The temperature dependence of the frequencies and linewidths of the Raman-active longitudinal optical (LO) phonons in GaAs and AlAs have been measured. The low-temperature lifetime of the LO phonon in AlAs is found to be 9.7 ps, very close to the corresponding GaAs value of 9.5 ps. This contradicts early theoretical predictions. The agreement between theory and experiment can be restored when the accidental degeneracy between the AlAs LO phonon frequency and a feature in the two-phonon density of states is taken into account. DOI: 10.1103/PhysRevLett.88.215502 PACS numbers: 63.20.Kr, 78.30.Fs The lifetime of the Raman-active longitudinal optical (LO) phonon in III-V semiconductor compounds is of critical importance for the coupled dynamics of carriers and phonons in these materials. Accordingly, an intense effort has been devoted to the study of this quantity for the past 20 years [1]. The recent availability of ab initio predictions of phonon lifetimes [2-5] represents a significant breakthrough, since they pinpoint the relevant LO phonon decay channels that may open the door for the engineering of phonon lifetimes.In spite of the technological relevance of the Al-Ga-As system, the lifetime t of LO phonons in AlAs has not been reported thus far. From time-domain measurements on an Al 0.3 Ga 0.7 As alloy sample, Tsen and Morkoç find t 7 6 1 ps (at 10 K) for both the GaAs-like and AlAs-like LO vibrations [6]. These results are quite puzzling since, as pointed out by Kash and Tsang [1], the phonon dispersion curves for GaAs and AlAs (which determine their possible LO decay channels) are quite different. A comparison of these curves for several semiconductors suggests that t AlAs should be significantly longer than t GaAs . The first ab initio estimates of phonon lifetimes added to the puzzle [3], since they predict t AlAs 12.7 ps . t GaAs 9.2 ps, in agreement with Kash and Tsang's qualitative argument. The reliability of the theoretical prediction was underscored by the good agreement with the existing experimental data for GaAs [7][8][9][10][11].The above discrepancies between theory and experiment could be dismissed by arguing that the GaAs-like and AlAs-like LO vibrations in AlGaAs alloys are not equivalent to the LO phonons in binary GaAs and AlAs crystals, so that a direct comparison is not legitimate. In this Letter, however, we report Raman measurements of the LO phonon frequencies and lifetimes in GaAs and AlAs. Our results show that the low-temperature lifetimes of LO phonons in GaAs and AlAs are indeed very similar. But we also find that these experimental results do not imply a failure of the ab initio method. Instead, they arise from a breakdown -in the case of AlAs -of the standard formalism used to compare theory with experiment. This breakdown is caused by the accidental degeneracy between the frequency of the AlAs LO phonon and a Van Hove singularity in the two-phonon density of states. Such degeneracies have been proposed in the past -most notably in the case of the transverse-optic phonon in ...