Ar(1s5) density in a co-axial argon plasma jet with N2–O2 shielding

Atmospheric-pressure microplasma jets (µAPPJs) are versatile sources of reactive species with diverse applications. However, understanding the plasma chemistry in these jets is challenging due to plasma-flow interactions in heterogeneous gas mixtures. Spatial metastable density profiles help to understand these physical and chemical mechanisms. This work focuses on controlling the shielding gas around a µAPPJ. We use a dielectric barrier discharge co-axial reactor where a co-flow shields the pure argon jet with different N2–O2 gas mixtures. A voltage pulse (4 kV, 1 µs, 20 kHz) generates a first discharge at the pulse’s rising edge and a second discharge at the falling edge. Tunable diode laser absorption spectroscopy measures the local Ar(1s5) density. A pure N2 (100%N2–0%O2) co-flow leads to less reproducible and lower peak Ar(1s5) density ( 5.8×1013cm−3 ). Increasing the O2 admixture in the co-flow yields narrower Ar(1s5) absorbance profiles and increases the Ar(1s5) density ( 6.9×1013 to 9.1×1013cm−3 ). The position of the peak density is closer to the reactor for higher O2 fractions. Absence of N2 results in comparable Ar(1s5) densities between the first and second discharges (maxima of 9.1×1013 and 9.3×1013cm−3 , respectively). Local Ar(1s5) density profiles from pure N2 to pure O2 shielding provide insights into physical and chemical processes. The spatially-resolved data may contribute to optimising argon µAPPJ reactors across the various applications and to validate numerical models.


Introduction
Low-temperature atmospheric-pressure microplasma jets (µAPPJs) extend plasma beyond physical barriers, providing a controlled method of plasma delivery [1].µAPPJs applications benefit from their portability and flexibility of gas mixtures, influencing the associated chemistry.The access to nonequilibrium plasma chemistry at ambient pressure (1 atm) and 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.
Noble gases also help to produce cooler plasma jets.Compared to molecular gases, like N 2 and O 2 , noble gases allow a lower operating gas temperature [31] by exhibiting energy transfer rates from electronically excited states to the gas translation mode lower than those observed from vibrationally and rotationally excited states of molecules.Moreover, noble gases facilitate breakdown at lower reduced electric fields due to higher energy tails in the electron energy distribution function (low-energy electronic and vibrational excited states of N 2 and O 2 deplete the tail), reduced recombination rates, and the occurrence of low-energy ionisation processes through step-wise ionisation of metastable states [32].Even so, there is considerable energy stored in the plasma due to the presence of highly excited species (e.g.Ar * at 11.55-15.75eV), which contribute to producing reactive oxygen and nitrogen species relevant for applications.
The unconfined nature of a plasma jet leads to interactions with the surrounding environment.In an argon µAPPJ, Gazeli et al [24] showed varying metastable density and lifetime with different gas flow rates.A high flow rate might induce turbulence, subsequently influencing the trajectory of the IWs [33] and increasing the quenching of excited species due to an enhanced mixing [34,35].The geometry of the reactor can also influence the IWs [11,19,32], potentially giving rise to plasma-induced flow perturbations [36,37].In co-axial reactors, factors like nozzle shape, flow rate, and co-flow rate are crucial in determining turbulence and gas mixing [38].Moreover, in such reactors, one can manipulate the nature of the shielding gas, affecting the production of reactive species [39,40].Lastly, electric parameters directly impact the IW.The shape of the voltage-pulse significantly influences the production of reactive species [15], and the amplitude of the voltage-pulse and frequency affect the reproducibility of the discharge [41][42][43][44].
Certain combinations of the previous factors can cause this transient system to reach a periodic stationary state [45], in which the trajectory of IWs is highly reproducible from pulse to pulse.This scenario is typically observed in helium APPJs, whereas it is less prevalent in argon, where IWs become filamentary in certain conditions [7,22,46].Filamentary IWs exhibit thinner streamers and a higher branching rate, challenging experimental measurements.Nevertheless, electric parameters can be adjusted to produce reproducible IWs in argon APPJs [7,22].Several studies have successfully measured fundamental quantities such as electron density and electric field [32] (Stark broadening [47][48][49][50], Thomson scattering [51][52][53]), as well as density and temperature of heavy chemical species, using emission and absorption spectroscopy [54,55].
Tunable diode laser absorption spectroscopy (TDLAS) is a non-intrusive diagnostic that allows for time-and spaceresolved absolute density of species [55].In argon µAPPJs, this diagnostic has been applied to measure the density of 4s states using line-of-sight or Abel inverse transform of the absorbance [6, 22-24, 26, 46, 56-58].In particular, the 811.531 nm line for the radiative transition Ar(2p 9 → 1s 5 ) can be used to probe the density of Ar(1s 5 ).Various studies have examined argon µAPPJs excited by a bell-shaped voltage pulse operating at 20 kHz and maximum amplitude of 6 kV [6,7,23,24,36].These electric conditions allowed for reproducible IWs with a characteristic diffuse light emission [22].These reactors' reproducible IWs are obtained for an argon flow rate above 400 sccm and below 1100 sccm, applied voltage above 4 kV, and repetition frequency above 10 kHz.Using a free-jet configuration at a 750 sccm flow rate, Essebbar et al [22] obtained an Ar(1s 5 ) density of 2.5 × 10 13 cm −3 , ∼4.8 mm from the nozzle using the line-of-sight method.In the case of µAPPJs impinging on targets at a floating potential, with flow rates ranging between 300 and 400 sccm, Gazeli et al [6,24] measured a maximum density of Ar(1s 5 ) between 1 × 10 13 and 4 × 10 13 cm −3 using the lineof-sight method.In another experiment, Gazeli et al [23] performed an Abel inversion of the peak absorbance profile in highly symmetric regions, resulting in an Ar(1s 5 ) maximum density ranging from 1 × 10 13 to 2 × 10 14 cm −3 for glass and alumina targets, respectively.
In this work, we build on previous research efforts to characterise µAPPJs, measuring the spatial profile of the maximum Ar(1s 5 ) density.By controlling the shielding gas mixture, we quantify how it affects the Ar(1s 5 ) density, providing hypotheses that relate to the plasma behaviour and fluid dynamic effects.
We use a co-axial cylindrical dielectric barrier discharge (DBD) reactor to produce an argon plasma jet (1 slm main jet, 3 slm N 2 -O 2 co-flow) by applying a square voltage pulse (4 kV amplitude, 20 kHz repetition frequency, 1 µs duration, 85 ns rise and 115 ns fall times).An in-house power source generates high-fidelity square pulses, producing two discharges, one at the rising edge and another at the falling edge of the applied voltage pulse.Both discharges are investigated with different shielding flow mixtures of N 2 and O 2 using a TDLAS setup similar to those of previous works [21][22][23][24]59].The current setup enables high spatial (28 µm horizontally and 48 µm vertically) and temporal (3 ns) resolution of the absorbance.Together with various data processing methods, this allows determining space-resolved absolute densities of Ar(1s 5 ).The different shielding gas mixtures allow one to isolate physical and chemical effects related to Ar-N 2 -O 2 chemistry.This paper continues with section 2, describing the materials and methods: first, the reactor setup, the gas flow parameters, and the power source are presented; then, the TDLAS experimental setup is outlined; the methods of laser absorption spectroscopy and data processing and analysis are described; and finally, the computational fluid dynamics methods are described.Section 3 presents the results and discussion, describing computational fluid modelling results; laminarturbulent jet characteristics; co-flow rate effect on Ar(1s 5 ) density; the reproducibility of Ar(1s 5 ) density values; and spatial profiles of the peak Ar(1s 5 ) density.Section 4 concludes with a summary.

