Bug running relaxed star in Minkowski metric

[msha0023]

Code fails when trying to evolve the star in isolation after the relaxation procedure.

Steps followed :-

~/phantom/scripts/writemake.sh grstar > Makefile (in Minkowski)
make
make setup
./phantomsetup star.setup
./phantom star.in

Error

WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 2
 WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 759
 WARNING! cons2primsolver [in step_extern_sph_gr (b)]: enthalpy did not converge on particle 2
 WARNING! cons2primsolver [in step_extern_sph_gr (b)]: enthalpy did not converge on particle 759
 WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 9
 WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 764
 WARNING! cons2primsolver [in step_extern_sph_gr (b)]: enthalpy did not converge on particle 9
 (buffering remaining warnings... enthalpy did not converge on particle 764) 
 WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 252
 WARNING! cons2primsolver [in step_extern_sph_gr (a)]: enthalpy did not converge on particle 501

 FATAL ERROR! maketree: NaN in particle position, likely caused by NaN in force on particle 2: x =        NaN

[msha0023]

Update: This issue does not happen when SYSTEM=gfortran is used for the compiler.

[Shigaharuki3012]

Update: This issue does not happen when SYSTEM=gfortran is used for the compiler.

Hi,I am also a user of PHANTOM.
When I used grstar setup,a different error from yours appeard even I have changed the compiler from ifort to gfortran.So have you successfully run the whole simulation after using gfortran compiler?
(I couldn't even pass the relax-star-process before I run the whole simulation(That is,for me ,tde.in).

[msha0023]

Update: This issue does not happen when SYSTEM=gfortran is used for the compiler.

Hi,I am also a user of PHANTOM. When I used grstar setup,a different error from yours appeard even I have changed the compiler from ifort to gfortran.So have you successfully run the whole simulation after using gfortran compiler? (I couldn't even pass the relax-star-process before I run the whole simulation(That is,for me ,tde.in).

Hi, I can get the simulations to work for both compilers. The way to run grstar is to use-> ~/phantom/scripts/writemake.sh grstar>Makefile, followed by make setup and make. Then use ./phantomsetup star, select the model and you can then relax a star. The file to use for tde.in is the star_00000.tmp. If you want to use GR in your TDE sims, then use make setup GR=yes METRIC=kerr and make moddump GR=yes METRIC=kerr. Then you can run ./phantommodump star_00000.tmp tde 0.0 . This will write tde.tdeparams and then tde.in which you can use for the run.

[Shigaharuki3012]

./phantomsetup star

Thank you for your reply.In fact when I use grstar setup all kinds of errors appear,even I had followed the steps.
But when I manually add GR=yes in the config.F90 in the ~/phantom/main folder then everything goes well but the units are in chaos(For example,c=686 but we need G=c=1 in GR=yes case)
So did you change anything in phantom?In fact these errors have bothered me for a long time...(Maybe it's because I made some mistakes in the command line...)
My commands were:

cd tde
 ~/phantom/scripts/writemake.sh grstar > Makefile
make 
make setup
./phantomsetup star     #Then I use polytrope setup and automatically-relax-star.Others are default
./phantomsetup star .setup

when in the last step(relax star),error appeared like :

ERROR! get_u0 in rho2dens: 1/sqrt(-v_mu v^mu) ---> non-negative: v_mu v^mu
 ERROR! get_u0 in rho2dens: 1/sqrt(-v_mu v^mu) ---> non-negative: v_mu v^mu
 ERROR! get_u0 in prim2cons: 1/sqrt(-v_mu v^mu) ---> non-negative: v_mu v^mu
 pmom =                     NaN                     NaN                     NaN
  pmom =                     NaN                     NaN                     NaN
  pmom =                     NaN                     NaN                     NaN
  rho* =   1.13353353642459
  en   =                     NaN
 FATAL ERROR! cons2prim: could not solve rootfinding on particle 6277
  rho* =   1.82053973411935
  en   =                     NaN

More details of some other error are shown in the discusion with Mr Price in #402.

[danieljprice]

I think some of this error is because we had ireconav = 1 by default in the .in file, while using reconstruction on the shock viscosity in GR is a very experimental feature.

Hence I have switched ireconav=-1 in #413 which should resolve the remaining issues here

Note also that since #413 it is no longer necessary to use grstar and moddump, one can simply use SETUP=grtde as a one-step relax-and-go procedure