Calculated beam quality correction factors for ionization chambers in MV photon beams

The beam quality correction factor, , which corrects for the difference in the ionization chamber response between the reference and clinical beam quality, is an integral part of radiation therapy dosimetry. The uncertainty of is one of the most significant sources of uncertainty in the dose determination. To improve the accuracy of available data, four partners calculated factors for 10 ionization chamber models in linear accelerator beams with accelerator voltages ranging from 6 MV to 25 MV, including flattening-filter-free (FFF) beams. The software used in the calculations were EGSnrc and PENELOPE, and the ICRU report 90 cross section data for water and graphite were included in the simulations. Volume averaging correction factors were calculated to correct for the dose averaging in the chamber cavities. A comparison calculation between partners showed a good agreement, as did comparison with literature. The values from TRS-398 were higher than our values for each chamber where data was available. The values for the FFF beams did not follow the same , relation as beams with flattening filter (values for 10 MV FFF beams were below fits made to other data on average by 0.3%), although our FFF sources were only for Varian linacs.

approach for determining k Q , based on Monte Carlo calculation of a single correction factor converting the mean absorbed-dose-to-air in the chamber cavity into absorbed dose-to-water at the reference measurement point. According to this alternative approach, the k Q factor for beam quality Q can be directly obtained as ratio where D chamber is the absorbed dose-to-air in the chamber cavity, D w is the dose-to-water at the reference point, and Q 0 refers to the reference radiation quality. (The symbol k Q is used instead of k Q,Q0 since the reference beam quality in this work is a Co-60 beam.) Secondly, improvements in computing power and availability of advanced Monte Carlo codes allowed to point out inaccuracies in determining some individual perturbation factors such as p wall and p dis ( p repl in the AAPM TG-51) that account for the non-water equivalence of the ionization chamber wall, and for the effects due to the replacement of water with the air cavity, respectively (Buckley and Rogers 2006, Wang and Rogers 2008, Wulff et al 2008a, Wang and Rogers 2009. Specifically for photon beams, Monte Carlo simulations by Buckley and Rogers (2006) demonstrated that the semi-empirical expression adopted to calculate p wall does not accurately describe the chamber wall effects. Furthermore, Monte Carlo calculations by Wang and Rogers (2009) revealed that there was an incorrect normalization in the old experimental data on which the TRS-398 p dis values were based. New refined experimental determinations provided p dis values in agreement with the calculated values both in Co-60 and in high-energy photon beams Johansson 2011, 2012), then confirming the reliability of the related Monte Carlo simulations. All these findings have posed the need for a re-evaluation of the recommended values of k Q . Muir and Rogers (2010) used the EGSnrc Monte Carlo system to directly calculate k Q factors for a large number of ionization chambers in photon beams. Their results showed that the Monte Carlo k Q values differed from the corresponding TRS-398 values up to about 1%. Depending on the chamber type and beam quality, these differences could be explained in terms of variations in the many parameters entering the traditional k Q expression when calculated using Monte Carlo simulations. A later analysis (Andreo et al 2013) of data available for a graphite walled ionization chamber widely used for reference dosimetry in radiation therapy, the NE 2571 type, showed that the Monte Carlo k Q values were systematically lower than the TRS-398 values with differences increasing up to about 0.8% at the highest photon beam energies. On the other hand, a comparison of the Monte Carlo calculated values with the experimental k Q data curve, obtained fitting results of measurements performed with different types of D w primary standards, also showed systematic differences, which indicates the need for further investigation.
In this context, in 2016 the IAEA started a process for the updating of the TRS-398 protocol, including the tables of k Q values therein. Importantly, the update of the TRS-398 should take into account new modalities of radiation therapy treatment, new ionization chambers on the market, and possible improvements in dosimetric accuracy. In order to establish new recommended values of k Q , laboratories around the world were invited to submit to the IAEA measured or Monte Carlo calculated k Q factors with their estimated uncertainties. Then, nine European laboratories interested in contributing to the TRS-398 update set up a consortium with the aim to improve robustness of their k Q determinations through cooperation in the framework of a metrology project named RTNORM. In this paper, Monte Carlo simulations for megavoltage photon beams made by four laboratories participating in the RTNORM project are described and results are discussed in comparison with literature data.
For k Q calculations, a number of reference-class cylindrical ionization chambers from different manufacturers, and with different sensitive volumes were considered (table 1). For sources, 20 photon beams from 6 MV to 25 MV, including flattening filter free (FFF) beams, produced by various types of clinical linear accelerators (linacs) as well as two Co-60 beams were used. Settings of radiation transport parameters were initially agreed and chosen so to perform the most accurate simulations. The recent recommendations of the ICRU report 90 (ICRU-90) on key data for reference dosimetry were adopted. Calculations were mostly performed with the EGSnrc code (Kawrakow et al 2017) but for two types of ionization chambers the PENELOPE code (Salvat 2015) was used and results compared to each other. The cooperation among the participating laboratories allowed an extensive investigation on approximations and parameters possibly affecting the Monte Carlo calculations. Specifically, the impact on k Q results of chamber modeling, type of input beam source and approximations in calculating point absorbed dose-to-water were investigated, and critical aspects in the Monte Carlo simulations have been identified and discussed.

