Role of volume and surface processes in the atomic oxygen loss frequency in oxygen glow discharges in Pyrex

Plasma Sources Science and Technology PAPER • OPEN ACCESS Role of volume and surface processes in the atomic oxygen loss frequency in oxygen glow discharges in Pyrex To cite this article: Pedro Viegas et al 2025 Plasma Sources Sci. Technol. 34 085003 View the article online for updates and enhancements. You may also like A fine-structure-resolved collisional radiative model for Ti plasma Nitish Ghosh and Lalita Sharma - Numerical investigation of stability of lowcurrent needle-to-plane negative corona discharges in air N G C Ferreira, P G C Almeida, A Eivazpour Taher et al. - Influence of large biased surfaces on the plasma parameters in the extraction region of a negative hydrogen ion source D Yordanov, D Wünderlich, C Wimmer et al. - This content was downloaded from IP address 194.210.228.226 on 19/12/2025 at 17:58 Plasma Sources Science and Technology Plasma Sources Sci. Technol. 34 (2025) 085003 (20pp) https://doi.org/10.1088/1361-6595/adf5e4 Role of volume and surface processes in the atomic oxygen loss frequency in oxygen glow discharges in Pyrex Pedro Viegas1,∗, Tiago Cunha Dias2and Vasco Guerra1 1Instituto de Plasmas e Fus˜ ao Nuclear, Instituto Superior Técnico—Universidade de Lisboa, Av. Rovisco Pais 1, 1049-001 Lisboa, Portugal 2Electrical Engineering and Computer Science Department, University of Michigan, Ann Arbor, MI, United States of America E-mail: pedro.a.vieg[email protected] Received 25 March 2025, revised 7 July 2025 Accepted for publication 30 July 2025 Published 7 August 2025 Abstract This work aims at understanding and finding the most accurate way to estimate an effective O surface recombination probability (γO) from the temporal evolution of O(3P) density measured in modulated current conditions in the positive column of oxygen glow discharges in Pyrex, which is relevant for experimental characterization and for model inputs. The procedure for deducing γOfrom the O loss frequency (νloss O) is not straightforward, since the processes determining the net losses of O(3P) are not known a priori. A global model describing plasma chemical kinetics is used in steady-state and current-modulated modes to assess the physical meaning of νloss O, for a total of 66 experimental conditions in the pressure range 0.4–7.5 Torr with 7.8 sccm flow rate and 20–40 mA currents. An optimal γOis derived from νloss Oand the accuracy of different hypotheses on the relevant processes for the net losses of O(3P) is addressed. The hypothesis that is found to be the most accurate indicates that the net loss of O(3P) is due not only to surface recombination, but also to recombination in volume and to flow losses. Volume processes account on average for 24% of O(3P) net losses, and that fraction tendentiously grows with pressure, reaching up to 71%. The verified hypothesis can be adopted for the estimation of γOfrom the measured νloss Oin both discharge (partial modulation) and post-discharge (full modulation) conditions. Finally, a discussion is had on what steps can be taken for a complete validation of the oxygen glow discharge model. Keywords: oxygen kinetics, recombination probability, global model, surface recombination 1. Introduction Low-temperature plasmas interact with surfaces in most of their applications, either in the active discharge phase or ∗Author to whom any correspondence should be addressed. Original Content from this work may be used under the terms of the Creative Commons Attribution 4.0 licence. Any further distribution of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI. in post-discharge. In several applications, this interaction is essential and delving into it enables to better control plasma properties and the design of effective processes (Kushner 2009, Murphy and Park 2017, Bogaerts et al 2022). Low-temperature plasmas alter solid surfaces by introducing charged particles and reactive gaseous species (Oehrlein 1997, Kersten et al 2001, Vanraes et al 2021). On their turn, the densities of these species in the plasma are affected by surface kinetics. As such, understanding the interaction between low-temperature plasmas and surfaces is crucial for several reasons: it not only deepens our fundamental knowledge of the relationship between these two media when 1 © 2025 The Author(s). Published by IOP Publishing Ltd Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al they are in direct contact, but it is also vital for optimizing processes such as plasma etching, surface modification, thin film deposition, atmospheric re-entry and plasma-tissue treatment. Oxygen-containing plasmas are widely used in material processing, biomedical applications and gas conversion technologies. One of the reasons for their interest is the key role of reactive oxygen species (ROS), and of the atomic oxygen O radical in particular. One example of its importance is O production in CO2conversion plasmas by molecular dissociation, leading to reactions that can enhance or hinder the final conversion. O atoms can react with CO2to increase dissociation, quench excitation, oxidize CO back to CO2, or recombine to form O2(Morillo-Candas et al 2019, van de Steeg et al 2021a, 2021b,2022). Among other applications, O plays a crucial role in diagnosing plasmas from its radiation (Booth and Sadeghi 1991, Caplinger and Perram 2020, Fiebrandt et al 2020, Viegas et al 2021) and in product separation from plasmas for electrochemical conversion (Rohnke et al 2004, Meiss et al 2008, Chen et al 2020, Pandiyan et al 2022). In pure oxygen plasmas, O atoms are easier to study, since they are produced mainly through electron impact dissociation of O2. At low pressures, O atoms are mostly lost due to surface recombination back into O2(Gousset et al 1989, Macko et al 2004, Annuˇ sová et al 2018, Dias et al 2023) or O3(Mazánková et al 2020,2024, Booth et al 2023, Meyer et al 2023). Therefore, the adsorption and recombination of O can determine the gas composition, the flux of ROS towards target surfaces and the availability of O for important volume reactions. Analysing oxygen plasmas is thus essential for a better understanding of plasmasurface interactions and the crucial role of atomic oxygen in these processes. Numerical modeling is an important tool to understand and optimize processes involving low-temperature plasmas. To accurately describe oxygen-based plasma systems using self-consistent macroscopic models, the understanding of O surface recombination is essential. In particular, an effective atomic oxygen surface recombination probability γOis often used in models to describe in a simple way the oxygen recombination at the surface. γOis defined as the portion of O flux to the wall that is lost. However, O surface interaction data are highly dependent on the specific working conditions, such as the type of surface, its temperature, the discharge current, pressure and gas temperature. Atomic oxygen surface recombination is thus a very complex process, that has been studied via kinetic modeling either analytically (Guerra 2007, Rakhimova et al 2009, Lopaev et al 2011, Booth et al 2019) or numerically (Gordiets et al 1996, Gordiets and Ferreira 1998, Cartry et al 2000, Guerra and Loureiro 2004, Guerra and Marinov 2016, Afonso et al 2024, Viegas et al 2024). Nevertheless, used either as direct model input or as benchmark for kinetic modeling, γOis often derived from fitting experimental measurements. The accuracy of models thus relies heavily on those measurements and their interpretation (Booth et al 2019,2020, 2023, Viegas et al 2023,2024). A recent review by Paul et al (2023) lists and compares surface recombination coefficient γOmeasurements for different diagnostic methods, types of discharges and surfaces. Inert materials like silica-based or glass-based materials are common not only in space vehicles used in atmospheric reentry conditions, where O recombination is important for the heat transfer to the surface, but also in plasma reactors. Several studies have experimentally assessed the dependence of γO on surface temperature, pressure, current, gas temperature and other parameters for Pyrex (Pagnon et al 1995, Macko et al 2004, Booth et al 2019,2020, Ziganshin et al 2025), for quartz (Kim and Boudart 1991, Balat-Pichelin et al 2003, Lopaev and Smirnov 2004, Cartry et al 2006, Lopaev et al 2011) and other silica-like materials (Cartry et al 1999, Rakhimova et al 2009, Kriˇ stof et al 2012, Ono and Kimata 2024). γOis consistently lower in post-discharge conditions than during the discharge. Several studies indicate that the oxygen loss probability on Pyrex may be about one order of magnitude higher when the surface is submitted to the plasma than in post-discharge (Magne et al 1993, Pagnon et al 1995, Macko et al 2004, Cartry et al 2006). Gordiets and Ferreira (1998) suggest that this may be due to the production of chemisorption sites on the surface by the plasma, approximately doubling its number during the plasma phase, when compared to post-discharge, which would promote higher surface recombination rates. Nevertheless, at a low pressure value of 0.7 Pa (5 mTorr), Bousquet et al (2007) find the difference between O surface recombination in discharge and post-discharge conditions to be relevant only in cases with humidity, not in pure O2and CO2plasmas. According to that work, H2O and OH molecules passivate sites in post-discharge, which are reactivated by the discharge due to desorption by ion bombardment. Despite the extensive list of studies, it is difficult to directly determine γOduring plasma operation in a nonintrusive way. Booth and Sadeghi (1991) developed the technique of time-resolved actinometry in modulated plasmas to allow to probe the kinetics of atom production and loss processes. This technique allows to measure a loss frequency of atomic oxygen, νloss O, constant during the modulation cycle. A priori, this quantity is equal to the surface loss rate divided by the atomic oxygen density, νloss O≃Lsurf/[O]. The measurement of νloss Othus allows to estimate the recombination probability γO, since, in a cylindrical container with a spatially uniform plasma, the portion of O flux to the wall that is lost, multiplied by the cylinder surface, must match the surface loss rate multiplied by the cylindrical plasma volume, such that: Lsurf ·πR2L=2πRL ·ΓO·γO=2πRL ·[O]·vth 4·γO,(1) γO=Lsurf ·2·R [O]·vth ≃νloss O·2·R vth ,(2) vth =√8kBTnw πmO ,(3) where ΓOis the flux of atomic oxygen towards the cylindrical wall, vth is the thermal velocity of O atoms at near-wall temperature Tnw,Ris the cylinder radius, Lis the cylinder length, kB is the Boltzmann constant and mOis the mass of atomic oxygen. As shown by equation (1), γOis a macroscopic parameter that is effectively written in the first order, i.e. as a constant, although the surface recombination processes themselves are 2 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al not necessarily independent of O density and other physical conditions. In fact, γOcan be dependent on O density and other parameters, despite being written as a constant for each condition (Guerra 2007). Using this technique, νloss Oand γOwere recently measured in oxygen and CO2plasmas by Booth et al (2019,2020), Morillo-Candás (2019), Morillo-Candas et al (2019) and Dias (2019) in both discharge (partial modulation) and postdischarge (full modulation) conditions. Some of these measurements were recently reproduced via mesoscopic numerical modeling by Afonso et al (2024) and Viegas et al (2024), which allowed to gain insights on the surface recombination phenomena determining the recombination of atomic oxygen. Inclusively, Afonso et al (2024) described reversible surface modification by fast particle impact at pressures below 1 Torr and its impact on the measured νloss Oand the estimated γO. Moreover, a global model for the positive column of the oxygen glow discharge, including O surface recombination, was recently presented and validated by Dias et al (2023). That work presented a reaction mechanism for oxygen plasmas, i.e. a set of reactions and corresponding rate coefficients, and validated them against benchmark experiments, a requirement for making sensible modeling predictions of the system under study (Casas 2023). The validation by Dias et al (2023) consisted in comparing simulation results of a self-consistent 0D model describing the positive column of an oxygen DC glow discharge, for gas pressures of 0.2–10 Torr and currents of 10– 40 mA, with experimental measurements of the main species densities, the gas temperature and the reduced maintenance electric field. Furthermore, that work argued that the development and validation of kinetic schemes for plasma chemistry should adopt a paradigm based on the comparison against standard validation tests. The comparison should ensure that the kinetic scheme is physically consistent and describes the observed tendencies of several quantities in a wide range of conditions, without aiming to put all simulation results on top of the experimental values. This work approaches the measurements of νloss Oand, given its importance for the interpretation of experiments and for setting up accurate models, questions the hypothesis νloss O≃ Lsurf/[O], present in equation (2), and aims to derive the most realistic γOfrom νloss O. Hence, it examines whether surface recombination is the only process determining the net loss of atomic oxygen during modulated conditions and investigates what processes are described by the measured νloss O. We start by presenting in section 2the possible interpretations of the measured O loss frequency and the possible formulas for deriving γOfrom it. This is done by listing and numbering a set of alternative hypotheses to be tested through numerical modeling. Then, in section 3the conditions studied in this work are presented, corresponding to the experimental measurements of νloss Oin pure oxygen glow discharge plasmas by Dias (2019). Moreover, the numerical models employed are described, both for use in conditions of steady-state discharge and of modulated current. The possibility of modeling the real conditions of measurement of νloss O, i.e. modulated current conditions, allows to find γOthat matches the νloss Omeasurements, as the most realistic model-dependent γO. Furthermore, the model can be used to assess the validity of the listed hypotheses. In section 4 the optimal γOis found. Moreover, the hypotheses are evaluated and the conclusions from that evaluation are taken to suggest corrections to the O loss probability estimation and to deduce the relevant O loss processes. This methodology is applied for both discharge and post-discharge conditions. Finally, in section 4.5 the next steps for model validation following this work and the work by Dias et al (2023) are discussed and in section 5the main conclusions of this work are presented. 2. O loss frequency interpretation and recombination probability estimation Atomic oxygen loss frequency measurements can be obtained, either using partially modulated current (for discharge conditions) or fully modulated current (for post-discharge conditions), by tracing the temporal evolution of O density. In the case of spatially uniform plasmas, such as the glow discharge positive column, this density, as other plasma parameters, is considered as global or volume-averaged, rather than local and space-dependent. The temporal evolution of O density can be traced by employing time-resolved actinometry and following the O(777 nm), O(845 nm) and Ar(750 nm) lines (Booth and Sadeghi 1991, Lopaev and Smirnov 2004). The O(777 nm)/Ar(750 nm) and O(845 nm)/Ar(750 nm) line ratios are related to the ground-state O(3P) density and thus allow to follow it. Under post-discharge conditions, there is no plasma emission and thus actinometry cannot be used, but the O density can be followed using absorption techniques, such as cavity ring down spectroscopy (CRDS) (Dias 2019, Morillo-Candás 2019). In partial modulation cases, the temporal evolution of the O(3P) density during current modulation is often very well fitted by exponentials, i.e. by the following expressions, respectively for the rise and fall half-periods (see figure 2) (Booth et al 2019, Dias 2019, Morillo-Candas et al 2019): [O(3P)]rise (t)=[O(3P)]max −([O(3P)]max −[O(3P)]min) ·exp(−νrise O(t−t0)),(4) [O(3P)]fall (t)=[O(3P)]min +([O(3P)]max −[O(3P)]min) ·exp(−νfall O(t−t0)),(5) where [O(3P)]min and [O(3P)]max are the minimum and maximum values of the O(3P) density and t0is the starting instant of the fit. Since the rise and fall frequencies are experimentally found to be very close, the O(3P) loss frequency is defined as (Booth et al 2019, Morillo-Candas et al 2019): νloss O=νrise O+νfall O 2.