Is electrospray emission really due to columbic forces?

Mass spectroscopy (MS) based on electrospray ionization (ESI) has been widely adopted over more than thirty years. When a fluid is exposed to a strong enough non uniform electric field, the induced charges force its surface to form a liquid jet.^1,2 The jet is initially highly collimated but, after a short fly-path, it disintegrates over a wide cone of tiny droplets. In the typical experimental set-up the liquid is flowing through a polarized needle, the injector, while a second electrode, the extractor, is a large plate located few centimeters away.

It has been argued³ that dielectrophoretic forces acting on liquid droplets cannot explain jet emission. In fact, induced dipoles should be attracted towards region of higher field, i.e. towards the needle tip. On the contrary, the movement of droplets always occurs from the needle towards the large extractor plate. This observation has been enough to deduce that the flying liquid drops must possess a net charge of the same sign as the needle electrode.^4,5 Partial corona discharge⁶ at the needle could cause necessary ionization. Jet disintegration has been usually explained in terms of columbic explosion occurring when, due to solvent evaporation, the net charge existing in each droplet exceeds the Rayleigh limit.^7–9 This is the current explanation of the electrospray mechanism adopted in the literature.¹⁰ 

On this basis, electrospray appeared a suitable ionization source for MS.^9–13 In 1988, Fenn reported ESI of poly(ethylene glycols)^14,15 for which he was awarded the Nobel Prize for Chemistry in 2002.

In spite of the undeniable success of ESI-MS, we have found that the current explanation for the electrospray-jet mechanism can be misleading.

In this paper we preliminarily show that the existence of net charges is not required for jet ignition nor for the formation of a plume.

Hence, we report the results from a first experiment, which has to be intended just as a plausibility test aimed to verify if the observed droplet trajectories are compatible with the effect of bare dielectrophoretic forces, in absence of any net charge.

Finally, we present a very simple experiment, performed under different field geometry. In such a configuration, the expected outcomes are dramatically different for charged and uncharged droplets. We will show that the obtained results are definitely unambiguous and model free, allowing to safely conclude that, at least at atmospheric pressure, the observed motion of liquid drops is not driven by columbic forces. We will suggest that, at the ejection time, the drops cannot be so charged as until now proposed and that additional mechanisms could play their role on the ionization process occurring in any ESI-MS experiment.

The currently accepted mechanism for electrospray assumes generation of gas-phase ions which are guided to the counter electrode by the electric field. Any measured electric current is ascribed to the flow of ions between the electrodes, in close analogy to what is observed in an electrolytic cell.¹⁶ However, the idea that the presence of net charges is a mandatory condition for electrospray generation³ comes from an incomplete description of the experimental conditions. The conclusion that any induced dipole must be driven towards regions of higher electric field is correct only under the condition of constant charges at the electrodes. Differently, any practical realization of electrospray is an experiment performed at constant potential conditions. Accounting for the experimental conditions, we follow the approach adopted in many textbooks.¹⁷ The process is described as the sequence of two distinct steps. Two electrodes are located in a volume, V, filled with a medium characterized by a dielectric constant ɛ₁. In the first step, a voltage source is connected to the electrodes in order to produce a charge distribution ρ(r) and the corresponding potential ϕ₁(r), then the voltage source is disconnected. When a volume V_d of a material with dielectric constant ɛ₂ > ɛ₁ is introduced in the volume V, a change in the potential, dϕ₁, is produced, corresponding to the energy change

During this step, the charges on each electrode remain unaltered, because the power source is disconnected. For any generalized displacement ξ, the force, (F_ξ)_Q, acting on the center of mass of the dielectric body is given by |$ - \left( {{{\partial W} \mathord{\left/ {\vphantom {{\partial W} {\partial \xi }}} \right. \kern-\nulldelimiterspace} {\partial \xi }}} \right)_Q$|$−∂W/∂ξQ$which, for ɛ₂ > ɛ₁, implies that the force is directed towards higher field regions (i.e. towards the needle tip in an electrospray experiment). Because of the experimental observation that the droplets move towards the extractor electrode, the conclusion has been drawn that the droplets must be charged with the same sign as the needle.^3–5 However, in ES experiment, the voltage supply maintains constant the electrode potentials (not their charges). The correct description has to include a second step in which the voltage source is connected again to the electrodes. Then, a new charge distribution, dρ, is induced, resulting in a change of the potential dϕ₂ = − dϕ₁. The energy change produced during the second step is

