Correlation of plasma generated long-lived reactive species in aqueous and gas phases with different feeding gases

Reactive species are believed to play an important role in the treatment effects in both direct and indirect plasma applications. To better understand plasma chemistry and reveal the reactive species generated in the aqueous and gas phases, this study investigated the reactive species (i.e. H2O2, NO2− and NO3− ) that were generated using two types of plasma jets with different feeding gases (i.e. air, CO2, and N2) and treatment time (i.e. 0, 5, 10 and 20 min) in both aqueous and gas phases. Under the same treatment conditions, the determinations of reactive species in ambient air (X-variable) were carried out using optical absorption spectroscopy (OAS), and the detection of reactive species in plasma-activated water (PAW, Y-variable) was using wet chemical-based colourimetric methods. For both measurements, OAS spectral features and the contents of reactive species in PAW varied from different feeding gases and treatment times. The correlations of X and Y were developed using partial least squares (PLS) regression modelling. High collinearity (R2CVs ∼ 0.955–1) was shown between the independent X and Y. Based on the PLS regression results, the selectivity ratio (SR) algorithm was used to select the most important spectral wavelengths that are highly related to the functional groups of these reactive species.

(Some figures may appear in colour only in the online journal) 4 www.ucd.ie/refrig and www.ucd.ie/sun * 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.

