nozzle and liquid effects on the spray modes in nanoelectrospray

Journal of Colloid and Interface Science 305 (2007) 111–123www.elsevier.com/locate/jcis

Nozzle and liquid effects on the spray modes in nanoelectrospray

Mark D. Paine ∗, Matthew S. Alexander, John P.W. Stark

Department of Engineering, Queen Mary, University of London, Mile End Road, London E1 4NS, UK

Received 16 June 2006; accepted 15 September 2006

Available online 20 September 2006

Abstract

Unforced nanoelectrospray can exhibit a number of stable spray modes. These include low frequency pulsations, high frequency pulsations, anda steady cone-jet. Experiments are reported here on such pulsations that have been observed in various salt loaded solutions of ethylene glycol,triethylene glycol and water. The spray current was monitored with 1 µs time resolution to show that spray regime characteristics depend on nozzlediameter and liquid conductivity. The frequency of pulsations was found to increase with both increased liquid conductivity and decreasing nozzlediameter. The charge ejected during a pulse is lower for smaller nozzles spraying higher conductivity liquids. Water solutions were observedundergoing high frequency pulsations, with these pulsations often occurring in lower frequency bursts. The frequencies of water pulsations wereas high as 635 kHz but the charge ejected by each pulsation was an order of magnitude lower than that observed in triethylene glycol. An unforcedelectrospray of water was also identified as being in the steady cone-jet mode with a higher degree of confidence than previously. The values forstable pulsation frequency and charge ejected observed in ethylene glycol lay between those of TEG and water.© 2006 Elsevier Inc. All rights reserved.

Keywords: Atomization; Nanoelectrospray; Electrospray; Pulsation; Nanospray; VMES

1. Introduction

Nanoelectrospray is the name given to a method of creat-ing aerosols of sub-micrometre sized droplets. It is a form ofthe electrospray ionisation technique, for which John Fenn [1]was awarded the Nobel Prize in 2002, whereby a strong electro-static field overcomes the surface tension of a liquid meniscusto form a liquid jet. Due to instabilities [2], the jet breaks upinto charged droplets which may contain molecules such aspharmaceutical compounds [3] or non-covalently bound bio-molecules [4]; these molecules may be introduced by liquidchromatography separation [5]. Recently nanoelectrospray hasbeen employed in a new technique which electrosprays solventat living tissue to produce gaseous ions for mass spectrometricanalysis [6]. Nanoelectrospray has found significant popular-ity as an analytical tool for ionising biomolecules due to itsease of use and the very small liquid droplets it produces. Thetechnique uses glass capillaries with micrometer-sized exits, thehydraulic resistance of which leads to flowrates in the nL/min

* Corresponding author. Fax: +44 208 983 1007.E-mail address: [email protected] (M.D. Paine).

0021-9797/$ – see front matter © 2006 Elsevier Inc. All rights reserved.doi:10.1016/j.jcis.2006.09.031

range [7]. Lower liquid flowrate results in the formation ofsmaller droplets [8,9] which evaporate more rapidly; the analytemolecules are more efficiently transferred into the mass spec-trometer [10,11]. Using this method, efficiency may be as highas one analyte molecule out of every eight leaving the needlebeing detected by the mass spectrometer; it is therefore a highsensitivity technique. Nanoelectrospray is also more tolerant tothe buffer salts found in the solution to be sprayed [12,13].

