Parameter: alpha

Definition

Mixing parameter $\alpha$ for dynamical mean-field theory self-consistent engine.

Type

Float, double precision

Default value

0.7

Component

ALL

Behavior

In the dynamical mean-field theory self-consistent engine, the new hybridization function and self-energy function are mixed using the following equations.

$$\begin{equation} \Delta_{\text{new}} \leftarrow \Delta_{\text{old}}(1-\alpha) + \alpha \Delta_{\text{new}} \end{equation}$$

$$\begin{equation} \Sigma_{\text{new}} \leftarrow \Sigma_{\text{old}}(1-\alpha) + \alpha \Sigma_{\text{new}} \end{equation}$$

Here, $\Delta_{\text{old}}$ and $\Sigma_{\text{old}}$ are obtained in the previous iteration, while $\Delta_{\text{new}}$ and $\Sigma_{\text{new}}$ are derived in the current iteration.

Comment

If you have trouble in achieving convergence, you can decrease it to 0.5 or even 0.2. If you are doing one-shot calculation (isscf = 1), this parameter is useless. It is useful only when isscf = 2.

See also isscf parameter for more details.