Local texture and fatigue crack initiation in a Ti-6Al-4V titanium alloy K. LE BIAVANT, S. POMMIER and C. PRIOUL Ecole Centrale Paris, Laboratory M.S.S.-Mat, CNRS UMR 8579, Chatenay Malabry, France Received in final form 27 September 2001 ABSTRACT Fatigue crack initiation was studied in a bimodal TA6V titanium alloy. A ghost structure inherited from the forging process, the scale of which is roughly 100 times the apparent grain size, was found to govern the initiation process. In these macrograins, that we have labelled macrozones, most of the primary alpha grains (a p ) are found to display the same crystallographic orientation. Fatigue cracks are initiated on the basal plane or, if basal slip is difficult, on the prismatic plane. Thus in macrozones, where basal or prismatic slip is easy, numerous neighbouring tiny cracks appear over the whole macrozone, which have the size of the primary a p grains. In these macrozones the contribution of crack coalescence to crack growth is consequently very significant. On the contrary, if basal and prismatic slips are both difficult in the macrozone, no crack can be found in the corresponding macrozone. The crack initiation process is thus highly heterogeneous at the scale of the macrozone. Furthermore, this microstructure is found to induce a large scatter in the fatigue life of notched samples. Keywords crack coalescence; fatigue crack initiation; microtexture; TA6V; titanium. NOMENCLATURE a  crack length da/dN  crack growth rate R  stress ratio a p  primary a-phase grains a s  secondary a-phase lamellae y  disorientation between the load axis and the direction normal to the slip plane l  disorientation between the load axis and the slip direction. g  disorientation between the slip direction and the direction normal to the surface ' 1 ,',' 2  Euler angles S  macroscopic nominal stress INTRODUCTION Fatigue crack initiation represents a major aspect in the design of industrial components. As a matter of fact it is usually found that the initiation of an engineering crack (the further extension of which can be predicted using linear elastic fracture mechanics), may consume more than 80% of the total fatigue life. This first stage of fatigue failure is very sensitive to defect distribution 1 microstructure, internal stresses 2 and local stress concen- tration 3 and is thus difficult to rationalize. Global tools, such as S±N curves determined on smooth samples, are therefore widely used to determine the life of compon- ents. However, it often appears that the data evaluated on smooth samples are not sufficient to predict the fatigue live of laboratory notched samples. Taking into account the stress field at the notch and using non-linear fracture mechanics to predict short crack growth at a notch may rationalize data obtained from smooth and notched samples. 4,5 If such a procedure is not sufficient, the discrepancy between smooth and notched specimen may reveal the heterogeneity of the defect distribution or of the material itself, at a scale comparable to the extent of the stress concentration of the notch. If that scale effect is not properly accounted for in notched samples, the prediction of the life of a real component from data collected on smooth samples is not reliable. Statistical methods 6 can help to solve this kind of ß 2002 Blackwell Science Ltd. Fatigue Fract Engng Mater Struct 25, 527±545 527 Correspondence: S. Pommier, Ecole Centrale Paris, Laboratory M.S.S.-Mat, CNRS UMR 8579, Chatenay Malabry 92290, France. E-mail: sylvie@mssmat.ecp.fr