I just started learning computational chemistry. I try to find many videos on TH-cam. But your videos are super good for beginners . I am going to watch all of them. Thanks Nicolas
Thank you very much for the explanation, Nicolas! My student and I are trying to do something similar for two nanocells, one formed by 6 Cu atoms and the other by 6 Ni atoms. However, we are having difficulty determining the values of: mult, nroot, norbs and nel. Could you please help us how we can determine these values?
These are difficult problems to solve. First, it depends on the total charge and therefore on the oxidation states of the atoms. If you have 6 Cu(0) atoms, their valence configurations are 3d10 4s1 (right?). So in principle you would need an active space of 6 x 6 orbitals and 6 x 11 electrons, which is quite large. The number of unpaired electrons is even, so you would have S(S+1)= 1, 3, 5, 7, etc, and you would need to calculate all this possibilities to see which ones are lower in energy. You will likely be unable to do this without approximations. So you could try an active space of 6 orbitals and 6 electrons, and test the lower allowable multiplicities. If that gives you different combinations of 4s orbitals, you are on a good track. Then you can increase it to 12 orb and 18 electrons, so you will get one 3d and one 4s orbital for each atom. But you will have to investigate further and see what happens. Of course if you have charged clusters you just add or subtract electrons from the active space, and change the multiplicity accordingly. The nroots is the number of excited states you want. It depends on what properties you would like to calculate. For a ground state calculation you just need one, but that is sometimes hard to converge. You can start with 6 or 12 roots, and see what you get (UV/vis transitions for example). And please use NEVPT2 or an equivalent perturbative correction. CASSCF alone is not accurate for energies.
@@niconeuman CASSCF Input: !def2-svp nofrozencore PAtom %casscf nel 6 norb 6 mult 1 nroots 2 PTMethod FIC-CASPT2 end end * xyz 0 1 N 0.0 0.0 0.0 N 0.0 0.0 1.09768 * Error : Unknown identifier in [CASSCF] block line 6 : Last token : PTMETHOD [file orca_main/maininp1.cpp, line 8873]:
@@niconeuman Input: !def2-svp nofrozencore PAtom %casscf nel 6 norb 6 mult 1 nroots 2 PTMethod FIC-CASPT2 end end * xyz 0 1 N 0.0 0.0 0.0 N 0.0 0.0 1.09768 * Error : Unknown identifier in [CASSCF] block line 6 : Last token : PTMETHOD [file orca_main/maininp1.cpp, line 8873]:
Input: !def2-svp nofrozencore PAtom %casscf nel 6 norb 6 mult 1 nroots 2 PTMethod FIC-CASPT2 end end * xyz 0 1 N 0.0 0.0 0.0 N 0.0 0.0 1.09768 * Error : Unknown identifier in [CASSCF] block line 6 : Last token : PTMETHOD [file orca_main/maininp1.cpp, line 8873]:
I'm sorry, but that is from an unfinished project, which I can't share. But what do you need? the input instructions for CASSCF are well covered in the CASSCF tutorial that can be found in the downloads section in the Orca Forum. Also on the Orca manual. There are many options, but the simple ones are covered in many examples.
@@niconeuman Dear Nicolas, Thank you for the answer. I actually managed to do it. One question, being a perturbation method, does it have any parameters related to energy level shifting (like CASPT2, for instance)? Because the energies I've obtained are around 0.5/0.8 eV below what I expected. I know that perturbation theory allows energy to lay below the "exact" value, but I was wondering if there are any parameters that I can apply in order to try to make it more accurate. Again, thanks for the attention and the answer.
you can select different active spaces. for a closed shell molecule, you will have an even number of electrons. so you can choose a 4,4 active space, which means two doubly occupied and two empty orbitals, or a 6,6 or 6,8 etc. the active space is always centred around the homo lumo gap. so if you want more occupied orbitals you increase at the same time the electrons and orbitals, and if you want more empty orbitals you only increase norb. what is best to choose I can't tell you. start with a small active space and see if you observe transitions corresponding to the experimental ones. and if not you can increase it. as you increase the size of the active space you sometimes need to increase the roots to see more transitions.
