Corona Discharge and Field Electron Emission in Ambient Air Using a Sharp Metal Needle: Formation and Reactivity of CO3−• and O2−•

CO3−• and O2−• are known to be strong oxidizing reagents in biological systems. CO3−• in particular can cause serious damage to DNA and proteins by H• abstraction reactions. However, H• abstraction of CO3−• in the gas phase has not yet been reported. In this work we report on gas-phase ion/molecule reactions of CO3−• and O2−• with various molecules. CO3−• was generated by the corona discharge of an O2 reagent gas using a cylindrical tube ion source. O2−• was generated by the application of a 15 kHz high frequency voltage to a sharp needle in ambient air at the threshold voltage for the appearance of an ion signal. In the reactions of CO3−•, a decrease in signal intensities of CO3−• accompanied by the simultaneous increase of that of HCO3− was observed when organic compounds with H–C bond energies lower than ∼100 kcal mol−1 such as n-hexane, cyclohexane, methanol, ethanol, 1-propanol, 2-propanol, and toluene were introduced into the ion source. This clearly indicates the occurrence of H• abstraction. O2−• abstracts H+ from acid molecules such as formic, acetic, trifluoroacetic, nitric and amino acids. Gas-phase CO3−• may play a role as a strong oxidizing reagent as it does in the condensed phase. The major discharge product CO3−• in addition to O2−•, O3, and NOx• that are formed in ambient air may cause damage to biological systems.


INTRODUCTION
Basic data on gas-phase and condensed-phase reactions of CO 3 −• and O 2 −• are of importance in aeronomy, environmental chemistry, biology, and medicine. In Scheme 1, the mechanisms for the formation of CO 3 −• , HCO 3 − , and NO 3 − via consecutive gas-phase ion/molecule reactions originating from O 2 , H 2 O, and N 2 are summarized. In gas phase reactions of CO 3 −• , O −• transfer reactions with the formation of CO 2 were found to be the major reaction channels for NO, NO 2 , SO 2 , 1) N 2 O 5 , 2) and 2,4,6-trinitrotoluene. 3) Fehsenfeld et al. 4) and van der Linde et al. 5) reported that CO 3 −• reacted with HNO 3 via a proton transfer reaction to form NO 3 − and HCO 3 • . In addition, van der Linde et al. 6) studied gas-phase reactions of CO 3 −• with formic acid (HCOOH) to form [HCOO − ····OH • ] using FT-ICR mass spectrometry. Ninomiya et al. 3) predicted that CO 3 −• reacts with H 2 O 2 to form the cluster ion, O 2 −• ····H 2 CO 3 . As of this writing, however, no H • abstraction by CO 3 −• in the gas phase has been reported, even though H • abstraction reactions are a major concern due to its potential for causing damage in biological systems. 7) us, it would be of interest to examine the issue of whether CO 3 −• also abstracts H • from organic compounds in the gas-phase. Kawashima  in the gas phase. In reactions of CO 3 −• in aqueous solutions, Elango et al. 9) reported that CO 3 −• reacted with aliphatic amines by (i) a H • abstraction to form HCO 3 − and (ii) an electron transfer to form CO 3 2− (one electron oxidation). e former is more probable in cases of primary amines, while tertiary amines reacted via electron transfer. Cli on and Huie 10) measured rate constants in aqueous solutions for H • abstraction reactions of CO 3 −• with several saturated alcohols and cyclic ethers. e Arrhenius pre-exponential factors ranged from 2×10 8 to 1×10 9 M −1 s −1 and the activation energies ranged from 3.8 to 6.9 kcal mol −1 (1 cal= 4.18 J). Crean et al. 11) investigated the oxidation of single-stranded oligonucleotides by CO 3 −• , leading to the generation of intrastranded cross-links between guanine and thymine bases that were separated by cytosines. Roginskaya et al. 12) studied the e cacy and site speci city of H • abstraction from DNA 2-deoxyribose by CO 3 −• and also evaluated the selectivity of damage in double-stranded DNA. Karmakar 14) and CH 3 CO 2 CH 3 , 14) [2] charge (electron) transfer reactions for CCl 2 F 2 , 15) CCl 3 F, 15) SF 6 , [16][17][18] 2,4,6-trinitrotoluene, 19) and O 3 , 20) [3] H + abstraction reactions for CF 3 SO 3 H, 21) HCl, 21,22) FSO 3 H, 23) and HNO 3 , 4) and [4] clustering reactions for CH 3 CN, 23) (CF 3 ) 2 CO, 14) H 2 C= CHCN, 14) (CH 3 ) 2 CO, 14) and higher hydrocarbons. 24

