Dear Muhammad
it depends on which properties you are interested, the level of theory, and the characteristics of the system you are interested in.
The number of atoms you mention are not a problem for the CRYSTAL code.
Let me give you some examples, referring not to feeling, but to published papers .
You might reproduce the data easily with the CRYSTAL code. We could provide you the input for most of these calculations.
A big difference with CRYSTAL (not for the other codes ) is SYMMETRY, that is fully exploited at ALL levels in the code.
Let me start from a LARGE MOLECULE WITHOUT ANY SYMMETRY
Elucidating the fundamental forces in protein crystal formation: the case of crambin
By: Delle Piane, Massimo; Corno, Marta; Orlando, Roberto; et al.
CHEMICAL SCIENCE Volume: 7 Issue: 2 Pages: 1496-1507 Published: 2016
Here you find a calculation of the crambin protein MOLECULE: ALL ELECTRON, HYBRID FUNCTIONAL,
fully optimized PLUS FREQUENCY calculations . 642 atoms. High accuracy.
In the same paper, the bulk structure plus about 172 water molecules are investigated, about 2000 atoms per cell. The celli s optimized.
A LARGE MOLECULE WITH A LOT OF SYMMETRY
Structural, electronic and energetic properties of giant icosahedral fullerenes up to C6000: insights from an ab initio hybrid DFT study
By: Noel, Yves; De La Pierre, Marco; Zicovich-Wilson, Claudio M.; et al.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS Volume: 16 Issue: 26 Pages: 13390-13401 Published: 2014
120 symmetry operators, All electron, B3LYP, 6000 atoms, 86000 basis functions, RUNNING ON A SINGLE CORE, the largest fullerene ever investigated……
A LARGE NANOTUBE WITH HIGH SYMMETRY: 5000 atoms, B3LYP, all electron……
Serpentine polymorphism: a quantitative insight from first-principles calculations
Raffaella Demichelis,*a Marco De La Pierre,a Mainak Mookherjee,b Claudio M. Zicovich-Wilsonc and Roberto Orlandod
CRYSTENGCOM 2016
A large cluster
Photoexcited carriers recombination and trapping in spherical vs faceted TiO2 nanoparticles Gianluca Fazio, Lara Ferrighi, Cristiana Di Valentin
Nanoenergy 2016
All the calculations were performed with the CRYSTAL14 [23] package where the Kohn–Sham orbitals are expanded in Gaussian type orbitals (the all-electron basis sets are O 8-411(d1), Ti 86-411 (d41) and H 511(p1)). The B3LYP [24,25] and HSE06 [26] hybrid functionals have been used throughout this work. The values of the optimized lattice parameters are 3.789 Å and 3.766 Å for a and 9.777 Å and 9.663 Å for c, respectively for and B3LYP and HSE06, which are comparable to the experimental values [27]. The TiO2 anatase bulk was modeled by a 6√2 ! 6√2 ! 1 bulk supercell with 864 atoms.
………Nanocrystals (NC) have been cut from a bulk anatase supercell following the procedure already described in our previous work [28]. Two stoichiometric nanocrystals (TiO2)159 " 4H2O (NCS) and (TiO2)_260 ……(from the computational section of the paper).
MASSIVE PARALLELISM of codes: not always the evidence is… evident.
CRYSTAL scales extremely well, also with hybrids, up to say 4096 or even more cores
A new massively parallel version of CRYSTAL for large systems on high performance computing architectures
By: Orlando, Roberto; Delle Piane, Massimo; Bush, Ian J.; et al.
JOURNAL OF COMPUTATIONAL CHEMISTRY Volume: 33 Issue: 28 Pages: 2276-2284 Published: OCT 30 2012
Roberto Dovesi
Unless you have unlimited access to some supecomputer facility, something above 3 nm (painful already) is too expensive for DFT. You still can try pure GGA functionals with RI (in ORCA for example) or DFTB (in Gaussian or DFTB+) or give up DFT and try semiempirical things like AM1, PM6 and so on.
SIESTA is usually one of the fastest DFT odes, but as Oleg said, 2nm will be ~2000 atoms, thats quite challenging for DFT
First of all you must ask yourself which properties you need? and after that maybe you realize that you don't need DFT calculation at all.
I am interested in calculating bang gap, optical, raman spectra and transport properties of quantum dots.
Due to the size of the system a much simpler approach like DFTB (implemented in Gaussian or DFTB+), as suggested by leg, is likely the only reasonable solution.
Yes, I can afford castep and supercomputer but you don't think that the use of plane wave code will be more costly to compute the properties of quantum dots?
Hi Muhammad,
What elements are involved in your quantum dots? The largest CASTEP calculation I've done personally is 4.3 nm x 4.3 nm x 4.3 nm of silicon (~4000 atoms), which takes about half an hour per geometry step on a reasonable supercomputer (say 500 Ivy Bridge cores with a good interconnect) -- but silicon has a low plane-wave cut-off energy, only 4 valence electrons per atom and isn't magnetic.
