Geometry optimization 3AGNR repeated 9 times in the z direction

[Luis M. Villamagua C.]

Hello all,

I am training myself on Graphen nanoribbon. So my first goal was to optimize a 3-AGNR repeated 9 times. I kindly ask you to check if what I have done until now is Ok. You can see the pictures inside the attached document because I still do not know how to insert them within the text.

FIRST STEP: Create a basic 3-AGNR (this structure is not repeated in the z direction)

I created this nanoribbon with the help of VNL following tutorials on youtube. See figure 1.

SECOND STEP: Find the optimal k-point mesh for the 3AGNR

I found the optimal k-point mesh for the nanoribbon. In figure 2 you can see how the energy converges to -1008.39972 eV for 1x1x18 mesh configuration. I have chosen 18 even though the system seems (at least visually) to converge at 9. From 18 onward the energy does not change at all, it remains the same even to the 5th decimal point. AM I CORRECT?

THIRD STEP: optimizing the basic 3-AGN

Again, following tutorials on youtube I set ATK as shown in figure 3 for the Geometry optimization. I did not change much but the k-point configuration to 1x1x18 in the new calculator tool; the rest was left as default (ANY ADVISE HERE?). In the geometry optimization tool I only changed:

Maximum Force to 0.005 eV/A
Maximum stress to 0.005 eV/A
Maximum step size to 0.1 A

I understand that this will improve the precision of the optimization (ANY DRAWBACK ABOUT IT) The rest was left as default.

FOURTH STEP: Create an extended 3-AGNR (the basic 3-AGNR is repeated 9 times in the z direction)

 I re-inserted the just optimized 3-AGNR into the Builder tool of VNL and asked it to repeat this structure 9 times (Figure 4).

FIFTH STEP: Optimize the extended 3-AGNR

I repeated the same procedure as in THIRD STEP, only changing the k-point configuration to 1x1x2. I read that I calculate the k-points only once and then is only matter of dividing them. What I did was 18/9 = 2. However, because I am very curious about this new topic I desided to check againd the k-points configuration. I find out that it did not converge (to the 5th decimal point as for the basic 3-AGNR) at 2 but 3. CAN YOU PLEASE EXPLAIN WHY IT HAPPENS?

I want to repeat  this procedure (if everything is ok) for a longer and wider nanoribbon in which I can introduce SW defect and study different properties. I will very appreciate your suggestions or comments.


Regards,
Luis M. Vilamagua

[Shinji Usui]

Dear Luis M. Villamagua C.

I think your procedure of the optimization is perfect.

However, you will get a complete　planar structure when you repeat the optimized one unit -C6H4- group.
Note that if two phenyl rings connect by single bond, the two plains will not　lie in the same plane,
because of the repulsion between neighbor Hydrogen atoms.
This is the reason you could not get the optimized structure in the fifth step.

Try to use biphenyl structure -C6H4-C6H4- (two unit cells) in the first step, and you will get a twisted ring
configuration by relaxation.   And then repeat the structure as the method you showed.


BR
Shinji
  

[Luis M. Villamagua C.]

Dear Shinji,

Thanks for you promp reply. This was an ultra-narrow graphene nanorribon (not relizable experimentally until now). I set the width to 3 atoms because I could run calculations from seconds to minutes. I will be working however with much wider nanoribbons like 16-AGNR, 30-AGNR or 40-AGNR. In these type of NR the hydrogen atoms are not so close between each other, hence the NR will not experience the twisting phenomena, right?. So my question is .....

Even for wider NR (like 16-AGNR) should I start with C6H4-C6H4 (as you are suggesting me)?

As an electronic engineer I am worried because we do not study this kind of theory and will be very difficult for me to accomplish this (at least not in a short term). If I create the NR with VNL is this still OK, I mean, would I still have good and accurate results?

Regards,
Luis

[Luis M. Villamagua C.]

Can you please clarify one more doubt?

Repeat a nanoribbon in the y and z directions with VNL
From youtube I learned to expand 8-AGNR 1nm to a 8-AGNR 10nm long (just an example). I mean repeating the basic structure in the z direction. Is it possible to repeat a nanoribbon in the y and z directions, for example if I have a 3-AGNR 1nm long, how would I do to get a 8-GNR 10nm long?

If this is possible, then I can optimize a very small GNR and then repeat it in the y z directions, right? Or is it always mandatory to create a nanoribbon with a specific width, optimize it and then repeat it in the z direction.

Thanks,
Luis M. Villamagua

[Shinji Usui]

Dear Shinji,

Thanks for you promp reply. This was an ultra-narrow graphene nanorribon (not relizable experimentally until now). I set the width to 3 atoms because I could run calculations from seconds to minutes. I will be working however with much wider nanoribbons like 16-AGNR, 30-AGNR or 40-AGNR. In these type of NR the hydrogen atoms are not so close between each other, hence the NR will not experience the twisting phenomena, right?. So my question is .....

Even for wider NR (like 16-AGNR) should I start with C6H4-C6H4 (as you are suggesting me)?

As an electronic engineer I am worried because we do not study this kind of theory and will be very difficult for me to accomplish this (at least not in a short term). If I create the NR with VNL is this still OK, I mean, would I still have good and accurate results?