Ion source assembly for the formation of CO 3
−• e mass spectrometric measurements were performed with a time-of-ight mass spectrometer (AccuTOF, JEOL, Akishima, Tokyo, Japan). Figure 1(a) shows the assembly of the cylindrical ion source tube that was used for the formation of CO 3 −• . e ambient air open distance between the terminal end of the cylindrical tube and the ion sampling ori ce of the mass spectrometer was 8 mm. When CO 2 was used as the reagent gas for the formation of CO 3 −• , O 2 −• with a relative intensity of about 20-30% compared to CO 3 −• was unavoidable, as shown in Fig. S1. e formation of O 2 −• may originate from O 2 contamination in the CO 2 reagent gas or O 2 formed by the decomposition of the CO 2 reagent gas in the corona discharge plasma. In contrast, when O 2 was used instead of CO 2 as the reagent gas, CO 3 −• was generated as the major ion and the formation of O 2 −• was negligible (see Fig.  2(a)). is indicates that O 3 −• produced by the DC discharge of the reagent O 2 gas was e ciently converted into CO 3 −• . e source of the carbon for the formation of CO 3 −• may be the impurity of CO 2 in the O 2 reagent gas, adsorbed CO 2 and/or CO on the wall of the gas pipe line. e use of a stainless steel throttle shown in Fig. 1(a) is essential for the generation of CO 3 −• as the major ion. When the throttle was not used, the back di usion of air into the inside of the tube could not be avoided, even when the tip of the needle electrode was recessed a distance of 10 mm from the exit of the glass tube. e back di usion of air was readily recognized by the detection of HCO 3 − originating from the moisture in the air (see Scheme 1(b)). When the tip of the needle electrode was recessed a distance of 30 mm from the exit of the mesh-covered glass tube with the throttle inserted near the exit, the back di usion of air was nearly completely avoided and CO 3 −• ions were formed as the only major reactant ion owing out of the glass tube. us, the back di usion of reactant vapor introduced into the plasma region of the ion source appears to be negligible in the present experimental setup.
e O 2 reagent gas with a ow rate of 3 L min −1 was ionized by a direct current (DC) corona discharge. e consecutive reactions (4)-(11) took place in ambient air (Scheme 1(a)). e major ion O 2 −• that was generated was formed by an electron attachment reaction (4) (11)).
e changes in the enthalpy for reactions (10) and (11) were calculated to be −37. 6 19) and −11.8 25) kcal mol −1 , respectively, and reactions (10) and (11) proceed with nearly collision rates. 23) O 2 −• formation using a sharp metal needle in ambient air e simple experimental setup for this process is shown in Fig. 1(b). e formation of O 2 −• as the major ion was only observed when a high-frequency (15 kHz) threshold voltage for the appearance of the ion signal was applied to the needle in ambient air. is phenomenon is attributed to the eld electron emission from the tip of the sharp needle (see the latter section).

Reactions of CO 3 −• with various molecules
By using the ion source shown in Fig. 1(a), reactions of CO 3 −• with hydrocarbons (n-hexane, cyclohexane, benzene, and toluene), alcohols (methanol, ethanol, 1-propanol, and 2-propanol), acetonitrile, acetone, water, and H 2 O 2 were examined. A 10 µL aliquot of a liquid sample was placed on the well of the heater shown in Fig. 1. e heater temperature was maintained at a temperature of about 30°C above the boiling point of the liquid.
As an example, experimental results obtained for nhexane are shown in Fig. 2. Figure 2 ese results clearly indicate that CO 3 −• abstracts H • from n-C 6 H 14 , with the formation of HCO 3 − in the gas phase (reaction (12)).
