−ky(0) = y(k)0 , k=1, 2, . . . , n
Caputo derivative: Dα∗ f(x) = Jn−αDnf(x), n=dαe
natural initial condition for DE: Dky(0) = y(k)0 , k=0, 1, . . . , n−1
K. Diethelm, Numerical Methods in Fractional Calculus – p. 4/21
![Page 9: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/9.jpg)
Grunwald-Letnikov Methods
Integer-order derivatives:
Dn y(x) = limh→0
1
hn
∞∑
k=0
(−1)k Γ(n + 1)
Γ(k + 1)Γ(n − k + 1)y(x−kh) (n ∈ N)
Fractional extension (Liouville 1832, Grunwald 1867, Letnikov 1868):
DαGLy(x) = lim
h→0
1
hα
∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh) (α > 0)
y defined only on [0, X]; thus assume y(x) = 0 for x < 0.⇒ Dα
GLy(x) = Dαy(x) if y is smooth.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 5/21
![Page 10: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/10.jpg)
Grunwald-Letnikov Methods
Integer-order derivatives:
Dn y(x) = limh→0
1
hn
∞∑
k=0
(−1)k Γ(n + 1)
Γ(k + 1)Γ(n − k + 1)y(x−kh) (n ∈ N)
Fractional extension (Liouville 1832, Grunwald 1867, Letnikov 1868):
DαGLy(x) = lim
h→0
1
hα
∞∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh) (α > 0)
y defined only on [0, X]; thus assume y(x) = 0 for x < 0.⇒ Dα
GLy(x) = Dαy(x) if y is smooth.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 5/21
![Page 11: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/11.jpg)
Grunwald-Letnikov Methods
Integer-order derivatives:
Dn y(x) = limh→0
1
hn
∞∑
k=0
(−1)k Γ(n + 1)
Γ(k + 1)Γ(n − k + 1)y(x−kh) (n ∈ N)
Fractional extension (Liouville 1832, Grunwald 1867, Letnikov 1868):
DαGLy(x) = lim
h→0
1
hα
bx/hc∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh) (α > 0)
y defined only on [0, X]; thus assume y(x) = 0 for x < 0.
⇒ DαGLy(x) = Dαy(x) if y is smooth.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 5/21
![Page 12: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/12.jpg)
Grunwald-Letnikov Methods
Integer-order derivatives:
Dn y(x) = limh→0
1
hn
∞∑
k=0
(−1)k Γ(n + 1)
Γ(k + 1)Γ(n − k + 1)y(x−kh) (n ∈ N)
Fractional extension (Liouville 1832, Grunwald 1867, Letnikov 1868):
DαGLy(x) = lim
h→0
1
hα
bx/hc∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh) (α > 0)
y defined only on [0, X]; thus assume y(x) = 0 for x < 0.⇒ Dα
GLy(x) = Dαy(x) if y is smooth.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 5/21
![Page 13: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/13.jpg)
Grunwald-Letnikov Methods
Exact analytical expression:
DαGLy(x) = lim
h→0
1
hα
bx/hc∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh)
for some h > 0.
Error = O(h) + O(f(0)) if f is smooth, i.e.
slow convergence if f(0) = 0,
completely unfeasible if f(0) 6= 0.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 6/21
![Page 14: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/14.jpg)
Grunwald-Letnikov Methods
Numerical approximation:
hDαGLy(x) =
1
hα
bx/hc∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh)
for some h > 0.
Error = O(h) + O(f(0)) if f is smooth, i.e.
slow convergence if f(0) = 0,
completely unfeasible if f(0) 6= 0.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 6/21
![Page 15: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/15.jpg)
Grunwald-Letnikov Methods
Numerical approximation:
hDαGLy(x) =
1
hα
bx/hc∑
k=0
(−1)k Γ(α + 1)
Γ(k + 1)Γ(α − k + 1)y(x − kh)
for some h > 0.
