Decreasing grain size has an advantage of both increasing the strength and improving the Impact toughness. Increase in strength is evident from Hall Patch Equation but what is reason for lowering of the transition temperature
Decrease in grain size results in increase in both yield strength and the fracture strength which results in decrease in transition temperature. While Hall - Petch relation is commonly known, second not so reported relation independently derived by Cottrell and Petch provides the fracture stress to be proportional to the ratio of surface energy by its square root of grain size for a given material. Thus decrease in grain size leads to decrease in transition temperature. Physically the grain size effect on fracture stress may be attributed to the extent of meanderings suffered by the growing crack. Smaller grain size implies larger meanderings, which requires larger stresses owing to the deflection of the crack plane from the favorable directions.
In low carbon steels, although the grain size has some influence on impact transition temperature, it is the thickness of the grain boundary carbides that exert the greatest effect. Fracture occurs in the Charpy test when a crack in a carbide is able to break-out into an adjacent ferrite grain. A critical carbide thickness is necessary for this process to be energetically favourable. For most heat treatments, the grain size and grain boundary carbide thickness are related - heat treatments that produce a fine grain size give thin carbides. Therefore, fine grain steels usually have a low impact transition temperature. If a fine grained steel was produced with thick grain boundary carbides its transition temperature would be much higher.
if due to any reason the grain boundary is weaker - segregation or presence of carbides that provides the easier crack propagation, then under the framework of cotrell - petch theory the influence of smaller grain size on transition temperature will be suppressed. This is because the surface energy term will dominate to reduce the fracture stress.
A very good publication to explain the initial question is this:
Ductile–Brittle Transition Temperature of Ultrafine Ferrite/Cementite Microstructure in a Low Carbon Steel Controlled by Effective Grain Size, T. HANAMURA, F. YIN and K. NAGAI, ISIJ International, Vol. 44 (2004), No. 3, pp. 610–617
This paper is a very good read and explaines in terms of the Yoffee diagram, how DBTT is decreased with decreasing effctive grain size (or unit crack path).
If somebody can not access this publication and is interested in more about this, please contact me.