Abstract:In the paper titled "New numerical approach for fractional differential equations" by A. Atangana and K.M. Owolabi [Math. Model. Nat. Phenom., 13(1), 2018], it is presented a method for the numerical solution of some fractional differential equations. The numerical approximation is obtained by using just local information and the scheme does not present a memory term; moreover, it is claimed that third-order convergence is surprisingly obtained by simply using linear polynomial approximations. In this note we show that methods of this kind are not reliable and lead to completely wrong results since the nonlocal nature of fractional differential operators cannot be neglected. We illustrate the main weaknesses in the derivation and analysis of the method in order to warn other researchers and scientist to overlook this and other methods devised on similar basis and avoid their use for the numerical simulation of fractional differential equations.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: when some numerical methods are used to solve Fractional Differential Equations (FDEs), due to the neglect of the nonlocality of fractional - order derivatives, the results are unreliable and completely wrong.
### Specific problem description
1. **Neglect of non - locality**
- Fractional - order derivatives are essentially non - local operators, which means they depend on the entire historical information, not just the information at the current point. The paper points out that in the method proposed by Atangana and Owolabi, the numerical approximation only uses local information and does not include the memory term, which violates the basic properties of fractional - order differential operators.
2. **Unreasonableness of convergence claims**
- This method claims that third - order convergence can be achieved through linear interpolation polynomials. However, according to the basic knowledge of numerical analysis, linear interpolation usually can only provide second - order convergence, unless there are special conditions or improved techniques. Therefore, this claim of third - order convergence is obviously unreasonable.
3. **Logical errors in method derivation**
- The paper analyzes in detail the derivation process of the Atangana and Owolabi method and points out multiple logical errors and inconsistencies. For example:
- In the derivation process, the grid points are confused with the degree of the interpolation polynomial.
- Unrealistic smoothness conditions are assumed (such as the boundedness of the third - order derivative of the function), while in fact the solutions of FDEs usually do not have such smoothness.
- In the error analysis, the influence of cumulative errors is ignored.
4. **Numerical experiment verification**
- To further prove the unreliability of this method, the author conducted simple numerical experiments. The results show that as the integration process progresses, the error increases rapidly, the numerical solution deviates from the exact solution, and finally leads to divergence.
### Conclusion
The main purpose of this paper is to warn other researchers and scientists not to use such methods based on local information to solve fractional - order differential equations. These methods are not only unreliable but also lead to completely wrong results. To correctly handle fractional - order differential equations, it is necessary to fully consider their non - local characteristics and adopt appropriate techniques to approximate the memory term.
### Related formulas
- **Caputo - type fractional - order derivative**
\[
{}^{C}D^{\alpha}_{0} y(t)=f(t, y(t)), \quad 0 < \alpha < 1
\]
where \({}^{C}D^{\alpha}_{0}\) represents the Caputo - type fractional - order derivative.
- **Volterra integral equation form**
\[
y(t)=y(0)+\frac{1}{\Gamma(\alpha)} \int_{0}^{t}(t - \lambda)^{\alpha - 1} f(\lambda, y(\lambda)) d\lambda
\]
- **Key formula in method derivation**
\[
y_{n + 1}=y(t_n)+\frac{f(t_n, y_n)}{h \Gamma(\alpha)} \left\{ \frac{2h}{\alpha (\alpha + 1)} t_n^{\alpha + 1}-t_{n - 1} t_n^{\alpha} \frac{1}{\alpha + 1}-h t_n^{\alpha}+\frac{t_n^{\alpha + 1}}{\alpha + 1} \right\}
\]
\[
+\frac{f(t_{n - 1}, y_{n - 1})}{h \Gamma(\alpha)} \left\{ \frac{t_n^{\alpha + 1}}{\alpha + 1}-h t_n^{\alpha}-\frac{t_{n - 1}^{\alpha + 1}}{\alpha + 1} \right\}
\]