How to calculate the bond force and bond length under different bias

[fangyongxinxi]

In this article http://prola.aps.org/abstract/PRB/v67/i19/e193104 (PRB 67, 193014, 2003) , the author did calculations using Siesta and got the Band force , Overlap population , and the Relaxed bond length under different bias .

We meet 2 questions when repeat this work by ATK .
1. How can we get the different band length under different bias ? In the article mentioned above , the author got the relaxed bond length and bond forces under different bias. I checked the ATK manual, didn’t find relevant command or ways to get such a result . Did the author used Siesta and  some other kind of software to accomplish this task  or just the ATK  could  bring this job to success ?
2.What is command CalculateAtomicForces used for ? Could we use it to calculate the atom force under different bias ?
I did some tests using the lih2li example in ATK. First , we did the Geo_Opt.
(The parameters :  froce tolerance is 0.05eV/Ang, Kpoint (1, 1 ,500 ) , basis set is DZP). Second , we set the Electrod Voltages (0, 0.1, 0.3 , 0.5 V) to calculate the relevant AtomicForces and Mulli_Popu and got the datas as below .
Bias (V)   Fz 1 (0.1eV/Ang)    Fz 2 (0.1eV/Ang)   Mulli population1    Mulli population2
0   -2.02               2.02                    1.0065   1.0065
0.1   -2.01               1.94                    1.0082   1.0045
0.3   -2.00               1.80                    1.0113   1.0011
0.5   -1.98               1.66                    1.0139   0.9983
The Fz 1 and Mulli_Popu1 are the force and Mulli_popu on H1 , the same to H2.  

Moderator edit: band->bond not to confuse anyone

[zh]

We meet 2 questions when repeat this work by ATK .
1. How can we get the different band length under different bias ? In the article mentioned above , the author got the relaxed bond length and bond forces under different bias. I checked the ATK manual, didn’t find relevant command or ways to get such a result . Did the author used Siesta and  some other kind of software to accomplish this task  or just the ATK  could  bring this job to success ?

What is the band length? It seems the bond length. Now ATK can not compute the force under bias. For  the last question, please contact one of the authors of your mentioned paper to ask the computational details.

2.What is command CalculateAtomicForces used for ? Could we use it to calculate the atom force under different bias ?
I did some tests using the lih2li example in ATK. First , we did the Geo_Opt.
(The parameters :  froce tolerance is 0.05eV/Ang, Kpoint (1, 1 ,500 ) , basis set is DZP). Second , we set the Electrod Voltages (0, 0.1, 0.3 , 0.5 V) to calculate the relevant AtomicForces and Mulli_Popu and got the datas as below .
Bias (V)   Fz 1 (0.1eV/Ang)    Fz 2 (0.1eV/Ang)   Mulli population1    Mulli population2
0   -2.02               2.02                    1.0065   1.0065
0.1   -2.01               1.94                    1.0082   1.0045
0.3   -2.00               1.80                    1.0113   1.0011
0.5   -1.98               1.66                    1.0139   0.9983
The Fz 1 and Mulli_Popu1 are the force and Mulli_popu on H1 , the same to H2. 

The "CalculateAtomicForces()" is used to compute the force exerting on atoms for the bulk or molecule systems. It can not be used to compute the force for a two-probe system under bias.  Please refer to the detailed explanation on the usage of "calculateOptimizedAtomicGeometry". Especially, please take a look at the geometry of a two-probe system:
http://quantumwise.com/documents/manuals/ATK-2008.10/ref.calculateoptimizedatomicgeometry.html

Indeed, the geometry optimization of a two-probe system is done by mapping the open two-probe system into a periodic atom configuration. One consequence of this is that two-probe optimization can only be done for an equilibrium two-probe calculation, i.e. a calculation done at zero bias.

[Anders Blom]

A small correction: you can actually compute the forces under bias. However it is correct that you cannot use the built-in functionality to run a geometry optimization for a two-probe system under bias.

It is however possible to do "manually", and I have recently written such a module, which thus allows relaxation under bias, and thus facilitates computation of the quantities sought in the original question (more or less). I will post it here later today.

The authors used TranSIESTA, which at that time could be considered the "academic pre-cursor" to ATK. ATK has similar but not identical capabilities as TranSIESTA; most notably ATK is not based on SIESTA, it just uses the same LCAO method for the electronic structure.

[Anders Blom]

I decided to publish the new module for relaxing a two-probe system in a new post. You can find it here: http://quantumwise.com/forum/index.php?topic=301.0

I hope it will enable you to extract the quantities you are interested in!