Hi Nicolas, I'm having problems with Rutenium (I) (Neutral), the calculation gives an error regarding to the multiplicity: [file orca_main/mainfcts.cpp, line 629]: Error : multiplicity (2) is even and number of electrons (464) is even -> impossible Can you please help me?
if you have an even number of electrons you can only have spin multiplicities of 1, 3, 5, etc, which are odd. so either your total charge is wrong, or you are asking for a doublet when you should have a singlet or triplet, or perhaps you have more or less atoms than you think. most likely, if you took your coordinates from a crystal structure, you may have missing or extra H atoms. or you could have a disorder which gives you some extra atoms you have to delete.
I am planning a CASSF/NEVPT2 on an open shell-iron complex (roughly 100 atoms) with non-innocent ligands. My experience is still limited and I am rather bad at guessing computational cost. Would a calculatin like this be feasible on a HPC or would you recommend to go inthe direction of DLNPO-NEVPT2?
In a cluster it would be feasible. You generally need a good amount of memory and disk for nevpt2. Start with def2-SVP on non metal atoms and you can see how long it takes. But that doesn't look that expensive.
@@niconeuman Thank you for your fast answer! The cluster has a handfull of nodes with 1.5TB of memory and up to 7TB of NVME scratch disks so I suppose that should be enough. What is your recommendation in regards of reaching Basis set limit?
@@Chem-iu5jx I don't have a recommendation on that. Depending on the calculations I've had very good results even with def2-SVP on the metal. In some cases def2-TZVP gave better results but I've never gone beyond. I would guess that def2-QZVP with NEVPT2 will be VERY costly
good video!! thanks. one question: i started learning zero-field calculations, for zero-field calculates, is necessary first calculate cassf? or is diferent ?
Hi, zero field splittings (ZFS) can be calculated by various methods. You can calculate them using DFT, choosing an option inside the %eprnmr module, or you can calculate them from casscf/nevpt2, with the keywords I showed in the example. For organic molecules, where the ZFS may come mainly from spin-spin magnetic interaction, DFT may do a good work. For transition metal complexes with high-spin metals such as Co(II), Fe(III), Mn, or lanthanides, DFT usually doesn't give good results and it is better to use CASSCF/NEVPT2. There are other theoretical methods such as coupled cluster, which I think also allow you to calculate ZFS, but I have never used them.
I haven't used caspt2. Probably it's similar but you'll have to check the manual. Also, this video was done in Orca 4.2 iirc. If you are using version 5.0x there are a few syntax changes. So first check that. In the manual it lists the main differences in definition of grids, and other keywords. Cheers
@@niconeuman sir, I also has a question about the nroot of CASSCF, I think the the number of root calculation is first search the T-S diagram to identify the how many such as 3T1、3A1、3E or 3T2 etc. Then, the overall root is 3T1*3+3E*2+1*A1... for the octahedral ligand field. sir, do you think my algorithm is reliable?
This might be a rather stupid question: how exactly do I get the total number of possible singlet, triplet configurations of a given d^n configuration?
hi, it's not a stupid question. there are formulas but I can't remember the name now. in the orca forum someone asked something similar and Prof. Neese mentioned the name of the formula. I remember that for cobalt(ii) only considering d orbitals there are in total 120 configurations, 40 of which are HS and 80 LS. that means 10 roots for quartet spin and 40 roots for doublet spin. the 120 comes from the ways of distributing 7 electrons in 10 spinorbitals: 10!/7!/3! for Ni(ii) for example you would have 10!/8!/2! = 45 and because you can have singlet and triplet you need 3x+1y = 45. the number of triplet configurations would be equal to distributing two holes in five orbitals, so it would be 5!/3!/2! = 10. so that would mean that there are ten triplet roots and therefore y = 15. please someone correct me if I'm wrong, I didn't write down the possibilities so maybe I did a mistake. hope this helps.
