fitting method for high-order force constants

[rfhari]

Hi @ttadano, I'm interested in calculating second- and third-order force constants. From the documentation, I understand that ALAMODE uses the ordinary least squares method (in the ALM module) to calculate these force constants. However, I was wondering if ALAMODE fits all the interatomic force constants (IFCs) simultaneously, or if it fits them iteratively based on the residuals, as shown in Equations 26–28 of the TDEP documentation. I couldn't find an explicit mention of this in the ALAMODE documentation. Could you please clarify this?

Thanks!

PS: Based on the sensing matrix of Eq 3 of this paper on CSLD, I think, unlike TDEP, Alamode fits is it simultaneously. But I'm not really sure.

[ttadano]

Both options are available. If you do not specify any options, all force constants will be fitted simultaneously. If you set the FC2XML option in the &optimize field, the second-order IFCs are fixed to the values in the given file, and only the anharmonic IFCs are fitted.

[ejmeitz]

When using the FC2XML flag I get an error. I've attached the input files and the log file from the run. Based on where the log-file ends I think the program is trying to read the xml file I have. I created the XML file using the ALM python library. A couple places stoi could be called but really hard to tell. Happy to debug this and open an issue, but I'm not sure I have the tools.

terminate called after throwing an instance of 'std::invalid_argument'
  what():  stoi

EDIT: Building the alm executable inside of ALM instead of ALAMODE I now get a more useful error:

Error in get_value_from_xml
The following entry could not be found in the XML file : Data.ForceConstants.HarmonicUnique.NFC2

Pretty sure this happens because I am manually loading in my own IFC's instead of calculating them with ALM first. So I dont end up with the unique IFCs I just have entries for all the possible IFCs for the primitive cell. Is there any way to generate the FCXML file in a way that will match the expected format of alm? Lookin at the source code here the XML key (HarmonicUnique) is not written by the fcsxml.py but is expected by the xml parser in ALM/ALAMODE (here)?

lj_alm.zip