One of the system benefits of nanoelectrospray is that theadvantageous low flowrates are obtained without the need forpumps. Instead, when no back pressure is applied, the flowrateis dictated by the nozzle geometry and the applied electric field.However, this could also be viewed as a weakness, in that smallvariations in the nozzle diameter can affect the flowrate withoutthe operator being aware, resulting in inconsistent performance[10]. In previously published research flowrate has been esti-mated either by measuring the time taken to spray a volume ofliquid [7,14–17] or by weighing a nozzle containing the sam-ple before and after a fixed spray time [11]. Neither of thesemethods captures the changes in flowrate associated with a timevarying signal intensity. Other researchers have used video mi-croscopy to observe the movement of a liquid meniscus either

http://www.elsevier.com/locate/jcis

mailto:[email protected]

http://dx.doi.org/10.1016/j.jcis.2006.09.031

112 M.D. Paine et al. / Journal of Colloid and Interface Science 305 (2007) 111–123

in a pressurised liquid vial [18,19] or the spray needle body[10] to calculate the liquid flowrate in real time. Gas pressurewas used to alter the liquid flowrate, but the effect of voltage onflowrate was deemed to be small. In our own previous work, wehave taken a rather different approach. In this we have measuredthe flowrate through emitters where a gas pressure was eitherapplied [20] or absent [21]. Using an in-line flow measure-ment system with pL/s accuracy, we have found a significantdependence of flowrate on the applied voltage, particularly atlow flowrates. In these reports we suggested the name voltagemodulated electrospray to communicate the idea that nanoelec-trospray is indeed sensitive to voltage.

All previous research into the electrospray process has fo-cussed on situations where the flowrate is either set using gaspressure or is controlled, typically by a syringe pump. The mostwidely understood form of electrospray is the cone-jet modewherein the electric field and surface tension balance to form aconical-like surface. The first hydrostatic solution predicting aconical meniscus was first derived by Taylor [22] and extendedto include the effect of the jet and spray in the 1980s. The latterhas resulted in the derivation of scaling laws relating flowrate,Q, conductivity, K , and the spray current emitted, I , from con-ventional emitters in this cone-jet mode [8,23]. These scalinglaws can be represented by the general form:

(1)I = const KaQb.

The stable cone-jet regime occurs within a rather narrowrange of flowrates and voltages that depends on the liquid prop-erties and the electrostatic setup. Beyond this range, there area number of other spray modes which are generally consideredunstable; these modes include multijet, microdripping, ramifiedjet and the spindle mode [24,25].

Further investigation into the spray modes that precede theonset to the cone-jet mode, has observed stable pulsationswhere the liquid cone periodically forms a transient jet. Bymeasuring the spray current using a fast current amplifier thepulsations were found to occur in two types of mode [26]. In thefirst mode (axial I) high frequency pulsations were grouped intolow frequency bunches which often disappeared at higher volt-ages, leaving only the high frequency oscillations (named axialmode II). The axial II pulsations were found to increase in fre-quency as voltage was increased. At a certain value of voltagethe spray switched to the cone-jet mode. High-speed photogra-phy revealed a correlation between the current oscillations andthe transient formation of a liquid jet [27]. It has been pointedout that these modes could have an effect on the signal obtainedin a mass spectrometer [28–30] and that the spray mode canbe automatically detected by monitoring the frequency of anypulsations. Each of these findings was found when the electro-spray was in a forced flow situation—either using a pump, or bythe application of gas pressure to maintain the flow. Pulsationfrequencies as high as 125 kHz have previously been reportedduring forced flow electrospray [30].

Some work has been performed previously to characterisethe effect of the applied voltage on the spray current for differ-ent sizes of emitter during unforced electrospray [16]. Howeverthe current was not analysed for fluctuations, so differing spray

modes may have been inadvertently presented in that work. Ourown recent work [21] investigated the flowrate and spray modesduring unforced electrospray, where the flowrate is dictatedpurely by the applied voltage. We again found that a highlystable pulsation mode occurred which was similar to the ax-ial II mode. The frequency of pulsations was higher than thoseobserved in forced flow. Importantly it was observed that in-creasing the pulsation frequency by increasing the voltage alsoincreased the liquid flowrate. Only limited data was presented,including a single aqueous solution of NaI, emitted through a10 µm nozzle and a single triethylene glycol solution, againdoped with NaI, emitted from a 50 µm nozzle. These data weretherefore unsuited to understanding a broad characterization ofthe basic phenomenon, in particular the effect of nozzle diame-ter and conductivity on the pulsation properties.