(6) The fits in equations (4) and (5) are the solution to the O(3P) continuity equation if it is written in the form (Booth and Sadeghi 1991): d[O(3P)] dt=S−L=A−νloss O·[O(3P)],(7) 3 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al where Sand Lrespectively represent all the O(3P) source and loss terms, and Arepresents a term which is independent of [O(3P)]. Since, during partial modulation, νloss Ois constant, then we can assume that the [O(3P)] temporal evolution is described by a net source term Aindependent of [O(3P)] and a net loss term B=νloss O·[O(3P)] linearly dependent on [O(3P)]. This is of course an approximation, as it is known, from the study of plasma volume kinetics (Dias et al 2023) and surface kinetics (Viegas et al 2024) that source and loss terms of O(3P) can depend non-linearly on its density (for instance, approximately quadratically in reactions with two O(3P) reactants). Nevertheless, the fact that equations (4) and (5) experimentally provide good fits for the [O(3P)] temporal evolution with constant νloss O, at least for partial modulation conditions, leads us to assume the form of equation (7) for the O(3P) continuity equation in the hypotheses in this work. It can be noticed from that equation that if the terms [O(3P)]min and [O(3P)]max correspond to a temporary steady-state, as is usually the case, then they are given by [O(3P)]min =Amin/νloss O and [O(3P)]max =Amax/νloss O. Thus, the independent source term Achanges with current, as can be expected to happen, for instance, to the electron-impact dissociation rate of O2, the most important source term of O(3P). To search how to interpret the measured νloss Oand how to calculate the effective O(3P) recombination probability γO, we assess what physical phenomena correspond to the independent and linearly dependent terms Aand B. We start by laying a set of alternative hypotheses: 1) If all the source terms are independent of [O(3P)] and all the loss terms are linearly dependent on [O(3P)], then A=Stot and B=Ltot. By dividing Ltot into surface and volume loss terms Lsurf and Lvol, the loss frequency and the recombination probability are given by: νloss O=Ltot [O(3P)] =Lvol +Lsurf [O(3P)] (8) γO=Lsurf ·2·R [O(3P)] ·vth =(νloss O·[O(3P)] −Lvol)·2·R [O(3P)] ·vth .(9) 2) By analyzing oxygen plasma kinetics (Dias et al 2023), different types of source terms of atomic oxygen can be distinguished, as O(3P) can be produced by electron impact reactions (with source term Sele), in particular by dissociation of O2, or by other forms of dissociation (with Sdiss), or by state exchanges (with Sstate), i.e. by reactions that convert O+, O−and mostly O(1D) into O(3P). From these, for relatively low dissociation fractions, dissociation source terms are expected to be approximately independent of [O(3P)], while state exchanges that depend on species created from O(3P) are expected to depend on its density. As such, the source term can be divided into independent and dependent sources as: Stot =Sind +Sdep;Sind ≃Sele +Sdiss; Sdep ≃Sstate. This hypothesis means that A=Sind and B= Ltot −Sdep. Then: νloss O=Ltot −Sdep [O(3P)] =Lvol +Lsurf −Sdep [O(3P)] (10) γO=Lsurf ·2·R [O(3P)] ·vth =(νloss O·[O(3P)] + Sdep −Lvol)·2·R [O(3P)] ·vth . (11) 3) By looking closer to O(3P) losses in oxygen plasma kinetics (Dias et al 2023), different types of loss terms can be distinguished. O(3P) can be lost by electron impact reactions (with loss term Lele), state exchanges (with Lstate), recombination reactions in volume (with Lrec) or at the surface, and through the convective axial flow (with Lflow). As such, the volume loss term can be written as Lvol = Lele +Lstate +Lrec +Lflow. The electron impact and state exchange losses of O(3P) mostly lead to the production of O(1D). On its turn, O(1D) is mostly lost due to de-excitation into O(3P), in what constitutes most of the O(3P) dependent source term. Hence, if we consider that O(1D) is produced from O(3P) excitation and neglect that it is also produced directly from O2electron impact dissociation, we conclude that Sstate ≃Lele +Lstate. This hypothesis leads to B=Ltot −Sdep ≃Lsurf +Lrec +Lflow, and thus: νloss O=Lsurf +Lrec +Lflow [O(3P)] (12) γO=Lsurf ·2·R [O(3P)] ·vth =(νloss O·[O(3P)] −Lrec −Lflow)·2·R [O(3P)] ·vth . (13) 4) The work by Booth et al (2019) analyzed O(3P) volume losses and concluded that the most important volume loss is the O+O+O2three-body recombination. That work, as well as that by Viegas et al (2024), thus used the approximation Lrec +Lflow ≃LO+O+O2, where this loss term can be estimated by knowing the experimental densities of O(3P) and O2(X) and the corresponding reaction rate coefficient kO+O+O2, that depends on the measured gas temperature, as LO+O+O2=2·[O(3P)]2·[O2(X)] ·kO+O+O2. From this hypothesis, B≃Lsurf +LO+O+O2, and: νloss O=Lsurf +LO+O+O2 [O(3P)] (14) γO=Lsurf ·2·R [O(3P)] ·vth =(νloss O·[O(3P)] −LO+O+O2)·2·R [O(3P)] ·vth . (15) 5) Finally, if it is assumed that surface recombination is by far the most important loss mechanism of O(3P), as is common in literature (Lopaev et al 2011, Morillo-Candas et al 2019, Ziganshin et al 2025), then the loss frequency and the recombination probability are simplified as: νloss O=Lsurf [O(3P)] (16) γO=Lsurf ·2·R [O(3P)] ·vth =νloss O·2·R vth .(17) 4 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Table 1. Summary of alternative hypotheses addressed in this work. Hypothesis νloss OγO 1Ltot [O(3P)] =Lvol+Lsurf [O(3P)] Lsurf·2·R [O(3P)]·vth =(νloss O·[O(3P)]−Lvol)·2·R [O(3P)]·vth 2Ltot−Sdep [O(3P)] =Lvol+Lsurf−Sdep [O(3P)] (νloss O·[O(3P)]+Sdep−Lvol)·2·R [O(3P)]·vth 3Lsurf+Lrec+Lflow [O(3P)] (νloss O·[O(3P)]−Lrec−Lflow)·2·R [O(3P)]·vth 4Lsurf+LO+O+O2 [O(3P)] (νloss O·[O(3P)]−LO+O+O2)·2·R [O(3P)]·vth 5Lsurf [O(3P)] νloss O·2·R vth The hypotheses are summarized in table 1. This work means to assess if any of these hypotheses is accurate. A conclusion on the accuracy of these hypotheses can tell us what is the physical meaning of νloss Oand which processes are responsible for the rise and fall of [O(3P)]. Moreover, it is instructive on how to accurately calculate a realistic condition-dependent effective recombination probability γOfrom the measured νloss O, which is very important for the interpretation of measurements and for input in models. 3. Methods and conditions 3.1. Discharge set-up and conditions The discharge conditions assessed in this work were obtained in a set-up similar to that studied experimentally and numerically by Booth et al (2019,2020,2022,2023), Western et al (2020), Dias et al (2023) and Viegas et al (2023,2024). The set-up is described in detail by Booth et al (2019,2020,2022, 2023). It consists of a DC glow discharge in O2, ignited in a Pyrex (borosilicate glass) tube of 2 mm thickness, 1.0 cm inner radius and 56 cm length. The discharge is ignited by electrodes located in side-arms separated by 52.5 cm. The conditions studied in this work correspond to the experimental data obtained by Dias (2019) and reported in the work by Viegas et al (2024), being referred to in that work as ‘data set 2’. They compose a total of 66 conditions with discharge current of either 20 mA or 40 mA and pressure values within the interval between 0.4 Torr (0.53 mbar) and 7.5 Torr (10 mbar). We should notice that the studied pressure range includes both the low (below 0.75 Torr) and high (above 0.75 Torr) pressure regimes identified by Booth et al (2019) and further studied by Afonso et al (2024), unlike the case in the work by Viegas et al (2024) where only the high pressure regime was addressed. The temperature of the cylindrical tube outer wall Twis kept constant at values of −20, 5, 25 or 50 ◦C. This is guaranteed by a water/ethanol mixture flowing through an outer envelope and connected to a thermostatic bath. The temperature drop across the Pyrex tube wall is considered to be negligible, of less than 2 K (Booth et al 2019). The gas flow rate is kept at 7.79 sccm and the air leak rate is expected to be very small, below 0.026 sccm. As a result, an approximately 52 cm long axially uniform and radially diffusive plasma column is studied with constant gas composition and cylindrical symmetry, in interaction with a cylindrical Pyrex vessel. The radially averaged gas temperature Tg,av was measured in the same way as by Booth et al (2019), from the rotational structure of the O2(b) optical emission at 762 nm, for Tw=5◦C and Tw=50 ◦C. For the remaining values of wall temperatures, Tg,av was calculated assuming a linear dependence of Tg,av with Tw. The near wall temperature Tnw was then obtained from the formula by Booth et al (2019). The electric field was measured in the same way as by Booth et al (2019), from voltage probes placed in the positive column of the glow discharge, and the reduced field E/Ng,av was then calculated from Tg,av and the ideal gas law. The radially-averaged density of atomic oxygen ground state O(3P) was measured via actinometry and its temporal variation was obtained from the temporal evolution of line emission intensity ratios, which led to obtaining the effective loss frequency of O atoms νloss O. The work by Dias et al (2023) shows that the experimental determination of the absolute value of [O(3P)]av and the corresponding gas fraction [O(3P)]av/Ng,av depend significantly on the experimental diagnostic (actinometry, VUV or CRDS), being variable within a 30% range. Likewise, it is highlighted by Viegas et al (2024) that different experimental campaigns for similar discharge conditions can obtain values of νloss Othat diverge up to a factor 3. Given that Lopaev and Smirnov (2004) report measurement errors below 5%, and following the experimental works by Booth et al (2019) and Dias (2019), it is assumed that the uncertainty of the νloss Omeasurement method is low. For that reason, we consider that the experimental uncertainty of νloss Oand [O(3P)]av/Ng,av is mostly due to reproducibility. To guide the eye, we add a 30% error bar to the measured quantities in the figures in this work. The experimental data set described here provides the input conditions (pressure, current, wall temperature, flow rate and loss frequency) for the models employed in this work and the experimental outputs (Tg,av,Tnw,E/Ng,av and [O(3P)]av) to compare with simulations. Despite the uncertainty associated to experimental reproducibility, the aims of this work of finding the most accurate way to interpret the measured νloss Oand finding the most accurate input parameters from experiments to models remain relevant for the understanding of this set of measurements and for future studies. 