you mean the maximum number of roots? that depends on the number of electrons, orbitals and multiplicity. There are formulas but I don't remember them. I just know that for Co(II) (7 el, 5 orb), there are 10 quartet and 40 doublet roots. But if you use a larger active space the number of roots will increase very fast. I usually use small numbers, in the 10-100 range, because I'm not interested in modelling ALL excited states, but only a few low-energy ones. Regarding the weights. These calculations are state-averaged CASSCF. This means that orbitals are optimized to describe best all possible roots included in the calculations. Normally all roots have the same weight (importance) in defining the optimal orbitals. But you can change those weights. For example if you use 100 roots, hypothetically that may lead to orbitals which are weird-looking for a certain calculation. Then you can change the weights so that only the first ten roots are used to optimize the orbitals. Then the other roots will be calculated (energies and compositions) using orbitals that are optimal for the first ten roots. This may improve the description of properties that only depend on low-lying excited states, such as magnetic properties, but may worsen other properties such as UV-vis absorption to high-lying roots. There is no general advice I can give, you will have to test for your systems and try to see what works best. I would not touch the weights unless I really had to. You have to provide weights for all roots that you choose, even if the values are zero for most of them. If you have 100 roots but mistakenly only provide 99 weights, you will get an error. Hope this helps. Cheers!
Can you please suggest some references to check formulas for root calculation? I am trying calculations with different configurations for which knowing about no. of roots is a must.
![[ORCA tutorial] H₂O geometry optimization in under 10 mins (2021) [ENG SUB]](http://i.ytimg.com/vi/onU2vPIGunE/mqdefault.jpg)
I just started learning computational chemistry. I try to find many videos on TH-cam. But your videos are super good for beginners . I am going to watch all of them. Thanks Nicolas
Thanks for the comment! I try to make the videos practical. It's nice that you find them useful
Thank you very much for the explanation, Nicolas!
My student and I are trying to do something similar for two nanocells, one formed by 6 Cu atoms and the other by 6 Ni atoms. However, we are having difficulty determining the values of: mult, nroot, norbs and nel. Could you please help us how we can determine these values?
These are difficult problems to solve. First, it depends on the total charge and therefore on the oxidation states of the atoms. If you have 6 Cu(0) atoms, their valence configurations are 3d10 4s1 (right?). So in principle you would need an active space of 6 x 6 orbitals and 6 x 11 electrons, which is quite large. The number of unpaired electrons is even, so you would have S(S+1)= 1, 3, 5, 7, etc, and you would need to calculate all this possibilities to see which ones are lower in energy. You will likely be unable to do this without approximations. So you could try an active space of 6 orbitals and 6 electrons, and test the lower allowable multiplicities. If that gives you different combinations of 4s orbitals, you are on a good track. Then you can increase it to 12 orb and 18 electrons, so you will get one 3d and one 4s orbital for each atom. But you will have to investigate further and see what happens. Of course if you have charged clusters you just add or subtract electrons from the active space, and change the multiplicity accordingly. The nroots is the number of excited states you want. It depends on what properties you would like to calculate. For a ground state calculation you just need one, but that is sometimes hard to converge. You can start with 6 or 12 roots, and see what you get (UV/vis transitions for example). And please use NEVPT2 or an equivalent perturbative correction. CASSCF alone is not accurate for energies.
Thank you very much, Nicolas!
This is an extremely informative video. I sincerely appreciate it.
Thank you very much! It's very nice to hear!
Hello Sir, I could not run caspt2 calculations after casscf even after following the manual. Can you please help me ?
Can you post the error you get? And the casscf part of your input file?
