Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Ulrik G. Vej-Hansen

Pages: [1] 2 3 ... 28
Questions and Answers / Re: OptimizationDevice Geometry
« on: January 10, 2020, 08:49 »
This means that the Poisson solver did not converge. You could try changing it, f.ex. to the Parallel Conjugate Gradient solver or increasing the mesh cutoff. You should also check that your gate and dielectric setup (if present) is reasonable.

Questions and Answers / Re: Parallelization in ATK 13.8
« on: May 21, 2019, 08:54 »
The 13.8 version is no longer supported. I suggest you take a look at the newest P-2019.03 version and the many improvements that have been made since 13.8.

This happens because you optimize the structure after calculating the Dynamical matrix and before calculating the phonon bandstructure. So what you do is this:

1. Calculate the dynamical matrix for your structure
2. Optimize, i.e. change the structure
3. Calculate the phonon bandstructure based on your DM and configuration, which no longer match.

You need to optimize the structure first, otherwise the rest does not really make sense.

Questions and Answers / Re: NLValueError
« on: March 27, 2019, 10:22 »
You are right that you should make sure to use consistent exchange-correlation functionals, not LDA for one and GGA for the other. However, it  is only the indirect reason for the error you see. Because you change the exchange-correlation type, you will get different default basis sets because you have not explicitly defined them, and this mismatch causes the error.

The important lines in the tutorial scripts ( and are these: - lines 116-123
Code: [Select]
basis_set = [

# Exchange-Correlation
exchange_correlation = LSDA.PZ - lines 16-20

Code: [Select]
    calculator = device_configuration.calculator()
    calculator = calculator(

Note how the basis set is explicitly defined in the first, spin-polarized, calculation, which is then read in and used as the starting point for defining the noncollinear one. This ensures that the calculation parameters are consistent.

P.S. Please attach your scripts instead of including the full text in the post. In that way, it will be much easier to read for everybody.

Questions and Answers / Re: NLValueError
« on: March 26, 2019, 10:33 »
It seems that you are using different basis sets for the initial and the Spin-Orbit calculations. If you attach both of the scripts I can make a suggestion that should solve the issue.

Questions and Answers / Re: How to solve this question?
« on: March 21, 2019, 10:32 »
It looks like the computing node(s) had problems with the connection to the license server. The final error could be a result of those connection problems, or it could be running out of memory.

The Twister plugin is not shipped with the installer, but must be downloaded from the AddOn manager. It is contained in the FunWithGraphene AddOn.

If you do it like you show, without any modifications, it will be the unrelaxed configuration. Study objects are not designed to be used sequentially in the same script.

However, it is easy to use the relaxed configuration instead. Just insert this line between the two study objects:

device_configuration = optimize_device_configuration.result()

There is a typo in the line where the energies are defined, where there is a trailing comma, which should not be there. If you remove it, it works. I will make sure to fix the script.

Note that you should also change the file ending to hdf5, but QuantumATK will automatically do that in the current version.

You can use the built-in methods to extract the matrix elements, see the manual page for details:
Note that we do not recommend, or support, accessing the hdf5 files 3rd-party software.

I will look into the other issue.

Sounds like you need the method .inverseRelaxationTime(), which can be invoked in a script - see the manual page for more details:

Does it look like the properties you need?

It will depend on the details of your device, so it is hard to give advice without seeing the script.

However, one thing you can do is to make sure threading is enabled. It will allow you to re-gain some of the performance that is lost when you reduce the number of processes. Also, make sure you distribute the processes equally on the node, to ensure that you can utilize the available memory efficiently. How to do this in practice depends on the hardware and queuing system, and cannot be done from inside QuantumATK.

The port restrictions are mentioned elsewhere in the installation guide:

I believe they apply to both the primary and secondary ports.

You can probably solve the problem by the approach mentioned here:

If not, we will need more information, and as Petr wrote, technical and license issues are better handled using the support e-mail.

Pages: [1] 2 3 ... 28