Error = O(h) + O(f(0)) if f is smooth, i.e.
slow convergence if f(0) = 0,
completely unfeasible if f(0) 6= 0.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 6/21
![Page 16: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/16.jpg)
Fractional Linear Multistep Methods
Generalization of classical linear multistep methods forfirst-order ODEs (Lubich 1983–1986)
General form:
ym =
dαe−1∑
j=0
y(j)0
xjm
j!+ hα
m∑
j=0
ωm−jf(xj, yj) + hα
s∑
j=0
wm,jf(xj, yj)
Computation of convolution weights directly from coefficientsof underlying classical linear multistep method
K. Diethelm, Numerical Methods in Fractional Calculus – p. 7/21
![Page 17: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/17.jpg)
Fractional Linear Multistep Methods
Generalization of classical linear multistep methods forfirst-order ODEs (Lubich 1983–1986)
General form:
ym =
dαe−1∑
j=0
y(j)0
xjm
j!+ hα
m∑
j=0
ωm−jf(xj, yj) + hα
s∑
j=0
wm,jf(xj, yj)
initial condition convolution weights starting weights
Computation of convolution weights directly from coefficientsof underlying classical linear multistep method
K. Diethelm, Numerical Methods in Fractional Calculus – p. 7/21
![Page 18: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/18.jpg)
Fractional Linear Multistep Methods
Generalization of classical linear multistep methods forfirst-order ODEs (Lubich 1983–1986)
General form:
ym =
dαe−1∑
j=0
y(j)0
xjm
j!+ hα
m∑
j=0
ωm−jf(xj, yj) + hα
s∑
j=0
wm,jf(xj, yj)
initial condition convolution weights starting weights
Computation of convolution weights directly from coefficientsof underlying classical linear multistep method
K. Diethelm, Numerical Methods in Fractional Calculus – p. 7/21
![Page 19: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/19.jpg)
Fractional Linear Multistep Methods
Computation of starting weights: Linear system of equations
Properties of linear system depend strongly on α:
Assume underlying classical method to be of order p, and let
A = {γ = j + kα : j, k = 0, 1, . . . , γ ≤ p − 1}.
Then, error of resulting fractional method = O(hp),
dimension of linear system = cardinality of A,
condition of linear system depends on distribution
of elements of A.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 8/21
![Page 20: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/20.jpg)
Fractional Linear Multistep Methods
Computation of starting weights: Linear system of equations
Properties of linear system depend strongly on α:
Assume underlying classical method to be of order p, and let
A = {γ = j + kα : j, k = 0, 1, . . . , γ ≤ p − 1}.
Then, error of resulting fractional method = O(hp),
dimension of linear system = cardinality of A,
condition of linear system depends on distribution
of elements of A.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 8/21
![Page 21: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/21.jpg)
Fractional Linear Multistep Methods
Computation of starting weights: Linear system of equations
Properties of linear system depend strongly on α:
Assume underlying classical method to be of order p, and let
A = {γ = j + kα : j, k = 0, 1, . . . , γ ≤ p − 1}.
Then, error of resulting fractional method = O(hp),
dimension of linear system = cardinality of A,
condition of linear system depends on distribution
of elements of A.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 8/21
![Page 22: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/22.jpg)
Fractional Linear Multistep Methods
Further comments on linear system for starting weights:
needs to be solved at every grid point,
coefficient matrix always identical,
coefficient matrix has generalized Vandermondestructure,
right-hand sides change from one grid point to another,
accuracy of numerical computation of right-hand sidesuffers from cancellation of digits.
