QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: renren123123 on August 30, 2013, 15:28

Title: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: renren123123 on August 30, 2013, 15:28
 In our free molecule optimation process,  in output file, we can find such as content in each step calculation:Calculating Eigenvalues : =====
Calculating Density Matrix : =====
# WARNING
# The computed multigrid residual is greater than the required accuracy.    
# Computed residual :   1.39945e-010                    
#Required accuracy :   1.00000e-012              
how to solve this error? thank you very much！
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: renren123123 on August 31, 2013, 13:35
The computed multigrid residual is greater than the required accuracy. why? we use SZ or SZP, this error appears!
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Anders Blom on August 31, 2013, 23:30
This happened in older versions sometimes, when using Neumann condition at the edges with metallic regions, but should not appear in 12.8.2 at least.
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Kim_W on December 16, 2017, 13:57
how to solve it? I did transmission calculation  with 2017.2.
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Ulrik G. Vej-Hansen on December 16, 2017, 21:42
Usually, this is caused by the metallic regions being too close to the edges of the cell, or the atoms. You can try changing the positions a little and/or try one of the other Poisson solvers.
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Kim_W on December 18, 2017, 06:31

Calculating Density Matrix : ==================================================

################################################################################
# WARNING                                                                      #
#                                                                              #
# The computed multigrid residual is greater than the required accuracy.       #
#                                                                              #
# Computed residual :   1.13381e-12                                            #
# Required accuracy :   1.00000e-12                                            #
#                                                                              #
################################################################################

Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Kim_W on December 18, 2017, 06:32
In addition, the results are similar to the same configuration without electric field.   Is that SG15 not suitable for band gap calculation under electric field in ATK? Can you help me to modify it?
Title: Re: how to solve "computed multigrid residual is greater than the required accuracy"
Post by: Petr Khomyakov on December 19, 2017, 11:17
I do not think the use of the SG15 basis set has much to do with this issue, unless you have a particular explanation for that, so there is no need to modify the basis set.   

I do not understand why you do H-passivation of H-atoms for the very central region part of the device only, not the entire device. It might be that your system has no band gap because there exist many dangling bonds giving rise to multiple defect levels inside the gap.

The electrode extensions appear to be too short. That might be a reason for the warning you got in the log file.   

I would suggest first doing electronic structure calculations for a single layer to understand what the band structure of your system looks like without the external field applied, and how it depends on the passivation method. I guess all that also depends on the actual atomic structure of the system, i.e., whether it is in equilibrium or strained.