Introduction
Plasma as the fourth state of matter has been widely employed as a non-thermal process in plasma medicine (such as wound healing, cancer therapy, blood coagulation, dermatology, stomatology and production/treatment of biocompatible materials), plasma agriculture (such as microbial decontamination, equipment sterilization, surface modification, functionality changes and stimulation of seed growth), and also plasmas in integrated circuit processing [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16]. Plasma has been found only about one hundred years ago, there are many areas of unknown remaining in plasma fundamentals and practice. To better demonstrate the mechanisms and optimize processing operations, it is essential to explore plasma-involved chemical reactions. While plasma chemistry is highly complex, involving various physical and chemical processes, there are more than 75 different species and 500 reactions found in air plasma chemistry, and the lifetime for most of the highly reactive species is quite short [1]. Among these, reactive nitrogen species (RNS) and reactive oxygen species (ROS) are considered the most important species in anti-cancer treatment, microbial decontamination and product functionality changes [17][18][19].
Plasma treatment was initially carried out in a direct way, where the plasma reactive species was directly in contact with the sample surface to complete the treatment [20][21][22][23][24]. In recent decades, indirect plasma approaches were applied by some researchers, where plasma is first discharged in or close proximity to water (or solutions) to create plasma-activated water (PAW), then these nitrogen-oxide-contained acidified solutions were used to immerse samples to complete plasma treatments [10,[25][26][27][28][29]. Plasma discharges generated within PAW are highly transient and involve rapid breakdown processes, which are mostly driven by laser pulsed, strong electric fields or by bubble implosions [30]. When plasma discharge is in contact with water or other aqueous solutions, the reactive species produced from gaseous plasma could be transferred towards the liquid phase of plasma through direct transforming to aqueous species or a series of plasma-liquids interactions in the bulk solutions [30][31][32][33][34]. Hydrogen peroxide (H 2 O 2 ), nitrite (NO − 2 ) and nitrate (NO − 3 ) are long-lived secondary products that could be preserved in PAW, extending the antimicrobial effect after plasma treatment [35,36]. Each of the above individual species in plasma is capable of microbial inactivation, which could be carried out simultaneously and the antimicrobial effect may be enhanced synergistically [37]. As reported by Traylor et al [35], PAW in the pH value range of 2-3 contains H 2 O 2 , NO − 2 and NO − 3 as well as other species, is effective in killing bacteria. Hydrogen peroxide (H 2 O 2 ) is a major ROS species existing in plasma; it can be catalysed to form highly reactive OH radicals which may induce cell death [1,18]. As predominant RNS species, nitrite (NO − 2 ) and nitrate (NO − 3 ) are responsible for the plasma-induced acidification of PAW. The nitrite ions (NO − 2 ) contained in solutions with a pH value less than 5 are called acidified nitrite, which has been reported to be antimicrobial [35].
Generally, the detection of reactive species could be carried out using the conventional wet chemistry-based colourimetric method and non-destructive spectroscopy-based methods which refer to fluorescence, electron spin resonance (ESR) spectroscopy, optical emission spectroscopy (OES) and optical absorption spectroscopy (OAS). ESR is particularly used in the detection of short-lived radical reactive species in aqueous solutions [1,32]. OES is based on the detection of the emission profiles due to energy transformation through the collisional and radiative process; it determines the presence of particular species meanwhile obtaining discharge temperatures [38,39]. If there is low spectral radiation, OAS could be a better alternative for the assessment of reactive species (i.e. ozone, hydroxyl radicals, argon metastable and helium metastable), where tungsten halogen lamps, xenon arc lamps and light-emitting diodes could be employed as probing spectral lights [40,41]. During OAS measurements, plasma particles work as absorbers for the electromagnetic waves to estimate the reactive species in the plasma beam [1,[42][43][44]. The capacity of OAS has been approved in the detection of absolute number densities of ROS and RNS (e.g. OH, NO 2 , O 3 , N 2 O 5 , H 2 O 2 and HNO 4 ) [45,46]. OAS diagnostics of various plasma reactive species were carried out in different wavelength ranges [44]. The absorption spectral features of O, N atoms and singlet delta oxygen O 2 ( 1 ∆g); ozone and hydroxyl radical; argon metastable and helium metastable, and fingerprint information of plasma species were observed in the wavelength ranges of vacuum ultraviolet, ultraviolet, visible and near-infrared, and mid-infrared, respectively [44]. The quantification of plasma-generated reactive species within food packages was also investigated using OAS by Patil et al [47]. According to the absorbance in the wavelength ranges of 385-415 and 250-260 nm; 220-240 nm, and 290-340 nm, the presence of NO 2 and O 3 ; N 2 O 5 , H 2 O 2 , HNO 4 ; OH radicals and water-bonded OH complexes were indicated, respectively. For the detection of long-lived species, simple modelling based on OAS was employed by Schmidt-Bleker et al [48] to investigate the reactive species generated using a cold atmospheric plasma jet with varied mixtures of oxygen and nitrogen at different humidity levels. The absorption peaks corresponding to ozone (O 3 ) and nitrogen dioxides (N 2 O 4 , NO 2 , and NO 3 ) were clearly observed in the wavelength ranges of 250. 13 [49] derived the concentration of ROS and RNS in PAW using UV-vis OAS through multiple measurements during plasma activation. Brubaker et al [50] also investigated the chemical kinetics of a cell culture medium during the treatment of argon-fed atmospheric pressure plasma using real-time ultraviolet OAS and colourimetric assays.
As some researchers pointed out that the reactive species in plasma can be transferred and preserved in PAW due to the plasma-liquid interactions [31,[51][52][53]. However, to the best of the author's knowledge, there is no researcher has tried to find out the correlations between the reactive species detected in the air using OAS and those preserved in PAW using wet chemistry methods through chemometric analysis. Therefore, this study investigated (1) diagnostics of reactive species that are generated by two types of atmospheric pressure plasma jets with different feeding gases (i.e. air, N 2 and CO 2 ) and treatment time (i.e. 0, 5, 10 and 20 min) using OAS and wet chemical-based colourimetric methods; (2) the correlations between X and Y using partial least squares (PLS); and (3) the determination of the most relevant OAS spectral wavelength to plasma reactive species using spectral variable selection algorithms.