Methods
ENEA, THM and STUK calculated the k Q values with EGSnrc user code egs_chamber (Wulff et al 2008b), while IST performed their calculations with PENELOPE user code penEasy (Sempau et al 2011, Salvat 2015. The chambers were modeled according to product catalogs and blueprints provided by the manufacturers. Every partner modeled their chambers independently. The chamber calculations were performed in a 30 cm × 30 cm × 30 cm water-phantom without walls. The source to surface distance (SSD) was 100 cm for linacs. The depth of the reference point of the chamber was at 5 g cm −2 for Co-60 sources, and 10 g cm −2 for linacs for THM, and at 5 cm and 10 cm for ENEA, STUK and IST. The depth differed, since the density of water in ICRU-90 was 0.9982 g cm −3 . The field size at 100 cm was 10 cm × 10 cm for each linac. The dose-to-water was scored to a 0.025 cm thick disk with 1 cm radius by IST and THM. For Farmer-type chambers, ENEA used the same disk dimensions, or a smaller disk radius of 0.25 cm for the CC13 chamber type. For STUK, the dose-to-water was scored in FC65-G and PTW 31021 chamber cavity shaped water volumes.
The chambers under investigation are listed in table 1. All chambers were waterproof, except the NE 2571, which had a 1 mm layer of PMMA on top. The PTW chamber walls consisted of an inner graphite layer surrounded by a thicker layer of PMMA. The charge collection efficiency was assumed to be homogeneous throughout the cavity volumes.
While the amorphous density of graphite used in the calculations was 1.7 g cm −3 , the evaluation of the density effect used the crystalline density of 2.265 g cm −3 . While ICRU-90 does not make an explicit recommendation, in this work the multiconfiguration Dirac-Fock re-normalization factors for photoelectric cross sections (mcdfxcom) were used.

Radiation sources
Point sources with linac energy spectrum, fully simulated linacs and Co-60 irradiators were used in the calculations. Part of the sources are presented in Mohan et al (1985), Picard et al (2010Picard et al ( , 2011Picard et al ( , 2013 and Czarnecki et al (2017). In addition, a Co-60 source and a Varian TrueBeam linac (Varian Medical Systems, Inc., Palo Alto, California, USA) were modeled with the EGSnrc user code BEAMnrc. Either phase-space files (PSF) or BEAMnrc shared library simulation sources (Kawrakow and Walters 2006), later referred to as BEAM sources, were used in egs_chamber as full linac sources. A list of the sources is in table 2.
ENEA, IST and THM used a BEAMnrc input file for Eldorado Co-60 irradiator available in the IAEA phasespace database Muir et al (2009), based on the work in Mora et al (1999), as the reference source. The input file was used to create a PSF at 80.5 cm distance from the source. STUK used a BEAMnrc model of their Gammabeam X200 Co-60 irradiator (GBX200) manufactured by Best Theratronics Ltd. (Ottawa, Canada) as the reference source. The source radius was 1 cm and it was encapsulated in a stainless steel container. The collimation consists of a stationary tungsten primary definer, and adjustable lead collimator leaves and tungsten trimmer bars. The scoring plane was at 95 cm from the source and the field size was 10 cm × 10 cm at 100 cm. No variance reduction techniques (VRTs) were used in the GBX200 simulations. A more detailed description and validation of the GBX200 BEAMnrc model will be discussed in a separate study. The SSD corresponded to the BEAMnrc scoring plane distance for both Co-60 sources.
BEAMnrc simulation sources for Varian TrueBeam linac photon beams were based on PSF provided by the manufacturer. The simulated radiation qualities were 6X, 6X-FFF, 10X, 10X-FFF and 15X. The PSFs were collected at a plane above the jaws, and were used as a source in subsequent BEAMnrc treatment head simulation. The input file contained the jaws, an approximation of the collimator baseplate, the MLC in park position, the light field reticle, and the interface mount with the geometrical and material details provided by the manufacturer. The scoring plane was 100 cm from the source for each beam. In all TrueBeam simulations EGSnrc code package version 2018 was used, with simulation parameters shown in table 3 with the exceptions ECUT = AE = 0.521 MeV, PCUT = AP = 0.01 MeV and Brems cross sections = NRC. The BEAMnrc linac head model for the Elekta Precise accelerator (Elekta Instrument AB, Stockholm, Sweden) was provided by PTB Germany. It was modeled according to the blueprints provided by Elekta. The linac head model and resulting phase space files were evaluated in a BIPM key comparison (Picard et al 2011). A Varian Clinac 10X-FFF BEAMnrc simulation source, described in Czarnecki et al (2017), was also used in the calculations.

