broken bonds in carbon nanotube when optimization

[perfetti]

Hi all!
I am optimizing a carbon nanotube clamped with 20 copper atoms in it. However, in the process of the optimization, the carbon nanotube has been broken and copper has been embedded into the carbon nanotube.
I am not sure this situation is reasonable, since the carbon-carbon sp2 bond is said to be stronger than that of diamond, so why the copper atoms can break the c-c bond? Is the carbon nanotube generated in this software give consideration of the physical properties of real carbon nanotube? If so,what mechanism is underlying? Hope anyone could clarify that for me. Discussion is also welcome.

The relaxation trajectory is attatched. Thanks.

[Anders Blom]

Method used? Input file would help.

[Shinji Usui]

Please check the covalent radius of Cu and C.  
Approximate values of them are
Cu = 1.32 Ang.
C = 0.73 Ang (sp2)
by Wiki.
So the initial bond length of the Cu-C for the optimization calculation should be around 1.32 + 0.73 =~ 2.0 Ang.

Looking your initial configuration for the optimization some Cu-C length are too short ( around 1.46 Ang)
and too much forces could be applied to the system and you found the C-C bond breaking.

You should make more relaxed structure as the initial configuration for the optimization.

[perfetti]

Dr. Blom,
      I used DFT. And the input file is attatched.

Method used? Input file would help.

[perfetti]

Hi Mr. Usui, thanks for your answer.
I am not sure about the initial distance. I did use 1.5 Angstrom as the initial one, but after relaxation, for device with less copper atoms, it always arrive at the value of around 2 Angstrom.
I am not sure if that's OK for these device results.
I mean, since it will always go through optimization, finally it would arrive at the equilibrium bond length.
Am I understanding it right?

Secondly, what do you mean by "You should make more relaxed structure as the initial configuration for the optimization"?

Please check the covalent radius of Cu and C.  
Approximate values of them are
Cu = 1.32 Ang.
C = 0.73 Ang (sp2)
by Wiki.
So the initial bond length of the Cu-C for the optimization calculation should be around 1.32 + 0.73 =~ 2.0 Ang.

Looking your initial configuration for the optimization some Cu-C length are too short ( around 1.46 Ang)
and too much forces could be applied to the system and you found the C-C bond breaking.

You should make more relaxed structure as the initial configuration for the optimization.

[Anders Blom]

I don't see any principal problem with the calculation. Most likely it's simply not a physical situation to have the Cu atoms that close to each other inside the nanotube. The trajectory clearly shows that in the very first steps the Cu atoms experience very large forces, as I interpret it due to the repulsion between Cu-Cu, and so they move outwards which in turn creates large forces on the C atoms, and then it all breaks. So Shinji is correct, the initial structure is the problem, it's not realistic.

[perfetti]

Thanks Dr. Blom.

Since the initial structure is not physical realistic, how could I justify it? From the result?

I did some research based the same initial distance yielding results, would that mean my research was not correct?

I don't see any principal problem with the calculation. Most likely it's simply not a physical situation to have the Cu atoms that close to each other inside the nanotube. The trajectory clearly shows that in the very first steps the Cu atoms experience very large forces, as I interpret it due to the repulsion between Cu-Cu, and so they move outwards which in turn creates large forces on the C atoms, and then it all breaks. So Shinji is correct, the initial structure is the problem, it's not realistic.

[Anders Blom]

That's a bit too big a question to answer for me, so I will not.

But as a general comment for any study, I would say that unless it's performed with the atoms in the equilibrium position, then one should be able to explain why not. And if the equilibrium positions cannot even be determined since the optimization does not work, well, then it's a bit of a problem for sure. I mean, you can dream up any system, but if it's realistic or relevant is a very different story.

[Shinji Usui]

If you are assuming Cu-C bonds as covalent ones the optimized bond length 2.0 Ang is reasonable.

My idea to relax the initial structure :
The Cu tube radius of your structure is too big, I mean the strain between CNT and Cu are too tight.
Making a smaller radius of Cu tube, you can construct more relaxed structure as the starting point of the optimization.

And also to enlarge the CNT radius is an alternative idea to release the strain in the system.

Hi Mr. Usui, thanks for your answer.
I am not sure about the initial distance. I did use 1.5 Angstrom as the initial one, but after relaxation, for device with less copper atoms, it always arrive at the value of around 2 Angstrom.
I am not sure if that's OK for these device results.
I mean, since it will always go through optimization, finally it would arrive at the equilibrium bond length.
Am I understanding it right?

