Why can not CNT (carbon nanotubes) absorb any atoms?

GROMACS version:2021
GROMACS modification: Yes/No
I work on the adsorption of ions and inhibitors on CNT (armchair structure). Recently, I did a molecular dynamics simulation with Gromacs 2021 software with a Charmm 27 force field. But CNT only interacts with water up to a distance of 0-2 nm a lot. Its interaction with other ions and inhibitors at a distance of more than 2 nm is minimal. I slightly modified the CNT charge with the DFT quantum method, but there was very little change in the species adsorption. The CNT absorbed both the inhibitor and the ion in my previous laboratory work, but absorption does not seem to occur in simulating molecular dynamics. I do not know precisely where the problem in my MD simulation is. Here is my Structure file and topology.
CNT.gro
Green Red Orange Magenta Azure Cyan Skyblue
160
1WAL C 1 5.331 5.781 5.000
1WAL C1 2 5.222 5.781 5.246
1WAL C2 3 5.102 5.781 5.315
1WAL C3 4 4.834 5.781 5.287
1WAL C4 5 4.732 5.781 5.195
1WAL C5 6 4.676 5.781 4.931
1WAL C6 7 4.732 5.781 4.805
1WAL C7 8 4.965 5.781 4.671
1WAL C8 9 5.102 5.781 4.685
1WAL C9 10 5.303 5.781 4.865
1WAL C10 11 5.324 5.661 5.069
1WAL C11 12 5.268 5.661 5.195
1WAL C12 13 5.035 5.661 5.329
1WAL C13 14 4.898 5.661 5.315
1WAL C14 15 4.697 5.661 5.135
1WAL C15 16 4.669 5.661 5.000
1WAL C16 17 4.778 5.661 4.754
1WAL C17 18 4.898 5.661 4.685
1WAL C18 19 5.166 5.661 4.713
1WAL C19 20 5.268 5.661 4.805
1WAL C20 21 5.331 5.541 5.000
1WAL C21 22 5.222 5.541 5.246
1WAL C22 23 5.102 5.541 5.315
1WAL C23 24 4.834 5.541 5.287
1WAL C24 25 4.732 5.541 5.195
1WAL C25 26 4.676 5.541 4.931
1WAL C26 27 4.732 5.541 4.805
1WAL C27 28 4.965 5.541 4.671
1WAL C28 29 5.102 5.541 4.685
1WAL C29 30 5.303 5.541 4.865
1WAL C30 31 5.324 5.421 5.069
1WAL C31 32 5.268 5.421 5.195
1WAL C32 33 5.035 5.421 5.329
1WAL C33 34 4.898 5.421 5.315
1WAL C34 35 4.697 5.421 5.135
1WAL C35 36 4.669 5.421 5.000
1WAL C36 37 4.778 5.421 4.754
1WAL C37 38 4.898 5.421 4.685
1WAL C38 39 5.166 5.421 4.713
1WAL C39 40 5.268 5.421 4.805
1WAL C40 41 5.331 5.300 5.000
1WAL C41 42 5.222 5.300 5.246
1WAL C42 43 5.102 5.300 5.315
1WAL C43 44 4.834 5.300 5.287
1WAL C44 45 4.732 5.300 5.195
1WAL C45 46 4.676 5.300 4.931
1WAL C46 47 4.732 5.300 4.805
1WAL C47 48 4.965 5.300 4.671
1WAL C48 49 5.102 5.300 4.685
1WAL C49 50 5.303 5.300 4.865
1WAL C50 51 5.324 5.180 5.069
1WAL C51 52 5.268 5.180 5.195
1WAL C52 53 5.035 5.180 5.329
1WAL C53 54 4.898 5.180 5.315
1WAL C54 55 4.697 5.180 5.135
1WAL C55 56 4.669 5.180 5.000
1WAL C56 57 4.778 5.180 4.754
1WAL C57 58 4.898 5.180 4.685
1WAL C58 59 5.166 5.180 4.713
1WAL C59 60 5.268 5.180 4.805
1WAL C60 61 5.331 5.060 5.000
1WAL C61 62 5.222 5.060 5.246
1WAL C62 63 5.102 5.060 5.315
1WAL C63 64 4.834 5.060 5.287
1WAL C64 65 4.732 5.060 5.195
1WAL C65 66 4.676 5.060 4.931
1WAL C66 67 4.732 5.060 4.805
1WAL C67 68 4.965 5.060 4.671
1WAL C68 69 5.102 5.060 4.685
