search

OrderN / CONQUEST-release

Full public release of large scale and linear scaling DFT code CONQUEST
http://www.order-n.org/
MIT License
96 stars 25 forks source link

Bug fix: correct the cell optimisation with constrained ratios #51

Closed davidbowler closed 4 years ago

davidbowler commented 4 years ago

The original version combined c/a and a/c into a single case, which gave the wrong behaviour. Now separated into individual cases.

davidbowler commented 4 years ago

@jackbaker1001 I found that this wasn't working as expected yesterday, and have made a correction - please check.

tsuyoshi38 commented 4 years ago

As I have not thought of these options (fixed c/a etc.), I don't understand why we need to distinguish fixed c/a and a/c cases. Although I have not tried the derivation, I can easily imagine that we need some condition between the change of c and change of a. But, I think we also need some transformation for the direction (a or c, b), from the stress tensor, but do we need to distinguish fixed (c/a) and (a/c) cases ?

davidbowler commented 4 years ago

Following the question from @tsuyoshi38 above, I have combined the a/c and c/a (and other) cases. @tsuyoshi38 @jackbaker1001 can you review please?

jackbaker1001 commented 4 years ago

I've taken a look and this now looks correct. Fixing c/a and a/c are equivalent, but, I had the search directions wrong!