secdec is hosted by Hepforge, IPPP Durham

Frequently Asked Questions (SecDec version 3)

For questions concerning pySecDec, please go to the pySecDec documentation page.

Tutorials: (SecDec-3)

Answers to common questions for SecDec-3:

  • "The sector decomposition seems to take forever"
    • You might have encountered a case where the heuristic sector decomposition algorithm X runs into an infinite recursion. Please choose one of the other strategies, e.g. G1 or G2.
  • I get a segmentation fault at the Mathematica stage. What happened?
      In case a segmentation fault occurs at the Mathematica stage and the {graph}Decomposition.log file ends without the statement "Time taken to do the decomposition: xxx secs", an infinite recursion might have occurred during the iterated sector decomposition. This can happen as soon as one of your functions which needs decomposition is higher than linear in several Feynman parameters.
      Please choose a different decomposition strategy.

  • How do I know if the result I get is correct?
    • For that you can check the errorprob (Probability that the uncertainty stated in the output file is erroneous) written to the result files *integrationfilenumber*x*epsorder*.out in the folder of the corresponding pole structure. For error probabilities close to or zero you can trust the correctness of the result. If the value is bigger, there might have been some convergence problems and it is advisable to redo the integration with different integration parameters (key, border, mindeviation and maxchisq, etc). For more info see the CUBA documentation to be found online and in {YourSecDecFolder}/Cuba-3.0/cuba.pdf.
    • You can also check the summed result files *point*results*epsorder.log. If you observe large cancellations among the individual results you may consider running the integration again, either with a larger grouping value or to a higher desired precision (decrease epsrel and epsabs respectively).

last updated: Sep 20, 2017