search

ajkerr0 / kappa

A python package to calculate thermal conductivity across molecular interfaces.
MIT License
6 stars 4 forks source link

Elements of Hessian matrix are unexpected in thermal conductivity calculation #66

Closed ajkerr0 closed 8 years ago

ajkerr0 commented 8 years ago

In the P_{i -> j} calculations, ALL of the corresponding kmatrix elements are zero when the kmatrix is constructed out of the analytical gradients. This is why the ballandspring hessian matrix is there for now.

We know physically this cannot be. Is this an indication that the hessian matrices aren't being calculated correctly?

ajkerr0 commented 8 years ago

We have observed the values after the inspect function was added (see above). We stopped getting zeros a while ago. Maybe it was an indexing error I fixed? (index vs. 3*index, for example)

ajkerr0 commented 8 years ago

We are getting mixed positive and negative terms in the Hessian were I wasn't expecting them. Test case for example: default cnt with default polyeth at indices 72, 78; only bond length interactions turned on.

Need to study to make sure this is allowed in general. If not, we have a problem!

ajkerr0 commented 8 years ago

These values may be allowed. Positive definiteness doesn't imply the only positives are on the diagonal.

ajkerr0 commented 8 years ago

I have in fact, confirmed that mixed positive and negative values are allowed in the off diagonal, in general.

ajkerr0 commented 8 years ago

See the above commit. These values I thought were wrong are actually allowed. It has to do with the equilibrium distance not discriminating between the x,y, and z positions.