1WAL C69 70 5.303 5.060 4.865
1WAL C70 71 5.324 4.940 5.069
1WAL C71 72 5.268 4.940 5.195
1WAL C72 73 5.035 4.940 5.329
1WAL C73 74 4.898 4.940 5.315
1WAL C74 75 4.697 4.940 5.135
1WAL C75 76 4.669 4.940 5.000
1WAL C76 77 4.778 4.940 4.754
1WAL C77 78 4.898 4.940 4.685
1WAL C78 79 5.166 4.940 4.713
1WAL C79 80 5.268 4.940 4.805
1WAL C80 81 5.331 4.820 5.000
1WAL C81 82 5.222 4.820 5.246
1WAL C82 83 5.102 4.820 5.315
1WAL C83 84 4.834 4.820 5.287
1WAL C84 85 4.732 4.820 5.195
1WAL C85 86 4.676 4.820 4.931
1WAL C86 87 4.732 4.820 4.805
1WAL C87 88 4.965 4.820 4.671
1WAL C88 89 5.102 4.820 4.685
1WAL C89 90 5.303 4.820 4.865
1WAL C90 91 5.324 4.700 5.069
1WAL C91 92 5.268 4.700 5.195
1WAL C92 93 5.035 4.700 5.329
1WAL C93 94 4.898 4.700 5.315
1WAL C94 95 4.697 4.700 5.135
1WAL C95 96 4.669 4.700 5.000
1WAL C96 97 4.778 4.700 4.754
1WAL C97 98 4.898 4.700 4.685
1WAL C98 99 5.166 4.700 4.713
1WAL C99 100 5.268 4.700 4.805
1WAL C100 101 5.331 4.580 5.000
1WAL C101 102 5.222 4.580 5.246
1WAL C102 103 5.102 4.580 5.315
1WAL C103 104 4.834 4.580 5.287
1WAL C104 105 4.732 4.580 5.195
1WAL C105 106 4.676 4.580 4.931
1WAL C106 107 4.732 4.580 4.805
1WAL C107 108 4.965 4.580 4.671
1WAL C108 109 5.102 4.580 4.685
1WAL C109 110 5.303 4.580 4.865
1WAL C110 111 5.324 4.459 5.069
1WAL C111 112 5.268 4.459 5.195
1WAL C112 113 5.035 4.459 5.329
1WAL C113 114 4.898 4.459 5.315
1WAL C114 115 4.697 4.459 5.135
1WAL C115 116 4.669 4.459 5.000
1WAL C116 117 4.778 4.459 4.754
1WAL C117 118 4.898 4.459 4.685
1WAL C118 119 5.166 4.459 4.713
1WAL C119 120 5.268 4.459 4.805
1WAL C120 121 5.331 4.339 5.000
1WAL C121 122 5.222 4.339 5.246
1WAL C122 123 5.102 4.339 5.315
1WAL C123 124 4.834 4.339 5.287
1WAL C124 125 4.732 4.339 5.195
1WAL C125 126 4.676 4.339 4.931
1WAL C126 127 4.732 4.339 4.805
1WAL C127 128 4.965 4.339 4.671
1WAL C128 129 5.102 4.339 4.685
1WAL C129 130 5.303 4.339 4.865
1WAL C130 131 5.324 4.219 5.069
1WAL C131 132 5.268 4.219 5.195
1WAL C132 133 5.035 4.219 5.329
1WAL C133 134 4.898 4.219 5.315
1WAL C134 135 4.697 4.219 5.135
1WAL C135 136 4.669 4.219 5.000
1WAL C136 137 4.778 4.219 4.754
1WAL C137 138 4.898 4.219 4.685
1WAL C138 139 5.166 4.219 4.713
1WAL C139 140 5.268 4.219 4.805
1TER H 141 5.386 5.875 5.000
1TER H1 142 5.266 5.875 5.278
1TER H2 143 5.119 5.875 5.367
1TER H3 144 4.818 5.875 5.339
1TER H4 145 4.688 5.875 5.227
1TER H5 146 4.622 5.875 4.931
1TER H6 147 4.688 5.875 4.773
1TER H7 148 4.949 5.875 4.619
1TER H8 149 5.119 5.875 4.633
1TER H9 150 5.347 5.875 4.833
1TER H10 151 5.378 4.125 5.069
1TER H11 152 5.312 4.125 5.227
1TER H12 153 5.051 4.125 5.381
1TER H13 154 4.881 4.125 5.367
1TER H14 155 4.653 4.125 5.167
1TER H15 156 4.614 4.125 5.000
1TER H16 157 4.734 4.125 4.722
1TER H17 158 4.881 4.125 4.633
1TER H18 159 5.182 4.125 4.661
1TER H19 160 5.312 4.125 4.773
10.00000 10.00000 10.00000
I cannot upload CNT.itp, I have to change .itp to .log until I upload it.
CNT.log (183.2 KB)

