Messages

1

General Questions and Answers / Re: Can we create MTP with dopant atom included?

« on: January 17, 2025, 23:21 »

Step 1 would be to make an MTP for GeTe. This should be relatively straightforward following the standard crystalline training protocol.
Step 2 is to use active learning to extend this MTP to a-GeTe
In step 3 you can then add C to either or both several GeTe crystal supercells and/or selected snapshots from the a-GeTe MD trajectory, and retrain on these. If you need to consider different kinds of C positions (interstitial, substitutional etc) make sure to cover all possibilities.
Finally, I would do a step 4 where you repeat the active learning for a-GeTe with C included.

Now, I am aware that the tutorials for MTP on the website are outdated. We are working on new ones which should be published very soon.

2

General Questions and Answers / Re: Question of : interface builder -> surface cell -> Some matches are degenerate?

« on: January 17, 2025, 23:16 »

Can you attach the structures, then I can better try to assist. Note that I don't think there is a problem or bug here, you just need to figure out a way to pick the most relevant structure.

3

General Questions and Answers / Re: How to extract effective potential and J from IVCharacteristics block?

« on: January 17, 2025, 23:14 »

Yes, the IVCharacteristics is a complex thing... Try to just print the full "file_data" to figure out better what's in the file, but also I realized that maybe the file you are looking at isn't actually the relevant one - the individual analysis output from the different biases etc are saved in separate files, isn't it?

4

General Questions and Answers / Re: Log file won't update

« on: January 17, 2025, 23:11 »

I think you can rename the log file to something.txt and then it can be attached. It has a few specific extensions that it allows only.

If it finished the optimization, then only reason it might fail would be that for some reason it cannot "nlsave" at the end.  You use the same filename for the MD trajectory and optimization trajectory as for saving the final result, this is generally not a good idea; I always make sure the MD and optimization trajectory files are separate files from the main results, as they can grow very large. So maybe that is already the issue, if this file has grown so large that it's slow to operate with.

You may not even need these trajectories if the only target here is to generate the melted/quenched structure.

5

General Questions and Answers / Re: Question of : interface builder -> surface cell -> Some matches are degenerate?

« on: January 16, 2025, 03:12 »

I guess we do things slightly differently, but I don't see this message. I took Si and Mo crystals from the Database in the Builder, cleaved 110 and 001, and dropped on the Interface Builder, see below.
I use W-2024.09.

6

General Questions and Answers / Re: difference between optimization using device LCAO and LCAO for bulk

« on: January 16, 2025, 03:07 »

In terms of results both methods should be similar, but using bulk relaxation before you convert to device is faster. What you need to do, in order to still  be able to convert the structure to a device, is to keep the atoms that will not relax fully fixed. This needs to be enough layers that the periodicity of the structure is still maintained. So, if there are 3 layers in the electrode (before relaxation) you need to keep 4 layers fixed, so that the algorithms notices that layers 1 and 4 are identical.

7

General Questions and Answers / Re: How to extract effective potential and J from IVCharacteristics block?

« on: January 16, 2025, 02:56 »

The IVCharacteristics is not an object saved to the HDF5 file, it's a framework for computing lots of individual sets of configurations and later analyze them, but each configuration, transmission spectrum, potential, etc are saved as individual objects in the HDF5 file. Therefore, you can also read them easily using the nlread command, but it can be somewhat tricky to identify which one is which...

If you use

Code

nlread("file.hdf5")

it will return ALL objects in the file, in a list. You can narrow this down by a particular class, so

Code

nlread("file.hdf5", DeviceConfiguration)

will return a list of all configurations. Finally, you can ask for a specific configuration etc, by using

Code

nlread("file.hdf5", object_id="name of object")[-1]

Note that nlread ALWAYS returns a list, even if only one object is found. This is why you need [-1] also when requesting a particular object. You can of course also use index=0 but this forum formats bracket-0-bracket funnily... It also smashes all text into one line when I use code blocks, making it hard to read, sorry.

The easiest way know the object name is to look in the GUI - you see it in your screenshots in the "name" column. Programmatically you can ask a file which object IDs it has using

Code

from NL.IO.NLSaveUtilities import nlinspect
file_data = nlinspect("Builder_Stash.hdf5")
for i in file_data:
    print(i.object_id)

You can in this way extract the converged state of a configuration at a specific bias (gate and source/drain) and compute additional quantities without ever having to redo the DFT+NEGF part. It is also possible to do additional analysis via the IVCharacteristics object itself, using the addAnalysis method (https://docs.quantumatk.com/manual/Types/IVCharacteristics/IVCharacteristics.html#NL.Study.IVCharacteristics.IVCharacteristics.addAnalysis).

It's hard to tell what went wrong with the file, hopefully it's not corrupt. Maybe try reading it from the command Python environment instead, or restart the NanoLab GUI to see if that helps.

