The observation of second-phase particles such as strengthening precipitates and nonmetallic inclusions on the fracture surface of steel specimens clearly indicates their importance in the overall ductile fracture process. Although voids can be initiated by dislocation or vacancy mechanisms, second-phase particles are the primary void initiators in practical engineering materials.

Reductions in volume fraction of second-phase particles, which can give smaller particles with no change in spacing or an increase in spacing with no change in size, or both simultaneously, have been shown in general to increase such things as fracture toughness, Klc, and true fracture strain, ef.

However, in wrought steels certain mechanical properties are very sensitive to shape and distribution of inclusions for a given volume fraction. For instance, fracture toughness, Klc, is the largest when the crack plane normal is in the longitudinal or primary rolling direction, intermediate when it is in the transverse direction, and smallest when it is in the short transverse or thickness direction.

The anisotropy in fracture toughness is large enough that, for instance in cylindrical pressure vessels, the steel plate’s longitudinal direction is oriented to support the larger hoop compared to the axial stress. Anisotropy decreases as cleanliness increases and thus provides an additional advantage to increased overall fracture toughness.

A metallographic examination of the three mutually perpendicular directions in a wrought steel, illustrates the nonuniform shape and distribution of inclusions and hence provides a qualitative explanation for the anisotropy in ductility and toughness. Events leading to such a distribution of Type II manganese sulfide (MnS) inclusions, the type commonly found in fully killed steels, and their results are being shown here.

The inclusions precipitate during solidification of the ingot in a "rod-shaped" morphology between dendrites. The resulting structure is referred to as interdendritic colonies of Type II MnS inclusions. Thus, even the original inclusion distribution is nonuniform. During subsequent hot rolling, the colonies are not only reoriented by rigid body motion so that their long dimension lies in the rolling direction in the rolling plane, but they are also plastically deformed.

Since the inclusion distribution and dimensions are so radically different in the three mutually-perpendicular directions, it is natural to expect the fracture surfaces from the three fracture plane orientations. Although different in appearance, all fractures are microscopically ductile (that is, fracture by microvoid coalescence).

Fracture surfaces from the three mutually perpendicular directions in a wrought steel plate are examined concurrently with micro-structure in an attempt to better understand void growth and anisotropy in fracture properties resulting from the shape and nonuniform distribution of inclusions. First, the microscopic aspects of void growth and coalescence in martensitic steel (ductile fracture) are discussed in general, then a description is given of ductile fracture in a wrought steel for the three mutually-perpendicular fracture plane orientations. From such observations, the inclusion distribution can be mathematically characterized and the anisotropy in fracture properties quantitatively appreciated. In many of the micrographs, the corresponding prior austenite grain size is illustrated for a better appreciation of relative size of fracture features with respect to important microstructural features.

A qualitative description of ductile fracture has already been provided by Cox and Low who studied plastic fracture (fracture under local plastic strain) in AISI 4340 and 18Ni maraging steel.

In AISI 4340 the process of plastic fracture was found to consist of void growth from inclusions first, followed by void growth from cementite (Fe3C) strengthening precipitates. Voids which form from strengthening precipitates in the highly constrained region between inclusion nucleated voids are believed to occur in sheets. In maraging steel the inclusion nucleated voids were found to grow until coalescence without being interrupted by void growth from strengthening precipitates. The strengthening precipitates in maraging steel, believed to be Ni3Mo, are several orders of magnitude smaller than the nonmetallic inclusions and do not initiate voids until they become much larger in size from, for instance, overaging.

This basic difference in the microscopic aspect of fracture is compatible with the observed higher fracture toughness of 18Ni maraging steel compared to AISI 4340 when both are heat treated to the same yield strength level.

It should be emphasized that although nearly all voids in practical engineering metals are nucleated by second-phase particles, all second-phase particles do not initiate voids. An illustration of this was the strain dependent "size effect". Another illustration, mentioned by Russ, concerns the "constraint relieving effect" on near-neighbor second-phase particles when a void forms and grows. For instance, consider that neighbor particles have equal probability of initiating a void. One will initiate a void first and, subsequently, cause a reduction of constraint on neighboring particles which, thereafter, may never initiate a void.

With only minor modifications, the principles of microscopically ductile fracture just discussed can be used to describe fracture in wrought steels which have inclusions that are anisotropic in shape and nonuniformly distributed.

It is possible, for grains away from the inclusions to slip first if they are more favorably oriented. Subsequent loading produces slip in grains between neighboring inclusions which ultimately results in a shear band.

After inclusion-nucleated void growth commences, whether it be matrix inclusion interface separation or inclusion fracture, strain is intensified in the shear band and, additionally, its width may be increased. Finally, small voids are initiated in the shear bands at the large Fe3C particles which themselves grow, coalesce, and produce macroscopic fracture. Experimental evidences of these events are present on ductile fracture surfaces as large inclusion-nucleated voids separated by smaller Fe3C-nucleated voids.

The stereo fractographs clearly illustrate that, because of inclusion orientation, long cylindrically-shaped voids initiate and grow from the inclusions and that a large amount of void growth occurs transverse to the loading direction.

From the foregoing descriptions of fracture for the three mutually-perpendicular fracture plane orientations, it becomes possible to describe a model which characterizes the inclusion distribution in wrought steel.

Such an observation is compatible with the low hardness steel being less "inclusion sensitive" than the higher hardness steel. Thus, in calculating fracture strain from its proposed functional relationship to volume fraction second-phase particles, it is necessary to use the effective volume fraction rather than the actual volume fraction base on random distribution of inclusions.

A more complete understanding of the effect of inclusion shape anisotropy on fracture strain must await a detailed mathematical analysis of multiaxial void growth in a noncontinuum.

The fractographic results of this investigation further illustrate that improvements in ductility and toughness at high strength levels will result from fewer and less severely deformed inclusions. Naturally, fewer inclusions imply improvements in steel-making practice, that is vacuum treatments such as vacuum induction melting, etc.

It has been known for some time that the amount of inclusion deformation depends on rolling temperature but the application of such knowledge has only recently been commercially available. Also, the addition of rare earth metal to the ladle gives rare earth sulfides, which are more difficult to deform upon rolling than the conventional Type II manganese sulfides. Thus, certain of the obvious methods of increasing toughness are now available.

Conclusions
Fracture surface examination of the three mutually-perpendicular fracture plane orientations resulted in the following observations which supplement existing information on microscopic aspects of fracture.

Void growth from inclusions occurred transverse to the loading direction. The extent of transverse void growth depended on fracture plane orientation but was observed to be as much as ten times the b0 dimension (that is, the smallest dimension) of the inclusion.

Of the hundreds of strengthening precipitates present per prior austenite grain, only a few participated in void nucleation and growth during fracture.

The nonuniform distribution of inclusions and the anisotropy in their shape were responsible for the differences in fractographic features on the three mutually-perpendicular fracture plane orientations.

The mathematical characterization of inclusion distribution illustrated that effective volume fraction of inclusions depended on fracture plane orientation and hence aided in the understanding of anisotropy in fracture strain.