We should add this more flexible geometric series (similar to even-tempered) as a Strategy. May also be worth investigating adding it as a starting guess.
Not the original article, but covered in https://doi.org/10.1063/1.1516801
\alpha_j = \alpha_0 \beta^{j-1}[1 + \gamma(j / N_{prim})^{\delta}]
For a total of four parameters to be optimised. While even-temp error (relative to CBS) can be 10 times that of a freely optimised basis, well-tempered is 2 or 3 times.
We should add this more flexible geometric series (similar to even-tempered) as a Strategy. May also be worth investigating adding it as a starting guess.
Not the original article, but covered in https://doi.org/10.1063/1.1516801
\alpha_j = \alpha_0 \beta^{j-1}[1 + \gamma(j / N_{prim})^{\delta}]
For a total of four parameters to be optimised. While even-temp error (relative to CBS) can be 10 times that of a freely optimised basis, well-tempered is 2 or 3 times.