Reactor setup
We use a cylindrical DBD reactor with an exposed electrode; see figure 1.This reactor produces an initial surface DBD (schematically represented in solid-line ellipsis in the figure), inducing a guided IW along the jet (schematically represented in dashed-line arrows in the figure).The reactor comprises a stainless steel tube, acting as the high-voltage electrode (anode), which is mechanically forced inside a quartz tube.A stainless steel sponge, the grounded electrode (cathode), sits between the quartz tube and a plastic (polymethyl methacrylate (PMMA)) tube.Argon flows from inside the stainless steel tube into the quartz tube, creating the main jet on exit.An N 2 -O 2 gas mixture flows between the quartz and the plastic tubes and through the stainless steel sponge, where it exits, creating the co-flow shielding jet.

Flow conditions
For the main jet, a thermal mass flow controller (GF40-SA46 from Brooks Instrument) maintains a 1 slm flow rate of argon (from AirLiquide, AlphaGaz 1, Ar ⩾ 99.999%, impurities in mole fractions: H 2 O ⩽3 ppm, O 2 ⩽2 ppm, C n H m ⩽0.5 ppm).For the co-axial jet, we explored a range of flow rates.Using Schlieren imaging, we found that a co-flow flux above 5 slm and below 2 slm is unsteady, and a flux of at least 2 slm is necessary to shield the argon jet's laminar portion fully.A flux between 3 slm and 4 slm led to more reproducible Ar(1s 5 ) absorbance profiles.Accordingly, we chose a total flux of 3 slm for the co-axial jet, which is set by two thermal flow controllers (red-y compact series, vögtlin): one for the oxygen gas (from AirLiquide, AlphaGaz 1, O 2 ⩾ 99.995%, impurities in mole fractions: H 2 O ⩽2 ppm, CO 2 ⩽0.5 ppm, CO ⩽0.5 ppm, C n H m ⩽0.5 ppm) and another for the nitrogen gas (from AirLiquide, AlphaGaz 1, impurities in mole fractions, N 2 ⩾ 99.999%, H 2 O ⩽3 ppm, O 2 ⩽2 ppm, C n H m ⩽0.5 ppm).To reduce air and water impurities inside the reactor [59], the jet flow runs for at least 30 min before acquiring measurements.

Power source
The power source creates a square-shaped voltage pulse.In this work, we apply voltage pulses with an amplitude of 4 kV, a repetition frequency of 20 kHz, and pulse duration set to 1 µs, with its full width at half maximum of 1.044 µs.For these conditions, the rise and fall times of the pulse are ∼85 ns and ∼115 ns, respectively.
At 20 kHz and 4 kV, the reactor takes about 20 min to heat up to a stable temperature, with the plastic tube and the quartz reaching 2 • C above room temperature and the stainless steel tube reaching 5 • C above room temperature (measurements made using an IR camera, FLIR i7).The power source also heats up (70 • C) during the first 30 min of operation, causing the pulse duration to reduce slightly (∼4% decrease) before attaining a stable value.Accordingly, the power source runs for at least 30 min before measurements are acquired.
Each voltage pulse produces two discharges, one during the rising and another during the falling edge of the pulse.Each discharge produces Ar(1s 5 ) metastables along the jet.The Ar(1s 5 ) absorbance decreases to noise levels after 1 µs, see figure 4. Accordingly, we chose a pulse duration of 1 µs to decouple the absorbance profiles of the two discharges.
The applied voltage, displacement current, and discharge current are described in figure 4 over one applied pulse.The applied voltage was measured with a voltage probe (MK-14KVAC from Lecroy), and the total current was measured with a current transformer (CT-B5.0from Magnelab).The signal's waveforms were acquired using a 2 GHz bandwidth digital oscilloscope (Lecroy 204MXi).The capacitance of the reactor was measured to be 0.7 pF.

