The imaginary part of two-dimensional Fourier-transform spectra in the rephasing and nonrephasing modes is used to analyze the homogeneous and inhomogeneous broadening of excitonic resonances in semiconductor nanostructures. Microscopic calculations that include heavy-and light-hole excitons as well as coherent biexcitonic many-body correlations reveal distinct differences between the rephasing and nonrephasing spectra. A procedure is proposed that allows separation of disorder-induced broadening in complex systems that show several coupled resonances.In the past decades, various optical techniques have been used to investigate and unravel the structure of electronic states in semiconductor nanostructures and other material systems. 1-5 Spatially resolved linear optical measurements give information about homogeneous and inhomogeneous broadening separately. Typically, however, they provide only general information, e.g., the total linewidth. On the other hand, nonlinear experiments have been applied successfully to obtain much more detailed information about the nature of excited states, the coupling among them, and many-body effects. In addition, different nonlinear optical techniques were used to investigate the amounts of homogeneous and inhomogeneous broadening ͑see Ref. 5 and references therein͒.Pump-probe measurements provide one-dimensional spectral information that cannot distinguish between homogeneous and inhomogeneous broadening. Hole burning can find the homogeneous contribution to the optical linewidth and, by comparing to the linear spectrum, provides an estimate of the inhomogeneous contribution. Four-wave-mixing ͑FWM͒ experiments show photon echoes in the timeresolved ͑TR͒ traces. 5,6 Their temporal width is determined by the inhomogeneous linewidth. However, for systems where more than a single resonance is simultaneously excited, the width of the echo is ill defined due to beating, 5,7,8 in particular, for small inhomogeneous broadening see ͑Fig.1͒. The time-integrated ͑TI͒ trace yields the homogeneous width, i.e., the dephasing rate. However, in the presence of more than just a single optical resonance, the decay parameter cannot uniquely be determined and a fitting procedure is needed.In semiconductor nanostructures, many-body Coulomb interaction strongly alters the nonlinear optical response. [3][4][5]9,10 Even at the Hartree-Fock level, e.g., the line shape of time-resolved FWM is significantly modified and signals for the wrong time ordering appear. [3][4][5]11,12 Additionally, already in the low-intensity third-order ͓ ͑3͒ ͔ limit, characteristic dependencies of the nonlinear transients and spectra on the polarization directions of the incident pulses and couplings among optically isolated resonances appear due to many-body correlations. [3][4][5]9,[13][14][15][16][17] A detailed microscopic description of interacting excitons in the presence of disorder is a formidable task. Thus, wellestablished knowledge is lacking on this topic. It was, however, shown that Hartree-Fock renormalizations...