3.2. 0D global model Lisbon Kinetics Boltzmann solver chemistry solver (LoKI-B+C) The calculations in this work were performed with the LoKIB+C 0D (volume-averaged) simulation tool, composed by two coupled modules: a LoKI-B for the two-term electron Boltzmann equation (EBE) (Tejero-del Caz et al 2019,2021) and a LoKI-C for the heavy-species kinetics (Alves and del Caz 2023, Dias et al 2023, Liu et al 2025, LoKI 2025). LoKI-B (Tejero-del Caz et al 2019) solves a spaceindependent form of the EBE under the usual two-term assumption, for a Legendre expansion of the electron distribution function in velocity space. LoKI-C solves a system of 5 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al 0D rate-balance equations for all the heavy-species densities (ns) in the plasma. Given the axial uniformity and radial diffusivity of the positive column of the low-pressure oxygen glow discharge, its plasma parameters can be described via a global or volume-averaged approach. The work by Viegas et al (2023) showed that 0D and 1D-radial models using the same data provide results in close agreement for these discharges for pressures up to 10 Torr. Namely, in what concerns the main plasma parameters (gas temperature, electron density, reduced electric field and dissociation fraction), the relative differences between the simulation results were found to be below 5%. In the rate-balance equations, the source-loss terms due to chemical reactions in volume (Schem s), transport and reactions at the wall (Ssurf s) and axial convective gas flow (Sflow s) are considered as spatially averaged, according to: ∂ns ∂t=Ss−Ls=Schem s+Ssurf s+Sflow s.(18) It should be noticed that the rate-balance equation for the electrons is not solved, since the calculations are performed for a given electron density consistent with the experimental discharge current. The source-loss term Schem scomprises electronimpact reactions, where the rate coefficients from LoKI-B are used, and chemical reactions between heavy particles. The transport and loss of neutral species at the wall is given by the negative term Ssurf s, considering the wall as the Pyrex tube, thus neglecting the interaction with the electrodes, that have much lower surface area and are not in direct contact with the glow discharge positive column. This term is described for most species by the model proposed by Chantry (1987) and explained by Guerra et al (2019). This approach takes into account both the characteristic time of diffusion of the species from the reactor volume to the wall (τdiff s) and the probability (γs) of destruction at the wall, by either recombination or de-excitation. This consideration of diffusion assumes concave density profiles. Nevertheless, the 1D-radial simulation results from the work by Viegas et al (2023) shows that the radial profiles of O(3P) density are flat or convex. Thus, unlike the previous works on oxygen discharges, in this article we assume flat profiles for O(3P), with [O(3P)]nw =[O(3P)]av, rather than using the model by Chantry (1987) with concave profile assumption. As such, the flux of O(3P) to the Pyrex wall is simply the thermal flux, in coherence with previous works on surface kinetics (Viegas et al 2024) and with the formula relating γOwith surface loss rates Lsurf (equation (2)). It should be noted that with low γO(of the order of 10−4−10−3) the concavity predicted by the Chantry (1987) model is small, and thus the difference between the two models is negligible. The radial transport of positive ions is assumed to be given by ambipolar diffusion, with coefficients obtained self-consistently in LoKI-B+C. The impact of different charged-particle transport theories on the discharge kinetics is discussed and quantified in the work by Alves and del Caz (2023). Since negative ions O−have relevant densities in the studied conditions, of the order of half of the electron density (Dias et al 2023), their effect on the electron density profile is taken into account, following (Guerra and Loureiro 1999). The diffusion loss term for a positive ion of density np is thus written as function of its mobility µp, the electron characteristic energy uk, the electron charge e, the electron density ne, the negative ion density nn, the cylinder radius Rand the λ function (Guerra and Loureiro 1999) as: Ssurf p=−np µpuk e λ(1+ne nn+ne) R2.(19) Sflow sentails the source and loss terms due to flow of particles entering and exiting the reactor (Sflow s=Sflow,in s− Lflow,out s). The inflow rate Sflow,in sis defined by the experimental conditions and is zero for all species except O2(X). In previous works the outflow rate Lflow,out swas determined so as to conserve mass in the discharge volume (Silva et al 2021, Dias et al 2023). The discharge under study is experimentally observed to be very stable in both volume and pressure, for instance under partial modulation conditions. For a better computational performance and a better physical description, in this work the outflow rate of every species is calculated as guaranteeing the pressure known from experiments (isobaric approximation). This is done by ensuring that the pressure pis constant, i.e.: dp dt=kBTg,av dNg,av dt+kBNg,av dTg,av dt=0,(20) where kBis the Boltzmann constant and Ng,av is the sum of all species densities, with temporal evolution determined by equation (18). As such, Lflow,out sis determined so as to immediately ensure equation (20), which results in: Lflow,out s=nsνflow,out s=ns(1 Ng,av (∑ s Schem s+Ssurf s+Sflow,in s) +1 Tg,av dTg,av dt).(21) The time-scale of pressure reposition (pressure wave propagation) due to heating/cooling or changes in particle number due to reactions is expected to be of the order of the reason between the plasma length and the speed of sound, which in this work is around 0.525/343 ≃1.5 ms, and thus is not immediate as assumed in the model. Nevertheless, this time-scale is much lower than the time typically required to reach steady-state, of hundreds of ms, and thus the approximation of immediate pressure reposition is appropriate. In addition to the equations for the species densities, LoKIC also solves a thermal model for the self-consistent calculation of the average gas temperature Tg,av and the near-wall temperature Tnw, assuming a parabolic profile Tg(r)between the peak temperature T0and Tnw. The thermal model is described in detail by Viegas et al (2023) and Dias et al (2023), and is based on the works by Pintassilgo et al (2014) and Pintassilgo and Guerra (2016). In the work by Dias et al (2023) it was found that the use of a heat convection coefficient between the gas and the wall of 100 W·m−2·K−1provides an excellent agreement between simulated and measured values of Tg,av. 6 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Figure 1. Workflow of LoKI-B+C, reproduced from Silva et al (2021). © IOP Publishing Ltd All rights reserved. and reproduced from Dias et al (2023). © The Author(s). Published by IOP Publishing Ltd CC BY 4.0. See the text of section 3.2 for details. For that reason, that value is taken also in this work, along with the remaining transport and thermodynamic data from Dias et al (2023). The reaction mechanism for the positive column of the oxygen glow discharge employed in this work is the same one that was developed, presented in detail and validated in the work by Dias et al (2023). However, two differences should be noticed between the reaction mechanisms of the current work and of Dias et al (2023). One difference is that in this work no vibrational states of O2(X) are considered. These are excluded for a better computational performance. As shown by Dias et al (2023), the populations of these vibrationallyexcited states are