Laser absorption spectroscopy setup
The current laser absorption spectroscopy setup is similar to those of previous works [6,[22][23][24] (please see Es-sebbar et al [22] figure 1 for a scheme).We used a tunable diode laser (DL100, TOPTICA photonics), in which the temperature and current are set to produce a beam of 811.523 nm wavelength.
After exiting the laser diode, the beam passes through: a neutral density filter to attenuate its intensity; a motorised shutter; a circular diaphragm to shape it; two mirrors to reflect it into the jet's x-axis (see figure 2) and a second diaphragm to reduce its width.Afterwards, a lens (35 mm focal length) focuses the laser into a narrower width, at which point it transverses the plasma jet.Then, two lenses (35 mm followed by 50 mm focal lengths) refocus the beam into a silicon photodiode (DC 125 MHz, Model 1801, New Focus), which measures its intensity.A band-pass filter of 810 ± 10 nm (810FS-25 Andover Corporation) is placed before the detector to block less relevant wavelengths.The beam waist was measured with the blade edge method [23].At the point it crosses the jet, the beam intensity varies from 90% and 10% over a distance ∆y = 28 µm and ∆z = 48 µm along the transverse and longitudinal directions, respectively, with a sigmoid-like intensity profile.At the lowest aperture, the second diaphragm did not allow for a ∆y < 30 µm beam waist.The ∆y beam-waist was achieved by y-direction diaphragm displacement, slightly slicing the laser beam along this direction.The y, z beamwaist difference is not problematic considering that the z-step (2 mm) is much larger (100 times) than the y-step (20 µm).
With the optical setup fixed, two servo micrometre motors move the plasma reactor in the y-z plane; see figure 2. We use a 2 mm step in the z direction and a 20 µm step in the y direction.A computer program controls the stepper motors, the motorised shutter, and data storage from the oscilloscope.

Laser absorption spectroscopy methods
We measured the density of the metastable Ar(1s 5 ) using laser absorption spectroscopy.In our setup, the laser beam, set at the wavelength of the Ar(2p 9 )→Ar(1s 5 ) transition λ 0 = 811.531nm, passes through the plasma and hits the photodiode detector.While passing through the jet, the variation of the intensity is related to the density of the absorbing species by the Beer-Lambert law: where λ 0 is the absorption wavelength, I 0 is the original beam intensity, I is the intensity after travelling a distance L through the absorbing medium, and k(λ) is the absorption coefficient.This relation can be written as where ρ(λ 0 , y) is the absorbance at a specific wavelength λ 0 , see figure 2.
The absorbance was generally axisymmetric in our µAPPJ; see section 3.In this cylindrical geometry, see figure 2, the Abel transform can provide a relation between the absorption coefficient and the absorbance, For an atomic line, the absorption coefficient reads [60], where we added the dependence in z, and where e is the elementary charge, f ij is the oscillator strength for the line, ϵ 0 is the vacuum permittivity, c is the vacuum speed of light, n is the number density of the atomic species absorbing the light, ν 0 is the frequency of the laser beam, and P(ν) is the normalised line profile ( ´P(ν)dν = 1).If the normalised line profile has a Lorentzian shape, it can be described as a function of the full width at half maximum (FWHM), having units inverse of its domain [61].At its peak, at a frequency ν 0 , the maximum is P(ν 0 ) = 2/(π ∆ν L ), with ∆ν L the FWHM.Using this profile on equation ( 4), the Ar(1s 5 ) density can be calculated from

Ar(1s 5 )→Ar(2p 9 ) line absorption profile
For an atomic transition, the absorption profile can be broadened by multiple processes [55,62,63].Various broadening mechanisms are a function of the gas temperature.Although we did not measure the gas temperature, previous works using a similar reactor with a different power source [22][23][24] measured a maximum gas temperature of 350 K.
In the following, we estimate the broadening by different processes.
The FWHW of the Doppler broadening is ∆λ [55,65], where T g = 300 K is the gas temperature, k B the Boltzmann constant, and m is the mass of an argon atom.Considering a gas temperature of 350 K, this FWHM increases by 8%.
From previous works in argon µAPPJs, using a similar reactor but with a bell-shaped applied voltage of amplitude 6 kV and repetition frequency 20 kHz (identical to the applied voltage used in [22]), the electron density reached at maximum n e = 2 × 10 14 cm −3 and the electron temperature corresponded to T e ≈ 2.9 × 10 4 K [53].This leads to an FWHM of the Stark broadening of ∆λ Stark = (1.05× 10 −4 T e + 4.86) ne[cm −3 ] 10 16 = 0.16 pm [66].Since the waveform of the applied voltage in the present work is not the same as in Santos Sousa et al [53], we can provide an upper estimate for this broadening assuming an electron temperature and electron density five times higher, leading to a ∆λ Stark = 2.01 pm.
The FWHM of the van der Waals broadening can be written as [62,67], where ᾱ is the mean atomic polarisability, R2 is the difference of the squares of the coordinate vectors of the upper and lower levels, T g = 300 K is the gas temperature, µ is the reduced mass, and N is the gas density.Konjević and Konjević [67] provide the various parameters for argon-argon broadening.In the case of T g = 350 K, the van der Waals broadening would increase by 10%.Considering instead N 2 or O 2 perturbers, the van der Waals broadening would be 40.96pm or 41.51 pm, respectively, with the polarisabilities of N 2 and O 2 obtained from [68].The FWHM of two convoluted Lorentzian profiles is given by the sum of the FWHM of the individual Lorentzian profiles.Accordingly, assuming a 80% (99%) Ar mixture with dry-air, the FWHM of the van der Waals broadening would slightly increase to 39.85 (39.57) pm, which is not as significant as the effect of the gas temperature.Experimentally, the absorption profile was determined by varying the temperature of the tunable diode laser, which yields a linear variation of its wavelength [24].Both the Voigt and Lorentzian profiles fit the absorption profile well.The former estimates the Gaussian variance at zero, probably due to the low contributions from Gaussian-like broadenings.The Lorentz fit, see figure 3, converges in an FWHM ∆λ L = 40.10± 0.52 pm, or ∆ν L = 18.27 ± 0.24 GHz (centred at 811.531 nm), which according to the previous estimates can be mainly attributed to van der Waals broadening.At 5 mm from the nozzle, between cases with a dry-air shielding co-flow and without co-flow, and between the centre and the edge of the plasma jet, the FWHM varied at most by 7%.Accordingly, a constant ∆ν L = 18.27GHz value is used for the density calculations.