The present paper therefore addresses the shortcomings ofour earlier description by spraying a range of liquids using arange of nozzle sizes.

2. Materials and methods

2.1. Electrospray configuration

In ESI-MS applications, nanoelectrospray is typically per-formed using so called “offline analysis” tips. In general thesetips are made from capillaries with inner diameters of 500 µmor more that reduce to a tip diameter of 1–4 µm. The sample isloaded using a fine pipette into the body of the needle.

The majority of the emitters used for the experiments re-ported here are similar to those used in ESI-MS; they are silicacapillaries, however with a 75 µm ID pulled to an exit diam-eter of either 8, 15 or 30 µm (New objective, MA). The outerdiameter of these at the emitter tip is approximately the sameas the internal diameter due to the taper used. The 75 µm boretips cannot be filled via pipettes. Instead, nitrogen was used topressure feed the liquid from a 100 µL plastic sample vial intothe tip. This was performed by connecting the spray capillar-ies to a feeding capillary of ∼50 cm length and 180 µm IDusing a stainless steel union (Valco). The union was of the zero-dead-volume type to minimise the possibility of deformable gasbubbles in the liquid connection. The feed capillary was fed intothe sample vial via a Swagelok tee-piece using a vespel ferruleto connect to the feed capillary and a rubber o-ring to connect tothe sample vial. Liquid was loaded into the sample vial by sy-ringe before fastening the o-ring fitting. The feed capillary exitwas submerged in the sample liquid. The tee-piece allowed N2gas pressure to be applied to the sample vial from a regulatorand measured using a digital manometer.

The liquid union was held in an insulator and the groundwiring connected the union to the fast current sensing equip-ment. This approach results in the liquid meniscus being heldat the ground potential via the conductivity of the liquid, ratherthan via a metallic coating at the tip exit. This reduces the oc-currence of corona discharge, a potential problem particularlywhilst spraying water [31].

The high voltage required to start the spray was applied toa polished aluminium disc held 3 mm away from the emitter

M.D. Paine et al. / Journal of Colloid and Interface Science 305 (2007) 111–123 113

Fig. 1. Schematic of the electrospray test rig.

on a separate insulator. The height of the electrode could beadjusted by micrometer. The majority of the emitter assemblywas shielded by a grounded metal cylinder in order to reducenoise (Fig. 1).

The spray equipment was initialised by the application ofgas pressure that forced the liquid into and through the spray tip.The application of a high potential difference meant the flowingliquid did not gather on the tip exit but was sprayed away fromthe tip. After any obvious bubbles were flushed through thisback pressure was removed and after a few minutes the voltageswitched off. The liquid was then held (by surface tension) atthe exit of the tip. The fluid surface in the liquid vial was heldat the same height as the liquid tip exit to ensure that there wasno net hydrostatic pressure acting on the liquid membrane (seeschematic in Fig. 1).

The electrospray current on the emitter was amplified fromthe nanoampere range using a variable gain high-speed currentamplifier (Laser Instruments, UK—model DHCPA-100) at again of 106 V/A at 1.6 MHz bandwidth. This signal was mea-sured by a digital storage oscilloscope (Wavetek, wavesurfer422) through 50 � DC coupling at 20 MHz bandwidth. Alldata was captured from a single scan with no averaging. In-dependent measurements of the average current at the extractorelectrode were obtained on-line using a non-grounded multi-meter. High voltage was applied to the collector to allow usto ground the emitter through the fast current amplifier. Thisallowed the monitoring of the emitted current rather than thecollected current with high temporal accuracy.

A high-resolution microscope monitored the shape of theliquid meniscus and determined the spray regime. The micro-scope consists of a Mitatoyu 10X infinity corrected objectiveon a Thales Optem 12.5x variable zoom, coupled with a SonyV500 CCD camera. The resolution of this video microscopewas ∼2 µm.