low and their influence on the plasma chemistry is very limited. These states are only relevant for the gas temperature calculations, due to strong V–T transfers through collisions with O atoms (Dias et al 2023). Therefore, to ensure proper energy conservation between the electron and chemical kinetics modules, we assume that the energy spent on vibrational excitation of O2by electron impact directly contributes to gas heating. As such, even on the value of Tg,av the influence of the full inclusion of vibrations is lower than 3%. The other difference between this work and the one by Dias et al (2023) is the O(3P) recombination probability γO. In both cases, γOis an input value to the model, dependent on experimental conditions. Dias et al (2023) took γOfrom the experimental measurements of νloss Oby Booth et al (2020) and the application of hypothesis 4 from section 2. In this work, γOis taken from the experimental measurements of νloss Oby Dias (2019) and by applying different hypotheses from section 2. Firstly, a constant value is taken as input γO, to evaluate those hypotheses. Finally, the workflow of LoKI-B+C, presented in several works (Guerra et al 2019, Silva et al 2021,2024, Dias et al 2023), should be recalled. A schematic figure of the workflow is provided by Silva et al (2021) (figure 1) and by Dias et al (2023) (figure 1) and is reproduced here in figure 1. LoKI-B+C calculates a self-consistent solution of both electron and heavy-species kinetics in the plasma for user-defined working conditions: initial mixture composition, gas pressure, discharge current, reactor radius, length and flow rate. For initial guesses of the mixture, average gas temperature, reduced electric field and electron density, LoKI-B calculates the space and time independent EEDF, the electronimpact rate coefficients, the electron power transferred into the different collisional channels and the electron transport parameters. Then, this information is used in LoKI-C to calculate the species densities and the gas temperature until steadystate is reached. If the steady-state E/Ng,av does not guarantee that the ion densities match the guessed electron density, its value, employed in the EBE solver, is adjusted so as to achieve quasi-neutrality in steady-state. Additionally, consistency is assured by updating the steady-state gas mixture and temperature calculated by LoKI-C into LoKI-B. Finally, the electron density is varied so as to obtain a calculated discharge current (I=evdπR2ne, where eis the electron charge and vdis the calculated drift velocity) equal to the experimental value given as input. In this work, the relative tolerances for quasi-neutrality, mixture composition and electron density cycles are all 10−3. Moreover, unlike previous works, no pressure cycle is considered here, since isobaric conditions are guaranteed by the outflow rate, which significantly speeds up computations. 3.3. Time-resolved model for modulated current In this work, a model is adapted from LoKI-B+C to be used not only in steady-state, but also for experimental conditions of time-resolved modulated current, i.e. with the experimental current profile I(t) taken as input. In this case, the reduced electric field and the gas mixture are time-dependent and not unique for each experimental condition. Nevertheless, for each condition, the steady-state solution is known by solving LoKIB+C in steady-state. For partially modulated current conditions, typically with current variations of the order of 20%, the EBE solution changes only slightly, for a given E/Ng,av, between the mixtures corresponding to the minimum and maximum current values. As such, in this work, we take the tabulated solution of the EBE through LoKI-B using the quasistationary assumption for a wide range of E/Ng,av for the mixture corresponding to the result of each steady-state simulation. This means that the coupling between LoKI-B and LoKIC is not fully self-consistent in this time-dependent model, but is computationally effective. For a given set of initial densities of neutral and charged species and of initial temperatures, the equations for temperatures and densities of every species are solved in time in isobaric conditions, as described in the previous section. The 7 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Figure 2. Temporal evolution of atomic oxygen density and reduced electric field, for conditions of wall temperature Tw=5◦C, pressure p=2 Torr, partially modulated current around I=40 mA (between 35 and 45 mA) and γO=1.226 ×10−3. (a): Density of O(3P) and E/ng,av during multiple cycles. (b): Fits of rise and fall curves of [O(3P)] resulting in an average νloss O=50.45 s−1. solution requires knowledge of E/Ng,av(t)at each time-step, for the computation of the electron-impact rate coefficients and the electron transport parameters. E/Ng,av(t)is found from current conservation and from the tabulation of the electron drift velocity as function of E/Ng,av(t),vd(E/Ng,av(t)), such that: vd(E Ng,av (t))=I(t) e·ne(t)·πR2.(22) Figure 2shows an example of the results obtained for an input current partially modulated between 35 and 45 mA, i.e. around I=40 mA, for a particular case with Tw=5◦C and p=2 Torr. The current takes the maximum value during 0.2 s and the minimum value for another 0.2 s, repetitively, as in experiments (Booth et al 2019, Dias 2019). Figure 2(a) represents the temporal evolution of the O(3P) density and of E/Ng,av along multiple current cycles, clearly showing that results are repeatable. As shown in the figure, the variations of E/Ng,av in time are sharp when there are shifts of current but smooth during the 0.2 s half-periods. As such, a time-dependent resolution of the EBE (Tejero-del Caz et al 2021) should only affect the results in the first instances after the current shifts, and the quasi-stationary assumption for the EBE resolution is justified. This is in agreement with the conclusions taken from the comparison of time-dependent and quasi-stationary resolution of the EBE in pulsed plasmas by Tejero-del Caz et al (2021). That work concluded that, for 1 Torr and dry-air, the results are similar for rise-times longer than the characteristic evolution time of the EEDF, of 20 µs. Figure 2(b) zooms in on a particular rise and fall cycle of [O(3P)] and shows the fits following equations (4) and (5), that match perfectly the density evolution, along with the fitted values of O(3P) rise and fall frequencies. As is visible, for small modulation amplitudes, νloss Ofrom rise and fall fits are very close, as in experiments (Booth et al 2019). Density jumps are visible in figure 2every time the current shifts between its minimum and maximum values, which is also the case in experiments (Dias 2019, Morillo-Candas et al 2019). This is due to sudden changes of temperature induced by the current modulation. In isobaric conditions, these produce sudden changes of densities. This explanation is confirmed by the fact that the measured temporal evolution of the intensity line ratios shown by Morillo-Candas et al (2019) and Dias (2019) presents these oscillations, but that of the fraction [O(3P)]av/Ng,av presented by Booth et al (2019) does not. Although the oscillations can be removed by examining the O(3P) fraction instead of its density and the simulated values of density and fraction loss frequencies are close (within a few %), in this work we analyze the loss frequencies of O(3P) density and not of its fraction, and thus we consistently assess [O(3P)]av. Instead of removing the oscillations by assessing fractions, in this work the fits of the numerical results leave out the first and last 5 ms of the 0.2 s half-periods, thus excluding the density jumps, as can be seen in figure 2(b). The fact that the oscillations are apparently produced in the same time-scale (a few ms) in experimental and numerical results confirms the appropriateness of our approach guaranteeing isobaric conditions, following equations (20) and (21). Since the time-scale of relevance in the analysis of the modulated current results is of 200 ms, much higher than the expected time-scale for pressure wave propagation, we find that the immediate pressure reposition approximation is valid also for the study of the time-dependent results. Finally, it has been confirmed that the results have only a small dependence on the choice of the current modulation amplitude. For instance, for the case of figure 2, applying current variations between 30 and 50 mA, instead of 35 and 45 mA, only decreases νloss Oby 0.1 s−1. Likewise, applying current variations between 38 and 42 mA only increases νloss O by 0.1 s−1. In this work, for 40 mA current cases, we choose to modulate the current between 35 and 45 mA. For 20 mA current cases, a modulation between 17.5 mA and 22.5 mA is taken. These choices of modulation are close to those in experiments (Booth et al 2019, Dias 2019, Morillo-Candas et al 2019). 