Efficient operator splitting and spectral methods for the time-space fractional Black-Scholes equation (Q2043837)

In this article, the authors suggest improved L1 operator splitting and spectral methods for solving Black-Scholes equation with fractional derivatives in both time and space. It is well known that traditional methods for American options include iterative methods such as the Brennan-Schwartz algorithm and the predicted SOR method. The authors propose the use of the improved L1 method and design its corresponding operator splitting method for the American option with non-smooth initial data so that order accuracy over time can be improved. An approximation of the time-fractional derivative is given, so that the linear complementarity problems (LCPs) is split into two easy sub-problems: a fractional boundary value problem and an algebraic system. Then the leading term from the convolution sum is separated so that it can be canceled in the second step. The numerical experiments (Tables 2 and 3) show that improved L1 operator splitting method is able to achieve the genuine convergence order of \(2-\gamma\), with \(\gamma \) being the order of fractional derivative in time, for both European and American options, while the original L1 method is only first-order accurate for American options. Overall, the new method is highly accurate and appears suitable for Black-Scholes systems with fractional derivatives.