K. Diethelm, Numerical Methods in Fractional Calculus – p. 9/21
![Page 23: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/23.jpg)
Fractional Linear Multistep Methods
A well-behaved example:p = 4, α = 1/3 ⇒ A =
{
0, 13, 2
3, 1, 4
3, 5
3, . . . , 3
}
.
card(A) = 10,
highly regular spacing of elements of A (equispaced)⇒ Matrix of coefficients can be transformed to usual
Vandermonde form
mildly ill conditioned system
special algorithms can be used for sufficiently accuratesolution (Bjorck & Pereyra 1970)
K. Diethelm, Numerical Methods in Fractional Calculus – p. 10/21
![Page 24: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/24.jpg)
Fractional Linear Multistep Methods
A badly behaved example:p = 4, α = 0.33
⇒ A = {0, 0.33, 0.66, 0.99, 1, 1.32, 1.33, . . . , 2.97, 2.98, 2.99, 3}
card(A) = 22,
spacing of the elements of A is highly irregular(in particular, some elements are very close together)
severely ill conditioned system
no algorithms for sufficiently accurate solution known(Diethelm, Ford, Ford & Weilbeer 2006)
K. Diethelm, Numerical Methods in Fractional Calculus – p. 11/21
![Page 25: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/25.jpg)
Quadrature-Based Methods
Direct approach:
Use representation Dαy(x) =1
Γ(−α)
∫ x
0
(x − t)−α−1y(t)dt
(finite-part integral) in differential equation
Discretize integral by classical quadrature techniques
Example: Replace y in integrand by piecewise linearinterpolant with step size h (Diethelm 1997)
⇒ product trapezoidal method; Error = O(h2−α)
Similar concept applicable to Caputo derivatives
K. Diethelm, Numerical Methods in Fractional Calculus – p. 12/21
![Page 26: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/26.jpg)
Quadrature-Based Methods
Indirect approach: (Diethelm, Ford & Freed 1999–2004)
Rewrite Caputo initial value problem as integral equation
y(x) =
dαe−1∑
k=0
y(k)0
xk
k!+ Jα[f(·, y(·))](x)
Apply product integration idea to Jα
Example 1: Product rectangle quadrature⇒ fractional Adams-Bashforth method
explicit method
Error = O(h)
K. Diethelm, Numerical Methods in Fractional Calculus – p. 13/21
![Page 27: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/27.jpg)
Quadrature-Based Methods
Indirect approach (continued):
Example 2: Product trapezoidal quadrature⇒ fractional Adams-Moulton method
implicit method
Error = O(h2)
P(EC)mE scheme: Combine Adams-Bashforth predictorand m Adams-Moulton correctors⇒ Error = O(hp), p = min{2, 1 + mα}
Similar concept applicable to Riemann-Liouville derivatives
K. Diethelm, Numerical Methods in Fractional Calculus – p. 14/21
![Page 28: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/28.jpg)
Quadrature-Based Methods
Common properties:
Coefficients can be computed in accurate and stable waywithout problems
Method is applicable for any type of differential equation
Reasonable rate of convergence can be achieved
K. Diethelm, Numerical Methods in Fractional Calculus – p. 15/21
![Page 29: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/29.jpg)
Example 1
Dα∗ y(x) = x4 + 1 +
24
Γ(5 − α)x4−α − y(x), y(0) = 1
exact solution on [0, 2]: y(x) = x4 + 1
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=2| vs. no. of nodes
10 100 1000
1e-10
1e-08
1e-06
0,0001
0,01
α = 0.50
K. Diethelm, Numerical Methods in Fractional Calculus – p. 16/21
![Page 30: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/30.jpg)
Example 1
Dα∗ y(x) = x4 + 1 +
24
Γ(5 − α)x4−α − y(x), y(0) = 1
exact solution on [0, 2]: y(x) = x4 + 1
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=2| vs. no. of nodes
10 100 1000
1e-10
1e-08
1e-06
0,0001
0,01
α = 0.33
K. Diethelm, Numerical Methods in Fractional Calculus – p. 16/21
![Page 31: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/31.jpg)
Example 1
Dα∗ y(x) = x4 + 1 +
24
Γ(5 − α)x4−α − y(x), y(0) = 1
exact solution on [0, 2]: y(x) = x4 + 1
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=2| vs. no. of nodes