8 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Figure 9. Atomic oxygen loss frequency νloss O, as function of pressure, for (a) Tw=25 ◦C and I=20 mA, (b) Tw=25 ◦C and I=40 mA, (c) Tw=50 ◦C and I=20 mA and (d) Tw=50 ◦C and I=40 mA. Simulation results obtained with an input γOestimated from the measured νloss Oby employing hypothesis 3 (see section 2). Measurements compared to corrected modulated simulation results and steady-state simulation results applying hypothesis 3. rate, it is relevant to discuss this paradox and to recognize that the validation of the oxygen plasma kinetics model is not complete. The first aspect to highlight is the uncertainty in loss frequency measurements. For similar borosilicate glass (Pyrex) tubes, under similar conditions, resulting in close O(3P) density and fraction measurements, and employing the same diagnostics, Booth et al (2019) and Booth et al (2020) find approximately factor 3 differences in νloss O. This is attributed by Booth et al (2020) to the dependence of νloss Oon the history of the surface material (Cartry et al 1999). Booth et al (2023) state that the dependence is especially a function of the discharge running time after an overnight shut-down. More generally, for models, this means that this high uncertainty should be taken into account. Another aspect to consider are the uncertainties in the data used in the model for the processes determining O(3P) sources and losses, from which we should highlight O2(X) electron-impact dissociation. The cross sections for these processes are taken in this reaction scheme from the IST-Lisbon database on LXCat (Pancheshnyi et al 2012, Alves 2014, Alves et al 2016), and are based on the work by Phelps (1985). By browsing through the LXCat databases, it is easy to verify that there are other dissociation cross section proposals: in the Itikawa database by Itikawa (2009); in the Phelps database attributed to Cosby (1993); in the Phelps database by Lawton and Phelps (1978) attributed to Hayashi; in the Triniti database by Ionin et al (2007) attributed to Klopovskii and Rakhimova. It has been pointed out by Kovalev et al (2005) that the dissociation cross-section reported by Phelps (1985) is likely to be overestimated. Both these uncertainties can justify the disparities observed in this section. Moreover, the model can be improved to include the surface production and destruction of ozone, studied recently by Booth et al (2023), Meyer et al (2023), Viegas et al (2024)and Mazánková et al (2020,2024). To complete the oxygen glow discharge model validation, one way to proceed in the future could be to separately validate surface recombination and O(3P) volume kinetics in conditions of known surface losses. This could be done by covering the reactor walls with fibres, guaranteeing γO=1 and then measuring O(3P) density via absorption techniques, but there would be the risk of a too low density. Alternatively, by igniting the glow discharge in a larger tube, there would be the possibility of keeping O(3P) far from the walls and guaranteeing that surface losses are negligible when compared to volume losses. Then, the measurement of O(3P) density would allow a direct comparison with a model based exclusively on volume kinetics and transport. With a validated volume kinetics, the model could assess which measurement of νloss Owould, by applying hypothesis 3, provide the input γOthat allows to 15 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Figure 10. Fraction of the terms in equation (12), from steady-state simulation results. Results as function of pressure, for both I=20mA and I=40mA, for (a) Tw=−20 ◦C, (b) Tw=5◦C, (c) Tw=25 ◦C and (d) Tw=50 ◦C. Simulation results obtained with an input γO estimated from the measured νloss Oby employing hypothesis 3 (see section 2). Figure 11. Temporal evolution of atomic oxygen density and corresponding fit in post-discharge, for conditions of wall temperature Tw=50 ◦C, pressure p=2 Torr, current I=40 mA and γO=1.78 ×10−4. match the measured O(3P) density and fraction. The model validation would of course be facilitated if the νloss Omeasurements were reproducible. In this respect, after concluding that νloss Ois dependent mostly on the discharge running time after an overnight shut-down, Booth et al (2023) and Ziganshin et al (2025) state that they operate the discharge for at least an hour before kinetic measurements are taken, as a way to achieve reproducibility. Nevertheless, Booth et al (2023) assessed γO in early post-discharge and the values are not directly comparable with those by Booth et al (2019,2020)and Dias (2019). Ziganshin et al (2025) addressed similar conditions to this work, with small admixtures of noble gases, and found measurements closer to those by Booth et al (2019) and Dias (2019) than to those by Booth et al (2020). 16 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Figure 12. Atomic oxygen fraction for the different conditions of current and wall temperature, as function of pressure. Measurements by Dias (2019) compared to steady-state simulation results employing the input γOfrom this work (hypothesis 3). The justification for the 30% errorbar in the measurements is given in the text. 5. Conclusions In this work, a global model describing plasma chemical kinetics is used in steady-state and current-modulated modes to address experimental conditions of O loss frequency (νloss O) measurement in the positive column of oxygen glow discharges in Pyrex, for a total of 66 conditions with pressures between 0.4 and 7.5 Torr, currents of 20 and 40 mA, and wall temperatures between −20◦C and 50◦C. The simulation results allow to derive the most realistic model-dependent and condition-dependent recombination probability γOPT that leads to matching νloss Ofrom simulations approaching the real measurements, i.e. from partial current modulation simulations, with the measured νloss O. Furthermore, the simulations are used to assess the accuracy of different hypotheses on the relevant processes for the net losses of O(3P) in current modulation conditions. The verification started by comparing a fixed surface recombination probability (γO), used as input for simulations, with γOobtained from the treatment of the simulation results under the different hypotheses. It then proceeded by comparing, for the set of 66 experimental conditions, νloss Ofrom steady-state simulations under the different hypotheses with νloss Ofrom current modulation simulations. The accurate interpretation of νloss Oand estimation of γOis relevant for experimental characterization and for model inputs. On its turn, the accurate simulation of O(3P) losses is important for model validation and predictability. The hypothesis that is verified to be the most accurate (hypothesis 3) indicates that the net loss of O(3P) during current modulation is due not only to surface recombination, but also to recombination in volume and to flow losses. It highlights the importance of volume processes for O(3P) losses. For the working conditions studied, these account on average for 24% of O(3P) net losses, and that proportion tendentiously grows with pressure, reaching up to 71%. The importance of considering volume processes for the O(3P) net loss grows with pressure and is expected to be higher at pressures above 7.5 Torr, and becomes negligible at pressures below 1 Torr, where hypothesis 3 becomes convergent with the hypotheses neglecting volume kinetics. The verification of hypothesis 3 changes the estimation of γOfrom the measured νloss O, which is relevant for experimental characterization and for model inputs. In previous works (Booth et al 2019, Viegas et al 2024), γOwas overestimated by considering an hypothesis accounting only for surface recombination and O+O+O2recombination. The simulations in this work found the rate of the latter reaction to compose on average 53% of volume recombination rates. As such, γOis corrected by applying hypothesis 3. The steady-state simulation and current modulation simulation results employing the corrected γOare shown to be coherent with each other and to agree well with the experimental measurements of νloss O. This work suggests that γOPT is the most realistic modeldependent recombination probability that can be derived from 17 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al the measured νloss O. Nevertheless, γOPT cannot be extracted in the absence of a complete validated volume kinetics model. Hence, this work recommends the adoption of hypothesis 3 for the estimation of γOfrom the measured νloss O, as long as sufficient experimental or simulation data is available