Data acquisition and processing
The oscilloscope records the electric signals (applied voltage, current, and transmitted light measured by the photodiode).Both current and transmitted light signals are corrupted by electromagnetic noise produced by the open plasma reactor.The transmitted light signal's noise was digitally removed by subtracting a noise baseline produced with the laser off.
For each voltage pulse and each (y, z) position, the temporal profile of the transmitted light signal contains two minima corresponding to the Ar(1s 5 ) metastable production at each of the two discharges (rise and fall edges of the voltage pulse).The oscilloscope then stores 1000 values of the two minima of the transmitted light, one full temporal profile of the transmitted light, the applied voltage, and the total current.The transmitted light minima corresponds to the maxima of absorbance.Using the 1000 maximum absorbance values, we compute the where A i (y, z) is the maximum absorbance for the pulse i at the position (y,z) and µ(y, z) is the arithmetic mean.This metric indicates higher or lower variation between pulses; see figure 9.However, in specific conditions, such as those near the nozzle, the distribution of the 1000 maximum absorbance values deviates significantly from a normal distribution.In such cases, the 5th, 50th (median), and 95th percentiles of the 1000 maximum absorbance values describe the dispersion better.
Using the median of the 1000 values of maximum absorption for each (y, z) position, one obtains 2D maps for the maximum absorbance for each of the two discharges.
The inverse Abel transform is used to obtain the Ar metastable density radial profile from the transverse profiles.We compared the algorithms available in PyAbel [69] to that of Beniaminy and Deutsch [70] used in previous µAPPJ works [23,59].A Gaussian function, for which an analytical inverse Abel transform exists, served as a benchmark.For this analytical function, we defined various discrete Gaussian profiles with different standard deviations and number of points and applied each Abel transform algorithm.The methods that yielded solutions closest to the analytical were the Basex and that of Benianimy and Deutsch.We chose the former due to performing better at fewer points.
We perform an inverse transform of the left-and the righthand side of the transverse profiles.By comparing the two profiles, one can evaluate if the transform profile is symmetric, a condition for the validity of the Abel transform.The left-and right-hand side inverted profiles should be continuous at r = 0, but it was not always possible to measure the absorbance at the centre of the profile.The reason is twofold.First, the reactor was mounted at an angle to the rail of the axial micrometre motor.Second, to correct this fault in the setup, the rail was slightly tilted to place the reactor perpendicularly to the table holding the laser.After this correction, when moving the reactor to acquire transverse absorbance profiles at different axial positions, the centre of the region under analysis is horizontally displaced by tens of µm.Accordingly, we define the centre by interpolating the transverse signal with a second-order spline and by iteratively calculating the position of the centre using a Newton-Raphson algorithm (tolerance of 35 µm), until the left and right radial profiles are continuous at r = 0.In a few cases of transverse profile asymmetry, the algorithm cannot converge; these cases are identified in the results section.

Computational fluid dynamic simulations
Computational fluid dynamics (CFD) simulations described the Ar molar fractions along the jet.We used a subsonic version of the simulation platform for aerothermodynamics radiation and kinetics (SPARK) [71], a finite-volume code solving the compressible Navier-Stokes equations.The subsonic version of SPARK differs from the original by including low-Mach preconditioning [72], the simple low-dissipation advection upstream splitting method inviscid flux solver [73], and primitive variables reconstruction using WENO-5 [74].The diffusive fluxes coefficients are calculated from the Wilke-Blottner-Eucken model as described in [75], where viscosity coefficients were obtained from [76].
Simulations were performed for a 2D axisymmetric geometry.The simulation domain adopted the reactor's dimensions extended until 40 mm longitudinally and 14 mm radially.We assume a parabolic velocity profile with a gas temperature of 300 K for the main central jet and the co-flow jet.A no-slip condition is assumed at the walls, a symmetry condition at the axis, and elsewhere, we adopt a subsonic outlet at a static pressure of 101 325 Pa.A sponge layer between 33 mm and 40 mm damps possible acoustic waves.Convergence was assumed once the flow variables remained stationary and after a drop of 10 −3 in the total residuals.

