I was wondering what mixer scheme exciting uses. I have noticed that the code seems to require many iterations (at least for the structure I'm working on). With the wien2k code I reach convergance after about 25 iterations, whereas with exciting I am now on cycle 48 and still have only have 0.61E-03 for the RMS change in effective potential.
My question then is if exciting uses the same mixing scheme as Wien2k, is there someway I can improve on speed (I have changed beta0 to 0.05 due to large oscillations which meade convergance take even longer).
Thanks
Michael
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
EXCITING uses a simple adaptive step size mixing scheme. You can adjust two parameters, namely beta0 and betamax. beta0 is the default mixing strength as well as the increment. betamax is the maximum mixing strength, and is set to 1 by default. See the Users' Guide for more details. The code used to have the option of Broyden mixing, but I removed it as it never seemed to work particularly well.
One of the reasons Wien2k converges so quickly is that it starts from a density which is very close to the converged solution. I have no idea how it does this because in principle both EXCITING and Wien2k start from a superposition of atomic charge densities, including the tails extending into the muffin-tins. If anyone has any ideas, then let me know.
The other reason is that the EXCITING convergence criterion is the RMS change in potential, whereas Wien2k (possibly) uses the change in total energy or charge density, which may converge faster.
Cheers
Kay.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I was wondering what mixer scheme exciting uses. I have noticed that the code seems to require many iterations (at least for the structure I'm working on). With the wien2k code I reach convergance after about 25 iterations, whereas with exciting I am now on cycle 48 and still have only have 0.61E-03 for the RMS change in effective potential.
My question then is if exciting uses the same mixing scheme as Wien2k, is there someway I can improve on speed (I have changed beta0 to 0.05 due to large oscillations which meade convergance take even longer).
Thanks
Michael
EXCITING uses a simple adaptive step size mixing scheme. You can adjust two parameters, namely beta0 and betamax. beta0 is the default mixing strength as well as the increment. betamax is the maximum mixing strength, and is set to 1 by default. See the Users' Guide for more details. The code used to have the option of Broyden mixing, but I removed it as it never seemed to work particularly well.
One of the reasons Wien2k converges so quickly is that it starts from a density which is very close to the converged solution. I have no idea how it does this because in principle both EXCITING and Wien2k start from a superposition of atomic charge densities, including the tails extending into the muffin-tins. If anyone has any ideas, then let me know.
The other reason is that the EXCITING convergence criterion is the RMS change in potential, whereas Wien2k (possibly) uses the change in total energy or charge density, which may converge faster.
Cheers
Kay.