Shape function parameters to fit non-periodic boundary conditions

[M. Zubair]

Hi Bachir,

I would like to set non-periodic boundary condition to simulate spherical nanoparticles, for that I know we do not mention any boundary condition in PDB file or main run.py file but "shape function must be there. Could you please elaborate a bit about setting "Shape function parameters" ?

Thanks.

Regards,

Muhammad Zubair

[Bachir Aoun]

Hi,
You are right, you should set boundary conditions to None.

The shape function computes and compensate for the loss of density due to the infinite boundary conditions. Mathematically speaking, a double fourier transform is computed to approximate the system's geometry ... default parameters should be good enough unless your particle is huge then you might need to play with rmax parameter ...
When you perform your stimulation, plot the constraint shape function along with your constraint, it should look decent with no obvious wiggles and your constraint should have a horizontal asymptomatic behavior at high q or R which is indicative of good parameters. Also you may look into the provided SiOx nanoparticle simulation provided in the examples ...

Hope this helps

[M. Zubair]

Hi Bachir,

Thanks for your last reply.

Can you comment on the attached image? I have plotted shape function constraint along with PDF constraint. It seems like my shape function parameters are not good enough. However, your insight into this will be highly appreciated. 

Thanks.

Regards,

Muhammad Zubair

[Bachir Aoun]

hi,

short hand answer, your shape function looks good to me. 

Keep in mind that a bad shape function will fail to straighten the calculated constraint at high R value and it is going to show high frequency wiggles that indicate poor sampling of the space.

With this being said it's always hard to accurately judge the quality of the calculated shape function. remember that the shape function fullrmc computes is very similar to the small angle pattern you get at doing small angle measurement. those big bumps that you see are most likely true and due to the shape of your nanoparticle.

i hate to influence your science but if you want my opinion don't sweat it too much your shape function seems good. 

if you want to have a feeling of this shape function, i recommend you playing with the parameters calculating the constraint data and plotting, no need to run the stochastic engine just compute constraint data ... you can also create different nanoparticle pdb files and repeat this exercise and you can plot the shape function in a log scale too and i bet that you will see a small angle diffraction pattern like especially if you are in Q space  ...

hope this helps

Bachir

[M. Zubair]

Hi Bachir,

Thanks a lot for your prompt response. It was elaborative enough to understand what is going on with shape function and how to check that.

Thanks for your time.

Regards,

Muhammad Zubair

[Andy Anker]

Hi Bachir,

I am in the same situation, where I have to validate my shape function. My solution has been to do a simulation of different parameters, to visually inspect the fit, shape function and error. AlsoI have tried to keep the parameters such that the Qmin and Rmin are kept low, Qmax and Rmax kept high and dr/dq is low aswell. Does this sound as a reasonable strategy?

Now I think, I have a reasonable shape function and it could be interesting to compare it with small-angle X-ray scattering data. Is it possible to do a comparison of those? In principle they should be the same, but I need to do a Fourier transformation of the shape function, or how?

Best,

Andy

[Bachir Aoun]

Hi Andy,

I would be more careful when trying to compare the shape function with small angle diffraction. There is a relationship between both entities but in all honesty i don't remember from the top of my head anymore.

if g(r) is the computed constraint before shape function subtraction and G(r) after subtraction then

G(r) = g(r) - 4*PI*r*N*T(r) where T(r) is the wide angle correlation function

if you want to read more about this topic i recommend you browsing this nice review https://pubs.acs.org/doi/abs/10.1021/acs.chemrev.5b00690

as far as choosing the shape function parameters i would use the default in fullrmc and let it figure it out but if you think it's not doing a good job this is the intuition behind the parameters

• rmax: it must be bigger than the biggest distance in your atomic model. normally i would take more than the double of that maximum distance value (shannon sampling)
• rmin: i would set it to 0 or close to 0
• dr: you got to experiment because a too small or too big value will give you wiggles. knowing that no atoms are closer than 0.5A fullrmc considers this threshold
• qmin and qmax: the default parameters in fullrmc should cover a big range of particle sizes unless your system is millions of atoms huge. 

I am not confident I should tell you what to chose this is up to you to look at the result and judge. Visual inspection is all what you need. At the end of the day you need to remember one very important thing, your experimental data is flattened visually using non physical shape functions. Fullrmc does things in more theoretical manner but this is being compared to experimentalist approach if you know what i mean.

don't sweat it, focus on analyzing the system after the simulation is done and testing different groups and moves upon those to get different configuration solutions...

best

  

 

[Andy Anker]

Hi Bachir,

Thanks for a really thorough answer!

Kind regards

Andy