Regards,
Luis

If you want to calculate wider GNR like 16-AGNR, you do not have to start with two units cell system.
The final structure of wider GNRs will have complete plane structure.

(Ultra narrow 3-AGNR is called polybenzene and we can synthesize it)

BR
Shinji

[Luis M. Villamagua C.]

Thank you so much Sir.

[Luis M. Villamagua C.]

Sir,

What constrain cell under geometry optimization widget is for? Can you please explain what will be the advantages or drawbacks of constraining or non constraining x, y and z when relaxing the lattice.

In your last replays you said that the optimization process I carried out was good. Can I create a device from this optimized structure. I mean ...

(1) I create a unit cell (i.e 6ZGNR) => Optimize it
(2) The just optimized cell is repeated  10 times in the z direction => Optimize it
(3) Insert the SW defect inside the optimized network => Optimize it
(4) Lastly create a device configuration out of the last optimized structure.


An extra question ...

Can I find the k-point grid with huckel model and then carry out the geometry optimization with DFT???


Regards,
Luis

[Umberto Martinez]

You can release the constrain only along z at step (1). Along x and y directions you have a vacuum region meaning no stress along these directions.

At step (2) you have the same structure which is already optimized.

at step (3) you may have stress along z if your nanoribbon is too short as the SW defect will influence the geometry of the surrounding.
However, you should have a nanoribbon which is long enough such that you defect is isolated (if this is what you want).
Therefore you do not need to release the z constrain here.

[Anders Blom]

About k-points: yes, that should work.

[Luis M. Villamagua C.]

So the steps to follow will be:

(1) create a unit cell (i.e 6ZGNR) => optimize it (THE Z CONSTRAIN IS RELEASED HERE)
(2) The just optimized cell is repeated 10 times in the z direction => (NO optimization is needed here)
(3) Insert the SW defect and relax again with or without constraining the z direction (this depends on the length of the ribbon)
(4) create the bulk.


In this tutorial (http://www.youtube.com/watch?v=t3xJNqP4X04) at minute 3:19 Nanna says ...

"It is most convenient optimize the ribbon as a device configuration because the periodicity is kept fixed automatically. If you optimize the ribbon as a bulk system, you must insure this manually"

Question 1
If I carry out the relaxation as I have been doing until now, must I ensure the periodicity manually?
Question 2
Isn't it more convenient to create the structure as in the tutorial and then carry out just one final relaxation with DFT? What will be the disadvantages of doing so??

Thanks again
Luis

[Umberto Martinez]

Of course Nanna is right.
Although for a single SW defects in the center of a 10times long nanoribbon both solutions are fine (optimize bulk conf. and create device - create device and optimize device conf.)

Note that in the electrode extension regions of a device configuration the positions of the atoms are kept fixed and they match the atom positions of the electrodes.
http://www.quantumwise.com/documents/tutorials/latest/ATKTutorialDevice/index.html/chap.input.html#sect1.input.device

In this respect, if you break the periodicity in a region close to the walls of the cell during the optimization of the bulk configuration, the "Device from Bulk" plugin will not recognize the periodicity that you need to create the electrodes.
Hence, you will have to do it manually.
(Actually, it may recognize the periodicity of a "distorted" structure.)
So, "If you optimize the ribbon as a bulk system, you MAY HAVE to insure this manually" always applies.
This should answer to your second question.

About the first question the answer is: no, if the plugin managed to do it automatically.

It is hard to give general and universal answers, you may need to adapt from case to case.
But I hope everything is more clear now, is it?

[Anders Blom]

Excellent answer, Umberto.

About the possible disadvantage of the device method, it's simply a heavier calculation so it takes longer time.

[Luis M. Villamagua C.]

Very very clear now  dear Humberto

About k-points: yes, it worked!

I computed the k-point grid for two different nanoribbons (bulk configuration), getting:

For 8ZGNR (12 times repeated  in the z direction) => with DFT (1x1x5) with Huckel (1x1x5)
For 16ZGNR, (12 times repeated in the z direction) => with DFT (1x1x6) with Huckel (1x1x7)

With Huckel took me a couple of minutes while with DFT took hours

When relaxing both geometries I used 1x1x9 in order to ensure a perfect sampling onto the BZ; I hope that this will not poor the accuracy of my calculations.

By the way, (correct me if I am wrong, please) the k-points should be found only once at the beginning. I mean, one must find the k-points of an specific nanoribbons (by changing the kc point and seeing where the energy converges) and then use this grid until the end of my project, right? For example if I want to insert a defect (SW defect in my case), should I still use the same k-point grid?

Regards,
Luis M. Villamagua

[Umberto Martinez]

Correct, as long as you do not chance the dimension of your device along the C direction you can keep the same k-points sampling in your project.

[Anders Blom]

The C kpoints are actually a bit special in a device calculation. They are only used in the electrode, which is normally a very small part of the total calculation time, so you might as well put a bit extra k-points here, to make it more accurate. But more importantly, there is a risk of having too few C k-points, as explained at http://quantumwise.com/publications/tutorials/mini-tutorials/214.