The answer lies in the name - GNDs are geometrically *necessary*, i.e. until the geometry of crystal is restored to a 'perfect' state, they must exist, else the crystal will become discontinuous.

The 'imperfect' state, relevant to GND, comes from bending of planes (such that some planes are stretched (tension) and some are compressed - important in single crystals) or from mismatch between crystals at interfaces (in polycrystals). So unless this imperfection is removed, the GNDs will survive. Annealing won't 'un-bend' planes (ignoring shape memory effects), and only in theory will it completely remove grain boundaries. So annealing and recovery processes primarily affect the SSDs, not the GNDs.

A possibly useful reference is below. You will find there that SSDs are treated using hardening and recovery terms (formation and annihilation), while GNDs are treated only with formation terms.

https://royalsocietypublishing.org/doi/pdf/10.1098/rspa.2007.0020?origin=publication_detail

Note: An unrelated but maybe helpful thing to remember is that plastic deformation is an irreversible process - the deformation is made possible (in terms of continuity) by GNDs -if we could annihilate every GND by annealing, the irreversibility would be lost.

I think that annihilation of SSDs takes place earlier and "more drastically" . My statement is based on the results of TEM observations. However, it is worth mentioning that the effect of crystallographic orientation and deformation is important. The misorientation of grain boundaries also has an important role in the creation and stability of dislocation structure (GND/SSD).

Greetings.

M. Wrobel

M. Wrobel thanks for response. Could you give me a source about your answer?

I think that any kind of dislocation that is generated sooner, annihilates easier. So, SSDs can be exhausted earlier during the annealing process.

Regards;

Amirsalar