Secondly, what do you mean by "You should make more relaxed structure as the initial configuration for the optimization"?

Please check the covalent radius of Cu and C.  
Approximate values of them are
Cu = 1.32 Ang.
C = 0.73 Ang (sp2)
by Wiki.
So the initial bond length of the Cu-C for the optimization calculation should be around 1.32 + 0.73 =~ 2.0 Ang.

Looking your initial configuration for the optimization some Cu-C length are too short ( around 1.46 Ang)
and too much forces could be applied to the system and you found the C-C bond breaking.

You should make more relaxed structure as the initial configuration for the optimization.

[perfetti]

Thanks Dr. Blom.
That's good enough.

That's a bit too big a question to answer for me, so I will not.

But as a general comment for any study, I would say that unless it's performed with the atoms in the equilibrium position, then one should be able to explain why not. And if the equilibrium positions cannot even be determined since the optimization does not work, well, then it's a bit of a problem for sure. I mean, you can dream up any system, but if it's realistic or relevant is a very different story.

[perfetti]

Thanks Mr. Usui.

If you are assuming Cu-C bonds as covalent ones the optimized bond length 2.0 Ang is reasonable.

My idea to relax the initial structure :
The Cu tube radius of your structure is too big, I mean the strain between CNT and Cu are too tight.
Making a smaller radius of Cu tube, you can construct more relaxed structure as the starting point of the optimization.

And also to enlarge the CNT radius is an alternative idea to release the strain in the system.

Hi Mr. Usui, thanks for your answer.
I am not sure about the initial distance. I did use 1.5 Angstrom as the initial one, but after relaxation, for device with less copper atoms, it always arrive at the value of around 2 Angstrom.
I am not sure if that's OK for these device results.
I mean, since it will always go through optimization, finally it would arrive at the equilibrium bond length.
Am I understanding it right?

Secondly, what do you mean by "You should make more relaxed structure as the initial configuration for the optimization"?

Please check the covalent radius of Cu and C.  
Approximate values of them are
Cu = 1.32 Ang.
C = 0.73 Ang (sp2)
by Wiki.
So the initial bond length of the Cu-C for the optimization calculation should be around 1.32 + 0.73 =~ 2.0 Ang.

Looking your initial configuration for the optimization some Cu-C length are too short ( around 1.46 Ang)
and too much forces could be applied to the system and you found the C-C bond breaking.

You should make more relaxed structure as the initial configuration for the optimization.

[perfetti]

Dear Dr. Blom,
        I have one more question about the C-C bond. As it shows that it can be broken at a certain setup, but I am not sure if the C-C bond reflected the real C-C sp2 bond length? Since it could be very strong to be broken. How did you generate the carbon nanotube, did you take the real strength properties into consideration? Thank you.
         Best regards.

       
           

Method used? Input file would help.

[Anders Blom]

The nanotube builder in VNL uses a graphene sheet and rolls it up. You can choose the bond length - the default is the "usual" one for graphene (and it's at least close to the expected on for CNTs too). The "bond strength" is not really a relevant parameter in this construction, since this is the required energy to break the bond. I think in your case you simply effectively have a very large built-in energy because you squeeze the Cu atoms together and put them close to the C atoms, and then let them go. The initial structure is simply not physical, or even close to physical, apparently (if it were, the optimization would just adjust the atoms a bit, and not make the structure explode like this).

[Nordland]

Whenever I encounter a problem with a relaxation, I always tries to see it from a MD point of view, to get an idea what the problem might be.

To assist your in your case, I did a similar calculation trying to estimate forces felt by the carbon nanotube.
I downloaded your trajectory and imported the starting configuration, and delete all the carbon atoms. I did this because I want to understand how much stress
the copper atoms were under in this initial configuration. I setup a MD configuration with no initial velocity, and made it run for 5000 steps. I used a classical potential, but the analysis would have been the same for DFT. After 5000 steps I would be able to calculate approximate temperature which corresponds to the potential energy being locked up in such a confined copper nanowire.

The estimated temperature of the copper wire is 2.6 million Kelvin, and regardless how strong the carbon-carbon binding is, I doubt it can sustain a 2.6 million degrees hot copper core.

[perfetti]

Thanks Dr. Nordland.
Thanks Dr. Blom.