Integrator issues

General
1

Hello there,

@Lidice and I have been coming across an error associated with the Integrator step size. Originally we believed it to be a problem with macOS, since Lidice can only get it to run on her linux machine. But I have had the same error appear on Linux, so it may be that Lidice has an older version of Quantics on Linux. Below is the error:


ERROR in subroutine Writestep (wrintegrat.f90):
Integration terminated. Integrator step too small :  0.00000001
------------------------------------

We’ve tried different integrators, different accuracies and changing the minstep tolerance in wrintegrat.f90. Has anyone come across this error before? or know of any fixes?

2

Looking at wrintegrat.f90, it seems your integration step is equal to the minimum step (minstep=1d-8).
I imagine it might come from the accuracy of the computer .

in wrintegrat.f90 try modifying that part:

! stepsize is exactly 0.0 in SIL diagonalisation
if ((abs(stepsize)<minstep).and.(abs(stepsize)>0).and.ll) then
routine = ‘Writestep (wrintegrat.f90)’
write(message,‘(a,f12.8)’) &
‘Integration terminated. Integrator step too small :’,stepsize
call errormsg
endif

by changing the format of the print command to ‘(a,f12.9)’ for example. Is the error message exactly?:
ERROR in subroutine Writestep (wrintegrat.f90):
Integration terminated. Integrator step too small : 0.000000010

or is it something lower ?

3

Hi Julien,

Yes, once making the change the exact error outputted is

ERROR in subroutine Writestep (wrintegrat.f90):
Integration terminated. Integrator step too small : 0.000000010
----------------------------------------
4

Hi Julien,

Yes, once making the change the exact error outputted is

ERROR in subroutine Writestep (wrintegrat.f90):
Integration terminated. Integrator step too small : 0.000000010
----------------------------------------

Actually I have also gotten smaller, 0.000000009 and 0.000000008 aswell