[Pw_forum] ab initio vs. first principles

Agostino Migliore amigliore at cmm.upenn.edu
Mon Dec 22 09:39:36 CET 2008


By itself, "ab initio", as directly derived and used from Latin, excludes
models and parametrization. Other thing is that, in general, we cannot
exactly describe a physical system, thus making use of approximations, so
of models. In the general use of computational physics and chemistry, the
expression includes models where empirical or semiempirical parameters are
not employed, so that a system is represented in a simplified way (->
model) while still resting on physics principles. One could argue on the
meaning of those approximations and the resulting "ab initio" nature of
the approach, thus also falling on linguistics arguments. I think it is
better to leave the expression with the meaning it acquired from the
general use (which is now its proper meaning in the proper computational
context). By rigorously speaking, my sentence reported below is right with
reference to the original meaning of "ab initio", and also agrees with its
acceptation in the computational world, once the "and" in "model and
fitting parameters" is interpreted as a logical AND.

Best regards,
Agostino Migliore

> A small doubt that I had while following this discussion, please someone
> correct-me if I'm wrong:
> Is the following definition right?

>> "[in ab-initio calculations we] do not make any assumption, such as
>> concerning models and/or fitting parameters."

> I'm saying this because we use, for example, the homogeneous electron gas
> model for LDA calculations, nonetheless it is still "ab-initio".
> I think that it's not the use of parameters or models that define a first
> principles method, but from where those parameters were taken.
> If your parameters came from experimental values, it's somewhat empirical
> (or semi-empirical at least), if you used just theoretical values (even
> if using a simplified model) it can be called ab-initio.

More information about the users mailing list