Plasma equipment
Two types of Atmospheric Pressure Plasma Jets (APPJ) systems were used in this study. Based on their relative beam temperature, they are named cold and hot plasma jets in this paper, respectively. The cold plasma system was sourced from the National Centre for Plasma Science and Technology at Dublin City University (Glasnevin, Dublin 9, Ireland) and the hot plasma system was supplied by the Diener electronic GmbH (PlasmaBeam Standards, Germany).
The cold plasma device is in essence a standard APPJ operating at 20 kHz. Different feeding gases (i.e. air, CO 2 , N 2 ) are supplied to the plasma system with a flow rate of 11 l min −1 . The high voltage electrode within the glass tube is pin-shaped, while the conical-shaped grounding electrode is at the end of the tube to provide a more stable operation for positioning the possible target. The plasma generated within the jet (with a diameter of 30 mm) is excited through a small gap in the ground electrode entering ambient conditions for treatment. The system also consists of a control panel, a high-voltage generator, and a gas compressor as shown in figure 1(a). More details about the cold plasma system were described by Charoux et al [54].
The hot plasma jet with a diameter of 32 mm is also operating at 20 kHz. The consumption rate of the multiple gases (i.e. air, CO 2 , N 2 ) is 1500 l h −1 with a pressure of 5-8 bar. The temperature of the plasma generated was between 150 • C and 200 • C. A jacket-wall cooling condenser with water circulation at 4.0 ± 0.5 • C was connected to the nozzle of hot plasma to cool down the plasma temperature to ∼20 • C. As shown in figure 1(b), the hot plasma system also consists of a control panel (230 V voltage, 50/60 Hz, approx. 450 W), a high voltage generator with a power of 300 W, and a multiple gas supply unit. More information about the plasma system can be found in the studies of Inguglia et al [55] and Charoux et al [56].

Detection of reactive species in PAW
Deionized water (25 ml) was placed in 150 ml plastic bottles and treated in triplicate using the cold or hot plasma jet for 5 min, 10 min and 20 min, respectively. For cold plasma treatment, the distance between the nozzle of the jet and the water surface was 85 mm. For hot plasma treatment, the cooling system at 0 • C was set up to control the plasma temperature at ca. 35 • C-40 • C. In order to obtain a complete reaction between plasma reactive species and deionized water, a silicone tube with a length of 100 mm was connected to the outlet of the cooling system at one end and the other end was immersed under the water surface for 5 mm to transfer the plasma beam. Then the concentrations of H 2 O 2 , NO − 3 , and NO − 2 in the PAW were measured in triplicate using the following wet chemistrybased colourimetric methods.
The concentrations of H 2 O 2 in PAW were estimated using the titanium oxysulfate (TiOSO 4 , Sigma-Aldrich) colourimetric method. As shown in equation (1), H 2 O 2 reacts with titanium sulfonate reagent to form pertitanic acid in yellow colour [57]. The calibration curve of H 2 O 2 concentrations (0.05-1.78 mM) and absorbance values were established using diluted solutions of 30% hydrogen peroxide standard (Sigma-Aldrich), and the absorbance values were acquired at 405 nm using a microplate reader (EPOCH-2, BioTek). Then 100 µl of TiOSO 4 solution was added into 1000 µl of PAW and incubated for 10 min in darkness before the measurement. Eventually, the H 2 O 2 concentrations in PAW after different plasma treatments were calculated using the developed standard curve.
The estimation of nitrate concentrations in PAW was carried out using the nitrate assay kit (Merck, product number 1.09713.0001) colourimetric method. The stock solution with a concentration of 100 mM was prepared using sodium nitrate, which was used to establish the standard curve of NaNO 3 in the concentration range of 0.1-0.14 mM; the absorbance value was read at 340 nm. Eight hundred microliters of reagent A in the nitrate assay kit was added into each Eppendorf tube (2 ml) first, 100 µl of the sample was carefully layout on reagent A (without mix), then 100 µl of reagent B was added into each Eppendorf tube and the content was mixed vigorously by the vortex. After 20 min of incubation in darkness, the absorbance value of the mixture was measured at 340 nm.
For the measurement of nitrite concentrations in PAW, Griess reagent (N-(1-naphthyl) ethylenediamine dihydrochloride) (Sigma-Aldrich, product number 03553) was used based on the Griess reaction [58]. Nitrite reacts with sulfanilic acid (HO 3 SC 6 H 4 NH 2 ) to form diazonium cation (HO 3 SC 6 H 4 -N≡N + ), which could subsequently react with aromatic amine 1-naphthuylamn (C 10 H 7 NH 2 ) to generate a red-violet coloured water-soluble azo dye (HO 3 SC 6 H 4 -N=N-C 10 H 6 NH 2 ) [59]. A standard curve of nitrite in the concentration range of 0.0025-0.05 mM was made using the dilutions of 100 mM sodium nitrite. Griess reagent was added into PAW with a ratio of 1:1 for reaction, and then the solutions were incubated for 30 min in darkness [60]. The absorbance value at 548 nm wavelength was acquired and used to calculate the concentrations of NO − 2 based on the standard curve [58]. Effects of plasma treatments with different feeding gases and treatment time on the concentrations of reactive species were analysed using one-way analysis of variance (ANOVA) in Statistical Package for Social Sciences (SPSS, v. 24, IBM statistical analysis). Average values were calculated from the triplicate measurements and the statistical significance was considered at P < 0.05.