The total energy change produced by the two steps is

which has the same amplitude than eq. (1) but opposite sign. For any generalized displacement, |$\xi,(F_\xi)_\phi = + \left( {{{\partial W} \mathord{\left/ {\vphantom {{\partial W} {\partial \xi }}} \right. \kern-\nulldelimiterspace} {\partial \xi }}} \right)_\phi = - (F_\xi)_Q$|$ξ,(Fξ)ϕ=+∂W/∂ξϕ=−(Fξ)Q$⁠. Due to the work made by the voltage source, the polarized dielectric body is pulled towards regions of lower electric field, i.e. towards the larger electrode.

This preliminary discussion clearly demonstrates that there is no a priori need for invoking the presence of net charges carried by the drops. At the same time it is quite clear that the above arguments cannot exclude their existence. Both dipoles and charges will exhibit similar motions in an electrospray experiment. Currently, it is assumed that the detection of electric current at the electrodes indicates a charge transfer between electrodes trough the jet. However, the observation of a very small current is not a safe indication of the neat charge carried by each drop. In absence of any electric charge, acceleration of a dielectric drop from one electrode towards the other requires a work. According to Eq. (3), this work is performed by the power supply. In order to maintain the electrodes at constant potentials, while the dielectric drops are moving, a redistribution of the charges on the electrodes is required. This implies a current through the power supply. This current can be measured but its detection does not implies a charge transfer between the needle and the droplets and hence a current through the jet. From the experimental point of view, substituting the metallic needle with a fused quartz capillary is a simple way for testing the above conclusions. When polarizing the dielectric needle, jet ignition is observed with no respect of the used fluid. In such a case no exchange of free electric charges between the capillary and the fluid can be hypothesized. In this perspective, electrospray shows strong analogy with another phenomenon, originally addressed as water bridge:^18,19 an intense electric field can coax water into leaping a tenths of millimeters gap between two glass beakers, forming a floating bridge. This phenomenon, first ascribed to some unknown properties of water, finally revealed to be originated just by dielectric forces.^20,21 In particular, the phenomenon is observed also in pure dielectric liquids where the deformation of the free liquid surface under an intense electric field is originated by the interaction of the electric field with the existing (both permanent and induced) dipoles while no free charge carriers can be hypothesized.

An infusion system PHD 2000 (HARVARD) has been used to control the liquid flux through a stainless steel (outer diameter 200 μm, inner diameter 100 μm). The extractor, a squared (5 × 5 cm2) stainless steel electrode or a second needle (depending on the selected field geometry, see below for details) has been located on a three axis goniometer to obtain the correct alignment. The goniometer has been located on a motorized linear stage (accuracy 1 μm) allowing for the selection of the injector extractor distance. A TREK 664 high voltage power supply has been used for electrode biasing. During each measurement, the current between the grounded electrode and the ground has been monitored by a Source Measure Unit (Keithley 236). In every experiment here reported the load current has been ever lower than 500 pA. Drop motion has been detected by a fast camera (Photron Fastcam SA4). A Rodenstok Componar 50 mm f/2.8 lens, mounted on an extension tube (from 10 to 40 cm long, depending on the required magnification ratio), has been used as the collecting optics. The camera has been located on a motorized linear stage, parallel to the linear stage supporting the extractor, allowing for changing the field of view along the field axis maintaining the same focus of the image. The whole experimental setup has been remotely controlled by a computer. For image analysis, the software provided by the camera maker has been used. In order to eliminate possible environment induced effects, each experiment has been performed within a box filled with nitrogen at atmospheric pressure. Some experiments have been performed after the substitution of the plate extractor with a needle shaped electrode. The diagram of the experimental geometry is reported in Fig. 1, when a plate extractor has been used. The change when the extractor is a second needle is straightforward.

