This work is motivated by the industrial need to improve the thermal treatments of carbon steels used for the production of files. It aims at clearly defining the role that carbon plays in hardness and wear resistance of files. Three steel grades with different carbon content have been studied. The samples underwent various austenizing treatments, including treatments under carburizing atmosphere, followed by a quenching process leading to an essentially martensitic microstructure. This microstructure has been studied by mechanical spectroscopy. Several complementary techniques have been used : microscopy, thermopower, calorimetry, measurement of the microhardness and of the file performance. The damping spectra show three relaxation peaks, attributed to the interaction of interstitial atoms with dislocations in martensite. Two of these peaks, M2 and M4, are related with the presence of interstitial carbon. Peak M2 is related with the migration of geometrical kinks along screw dislocations, and peak M4 with the kink-pair formation and migration on screw dislocations. Both peaks are controlled by carbon dragging. This interpretation is supported by the measured activation enthalpies. The amplitude of peak M4 depends on the interstitial-carbon content, which apparently contradicts its physical interpretation. This dependence is attributed to the increase of the quench-induced dislocation density with interstitial-carbon content. The peak amplitude is in fact proportional to the dislocation density. The mechanical spectroscopy measurements are also sensitive to tempering effects. In particular, the precipitation of transition carbides at 380 K is associated with a decrease of the damping background, attributed to the depletion of martensite from interstitial carbon. The decomposition of retained austenite, occurring at 520 K or 670 K depending on the grade, is associated with a modulus anomaly, attributed to a sudden decrease of the dislocation density in martensite. A damping maximum coinciding with this transformation can also be observed at low frequency. This study lead to the use of new thermal treatments under carburizing atmosphere. The used gas mix gives rise to an efficient carbon enrichment of the samples. However, the increase of interstitial-carbon content may also lead to a higher content of retained austenite, thus reducing hardness. Calorimetry measurements allow the quantification of such content, which varies according to the steel grade and the thermal treatment. Martensite hardness and wear resistance of the files clearly increase with the interstitial-carbon concentration, that can be quantified by thermopower measurements, after correction of the effect of retained austenite. Long carburizing treatments can cause the formation of carbides on the sample surface. The presence of carbides in a high-carbon martensite does not seem to improve the mechanical performances, and can even be detrimental. The present study leads to the assessment of an optimal treatment duration of 30 min, which produces a hard martensite with maximum interstitial-carbon content and limits the amount of retained austenite and carbides. This carburizing time seems to produce the saturation of both the content of interstitial carbon at the steel surface and of the amplitude of damping peak M4. The amplitude of this peak seems to be an effective criterion for the evaluation of the ideal carburizing time. Indeed, the peak amplitude increases with the content of interstitial carbon and decreases with the content of retained austenite and carbides. The estimated optimal carburizing time is compatible with production demands. A cryogenic treatment proves to lower efficiently the amount of retained austenite.