Detection of reactive species in the air using OAS
As shown in figure 2, the OAS absorbance spectra of both hot and cold plasma were collected in a sealed dark cabinet using a Czerny-Turner charge-coupled device (CCD) spectrometer (CCS200, Thorlabs, USA) with an sub-miniature asssemly (SMA-SMA) fiber patch cable probe (M92L01, Thorlabs, USA) in the wavelength range of 350-900 nm at 1 nm increment. A tungsten halogen lamp (QTH10/M, Thorlabs, USA) was used to enhance the intensities of spectra. One head of the fibre optic probe was connected to the CCD spectrometer and the other head was fixed on a probe stand, which was built underneath the nozzles of plasma jets. The probe was placed under the plasma nozzle vertically with a distance of 80 mm for the hot plasma jet, and 5 mm for the cold plasma jet. The aperture of the probe silicon core (Ø200 µm) was adjusted to face towards the tungsten halogen lamp to achieve maximum spectral signal-to-noise ratio. Measurements were carried out in duplicate at ambient temperature (∼20 • C). The spectra of gas bases (only with the gas flow but without plasma generation) and the spectra of different gases-generated plasma at 5 min, 10 min and 20 min were collected, respectively.
Spectral instrument control, data acquisition and file conversion were performed using Thorlabs OAS v.2.85 (Thorlabs, USA). The acquired spectral data were saved as .csv files and imported into MATLAB R2019a (The Mathworks, Natick, MA, USA) for asymmetric least squares baseline correction and to resolve smoothness on the baseline corrected spectra with every five channels for convolution.

Develop a correlation between reactive species in PAW and air
The correlation between X and Y was developed using PLS regression modelling to investigate the transforming/interaction of reactive species from/between the plasma gas phase to the liquid phase. PLS regression is a commonly used method for spectral analysis, which aims to develop the linear relation between the spectra (Xs) and the reference parameters (Yr) [61]. Within PLS analysis, Xs and Yr matrices are transformed into new spaces first, then the obtained data named Xs scores and Yr scores are selected and correlated in an attempt to get the maximum interpretation of Yr scores by Xs scores. After that, the predicted Yp scores are subsequently employed to produce the prediction of Yp. More details of PLS regression can be found in the study of Geladi and Kowalski [62]. The performance of PLS modelling was evaluated using the values of the regression coefficient of determination (R 2 CV) and root mean square error (RMSECV) from cross-validation. The regression coefficient intensities derived from the best-performed PLS modelling were highly correlated with the reactive species that were generated from plasma treatments and the wavelengths with the most correlation to the reactive species were selected using the selectivity ratio SR method [63].