Numerical calculations of the electric field under different experimental geometries has been performed by finite elements analysis using the software Maxwell (Ansoft, version 13.0.0). The software allows for the proper design of the electrodes and for assigning the material properties and excitations (electrode biasing). When biasing is defined, it is possible to design a body, to assign its electric properties and eventually to calculate the force acting on the center of mass of the body when it is located in between the electrodes. Maxwell disregards any work necessary for maintaining a constant potential, so it gives the correct force amplitude but with the wrong direction (in agreement with our Eqs. (1)–(3)). Taking into account this detail, spherical bodies with radius coincident with the experimentally measured radius of the liquid drops have been located in different positions in between the electrodes. This allowed the estimate of the acceleration field in the whole volume of interest and hence the drop trajectory has been numerically calculated. To simplify the calculation, the numerical evaluation of drop velocity as a function of the distance from the needle tip, reported in Fig. 2(c), has been obtained after the following assumptions: at time t = 0 the drop center of mass is at rest at a distance of 50 μm from the needle tip and on the symmetry axis of the field (the initial velocity of the fluid through the needle is at least two orders of magnitude lower than the velocity produced by the acceleration due to the electric field after a path shorter than 1 mm). Gravity effects are disregarded. However, this approximation does not imply large errors because the initial acceleration is two orders of magnitude larger than the gravity acceleration. The relative error produced by these approximations is within 0.02, lower than our experimental uncertainty (see error bars in Fig. 2(c)).

According to this, the velocity plot reported in Fig. 2(c) refers to a trajectory which is a straight line along the symmetry axis of the electric field.

The first experiment is performed under the usual geometry of any electrospray experiment (see Fig. 1). In this experiment the liquid flows through a metallic needle (the injector). A second electrode (the extractor) is placed oppositely than the injector at a distance of few centimeters. When this configuration is adopted, different working regimes can be established (from pulsed emission of discrete drops to the formation of stable liquid jets) by the appropriate tuning of the experimental parameters (injector-extractor distance, flow rate of the liquid trough the needle, applied voltages at the electrodes) and depending on the adopted fluid. Aim of this experiment has been to compare the experimental drop trajectory with the results of calculations based on the assumption of charge-free drops. The trajectory of a drop can be directly observed by a fast camera (PhotronFastcam SA4). The request for observing the trajectory of a single drop imposes some experimental conditions: 1) the average size of each drop must be larger than the experimental spatial resolution; 2) the emitted drops must be almost mono-dispersed in size. In principle, both the experiment and the calculations can be performed with different liquids. However, taking into account that the role of this experiment is merely a plausibility test, only one sample has been investigated into details. The selected sample was bi-distilled and deionized water. The injector was a stainless steel needle with a channel of 100 μm (outer diameter 200 μm). The extractor was a stainless steel plate (5 × 5 cm²). The required condition for a pulsed emission of drops of almost uniform size has been obtained at the following experimental conditions: injector-extractor distance 2 cm; flow rate 20 μl/min; injector potential −6.5 kV; extractor at ground. The images of the flying drops have been captured at a frame rate of 8000 frames/s, over a field of view allowing the observation of a trajectory up to a distance of 5 mm from the needle tip. The spatial resolution has been 8.7 μm/px.

