Minimizing molecules on Si(100)

[LUMOnosity]

Hello, novice ATK user here with hopefully a simple set of questions (I’ll emphasize I’m also novice at computational modeling in general, synthetic chemist really).  I’ve searched the postings and feel that nothing adequately addresses the issues I’m experiencing.  However, if I’ve missed something please accept my apologies and feel free to direct me to the appropriate posting.

Essentially I’m attempting to reproduce the work in Chemical Physics Letters 400 (2004) 347–352 [10.1016/j.cplett.2004.10.140] involving two organic molecules bonded to the Si(100)[2x1] surface.  My intention is to investigate similar systems with other organic structures.  I’ve had some luck reconstructing the structure in ATK and feel my representation is seemingly equivalent to that reported. 

However, when I attempt to minimize the structure my model converges to a significantly different conformation within the bonded molecules (the Si slab being fixed).  Curiously, during the minimization the confirmation seems to approach that in the literature, then move away.  Please see the attached images for comparison of final conformations.  To the best of my knowledge, I’m using as equivalent an exchange correlation as I can, that being local spin density and the VWN potential.  Please see my attached script.  I’m treating the system as a unit cell in ATK, but perhaps I should run the minimization as a discrete molecule instead?  Do I have an unreasonably small number of K points, I’ve tried sets in the range of 3,3,1 and 4,4,1 and 5,5,1? 

Understandably, my attempts at replicating the DOS and molecular orbital wavefunctions near the HOMO-LUMO gab have yielded significant variance to those in the report as well.  I would also like to get a better handle on this, but feel accurately minimizing the structure should be addressed first.

Any insight would be welcome.

[Umberto Martinez]

Indeed, it looks like you managed to reproduce the structure in your Ref.
Although, you need to increase the vacuum region around your structure (increase the lattice parameters with Bulk Tools>Lattice Parameters and center the configuration by Coordinate Tools>Center) or even better, convert to a MoleculeConfiguration.
You can easily do that clicking on the icon on the left side of the Builder.

In both cases you do not need k-points, (1,1,1)

The reason is that you are using a cluster approach to simulate a surface.
Common and good approximation in 2004. Now you can properly simulate a surface with a periodic approach if you wish.
See for example http://quantumwise.com/publications/tutorials/mini-tutorials/190

[LUMOnosity]

Umberto,

Your response is much appreciated and quite helpful.  I took your advice to convert the structure to a MolecularConfiguration.  I was able to minimize the structure using this criteria and the final conformation of the hexylnapthyl groups bonded to the Si(100) slab was much closer to that reported in the literature study I am trying to reproduce.

However, as you can see from the images I’ve included in this update, the final conformations of the bonded molecules still do not match those from the literature report.  While not being completely distorted as in my first attempt, they do adopt a torsional angle along the Z-axis between the molecules and the underlying Si(100) slab of about -10°.  This is in contrast to the literature report which depicts the molecules being rotationally aligned with the underlying Si slab

As a means to check if the rotation was induced by the napthyl groups I ran a minimization with just two hexyl groups and again observed a torsional rotation of about -10°, see attached images. 
Is this possibly a more accurate final structure or is there something biasing this rotation in my initial geometry? Any insight as to what is causing this discrepancy or rotation would be appreciated. I’ve also attached my script.

[Umberto Martinez]

you can have more accurate results by increasing the mesh cut-off, changing the basis set...

or, maybe your calculation is more accurate!

[LUMOnosity]

Hello

I am reposting to this thread after successful replication of literature results and extension of the method to new structures.  This success was a direct result of Umberto’s suggestion to carry out the calculations with a periodic approach as opposed to a cluster approach as a means to model the Si(100) crystalline surface. 

At this point I would like to further exploit this capability.  So far I have been able to minimize the structure and calculate the molecular energy spectrum, density of states, and eigenstates for a series of 1, 2, 3, and 4 molecules bonded to the Si(100) surface.  Going forward, I would like to model a continuum of bonded molecules in the same manner as used to model the crystalline slab.   However, when I generate a unit cell as seen in the attached image, the calculations yield results equivalent to a single molecule attached to an infinite underlying crystalline surface.  So I’m curious, does ATK understand this unit cell as a representing a single molecule attached to an infinite periodic plane, or does it understand it as a periodic extension of every element in the unit cell, i.e. both the crystalline slab and the bonded molecule are periodic in A and B dimensions?  What is the appropriate way to setup a calculation where both elements are periodic in A and B dimensions?

Any assistance is greatly appreciated.