I want to mention that the exchange will depend on very small changes in the Cu-Cu distance. There is a discrepancy between calculations and experimental exchange which I still am not sure where it comes from. I need to do further calculations and make a new video where I clarify some things which are not right from the last videos in the series. I still don't completely understand what's going on! So take this information with a grain of salt! Check comments below, some viewers have made important comments!
Actually, the Cu-Cu distance is not the main factor affecting the exchange magnitude in copper paddlewheels, since the exchange primarily occurs through bridging ligands rather than direct Cu-Cu orbitals overlap. Yet it is strange why the Cu-Cu distance changes so drastically upon optimization.
I tried to optimize the same structure by means of B3LYP / def2-tzvp. In my case, the Cu-Cu distance retains at ca. 2.62 Å, but the apical Cu-O bond increases up to 2.3 Å. The resulting BS-DFT solution yields the J value of ca. -400 cm-1 that is definitely inacceptiple. Where could the reason lie in, anc is it ok to use B3LYP for bs-dft calculations?
@@MikhailKendin maybe you are right. I haven't done studies in these type of complexes to separate the different contributions. It probably has an indirect effect because the bite angle of the carboxylates may change. How big it is I wouldn't know.
@@MikhailKendin I believe the packing effects have some influence. But also dispersion corrections are usually necessary to get realistic structures. The apical bond distances in copper complexes are influenced by the Jann Teller distortion and therefore I think that good functionals are necessary. B3lyp should be ok. I don't use it though, my workhorse is PBE0.
Are you sure that the experimental value that you are showing for the byp complex is not using the -JS1S2 hamiltonian? If so, orca is giving the results using -2JS1S2, which means that you have to divide 172 cm-1 by 2 to be able to compare the calculated and the experimental value, if I am not mistaken. And I would not recommend going directly to the optimized structure (if you have a good and reliable X-ray structure!) to calculate the exchange constant using BS-DFT because of several reasons, but mainly because even if you think that a change in 2 degrees of the Cu-O-Cu angles is not that large, it has indeed a HUGE effect in the J coupling. Thanks for the video! I am always very interested in magnetism :)
Hi, in the paper by Hatfield they use the -2JS1*S2 for the HDVV Hamiltonian, but they report the 2J = +172 cm^-1 value, and they refer to it as singlet-triplet splitting. Orca uses -2JS1*S2 but they report J, so that appears to be a different definition. I may have gotten this wrong! I asked in the Orca forum about this and will issue a clarification/rectification once I'm sure. Regarding the crystal/DFT geometry I agree with you. However in the previous video people had criticized the fact I was using crystal unoptimized geometries, and as I (perhaps), was looking at the wrong values for the singlet-triplet gap, I thought that indeed the results were bad because of the lack of optimization. I think that some geometry optimization may be necessary in some cases. If very subtle structural features are to be studied, perhaps optimizations with some constrains could be useful. But a good study should explore a set of different conditions and options to see how much they alter the results, to see if a result is really robust, or what it depends on.
@@niconeuman Thanks for the answer! I agree with the optimization with some constrains, or maybe just hydrogen optimization could also be a good idea. But in some cases it's known that the position of the hydrogen of the bridging OH also changes the calculated J! Like in the case of the complex with the byp. So, it depends on the situation I guess. But generally if I have good quality x-ray, I only go for the hydrogen opt or no optimization at all. Regarding the experimental value, I am very curious, because I remember working with it and I think I used J=86 cm-1 or something like this. Good videos as usual! Keep going 🙂
Hello, again it was helpful a lot. Could you consider to make a video to extract (individual) exchange coupling constants of multinuclear, say tetranuclear transition metal complexes?
hi, sorry for the delay. it will take me some time because I don't know how to practically do it. also I have to search a suitable example that is not too computationally expensive. transition metal clusters are usually large systems.
Hi, unfortunately not for some time. I have no experience with setting up force fields and haven't had time to learn. I have tested the qmmm module using qm/qm, with DFT on the solute and semiempirical methods such as XTB on the solvent. I may do a video on that in a month or so.
The crystal structure Cu positions and the Cu-Cu distances are likely to be right, because they are the strongest scatterers in the crystal. Maybe the calculations are missing something? I believe that the experimental Cu-Cu distance are more likely to be correct, not the geometry optimized distances!
I agree that the experiment is correct. DFT will only roughly reproduce bond lengths and angles unless the chemical environment is realistic. in crystals it's the packing. I will eventually study if I can optimize a cluster of molecules in the crystal and see how much of a difference it makes.
Yep. Most likely the H2O molecules are doing H-bond with the next molecule in the crystal. And when I optimize without the surrounding molecules the H2O attempts to H-bond intramolecularly. Those are some of the problems of trying to model a crystal with a single molecular complex. It is worse when you have a covalent extended material and you need to break bonds and replace functional groups. I could add some surrounding complexes and fix their coordinates, but I'm afraid that will have to wait. I'm quite busy at the moment.
@@niconeuman Cu-O-H should me "more tetrahedral", maybe CPCM would be enough to do the job for extended materials, maybe nwchem with plane waves would be better but unfortunately I have almost zero experience with it (I only find DFT hybrids are unbelievably slow when using plane waves)
I worked with F.R.L. Schoening, who did the copper acetate structure, during my Ph.D.
That's very cool! I worked with a Prof. for many years who did a postdoc with Stevens, who created the equivalent crystal field operators.