In Fig. 2 the experimental results are reported and compared with the outcomes from finite elements calculation. In the panel a of Fig. 2, the calculated electric field is reported over a region extended up to 5 mm from the injector (matching the experimental field of view). In the calculation, the size of the electrodes, their distance and their biasing reproduce the experimental conditions. In Fig. 2(b) (Multimedia view), four consecutive frames of a drop leaving the injector are reported. The images in Fig. 2(b) are cuts of the 6 mm wide original frames. From the images, we estimated an average droplet diameter of 70 μm. They were produced at a rate of 1800 drops/s, which gives a flow rate of 19.4 μl/min. This value is very close to the value set up at the needle (20 μl/min) ensuring us about the accuracy of our estimates. To calculate the field of force acting on the flying droplet, a probe sphere (70 μm diameter, ɛ = 80) was located at different positions (see panel a in Fig. 2). Hence, the drop trajectory has been calculated and compared with the experimentally observed one (panel c in Fig. 2). The agreement between calculations and experiment is satisfactory. The apparent over-estimate of the velocity at large distance from the needle can be easily rationalized taking into account that simulated droplets are rigid spheres where any energy loss due to viscosity and internal degrees of freedom (rotations and surface waves) are disregarded. This drawback is clearly evident from Fig. 2(d), where we report a higher resolution sequence (4μ/px, 20000/frames/s) of the emission of a water droplet under the same biasing condition of Fig. 2(b), but at a flow rate of 2 μl/min. The last frame in the sequence shows that, at least in the first part of its fly, the drop experiences large vibrations (this effect can be better seen in the on-line movies). In addition, the sequence shows that some jets are departing from the liquid surface. This is partially visible in the first frame of the sequence in the Fig. 2(b), but it becomes clearly evident after changing the flow rate through the needle (see Fig. 2(e), which is an enlarged view from the sequence reported in the panel d of Fig. (2)). It has already reported that the experimental conditions affects the formation of secondary jets, with production of satellite droplets affecting the drop polidisperisity.²² It was also shown that a proper selection of the experimental parameters can allow for avoiding (or, at least, minimizing) secondary jets in order to produce almost monodispersed drops.²² The presence of tiny drops within the jet (whose trajectory is undetectable) could have some effects that cannot be evaluated in the proposed calculation.

We wish to point out that the results of this experiment and the comparison with the reported numerical results are reported only to demonstrate that the observed trajectories are qualitatively in agreement with our hypothesis for the absence of net charges. However there is no way to obtain unambiguous answers from this kind of experiments. It is quite evident how a number of very accurate calculations,^22–24 based on the assumption for charged drops, can reproduce the experimental observations into details. Both dipoles and charges are accelerated in the same direction in the non-uniform electric field obtained with the usual experimental geometry. If the ionization efficiency is low (as it is currently assumed) the differences between the two situations could be too small to be clearly detected. In addition, concurrence of the two effects can be plausible. In summary, to distinguish between the two situations through a numerical calculation would require the a priori knowledge of the charge carried by each drop.

In principle, one can be led to assume that the accurate measurement of the current between the extractor and ground should answer the question if drops are charged or not. However, the arguments discussed in sec. II suggest that detecting a very low current cannot be taken as the safe indication for a flow of charges between the electrodes. In our experiments, the monitored current never exceeded 500 pA. With reference to our Fig. 2(c), it can be calculated that, in order to accelerate an uncharged water drop (diameter 70 μm) from the needle tip to the extractor plate a work of about 1.1 nJ is required. To extract 1800 drops in one second requires a power of about 1.9 μW which, at the applied difference of potential of 6.5 kV, corresponds to a current of about 300 pA. If we would interpret this current as the current flowing through the needle channel (100 μm diameter), we should deduce a current density of about 4 μA/cm². In the above described situation, such an interpretation would be erroneous: the detected current is merely indicating that a work is required for maintaining the constant difference of potential between the electrodes as the droplets are flowing.

It is known that the measurement of the current in a spray-jet experiment is an extremely delicate task and several difficulties, generated by the existence of different competing contributions, have been already pointed out.²⁵ However, at our knowledge, accounting for the contribution due to the work required to maintain the potentials of the two electrodes constant has never been addressed.

In view of the above discussion, we concluded that, in order to remove any ambiguity, a different experiment must be performed.