Electric signals and absorbance
Our cylindrical DBD reactor, see figure 1, produces two discharges per applied voltage pulse, each one producing Ar(1s 5 ) species, see figure 4.This figure shows the time evolution of the discharge current and the plasma absorbance over one applied voltage pulse.The discharge current and the absorbance present two peaks: one during the rising edge and another on the falling edge of the applied voltage pulse.Fast imaging diagnostics show an IW propagating downstream along the jet after the first current peak, which is consistent with previous works using a bell-shaped applied voltage pulse [6,[22][23][24].Top: applied voltage pulse (red), discharge (black) and displacement (blue) currents; bottom: absorbance temporal profiles, measured at z = 10 mm and y = 0 µm for 100%O 2 (blue) and 100%N 2 (orange) shielding gas.Absorbance starts to increase during the passage of the streamer, which at z = 10 mm occurs at ∼200 ns (varying with the shielding gas mixture) after the beginning of the discharge.The maximum absorbance is marked for each discharge: × for the first discharge; for • the second discharge.
During the plateau of the applied voltage pulse, the discharge current oscillations decrease to almost null amplitude, which is coherent with the behaviour of a DBD reactor where surface charges deposited on the dielectric cancel out the electric potential difference to the anode.The surface charges are subsequently released upon the decrease in the applied voltage during the falling edge of the pulse.This produces a second discharge, which also propagates downstream.While the first discharge is an IW, being filamentary or diffuse depending on the position along the argon jet [22], the second discharge is always diffuse.The propagation of the IW downstream along the argon jet, see figure 1, should be related to the plasma electronegative shielding [30,59], where anions clustering around the jet reduce the conductivity in the jet/shielding boundary, thus focusing the IW into the centre of the jet.Although this mechanism is usually associated with IWs (the first discharge in our case), it should also be present in the second discharge, starting only 1 µs after the IW.For a 100% N 2 shielding gas, see figure 4, the Ar(1s 5 ) density is particularly low, suggesting a less effective electronegative shielding due to a much lower anion production.

Computational fluid dynamic simulations
Since the mixing rates between the argon jet and the co-flow gas may influence the IW trajectory and the quenching rates of Ar(1s 5 ), we performed CFD simulations to evaluate the Ar molar fraction for the different shielding gas mixtures.In laminar flows, the mixing depends mainly on the diffusion coefficients, or equivalently, the viscosity of the gas [77], which is comparable for N 2 , O 2 , and humid air with 2%H 2 O [78].Still, gaseous oxygen is slightly more viscous than nitrogen, leading to a faster mixing with the argon jet and lower axial argon fraction.However, at the axis, between 100%N 2 and 100%O 2 shieldings, the relative difference in Ar molar fraction is low: 1% at 12 mm, 5% at 22 mm, and 10% at 33 mm.The mixing starts from the jet/shielding boundary towards the jet's interior.If we delimit the locations with a 90% (99%) Ar mole fraction, we obtain a cone-like shape of about 20 mm (10 mm) in length, see figure 5. Accordingly, contour regions of higher argon fractions are cone-shaped, with smaller areas for higher argon purity.This is consistent with the cone-shaped density profiles of argon metastables measured in this work with TDLAS.

Laminar to turbulent flow transition
Complementary Schlieren imaging was performed to evaluate the laminar length of the argon jet.These experiments were performed in an identical experimental setup to that of the work of Darny et al [36].Without plasma and in the absence of co-flow (no co-flow), images show a laminar and almost cylindrical argon jet flowing from the nozzle, which eventually transitions into a turbulent flow at about 39.4 ± 2.2 mm from the nozzle.Without plasma and with co-flow, a shielding gas envelopes the argon jet, starting from the edge of the co-flow tube until the turbulent region.In this case, the average laminar length is of: The decrease in laminar length points to a more unstable jet for higher oxygen-fraction shieldings.Although we did not perform Schlieren measurements for a 4 kV pulse of 1 µs duration, for a 6 kV pulse of 500 ns of duration at 20 kHz of repetition frequency, the average laminar length, compared with the plasma off case, was reduced by: • ∼8 mm without co-flow shielding; The decrease in laminar length with plasma ON is expected as plasma-flow perturbations have been shown to disrupt the laminar flow [36,37].Higher oxygen-fraction shieldings also seem to enhance plasma-flow perturbations, although this effect may be due to an already more unstable jet, as observed in the plasma off case.A lower laminar length may affect the Ar(1s 5 ) density profiles due to the increased mixing, leading to a higher Ar(1s 5 ) quenching rate.The Ar(1s 5 ) quenching rate (in units of 10 −11 cm 3 s −1 ) is of 3.6 by N 2 [79], 21 by O 2 [79] and 78 by H 2 O [58].