Beam quality specifier
The beam quality specifier chosen for this work was the TPR 20,10 , which was calculated with equation from Followill et al (1998) TPR 20,10 = 1.2661PDD 20,10 − 0.0595, where PDD 20,10 is the ratio of dose-to-water at depths 20 cm and 10 cm with 100 cm SSD at 10 cm × 10 cm field size. The reason for using the conversion from PDD 20,10 to TPR 20,10 was that most of the PSF were scored at a distance larger than 80 cm from the source, and hence the TPR 20,10 could not be calculated by definition. Since (2) was derived from with-flattening-filter beam data, TPR 20,10 for the TrueBeam FFF sources was calculated also by definition. The scoring volume for TPR 20,10 calculations was a 0.25 cm thick disk with radius of 0.5 cm for ENEA and STUK and the 0.025 cm high disk with 1 cm radius for others.

Simulation parameters and variation reduction techniques in EGS
In the k Q simulations in EGSnrc, the photons were simulated until the energy was below 1 keV (PCUT and AP) and electrons until 512 keV (ECUT and AE), corresponding to 1 keV kinetic energy. A list of the Monte Carlo parameters, unless mentioned otherwise, is in table 3. The egs_chamber user code provides multiple effective VRTs that speed up the calculations significantly, and enable extracting more information from a limited number of particles. These techniques are presented in Wulff et al (2008b). One such technique is the photon cross-section enhancement (XCSE), where the photon total cross section is increased by an enhancement factor n in a volume surrounding the chamber cavity. The secondary particle weights are reduced by the same factor and photon interactions are carried out with probability 1/n. This leads to a situation where higher number of electrons, with reduced weight, are created along the track of a photon.
In Russian Roulette (RR), when an electron is created, the program checks whether a user defined geometry is within the electron range. If not, the electron is simulated only with probability 1/m, where m is a rejection factor, and the weight is multiplied by m. Simulating part of the electrons will take into account the contribution to dose from x-rays generated by the electrons. If the electron is inside the volume, it is terminated and energy deposited  in the current region if the electron energy is below an Esave parameter. The material with which the electron ranges are calculated is given as a input and it should be the least attenuating material around the chamber cavity. If there is air outside the chamber cavity, it is more effective to expand the RR region around those areas than define the rejection material as air.
In intermediate phase-space scoring (IPSS), an intermediate phase-space is scored on a surface of a defined geometry. These scored particles are then transported to every other geometry in the scoring options block in the input file. This can be convenient when simulating multiple geometries at the same position.
ENEA used an XCSE region surrounding the chamber geometries by 2 cm, and an enhancement factor of 128 for Co-60 and 6 MV beams and 64 for other beams. The RR rejection factor was equal to the enhancement factor and the Esave parameter was 521 keV. STUK had a XCSE enhancement factor of 64 and rejection factor of 128 for all calculations. The Esave parameter was 512 keV. THM applied for all simulations a XCSE factor of 128 (equal to the RR rejection factor) within a region of 2 cm surrounding the individual chamber geometries and the ESAVE parameter was 521 keV. To save computation time, STUK and THM applied the IPSS technique and simulated all the chambers in one run. The IPSS volume encompassed all XCSE volumes.