Gas-phase H • abstraction reactions by CO 3 −• have not been extensively explored, which may be due to the fact that reactant molecules are not detected as ions in H • abstraction reactions. In this work, H • abstraction reactions were also detected for cyclohexane, toluene, methanol, ethanol, 1-propanol, and 2-propanol (data not shown). However, H 2 O, benzene, acetone, and acetonitrile did not show any noticeable reactivity toward CO 3 −• under the present experimental conditions. From the value for the heat of formation of HCO 3 − (−173.9 kcal mol −1 ), 26) the bond energy of H • ····CO 3 − was estimated to be 105.7 kcal mol −1 . did not show any noticeable reactivity toward acetonitrile, but there may be a substantial entropy barrier for this reaction. In the present work, occurrence/non-occurrence experiments on the reactivity of CO 3 −• in the gas phase were conducted for the rst time, though no quantitative information such as rate constants was obtained.
In order to determine whether or not CO 3 −• can abstract a hydrogen atom from organic molecules (Ms), DFT calculations were carried out for the reaction of CO 3 −• with three representative molecules, namely, benzene, n-hexane and toluene. e computational method is M06-2X/6-311+ +G(2p,d), which gives very accurate thermochemical data and activation energies. 27) Using this approach, the geometries of the transition state (TS) were determined. Subsequently, the intrinsic reaction coordinate (IRC) was traced to obtain the geometries of weakly bound complexes, In our previous work, 3) based on theory, we predicted that CO 3 −• reacts with H 2 O 2 to form O 2 −• ····H 2 CO 3 cluster ions. However, this cluster ion was not detected experimentally in our previous study. 3) is may be due to the relatively low abundance of CO 3 −• generated by the ambient-air corona discharge. Figures 3(a) and 3(b) show the mass spectra before and a er the introduction of H 2 O 2 into the ion source shown in Fig. 1(a), respectively. In Fig. 3(b)

Field electron emission by the application of highfrequency voltage to the sharp metal needle
As shown in Scheme 1 in the ambient-air corona discharge, O 2 −• as the intermediate ion is rapidly converted into CO 3 −• as the terminal product ion via reactions (4)- (11). us, investigating the reactivity of O 2 −• as the single reactant ion is di cult using corona discharge as the ion source. In fact, previous investigations of the reactions of O 2 −• have primarily been conducted using owing a erglow techniques. 23) However, we discovered that O 2 −• was only formed as the product ion only when a high frequency voltage, but not a DC voltage, was applied to the needle electrode in the open ambient air (see Fig. 1(b)). (m/z 60) indicated that gas breakdown followed by the generation of a corona discharge did not occur at this threshold voltage of ±1150 V. e formation of O 2 −• as the predominant ion suggests that free electrons were generated and they were converted into O 2 −• in ambient air by the electron attachment reaction (4). It is likely that free electrons were generated by the eld electron emission from the tip of the needle electrode. e conceptual scenario for the tunneling electron emission is depicted in Fig. S4. As shown in Fig.  4(b), O 3 −• and CO 3 −• started to be detected when the AC voltage was increased to ±1250 V. is suggests that the AC corona discharge started to contribute to ion formation in addition to eld electron emission at ±1250 V. With a further increase in the AC high voltage to ±1600 V, NO x − ions (x=2, 3) originating from the decomposition of N 2 (in Scheme 1(c)) were detected.

Figures 4(e)-4(h)
show negative-mode mass spectra when a negative DC high voltage was applied to the needle. Figure  4(e) shows the mass spectrum when a threshold voltage of −1500 V was applied to the needle for the observation of ion signals. It should also be noted that the threshold voltage of −1500 V is considerably higher (more negative) than that of the negative-phase AC voltage of −1150 V shown in Fig. 4(a). It is apparent that in the DC mode of operation, eld electron emission is largely suppressed and a corona discharge is directly generated at −1500 V. Furthermore, in the DC mode, the eld electron emission must occur at around −1150 V at the precise moment of when a high voltage is applied to the needle. However, due to the continuous application of a negative DC high voltage to the needle, free electrons were emitted and O 2 −• formed by electron attachment may have accumulated near the tip of the electrode. e accumulated negative charge near the tip of the needle should result in the formation of a space-charge eld that shields the electric eld at the tip of the needle. Due to the decrease in the electric eld, the eld strength at the needle tip becomes lower than that needed for the eld electron emission. Such a build-up of the space charge eld can be avoided by the application of an AC voltage because the free electrons and resulting O 2 −• that are produced in the negative voltage phase can be completely scavenged by the metal needle by the subsequent positive phase high voltage that is applied, as shown in Fig. 5.
