Purpose: The average energy expended by an energetic electron to create an ion pair in dry air, W air , is a key quantity in radiation dosimetry. Although W air is well established for electron energies up to about 3 MeV, there is limited data for higher energies. The measurements by Domen and Lamperti [Med. Phys. 3, 294-301 (1976)] using electron beams in the energy range from 15 to 50 MeV can, in principle, be used to deduce values for W air , if the electron stopping power of graphite and air are known. A previous analysis of these data revealed an anomalous variation of 2% in W air as a function of the electron energy. We use Monte Carlo simulation techniques to reanalyze the original data and obtain new estimates for W air , and to investigate the source of the reported anomaly. Methods: Domen and Lamperti (DL) reported the ratio of the response of a graphite calorimeter to that of a graphite ionization chamber for broad beams of electrons with energies between 15 and 50 MeV and at different depths in graphite (including depths well beyond the range of the primary electrons, i.e., in the bremsstrahlung photon regime). Using a detailed EGSnrc model of the DL apparatus, as well as up-to-date stopping powers, we compute the dose ratio between the ionization chamber cavity and the calorimeter core, for plane-parallel electron beams. This dose ratio, multiplied by the DL measured ratio, provides a direct estimate for W air . Results: Despite an improved analysis of the original work, the extracted values of W air still exhibit an increase as the mean electron energy at the point of measurement decreases below about 15 MeV. This anomalous trend is dubious physically, and inconsistent with extensive data for W air obtained at lower energies. A thorough sensitivity analysis indicates that this trend is unlikely to stem from errors in extrapolation and correction procedures, uncertainties in electron stopping powers, or bias in calorimetry or ionization chamber measurements. However, we find that results are quite sensitive to the intrinsic graphite mass thickness of the detectors and to the incident beam energy.
Conclusions:The DL experiment provides data in an energy regime where the electron stopping power is insensitive to the mean excitation energy of graphite -an issue plaguing W air experiments at lower energies. Unfortunately, state-of-the-art scrutiny of the original data cannot explain the anomalous trend in terms of perturbation effects or extrapolation bias. It can only be understood in terms of speculative offsets in graphite mass thickness or beam energy. Therefore higher accuracy measurements for electron energies above 15 MeV are recommended to further resolve the value of W air .