Reactive species in PAW
Results of reactive species in PAW generated using cold and hot plasma jets are shown in figure 3. The reactive species (i.e. H 2 O 2 , NO − 3 , and NO − 2 ) of the cold plasma generated PAW were in the concentration ranges of 0.0126-0.0307, 0.0138-0.1572 and 0.0061-0.0422 mM, respectively. While for hot plasma generated PAW, the concentration ranges of these reactive species were 0.0119-0.1058, 0.0106-0.0287 and 0-0.0008 mM, respectively. The concentrations of reactive species in both cold and hot plasma generated PAW significantly (P < 0.05) increased after plasma treatment, compared with the control samples.
In cold plasma-generated PAW, the concentrations of reactive species positively correlated with plasma treatment time, except for the concentration of NO − 2 in N 2 plasma-treated PAW, the highest concentration was found in the samples treated by the plasma for 10 min. The concentration levels of reactive species varied with the use of plasma based on different feeding gases. The reactive species with the highest concentration in air PAW is NO − 3 after the treatment of 20 min. In  In PAW, oxidation reactions initiated by the ROS and RNS are mainly responsible for their various application effects [1,25]. Atomic oxygen (O) and hydroxyl radicals (·OH) as the primary products in PAW are the most important species that can react with almost all cell components, leading to etching and oxidation processes [36,37]. The generation of ·OH· is based on electronic dissociation or the collisions with longlived species, as shown in the following reactions [64]: Then with the presence of the primary species, hydrogen peroxide H 2 O 2 as the secondary species (e.g. H 2 O 2 , O 3 and ONOOH) was produced by the following reaction [36,65]: If ambient air or nitrogen is used for plasma generation, nitric oxide (NO) was formed through the reaction with O 2 − , and then further oxidized to NO 2 and NO 3 , as shown in equations (8)-(12) [37] Some interesting results were shown in this study. Firstly, the concentration of NO − 2 decreased at the longest treatment time in the cold plasma in nitrogen. It may result from two pathways, one is the reaction with H 2 O 2 to produce ONOOH, and the other is the transformation into NO − 3 [66]. These reactions may also contribute to the increase of NO − 3 in the PAW. Secondly, the concentration of H 2 O 2 generated by cold plasma in CO 2 is higher than in the air and N 2 . While the reactive species (i.e. H 2 O 2 , NO − 3 , and NO − 2 ) generated by hot plasma in CO 2 was lower than in the air and N 2 . As the effect of CO 2 has been rarely investigated compared with compressed air and nitrogen, the possible influence of CO 2 on the production of the determining reactive species is still unknown [67]. Thirdly, the concentration of H 2 O 2 generated by both plasma systems increases with the treatment time, a similar trend has also been reported in another study, where the concentration of H 2 O 2 in plasma-treated water was increased from 0 to 0.01 mM for about 5 min treatment [68]. However, the reason why the concentrations of H 2 O 2 generated by cold plasma in CO 2 and hot plasma in nitrogen are obviously higher than the others still needs to be further explored.

OAS spectra of the plasma.
Plasma spectra (390-600 nm) of different gases and treatment times are shown in figures 4 and 5. The spectra in the wavelength range of 390-500 nm consist of more identified absorbance peaks than that in the wavelength range of 500-600 nm. The general spectral shapes were stable with different plasma treatment times. The absorbance peaks that correlated to reactive species generated by different plasma were identified in tables 1-3, with the information on energy level transitions, which agree with the findings of some previously published studies [69,70].

