Lower bainite (temperature range 400-250°C) appears more acicular than upper bainite, with more clearly defined individual plates adopting a lenticular habit. Viewed on a single surface they misleadingly suggest an acicular morphology.

However, two-surface optical microscopy of lower bainite indicates that the ferrite plates are much broader than in upper bainite, and closer in morphology to martensite plates. While these plates nucleate at austenitic grain boundaries, there is also much nucleation within the grains, i.e. intragranular nucleation, and secondary plates form from primary plates away from the grain boundaries.

Electron microscopy shows that the plates have a similar lath substructure to upper bainite, with the ferrite subunits about 0.5 μm wide and slightly disoriented from each other. The plates possess a higher dislocation density than upper bainite, but not as dense as in martensites of similar composition.

The crystallography of the plates seems to depend both on the temperature of transformation, and on the carbon content of the steel. Moreover they showed that the phenomenological theory of martensite could be used for lower bainite to give satisfactory agreement between theory and experiment.

Ohmori and coworkers have found that, in a 0.1% C steel, bainite formed near the Ms has a {011}α habit plane and a <111>α growth direction similar in behavior to low carbon lath martensite. However, on increasing the carbon to 0.6-0.8% C, the habit of the bainitic ferrite plates changes to {122}α//{496}γ, which is not the same as for martensite of the same composition which has a {225}γ habit plane. Because of such variations, it has been suggested that lower bainite is not a true martensitic reaction. However, there is no reason to expect the transformations to be identical, and anyway the inhomogeneous shear would be expected to occur by slip in lower bainite, whereas twinning is the mode adopted in higher carbon martensites.

However, in contrast to tempered martensite the cementite particles in lower bainite exhibit only one variant of the orientation relationship, such that they form parallel arrays at about 60° to the axis of the bainite plate. This feature of the precipitate suggests strongly that it has not precipitated within plates supersaturated with respect to carbon, but that it has nucleated at the γ/α interface and grown as the interface has moved forward. It thus appears that the lower bainite reaction is basically an interface-controlled process leading to cementite precipitation, which then decreases the carbon content of the austenite and enhances the driving force for further transformation.