In these experiments, the extractor plate was substituted by a needle shaped extractor. When both the injector and the extractor are needles the field strength has two maxima at the needle tips. In Fig. 3(a), we report the result of finite elements calculation in the new configuration. The injector is on the left while the extractor is on the right. The distance between the needle tips is 2 cm and the electrodes are biased as in Fig. 2(a). The observation of the drop motion under such a field configuration immediately allows for discriminating between charged and uncharged drops. A charged drop (carrying a charge of the same sign of the injector) will be repelled by the injector. At the same time, it will be attracted by the extractor (charged with the opposite sign). Due to the action of columbic forces, a charged drop should accelerate following the direction of the force field. Its trajectory will be a curve that moves from the injector tip to the extractor. An uncharged drop will follow a different trajectory. When leaving the injector, it will be repelled similarly to a charged drop (see the arrow close to the injector tip, on the left in Fig. 3(a), which represents the direction of the force in a position close to the electrode). But, after reaching the half way between the electrodes, it will enter a region of increasing field strength and, according to Eqs. (1)–(3), it will be repelled from the extractor tip (the arrow close to the extractor, on the right in Fig. 3(a), represents the direction of the force in that position).

The interpretation of the results does not require any calculation nor any underlying assumption, at least until the motion of well separated and optically resolved particles is analyzed. In performing these measurements, care was taken in adopting experimental conditions (flow rate and field strength) suitable for avoiding any emission of secondary jet from drop surface.²² This allows to rule out any possible shielding effect due to the accumulation of ionized and nebulized matter around the extractor.³ Under such conditions, we have just to observe if the emitted drops hit the extractor needle or, on the contrary, they are repelled and back-reflected. The experiment was performed on several, both apolar and polar, liquids including ionic solutions. Care was taken in exploring several experimental conditions, repeating the observation alternatively with the injector at ground and the extractor biased and under reverse polarization. The experiment was also repeated under different voltages, different field strengths and at different fluxes of the liquid through the needle. In any case the same observation was made: no liquid drop reaches the extractor.

Some few examples are summarized in Fig. 3 (see the figure caption for details about the different experimental parameters). In panels b (Multimedia view) and c (Multimedia view), two sequences of images are reported showing the motion of water droplets coming from the injector (on the left). The horizontal size of each frame is 1.5 mm. In each panel the time sequence is directed downwards and the frames of each sequence are delayed of a fixed time, Δt (the Δt values corresponding to each sequence are indicated in Fig. 3). The yellow points in the first frame of each sequence mark the successive positions of a drop in the sequence, while the red line is the observed portion of the trajectory. In panel b, the field of view is centered at midpoint between injector and extractor. It is quite evident that the droplets are deviated upwards and then back reflected following a vortex line. In panel c, the field of view is centered at the extractor needle. No drop is observed hitting the needle and only few of them are able to enter the field of view. The one marked in the reported sequence clearly decelerates in approaching the extractor needle and eventually reverts its motion, flying back towards the injector.

It could be possible that only a very small charge is carried by large drops, so that electrophoretic effects can overcome the effect of the columbic attraction at drop-needle distances comparable with the drop size.^26,27 To exclude such a possibility the observation must be repeated under stable jet conditions, when the droplet size decreases down to nanoscale at the plume formation. Panel d (Multimedia view) refers to a stable jet of ethanol, as an example. Also under these conditions no droplets from the plume are reaching the extractor needle. The initially collimated jet spreads out in a cloud of tiny droplets. But it can be observed that the cloud of nebulized liquid reverses its motion being repelled from the extractor needle. Panel e (Multimedia view) shows a drop of an aqueous solution of NaCl (5% w/w): also in this case the drop decelerates and abruptly deviates downwards when approaching the extractor needle. Even if the sample consists of a ionic solution, we cannot observe any sign of columbic forces acting on a possible net charge (see supplementary materials²⁸ for a further example).

