Hello,\nJust opened the box and looked at structures.csv .\nThe third molecule (dsgdb9nsd_000003) is water.\nI calculated the distances and the angles seen fro the oxygen.\nThe distances and the angle are not exactly as expected.\nAny idea why?\nWould it be because liquid phase is assumed?\nThanks,\nMichel
", "messages": [ { "id": "546119", "postDate": "06/06/2019 09:17:42", "content": "Hello,\nJust opened the box and looked at structures.csv .\nThe third molecule (dsgdb9nsd_000003) is water.\nI calculated the distances and the angles seen fro the oxygen.\nThe distances and the angle are not exactly as expected.\nAny idea why?\nWould it be because liquid phase is assumed?\nThanks,\nMichel
", "rawMarkdown": "Hello,\nJust opened the box and looked at structures.csv .\nThe third molecule (dsgdb9nsd_000003) is water.\nI calculated the distances and the angles seen fro the oxygen.\nThe distances and the angle are not exactly as expected.\nAny idea why?\nWould it be because liquid phase is assumed?\nThanks,\nMichel", "votes": null }, { "id": "546164", "postDate": "06/06/2019 10:07:29", "content": "the way I see it, the atoms presented in the structures.csv file doesn't necessarily mention all the atoms present in the molecule. It just contains some of the atoms along with their positions.
eg: 2 Hydrogen atoms and 1 Oxygen atom can be present in H2SO4, CH3OH and so on.
", "rawMarkdown": "the way I see it, the atoms presented in the `structures.csv` file doesn't necessarily mention all the atoms present in the molecule. It just contains some of the atoms along with their positions. \n\neg: 2 Hydrogen atoms and 1 Oxygen atom can be present in H2SO4, CH3OH and so on.", "votes": null }, { "id": "546564", "postDate": "06/06/2019 17:28:44", "content": "Why do you think so? I would be very surprised...
\n\nI would actually guess, the reason the angle is a bit off to the usual number, is the fact, that the molecules are calculated at 0 K and then the angle is a bit different from the usual ~300 K (also see this thread).
", "rawMarkdown": "Why do you think so? I would be very surprised...\n\nI would actually guess, the reason the angle is a bit off to the usual number, is the fact, that the molecules are calculated at 0 K and then the angle is a bit different from the usual ~300 K (also see [this thread](https://www.kaggle.com/c/champs-scalar-coupling/discussion/94117#latest-542356)).", "votes": null }, { "id": "546702", "postDate": "06/06/2019 20:27:44", "content": "Structures.csv and the individual structure .xyz files do appear to contain all the atoms in the molecule.
\n\nAs Berni says, the temperature and pressure from the QM calculation that generated the structure was probably not standard temperature and pressure.
", "rawMarkdown": "Structures.csv and the individual structure .xyz files do appear to contain **all** the atoms in the molecule.\n\nAs Berni says, the temperature and pressure from the QM calculation that generated the structure was probably not standard temperature and pressure.", "votes": null }, { "id": "546712", "postDate": "06/06/2019 20:39:31", "content": "Hi\nYou're right it's water! My guess is that the structure file is not the ideal (perfect) structures but rather structures calculated by first-principles (DFT or CCSD or Hartree-Fock or whatever). Since these are approximations of the real solution of the Schrodinger equation, we can expect some small differences.
\n\nThat's my guess :)
", "rawMarkdown": "Hi\nYou're right it's water! My guess is that the structure file is not the ideal (perfect) structures but rather structures calculated by first-principles (DFT or CCSD or Hartree-Fock or whatever). Since these are approximations of the real solution of the Schrodinger equation, we can expect some small differences.\n\nThat's my guess :)", "votes": null }, { "id": "546821", "postDate": "06/06/2019 23:36:52", "content": "what evidence do you have to prove that molecules in structures.csv happen to have all the molecules present in that specific molecule? By observing the data this seems so, but how can we be 100% sure?
Well, this kernel has gone through and managed to assign bonds to all structures: https://www.kaggle.com/aekoch95/bonds-from-structure-data
\n\nSince there are no dangling bonds, all atoms must be present in each specific molecule.
