Hi,

In general, vasp has a more complete basis set library, in addition, vasp also supports post-Hartree-Fock method, supports molecular modeling on GPUs.

I am not sure about "better" means what. In my opinion, both can done well on defects.

Good luck!

Thank you very much

There is no difference as such. One can use any package. However, every packages has its own advantages in calculating different parameter.

VASP has complete library and exact and well tested pseudo-potentials.

I have not used castep and I am not aware about this.

Good luck.

Dear Kareem,

I am a CASTEP developer, but I'll try not to be (too) biased in my answer. The underlying theory is almost identical, both use a plane-wave basis with pseudopotentials and both support a wide range of XC functionals etc. CASTEP uses norm-conserving and Vanderbilt ultrasoft pseudopotentials, and VASP uses norm-conserving and Bloechl PAW pseudopotentials -- VASP also has some ultrasoft pseudopotential support, but historically at least these were not generated with Vanderbilt's scheme and were rather inaccurate. CASTEP used to have much better pseudopotentials than VASP, but the VASP people have worked quite hard to catch up in recent years and I don't think there's much to choose between them now as long as you use the latest ones.

VASP has more post-DFT options than CASTEP, and CASTEP has phonons via perturbation theory and better parallel scaling. We are working to add more post-DFT options to CASTEP at the moment. We try quite hard to make CASTEP user-friendly, but I'm not the right person to judge whether we've succeeded!

I'm confused by your statement about VASP being better for defects, CASTEP is fine for these (as is Quantum Espresso and pretty much any similar package).

There are ports of CASTEP to GPUs (and Xeon Phis), if that's important to you, though neither VASP nor CASTEP get the enormous speed-ups that classical MD codes get. The fastest plane-wave ab initio code on a GPU is almost certainly Filippo Spiga's port of Quantum Espresso.

I've tried to avoid going into too much detail, but I'm happy to provide more if you like.

@QiangZhao: What do you mean by"In general, vasp has a more complete basis set library"?

All the best,

Phil Hasnip

(CASTEP developer)

Thank you Philip for the detailed answer. Very informative.

I know CASTEP can produce partial density of states for atomic orbitals (s, p, d orbitals), but can it produce Projected density of states for molecular orbitals (pi, sigma orbitals)?

This isn't something I've been involved with personally, but I think the CASTEP post-processing tool MolPDOS can provide that information. This was developed by Karsten Reuter (Munich, Germany) and Reinhard Maurer (Warwick, UK). It also allows a kind of DFT+U with the U applied to a molecular orbital.

Alright. Thank you!

Kareem, if you are interested in pDOS and molecular orbitals (and things like COOP) you may have a look at periodic DFT codes using atomic orbitals rather than plane waves. E.g. CRYSTAL, FHI-AIMS, BAND.

By construction the LCAO code will give much easier access to AO/MO-like properties without the need to Localize/Wannierize etc.

(CASTEP) is a business programming bundle which utilizes thickness practical hypothesis with a plane wave premise set to compute the electronic properties of crystalline solids, surfaces, atoms, fluids and undefined materials from first principles. But, (VASP,) is a bundle for performing abdominal ab initio quantum mechanical calculations utilizing in projector expanded wave technique and a plane wave premise set etc.

Muzammal Hameed Tariq If I've understood you correctly, you're saying CASTEP is a commercial program; this is not entirely true. CASTEP is developed solely by academic research groups and free for academic use (research and teaching) worldwide, but it is true that it is sold (via BIOVIA) to industrial users for commercial use.

See here for details on how to get CASTEP: http://www.castep.org/CASTEP/GettingCASTEP

In fact, VASP was initially developed from an early CASTEP version that Jürgen Hafner borrowed from Mike Payne. But then both programs grew apart, VASP got ultrasoft pseudopotentials and then PAW potentials. The power of VASP (according to Karl Schwartz, developer of Wien2K) is that using only two PAW projectors can get enough precison to be fast and stable compared to Wien2K. But since we do not have access to the VASP pseudopotential program we cannot verify that.

José Luis Ricardo Chávez I'm sure that's right, that's the experience we have in CASTEP as well. 2 ultrasoft or PAW projectors per channel seems to be an excellent balance between computational cost and accuracy.

These two program usually use for different purposes

Md. Masud Parvez What do you mean?

Dear Philip James Hasnip

(CASTEP) is a business programming bundle that utilizes thickness practical hypothesis with a plane wave premise set to compute the electronic properties of crystalline solids, surfaces, atoms, fluids, and undefined materials from first principles.

But, (VASP,) is a bundle for performing abdominal ab initio quantum mechanical calculations utilizing in projector expanded wave technique and a plane wave premise set, etc. Which already mentioned by Muzammal Hameed Tariq

Md. Masud Parvez This description looks like it was automatically translated from another language, and the translation is a little confused. Here's my attempt at correcting it, with notes:

Original: "(CASTEP) is a business programming bundle"

Translation: CASTEP is a commercial program

Notes: Whilst CASTEP is sold commercially to industry, it is free for non-commercial use worldwide. I don't think VASP is free to anyone outside Austria.

Original: "that utilizes thickness practical hypothesis with a plane wave premise set"

Translation: that uses density functional theory with a plane-wave basis set

Notes: This is correct, both VASP and CASTEP use DFT with a plane-wave basis set

Original: "(VASP,) is a bundle for performing abdominal ab initio quantum mechanical calculations"

Translation: VASP is a program for performing ab initio quantum mechanical calculations

Notes: This is correct, both VASP and CASTEP perform ab initio quantum mechanical calculations. I have no idea what "abdominal" is supposed to be.

Original: "utilizing in projector expanded wave technique and a plane wave premise set"

Translation: using projector-augmented wave (PAW) pseudopotentials and a plane-wave basis set

Notes: VASP typically uses PAW pseudopotentials, CASTEP typically uses ultrasoft pseudopotentials. The two formalisms are almost identical, and I'm not aware of any practical advantage of one or the other.

Thank You very much Philip James Hasnip for your clarification.