Vector gradient #78
Vector gradient #78
Conversation
a function to align for dipole moment derivatives
a function to align for dipole moment derivatives
…o vectorGradient Conflicts: qcelemental/models/align.py qcelemental/molutil/test_align.py
qcelemental/models/align.py
Outdated
| al_mu_x[self.atommap[at]][:] = Datom[0,:] | ||
| al_mu_y[self.atommap[at]][:] = Datom[1,:] | ||
| al_mu_z[self.atommap[at]][:] = Datom[2,:] | ||
| return (al_mu_x, al_mu_y, al_mu_z) |
There was a problem hiding this comment.
since inputs and outputs are xyz components, would it make sense to either do three separate calls or pass as (3 * nat, 3)? like to take better advantage of broadcasting.
There was a problem hiding this comment.
What I like about the 3 explicit arguments is that it is obvious to any user what should be passed, particularly for a mixed-derivative like this - versus the challenge of a new person (or myself in 6 months:) trying to figure out what the single variable data structure should be. For the Hessian you would guess correctly, maybe not for this. But I'm fine changing it if you think there's a compelling reason and relying on examples, tests, and documentation.
There was a problem hiding this comment.
sorry for the delay -- it's been a crazy month. talking it over, we think the packed arrangement would be more in keeping with the other functions. the first line of the fn can be mu_x, mu_y, mu_z = *mu and the return can be return al_mu and the test call can be p2c_dipder_x, p2c_dipder_y, p2c_dipder_z = mill.align_vector_gradient(...)
qcelemental/molutil/__init__.py
Outdated
| @@ -1 +1,3 @@ | |||
| from .align import B787, kabsch_align, compute_scramble | |||
| from .test_align import test_hessian_align, test_vector_gradient_align | |||
There was a problem hiding this comment.
this shouldn't be needed as we don't want the tests to be part of the module proper. if you let me know the error message that led to this addition, I can probably suggest an alternate solution.
There was a problem hiding this comment.
I just wanted to be able to run the tests from an external directory. If they are not here, how do you do that? I was just testing with:
sys.path.insert(0, '/home/rking/programs/QCElemental')
import qcelemental as qcel
qcel.molutil.test_vector_gradient_align()
There was a problem hiding this comment.
Try just “pytest” once “import qcelemental as qcel” works.
I added function to align the derivatives of vector components. For example, the code
will align dipole moment derivatives. It can convert CFOUR's DIPDER with PSI4's file17 values.
A remaining danger is that CFOUR reorders atoms when computing the gradients. Unless one attends carefully to the atomic numbers in DIPDER, or processes carefully the symmetry sorting presented in the output file, one might not even notice. I've attached a sample output and dipder file. The geometry does not appear to be printed anywhere in the order that was used in DIPDER, which means reliable atom mapping will be a challenge.
c4-vib.txt
c4-dipder.txt
The dipole moment derivatives are presented in the output file too. Only to 6 decimals.
Does the same problem emerge in a Hessian computation? Perhaps you have solved it.