On-demand switching of oxidation

Fig. 3 shows the switching between nonoxidizing and oxidizing conditions by modulating the HV applied to the emitter. Using spray current as the feedback signal, the HV was feedback controlled with a personal computer using a custom program written in C# (source code in the ESI^†). The spray current was step-modulated between 200 and 600 nA (Fig. 3a). The ratio of the EIC of the oxidized peak to the EIC of the non-oxidized peak responded near-instantly (in less than 1 s) to the spray current (Fig. 3b). This indicated that the oxidation was strongly dependent on the spray current, hence the flow rate. As the flow rate was switched to 8.4 nL min⁻¹ from 0.9 nL min⁻¹, the oxidized species also depleted immediately back to the mass spectrum shown in Fig. 3c(i) and (ii).

Open in a separate window

Fig. 3

Step modulation of spray current. On-demand oxidation by step-modulation of spray current for 10 μM melittin in 1 M ammonium formate. (a) Waveform of the measured spray current. The HV applied to the emitter is feedback-controlled using spray current as the feedback signal. The preset values for the spray current are 200 and 600 nA. (b) The ratio of total oxidized ion species to the overall ion species. (c) Mass spectra of melittin acquired at (i) high flow rate (600 nA) and (ii) low flow rate (200 nA). Asterisk denotes the unoxidized peak.

It is well known that sustained electrospray is associated with the inherent electrochemical reaction (ECR), thus one may wonder if the ECR was involved in the observed oxidation. Here, an inserted platinum wire was used as the charging electrode, and the distance between the distal end of the electrode and the tip of the ESI emitter was approximately 10 mm. Under the operating flow rate of several to several tens nL min⁻¹, it would take more than 10 minutes to transport the oxidation products generated by the ECR at the electrode to the Taylor cone. Likewise, it would take several minutes to consume the ECR product when the oxidation was turned OFF. The near instantaneous response of oxidation to the change in spray current provided clear evidence that the observed phenomenon was not relevant to ECR at the electrode. Instead, the oxidation was likely to take place at the tip of the Taylor cone, on/inside the charged droplets, or in the gas phase. It is noted that as the solutions used in this study were highly conductive, the spraying and the oxidation condition was not affected by the position of the platinum electrode.

Dependence on the electric field

The oxidation was believed to involve the reactive oxygen species (ROS). The formation of ROS here was not spontaneous, i.e., it needed external activation or energization. We hypothesize that the strong electric field on the tip of the Taylor cone or the surface of the initially charged droplet was responsible for the observed oxidation phenomenon. Under the condition near the lowest flow rate and spray current, the formed Taylor cone was the sharpest, generating the finest charged droplet. Here the correlation between the electric field and the threshold to induce oxidation was evaluated using cytochrome c in ammonium acetate and ammonium formate solutions of different electrical conductivities. The measurement began by operating the ESI under nonoxidizing conditions at a high current and high flow rate. The naturally occurring oxidized species detected at the high flow rate were defined as the baseline. The flow rate was gradually decreased by reducing the emitter voltage until the ion intensity of the total oxidized species accounted for 10% of the TIC of cytochrome c. The measured spray current at that point was assigned as the threshold current for the estimation of the electric field that triggers the oxidation. Following the scaling law,²⁸ the radius of the initially produced charged droplet R_o can be approximated using²⁸

2

where ε_o is the vacuum permittivity. The flow rate Q can be readily calculated from the spray current following eqn (1). For an order of magnitude analysis, we assume the stress owing to the surface tension equals that of the electrostatic field (i.e., at Rayleigh limit). In that case, the electric field strength on the initial droplet surface E_o is

3

The value estimated using eqn (3) should reflect the order of magnitude of the electric field on the droplet surface as well as on the tip of the Taylor cone when the droplet was generated. In terms of spray current, E_o is

4

Depending on the electrical conductivity, the spray currents (or the flow rates) required to observe the oxidation were different. However, the plot of the radius R_o and electric field E_o in Fig. 4 shows an interesting result that the initialization of oxidation takes place at an average R_o of ~19.2 nm and an E_o of 1.3 V nm⁻¹, independent of the conductivity and the choice of the buffer solution. This suggested that the turn-on of oxidation was strongly linked to the electric field associated with the nanoESI.

