Purpose: There are currently several commercially available radiotherapy treatment units without a flattening filter in the beam line. Unflattened photon beams have an energy and lateral fluence distribution that is different from conventional beams and, thus, their attenuation properties differ. As a consequence, for flattening filter free (FFF) beams, the relationship between the beam-quality specifier TPR20,10 and the Spencer-Attix restricted water-to-air mass collision stopping-power ratios, ((L) over bar/rho)(air)(water), may have to be refined in order to be used with equivalent accuracy as for beams with a flattening filter. The purpose of this work was twofold. First, to study the relationship between TPR20,10 and ((L) over bar/rho)(air)(water) for FFF beams, where the flattening filter has been replaced by a metal plate as in most clinical FFF beams. Second, to investigate the potential of increasing the accuracy in determining ((L) over bar/rho)(air)(water) by adding another beam-quality metric, TPR10,5. The relationship between ((L) over bar/rho)(air)(water) and % dd(10)(x) for beams with and without a flattening filter was also included in this study. Methods: A total of 24 realistic photon beams (10 with and 14 without a flattening filter) from three different treatment units have been used to calculate ((L) over bar/rho)(air)(water), TPR20,10, and TPR10,5 using the EGSnrc Monte Carlo package. The relationship between ((L) over bar/rho)(air)(water) and the dual beam-quality specifier TPR20,10 and TPR10,5 was described by a simple bilinear equation. The relationship between the photon beam-quality specifier % dd(10) x used in the AAPM's TG-51 dosimetry protocol and ((L) over bar/rho)(air)(water) was also investigated for the beams used in this study, by calculating the photon component of the percentage depth dose at 10 cm depth with SSD 100 cm. Results: The calculated ((L) over bar/rho)(air)(water) for beams without a flattening filter was 0.3% lower, on average, than for beams with a flattening filter and comparable TPR20,10. Using the relationship in IAEA, TRS-398 resulted in a root mean square deviation (RMSD) of 0.0028 with a maximum deviation of 0.0043 (0.39%) from Monte Carlo calculated values. For all beams in this study, the RMSD between the proposed model and the Monte Carlo calculated values was 0.0006 with a maximum deviation of 0.0013 (0.1%). Using an earlier proposed relationship [Xiong and Rogers, Med. Phys. 35, 2104-2109 (2008)] between % dd(10) x and ((L) over bar/rho)(air)(water) gave a RMSD of 0.0018 with a maximum deviation of 0.0029 (0.26%) for all beams in this study (compared to RMSD 0.0015 and a maximum deviation of 0.0048 (0.47%) for the relationship used in AAPM TG-51 published by Almond et al. [Med. Phys. 26, 1847-1870 (1999)]). Conclusions: Using TPR20,10 as a beam-quality specifier, for the flattening filter free beams used in this study, gave a maximum difference of 0.39% between ((L) over bar/rho)(air)(water) predicted using IAEA TRS-398 and Monte Carlo calculations. An additional parameter for determining ((L) over bar/rho)(air)(water) has been presented. This parameter is easy to measure; it requires only an additional dose measurement at 5 cm depth with SSD 95 cm, and provides information for accurate determination of the ((L) over bar/rho)(air)(water) ratio for beams both with and without a flattening filter at the investigated energies. (C) 2014 American Association of Physicists in Medicine.