[Teresa Sun]

Error message

tmp<N4Foam14GeometricFieldINS_6TensorIdEENS_12fvPatchFieldENS_7volMeshEEE> deallocated

    From const T& Foam::tmp<T>::cref() const [with T = Foam::GeometricField<Foam::Tensor<double>, Foam::fvPatchField, Foam::volMesh>]
    in file /cvmfs/hpc.rug.nl/versions/2023.01/rocky8/x86_64/amd/zen3/software/OpenFOAM/v2206-foss-2022a/OpenFOAM-v2206/src/OpenFOAM/lnInclude/tmpI.H at line 217.

FOAM aborting

#0  Foam::error::printStack(Foam::Ostream&) at ??:?
#1  Foam::error::simpleExit(int, bool) at ??:?
#2  Foam::tmp<Foam::GeometricField<Foam::Tensor<double>, Foam::fvPatchField, Foam::volMesh> >::cref() const at ??:?
#3  Foam::tmp<Foam::GeometricField<Foam::innerProduct<Foam::Vector<double>, Foam::Tensor<double> >::type, Foam::fvPatchField, Foam::volMesh> > Foam::operator&<Foam::Vector<double>, Foam::Tensor<double>, Foam::fvPatchField, Foam::volMesh>(Foam::dimensioned<Foam::Vector<double> > const&, Foam::tmp<Foam::GeometricField<Foam::Tensor<double>, Foam::fvPatchField, Foam::volMesh> > const&) at ??:?
#4  Foam::LESModels::scalarQR<Foam::IncompressibleTurbulenceModel<Foam::transportModel> >::k(Foam::tmp<Foam::GeometricField<Foam::Tensor<double>, Foam::fvPatchField, Foam::volMesh> > const&) const at ??:?
#5  Foam::LESModels::scalarQR<Foam::IncompressibleTurbulenceModel<Foam::transportModel> >::correctNut() at ??:?
#6  ? at ??:?
#7  __libc_start_main in /lib64/libc.so.6
#8  ? at ??:?
Aborted (core dumped)

Solution
In .C file, change the red part of the code

template<class BasicTurbulenceModel>		
tmp<volScalarField> scalarQR<BasicTurbulenceModel>::k
(			
    const tmp<volTensorField>& gradU	
) const
{                                             
    volSymmTensorField  D(symm(gradU));       
    const volVectorField& U = this->U_;
    const auto& rhok = U.mesh().lookupObject<volScalarField>("rhok");
    volScalarField	buoyancy = 1/4 * gradU & fvc::grad(rhok) & g_;

}

to

    volScalarField	buoyancy = 1/4 * fvc::grad(U) & fvc::grad(rhok) & g_;

It's trivial matters, I post it to remind myself. It would be nicer if it helps someone else

[olesen]

If the first line were changed to have
volSymmTensorField D(symm(gradU())); then you wouldn't lose the temporary there and could leave the buoyancy term as-is without recalculating the gradient (again)

[Teresa Sun]

Hi Mark,

Thanks for your advice and explanations. It worked and I guess it saves computational cost since the gradient is not recalculated again. But in the solver information, the deltaT is almost the same as the recalculated version. Anyway, your methods seem more reasonable.

Thanks again for the suggestion.
Best regards
Teresa

[olesen]

It is also generally preferable to construct variables directly, not by assignment (=) since this can be ambiguous when creating a field from a tmp type.

// not so good
volScalarField buoyance = 1/4 * ...;

 // better
volScalarField buoyance
 (
    1/4 * ...
);

gcc won't care, but clang will and when you start trying to share code and other people can't get it to compile...

[Teresa Sun]

Hi Mark,

Thanks for your suggestion. It's very helpful. Indeed we plan to share the code when it's ready.
Thank you very much.

Best,
Teresa