Going back to the original beam of Figure 1, which has stretching or tensile stresses along the upper surface, any time the stress exceeds the strength, a crack can open. The crack now makes the loading even more uneven. The stresses no longer vary gradually from tensile to compressive as you go from a top to bottom surface layer. The maximum tensile stress is now experienced at the crack tip, which may be anywhere in the thickness. The original top surface layers now have a near zero stress. But the crack tip location does not experience the same level of stress as the material at the originally intact surface did. The crack tip location has a much higher stress condition due to concentration of the stresses. Most of the shape change (strain) associated with the stress is forced into the small area at the crack tip instead of being spread out along the top surface. The atoms along the crack tip are “really stressed out!”
There is another major reason that cracks can change direction. Instead of a block of cheese, imagine in your hands a pack of cards. If you look closely while bending or flexing the pack in your hands, the edges of the cards at the ends perpendicular to the length of the bend are now forming tiny steps. This is because the cards have slid a little bit with respect to each other due to the bending. The solid part can’t slide as easily as the cards, but the same stresses are acting to make this happen in the solid parts as well. Shear stress is the technical name for this. When an intact beam is loaded in bending, the shear stresses are parallel to the length of the beam, just like in the pack of cards. Because of the restraint (steps can’t form in a solid block) the shear stresses are highest near the center of the beam thickness. Occasionally, I have seen a crack make a “hard left” turn, usually in the center half of the “height” (or thickness) of the beam, due to shear stresses “taking over.” Cracks due to shear stresses in bending are at 90 degrees to the cracks created by the tensile stresses.
Figure 4 shows a round bar that was notched with a band saw to make it easier to break at a later time. The saw cut simulates the separation due to an initial macro-brittle crack due to tensile stresses. The actual crack is seen to be at an angle to the length of the round “beam.” When the stress is high enough, the remaining ligament separates due to shear stresses from compressive or tensile loading. This effect is often called a “compression hinge.”