Simulation parameters and variation reduction techniques in PENELOPE
The penEasy version in the calculations was 2015-05-13 (compatible with the PENELOPE 2014 code package). The use of PENELOPE and penEasy for calculating k Q and TPR 20,10 factors has been studied for example in Sempau and Andreo (2006), Erazo and Lallena (2016), Gomà et al (2016). Material files in accordance with ICRU-90 were used and the mean excitation energies were set to 81 eV and 78 eV for graphite and water, respectively.
The detailed geometries of the ionization chambers were designed by means of quadratic surfaces using the PENELOPE package PENGEOM (Salvat 2015). The chambers implemented in PENELOPE were NE 2571 and Exradin A12S. Both chamber geometries were modeled according to manufacturer's specifications and literature Lallena 2013, Pimpinella et al 2019). The tallyEnergyDeposition was used to calculate the energy deposited in the chamber cavity and in water.
PENELOPE uses various parameters for electron and positron transport: C1 (average angular deflection produced by multiple scattering), C2 (maximum average fractional energy loss between consecutive hard inelastic events), Wcc (cutoff energy for hard inelastic collisions) and Wcr (cutoff energy for hard bremsstrahlung emission). DSMAX refers to the maximum step length of charged particles in the materials. Typically, this parameter is set to approximately 0.1 times the material thickness (Salvat 2015). For both chambers, detailed simulation (C1 = C2 = Wcc = Wcr = 0) was performed in the active volume, as well as in all regions sharing a surface with it. For other materials in the chamber and water present up to 2 cm around the chamber, the parameters C1 and C2 were set to 0.05. In the water phantom and surrounding air, C1 and C2 were 0.1. Wcc and Wcr had values of 10 keV and 1 keV, respectively, for all materials, except the ones where detailed simulation was performed. Additionally, DSMAX was set to approximately one tenth of material thickness. For photons, cutoff energy was 1 keV in all materials. For electrons and positrons, cutoff energies were set to 1 keV for all chamber materials, 10 keV for water present up to 2 cm around the chamber and 200 keV for other regions. The simulations to obtain D w were performed analogously to the chamber simulations. A 0.5 mm water skin enveloped the dose-to-water scoring volume. Detailed simulation was performed in both of these bodies. The C1, C2, Wcc, Wcr and DSMAX were as in chamber simulations with the region of C1 = C2 = 0.05 enveloping the disk by up to 2 cm.
For all simulations, simple particle splitting was implemented in the sensitive volume, with a factor of 10. Additionally, a SPLITTING FACTOR was used for the phase space sources, equal to 10 for the Varian and NRC's Elekta beams and 200 for the Co-60 beam (the Eldorado Co-60 PSF was much smaller than the other PSF). During the simulation, each particle in the PSF was split and run SPLITTING FACTOR times and the statistical weight was reduced accordingly. However, a very large SPLITTING FACTOR may lead to a stack overflow. To solve this problem, the parameter NMS was modified to 1000 000 in the PENELOPE source files (penEasy v.2015-05-13 documentation, source phase space file. J. Sempau). The splitting factors presented in this section are the result of several trial and error tests performed in an effort to improve simulation efficiency without biasing the simulation results.

Difference in dose profiles between phase space files and beam sources
When calculating the dose profiles with EGSnrc for the TrueBeam, Eldorado and GBX200 PSFs, the variance between single simulated doses was larger than what would have been predicted by the uncertainties. The effect was observed regardless of whether the PSF were restarted (the PSF run multiple times without reducing the particle weights) or not, making the latent variance of the PSFs an unlikely reason for the behavior (Sempau et al 2001, Alhakeem andZavgorodni 2018). When making identical simulations with BEAM sources, the variance between the points was lower and the results were as expected. THM assumed that the dimensions of the chambers were large enough to average most of the unexpected variance in the profiles and BEAM sources were used whenever possible. Preliminary investigation with PENELOPE did not show unexpectedly large variance in the profile calculations.