Open in a separate window

Fig. 4

Threshold electric field. Radius and the electric field of the initial droplet at the threshold for the detection of oxidized species. The sample is cytochrome c (25 μM) in different buffer solutions of various electrical conductivities. Blue square: ammonium acetate. Red circle: ammonium formate. The average threshold radius and electric field are ~19.2 nm and 1.3 V nm⁻¹.

One may argue that the phenomenon was identical to the commonly encountered corona discharge co-generated by the ESI at higher HV. An important discrepancy from the corona discharge-induced phenomenon was that the observed oxidation species were at their highest abundance at the lowest HV and spray current. Also, under the super-atmospheric pressure, an apparent corona discharge was absent as verified by microscopic inspection of the Taylor cone and the measurement of spray current. Still, the possibility of partial discharge could not be excluded because the current was at the same order of magnitude or lower than the electrospray current. We conjecture that despite the absence of apparent discharge, the local electric field was strong enough to accelerate electrons and ions in proximity to the liquid surface for the formation of reactive oxygen species. In the following sections, this oxidation taking place at a low flow rate is referred to as strong electric field-induced oxidation (SEFO). It is noted that the flow rate in eqn (2) or the spray current in eqn (4) cannot be reduced to an arbitrarily small value because to sustain a steady cone–jet mode, there exists a minimum flow rate Q_min which is given as²⁸

5

where ρ is the density of the liquid.²⁸ This minimum flow rate limits the highest possible electric field E_max that could be achieved for a given solution to

6

Eqn (6) implies that for an aqueous solution at room temperature where γ = 0.072 N m⁻¹, ρ = 997 kg m⁻³, ε_r = 80, it requires a minimum electrical conductivity of 2.3 S m⁻¹ to attain the threshold electric field of 1.3 V nm⁻¹ at the minimum flow rate. In practice, K needed to be greater than 3.8 S m⁻¹ so that the operation above Q_min could still produce pronounced oxidation.

Oxidation of the molecular ion

The strong electric field-induced oxidation of caffeine in Fig. 5 shows some unusual results not found in typical ESI measurements. The solvent here was 200 mM HCl aqueous solution with an electrical conductivity of ~7 S m⁻¹. Using a highly conductive solution, the molecular ion of caffeine (M⁺) was detected along with the protonated species ([M + H]⁺) at a relatively higher flow rate (10 nL min⁻¹ < Q < 30 nL min⁻¹). Although the detection of the molecular ion is common for electron ionization and field ionization, caffeine is usually detected almost solely as the protonated species using ESI or APCI. Due to the high electric field, the presence of M⁺ may indicate the presence of field ionization. Another possible ionization route is via charge transfer from the water radical cation which is reportedly present near/at the droplet–air interface under a strong electric field.²⁴ The incorporation of oxygen into the molecular ion was also observed by decreasing the spray current to <10 nL min⁻¹. The abundance of M⁺ was found to decrease with the rise of [M + O]⁺ at m/z 210, and [M + HO₂]⁺ at m/z 227. The intensity ratios of M⁺/[M + H]⁺, [M + O]⁺/[M + H]+, and [M + HO₂]⁺/[M + H]+ are shown, respectively in the subplots of Fig. 5.

Open in a separate window

Fig. 5

Detection of caffeine molecular ion. Detection of caffeine (25 μM) in 200 mM HCl under different spray currents. Red: Intensity ratio of M⁺/[M + H]⁺. Blue: [M + O]⁺/[M + H]⁺. Green: [M + HO₂]⁺/[M + H]⁺. Mass spectra in the inset are taken at 1000 and 200 nA.