In each of the data sets, for a given nominal tip diameter, twodifferent emitters were used. Whilst it would be expected thatthe measured spray properties should be consistent within theband of measurement uncertainty, it was found that such mea-

Table 1Solutions used in the experiments

Name Liquid Dopant Conductivity (S/m)

T1 TEG 1.5 g/L NaI 2.4e−3T6 TEG 6 g/L NaI 8.7e−3T25 TEG 25 g/L NaI 3.3e−2E05 EG 0.5 g/L NaI 2.9e−3E5 EG 5 g/L NaI 2.7e−2W70 Water 70 µM NaI 8e−4W700 Water 0.7 mM NaI 7.3e−3W7000 Water 7 mM NaI 7.4e−2

surements seemed to lie outside the measurement errors. Webelieve that this is due to the detailed variation in the emittersas supplied, particularly in the internal emitter profile, since thedata we are taking is expected to be dependent on the internaland external properties of the emitters. As a result we have plot-ted the values determined for frequency, peak currents, etc., forboth sets of emitter measurements in all the data presented inSection 3.

2.2. Electrospray solutions

Ethylene glycol (EG), triethylene glycol (TEG) and distilledwater, were used as base solvents. To be stable in nanoelectro-spray mode at a flowrate of order 1 nL/min, a solution musthave conductivity greater than ca. 10−2 S/m. Pure solventsmust therefore be doped with an ionic compound. In the presentwork, EG, TEG and distilled water solutions containing varyingconcentrations of NaI were prepared. To avoid contaminationof the EG and TEG solutions with water vapour these solutionswere prepared in a dry box. The conductivity of the solutionsused in our tests is shown in Table 1. Conductivity was deter-mined using a novel triangular waveform method [32].

3. Results

All electrospray experiments were performed with no netpressure applied to the fluid to force fluid flow. The majority ofour attention here is on the mode previously identified as a vari-ant on the forced flow mode, termed axial mode II. These resultsare reported in Sections 3.1–3.3. However other modes werealso observed and these are reported in Sections 3.4 and 3.5.

The experimental method, followed for all the solutions, wasas follows. The voltage on the extractor was increased from zerountil steady oscillations were observed; this voltage is U0, theonset voltage of oscillations. For many nozzles, this point waspreceded by the sporadic appearance of current spikes with nodiscernable frequency. Corona discharge did not occur at suchlow voltages. These spikes were neglected. At each of the mea-surements taken above U0, the current trace was stored and animage taken of the meniscus, using the video microscope, inorder to identify any distinctive features. The period of the os-cillations and time averaged collector current were recorded.Corona discharge was ruled out by observing those sprays ob-tained at high electrical field using long CCD exposure times.


Fig. 2. Spray current oscillations for T25 on a 15 µm emitter.

3.1. General pulsation characteristics

Typical current waveforms obtained for TEG solution T25sprayed from a 15 µm diameter tip, are shown in Fig. 2. Thelegend indicates the voltage at which the trace was obtained.Only a few waveforms are shown to preserve clarity. The tracesshow that as the voltage increases the current peaks associatedwith the oscillations become closer. The data presented in thesecurves also shows, in this case, that the maximum current, Ipeakalso becomes larger, as the voltage is increased. These stableoscillations in unforced mode were originally identified in ourprevious work [21].

The time-averaged current measured with the multi-meter,Iave, increases in a near linear fashion with voltage throughoutthe pulsation regime. As the electrospray mode transforms intothe steady state cone jet regime, there was a noticeable increasein this average current. During the cone-jet mode the averagecurrent then continues to increase linearly with voltage.

In the majority (85%) of the tests undertaken with TEG solu-tions, the pulsation regime switched to a steady state operationof stable cone-jet mode. At a certain threshold voltage the cur-rent pulses changed to a steady current having a lower valuethan the maximum pulse peak currents. No oscillations couldbe observed in this state. Observation of the liquid meniscus re-vealed the cone apex and jet (the latter only visible for lowerconductivities) to be non-fluctuating.