A separate experiment was performed to examine the electron scavenging e ect suggested above.   Fig. S5(e). It is evident that a bias voltage of +500 V is effective for scavenging electrons that are emitted by the tunneling e ect and this suppresses the occurrence of corona discharge. However, a pulse width of 1 µs in Fig. S5(f) is suf-cient for gas breakdown to occur as discharge product ions such as CO 3 −• start to be detected. e scavenging e ect of the build-up of negative charges near the needle tip in the positive phase of AC high voltage should be dependent on the frequency of the AC high voltage. Fig. S6 shows mass spectra obtained when the frequency of the AC high voltage was changed in the range of 20 kHz to 5 kHz measured at the threshold voltage for the eld electron emission. Cl − , HCOO − , and CH 3 COO − product ions were formed by the reactions of O 2 −• with HCl, formic acid, and acetic acid contaminants, respectively, that are present in the laboratory air (see the latter section). ere seems to be no noticeable frequency dependence on the eld electron emission in the range of 5-20 kHz. In addition to the use of a stainless steel acupuncture needle, various other metal needles were tested as electrodes. Metal wire, with a diameter of 0.1 mm, was cut tangentially by a nipper and then sharpened using Emery paper (# 1000) and was used for an emitter. Among the tested metals (Ti, W, Cr, Co, Mo, Pt, Pd, Fe, Au, Ni, Ir, Cu, constantan (Cu/Ni alloy), Pd/Pt(1/9)), Ti, Pd, constantan, Pd/ Pt(1/9), and Cr were found to be appropriate for eld electron emission. ere is a crude trend that metals with lower work functions are better-suited as eld electron emitters.
In our previous paper, an AC corona discharge was applied to an atmospheric-pressure chemical ionization (APCI) ion source for the rst time. 29) e AC corona discharge was found to be superior to a DC corona discharge  for various reasons 3,29) : (i) corrosion of the needle electrode by the AC corona is much less than that for a DC corona, (ii) both positive and negative ions can be detected without changing the polarity of the high voltage power supply, (iii) an AC corona gives as strong positive and negative ion signal intensities as a DC corona even though an intermittent plasma is generated in the AC corona, (iv) ionization by an AC corona is milder than that for a DC corona, (v) transition to arc discharge for an AC corona is largely suppressed compared to that for a DC corona. ese characteristic di erences between AC and DC corona discharges can be envisaged by the observation of positive-mode mass spectra obtained by AC and DC corona discharges. Figure 6 shows the positive-mode mass spectra for ambient air measured under the same experimental conditions as in Fig. 4. As suggested in Fig. 4(b), the corona discharge started at the threshold voltage of ±1250 V in the positivemode for the AC corona discharge in Fig. 6(a). e signal intensities for [(H 2 O) n +H] + (n=2, 3) increase only gradually with increasing AC voltage from ±1250 V to ±1600 V. In contrast, for the DC corona discharge as shown in Figs. 6(d)-6(f), the ion signal intensities increase steeply with increasing applied DC voltage from the threshold voltage of +1900 V to +2100 V. With a further increase in +DC high voltage, a transition to arc discharge was anticipated.
It should be noted that the threshold voltage for the DC corona discharge (+1900 V) was much higher than that for the AC corona discharge (±1250 V). is indicates that AC and DC corona discharges are based on quite di erent breakdown mechanisms. Plasma is an electrically conducting media composed of positive and negative charges and is generated by an electron avalanche induced by electrons accelerated in a high electric eld in an insulating media.
In the positive-mode DC discharge, nascent electrons that trigger the discharge are generated accidentally by cosmic rays or photoelectrons. e greater fraction of the nascent electrons are attracted to the anode, where they are annihilated by the high electric eld near the anode. Due to the paucity of electrons that act as the seeds for gas discharge breakdown, a high threshold voltage is necessary for initiating a positive-mode DC corona discharge. In addition, the incidental generation of nascent electrons may lead to the positive-mode corona discharge being unstable. Owing to the application of a positive potential to the anode for the positive-mode DC corona, positive ions accumulate near the tip of the needle electrode leading to the formation of a Debye sheath that shields the potential applied to the needle. is explains why a high positive potential is necessary for maintaining a stable positive-mode DC corona discharge. In contrast, in the AC corona discharge, electrons are supplied to the insulating medium by the eld electron emission in the negative-phase voltage, enabling the maintenance of discharge with a much lower applied voltage. In summary, the AC corona discharge can be maintained with a much lower voltage than DC corona discharge for both positive-and negative-mode of mass spectrometric operation.