While the conclusion that columbic forces are not playing the major role can be safe in the case of micrometric and well separated drops, some caution must be taken when plumes from stable jets are considered. Interpretation of the experiments performed on stable jets should take into account the possibility of shielding effects due to the accumulation of ions around the extractor needle. Such an effect was originally proposed by Pohl³ to explain the observed repulsion of the liquid surface placed below a polarized needle electrode. Even if the concept of “ionic cloud” was introduced as the result of the erroneous evaluation of the problem, we cannot exclude, in principle, that no ionized and nebulized matter is accumulated around the extractor needle. However, if this should happen, the ionized liquid should start wetting the needle surface. Instead, after several minutes of continuous operation, the outer surface of the needle remains clean and dry. In our opinion, this makes the hypothesis for the formation of ion clouds around the extractor, at least, improbable. In addition, the existence of ionized gas in the proximity of the extractor needle (due to the presence of ionic species or to a partial corona discharge) should be detected by a higher current when the two needle experiment is compared against experiments performed under the conventional geometry. This is not the case: the measured current is lower than that observed under the usual configuration and easily becomes negligible. The drops repelled by both the electrodes eventually fall down due to gravity. Gravity produces work in extracting the drop from the volume in between the electrodes. So a steady state condition is reached, where no further redistribution of the charges on the electrode surfaces is required.

Summarizing our results, we have detected that, contrary to the current belief, no net charge is necessary to explain the observed motion of the drops in a usual electrospray experiment. Such belief originates from an erroneous evaluation of all the energy contributions.

We have shown that any measurement of a small current cannot be taken as the safe indication for the charge carried by the droplets crossing the gap between the electrodes, at least in absence of any accurate evaluation of the work required for maintaining the difference of potential constant. The results from experiments performed under symmetric non uniform electric fields shows that micrometric drops coming from the extractor are uncharged or, at least, are so weakly charged to make columbic forces negligible. Even under spray-jet conditions, significant charging of nebulized liquid from the vertex of the Taylor cone seems improbable (at least, at atmospheric pressure).

This imply that the current description of the forces acting during electrospray must be reconsidered. This can imply that ES is a ionization method even less efficient than it was until now supposed.

The results from this simple experiment have implications in a wide range of fields, including mass spectrometry, aerosol research, ion thruster research, and pesticide deposition. As an example, if the initially formed drops are not charged (or only very weakly charged), the accepted mechanism of droplet disintegration^7–10 (cycles of charged droplet evaporation followed by Coulomb fission) cannot be assumed. A possible alternative explanation of the phenomenon can be obtained through the theoretical nonlinear analysis²⁹ of deformation and breakup of dielectric drops in the presence of a quadrupole electric field (compare, as an example, panel e in our Fig. 2 with Fig. 8 in ref. 29).

A further possible explanation for the formation of an expanding plume follows from our Fig. 4. There we report the results from finite element calculations when a “cloud” of tiny water drops is located close to the needle tip. Calculations were performed under the same electrode geometry and biasing adopted for Fig. 2(a). Color changes, from red to blue, map decreasing values of the field strength. It can be seen that the field intensity is higher in between the drops, while it decreases outside the cloud of particles. Following Eqs. (1)–(3), it becomes quite clear that dielectrophoretic forces act expanding the cloud, with no need for invoking columbic explosion.

As a final comment, the validity of ESI-MS technique is beyond any doubt (the occurrence that Paul traps really capture ions cannot be undetermined) but we claim that the widely accepted explanation for ESI should be deeply revised. In view of our results, the generally accept idea that electrospray emission is driven by columbic forces must be ruled out. At atmospheric pressure, any possible ionization mechanism is so poorly efficient to make any effect of columbic forces negligible. It could be that different mechanisms start to play a role when the gas dispersion of nanosized drops enters in low pressure volumes while approaching the counter electrode, as in any ESI-MS experiment.

Work supported by the CNR-PAN bilateral agreement for the years 2014–2016. Project title: Experimental and theoretical aspects of electrospray and its application for thin film deposition.