I belive first of all there should be approppriate loading and temperature conditions as well as the small grain size. Of course you know that there is a rerlationship between strain rate and grain size during GBS. Therefore grains should be free of any dislocations. However, because it is a post-mortem investigation then some of them may still contain some. Also dynamic rrecrystallization may be interfere with the grain structure. I was observing some time ago something like stratified nanocrystalline sized grain structures. It looked like a quasi-fluid flow when small grains formed sublayers inside the large layer flowing between two plates moving at different velocities (Couette flow). 

In addition, grain dislocations form steps at grain boundaries (GB) to which vacancies migrate (cavitation). Cavities merge during the GB sliding process and finally result in micro-cracks. Thus there are normally no dislocations anymore at the interface with the GB. But this depends also of microstructural features resulting from heat treatment, precipitation of carbides, dynamic recrystallization etc. Also a zone of denuded γ’ / Ni3 (Al, Ti) phase can be formed adjacent to the GB providing barrier to dislocation motion.

I appreciate your valuable comments, Sergei and Guibert. I found dislocations stacking inside grains and denuding/absorbing into grain boundary, i.e., less dislocation density around the grain boundary as compared to the grain interior.

Dear Namhyun,

Could you check the causes of dislocations pile-ups inside the grains (solute atmosphere, ...) ? 

It was actually springs stress-relaxed at 500C for 300h. I think that dislocations probably pile-up for drawing process of springs and stress-relaxation. This is the process how the springs produced: Drawing  =>  Aging at 620-730C for 16h  =>  Stress relaxation at 500C for 300h.