A numerical Bernstein splines approach for nonlinear initial value problems with Hilfer fractional derivative
Niels Goedegebure, Kateryna Marynets · 2025
Read A numerical Bernstein splines approach for nonlinear initial value problems with Hilfer fractional derivative by Niels Goedegebure, Kateryna Marynets (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-22335.txt.
Introduction
Niels Goedegebure and Kateryna Marynets present a numerical method for equations with a broad fractional derivative, in a 2025 paper. The Hilfer derivative blends two classic fractional forms. Solutions to such equations often spike near the start time. That spike makes accurate computing hard.
Their method uses Bernstein spline curves to trace the solution. They state clear conditions for the method to converge, with rates. The code runs in a vectorized, parallel way for speed. Tests matched the theory and beat a standard predictor method on accuracy by a wide margin. Fractional equations model memory effects in materials and fluids, so sharper solvers widen their use.
Catalog record
| CREATOR | Niels Goedegebure, Kateryna Marynets |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 54 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-22335
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A numerical Bernstein splines approach for nonlinear initial value problems with Hilfer fractional derivative Niels Goedegebure, Kateryna Marynets Abstract The Hilfer fractional derivative interpolates the commonly used Riemann-Liouville and Caputo fractional derivative. In general, solutions to Hilfer fractional differential equations are singular for t ↓ 0 t\downarrow 0 and are difficult to approximate with high numerical accuracy. We propose a numerical Bernstein splines technique to approximate solutions to generalized nonlinear initial values problems with Hilfer fractional derivatives. Convergent approximations are obtained using an efficient vectorized solution setup with few convergence requirements for a wide range of nonlinear fractional differential equations. We demonstrate efficiency of the developed method by applying it to the fractional Van der Pol oscillator, a system with applications in control systems and electronic circuits. Keywords: fractional initial value problems, splines, numerical approximation, Hilfer fractional derivative, fractional Van der Pol oscillator. 1 Introduction Fractional calculus offers extension of classical differential and integral operators to the non-integer orders. Particularly of interest are fractional derivatives, generalizing \alpha -order differentiation for \alpha\in\mathbb{R}_{>0} . Fractional differential equations (FDEs) involving these fractional derivatives have gained considerable interest from applications in control theory [ 1 , 2 , 3 ] and physical systems [ 4 , 5 , 6 ] . In the focus of this paper will be a d -dimensional fractional initial value problem (FIVP): \displaystyle\begin{cases}{{}_{0}}\mathrm{D}_{t}^{{\alpha},{\beta}}\mathbf{y}(t)=\mathbf{f}(t,\mathbf{y}(t)),\quad 0<t\leq T,\\ {{}_{0}}\mathrm{I}_{t}^{1-{\gamma}}\mathbf{y}(0)=\mathbf{\tilde{y}}_{0},\end{cases} (1) where \mathbf{y}:(0,T]\to\mathbb{R}^{d} are the unknown vector-valued solution functions for initial values \mathbf{\tilde{y}}_{0}\in\mathbb{R}^{d} and \mathbf{f}:(0,T]\times\mathbb{R}^{d}\to\mathbb{R}^{d} is a given vector-valued function, in general nonlinear with respect to both of its arguments. Here, {{}_{0}}\mathrm{D}_{t}^{\alpha,\beta} denotes the Hilfer fractional derivative of order \alpha\in(0,1) and type \beta\in[0,1] as introduced in [ 4 ] , and {{}_{0}}\mathrm{I}_{t}^{1-\gamma} stands for the fractional Riemann-Liouville integral with \gamma=\alpha+\beta-\alpha\beta . Note that in this work, operations {{}_{0}}\mathrm{I}_{t}^{\alpha} , {{}_{0}}\mathrm{D}_{t}^{\alpha,\beta} , |\cdot| and \leq are applied component-wise. A general challenge for the modeling and interpretation of FDEs is the type of required initial conditions. Depending on the chosen fractional differential operator, FDEs can result in fractional derivative or integral initial values, as discussed in [ 7 ] . The Hilfer fractional differential operator parametrizes this operator choice by interpolating between two commonly used fractional differential operators: Riemann-Liouville for \beta=0 and Caputo for \beta=1 . These operators and their properties are extensively discussed for instance in [ 7 , 8 ] . Here, the Riemann-Liouville operator results in fractional initial values whereas the Caputo derivative operator yields integer order initial conditions. We highlight the results of Furati et al. 2012 , where the authors prove existence and uniqueness of solutions of FIVP ( 1 ) with Hilfer fractional derivatives for a scalar-valued setting (d=1) . This results can be also expended to the higher dimensional systems for general d\in\mathbb{N} . For \beta\not=1 and \mathbf{\tilde{y}}_{0}\not=\mathbf{0} however, the solutions to this system are in general singular for t\downarrow 0 . This provides challenges in implementation for the numerical approximations to solutions compared to the Caputo case of \beta=1 , for which numerous established methods exist [ 10 , 11 , 12 ] . Recently, some numerical methods have been proposed to approximate solutions to the general case for Hilfer fractional differential equations. [ 13 ] provide an approach yielding mild solutions under conditions for integro-differential equations. Admon et al. 2023 proposes a method for linear relaxation-oscillations, applying integration on both sides of ( 1 ) and taking a Caputo approach. Another linear Hilfer integro-differential is considered by Naveen et al. 2023 , taking a Lagrange polynomial approach. In this paper, we propose a new numerical method to efficiently approximate solutions to the nonlinear FIVP ( 1 ) with Hilfer fractional derivatives. The method is expanded from thesis work [ 16 ] , using a Bernstein splines approach as motivated by recent results in [ 17 , 18 ] . To overcome numerical difficulties and singularities, we apply a time-domain transformation of solutions using a weighing approach as it was done in [ 9 ] for its weighed norm, together with a time domain transformation resulting in solutions starting from \varepsilon>0 small. The new method shows highly satisfactory results with an efficient vectorized implementation and allows for convergence for a wide range of nonlinear problems. The outline of the paper is as follows. In Section 2 we provide preliminary results on fractional calculus to establish existence and uniqueness of analytical solutions for FIVP ( 1 ). In Section 3 , we present preliminary Bernstein spline approximation results. We use these sections to prove the main convergence result on Hilfer Bernstein splines for FIVP’s of Theorem 4.1 in Section 4.1 , with a detailed numerical implementation in Section 4.2 . We show convergence of numerical experiments in Section 5.1 , together with applications to the fractional Van der Pol oscillator in Section 5.2 , a system with applications in control theory and electronic circuits as recently studied in [ 19 , 20 , 21 ] . 