Volume correction factors
Volume correction factors k vol , according to the recommendation of the IAEA report TRS-483 (IAEA 2017) resulting from inhomogeneous dose profiles, were applied to the k Q by IST and THM. The correction factor converts the average dose throughout the dose calculation volume into the dose at the reference point. The formulae for calculating the k vol factors are presented for example in Delaunay and Ostrowsky (2007) for the simplest methods and analyzed more thoroughly in IAEA TRS-483. The equation for the volume correction factor from the TRS-483 is where w(x, y) is the weighting factor that takes into account the calculation volume thickness in the beam direction, and OAR is the off-axis-ratio, meaning the ratio of dose at position (x, y) and at the center of the beam. The simplest method to calculate the correction factor for a cylindrical ionization chamber is a line integral along the chamber axis where L is the chamber length and the chamber is aligned along the y -axis. Equation (4) assumes that the dose averaging in radial direction is negligible. For the disk shaped D w calculation volume, this assumption is not valid and the volume correction factor was calculated assuming cylindrical symmetry of the profile within the disk radius where the weighting factor w(r) ≡ 1 since the thickness is constant, R is the disk radius and r the distance from the beam axis.
To obtain the OAR, THM calculated dose-profiles for the sources they used. The volume correction factors were applied to chambers simulated by THM and the profiles were calculated with the egs_chamber user code. The water voxel for the profile calculations was a disk with a radius of 0.1 cm and height of 0.2 cm centered around the measurement depth 10 cm for the high energy photon beams and 5 cm for Co-60. To save calculation time, the calculations were performed applying an ECUT value of 0.521 MeV instead of 0.512 MeV.
Additionally, the profiles of the TrueBeam 6X-FFF and 10X-FFF beams at the 10 cm depth were calculated using a BEAM source and the volume correction factors obtained from these calculations were used by IST. The water voxels for the profile calculations were rectangular parallelepiped with dimensions of 0.1 cm × 0.3 cm × 0.1 cm in xyz-directions. The variance reduction parameters were as in section 2.3. Both THM and IST used polynomial fits to the simulated profiles to get the OAR function.  (4) and (5), and the correction factors applied to the k Q calculated with average doses over the dose scoring volumes calculated with (7). The partner who applied the correction factor in question is given in brackets.

Beam
Chamber k vol k vol The measured dose-to-water at reference depth in beam quality Q is given by where N D,w,Q0 is the calibration coefficient for the dose-to-water measured in a Co-60 beam and M Q is the dosimeter reading corrected for influence quantities. If these quantities include the volume averaging in the cavity, the D chamber in (1) has to be corrected for the volume averaging. In this case, k Q corrects only for the difference of the chamber response between the radiation spectra of Q and Q 0 . If the volume averaging correction is not included in M Q , then the average dose over the cavity volume is used in the calculation of k Q . The dose-towater is defined at the reference point in both cases, meaning that the volume averaging in the D w scoring volume is corrected for. The correction factor applied to the k Q values calculated with simulated (average) doses over both D w and D chamber scoring volumes can then be obtained with Note that (7) assumes the volume correction factors for the Co-60 beam to be unity.

Comparison of beam quality correction factors for NE 2571 chamber
Each partner calculated the k Q factor for the NE 2571 chamber with the Eldorado Co-60 PSF and Elekta Precise 6 MV PSF from NRC (Picard et al 2010) for validation of the calculation methods. Every partner restarted the PSF (the particles in the file were run multiple times without reducing the weight). The results of the comparison are in  difference between the results might be due to variations in library data. The estimation of type B uncertainties is studied for example in Muir and Rogers (2010), Wulff et al (2010), Czarnecki et al (2018), and the subject is beyond the scope of this publication. However, based on these studies the type B uncertainty of the Monte Carlo k Q values can be estimated below 0.5% (k = 1). The differences between the comparison results were deemed small enough to regard the calculation methods of each partner valid.

Volume correction factors
The volume averaging was most significant in the FFF-beams, and in PTB's Elekta Precise 10 MV and 25 MV beams. Because of the more significant dose averaging in the 1 cm radius disk D w scoring volume than in the cavity, the corrections to the k Q calculated with averaged doses were largest for the small chambers. Some of the calculated volume correction factors are listed in  account even though FFF beams were used, since the correction factor was the same for the chamber and the cavity shaped dose-to-water scoring volume. The shapes of NE 2571 and FC65-G chamber cavities were close enough to assume the difference between the volume correction factors to be insignificant. ENEA's calculations of k Q factors showed equal results (within 0.1%) when using dose-to-water scoring volumes matching the chamber cavity or using a disk with diameter matching the chamber cavity length (Pimpinella et al 2019). These calculations allowed ENEA to establish that the volume averaging effect can be neglected for the sources and the dose-to-water scoring volumes that they used. Although (4) uses the simple weighting factor, the k vol factors according to TRS-483 are close to those calculated with more accurate methods. This was verified for the FFF beams by calculating the volume correction factors for NE 2571, FC65-G and A12S chambers with a model that takes into account the chamber cavity shape and the central electrode. The factors agreed to within 0.0003.