Spectra of plasma for different treatment times.
Under different plasma treatment times (i.e. 5, 10 and 20 min), the spectra of cold and hot plasma that were generated with a specific plasma gas (i.e. air, CO 2 or N 2 ) are shown in figures 4 and 5, respectively. For both cold and hot plasma spectra, higher absorbance intensities of the gas bases were found in all the peak areas, especially for air and N 2 plasma. However, when using CO 2 as feeding gas, the differences in absorbance intensity were not obvious for different treatment times. The results revealed that the spectral intensity pattern aligned in graphs following the logical order of plasma treatment time. This phenomenon might be explained as the plasma was generated and measured in a sealed dark cabinet, and the concentrations of reactive species in the air changed with the operations of plasma treatment time. Before the wavelength of 464.40 nm (related to O + ), spectral intensities of the gas base are slightly higher than those of the plasma operated; while after this critical wavelength, the situation is inverse, as shown in figures 4(a) and 5(a).
As shown in figure 4(a), obvious spectral features of cold air plasma were found at 425, 436, 438, 443 and 464 nm that related to N 2 + , O 2 + , O, N + and O + , respectively. For cold N 2 plasma, relatively high spectral absorbance intensities can be observed at 416, 425 and 444 nm, which were related to N, N 2 + and N 2 , respectively (figure 4(c)). When using CO 2 as the feeding gas, the absorbance spectral features of cold CO 2 plasma were observed at 434 and 436 nm that were related to O 2 + and C 2, respectively. While fewer variations of absorbance intensities for different treatment times were shown in the spectra of cold CO 2 rather than that of air and N 2 plasma ( figure 4(b)).
Similar to the spectra of cold plasma, high absorbance intensities of hot air plasma were found at 395, 400  N, respectively (figure 5(a)); obvious absorbance intensity of hot CO 2 plasma were observed at 434 and 435 nm that related to O 2 + and C 2 , respectively (figure 5(b)); for hot N 2 plasma, the obvious spectral absorbance intensities were found at 415, 425 and 444 nm are related to N, N 2 + and N 2 , respectively ( figure 5(c)).

Spectra of plasma for different feeding gases.
The absorbance spectra of cold and hot plasma generated with different feeding gases (i.e. air, CO 2 , and N 2 ) at a specific treatment time (i.e. 5, 10 and 20 min) are shown in figure 6. For cold plasma, the spectral absorbance intensities vary with different plasma treatment times. No variations of the three spectra were observed for 5 min plasma treatment in figure 6(a-1); while obvious differences in spectra were revealed when the treatment time increased to 10 and 20 min, as shown in figure 6(a-2) and (a-3). Different absorbance peaks of the three cold plasmas were mainly found from 430 to 500 nm that related to O 2 + , O, N + , O + of air, O 2 + , C 2 of CO 2 , and N 2 + , N 2 of N 2 plasma, respectively. For hot plasma, the differences in spectral intensities of different plasma gases were shown for 5, 10 and 20 min treatments (figure 6(b-1), (b-2) and (b-3)), which were mainly found from 390 to 520 nm that related to O, N + , N + , N 2 + O 2 + , O, N + , O + , N + , O + of air, CO 2, O 2 + , C 2 of CO 2 and N, N 2 + , N 2, N + , N + of N 2 . There were no obvious differences shown in the spectra of different gas-generated plasma with the increasing treatment time.