@@niconeuman
CASSCF Input:
!def2-svp nofrozencore PAtom
%casscf nel 6
norb 6
mult 1
nroots 2
PTMethod FIC-CASPT2
end
end
* xyz 0 1
N 0.0 0.0 0.0
N 0.0 0.0 1.09768
*
Error :
Unknown identifier in [CASSCF] block line 6 :
Last token : PTMETHOD
[file orca_main/maininp1.cpp, line 8873]:
@@niconeuman
Input:
!def2-svp nofrozencore PAtom
%casscf nel 6
norb 6
mult 1
nroots 2
PTMethod FIC-CASPT2
end
end
* xyz 0 1
N 0.0 0.0 0.0
N 0.0 0.0 1.09768
*
Error :
Unknown identifier in [CASSCF] block line 6 :
Last token : PTMETHOD
[file orca_main/maininp1.cpp, line 8873]:
Input:
!def2-svp nofrozencore PAtom
%casscf nel 6
norb 6
mult 1
nroots 2
PTMethod FIC-CASPT2
end
end
* xyz 0 1
N 0.0 0.0 0.0
N 0.0 0.0 1.09768
*
Input:
!def2-svp nofrozencore PAtom
%casscf nel 6
norb 6
mult 1
nroots 2
PTMethod FIC-CASPT2
end
end
* xyz 0 1
N 0.0 0.0 0.0
N 0.0 0.0 1.09768
*
Error :
Unknown identifier in [CASSCF] block line 6 :
Last token : PTMETHOD
[file orca_main/maininp1.cpp, line 8873]:
Hi. Very nice video! Could you please share the input file example?
I'm sorry, but that is from an unfinished project, which I can't share. But what do you need? the input instructions for CASSCF are well covered in the CASSCF tutorial that can be found in the downloads section in the Orca Forum. Also on the Orca manual. There are many options, but the simple ones are covered in many examples.
@@niconeuman Dear Nicolas, Thank you for the answer. I actually managed to do it. One question, being a perturbation method, does it have any parameters related to energy level shifting (like CASPT2, for instance)? Because the energies I've obtained are around 0.5/0.8 eV below what I expected. I know that perturbation theory allows energy to lay below the "exact" value, but I was wondering if there are any parameters that I can apply in order to try to make it more accurate. Again, thanks for the attention and the answer.
I want to run CASSCF on a simple BODIPY molecule. Can you help me how to select the orbitals and electrons?
you can select different active spaces. for a closed shell molecule, you will have an even number of electrons. so you can choose a 4,4 active space, which means two doubly occupied and two empty orbitals, or a 6,6 or 6,8 etc. the active space is always centred around the homo lumo gap. so if you want more occupied orbitals you increase at the same time the electrons and orbitals, and if you want more empty orbitals you only increase norb. what is best to choose I can't tell you. start with a small active space and see if you observe transitions corresponding to the experimental ones. and if not you can increase it. as you increase the size of the active space you sometimes need to increase the roots to see more transitions.
Hi Nicolas, I'm having problems with Rutenium (I) (Neutral), the calculation gives an error regarding to the multiplicity:
[file orca_main/mainfcts.cpp, line 629]: Error : multiplicity (2) is even and number of electrons (464) is even -> impossible
Can you please help me?
if you have an even number of electrons you can only have spin multiplicities of 1, 3, 5, etc, which are odd. so either your total charge is wrong, or you are asking for a doublet when you should have a singlet or triplet, or perhaps you have more or less atoms than you think. most likely, if you took your coordinates from a crystal structure, you may have missing or extra H atoms. or you could have a disorder which gives you some extra atoms you have to delete.
I am planning a CASSF/NEVPT2 on an open shell-iron complex (roughly 100 atoms) with non-innocent ligands. My experience is still limited and I am rather bad at guessing computational cost. Would a calculatin like this be feasible on a HPC or would you recommend to go inthe direction of DLNPO-NEVPT2?
In a cluster it would be feasible. You generally need a good amount of memory and disk for nevpt2. Start with def2-SVP on non metal atoms and you can see how long it takes. But that doesn't look that expensive.
@@niconeuman Thank you for your fast answer! The cluster has a handfull of nodes with 1.5TB of memory and up to 7TB of NVME scratch disks so I suppose that should be enough. What is your recommendation in regards of reaching Basis set limit?
@@Chem-iu5jx I don't have a recommendation on that. Depending on the calculations I've had very good results even with def2-SVP on the metal. In some cases def2-TZVP gave better results but I've never gone beyond. I would guess that def2-QZVP with NEVPT2 will be VERY costly
good video!! thanks. one question: i started learning zero-field calculations, for zero-field calculates, is necessary first calculate cassf? or is diferent ?
Hi, zero field splittings (ZFS) can be calculated by various methods. You can calculate them using DFT, choosing an option inside the %eprnmr module, or you can calculate them from casscf/nevpt2, with the keywords I showed in the example. For organic molecules, where the ZFS may come mainly from spin-spin magnetic interaction, DFT may do a good work. For transition metal complexes with high-spin metals such as Co(II), Fe(III), Mn, or lanthanides, DFT usually doesn't give good results and it is better to use CASSCF/NEVPT2. There are other theoretical methods such as coupled cluster, which I think also allow you to calculate ZFS, but I have never used them.
Clarísimo! Muchas gracias
good video!! thanks. But I have one question: the CASPT2 input file in ORCA is similar to CASSCF? I use your input file but it reports an error.
I haven't used caspt2. Probably it's similar but you'll have to check the manual. Also, this video was done in Orca 4.2 iirc. If you are using version 5.0x there are a few syntax changes. So first check that. In the manual it lists the main differences in definition of grids, and other keywords.
Cheers
@@niconeuman Thanks for your kind response!!
@@niconeuman sir, I also has a question about the nroot of CASSCF, I think the the number of root calculation is first search the T-S diagram to identify the how many such as 3T1、3A1、3E or 3T2 etc. Then, the overall root is 3T1*3+3E*2+1*A1... for the octahedral ligand field. sir, do you think my algorithm is reliable?
This might be a rather stupid question: how exactly do I get the total number of possible singlet, triplet configurations of a given d^n configuration?
hi, it's not a stupid question. there are formulas but I can't remember the name now. in the orca forum someone asked something similar and Prof. Neese mentioned the name of the formula. I remember that for cobalt(ii) only considering d orbitals there are in total 120 configurations, 40 of which are HS and 80 LS. that means 10 roots for quartet spin and 40 roots for doublet spin. the 120 comes from the ways of distributing 7 electrons in 10 spinorbitals: 10!/7!/3!
for Ni(ii) for example you would have 10!/8!/2! = 45 and because you can have singlet and triplet you need 3x+1y = 45. the number of triplet configurations would be equal to distributing two holes in five orbitals, so it would be 5!/3!/2! = 10. so that would mean that there are ten triplet roots and therefore y = 15. please someone correct me if I'm wrong, I didn't write down the possibilities so maybe I did a mistake. hope this helps.
sir, how do we determine no. of roots from multiplicity? And, what is the meaning of weights, how do we assign the weights in a calculation?
you mean the maximum number of roots? that depends on the number of electrons, orbitals and multiplicity. There are formulas but I don't remember them. I just know that for Co(II) (7 el, 5 orb), there are 10 quartet and 40 doublet roots. But if you use a larger active space the number of roots will increase very fast. I usually use small numbers, in the 10-100 range, because I'm not interested in modelling ALL excited states, but only a few low-energy ones.
Regarding the weights. These calculations are state-averaged CASSCF. This means that orbitals are optimized to describe best all possible roots included in the calculations. Normally all roots have the same weight (importance) in defining the optimal orbitals. But you can change those weights. For example if you use 100 roots, hypothetically that may lead to orbitals which are weird-looking for a certain calculation. Then you can change the weights so that only the first ten roots are used to optimize the orbitals. Then the other roots will be calculated (energies and compositions) using orbitals that are optimal for the first ten roots. This may improve the description of properties that only depend on low-lying excited states, such as magnetic properties, but may worsen other properties such as UV-vis absorption to high-lying roots. There is no general advice I can give, you will have to test for your systems and try to see what works best. I would not touch the weights unless I really had to.
You have to provide weights for all roots that you choose, even if the values are zero for most of them. If you have 100 roots but mistakenly only provide 99 weights, you will get an error.
Hope this helps. Cheers!
Thank you, sir. Now, I have a better idea about weights and roots.
Can you please suggest some references to check formulas for root calculation? I am trying calculations with different configurations for which knowing about no. of roots is a must.