The aim of the present paper is to present the results obtained in May 1994 at the French National Synchrotron Facility LURE on cold rolled ultra-low carbon steel. We observed on several lines at various 2* positions, a splitting of the diffraction lines into two peaks, one with a high intensity and a small width (peak 1) and another with much lower intensity but a larger width (peak 2). The angular distance between the two peaks is about 0.7 2$\theta$.
Three samples have been annealed and plastically strained : skin-pass 2.5%, skin-pass followed by tensile straining at 1.5% and 20%. A fourth sample has been cold-rolled at 45% reduction under standard factory conditions. An acquisition at high 2$\theta$ angle on {321} planes has been performed in order to determine the residual stresses on each peak and we also made acquisitions on 2 orders of {110} planes to make a Warren- Averbach analysis. The two peaks have been fitted simultaneously, peak 1 being modelized by a modified asymmetric Pseudo-Voigt function and peak 2 by a Gaussian function. We have shown that under our experimental conditions the instrumental broadening is mainly due to spectral dispersion. This enabled us to perform the Stokes correction. According to T.E.M. observations, theoretical modelling of plastic deformation, and to the fact that the ferritic matrix of our steel is composed of almost pure iron, we interpret our observations as the consequence of a bimodal dislocation structure. Peak 1 would then come from low dislocation density volumes and peak 2 from high density volumes. This assumption is supported by the Warren-Averbach analysis of peak 1. The average size of the coherent domains is very large (about 500 nm) and decrease slowly with plastic deformation. Except for small column length (< 100 nm), the elastic distortion is nearly constant ; its r.m.s. value increase from 15 10-6 to 24 10-6 with plastic deformation. On peak 2 the Warren-Averbach analysis was not possible but we observed that the peak width increase with plastic straining. Another interesting result is that the intensity of peak 2 increase more rapidly with the deformation than that of peak 1.
Our results support the following mechanism of plastic deformation : the macroscopic strain is mainly localized in the "soft phase", i.e. the low dislocation density part, but the dislocations produced move to the "hard phase" so that the elastic distortion in the soft phase remains low and the domain size hardly changes. The intensities and peak width evolution indicate that the stacking of dislocations in the hard phase consist of volume increase rather than density increase.
The experiments presented here were performed at Laboratoire pour l'Utilisation du Rayonnement Electromagn\'{e}tique (LURE) at Orsay, France, on Beam Line D23A managed by Michel BessiĞre.