Water is a common solvent for many electrospray appli-cations however, its properties differ considerably from tri-ethylene glycol, in particular its surface tension is much higherand viscosity is much lower. Pulsations of the same form asthose observed in TEG solutions, pulsation mode axial II werealso observed. A comparison between the raw pulse data re-veals that in water the pulse durations are more than an orderof magnitude shorter than for the TEG solutions; thus in waterpulse durations are typically of ∼2 µs, in comparison to TEGpulses lasting ∼50 µs. The shorter pulse duration is also asso-ciated with a much higher frequency pulsations.

The way the frequency of pulsations in water changes withthe applied voltage has another feature which distinguishes itfrom TEG. Thus in water there is a clear step from a low fre-quency albeit at 50 kHz to a very high frequency 200 kHzpulsation mode. Whilst this rapid frequency rise was shown inour previous work, for the tip used in that work no cone-jetmode was obtained. In two thirds of the water solutions testedin this work a transition from pulsation to a stable cone-jet doestake place under VMES control. Of those combinations whichentered a cone-jet mode 75% sustained the mode over a widevoltage range.

Ethylene glycol is similar to TEG in many respects, althoughits viscosity is ∼50% lower. A smaller number of experimentswere performed using two EG solutions, whose conductivityvalues span an order of magnitude difference. Fluid propertiesfor these solutions are also identified in Table 1. The generalcharacteristics of EG pulsations are similar to those observed inTEG, with there being no high frequency transition.

3.2. Axial mode II pulsation dependence upon applied voltage

A greater range of results was obtained using the solventTEG. This was because this solvent has the lowest surface ten-sion of the three liquids and as a result onset occurs at lowervoltage for a given tip size. The lower voltage in turn reducesthe risk of corona discharge.

Investigation of the effect of conductivity on the observedpulsation properties was examined by electrospraying the liq-uids T1, T6 and T25. This range of liquids provides a variationin conductivity over more than an order of magnitude. The on-set voltage for stable pulsations was identified to be a functionof the liquid/emitter combination. As a result, in order to com-pare results, rather than using the applied voltage Ua it is morephysically insightful to plot measured parameters as a functionof voltage above this onset voltage, Ua − U0. We define this tobe the voltage excess. The dependence of pulsation frequency,as a function of voltage excess for each solution is shown in


Fig. 3. In each of the data sets the emitter used was one havingan exit diameter of 15 µm. Error bars are included to reflect thefact that the period of the oscillations has some slight variation.This fluctuation is more noticeable at voltages close to U0 andin low conductivity solutions. The regular increase of pulsationfrequency with voltage excess as shown indicates that through-out the voltage range the pulsation mode is indeed axial II.

The frequency of the stable spray oscillation varies overmore than an order of magnitude for these three solutions. Theincrease in frequency appears to be linear with the applied volt-age. Comparison of the gradients for the best fit linear trendfor these data sets �f/�(Ua − U0), in the different liquidsalso shows that as the fluid conductivity increases, there is acorresponding increase in the rate with which the pulsation fre-quency increases with applied voltage. Indeed for this overalldata set, albeit comprising of only 3 gradient values, there ap-

pears to be a good correspondence between best fit of the gra-dient value �f/�(Ua − U0) versus conductivity K , with therebeing a linear trend, with a regression coefficient of 0.98. As aresult we conclude that the frequency of the pulsations obtainedfor a specific tip is higher for a higher conductivity liquid.

Investigation of the sensitivity of the peak current during apulse upon the applied voltage was also undertaken. Some fluc-tuation in the value of the peak current, Ipeak was observed inthe pulsations at a fixed value of voltage excess. As a result, inorder to get a measure for this important parameter the valueof Ipeak for typically up to 10 pulses were used. The values soobtained are plotted in Fig. 4, wherein the measurement fluc-tuation is indicated by the plotted error bars. These data wereobtained from 15 µm diameter tips. From this data the voltagedependence of the magnitude of Ipeak observed is rather un-clear. Thus in the highest conductivity liquid tested (T25) there

Fig. 3. Oscillation frequency against voltage excess for T1, T6 and T25 on 15 µm emitters.

Fig. 4. Peak pulse currents against voltage excess for T1, T6 and T25 on 15 µm emitters.


appears to be a linearly increasing trend in current with volt-age excess; the regression coefficient for this data is 0.991. Thegradient of current with voltage is however modest, and the to-tal range of peak current for this liquid varies by less than 25%of the mean value. The lower conductivity solutions show nodiscernable trend with applied voltage.

We conclude that the sensitivity of peak current to voltageis weak for the TEG solutions tested, implying that the maxi-mum rate that charge is removed during the pulsation is ratherinsensitive to the applied field.

As with the TEG data, both the water and EG experimentsshowed that decreasing the liquid conductivity results in lowerpeak currents. For the specific case of water the W70 solutionhad peak currents typically only 25% of those achieved withW7000. The dependence of Ipeak both in water and EG withapplied voltage again has a similar characteristic to those de-scribed for TEG, wherein sensitivity was more notable in thehigher conductivity solutions. This suggests that Ipeak does in-deed increase with applied voltage, however the quality of thedata at present is insufficient to resolve fully the nature of thedependence.

3.3. Axial mode II pulsation dependence on tip diameter

Experimental data was also obtained to identify how the tipdiameter affects the properties of the observed pulsations. Theproperties of interest are the pulsation frequency, the peak cur-rent and the total charge extracted during a pulse. As we haveseen from the preceding section the pulsation characteristics foreach liquid are dependent upon both the applied voltage and thesolution conductivity. In order therefore to make comparisonsbetween data sets it is necessary to identify specific conditionsfor these comparisons.

All liquids investigated demonstrated that the highest fre-quency of pulsations was always obtained at a voltage excess

just below that at which the pulsation mode was replaced bysome other spray regime. In many cases, including data ob-tained for water, this would be a transition to stable cone-jetmode. In certain examples, such as those taken on the largestemitter tip size, the spray mode could change to either a multi-jet mode or even a corona discharge. As a result, when mak-ing detailed comparison between liquids we have selected themaximum frequency, fmax as an appropriate way to capture fre-quency dependence. This data is collected for all the solutionsin Fig. 5, for each tip/liquid combination.

The overall data for the three TEG solutions shows that fmax

increases with both increasing conductivity and decreasing tipdiameter over the complete range of liquids and tip size.

These two trends for each solvent are also evident within thewater and EG data sets. It is also apparent that the highest fre-quency oscillations are obtained from high conductivity watersolutions sprayed from the smaller diameter tips. The highestfrequency pulsation observed was 0.63 MHz. We note that wa-ter is the lowest viscosity solvent tested, and that there is ageneral trend through the data sets that higher frequency pul-sations are observed for lower viscosity solutions.

We have already noted that for the highest conductivity TEGsolution tested, the peak current shows some sensitivity to theapplied voltage applied from one particular tip. However wehave concluded from the data presented in Fig. 4, that over-all this sensitivity is modest. As a result, but noting this asan approximation, we characterize here the peak current dur-ing a pulsation, for each solution, by the average value of Ipeak

observed over the entire voltage range for which stable axialmode II pulsations occur. This average value 〈Ipeak〉, as a func-tion of tip diameter, is plotted in Fig. 6 for the TEG data. Thesedata show a significant correspondence of 〈Ipeak〉 with both liq-uid conductivity and tip diameter. Thus on a given tip, as theconductivity of the solution increases, there is an increase in

Fig. 5. The effect of liquid conductivity and tip diameter on the maximum frequency.


Fig. 6. The effect of liquid conductivity and tip diameter on the average peak current during a pulse.

〈Ipeak〉. Additionally as the tip size increases, for a given solu-tion the value of 〈Ipeak〉 also increases.

In water, as with the TEG, the effect of reducing the tip di-ameter was again to lower the peak current during a pulse. Theaverage peak currents when spraying W7000 were 172, 73 and53 nA for 30, 15 and 8 µm tips, respectively.

There are two issues now to consider in relation to the com-bination of frequency sensitivity data and current sensitivitydata. The peak current identifies the maximum charge extrac-tion rate from the fluid meniscus, whereas the total charge ex-tracted from the meniscus, that is the integral of current throughthe pulse, gives an indication of the amount of material whichmay be removed from the meniscus during the pulsation, ifone assumes that the charges extracted are indeed solvated. Al-though the peak heights of the current pulses increase with both

conductivity and tip diameter the pulse duration was observedto decrease with conductivity and increase with tip diameter.

The data for all solutions tested for the pulse duration, Ton,is plotted in Fig. 7. Here, the on time, Ton has been defined asthe width of the pulse peak when the current is greater than0.25 × (Ipeak − Ibase) + Ibase. The longest pulse duration was159 µs, for T1 sprayed from a 30 µm needle, whilst the shortestpulse duration for TEG was 16 µs for T25 sprayed from a 4 µmnozzle.

Let us then approximate the charge ejected during one cur-rent pulsation to be given by Ipeak ·Ton. This approach has beenvalidated by comparing this value against that obtained for spe-cific measured waveforms by numerically integrating the pulseshape itself. This comparison revealed that there is good agree-ment between the two methods to within typically 10%. The

Fig. 7. The pulse on-time for all liquids.


Fig. 8. The effect of liquid conductivity and tip diameter on the charge ejected by a pulse.

calculated charge ejected during a pulsation for the solutionsis plotted in Fig. 8 against tip diameter. We reemphasize thatwe have used the average value of Ipeak for these calculatedvalues and data plotted in this figure is therefore an averagedpulse charge over the full voltage range over which stable pul-sation occurs. The data plotted reveals a strong trend whereinthe charge ejected during a pulse increases with the diameter ofthe tip.

In almost every case the charge ejected from a tip sprayingwater solutions is an order of magnitude lower than for the samesized tip spraying TEG solutions. This trend is also visible inthe EG solutions, with the charge emitted during a pulse beingmore comparable to the TEG solutions. It is interesting to notethat the EG data falls between that for TEG and that for water.

Although the data demonstrates some scatter, herein we haveonly plotted error bars for the noisiest data set in order to main-tain clarity, the charge ejected, for a given solvent, appears tobe independent of conductivity. This is most clearly seen in theTEG data.

The voltage, UCJ, at which the spray became a stable cone-jet, was dependent on the tip diameter, with no discernableinfluence from the liquid conductivity. The average onset volt-age excess, �Vave = 〈UCJ − U0〉, for all data from each nozzletip diameter were: 278, 495 and 717 V for 8, 15 and 30 µm tips,respectively. Clearly then the range over which pulsations occuris greater for a larger tip diameter. The cone-jet onset also takesplace at higher voltage for larger tips. This is in accordancewith the standard electrospray onset voltage model popularizedby Smith [33].

The onset of cone-jet mode shows a correlation with the pul-sation duty cycle, defined by pulse duration divided by the pe-riod Tperiod, associated with the pulsation frequency. The max-imum duty cycle is difficult to obtain precisely as the stabilityof the spray frequency is reduced as stable cone-jet operation is

approached. However, some simple observations can be made.The maximum duty cycle in all cases is always of the order of40–50%. We have not seen any evidence of a pulsing VMEStransitioning to a stable cone jet when the duty cycle is below20%. Similarly, we have not observed a pulsating electrospraywith a duty cycle greater than 59%. It appears that the pulsatingmode is unstable if the pulse duration is very close to the timebetween oscillations.

The onset voltage of the pulsations, U0, varied with the noz-zle diameter. For TEG the average U0 was 1044, 1443 and1753 V for 8, 15 and 30 µm diameter tips, respectively. Val-ues for EG were very similar. For water the average U0 was1423, 1782 and 2140 V for 8, 15 and 30 µm diameter tips, re-spectively, this reflects the higher surface tension of water.