Calculated beam quality correction factors
The calculated beam quality correction factors and comparison to previously published data are shown in figures 1-7. In each figure, the error bars represent the type A standard uncertainty. The uncertainty of the k Q was calculated by assuming that the uncertainties of individual dose calculations were uncorrelated and the correlated sampling option in egs_chamber was not used.
where a, b and c are fitting parameters, was used to obtain fits to the k Q data. The parameters are given in table 6. The fits do not have any physical meaning and may not be valid outside the TPR 20,10 range of the sources used. The FFF-beam data does not follow the same fit as the with flattening filter (WFF) beam data for the 10 MV FFF beams and the FFF points are below the fit. Due to the difference, the FFF-beam results were not used when calculating the fitting parameters. The difference of the FFF-beam k Q values to the value of the fit at the same TPR 20,10 for the 10X-FFF beams averaged per chamber ranged from 0.2% to 0.5% and the average difference over all the chambers was 0.33%. The difference between the WFF and FFF beam k Q values was not estimated for the 6X-FFF beam since the fit would have had to be extrapolated. If the volume averaging in the chamber cavity had not been corrected for, the FFF and WFF data would be in better agreement. However, we can make no general conclusion of the validity of using the same TPR 20,10 , k Q fit for FFF-beams from the results of this work, especially since all the FFF data is for Varian linacs; other manufacturers realise the FFF beams in a different way (de Prez et al 2018).
The difference between the TPR 20,10 for FFF beams calculated with (2) and by definition was less than 0.004. This difference is well within the expanded uncertainty of approximately 0.009 of the TPR 20,10 , PDD 20,10 relation given in Followill et al (1998). When using the fitting parameters for the NE 2571 chamber from table 6, the changes in k Q between the TPR 20,10 calculated by the two different methods were less than 0.05%. Therefore the TPR 20,10 values calculated with (2) for the FFF beams used are considered accurate enough for the purposes of this work.
For every chamber, the results between partners are consistent, especially with acceleration voltages of 10 MV and lower, and difference to considered literature, not including TRS-398, is less than approximately 0.5%. For the NE 2571 chamber, the agreement with other calculated k Q values is good, with difference in figure 2 never exceeding 0.2%. The same level of agreement is found with the experimental determinations from McEwen (2010). Difference to measured values from Andreo et al (2013) and de Prez et al (2018) is larger, but always below 0.4%. For the PTW 30013 ( figure 3(a)), there is a systematic difference between our results and the fit from Muir and Rogers (2010), while there is a good agreement with more recent data from Mainegra-Hing and Muir (2018)  Phys. Med. Biol. 65 (2020) 075003 (15pp) who reportedly modeled in more detail the stem of the chamber, compared to Muir and Rogers (2010). In this study we also paid particular attention to the geometry of this chamber type, and this may possibly explain the better agreement between our results and those of Mainegra-Hing and Muir (2018). This study also shows a systematic difference for PTW 31010 (figure 3(b)) between our results and those of Wulff et al (2008a) and Muir and Rogers (2010) at higher TPR 20,10 values, even though EGSnrc was used in both studies. Of note is that agreement with experimental results from McEwen (2010) is very good for both chambers. The TRS-398 values are higher than ours for each chamber. For the IBA CC13 chamber, the difference is most significant in the TPR range of 6 MV to 10 MV beams. However, our results agree with the data in TRS-398 within the TRS-398 stated standard uncertainty of 1%. The differences are at least partly explained by issues discussed in section 1. For PTW 31021, our results agree well with the results in Delfs et al (2019), but the data from STUK is systematically lower than the data from THM. The adaptation of the new ICRU-90 cross sections does not affect the k Q values significantly based on Czarnecki et al (2018) and Mainegra-Hing and Muir (2018).

Conclusions
We calculated an extensive dataset of k Q factors for 10 ionization chambers adding to 147 k Q values in 20 linac beams. The validity of the results was improved by comparing partner's calculated k Q and TPR 20,10 values to each other, and to literature. No systematic differences across chambers to calculated nor measured literature results were observed, with the exception of TRS-398 data, where the k Q were higher for each chamber where data was available. Part of the chambers have little existing k Q data and are not included in the current version of TRS-398, for example PTW 31013 and the PTW Semiflex 31021 3D. We used the ICRU-90 cross sections for water and graphite, which only few publications have provided k Q data with. The FFF-beam results, calculated for two Varian linacs, do not follow the same fit as beams with a flattening filter, at least when the volume averaging in the chamber cavity is corrected in the determination of k Q . Furthermore, our FFF beams followed the conversion from PDD 20,10 to TPR 20,10 from Followill et al (1998) within the stated uncertainties of that work.