The charge assignment makes no sense to me. Each C is equivalent in an CNT; the charges you have obtained are likely a geometry-specific solution within the DFT calculation that are incorrect for use in the condensed phase (and inconsistent with the methods of charge assignment used in the CHARMM force field that you say you’re using). I don’t follow CNT simulations much but I do not see any justification for assigning partial charge at all. Each C is equivalent, so essentially an LJ sphere with no charge. Of course, this is a very crude model and I would assume an additive approximation would have serious limitations and a polarizable model would be better (this has been demonstrated).

In any case, any attraction that a species would have with the CNT would be weak, especially in aqueous solution with ions. The magnitudes of charges you’ve assigned (again I think they’re flawed) are generally very small and unlikely to drive any electrostatic interaction, it’s essentially all LJ, which is much weaker in solution that electrostatics.

I have used CNTs generated from CHARMM-GUI and they seem to work fine in sorption of charged and uncharged species. I have not specifically looked into sorption of ions but pi-cation interacttions could me modelled in files generated by CHARMM-GUI. You can get Gromacs-compatible files, among other MD program. One difference from your work is that I used CHARMM36 and not CHARMM27.

Thanks, jalemkul and rpsingh. Dear rpsingh, if possible, send me the topology file uncharge, its partial charge part. As it turns out, I already sent my itp file in Log format. I’m skeptical about the CNT charge right now. If possible, send me your partial CNT unloaded parts for comparison. I already registered on charmm GUI, but I do not know precisely how it works.

How to understand pi-cation interaction works correctly in MD simulation?

In a finite CNT, the ends would carry some dipolar charges one way or another, assuming of course this is where the charges are in this case… Although copy-pasting from DFT is a bit funny (guilty of it myself, but in somewhat more complex situations: Phys. Rev. Lett. 127, 138103 (2021) - Nanopores in Atomically Thin 2D Nanosheets Limit Aqueous Single-Stranded DNA Transport), is CHARMM’s charge assignment all that different from the venerable Hartree-Fock-derived stuff in, say, OPLS-AA?

Sorry. I don’t understand why the DFT charge is invalid with CHARMM. Should I use Hartree-Fock-derived for modifying partial charge in the OPLS-AA force field? Why??

You are simulating an armchair tube, which is metallic and thus “absorption” will not be described correctly, unless you have a properly parameterized dynamic polarization scheme for the CNT carbons. Simulation of these structures requires knowledge of solid-state physics. See for example https://pubs.acs.org/doi/abs/10.1021/acs.jpcc.7b08891

Hartree-Fock or not is not the issue. The non-polarizable tube is. :)

Force fields are self-consistent entities. You have to derive parameters for new species in a consistent way. You can’t parametrize some aspects (like small molecules) using some model chemistry, then choose another model chemistry for another species. If you’re using CHARMM, charge assignment is done using overestimated MP2/6-31G* dipole moments of model compounds, then validated against scaled HF/6-31G* water interactions. I recently gave a webinar on the history and technical details of CHARMM development that you might find useful in this regard: BioExcel Webinar #58: CHARMM Force Field Development History, Features and Implementation in GROMACS - YouTube

Obviously all of that is difficult with something like a CNT, because suitable model compounds have no net dipole moment, and the properties of the entire thing will (as noted above) lead to a geometry-specific solution and you will get a fictitious result. Moreover, you can’t take charges directly from a gas-phase QM calculation and apply them in the condensed phase, because doing so neglects the polarization response in water.

The current situation with MD modeling of nanomaterials departs from any HF-based charges, which would be wildly inaccurate for solids.
Many if not most papers invent self-serving parameters, while relatively rare examples of meaningful work use the DDAP scheme within PBE-based DFT. I wanted to give specific examples from Blankschtein’s group, but, probably with the help of some local online vigilante, this forum decided that I am spamming with links. ;)
The DDAP-based approach is essentially what I have adopted. And this is not including the polarization stuff… As I’ve said many times before on this forum’s predecessor, the disconnect between the biophysics community and solid-state/nanomaterials folks is just staggering and I see no clear solution in mind.

@Sasha if you or anyone else in this domain are interested in contributing to developing polarizable models for materials like CNTs, graphene, etc. I have an interest in doing the parameter work. It’s just been a bit daunting to think about jumping into a field in which I have no experience. The parameter side honestly shouldn’t be too bad using our Drude model (famous last words, right?) and there are a number of interesting applications that I could see for our protein and nucleic acid model in conjunction with these kinds of species. Let me know if you’re interested.

Hey Justin. I would hate to be the guy who starts the conversation and then promptly disappears when the time comes to put the proverbial money where the mouth is. In my second “ignored” post, you can see what they did at MIT using LAMMPS. To avoid overpolarization, they don’t use the simple Drude model, but I suppose it all goes under the Drude “umbrella.”

My main concern is the lack of scenarios that could be used for validation and I’d like to hear your opinion on that. There should be another paper coming up specifically with polarizable CNTs and a comparison with experimental data, so hopefully it’s out soon. Without revealing anything that was shared privately, at the moment the only way to “test” the effects of polarization in CNTs is that experimentally narrow tubes are a lot more open to things going inside them. MD simulations without polarization disagree and it’s pretty easy to see why.

At any rate, I am more than happy to share carefully designed non-polarizable versions of graphene, etc. Parameterization of an “infinite” sheet of graphene is quite easy to test, because at a suitable distance from the plane an ion would interact with its image. For the simplest Drude case, obtaining the spring parameter actually requires no MD simulations, because the whole thing can be done analytically. For more complex scenarios and/or with semiconductors it will be a lot more complicated.

Hi Reza,

Sorry for the delay in response.

CHARMM-GUI has a nanotube modeller which allows producing CNTs with different chiralaties. This is basically based on a CHARMM-compatible INTERFACE FF. I have generated (10,10) and (6,6) CNTs to model adsorption of diverse organic compounds and qualitatively (and, semi-qyantitatively) the results agree with experiment.