Ar(1s 5 ) density for increasing dry-air co-flow rate
Our TDLAS setup and methods allow us to measure for each point (r, z), the maximum Ar(1s 5 ) density of the first and second discharges, which can be displayed in the form of 2D maps for the whole jet, as shown in figure 6.
For the first discharge, without co-flow, the spatial profile of the Ar(1s 5 ) density is cylindrical with increasing density values along z up to 5.2 × 10 13 cm −3 at 14 mm from the nozzle, decreasing to below 1 × 10 13 cm −3 at 20 mm.With a 0.5 slm dry-air co-flow, the maximum density increases in the jet's centre, reaching a maximum value of 7.2 × 10 13 cm −3 at 4 mm from the nozzle.The profile becomes shorter (densities around 1.5 × 10 13 cm −3 near 18 mm) and less symmetric.Increasing to a 1 slm dry-air co-flow leads to the maximum density in the jet's centre to rise again, reaching a maximum value of 8.0 × 10 13 cm −3 at 10 mm from the nozzle.The profile is more asymmetric, longer (densities around 1.7 × 10 13 cm −3 near 26 mm), and partially cone-shaped, being broader near the nozzle and narrower at 14 and 16 mm, before broadening again at 20 mm.At a 3 slm co-flow rate, the density increases further to a maximum of 8.9 × 10 13 cm −3 at 14 mm.The profile is now symmetric, longer (densities around 3 × 10 13 cm −3 at 26 mm), and again cone-like, with a narrower position near 16 mm.
The profile is cone-like for the second discharge without co-flow, with the density increasing to a maximum value of 3.4 × 10 13 cm −3 at 6 mm and then decreasing to a minimum value of 4.6 × 10 12 cm −3 at 16 mm from the nozzle.The asymmetry increases for a 0.5 slm dry-air co-flow, where the density increases up to 7.2 × 10 13 cm −3 at 4 mm, decreasing downstream to 7.4 × 10 12 cm −3 at 16 mm from the nozzle.With a 1 slm co-flow, the profile is still asymmetric, and the Ar(1s 5 ) density reaches a peak value of 6.3 × 10 13 cm −3 around 12 mm, decreasing to 3.9 × 10 12 cm −3 at 18 mm from the nozzle.With a 3 slm co-flow, the profile becomes symmetric from 8 mm onwards, funnelling until 14 mm, when it becomes cylindrical.For this co-flow rate, the Ar(1s 5 ) density Figure 6.Spatial profiles of the maximum Ar(1s 5 ) density at different (r, z) positions for the first discharge (first row) and the second discharge (second row), and for different dry-air co-flow rates (depicted in different columns).Negative radial positions are for data obtained from Abel inversion of the left-hand side of the transverse absorbance profiles, while positive radial positions are for the corresponding right-hand side.Comparison between profiles at negative and positive radii allows for estimating the symmetry of the absorbance and, thus, the reliability of the Ar(1s 5 ) density measurement.reaches a maximum value of 5.8 × 10 13 cm −3 around 14 mm, decreasing to 5.7 × 10 11 cm −3 at 18 mm from the nozzle.
Dry-air shielding reduces the concentration of H 2 O surrounding and in the plasma jet, which may lengthen and broaden the spatial profile of the Ar(1s 5 ) maximum density.Compared to dry air, humid air has a higher attachment rate, which can hinder the formation and propagation of the streamer [80,81] by reducing the accumulation of charges on the streamer head, lowering its radius and velocity and increasing the threshold field for propagation.Also, the quenching rate of Ar(1s 5 ) by H 2 O is higher than by O 2 (∼4-fold) or N 2 (∼21-fold), see section 3.3, which could lead to lower Ar(1s 5 ) peak densities and shorter profiles.
The lower Ar(1s 5 ) densities obtained for co-flows of 1 slm and 0.5 slm may result from incomplete shielding, leading to a higher H 2 O concentration in the argon jet.The lower densities could also be caused by a non-stationary flow, leading to less reproducible IW trajectories.The asymmetric Ar(1s 5 ) density profiles for 0.5 and 1 slm co-flows supports this hypothesis, and Schlieren imaging also shows an asymmetric flow pattern in the case of a 1 slm co-flow.

Reproducibility of the absorbance with different N 2 /O 2 co-flow gas fractions
While sweeping the co-flow rate between values of 0.5 slm to 5 slm, we found a higher reproducibility of the Ar(1s 5 ) absorbance profile for co-flows between 3 and 4 slm.Accordingly, a 3 slm co-flow rate was used to measure Ar(1s 5 ) density between different N 2 /O 2 co-flow gas fractions.
Molecular oxygen in the co-flow mixture affected the variability and the shape of the absorbance profiles from pulse to pulse; see figure 7. Near the nozzle, the variability is much higher for both discharges and all co-flow mixtures.Moving away from the nozzle, the variability decreases, remaining higher for a 100%N 2 (0%O 2 ) co-flow, which also shows a broader profile.The admixture of 2% O 2 in the co-flow is enough to considerably reduce the variability and narrow the profile to shapes similar to those of a 100% O 2 co-flow mixture.
Spatial profiles of the RSD of the maximum absorbance are presented in figure 9.The RSD is always below 40%.The RSD reduces to 15% from 10 mm onwards for co-flows with molecular oxygen and from 14 mm onwards for a 100%N 2 (0%O 2 ) co-flow.For the first discharge, the decrease in the RSD coincides with the beginning of the symmetric region for the profiles of the Ar(1s 5 ) maximum density, see figure 10.
The broader absorbance profiles and higher RSD observed during the first discharge for every co-flow mixture near the nozzle can be due to the random trajectory of the streamer as it propagates from the dielectric surface to the axis of the argon jet [30].These observations are coherent with the hollow electron density profiles measured with Thomson scattering [53] in that region in a similar reactor without co-flow.In our reactor, fast-imaging shows that the IW is not axisymmetric during the surface-to-volume transition, resembling a thin streamer [22].This filament leaves the nozzle at varying azimuthal positions from pulse to pulse, which could explain both the high 95th (passing filament) and low 5th (no filament) percentiles of the absorbance values at z = 0 mm, see figure 8.At the nozzle, the even broader absorbance profile of the 100%N 2 case could be related to a lower electronegative shielding.For higher O 2 concentrations in the co-flow, the narrower profile indicates a better focusing of the IW [30,59] closer to the axis.
Furthermore, in N 2 -O 2 streamers, a lower concentration of molecular oxygen has been associated with higher streamer branching probabilities [82,83], possibly due to a lower rate of photoionisation, which could explain both the higher RSD and the broader profiles of the 100%N 2 case even away from the nozzle.Note that the 100%N 2 co-flow case can contain O 2 impurities (⩽2 ppm) from the Ar and N 2 gas bottles so that photoionisation exists.Further downstream, the absorbance is usually more reproducible in the jet's centre, where the IW is more diffuse [22], and less so at the edges.
For the second discharge, further away from the nozzle, the peak absorbance is lower for 2%O 2 and 100%N 2 (0%O 2 ) coflow, see figure 7, which translates to a lower Ar(1s 5 ) density, see figure 10.This could be related to a lower electronegative shielding and a less efficient focusing of the discharge into the jet axis [30], which is also evident from the spread of the RSD of the peak absorbance at 100%N 2 (0%O 2 ), see figure 9.