10 100 1000
1e-10
1e-08
1e-06
0,0001
0,01
α = 0.10
K. Diethelm, Numerical Methods in Fractional Calculus – p. 16/21
![Page 32: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/32.jpg)
Example 2
Dα∗ y(x) =
40320
Γ(9 − α)x8−α − 3
Γ(5 + α/2)
Γ(5 − α/2)x4−α/2 +
9
4Γ(α + 1)
+(
1.5xα/2 − x4)3
− y(x)3/2, y(0) = 0
exact solution on [0, 1]: y(x) = x8 − 3x4+α/2 + 2.25xα
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=1| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.50
K. Diethelm, Numerical Methods in Fractional Calculus – p. 17/21
![Page 33: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/33.jpg)
Example 2
Dα∗ y(x) =
40320
Γ(9 − α)x8−α − 3
Γ(5 + α/2)
Γ(5 − α/2)x4−α/2 +
9
4Γ(α + 1)
+(
1.5xα/2 − x4)3
− y(x)3/2, y(0) = 0
exact solution on [0, 1]: y(x) = x8 − 3x4+α/2 + 2.25xα
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=1| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.33
K. Diethelm, Numerical Methods in Fractional Calculus – p. 17/21
![Page 34: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/34.jpg)
Example 2
Dα∗ y(x) =
40320
Γ(9 − α)x8−α − 3
Γ(5 + α/2)
Γ(5 − α/2)x4−α/2 +
9
4Γ(α + 1)
+(
1.5xα/2 − x4)3
− y(x)3/2, y(0) = 0
exact solution on [0, 1]: y(x) = x8 − 3x4+α/2 + 2.25xα
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=1| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.10
K. Diethelm, Numerical Methods in Fractional Calculus – p. 17/21
![Page 35: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/35.jpg)
Example 3
Dα∗ y(x) = −x1−αE1,2−α(−x) + exp(−2x) − y(x)2, y(0) = 1
exact solution on [0, 5]: y(x) = exp(−x)
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=5| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.50
K. Diethelm, Numerical Methods in Fractional Calculus – p. 18/21
![Page 36: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/36.jpg)
Example 3
Dα∗ y(x) = −x1−αE1,2−α(−x) + exp(−2x) − y(x)2, y(0) = 1
exact solution on [0, 5]: y(x) = exp(−x)
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=5| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.33
K. Diethelm, Numerical Methods in Fractional Calculus – p. 18/21
![Page 37: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/37.jpg)
Example 3
Dα∗ y(x) = −x1−αE1,2−α(−x) + exp(−2x) − y(x)2, y(0) = 1
exact solution on [0, 5]: y(x) = exp(−x)
Grunwald-Letnikovdirect quadratureindirect quadraturemultistep (BDF4)
|Error at x=5| vs. no. of nodes
10 100 10001e-10
1e-08
1e-06
0,0001
0,01
α = 0.10
K. Diethelm, Numerical Methods in Fractional Calculus – p. 18/21
![Page 38: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/38.jpg)
Summary
Grunwald-Letnikov: reliable but slowly convergent
Linear multistep methods: very bad for some values of α,very good for others
Quadrature-based methods: useful compromise,in particular because of simple speed-up possibility
All types of basic routines available inGNS Numerical Fractional Calculus Library(presently FORTRAN77)
K. Diethelm, Numerical Methods in Fractional Calculus – p. 19/21
![Page 39: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/39.jpg)
Extensions
Application of methods to
time-/space-fractional partial differential equations,
fractional optimal control problems.
Common problem: High arithmetic complexity due tonon-locality of fractional operators.Solution for quadrature methods: Nested mesh concept(Ford & Simpson 2001).
K. Diethelm, Numerical Methods in Fractional Calculus – p. 20/21
![Page 40: A Survey of Numerical Methods in Fractional Calculus...in Fractional Calculus Fractional Derivatives in Mechanics: State of the Art CNAM Paris, 17 November 2006 Kai Diethelm diethelm@gns-mbh.com,](https://reader030.vdocuments.us/reader030/viewer/2022040115/5e6da84ad29167699605ded9/html5/thumbnails/40.jpg)
Merci pour votre attention!
contact: [email protected]
K. Diethelm, Numerical Methods in Fractional Calculus – p. 21/21