1: \begin{abstract}

2: It is demonstrated is this letter that linear multistep methods for 

3: integrating ordinary differential equations can be used to develop a family of 

4: fast forward scattering algorithms with higher orders of convergence. Excluding the cost of 

5: computing the discrete eigenvalues, the 

6: nonlinear Fourier transform (NFT) algorithm thus obtained has a complexity 

7: of $\bigO{KN+C_pN\log^2N}$ such that the error vanishes as $\bigO{N^{-p}}$ where 

8: $p\in\{1,2,3,4\}$ and $K$ is the number of eigenvalues. Such 

9: an algorithm can be potentially useful for the recently proposed 

10: NFT based modulation methodology for optical fiber communication. The exposition

11: considers the particular case of the backward differentiation formula ($C_p=p^3$) 

12: and the implicit Adams method ($C_p=(p-1)^3$) of which the latter proves to be

13: the most accurate family of methods for fast NFT.

14: \end{abstract}

15: