Dear Andy, 1. Update to siesta 4.1-b3.
2. Convergence of transiesta is not guaranteed. You should continue to "play" with the parameters to achieve convergence. 3. One way of improving convergence may be to create smaller differences of bias'. I.e. start with V=0, copy TSDE to V=0.25, then copy TSDE to V=0.5 and so on. The closer the previous density matrix is to the current, the better convergence. It seems you copied TSDE from a V=1 and started a V=2 calculation. This is an extreme bias to "suddenly" apply to a junction. Hence the above suggestion may be what you need. Again, it isn't guaranteed to converge, so you have to keep on trying. 2017-07-18 15:32 GMT+02:00 孙玉 <[email protected]>: > > > *Hello siesta users:* > * I am using siesta4.1-b2 to calculate transport **properties of the > SiP* > > * structure.* > > * When the bias Voltage is 1.0V, SCF can converge .But when I set bias > voltage* > * to 2.0V , SCF did not converge in maximum number of steps. And dHmax > shows* > *no convergent tendency.* > * To slove this problem, I have tried every method to improve the > properties* > *of SCF convergence, ie. decrease DM.MixingWeight to 0.02 ( or 0.01), set * > *DM.NumberPulay to 5, increase MeshCutoff to 300 Ry, and UseSaveDM is * > *true.* > * Could you tell me how to achieve the convergence of SCF when the > bias voltage* > *is higher.* > * Here I also attach my fdf and output file.* > > *Any suggestions are appreciated.* > > * THANKS* > > -- > 学然后知不足。 > Best Regards > from Andy_孙玉 > > > > -- Kind regards Nick