8

General Questions and Answers / Re: IV stop

« on: January 16, 2025, 02:43 »

If you are using the IVCharacteristics method, the script is restartable as-is. That is, just rerun the script, without modifications, and it will skip all tasks already done.

9

General Questions and Answers / Re: Eigenvalues and eigenstates

« on: January 10, 2025, 08:26 »

The first step is probably to think about what the purpose of creating the plot it, what data are you looking to extract. Maybe an isosurface is not the best choice?

In any case, what you are plotting is basically a 4-dimensional dataset, a function f(x,y,z). Since we have no real way to doing a 4D plot, we need to use various 3D projections, and the isosurface is one such option. But all isovalues are equally relevant or "correct", so we cannot pick one for you. An isosurface is this the best option for the plot if you are interested in knowing where in space the quantity takes certain values.

It might help to consider the 2D analogy of height curves on a map.. Instead of seeing the mountain as a real 3D surface, you plot isolines that connect points in the geography that lie on the same altitude. By looking at that plot you can trace the contours, which for instance help you pick a suitable walking path, or at least tell you which parts of the trail will be steep or flat. But it would not be helpful if someone just picked one single isoline and didn't show the other ones.

The isosurface is similar, but in one higher dimension which unfortunately means we can only look at one curve (surface) = isovalue at the time. An alternative might be to use a point could plot instead, where color is used as a 4th dimension to indicate where the function is high or low. In that case you can plot all data at once, but it might also be hard to see what goes on in the middle of the system. Finally, you can also use cut-planes to plot the data in a specific plane, and then move this plane along the structure.

10

General Questions and Answers / Re: Question of : interface builder -> surface cell -> Some matches are degenerate?

« on: January 8, 2025, 20:58 »

This is a rather new feature and I haven't yet encountered a case where it showed up. Would be helpful with some details about the configurations to reproduce and advise.

11

Scripts, Tutorials and Applications / Re: ElectricFieldConstraint support for BaTiO3 structure

« on: January 8, 2025, 20:51 »

First of all, I would make sure you have a working setup for GGA, before trying this extremely time-consuming calculation with HSE. Also, use the update strategy "FixElectricFieldCorrection" to avoid recalculating the Born charges each time, before you have a working methodology. This is just to make sure the principle of the calculation works. Now, note that 10000 V/cm is actually not a large field, it corresponds to 1 mV applied over 1 nm (typical unit cell size, roughly), so that might explain what you both observed.

12

General Questions and Answers / Re: Regarding the problem of converting the structure from atkpython to cif

« on: January 8, 2025, 20:42 »

What you need is either

Code: python

from NL.QuantumATK.ConfigurationConverters.CIFConverters import configurationToCIF
with open("myfile.cif", 'w') as f:
    f.write(configurationToCIF(configuration))

or

Code: python

from AddOns.VASPPlugins.IO import writePOSCAR
writePOSCAR(configuration, "POSCAR")

NOTE: In case someone wants to use this export function for more advanced crystallography purposes, bear in mind that the QuantumATK CIF exporter is rather primitive and can only save structure as P1, so it will not use the correct spacegroup symmetries to reduce the atom list to the irreducible basis.

13

General Questions and Answers / Re: Nothing appear on the I-V curve

« on: January 8, 2025, 20:31 »

Impossible to advise without more info. Did the calculation actually finish successfully? What data did it produce in the HDF5 file?

14

General Questions and Answers / Re: Missing add-on (sai export)

« on: January 8, 2025, 20:25 »

Sorry, this tutorial is 8 years old, and is very much outdated and should be removed. Also the Sentaurus tool "ATD" that would read the sai file no longer exists.

If you describe in more detail what you are trying to achieve (beyond reproducing the tutorial) by using an atomistic band structure in TCAD, we can probably suggest other methods to couple these tools.

15

General Questions and Answers / Re: Request for Assistance: Converting MoS2 Structure into Onion Shape

« on: January 8, 2025, 20:17 »

See https://forum.quantumatk.com/index.php?topic=12074.msg40759, maybe it helps, but  the TubeWrapper doesn't make a sphere, just a tube. I must confess it seems highly unlikely that a spherical ball of MoS2 would be stable at all... For C60 we need to combine 5-rings and 6-rings, and still that corresponds to graphene. Adding the extra layers of Sulphur atoms seems almost impossible. If you are thinking about structures like in https://pubs.acs.org/doi/10.1021/acs.jpcc.7b07784, it seems to me these are nanoparticles rather than fullerene-like structures, i.e. you would just make a large supercell of MoS2 bulk crystal and cut it with a sphere (remove all atoms outside a certain distance from the center). It might then self-organize into the onion-like structure upon thermal annealing with MD.