Ar(1s 5 ) maximum density at different N 2 /O 2 co-flow gas fractions
Figure 10 shows the spatial profiles of the Ar(1s 5 ) maximum density, measured with the TDLAS setup for the two discharges, and for different N 2 /O 2 co-flow gas fractions, including the Abel inversion of the left-and the right-hand sides of the transverse profiles.This was used to verify the cylindrical symmetry of the discharge, a necessary condition for the reliability of the Abel inversion.The non-symmetric regions correlate well with those of higher RSD; see figure 9.The second discharge with a 100%N 2 co-flow showed no symmetric region.For both discharges, the spatial density profile funnels downstream.This is consistent with the progressive mixing of the shielding gas into the argon jet, as determined by CFD simulations, which favors the quenching of Ar(1s 5 ).
For 50% and 20% of O 2 , the profiles reach a narrow point at 14 mm, after which the profile broadens again.This shape is seen to a lesser degree at 16 mm for 80%O 2 co-flow.We do not have a definite explanation for this peculiar shape.A possible reason is the filamentary off-axis trajectory of the IW after the end of the reproducible zone, as observed in fast imaging diagnostics [22].However, in non-filamentary helium plasma jets, Hübner et al [20] and Douat et al [59] also observed a similar shape for the electron and metastable density profiles, respectively.Further experiments should be conducted to properly analyse this effect.For 100%O 2 co-flow, there is no narrowing/broadening point for the profile of the Ar(1s 5 ) maximum density.However, one observes the increase of the RSD after 20 mm; see figure 9.
In the symmetric region, the maximum density is always in the centre of the radial profile, where the higher Ar fraction region is also found.In all cases, the Ar(1s 5 ) maximum density is of the order of 10 13 cm −3 , with the highest value measured for 100%O 2 , and the lowest for 100%N 2 (0%O 2 ), see table 1.
The lower peak densities for 100%N 2 (0%O 2 ) shielding could be related to a more filamentary discharge, again due to higher branching ratios [82,83] and less reproducible streamer  trajectories.The overall higher RSD supports this possibility for the 100%N 2 shielding; see figure 9.A more filamentary discharge would have a twofold effect of broadening the absorbance profile and lowering the median absorbance over 1000 pulses, decreasing the maximum value of the Ar(1s 5 ) radial profile.The high Ar(1s 5 ) density for 20%O 2 shielding, see table 1, cannot be explained by the previous hypothesis.One possibility is that this co-flow gas mixture, with a relatively lower O 2 fraction, mixes less with argon (due to the lower viscosity of the nitrogen-dominated gas shielding [77,78] and due to a less turbulent flow as evidenced by a longer laminar length).In this case, the quenching ratios are lower relative to higher O 2 fraction [79] and the reproducibility higher than in the case of 100%N 2 (0%O 2 ), see figure 9.This combination of effects could explain the increase in the maximum Ar(1s 5 ) density for 20%O 2 co-flows, but this hypothesis cannot explain the even higher Ar(1s 5 ) maximum density at 100%O 2 co-flows.It is possible that pure molecular oxygen shielding presents more substantial memory effects, with anions functioning as electron reservoirs and limiting the diffusion loss of electrons between pulses, which would be coherent with the highest Ar(1s 5 ) density on both discharges, see table 1 and figure 10.Still, this suggests a missing mechanism involving N 2 , which cannot be explained with the current analysis.
The position of the maximum density shifts upstream from 14 mm to 10 mm with increasing O 2 fraction in the co-flow.This could be due to higher Ar/shielding-gas mixing, owing to the higher O 2 viscosity [78], which along with higher quenching rates by O 2 [79] could limit the maximum Ar(1s 5 ) density position closer to the nozzle.The shift of the peak density position could also be associated with the decrease in laminar length observed for increasing O 2 fraction co-flows; see section 3.3.Turbulence may increase the mixing between argon and the shielding gas.A shorter laminar length may lead to higher quenching rates closer to the nozzle, thus shifting the position of Ar(1s 5 ) maximum density upstream.
Compared to the first discharge, the maximum density for the second discharge is lower for co-flow mixtures between 0% to 80%O 2 , see table 1.The exception is 100%O 2 , where the maximum density is comparable for the two discharges.For the second discharge, the position of the maximum metastable density is always around 12 mm, whatever the co-flow gas mixture.
Unfortunately, for the second discharge, the absorbance for 100%N 2 (0%O 2 ) is too low and asymmetric, so the Abel inversion could not be performed.The low metastable density at 100%N 2 (0%O 2 ) could be related to a higher plasma conductivity near the nozzle, due to the lower anion density, causing a higher electrons losses, leading to a shorter discharge propagation downstream along the axis.
The spatial density profiles also funnel downstream for the second discharge, see figure 10, which is again consistent with the progressive mixing of the shielding gas into the argon jet, see figure 5.The cone-like shape is more pronounced for 100% to 50% O 2 shieldings, approaching a cylindrical shape for 20%O 2 and below, possibly due to lower electronegative shielding.Apart from 100%N 2 (0%O 2 ), the position at which the spatial profile of the second discharge ends is near 20 mm, which for > 50%O 2 co-flows is near the laminar to turbulent transition.
The high Ar(1s 5 ) maximum density occurring during the second discharge at 100%O 2 co-flow is not easy to explain.The enhanced electronegative shielding for a pure O 2 co-flow may account for the higher density, but this does not explain the significant difference between the maximum density of the metastable for 100%O 2 and the other oxygenated shielding gas mixtures.This behaviour at 100%O 2 (0%N 2 ) suggests a mechanism involving N 2 , which could not be identified in this analysis.

Conclusion
The argon µAPPJs reactor studied in this work produces two discharges, one at the rising edge and the other at the falling edge of the applied voltage pulse.TDLAS allowed measuring Ar(1s 5 ) absolute density radial profiles for each discharge.The first discharge is longer, broader, and less reproducible than the second discharge, generally leading to higher Ar(1s 5 ) densities.In both discharges, the Ar(1s 5 ) density profile funnels downstream, coherent with the shape of higher argon molar fractions regions found by the CFD simulations.The shielding gas significantly affects the Ar(1s 5 ) density profiles.A dry-air co-flow increases the Ar(1s 5 ) density compared with the ambient case.Since the argon jet flow is not significantly modified, this is possibly due to the lower H 2 O concentration.The absence of O 2 in the shielding gas leads to lower reproducibility and broader Ar(1s 5 ) density profiles.O 2 admixture significantly increases both discharges' reproducibility and Ar(1s 5 ) density.The absence of N 2 leads to the highest Ar(1s 5 ) density in both discharges.
Even with the complementary Schlieren and CFD analysis, it is not easy to quantify the interaction between plasma and flow dynamics.Furthermore, electrodynamic and chemical processes can vary significantly with different co-flow gas mixtures.Expanding this analysis may require a comprehensive Ar-N 2 -O 2 reactive scheme to be compared with detailed local and temporal Ar(1s 5 ) density profiles.Even though our analysis is coherent with the current hypothesis on APPJ dynamics, it poses questions unanswered within the existing literature, such as how can the absence of N 2 in the shielding gas cause a significantly higher Ar(1s 5 ) density in both discharges compared with other shielding mixtures?This work also provides spatially accurate data of Ar(1s 5 ) densities in limiting conditions, i.e.Ar-N 2 and Ar-O 2 gas mixtures, and details the behaviour in intermediate cases.This data is essential to plan future experimental studies and to validate numerical models.It may also aid in developing shielded argon µAPPJs for different applications.

Figure 1 .
Figure 1.Scheme of the co-axial reactor including relevant dimensions.The origin of the transverse axis, y, is at the reactor's centre line, and the axial axis's origin, z, is at the nozzle exit.Solid-line ellipsis are a possible position for surface charges and the initial surface DBD.Dashed-line arrows represent the IW's trajectory while leaving the nozzle.The expected electrostatic field lines outside the nozzle are in dotted-line arrows in a case without discharge.

Figure 2 .
Figure 2. Schematic of the laser passing through the plasma, including the referential.Representative absorbance profile along the transverse direction.Stepper motors move the reactor along the transverse direction y and the axial direction z.

Figure 3 .
Figure 3. Normalised line absorption profile (P(ν), ´P(ν)dν = 1) at 5 mm from the nozzle and at the centre of the plasma jet with a dry-air co-flow at 3 slm.The profile was fitted with a Lorentzian function with a constant baseline.

Figure 4 .
Figure 4.Top: applied voltage pulse (red), discharge (black) and displacement (blue) currents; bottom: absorbance temporal profiles, measured at z = 10 mm and y = 0 µm for 100%O 2 (blue) and 100%N 2 (orange) shielding gas.Absorbance starts to increase during the passage of the streamer, which at z = 10 mm occurs at ∼200 ns (varying with the shielding gas mixture) after the beginning of the discharge.The maximum absorbance is marked for each discharge: × for the first discharge; for • the second discharge.

Figure 5 .
Figure 5. Contours of the Argon molar fraction in the case of no co-flow, i.e. pure argon jet into a quiescent dry-air + 2%H 2 O gas mixture.

Figure 8 .
Figure 8. Fast-imaging for 1000 sccm argon jet, no co-flow, 20 kHz repetition frequency, 200 ns pulse duration.Images show one exposure of the first discharge in different pulses.The argon jet and the ionisation wave propagate from top to bottom.The surface-to-volume streamer passage occurs at different azimuth angles.Although these images provide only 2D projections of the IW trajectory, the thickness was measured to be 380 ± 87 µm at z = 0 mm and 793 ± 12 µm at z = 4.5 mm.The fast-imaging experimental setup is similar to those of previous works [22].

Figure 9 .
Figure 9. Relative standard deviation for the 1000 measures of peak absorbance, calculated at different (y,z) positions for the first discharge (first row) and the second discharge (second row), and for different N 2 /O 2 co-flow gas fractions (depicted in the different columns).

Figure 10 .
Figure 10.Spatial profiles of the maximum Ar(1s 5 ) density at different (r,z) positions, for the first discharge (first row) and the second discharge (second row), and for different N 2 /O 2 co-flow gas fractions (depicted in the different columns).The negative/positive radial positions are as in figure 6.The second discharge with a 100%N 2 (0%O 2 ) co-flow showed no symmetric region.

Table 1 .
Maximum density of Ar(1s 5 ) (units of (10 13 cm −3 ) in the first and second discharges for different O 2 percentages in the shielding gas.
a A symmetric Abel inversion could not be performed for this case.