The segregation of group IVB to VIB elements to prior austenite grain boundaries causes temper embrittlement phenomenon accompanies the intergranular fracture mode. The area fraction of intergranular facets on the fracture surface is a widely-accepted tool used to measure the embrittlement susceptibility. However, the occurrence of intergranular fracture is sensitive not only to the grain boundary composition, but also to the test temperature, etc. A fractographic survey of partially embrittled specimen show an inhomogeneous distribution of the intergranular facets on the fracture surface, and the fracture mode may vary systematically along the fracture path. In 1975, Viswanathan and Joshi showed that the local fracture mode changed as a function of distance from the notch root and the average fraction of intergranular fracture was the highest at a test temperature which depend on microstructure and matrix hardness in the 1CrMoV steel. The reason why the maximum amount of intergranular fracture occurred at other than the lowest test temperature was not well understood. In 1985, Jin Yu and McMahon, Jr. showed the experimental results of temper embrittled 2.25Cr-1Mo steel that intergranular fracture occurred most prominently in the region where cracking initiated, but that the fracture mode tended to change to cleavage as the cracking propagated and accelerated in the slow bending test of notched specimens at various temperature. Also, they noted that the area fraction of intergranular fracture had a maximum in plot as a function of test temperature and the temperature of this maximum tended to increase with specimen hardness. In above instances, fracture mode variation is important. The main idea of this experiment is that the variation of specimen geometry may affect the fracture mode variation. In this experiment, specimen of different shape (e.g. CT, DCB, constant-K specimen) is tested to measure critical stress intensity factor in liquid nitrogen (77K). In comparison with former works, each specimen in this experiment had a sharp crack, and the stress intensity factor of each specimen was able to calculate. Fractographic survey by SEM showed a different fracture mode in each specimen. The result of this experiment is the evidence that the geometry of specimen affect the fracture mode variation. On the basis of difference in fracture mode in this experimental results, fracture mechanics can be applied to analysis of fracture mode variation.