1: \begin{abstract}

2:   The two-step backward differential formula (BDF2) implicit method with unequal time-steps

3:   is investigated for the Cahn-Hilliard model by focusing on the numerical influences

4:   of time-step variations.

5:   The suggested method is proved to preserve a modified energy dissipation law

6:   at the discrete levels if the adjoint time-step ratios fulfill

7:   a new step-ratio restriction $0<r_k:=\tau_k/\tau_{k-1}\le r_{\mathrm{user}}$

8:    ($r_{\mathrm{user}}$ can be chosen by the user such that $r_{\mathrm{user}}<4.864$),

9:   such that it is mesh-robustly stable in an energy norm.

10:   We view the BDF2 formula as a convolution approximation of the first time derivative

11:   and perform the error analysis by using the recent suggested discrete orthogonal convolution kernels.

12:   By developing some novel convolution embedding inequalities with respect to the orthogonal convolution kernels,  an $L^2$ norm error estimate is established at the first time under

13:   the updated step-ratio restriction $0<r_k\le r_{\mathrm{user}}$.

14:   The time-stepping scheme is mesh-robustly convergent in the sense that the convergence constant (prefactor)

15:   in the error estimate is independent of the adjoint time-step ratios.

16:   On the basis of ample tests on random time meshes,

17:   a useful adaptive time-stepping strategy is applied to efficiently capture

18:   the multi-scale behaviors and to accelerate the long-time simulation approaching the steady state.\\

19:   \noindent{\emph{Keywords}:}\;\; Cahn-Hilliard  model;  adaptive BDF2 method; discrete energy dissipation law;

20:   orthogonal convolution kernels, discrete convolution embedding inequality; error estimate\\

21:   \noindent{\bf AMS subject classifications.}\;\; 35Q99, 65M06, 65M12, 74A50

22: \end{abstract}