I want to mention that the exchange will depend on very small changes in the Cu-Cu distance. There is a discrepancy between calculations and experimental exchange which I still am not sure where it comes from. I need to do further calculations and make a new video where I clarify some things which are not right from the last videos in the series. I still don't completely understand what's going on! So take this information with a grain of salt! Check comments below, some viewers have made important comments!
Actually, the Cu-Cu distance is not the main factor affecting the exchange magnitude in copper paddlewheels, since the exchange primarily occurs through bridging ligands rather than direct Cu-Cu orbitals overlap. Yet it is strange why the Cu-Cu distance changes so drastically upon optimization.
I tried to optimize the same structure by means of B3LYP / def2-tzvp. In my case, the Cu-Cu distance retains at ca. 2.62 Å, but the apical Cu-O bond increases up to 2.3 Å. The resulting BS-DFT solution yields the J value of ca. -400 cm-1 that is definitely inacceptiple. Where could the reason lie in, anc is it ok to use B3LYP for bs-dft calculations?
@@MikhailKendin maybe you are right. I haven't done studies in these type of complexes to separate the different contributions. It probably has an indirect effect because the bite angle of the carboxylates may change. How big it is I wouldn't know.
@@MikhailKendin I believe the packing effects have some influence. But also dispersion corrections are usually necessary to get realistic structures. The apical bond distances in copper complexes are influenced by the Jann Teller distortion and therefore I think that good functionals are necessary. B3lyp should be ok. I don't use it though, my workhorse is PBE0.
Are you sure that the experimental value that you are showing for the byp complex is not using the -JS1S2 hamiltonian? If so, orca is giving the results using -2JS1S2, which means that you have to divide 172 cm-1 by 2 to be able to compare the calculated and the experimental value, if I am not mistaken.
And I would not recommend going directly to the optimized structure (if you have a good and reliable X-ray structure!) to calculate the exchange constant using BS-DFT because of several reasons, but mainly because even if you think that a change in 2 degrees of the Cu-O-Cu angles is not that large, it has indeed a HUGE effect in the J coupling.
Thanks for the video! I am always very interested in magnetism :)
Hi, in the paper by Hatfield they use the -2JS1*S2 for the HDVV Hamiltonian, but they report the 2J = +172 cm^-1 value, and they refer to it as singlet-triplet splitting. Orca uses -2JS1*S2 but they report J, so that appears to be a different definition. I may have gotten this wrong! I asked in the Orca forum about this and will issue a clarification/rectification once I'm sure.
Regarding the crystal/DFT geometry I agree with you. However in the previous video people had criticized the fact I was using crystal unoptimized geometries, and as I (perhaps), was looking at the wrong values for the singlet-triplet gap, I thought that indeed the results were bad because of the lack of optimization.
I think that some geometry optimization may be necessary in some cases. If very subtle structural features are to be studied, perhaps optimizations with some constrains could be useful. But a good study should explore a set of different conditions and options to see how much they alter the results, to see if a result is really robust, or what it depends on.
@@niconeuman Thanks for the answer! I agree with the optimization with some constrains, or maybe just hydrogen optimization could also be a good idea. But in some cases it's known that the position of the hydrogen of the bridging OH also changes the calculated J! Like in the case of the complex with the byp. So, it depends on the situation I guess. But generally if I have good quality x-ray, I only go for the hydrogen opt or no optimization at all. Regarding the experimental value, I am very curious, because I remember working with it and I think I used J=86 cm-1 or something like this. Good videos as usual! Keep going 🙂
Thank you very much. Always so clear
Crystal packing forces for molecular crystals are generally very weak unless there are intermolecular hydrogen bonds.
Hello, again it was helpful a lot. Could you consider to make a video to extract (individual) exchange coupling constants of multinuclear, say tetranuclear transition metal complexes?
hi, sorry for the delay. it will take me some time because I don't know how to practically do it. also I have to search a suitable example that is not too computationally expensive. transition metal clusters are usually large systems.
Please. Would there be any tutorial for calculations with QM/MM?
Hi, unfortunately not for some time. I have no experience with setting up force fields and haven't had time to learn. I have tested the qmmm module using qm/qm, with DFT on the solute and semiempirical methods such as XTB on the solvent. I may do a video on that in a month or so.
@@niconeuman It would definitely be quite interesting too.
so finally I uploaded a QM/QM example using XTB. it took me a while. if you are still interested check it out!
The crystal structure Cu positions and the Cu-Cu distances are likely to be right, because they are the strongest scatterers in the crystal. Maybe the calculations are missing something? I believe that the experimental Cu-Cu distance are more likely to be correct, not the geometry optimized distances!
I agree that the experiment is correct. DFT will only roughly reproduce bond lengths and angles unless the chemical environment is realistic. in crystals it's the packing. I will eventually study if I can optimize a cluster of molecules in the crystal and see how much of a difference it makes.
The waters of paddlewheel complex still doesn't look very good, maybe some solvent correction should be applied.
Yep. Most likely the H2O molecules are doing H-bond with the next molecule in the crystal. And when I optimize without the surrounding molecules the H2O attempts to H-bond intramolecularly. Those are some of the problems of trying to model a crystal with a single molecular complex. It is worse when you have a covalent extended material and you need to break bonds and replace functional groups. I could add some surrounding complexes and fix their coordinates, but I'm afraid that will have to wait. I'm quite busy at the moment.
@@niconeuman Cu-O-H should me "more tetrahedral", maybe CPCM would be enough to do the job
for extended materials, maybe nwchem with plane waves would be better but unfortunately I have almost zero experience with it (I only find DFT hybrids are unbelievably slow when using plane waves)