to perform that calculation (which mostly requires O, O2and O3 densities, gas temperature and rate coefficients). The results allow to conclude how to compare simulations with input γOwith νloss Omeasurements. Indeed, νloss O can be obtained not only from modulation simulations, but also from steady-state simulations by applying hypothesis 3 as post-treatment, and the obtained values can be compared with νloss Omeasurements. Furthermore, it is confirmed that hypothesis 3 is also accurate for the treatment of νloss Oand γOin post-discharge conditions. According to the simulations in this work, if νloss Ois measured in full modulation experiments, γO for post-discharge can be accurately estimated by applying hypothesis 3. Finally, it is pointed out that the oxygen glow discharge model validation is not complete, since, with the current model and measurements, the input surface recombination probability γOthat allows simulations to match the measured O(3P) loss frequency νloss Ois different from that which allows to find a good agreement with the measured O(3P) density or fraction. A discussion is had on what steps can be taken for a complete model validation that builds up on the validation undergone by Dias et al (2023). Using the approach in this work, of describing atomic oxygen surface recombination via an effective recombination probability γO, implies setting different input γOfor different conditions of wall temperature, pressure, current and modulation (partial modulation corresponding to discharge conditions and full modulation to post-discharge). Those different input γOcan be estimated, either by optimization or by applying hypothesis 3, but only if νloss Ois measured in the corresponding conditions. This implies that the model with input γOis dependent on reliable measurements for each condition under study and hence is not validated for use in different conditions. Alternatively, atomic oxygen surface adsorption and recombination can be described in a global model, such as the one in this work, via the inclusion of mesoscopic modeling of surface kinetics, as recently developed and validated by Afonso et al (2024) and Viegas et al (2024). Besides being more independent of νloss Omeasurements, the mesoscopic simulation of surface kinetics, by either deterministic or Monte Carlo methods, also allows to assess the importance and dependencies of the different surface recombination mechanisms, improving the understanding of plasma-surface interactions. The verification of hypothesis 3 for the treatment of νloss Owith surface kinetics modeling instead of input γO will be the object of future work. In future work, we would also like to go beyond the pure oxygen plasma and assess the physical meaning of νloss Oand γOin other oxygen-containing plasmas, such as in the CO2glow discharge. In particular, it would be interesting to verify the accuracy of hypothesis 3 in that case. Data availability statement All data that support the findings of this study are included within the article (and any supplementary files). Acknowledgment This work was supported by the Portuguese FCT—Fundaç˜ ao para a Ciˆ encia e a Tecnologia, under project reference UIDB/50010/2020 (https://doi.org/10.54499/UIDB/50010/ 2020), project reference UIDP/50010/2020 (https://doi. org/10.54499/UIDP/50010/2020) and project reference LA/P/0061/2020 (https://doi.org/10.54499/LA/P/0061/2020). This work was further supported by FCT under project SYAMESE (https://doi.org/10.54499/2023.15276.PEX) and by the European Union under Horizon Europe project CANMILK (DOI:https://doi.org/10.3030/101069491). PV acknowledges support by project CEECIND/00025/2022 of FCT. The authors acknowledge useful discussions with Blandine Berdugo, Dr Olivier Guaitella and Dr Jean-Paul Booth from Laboratoire de Physique des Plasmas. ORCID iDs Pedro Viegas 0000-0002-3820-3300 Tiago Cunha Dias 0000-0002-2179-1345 Vasco Guerra 0000-0002-6878-6850 References Afonso J, Vialetto L, Guerra V and Viegas P 2024 Plasma-induced reversible surface modification and its impact on oxygen heterogeneous recombination J. Phys. D: Appl. Phys. 57 04LT01 Alves L, Coche P, Ridenti M A and Guerra V 2016 Electron scattering cross sections for the modelling of oxygen-containing plasmas Eur. Phys. J. D 70 124 Alves L L 2014 The IST-Lisbon database on LXCat J. Phys.: Conf. Ser. 565 012007 Alves L L and del Caz A T 2023 Charged-particle transport models for global models Plasma Sources Sci. Technol. 32 054003 Annuˇ sová A, Marinov D, Booth J P, Sirse N, da Silva M L, Lopez B and Guerra V 2018 Kinetics of highly vibrationally excited O2(X) molecules in inductively-coupled oxygen plasmas Plasma Sources Sci. Technol. 27 045006 Balat-Pichelin M, Badie J M, Berjoan R and Boubert P 2003 Recombination coefficient of atomic oxygen on ceramic materials under earth re-entry conditions by optical emission spectroscopy Chem. Phys. 291 181–94 Bogaerts A, Neyts E C, Guaitella O and Murphy A B 2022 Foundations of plasma catalysis for environmental applications Plasma Sources Sci. Technol. 31 053002 Booth J P and Sadeghi N 1991 Oxygen and fluorine atom kinetics in electron cyclotron resonance plasmas by time-resolved actinometry J. Appl. Phys. 70 611–20 Booth J P, Guaitella O, Chatterjee A, Drag C, Guerra V, Lopaev D, Zyryanov S, Rakhimova T, Voloshin D and Mankelevich Y 2019 Oxygen (3P) atom recombination on a pyrex surface in an O2plasma Plasma Sources Sci. Technol. 28 055005 18 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al Booth J-P et al 2020 Determination of absolute O(3P) and O2(a1∆g) densities and kinetics in fully modulated O2dc glow discharges from the O2(X3Σ− g) afterglow recovery dynamics Plasma Sources Sci. Technol. 29 115009 Booth J P et al 2022 Quenching of O2(b1Σ+ g) by O(3P) atoms. Effect of gas temperature Plasma Sources Sci. Technol. 31 065012 Booth J-P, Guaitella O, Zhang S, Lopaev D, Zyryanov S, Rakhimova T, Voloshin D, Chukalovsky A, Volynets A and Mankelevich Y 2023 Oxygen atom and ozone kinetics in the afterglow of a pulse-modulated DC discharge in pure O2: an experimental and modelling study of surface mechanisms and ozone vibrational kinetics Plasma Sources Sci. Technol. 32 095016 Bousquet A, Cartry G and Granier A 2007 Investigation of O-atom kinetics in O2, CO2, H2O and O2/HMDSO low pressure radiofrequency pulsed plasmas by time-resolved optical emission spectroscopy Plasma Sources Sci. Technol. 16 597–605 Caplinger J E and Perram G P 2020 The importance of cascade emission and metastable excitation in modeling strong atomic oxygen lines in laboratory plasmas Plasma Sources Sci. Technol. 29 015011 Cartry G, Magne L and Cernogora G 1999 Atomic oxygen recombination on fused silica: experimental evidence of the surface state influence J. Phys. D: Appl. Phys. 32 L53–L56 Cartry G, Magne L and Cernogora G 2000 Atomic oxygen recombination on fused silica: modelling and comparison to low-temperature experiments (300 K) J. Phys. D: Appl. Phys. 33 1303–14 Cartry G, Duten X and Rousseau A 2006 Atomic oxygen surface loss probability on silica in microwave plasmas studied by a pulsed induced fluorescence technique Plasma Sources Sci. Technol. 15 479–88 Casas P A F 2023 Continuous process for validation, verification and accreditation of simulation models Mathematics 11 845 Chantry P J 1987 A simple formula for diffusion calculations involving wall reflection and low density J. Appl. Phys. 62 1141–8 Chen G, Buck F, Kistner I, Widenmeyer M, Schiestel T, Schulz A, Walker M and Weidenkaff A 2020 A novel plasma-assisted hollow fiber membrane concept for efficiently separating oxygen from CO in a CO2plasma Chem. Eng. J. 392 123699 Cosby P C 1993 Electron-impact dissociation of oxygen J. Chem. Phys. 98 9560–9 Dias T C 2019 Atomic oxygen kinetics in CO2plasmas MSc Thesis Instituto Superior Técnico - Universidade de Lisboa Dias T C, Fromentin C, Alves L L, del Caz A T, Silva T and Guerra V 2023 A reaction mechanism for oxygen plasmas Plasma Sources Sci. Technol. 32 084003 Fiebrandt M, Bibinov N and Awakowicz P 2020 Determination of atomic oxygen state densities in a double inductively coupled plasma using optical emission and absorption spectroscopy and probe measurements Plasma Sources Sci. Technol. 29 045018 Gordiets B, Ferreira C M, Nahorny J, Pagnon D, Touzeau M and Vialle M 1996 Surface kinetics of N and O atoms in N2-O2 discharges J. Phys. D: Appl. Phys. 29 1021–31 Gordiets B F and Ferreira C M 1998 Self-consistent modeling of volume and surface processes in air plasma AIAA J. 36 1643–51 Gousset G, Touzeau M, Vialle M and Ferreira C M 1989 Kinetic model of a DC oxygen glow discharge Plasma Chem. Plasma Process. 