", "rawMarkdown": "Well, this kernel has gone through and managed to assign bonds to all structures: https://www.kaggle.com/aekoch95/bonds-from-structure-data\n\nSince there are no dangling bonds, all atoms must be present in each specific molecule.", "votes": null }, { "id": "558001", "postDate": "06/21/2019 17:07:33", "content": "The above mentioned dsgdb9nsd_000003 has only 1 entry in the train/test with atom_index_0=1 and atom_index_1=2. Why is that strange combination of atom_index_x and why only one pair? Should it be 2 pairs?\ndo actually atom_index_x mean anything? is there some real logic behind there values? I thought they cover all permutations of atoms?
", "rawMarkdown": "The above mentioned dsgdb9nsd_000003 has only 1 entry in the train/test with atom_index_0=1 and atom_index_1=2. Why is that strange combination of atom_index_x and why only one pair? Should it be 2 pairs?\ndo actually atom_index_x mean anything? is there some real logic behind there values? I thought they cover all permutations of atoms?", "votes": null }, { "id": "558008", "postDate": "06/21/2019 17:23:54", "content": "Why should they ask for couplings of every possible combinations of atoms? The train data does not describe the molecules! The molecules are described in structures.csv. There you'll also find which atoms the indices in the couplings correspond to. (And as only couplings with at least one hydrogen atom are part of this challenge, there is only one possible coupling in H2O.)
the way I see it, the atoms presented in the structures.csv file doesn't necessarily mention all the atoms present in the molecule. It just contains some of the atoms along with their positions.
eg: 2 Hydrogen atoms and 1 Oxygen atom can be present in H2SO4, CH3OH and so on.
", "votes": null, "replies": [ { "id": 546564, "author_name": "bernir", "author_url": "", "post_date": "06/06/2019 17:28:44", "content": "Why do you think so? I would be very surprised...
\n\nI would actually guess, the reason the angle is a bit off to the usual number, is the fact, that the molecules are calculated at 0 K and then the angle is a bit different from the usual ~300 K (also see this thread).
", "votes": null, "replies": [] }, { "id": 546702, "author_name": "phucco", "author_url": "", "post_date": "06/06/2019 20:27:44", "content": "Structures.csv and the individual structure .xyz files do appear to contain all the atoms in the molecule.
\n\nAs Berni says, the temperature and pressure from the QM calculation that generated the structure was probably not standard temperature and pressure.
", "votes": null, "replies": [] }, { "id": 546821, "author_name": "ranitha", "author_url": "", "post_date": "06/06/2019 23:36:52", "content": "what evidence do you have to prove that molecules in structures.csv happen to have all the molecules present in that specific molecule? By observing the data this seems so, but how can we be 100% sure?
Well, this kernel has gone through and managed to assign bonds to all structures: https://www.kaggle.com/aekoch95/bonds-from-structure-data
\n\nSince there are no dangling bonds, all atoms must be present in each specific molecule.
", "votes": null, "replies": [] } ] }, { "id": 546712, "author_name": "borisdee", "author_url": "", "post_date": "06/06/2019 20:39:31", "content": "Hi\nYou're right it's water! My guess is that the structure file is not the ideal (perfect) structures but rather structures calculated by first-principles (DFT or CCSD or Hartree-Fock or whatever). Since these are approximations of the real solution of the Schrodinger equation, we can expect some small differences.
\n\nThat's my guess :)
", "votes": null, "replies": [] }, { "id": 558001, "author_name": "snovik1975", "author_url": "", "post_date": "06/21/2019 17:07:33", "content": "The above mentioned dsgdb9nsd_000003 has only 1 entry in the train/test with atom_index_0=1 and atom_index_1=2. Why is that strange combination of atom_index_x and why only one pair? Should it be 2 pairs?\ndo actually atom_index_x mean anything? is there some real logic behind there values? I thought they cover all permutations of atoms?
", "votes": null, "replies": [ { "id": 558008, "author_name": "bernir", "author_url": "", "post_date": "06/21/2019 17:23:54", "content": "Why should they ask for couplings of every possible combinations of atoms? The train data does not describe the molecules! The molecules are described in structures.csv. There you'll also find which atoms the indices in the couplings correspond to. (And as only couplings with at least one hydrogen atom are part of this challenge, there is only one possible coupling in H2O.)