[Roler]

Here is my paradox, I'm struggling with my molecular simulation because of this, here it is:

Look at the molecule H2O(water)

Hydrogens van der waals radius is 120 pm
The O-H bond length is 96 pm

That means that if you use the van der waals radius of hydrogen to represent the atom as a sphere, the sphere of both the hydrogens would go through the center of the oxygen sphere, and the hydrogen spheres would therefor overlap.
By definition every atom that overlaps their van der waals radius are bound together.
Therefore my conclusion:The hydrogen atoms in water are bound to each other.

Prove why this is wrong logic.

[pm133]

I'll have a go at this.

The hydrogen atoms are not behaving as spheres in the water molecule.

The moment an oxygen appears in the vicinity the electron density of the hydrogen undergoes an anisotropic transformation with electron density being pulled into the oxygen atom for electronegativity reasons as seen by the presence of the dipole moment.

At this point the electron density which was part of the hydrogen is now belongs to the oxygen and when a second hydrogen comes along, it'll bond with that oxygen electron density.

That would be my explanation.

[pm133]

It's also fair to say that in molecules, there is a fair bit delocalisation of electron density across the molecule anyway and the idea of atoms "owning" particular electron density starts becoming a bit hazy.
Canonical molecular orbital calculations will usually show contributions from many atoms within each molecular orbital.
I think confusion on this comes from the fact that ideal atomic models are being applied to real world molecular environments and the fit is not always a good one.

[Roler]

For my simulation I would need a table over linear transformations I should apply to the spheres volume, when they bind.

I do not want to calculate force field, since that makes it too computationally expensive.

Where can I find a table over how much of its volume a hydrogen looses when it binds oxygen etc( that is, I need the new  van der waals radius).
I don't want to look at the molecule as a whole, I want the atoms to be separate entities.

Thank you for the answers.

[Irlanur]

Usually in MD simulations you don't look at vdW interactions between atoms of around 1-3 bonds apart.

[Roler]

From wikipedia I found this:

Non-uniform scaling (anisotropic scaling) is a non linear transformation.

That backs up what you say, the hydrogen is no longer a perfect sphere. But lets say I'll keep it as a perfect sphere, because it not that far from the truth. Would that be possible?

So in the way I am thinking, should I just increase the sphere of the oxygen and decrease the hydrogen sphere a little(That is the van der waals radius for the spheres), would that be an approximation you would approve ?

Thanks.

[pm133]

I shall back out at this stage because I don't have experience in MD simulations but I would say that the OH bond will not look anything like two overlapping spheres. It would look like a guitar plectrum with most density on the O atom.
In ab initio we add polarisation to capture this anisotropy but no idea what MD needs to do. I guess the validity of any approximation will depend on the property you are trying to calculate and how accurate you need it.

[Irlanur]

What you want is unphysical. That's ok, MD is unphysical, too. But in order to represent the real world by any means I think there is no other solution than calibrating your model (in a way that MD-ists calibrate their force-fields.). Is there any literature about your approach? I know only about more or less "classical" MD.

also:

Hydrogens van der waals radius is 120 pm

where is that value from? that's more than the typical C-H bond length...

[Roler]

The 120 pm is standard, look here:

https://en.wikipedia.org/wiki/Van_der_Waals_radius

top right corner it gives some values.

Okey, so they would look more like a guitar than a double sphere with different radii, I have not seen any literature with this approach, because force field is the s#*$ in MD these days, I just want to see if I can make a much more computer effective method, that can get "Approximate results" for my simulation faster. 

[Irlanur]

Yeah sorry. of course this all makes no sense. vdW radii make only sense between non-bonded atoms. otherwise your bonds would all be way too long.

[Roler]

Okey, my conclusion:

If I want to keep all atoms in a molecule as sphere, then all the sphere need to decrease in size to have a new vdW radii equal the covalent radii ?

I can make it more accurate by:

if I am going to be able to simulate the atoms by them self but not as sphere(rather the real shape), I need to know the vdW sphere of the molecule and put the form (guitar etc) to each atom in the molecule, so that they create the complete form of the molecule in the end.

Therefor I am gonna look up in a table how the form should be, and let this form oscillate to simulate atomic vibrations.

Thank you, I think I understand now, add comments if you have any

[Irlanur]

Can I maybe ask what exactly you want to simulate and what properties you want to extract? It might help quite a bit. There are more pitfalls than you might expect, and there is no need to repeat all the mistakes that have been done in the past. (although this is the path where you learn the most....)

[Roler]

I just want it to act as physical correct as possible, so i can put in molecules and they will behave as in the real world, I dont want any output(except the graphics of course), just watch how things happens on the small scales