Comparison with other oxidation methods

For comparison, the oxidation of cytochrome c was performed using various in situ oxidation devices operated together with the same nanoESI source. Fig. 6 shows the results of (i) atmospheric pressure corona discharge (APCD), (ii) in-emitter electrochemical (EC) reactor, (iii) dielectric barrier discharge (DBD), and (iv) ozone generated from DBD. Mass spectra in row (a) were taken when oxidation devices were in the initial OFF condition. Those in row (b) were taken after the oxidation devices were turned ON and sufficient time had passed for the ion signal to reach a steady state condition. For APCD in column (i), the result was taken by placing a corona discharge needle in between the high-pressure ion source outlet and the MS ion inlet. This arrangement which separated the ESI source from the discharge source was chosen to preserve the stability of the electrospray. The discharge current was >10 μA. Surprisingly, no obvious oxidation was observed. Possible explanations are a low concentration of ROS and poor interaction between the charged droplets/ions and the ROS. The repulsive force may prevent the charged droplet from approaching the ROS-rich high electric field region. Changing the polarity of the corona discharge shifted the charge state distribution to a low charge without any improvement in the oxidation performance. Fig. S5^† shows another APCD measurement in which the corona discharge was co-generated at the stainless-steel emitter of an atmospheric pressure ESI source. The HV was increased up to the level before the fluctuation of the spray but only a weak oxidation peak was detected.

Open in a separate window

Fig. 6

Various oxidation sources. Oxidation of cytochrome c using (i) atmospheric pressure corona discharge (APCD), (ii) in-emitter electrochemical reactor (EC), (iii) dielectric barrier discharge (DBD), and (iv) ozone generated from DBD. (a) Oxidation OFF. (b) Oxidation ON. The red circle denotes the peak due to the incorporation of oxygen into cytochrome c. Asterisk denotes the unoxidized peak. # Sign denotes the peak due to the adduction of HNO₃. Sample: 10 μM cytochrome c in 100 mM ammonium formate.

Column (ii) shows the electrochemical oxidation performed inside the capillary of the emitter. In addition to the primary HV electrode for electrospray, another wire with insulator cladding was inserted into the ESI capillary as the second electrode. A ~3 V potential difference was applied between those electrodes. The HV electrode was the cathode and the second electrode was the anode. The choice of the second electrode was important to prevent anodic oxidation and the formation of metal ions in abundance (Fig. S6^†). Mass spectra in Fig. 6(ii) were taken using a platinum wire with Teflon cladding as the second electrode. After turning ON of the EC cell, the oxidation peak increased gradually and reached a steady state after 20 min. Only one oxygen attachment was observed.

Dielectric barrier discharge was used as shown in Fig. 6(iii) by placing the DBD source in between the ion source and the MS inlet. Glass was used as the barrier material. The discharge current for DBD was in the order of mA, significantly higher than the corona discharge source in (i), and the odor of ozone can be felt from 1 m away from the source. The degree of oxidation was much higher than corona discharge (i) and SEFO, well exceeding the number predicted by Table S1.^† However, the tunability for the degree of oxidation using this method was not easy and the mass spectrum was complicated by the presence of adduct species such as n(HNO₃). Fig. 6 column (iv) shows the oxidation by ozone generated from the DBD. A cylindrical DBD source made of a quartz tube was added in between the ion source and the gas supply line. The positive and negative ions from the DBD were annihilated by the metallic transfer tube and the inner wall of the ion source. This was verified by the observation that no background ion was detected when the DBD was turned ON, and ESI was turned OFF. By electrospraying the solution in the ozone-containing air, the oxidation peak gradually appeared and the intensity reached a plateau after ~5 min. The most dominant oxidation peak was the one with 1 oxygen atom. Even when the DBD was turned OFF, it took approximately 30 min for the mass spectrum to return to the oxidation-free condition, greatly exceeding the time needed to flush the ozone. This indicated that ozone has dissolved into the solution phase via the Taylor cone.

Another comparison with ozone oxidation is shown in Fig. 7 for peptides. The mass spectrum of the ozone-oxidized melittin in Fig. 7a shows a dominant peak with the incorporation of three oxygen atoms. The finding by Richards et al. also showed that the ozonated protein exhibits a predominant peak with a three-oxygen adduct.³³ The result of SEFO (Fig. 3c(ii)), however, shows a different oxygen distribution. The ion with the incorporation of one nitrogen was also detected which might be due to the interaction with other neutral species from DBD. The distributions of oxygen generated from ozone and SEFO were obviously quite different. Another measurement was performed using angiotensin I which was reportedly not to be oxidized by ozone.³³ The same finding was reproduced in our experiment (Fig. 7b) using ozone from DBD. Fig. 7c shows that the oxidation of angiotensin I was still possible by SEFO. In sum, the result in Fig. 7 suggests that the main ROS involved in the SEFO was not ozone.

Open in a separate window

Fig. 7

Oxidation by ozone. (a) Oxidation of 10 μM melittin in 100 mM ammonium formate using ozone produced from the DBD source. (b) Oxidation of 20 μM angiotensin I in 50 mM HCl using ozone. (c) Strong electric field-induced oxidation of angiotensin I in 500 mM HCl. Asterisk denotes the unoxidized peak.

Source of oxygen atoms

The source of oxygen atoms attached to the analytes depends on the origin of ROS, and the subsequent interaction with the water and the analytes. If the phenomenon is discharge-like, the formation of ROS should take place in the gas phase. On the other hand, the strong electric field established by the microdroplet electrical double layer (also reported to be in the order of 1 V nm⁻¹) was recently postulated by Zare and others to generate hydroxyl radicals in the liquid phase which contributed to the observed spontaneous formation of hydrogen peroxide.^25,34 Isotope labeling method was used to determine the origin of the oxygen atom. The analyte was l-phosphatidylcholine (PC 18:1/18:1) with two double bonds (one at each fatty acid chain). Two sets of water solutions were used. One was prepared in normal pure water and another was prepared in H₂¹⁸O (97% atom). The measurement was performed using air and pure nitrogen as the operating gases for the ion source.

The mass spectra acquired using normal water solution and air are shown in Fig. 8a(i) and b(i). Two dominant oxidation peaks were detected owing to the preferential attack of the double bond.²⁶ The result obtained using a water-¹⁸O solution with air as the operating gas is shown in Fig. 8a(ii) and b(ii). The magnified mass spectrum in ​in8b8b shows that even with water-¹⁸O, the dominant oxidation peaks were due to those of [M + H + ¹⁶O]⁺ and [M + H + 2¹⁶O] ⁺. Ion species of [M + H + ¹⁸O]⁺, [M + H + ¹⁶O + ¹⁸O] ⁺, and [M + H + 2¹⁸O]⁺ were detectable but with much weaker ion intensities (see intensity comparison in Fig. 8c). Changing the operating gas to high-purity nitrogen (99.99995%) reduced the intensity of the oxidation peak as shown in Fig. 8a(iii) and 8c. Even under that condition, the oxidation was still dominated by ¹⁶O. A similar result was also observed for the melittin sample prepared in water-¹⁸O (Fig. S7^†).

Open in a separate window

Fig. 8

Measurement using the oxygen isotope. (a) Mass spectra obtained under oxidized conditions for 5 μM l-phosphatidylcholine (PC 18:1/18:1) in 500 mM HCl prepared using normal water (i) and water-¹⁸O (ii & iii). The operating gases are air (i & ii) and high-purity nitrogen (iii). Asterisk denotes the peak without oxidation. (b) Enlarged view of (a) (i) and (a) (ii). (c) Intensity comparison for different oxidation species.

With sufficient acceleration of seed electrons near the charged liquid surface, the oxygen molecules in the proximity could be split into oxygen atoms which could eventually form ozone when combined with other oxygen molecules. Although the contribution by the ozone in SEFO could not be completely ruled out, the mass spectra obtained using external ozone (Fig. 7a, b and 6(iv)) clearly showed that it was not the primary oxidation route. The ozone and/or the split oxygen atoms could react with the water to form hydroxyl radicals (˙OH). The detection of ¹⁸O-related peaks in the water-¹⁸O sample supports the formation of ¹⁸˙OH. The detection of H₂O₂ (which can be formed by ˙OH) in Fig. S8^† also supports the formation of hydroxyl radicals. However, despite its presence, the low abundance of ¹⁸O-related ions indicated that the hydroxyl radical oxidation was unlikely the main oxidation route in SEFO because the reaction of gas phase oxygen species with water-¹⁸O should result in equal amounts of ¹⁸˙OH and ¹⁶˙OH. In sum, the primary source of oxygen in the SEFO was the gas phase ¹⁶O₂, suggesting a possibility of direct oxidation of analytes by the oxygen atoms generated by the strong electric field. It is noted that oxidation by the atomic oxygen generated from the atmospheric plasma was also reported to be possible for the analyte in the solution.³⁵

This work was partially supported by Grants-in-Aid for Scientific Research (KAKENHI) from the Japan Society for the Promotion of Science. Z. H. acknowledges the financial support from the China Scholarship Council (CSC).