9189–206 Guerra V 2007 Analytical model of heterogeneous atomic recombination on silicalike surfaces IEEE Trans. Plasma Sci. 35 1397–412 Guerra V and Loureiro J 1999 Kinetic model of a low-pressure microwave discharge in O2-N2including the effects of O−ions on the characteristics for plasma maintenance Plasma Sources Sci. Technol. 8110–24 Guerra V and Loureiro J 2004 Dynamical monte carlo simulation of surface atomic recombination Plasma Sources Sci. Technol. 13 85–94 Guerra V and Marinov D 2016 Dynamical Monte Carlo methods for plasma-surface reactions Plasma Sources Sci. Technol. 25 045001 Guerra V, del Caz A T, Pintassilgo C D and Alves L L 2019 Modelling N2-O2plasmas: volume and surface kinetics Plasma Sources Sci. Technol. 28 073001 Ionin A A, Kochetov I V, Napartovich A P and Yuryshev N N 2007 Physics and engineering of singlet delta oxygen production in low-temperature plasma J. Phys. D: Appl. Phys. 40 R25–R61 Itikawa Y 2009 Cross sections for electron collisions with oxygen molecules J. Phys. Chem. Ref. Data 38 1–20 Kersten H, Deutsch H, Steffen H, Kroesen G M W and Hippler R 2001 The energy balance at substrate surfaces during plasma processing Vacuum 63 385–431 Kim Y C and Boudart M 1991 Recombination of oxygen, nitrogen and hydrogen atoms on silica: kinetics and mechanism Langmuir 72999–3005 Kovalev A S, Lopaev D V, Mankelevich Y A, Popov N A, Rakhimova T V, Poroykov A Y and Carroll D L 2005 Kinetics of O2(b1Σ+ g) in oxygen RF discharges J. Phys. D: Appl. Phys. 38 2360–70 Kriˇ stof J, Macko P and Veis P 2012 Surface loss probability of atomic oxygen Vacuum 86 614–9 Kushner M J 2009 Hybrid modelling of low temperature plasmas for fundamental investigations and equipment design J. Phys. D: Appl. Phys. 42 194013 Lawton S A and Phelps A V 1978 Excitation of the b1Σ+ gstate of O2by low energy electrons J. Chem. Phys. 69 1055–68 Liu Y, Silva T, Dias T C, Viegas P, Zhao X, Du Y, He J and Guerra V 2025 Gas heating and plasma chemistry in low-pressure CO2plasmas J. CO2Util. 97 103128 LoKI 2025 Lisbon KInetics (LoKI) by N-PRiME (available at: nprime.tecnico.ulisboa.pt/loki/) Lopaev D V and Smirnov A V 2004 Diagnostics of heterogeneous processes with the participation of radicals by time-resolved actinometry Plasma Phys. Rep. 30 882–93 Lopaev D V, Malykhin E M and Zyryanov S M 2011 Surface recombination of oxygen atoms in O2plasma at increased pressure: I. The recombination probability and phenomenological model of surface processes J. Phys. D: Appl. Phys. 44 015201 Macko P, Veis P and Cernogora G 2004 Study of oxygen atom recombination on a Pyrex surface at different wall temperatures by means of time-resolved actinometry in a double pulse discharge technique Plasma Sources Sci. Technol. 13 251–62 Magne L, Coitout H, Cernogora G and Gousset G 1993 Atomic oxygen recombination at the wall in a time afterglow J. Phys. III 31871–89 Marinov D, Teixeira C and Guerra V 2017 Deterministic and Monte Carlo methods for simulation of plasma-surface interactions Plasma Process. Polym. 14 1600175 Mazánková V, Trunec D, Krzyˇ zánková A, Jurmanová J and Krcˇma F 2020 Study of oxygen and ozone reactions on surfaces previously treated by ozone Japan. J. Appl. Phys. 59 SHHA02 Mazánková V, Trunec D, Staneˇk F and Krcˇma F 2024 Study of ozone destruction reaction on surfaces Plasma Phys. Technol. 11 36–38 Meiss S A, Rohnke M, Rettig F, Moos R and Janek J 2008 Ion-conducting probes for low temperature plasmas Contrib. Plasma Phys. 48 473–9 Meyer M, Foster J and Kushner M J 2023 A surface mechanism for O3production with N2addition in dielectric barrier discharges Plasma Sources Sci. Technol. 32 085001 Morillo-Candas A S, Drag C, Booth J-P, Dias T C, Guerra V and Guaitella O 2019 Oxygen atom kinetics in CO2plasmas 19 Plasma Sources Sci. Technol. 34 (2025) 085003 P Viegas et al ignited in a DC glow discharge Plasma Sources Sci. Technol. 28 075010 Morillo-Candás A S 2019 Investigation of fundamental mechanisms of CO2plasmas PhD Thesis ´ Ecole Polytechnique (available at: www.theses.fr/2019SACLX091) Murphy A B and Park H 2017 Modeling of thermal plasma processes: the importance of two-way plasma-surface interactions Plasma Process. Polym. 14 1600177 Oehrlein G S 1997 Surface processes in low pressure plasmas Surf. Sci. 386 222–30 Ono R and Kimata Y 2024 Surface loss probability of atomic oxygen on silica under atmospheric-pressure CO2J. Phys. D: Appl. Phys. 57 505202 Pagnon D, Amorim J, Nahorny J, Touzeau M and Vialle M 1995 On the use of actinometry to measure the dissociation in O2DC glow discharges: determination of the wall recombination probability J. Phys. D: Appl. Phys. 28 1856–68 Pancheshnyi S, Biagi S, Bordage M, Hagelaar G, Morgan W, Phelps A and Pitchford L 2012 The LXCat project: Electron scattering cross sections and swarm parameters for low temperature plasma modeling Chem. Phys. 398 148 Pandiyan A, Kyriakou V, Neagu D, Welzel S, Goede A, van de Sanden M C M and Tsampas M N 2022 CO2conversion via coupled plasma-electrolysis process J. CO2Util. 57 101904 Paul D, Mozetic M, Zaplotnik R, Primc G, Donlagic D and Vesel A 2023 A review of recombination coefficients of neutral oxygen atoms for various materials Materials 16 1774 Phelps A 1985 Tabulations of collision cross sections and calculated transport and reaction coefficients for electron collisions with O2(JILA Information Center) p 28 Pintassilgo C D and Guerra V 2016 Power transfer to gas heating in pure N2and in N2-O2plasmas J. Phys. Chem. C 120 184–201 Pintassilgo C D, Guerra V, Guaitella O and Rousseau A 2014 Study of gas heating mechanisms in millisecond pulsed discharges and afterglows in air at low pressures Plasma Sources Sci. Technol. 23 025006 Rakhimova T V, Braginsky O V, Kovalev A S, Lopaev D V, Mankelevich Y A, Malykhin E M, Rakhimov A T, Vasilieva A N, Zyryanov S M and Baklanov M R 2009 Recombination of O and H atoms on the surface of nanoporous dielectrics IEEE Trans. Plasma Sci. 37 1697–704 Rohnke M, Janeka J, Kilner J A and Chater R J 2004 Surface oxygen exchange between yttria-stabilised zirconia and a low-temperature oxygen rf-plasma Solid State Ion. 166 89–102 Silva A F, Morillo-Candás A S, del Caz A T, Alves L L, Guaitella O and Guerra V 2021 A reaction mechanism for vibrationally-cold low-pressure co2plasmas Plasma Sources Sci. Technol. 29 125020 Silva T, Bera S, Pintassilgo C D, Herrmann A, Welzel S, Tsampas M N, van de Sanden M C M, Alves L L and Guerra V 2024 Unraveling NO production in N2-O2plasmas with 0D kinetic modeling and experimental validation J. Phys. Chem. A 128 7235–56 Tejero-del Caz A, Guerra V, Gonçalves D, da Silva M L, Marques L, Pinh˜ ao N, Pintassilgo C D and Alves L L 2019 The LisbOn KInetics Boltzmann solver Plasma Sources Sci. Technol. 28 043001 Tejero-del Caz A, Guerra V, Pinh˜ ao N, Pintassilgo C D and Alves L L 2021 On the quasi-stationary approach to solve the electron boltzmann equation in pulsed plasmas Plasma Sources Sci. Technol. 30 065008 van de Steeg A, Viegas P, Silva A, Butterworth T, van Bavel A, Smits J, Diomede P, van de Sanden M and van Rooij G 2021a Redefining the microwave plasma-mediated CO2reduction efficiency limit: The role of O-CO2association ACS Energy Lett. 62876–81 van de Steeg A W, Vialetto L, Silva A F, Peeters F J J, van den Bekerom D C M, Gatti N, Diomede P, van de Sanden M C M and van Rooij G J 2021b Revisiting spontaneous Raman scattering for direct oxygen atom quantification Opt. Lett. 46 2172–5 van de Steeg A W, Vialetto L, da Silva A F S, Viegas P, Diomede P, van de Sanden M C M and van Rooij G J 2022 The chemical origins of plasma contraction and thermalization in CO2 microwave discharges J. Phys. Chem. Lett. 13 1203–8 Vanraes P, Venugopalan S P and Bogaerts A 2021 Multiscale modeling of plasma-surface interaction-general picture and a case study of Si and SiO2etching by fluorocarbon-based plasmas Appl. Phys. Rev. 8041305 Viegas P, Vialetto L, van de Steeg A W, Wolf A J, Bongers W A, van Rooij G J, van de Sanden M C M, Diomede P and Peeters F J J 2021 Resolving discharge parameters from atomic oxygen emission Plasma Sources Sci. Technol. 30 065022 Viegas P, Dias T C, Fromentin C, Chukalovsky A, Mankelevich Y, Proshina O, Rakhimova T, Guerra V and Voloshin D 2023 Comparison between 1D radial and 0D global models for low-pressure oxygen DC glow discharges Plasma Sources Sci. Technol. 32 024002 Viegas P, Silveira J, Dias T C, Guaitella O, Morillo-Candás A S and Guerra V 2024 Surface recombination in Pyrex in oxygen DC glow discharges: mesoscopic modelling and comparison with experiments Plasma Sources Sci. Technol. 33 055003 Western C M, Booth J-P, Chatterjee A and de Oliveira N 2020 Rydberg spectra of singlet metastable states of O2Mol. Phys. 119 1–2 Ziganshin I I, Galiullin K R, Lopaev D V, Kirillov E A and Rakhimov A T 2025 VUV radiation impact on O(3P) atom surface loss on Pyrex in DC discharge in mixtures of O2with Ar, Kr and Xe Plasma Sources Sci. Technol. 34 035007 20