2 Preliminary results on fractional calculus Definition 2.1 (Riemann-Liouville fractional integral, [ 11 ] ) . For \alpha\geq 0 and a\leq t\leq b , define the Riemann-Liouville fractional integral as: \displaystyle{{}_{a}}\mathrm{I}_{t}^{\alpha}f(t)={\frac{1}{\Gamma(\alpha)}}\int_{a}^{t}\left(t-s\right)^{\alpha-1}f(s)\,\mathrm{d}s. (2) Furthermore, we use the notational convention {{}_{a}}\mathrm{I}_{t}^{\alpha}f(a):=\raisebox{2.15277pt}{\scalebox{0.8}{$\displaystyle\lim_{t\downarrow a}\;$}}{{}_{a}}\mathrm{I}_{t}^{\alpha}f(t) . This operator has a number of useful properties: Proposition 2.1 (Semigroup property of fractional integration, [ 11 ] ) . For \alpha,\beta>0 , if f\in L^{p}[a,b] with 1\leq p\leq\infty , we have: \displaystyle({{}_{a}}\mathrm{I}_{t}^{\alpha}{{}_{a}}\mathrm{I}_{t}^{\beta})f(t)={{}_{a}}\mathrm{I}_{t}^{\alpha+\beta}f(t), (3) almost everywhere for t\in[a,b] . Furthermore, if \alpha+\beta>1 , the equality ( 3 ) holds pointwise. The following boundedness and continuity result will be useful for further results. Proposition 2.2 ( L^{p} boundedness in f , [ 11 ] ) . The Riemann-Liouville integral is bounded in L^{p}[a,b] for f\in L^{p}[a,b] , with 1\leq p\leq\infty , satisfying: \displaystyle||{{}_{a}}\mathrm{I}_{t}^{\alpha}f||_{L^{p}[a,b]}\leq\frac{(b-a)^{\alpha}}{\Gamma(\alpha+1)}||f||_{L^{p}[a,b]}. (4) By linearity of integration, this result also provides continuity in L^{p}[a,b] with regards to f . Now, let us introduce two special functions. Definition 2.2 (Beta function, [ 11 ] ) . For a,b>0 , denote the (complete) beta function as: \displaystyle\mathrm{B}(a,b)=\int_{0}^{1}\vartheta^{a-1}(1-\vartheta)^{b-1}\,\mathrm{d}\vartheta. (5) Definition 2.3 (Incomplete beta function, [ 22 ] ) . For a,b>0 and 0\leq z\leq 1 , denote the incomplete beta function as: \displaystyle\mathrm{B}_{z}(a,b)=\int_{0}^{z}\vartheta^{a-1}(1-\vartheta)^{b-1}\,\mathrm{d}\vartheta. (6) We now use the beta functions as introduced above to explicitly calculate the fractional integral of a locally supported monomial. Proposition 2.3 (Fractional integral of monomials) . For \alpha>0 , k>-1 and 0<b\leq t , the \alpha -fractional integral of a monomial with local support \mathds{1}_{[0,b]}s^{k}:[0,b]\to\mathbb{R} is given by: \displaystyle{{}_{0}}\mathrm{I}_{t}^{\alpha}\{\mathds{1}_{[0,b]}s^{k}\}(t)=\frac{t^{\alpha+k}}{\Gamma(\alpha)}\mathrm{B}_{b/t}(k+1,\alpha). (7) Proof. Substituting z:=\frac{s}{t} gives: \displaystyle\frac{1}{\Gamma(\alpha)}\int_{0}^{b}(t-s)^{\alpha-1}s^{k}\,\mathrm{d}s=\frac{t^{\alpha+k}}{\Gamma(\alpha)}\int_{0}^{b/t}(1-z)^{\alpha-1}z^{k}\,\mathrm{d}z=\frac{t^{\alpha+k}}{\Gamma(\alpha)}\mathrm{B}_{b/t}(k+1,\alpha). (8) ∎ Remark 2.1 . Note that by Proposition 2.3 we have: \displaystyle{{}_{0}}\mathrm{I}_{t}^{\alpha}\mathds{1}_{[0,b]}s^{k}=\begin{cases}\frac{t^{\alpha+k}}{\Gamma(\alpha)}\mathrm{B}_{1}(k+1,\alpha)\quad&\textrm{if}\quad t\leq b\\ \frac{t^{\alpha+k}}{\Gamma(\alpha)}\mathrm{B}_{b/t}(k+1,\alpha)\quad&\textrm{if}\quad t>b\end{cases}=\,\frac{t^{\alpha+k}}{\Gamma(\alpha)}\mathrm{B}_{1\land b/t}(k+1,\alpha). (9) Here, x\land y:=\min(x,y) . Hence, the fractional integral of a monomial with local support is not constant after its support, in contrast to the integer-order integration case. Now, let us establish fractional differentiation. Definition 2.4 (Hilfer derivative, [ 4 ] ) . Define the Hilfer fractional derivative of order 0<\alpha<1 and type 0\ \leq\beta\leq 1 for t\geq 0 as: \displaystyle{{}_{0}}\mathrm{D}_{t}^{\alpha,\beta}={{}_{0}}\mathrm{I}_{t}^{\beta(1-\alpha)}\frac{\,\mathrm{d}}{\,\mathrm{d}t}{{}_{0}}\mathrm{I}_{t}^{(1-\beta)(1-\alpha)}. (10) Remark 2.2 (Interpolation of Caputo and RL-derivative) . The Hilfer fractional derivative interpolates two of the most commonly used fractional derivative operators: \displaystyle{{}_{0}}\mathrm{D}_{t}^{\alpha,\beta}=\begin{cases}\frac{\,\mathrm{d}}{\,\mathrm{d}t}{{}_{0}}\mathrm{I}_{t}^{1-\alpha},\quad\text{for}\quad\beta=0,\quad\text{(Riemann-Liouville derivative)},\\ {{}_{0}}\mathrm{I}_{t}^{1-\alpha}\frac{\,\mathrm{d}}{\,\mathrm{d}t},\quad\text{for}\quad\beta=1,\quad\text{(Caputo derivative)}.\end{cases} (11) To obtain solutions to Hilfer fractional differential equations, Furati et al. 2012 introduce the following space of