Actually I have done a larger calculation by volume (12 nm x 17 nm x 10 nm), but it was a DNA strand in a large box so not very many atoms (~1200 I think) or electrons.
All the best,
Phil Hasnip
(CASTEP Developer)
I have to study Si and Ge QDs, and the 2.2 nm size (diameter) QD passivated with hydrogen consists of 281 Si and 172 H atoms.
Could you please give details of your calculations? For example, which correlation exchange functional and pseudopotentials (norm conserving, ultrasoft or on the fly) were used for geometry optimization? I have seen many papers where Becke-Lee-Yang-Parr correlation exchange functional at the generalized gradient approximation level was used for geometry optimization. But I couldn't find this correlation exchange functional in CASTEP.
Please comment.
Germanium isn't much harder than silicon so you should be fine. My 4000 atoms were with LDA and norm-conserving pseudopotentials. Ultrasofts (USPs) are fine too, though at some point over 3000 atoms I ran into some memory problems with USPs because of the extra projectors. I've used LDA and PBE, both are similar in time and memory.
Dear Muhammad
it depends on which properties you are interested, the level of theory, and the characteristics of the system you are interested in.
The number of atoms you mention are not a problem for the CRYSTAL code.
Let me give you some examples, referring not to feeling, but to published papers .
You might reproduce the data easily with the CRYSTAL code. We could provide you the input for most of these calculations.
A big difference with CRYSTAL (not for the other codes ) is SYMMETRY, that is fully exploited at ALL levels in the code.
Let me start from a LARGE MOLECULE WITHOUT ANY SYMMETRY
Elucidating the fundamental forces in protein crystal formation: the case of crambin
By: Delle Piane, Massimo; Corno, Marta; Orlando, Roberto; et al.
CHEMICAL SCIENCE Volume: 7 Issue: 2 Pages: 1496-1507 Published: 2016
Here you find a calculation of the crambin protein MOLECULE: ALL ELECTRON, HYBRID FUNCTIONAL,
fully optimized PLUS FREQUENCY calculations . 642 atoms. High accuracy.
In the same paper, the bulk structure plus about 172 water molecules are investigated, about 2000 atoms per cell. The celli s optimized.
A LARGE MOLECULE WITH A LOT OF SYMMETRY
Structural, electronic and energetic properties of giant icosahedral fullerenes up to C6000: insights from an ab initio hybrid DFT study
By: Noel, Yves; De La Pierre, Marco; Zicovich-Wilson, Claudio M.; et al.
PHYSICAL CHEMISTRY CHEMICAL PHYSICS Volume: 16 Issue: 26 Pages: 13390-13401 Published: 2014
120 symmetry operators, All electron, B3LYP, 6000 atoms, 86000 basis functions, RUNNING ON A SINGLE CORE, the largest fullerene ever investigated……
A LARGE NANOTUBE WITH HIGH SYMMETRY: 5000 atoms, B3LYP, all electron……
Serpentine polymorphism: a quantitative insight from first-principles calculations
Raffaella Demichelis,*a Marco De La Pierre,a Mainak Mookherjee,b Claudio M. Zicovich-Wilsonc and Roberto Orlandod
CRYSTENGCOM 2016
A large cluster
Photoexcited carriers recombination and trapping in spherical vs faceted TiO2 nanoparticles Gianluca Fazio, Lara Ferrighi, Cristiana Di Valentin
Nanoenergy 2016
All the calculations were performed with the CRYSTAL14 [23] package where the Kohn–Sham orbitals are expanded in Gaussian type orbitals (the all-electron basis sets are O 8-411(d1), Ti 86-411 (d41) and H 511(p1)). The B3LYP [24,25] and HSE06 [26] hybrid functionals have been used throughout this work. The values of the optimized lattice parameters are 3.789 Å and 3.766 Å for a and 9.777 Å and 9.663 Å for c, respectively for and B3LYP and HSE06, which are comparable to the experimental values [27]. The TiO2 anatase bulk was modeled by a 6√2 ! 6√2 ! 1 bulk supercell with 864 atoms.
………Nanocrystals (NC) have been cut from a bulk anatase supercell following the procedure already described in our previous work [28]. Two stoichiometric nanocrystals (TiO2)159 " 4H2O (NCS) and (TiO2)_260 ……(from the computational section of the paper).
MASSIVE PARALLELISM of codes: not always the evidence is… evident.
CRYSTAL scales extremely well, also with hybrids, up to say 4096 or even more cores
A new massively parallel version of CRYSTAL for large systems on high performance computing architectures
By: Orlando, Roberto; Delle Piane, Massimo; Bush, Ian J.; et al.
JOURNAL OF COMPUTATIONAL CHEMISTRY Volume: 33 Issue: 28 Pages: 2276-2284 Published: OCT 30 2012
Roberto Dovesi
- Badges
- Science topic