1: \begin{abstract}

2: We present a new mimetic finite difference method able to deal with curved faces.

3: Notably, it yields a symmetric discrete problem which uses only one degree of freedom for each curved face.

4: The principle at the core of our construction is to abandon the standard definition of local consistency of mimetic finite difference methods.

5: Instead, we exploit the novel and global concept of $P_0$-consistency.

6: Numerical examples confirm the consistency and the optimal convergence rate of the method for real-case problems having also curved boundary geometries.

7: \end{abstract}

8: