2. Experimental

2.1. Materials

The liquids, commercial sources, full anion names and key physical properties available in the literature are listed in table 1. The first three were selected based on their singularly low viscosities and high electrical conductivities. EMI-C(CN) $_3$ was included as a probably less viscous and more electrically conducting salt than the 1-butyl-3-methylimidazolium-tricyanomethanide successfully investigated by Caballero-Perez et al. (Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025). All the experiments presented here have been performed in negative mode for consistency with prior work (Perez-Lorenzo Reference Perez-Lorenzo2022), and to highlight the relevance of the anion in the emission performance.

Table 1. Properties at 25 $\,^\circ \mathrm{C}$ of the ILs investigated: $\unicode{x03BC}$ = viscosity; $\gamma$ = surface tension; K = electrical conductivity; $\rho$ = density; $\text{EMI}$ = 1-ethyl-3-methylimidazolium, $\textrm{SCN}$ = thiocyanate; $\text{N}(\text{CN})_2$ = dicyanamide, $\text{BF}_4$ = tetrafluoroborate, $\text{C}(\text{CN})_3$ = tricyanomethanide. $^{1}$ Seki et al. (Reference Seki, Kobayashi, Kobayashi, Takei, Miyashiro, Hayamizu, Tsuzuki, Mitsugi and Umebayashi2010), $^{2}$ Zhang et al. (Reference Zhang, Li, Zhang and Wu2014), $^{3}$ Wang et al. (Reference Wang, Xing, Zhang, Liu and Zhang2022), $^{4}$ Safarov et al. (Reference Safarov, Nachhu, Hassel and Müller2021), $^{5}$ Klomfar, Souckova & Patek (Reference Klomfar, Souckova and Patek2011), $^{6}$ Schreiner et al. (Reference Schreiner, Zugmann, Hartl and Gores2010), $^{7}$ Neves et al. (Reference Neves, Kurnia, Coutinho, Marrucho, Lopes, Freire and Rebelo2013), $^{8}$ Souckova, Klomfar & Patek (Reference Souckova, Klomfar and Patek2011), $^{9}$ Vila et al. (Reference Vila, Ginés, Pico, Franjo, Jiménez, Varela and Cabeza2006), $^{10}$ Souckova, Klomfar & Patek (Reference Souckova, Klomfar and Patek2015), $^{11}$ (Zubeir et al. Reference Zubeir, Rocha, Vergadou, Weggemans, Peristeras, Schulz, Economou and Kroon2016; Musiał et al. Reference Musiał, Cheng, Wojnarowska and Paluch2020). $^{*}$ Value used is the average between these two studies. $^{12}$ Freire et al. (Reference Freire, Teles, Rocha, Schröder, Neves, Carvalho, Evtuguin and Santos2011), $^{13}$ Fröba et al. (Reference Fröba, Rausch, Krzeminski, Assenbaum, Wasserscheid and Leipertz2010).

2.2. Apparatus

Our instrument offers the unique ability to measure energy and velocity distributions in series, for the whole beam and with angular resolution. This permits an unusually detailed characterisation of the drops and ions produced, even in fairly complex mixtures. Alternative arrangements to infer such information have been previously given by Gamero and colleagues (Gamero-Castaño Reference Gamero-Castaño2008; Gamero-Castaño & Cisquella-Serra Reference Gamero-Castaño and Cisquella-Serra2021) and Prince and colleagues (Miller et al. Reference Miller, Ulibarri-Sanchez, Prince and Bemish2021). The electrosprays are generated from a fused silica capillary (Polymicro Technologies), with inner diameter (ID) 50 $\,\unicode{x03BC} \text{m}$ (40 $\,\unicode{x03BC} \text{m}$ for C(CN) $_3$ ) and length $\sim 0.5{\rm \,m}$ . The emitting end is pulled under a flame, mechanically sharpened to an outer diameter of approximately 36 $\,\unicode{x03BC} \text{m}$ (30 $\,\unicode{x03BC} \text{m}$ for C(CN) $_3$ ) (figure 1), and sputter coated with a thin layer of iridium (Caballero-Perez et al. Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025). The opposite end of the capillary is immersed in the liquid sample, contained in a pressurised reservoir that also provides electrical contact through a platinum electrode. This electrode is connected through a current meter to a negative high voltage power supply. Because there is a non-negligible voltage drop through the length of the capillary, the voltage at the iridium-coated emitter tip is determined separately by an electrostatic voltmeter (Advanced Energy Trek 370) without either physical contact with the meniscus or charge transfer. This arrangement avoids the chemical degradation of the thin metallic coating at the electrospraying tip (see page 28 of Perez-Lorenzo Reference Perez-Lorenzo2022) since electrochemical reactions take place instead in the liquid reservoir. The effect of these reactions on the purity of the IL sample is minimal because the reservoir volume (approximately 1 $\textrm {cm}^{3}$ ) is much larger than the liquid volume sprayed during a typical experiment. As a test of the reliability of the electrostatic voltmeter, we verified that its reading coincided with the voltage applied to the platinum electrode when the spray current was cancelled by stopping the liquid flow. The iridium-coated emitter tip is in mechanical contact with a stainless steel sheath tube, which is insulated from the chamber through a PEEK sheath on the vacuum side, and an alumina sheath $\sim \kern-2pt 100$ mm long on the atmospheric side. The whole side of the vacuum chamber, including the 10’ vacuum flange and extractor where the emitter is located, and the silica sheathed section of the capillary, can be heated or cooled in the range $0{-}140^{\,\circ} \text{C}$ . The chamber flange and the extractor electrode are in thermal contact, limiting radiative heat loss of the capillary’s cantilevered section to just the extractor orifice region, which is small with respect to the assembly size. With regard to thermal conductivity, the heated/cooled section of the capillary is composed of aluminium, alumina, stainless steel and silica of reasonable thermal conductivity. For the liquid flowing through the silica capillary, the thermal relaxation time $\tau _{\textit{TH}} \sim C_v r^2/\lambda$ , where $C_v$ is the volumetric heat capacity, $r$ is the pipe radius and lambda is the thermal conductivity) is much shorter than that of the residence time $\tau _{\textit{RES}} = \textit{LQ}/A$ , where $L$ is the pipe length, $Q$ is the liquid flow rate and $A$ is the pipe cross-section. For a representative case of EMIBF $_4$ , these values differ from approximately 1 ms to 100 s, suggesting an excellent thermal homogenisation prior to emission.

Figure 1. Detail of the electrospray tip with (a) and without (b) IL. The distance between the small marks is 10 $\,\unicode{x03BC} \text{m}$ .

Figure 2. Detail of the experimental set-up. Here, $\textrm{RPA}_V$ is stopping potential or $V_G$ , $\textrm{RES}_V$ reservoir voltage, $\textrm{RES}_P$ reservoir pressure, $\textrm{ESVM}$ electrostatic voltmeter, $\textrm{EXT}$ extractor (current meter), $\textrm{ES}_I$ electrospray (current meter), $\textrm{OX}$ oscilloscope (8 channel), $\textrm{OPA}$ operational amplifier (current feedback, inverting configuration), $\textrm{ETRAP}_V$ electron trap collector bias voltage, $\alpha _{\textit{min}, \textit{max}}$ open angle corresponding to a collector, $ \textrm{SWITCH} $ electronic MOSFET alternating between $\textrm{RPA}_V$ and $ \textrm{GND} $ ground. During regular operation the electrospray voltage is constant. On a data capture event, the gate’s central electrode is initially at $\textrm{RPA}_V$ and then the switch rapidly connects it to $ \textrm{GND} $ , the electrometer records current over time synchronised with the switching operation.

The experimental set-up determining the particle velocity, energy and angular distribution has been thoroughly described (Perez-Lorenzo & Fernandez de la Mora Reference Perez-Lorenzo and Fernandez de la Mora2022). It is also schematically shown in figure 2. In essence, it consists of a planar gate electrode arrangement placed perpendicularly to the spray axis. The gate is rapidly pulsed between ground and progressively higher absolute values $V_G$ of the negative voltage, allowing only the passage of negatively charged particles with energies above $V_G$ . In this planar gate configuration, particles moving off-axis experience some deflection during the gate transition. Their axial velocity is first slowed and then accelerated by the gate. But the radial component remains unperturbed, resulting in a deflection away from the axis. Although the deflection is drastic for particle energies very close to or below the gate potential, it is relatively modest even at energies only 1 % above the gate potential (see figure 4.7 of Perez-Lorenzo Reference Perez-Lorenzo2022). The narrow energy distributions found for ions confirm that these deflections are inconsequential. Additionally, this gate configuration leads to other minor issues, including transparency loss due to material accumulation within the grids, material backscattering into the emitter and energy alterations for the material trapped between the grids during switching. Note that for terminological convenience the term `energy’ here means energy per unit charge, and is used as synonymous with `retarding voltage’. The charge collector consists of six planar electrically insulated concentric rings, also perpendicular to the spray axis. Each ring is connected to an independent electrometer, which records the current versus the time t after the gate opens or closes. The flight times $t$ and the stopping potentials $V_G$ associated with the charged particles received by each collector determine the particle velocity and energy as $u = L/(t \cos\alpha )$ and $\xi = V_G/\cos^2\alpha$ , where $\alpha$ is the polar angle between the axis and the mid-radius of each ring, with the capillary tip used as the origin of coordinates. Here, $L$ is the axial distance from the collector to the gate. The electrometer electrodes are biased to + 10 V with respect to a grounded grid placed immediately in front of them to reject backscattered electrons back into the collectors. The outermost collector ring is symmetrically divided into three $120\,^\circ$ sections to diagnose any mismatch between the plume and the geometric axes. With this set-up, approximately $50\,\%$ of the emitted current is directly collected by the electrometer, and almost all of it is accounted for when correcting for the transparencies of the various components (masks and meshes). The emitted flow rate is calculated based on the known pressure drop applied between reservoir and vacuum, corrected for the temperature-dependent viscosity along the various sections of capillary, in addition to the effects of reservoir height and tip geometry. Interestingly, the measured voltage drop (reservoir–tip) with electrospray current and material conductivity results in an effective capillary ID, of a smaller value of $40-44 \,\unicode{x03BC} \text{m}$ rather than 50 (and 35 instead of 40 for $\text{C}(\text{CN})_3$ ). Regardless of this disparity, the emitted flow rate has been calculated based on the manufacturer ID. Prior to acquiring time of flight (TOF) spectra for the various liquids studied, the lowest stable current versus backpressure $I(P)$ are determined by pre-screening current/voltage conditions producing emissions with stable currents and visually steady menisci. These two indicators of stability are continuously monitored during acquisition $\sim 20 {\rm \,min}$ . The $I(P)$ traces, obtained from high pressure down to 0 and back up can be observed to overlap reasonably, with a small hysteresis at the lowest pressures (likely due to unsteadiness of the spray after flow starvation).

4. Relation between the QI and PI regimes

Details of the transition between QI and PI regimes are illuminated by a remarkable recent study of several ILs electrosprayed in vacuo in positive mode at room temperature (Caballero-Perez et al. Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025). By finely varying and accurately measuring the flow rate, they cleanly observe the QI $\rightarrow$ PI transition for BMI-C(CN) $_3$ . Unexpectedly, they find that the smallest flow rate at which a steady emission may be stabilised decreases substantially as the capillary tip shrinks. By using decreasing tip diameters down to $15\,\unicode{x03BC} \text{m}$ Caballero-Perez et al. (Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025) are able to stabilise their emitters at mass flow rates as small as $2 {\times}10^{-12} \,\textrm{kg}\,\textrm{s}^{- 1}$ , which results in a clear jump in the drop current from a value of approximately $2.5\,\%$ of the total current down to zero at a critical mass flow rate of $4 {\times}10^{-12} \,\textrm{kg}\,\textrm{s}^{- 1}$ . No unambiguous jump is visible in the total current I at the transition point, either because there is none, or because it is so small as to be buried in the noise associated with the measurement of the tiny flow rates involved. Nevertheless, there is little ambiguity in the fact that a transition arises at a critical flow rate, because the drop current fraction is almost constant at approximately $2.5\,\%$ for a fair range of flows above the critical value, and falls to zero below the critical value. This jump is also clearly manifested in a discontinuity in the thrust efficiency, which is highly sensitive to small drop flow rates. In a previously reported PI transition (Romero-Sanz et al. Reference Romero-Sanz, Bocanegra, Fernandez de la Mora and Gamero-Castaño2003; Romero-Sanz & Fernandez de la Mora Reference Romero-Sanz and Fernandez de la Mora2004), no efficiency or relative drop current data versus flow rate were included, so the existence of a sharply defined critical flow rate for a PI transition cannot be assessed. A PI transition for EMI-Im has been previously reported in a capillary with a tip diameter of $20 \,\unicode{x03BC} \text{m}$ (Caballero-Perez & Gamero-Castaño Reference Caballero-Perez and Gamero-Castaño2024). However, Caballero-Perez et al.(Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025) could not reproduce this feat in their 2025 study. Other occasional failed attempts by others to achieve PI emissions with EMI-BF $_4$ suggest similarly that the QI mode and its PI transition are highly unusual creatures, depending on certain peculiarities that are difficult to reproduce experimentally. These peculiarities are evident also in the known effectiveness of externally wetted versus capillary sources to achieve PI operation. They also manifest themselves in porous variants of ILIS, which sometimes operate in QI rather than PI mode (Villegas-Prados et al. Reference Villegas-Prados, Cruz, Wijnen, Fajardo and Navarro-Cavallé2024). One of the unusual features just noted is that the total current is barely changed at the transition. This is particularly significant because the structure of cone jets, when they form, depends greatly on this current. If there is a drastic transition from a PI situation without a jet to another with a jet, one would expect completely different currents. The proximity of the currents in both regimes leads to the suspicion that the corresponding meniscus geometries are close to each other. Because all available calculations of PI menisci (Higuera Reference Higuera2008; Coffman Reference Coffman2016; Gallud & Lozano Reference Gallud and Lozano2022; Magnani & Castaño Reference Magnani and Gamero-Castaño2023) include no jet, the hypothesis that there is really no QI jet offers itself naturally. Another notorious peculiarity is the small value of the QI drop current $I_d$ , which is an almost direct measure of the jet current. In our own measurements $I_d\sim 0.5\,\%$ of approximately 1000 nA, namely $\sim 5$ nA. In Caballero-Perez et al. (Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025), the total current of BMI-C(CN) $_3$ at the transition point is slightly less than 200 nA, with a drop current estimated also at approximately 5 nA. There is no known precedent for stable cone jets of highly conducting liquids supported on capillaries and carrying such small currents, suggesting again the absence of a QI jet. Even if one did not trust calculations indicating that there is no jet in the PI regime, it would be difficult to imagine a stable jet-like protrusion ejecting so little current from a highly charged meniscus. One is almost forced to conclude that, whatever QI jet is present, it must be very short, and must carry no mass flow most of the time. In this scenario, the QI jet would be so short as to be the mere tip of the meniscus. And whatever part of the incoming liquid flow that is not steadily emitted as ions would accumulate quasi-steadily in the meniscus, and be periodically ejected in a burst of nanodrops. As the total flow rate is decreased, a point would be reached when there is no excess flow, and the PI regime would set in. The issue of stability is apparently key in this hypothetical scenario, because the quasi-steady meniscus would need to be a stable ion emitter through the whole period of being filled and bursting. It makes sense that its stability would depend sensitively on geometric details such as capillary shape and dimensions.

Additional evidence for the hypothesis that drops are emitted in brief bursts is given by the anomaly already noted that the drop energy falls between the jet voltage and half its value. Typical voltage losses for the formation of a steady jet are a modest fraction of the emitter voltage, not $50\,\%$ of it. This high level of dissipation is difficult to rationalise in a steady situation. However, if a transient jet is briefly formed, it is bound to consume more electrical energy than a steady jet. This is clear from photographs of Taylor cone jets operating in the unsteady mode, where substantial surface and kinetic energy remains in the meniscus immediately after the jet detaches, which cannot be recovered as electrical energy.

In conclusion, all available evidence supports the hypothesis that the QI and PI modes involve similar menisci with no substantial protruding jet, and similar steady or quasi-steady PI currents. Their key difference is that QI menisci are fed at flow rates higher than can be ejected as pure ions. The excess QI liquid accumulates in the meniscus through a relatively long part of the cycle, is briefly ejected at the end of the period in the form of nanodrops (probably through a transient jet) and the process is again repeated. In contrast, PI menisci can eject all the incoming flow rate in the form of ions, and therefore operate steadily. In this scenario, ILs capable of reaching the QI regime would similarly attain the PI mode if the flow rate could be reduced sufficiently. The question of the measurability of these hypothetical transients and their frequency is of considerable interest. An upper bound for the time between bursts (the period) could conceivably be defined by analogy to what is observed in conventional electrified cone jets operating slightly above the maximal stable flow rate. In that case the meniscus is seen to slowly grow into an acorn shape, and then suddenly burst. The volume of the bursting meniscus is of the order of twice that of a steady meniscus, $V_{m}$ . Therefore, the period would be ${V_{m}}/{Q}$ . For a characteristic length of 30 $\,\unicode{x03BC} \text{m}$ and a typical small flow rate of 10⁻¹² $\textrm {cm}^{3}$ s⁻¹ this time is several ms, too large to have an effect on the observed dynamics. Furthermore, such a meniscus evolution would be visible, but is not observed. We should note here that our TOF measurements can detect current changes at frequencies of 1 MHz, such that any transient phenomenon occurring at frequencies below 10 MHz would be noticeable in our TOF curves. In a forthcoming study, we report on current oscillations observed in TOF spectra at frequencies of the order of 0.1 MHz. These correspond to the natural vibration frequency of a drop some 10 $\,\unicode{x03BC} \text{m}$ in diameter, although they are not visually observable at the tip of our meniscus. They arise first at the onset of ion evaporation from the meniscus neck, under conditions when a large fraction of the emission current is still in the form of drops. The frequency of transient jet formation in the QI regime must be much greater than either the estimated 1 kHz or the observed 100 kHz values, and is unquantifiable in our facility. We note finally that there is no guarantee that any IL held on a capillary source will be able to reach the QI mode. Figure 7 of Caballero-Perez et al. (Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025) provides three examples of ILs not doing so, even from $15\,\unicode{x03BC} \text{m}$ emitter tips. The 4 ILs here discussed that do successfully operate in the QI mode are a special subgroup of the 8 ILs we have studied, four of which produce a substantial drop current under all conditions investigated.

5. Relation between capillary tip diameter and minimal flow rate

The discovery that smaller tip diameters $d_n$ stabilise reduced minimal flow rates $Q_{\textit{min}}$ of ILs (Caballero-Perez et al. Reference Caballero-Perez, Galobardes-Esteban and Gamero-Castaño2025) deserves more discussion, as it greatly facilitates achieving PI emissions from capillary emitters. A rather strong dependence of $Q_{\textit{min}}$ on $d_n$ has previously been established for electrified glycerol jets, and attributed to its high viscosity (Scheideler & Chen Reference Scheideler and Chen2014). The conventional explanation for why Taylor cone jets are most frequently governed by laws unaffected by the emitter diameter $d_n$ is the large disparity between the nozzle and the jet diameters $d_j: d_n/d_j\gt \gt 1$ (Fernandez de la Mora & Loscertales Reference Fernandez de la Mora and Loscertales1994). This large scale ratio remains true for the glycerol jets of Scheideler & Chen (Reference Scheideler and Chen2014), and also under conditions typical of IL cone jets. However, for sufficiently viscous fluids, the viscous length $L_\mu =(\mu Q/\gamma )^{1/2}$ may take values intermediate between $d_n$ and $d_j$ . This may create conditions when $d_n$ is not very large compared with $L_\mu$ , and therefore $d_n$ ceases to be irrelevant. This possibility is illustrated in the image of a glycerol cone jet provided by Scheideler & Chen (Reference Scheideler and Chen2014), where the meniscus structure does not rapidly evolve from a thin jet into the usual conical meniscus expected from Taylor’s static equilibrium (Taylor Reference Taylor1964). Rather, there is an extended concave–convex meniscus evolving from the thin jet to the large capillary support, presumably extending over a characteristic length $L_\mu$ . This lack of a cone in electrified menisci of glycerol was already prominent in figure 7 of Zeleny (Reference Zeleny1917), as well as in figure 9 of Taylor & Van Dyke (Reference Taylor and Van Dyke1969), although without establishing a connection with $Q_{\textit{min}}$ . A reduction of $Q_{\textit{min}}$ with capillary diameter has also been observed by Ponce-Torres et al. (Reference Ponce-Torres, Rebollo-Muñoz, Herrada, Gañán-Calvo and Montanero2018) for liquids such as 1-octanol, 1-decanol and 1-butanol, all having substantially less extreme viscosities than glycerol. Their study includes extensive theoretical and numerical considerations.

The hypothesis that glycerol and the new IL observations are both related to viscous effects finds further qualitative support in the fact that ILs tend to be substantially more viscous than liquids typically used in electrospray applications.

6. Have we reached the purely ionic regime?

Figure 6. Comparison of two cases of EMI-SCN at fixed current and flow rate, although for two different flight lengths $L$ . The horizontal scale is the flight time ( $\tau$ ) in (a) and the ion velocity ( $u= \tau /L$ ) in (b). The insets show the detail on the small step associated with the putative QI droplets.

Table 2. Emitter temperature, voltage, current and calculated flow rate for the various experiments from Figure 3.

During the review process, the peculiar fact that the small particle peak observed in the QI spectra of figure 3 (b) always appears at the same flight time has led one referee to wonder if this minor step might not be an artefact. One plausible explanation of this coincidence would be the fact that the transient jet would tend to have a fixed size, with only its frequency changing as a function of flow rate. Nevertheless, to test the referee’s hypothesis, we have reviewed two existing TOF spectra taken at two different flight distances $L$ under very similar spraying conditions. As shown in figure 6(a), all major features in these spectra coincide when plotted versus $t/L$ . However, the two small drop steps appear at exactly the same time in both spectra, and therefore at different $t/L$ values (figure 6 b). It is hence quite probable that these drop steps might not be real, in which case we would have reached the PI regime. We have unfortunately no credible hypothesis to explain an artificial step of a few nA at approximately 40 $\unicode{x03BC}$ s, nor are we able to carry out new experiments to check if the PI regime had in fact been achieved. Fortunately, this ambiguity forces no revision of our conclusions relating to the QI regime.