How to cancell imaginary frequency when calculating dynamic matrix?

[wot19920302]

Dear  Quantumwise staffs
            I use empirical force  to optimize gamma unitcell of graphyne nanosheet and calculate phonon band in ATK. However, the calculated band has a lot of imaginary frequency.  The  adopted empirical force is  Brenner, which has been used in previsou work (see Figure 1 in attachment, doi: 10.1103/PhysRevB.85.235436 ). I have tried a lot (improve  convergence criterion of force and strss on cell to 1e-05*eV/Ang and 1e-07*eV/Ang**3, respectively, or increase repeatition in dynamic matrix to (9,9,1), even  decrease  max_interaction_range to 5.0*Angstrom) to cancell imaginary frequency but can't make it. I hope anyone can kindly help me.
           The script is attached to bottom (gamma.py).
           Best wishes

[Petr Khomyakov]

You may need using much larger supercells (e.g., try 17x17x1 or larger) for getting rid of the negative frequencies. Adopting a different force field potential might also be an option. Are you sure that you are using exactly the same potential as mentioned in the paper? For each potential, there is a reference to the paper from which the potential parameters were taken.

[wot19920302]

Thanks for your help. I will do as you suggested.     In addition, I am sure that I  use exactly the same potential as mentioned in the paper because the reference the paper provides (Fig.1)  completely coincides with the that in VNL (Fig.2)