Correlation between reactive species in gas and PAW
PLS regression modelling was developed to investigate the correlation between X and Y. The SR method was used in this study to search the important spectral variables in the prediction property of PLS regression modelling. The most important spectral wavelengths were selected based on the calculated SR scores, which were defined by the ratio between the explained and unexplained variances for each spectral variable through the calculation of both the predictive ability (regression vector) and the explanatory ability (spectral variance/covariance matrix). An F-test was used to decide the critical value of the F distribution with 95% confidence (F c ) in order to define a boundary between variable regions with high discriminating ability and less interesting regions. The spectral variables (with SR scores >F c ) were eventually selected as the most relevant spectral variables that related to the reactive species (i.e. H 2 O 2 , NO − 3 and NO − 2 ). The plot of the PLS regression model and the mean spectra with SR selected wavelengths for the prediction of each reactive species are shown in figure 7.
The PLS cross-validation plot and the mean spectrum with SR selected spectral variables (highlighted in black) for the prediction of H 2 O 2 that was generated using cold air plasma with 5, 10 and 20 min of treatment time are shown in figures 7(a-1) and (a-2). Results of the PLS model ( figure 7(a-1)) revealed R 2 CV of 1.000 and RMSECV of 1.243 × 10 −5 mM; the model was developed using 2 latent variables and 20 SR selected wavelengths. The results demonstrated the collinearity between the OAS spectral data and the measured H 2 O 2 . The SR selected wavelengths (n = 20) that are highly related to H 2 O 2 in PAW which was treated using cold air plasma located in 438-464 nm ( figure 7(a-2)). The wavelengths at 438, 443 and 464 nm, were also corresponding to the chemical bonds of O, N 2 and O + , respectively (table 1). Additionally, Bogumil et al [71] also reported that the absorption cross-section spectral features were observed in the wavelength range of 234-440 nm that related to O 2 .  that was generated using hot air plasma with 5, 10 and 20 min of treatment time with R 2 CV of 1.000 and RMSECV of 4.491 × 10 −6 mM. SR selected spectral variables (n = 11) for the PLS modelling were mainly located in the wavelength range of 353-380 nm and also at 442 and 447 nm ( figure 7(b-2)). It can be concluded that these wavelengths are related to the functional group of NO − 3 ; which has also been reported by other publications based on absorption cross-section [72][73][74].
For the prediction of NO − 2 that was generated using hot plasma with different feeding gases for 10 min, the PLS regression model was developed using 1 latent variable and 55 SR selected spectral variables with R 2 CV of 0.955 and RMSECV of 1.572 × 10 −3 mM (figure 7(c-1) and (c-2)). The SR selected wavelengths (n = 55) were in the ranges of 356-376 nm and 425-456 nm. The specific wavelengths at 425 and 444 nm were corresponding to N 2 + and N 2 , respectively (table 3). The absorption cross-section peaks in these wavelength ranges were reported to be assigned to NO − 2 [75][76][77]. As shown in figures 7(a-1), (b-1) and (c-1), high collinearities (R 2 CVs ∼ 0.955-1, RMSECVs < 0.001) were found between the OAS spectral data of reactive species (i.e. H 2 O 2 , NO − 3 and NO − 2 ) existing in ambient air and the wet chemistry colourimetric results of these reactive species in PAW after the analysis using PLS regression modelling with the SR algorithm. The strong correlations between the OAS spectral results and the wet chemistry-based colourimetric results revealed the phenomenon that excited reactive species generated from plasma can be transferred into or interacted with the liquid and then get preserved. It was consistent with the previous studies of Ma et al [25] and He et al [31].

Conclusions
Diagnostics of reactive species generated using cold and hot plasma jets were investigated using OAS and wet chemistrybased colourimetric methods. The concentration of reactive species (i.e. H 2 O 2 , NO − 3 , NO − 2 ) in PAW varied when using different feeding gases (i.e. air, CO 2 , and N 2 ) to generate plasma; while it is positively correlated to plasma treatment time (i.e. 5, 10, and 20 min). The OAS spectral features of plasma show obvious differences when using different feeding gases and treatment times, especially at the wavelength range of 390-500 nm. PLS regression modelling was used to establish the correlation between OAS spectral data and the estimated concentrations of reactive species in PAW. The SR scores derived from PLS regression modelling were used to select the most relevant spectral variables that related to the reactive species. In the current study, both quantitative and qualitative investigations on plasma reactive species were exploited for the diagnostics of the specific plasma reactive species, which helps to understand the fundamentals of plasma processing. However, the mechanisms of plasma-induced reactions are still incompletely understood and need to be further explored.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).