A negative-mode corona discharge is bene cial for the detection of molecules that have positive electron a nities because all electrons are eventually converted into negative ions by electron attachment reactions such as the formation of O 2 −• in this experiment. As described in the introduction, CO 3 −• causes oxidative damage to biological systems such as DNA and proteins. As shown in Fig. 4 , O 3 , etc. ere are many commercially available household appliances that use a corona discharge for the sterilization of bacteria and virus in air. To examine the kinds of ions that are formed by sterilizers that use a corona discharge, the plasma-activated air owing out from a commercial air sterilizer (USB type, Air Success Mini, Air Success, Kanagawa, Japan) was measured. Figure 7 shows a mass spectrum for air ionized by the negative DC mode multiple-ring corona discharge that is installed in the Air Success Mini. e mass spectrum is very similar to those shown in Figs. 4(e)-4(h) and oxidative CO 3 −• was detected as one of the major ions.

Reactions of O 2
−• with various molecules e reactions of O 2 −• with hydrocarbons, alcohols, acetone, and acetonitrile were examined by placing 10 µL liquid samples in the heater shown in Fig. 1(b). Neither H • nor H + abstraction reactions were observed for these compounds. However, when 10 µL formic acid, acetic acid, and tri uoroacetic acid were introduced into the ion source, the respective deprotonated carboxylate ions HCOO − , CH 3 COO − , and CF 3 COO − were clearly detected as the major ions. A er these three measurements, a mass spectrum for laboratory air contaminated by these three acids was collected, as shown in Fig. S7. All three acids were clearly detected indicating that the present eld-electron-emission type ion source is suitable for the detection of trace amounts of acids. Deprotonated ions were also detected for several amino acids (leucine, isoleucine, alanine, and phenylalanine). Figures 8(a) and 8(b) show the mass spectra before and a er introducing phenylalanine (Phe) into the ion source. Approximately 10 s a er the deposition of a 10 µL aqueous solution of a 10 −3 M phenylalanine on the heater at 140°C, deprotonated [Phe−H] − and [Phe+ O 2 ] −• cluster ions started to be detected. e ion signals continued to be detected for much longer than 10 s due to the slow evaporation of the phenylalanine at 140°C (melting point: 283°C, boiling point: 295°C). Fig. S8 shows the results obtained for nitric acid. Fig. S8(a) shows the mass spectrum before sample introduction, in which O 2 −• is detected as the only major ion. Fig. S8(b) shows the mass spectrum obtained when a cotton ball that was wetted by a 30% aqueous nitric acid was positioned in close proximity to the ion source at 0. 24

CONCLUSION
In gas-phase reactions of CO 3 −• , O −• transfer, O 2 −• transfer, and H + abstraction reactions with inorganic and organic molecules have been studied to date. However, H • abstraction reactions with organic molecules, although of interest, have not been reported. In this work, occurrence/nonoccurrence experiments of H • abstractions of CO 3 −• with various molecules in the gas phase are reported for the rst time. H • abstraction was observed for n-hexane, cyclohexane, methanol, ethanol, 1-propanol, 2-propanol, and toluene, but no reactions were observed for acetonitrile, acetone, benzene, and H 2 O. DFT calculations clearly demonstrated the reason for this contrast between the occurrence for toluene and n-hexane and the nonoccurrence for benzene. In biological systems, CO 3 −• is capable of causing serious oxidative damage to proteins and DNA molecules via H • abstraction reactions. It should therefore be assumed that air sterilizers with the function via the use of a corona discharge ion source evolve CO 3 −• ions as the major ions, which could be harmful to mucous membranes such as lungs.
When an AC high voltage was applied to the sharp metal needle electrode in ambient air, tunneling electron emission from the tip of the needle was observed and the generation of electrons were detected as O 2 −• by an electron attachment reaction. O 2 −• did not show any reactivity toward hydrocarbons or alcohols but it abstracts H + from acid molecules such as formic acid, acetic acid, nitric acid and amino acids. By investigating the threshold behavior of ion formation for