3.4. Axial I mode in VMES

As we have noted not all the liquids show the same pulsationnature across the range of applied voltages wherein stable pul-sation modes may be observed. Thus particularly when spray-ing low conductivity water solutions on the larger tips, directcomparison of data is made more complex by the appearanceof new pulsation modes. Fig. 9 shows 2 sample waveforms ob-tained when spraying W70 on 30 µm tips. Both waveforms arereminiscent of the axial I pulsations described by Juraschekand Rollgen [26] in that there are very high frequency pulsa-tions (∼100 kHz) occurring in much lower frequency groupings(∼3 kHz). However this similarity is perhaps superficial dueto the following: (a) Juraschek and Rollgen’s findings were inforced, rather than unforced spray conditions, (b) in our newdata significantly higher frequencies but with a smaller num-ber of pulses form the pulse envelope. This is the first report ofaxial I pulsations during unforced nanoelectrospray or VMES.This mode of spraying was also observed in the EG solutions,


but only on the largest emitter having a tip diameter of 150 µm.The E5 solution exhibited double peaks only, whilst E05 exhib-ited a very large number of bunches of pulsations at frequenciesas low as 20 Hz. No axial mode I pulsations were observed inthe TEG solutions however.

This mode will only occur for the appropriate combinationof liquid and nozzle; the data obtained suggests a low value ofhydraulic resistance is required. The low viscosity of water cou-pled with the larger tip diameter means that small fluctuationsin pressure can result in relatively large liquid flowrates intothe cone. Since the mechanism behind axial mode I pulsationis thought to be the depletion and replenishment of the entireliquid cone [26,27], any disturbances may lead to relativelylarge-scale mechanical oscillations in the liquid meniscus.

3.5. The axial IIB mode

The calculated charge lost during a pulsation in Section 3.3is based on charge being emitted only during the ‘on-time.’A different measure can be obtained by integrating the currentwaveform over some period of time, not specifically related toany of the frequency characteristics of the data, say the data cap-ture time and then dividing this charge by the number of pulsescaptured; this calculation yields the charge ejected per pulse cy-cle, �Q. This approach fully includes any charge ejected in thetrailing edge of a pulse. A measure of current, termed here IDCmay be derived from this total charge, �Q being divided by thepulse on time, Ton. Fig. 10 plots IDC against voltage excess forthe TEG solutions on a 30 µm tip.

Fig. 9. The occurrence of axial I pulsations during VMES of W70 on a 30 µm tip.

Fig. 10. IDC against voltage excess for T1, T6 and T25 on a 30 µm emitter.


IDC increases with voltage excess for these solutions untila maximum is reached. This mode was named axial mode IIBin our previous work, however, it does not always occur. Dur-ing all the experiments undertaken here, this mode seems moreprevalent at higher conductivities and larger nozzle diameters.The axial IIB mode was also observed for some of the EG data,but was absent for all water solutions.

Low temporal resolution images taken of the liquid meniscussuggest a possible physical mechanism for this mode, as shownin Fig. 11. A larger nozzle was used for these images to allowthe change in meniscus shape to be seen clearly.

The image in Fig. 11a shows the meniscus deforming dueto electric stress, although in this condition there is no liquidejection. In the remaining images the meniscus undergoes sta-ble pulsations in either axial mode II or IIB, although the jet isnot discernable in these images. The values for IDC correspond-ing to images (b), (c), and (d) together with the charge ejectedduring a pulse, 〈Q〉, are given in Table 2.

Table 2The average current during a pulse, IDC and total charge ejected during a pulse,Q, for the sprays in Fig. 11

Voltage (kV) IDC (nA) 〈Q〉 (1 × 10−12 C)

2.5 207 5.82.7 146 2.82.9 147 2.5

The images show the size of the liquid cone decreasing as themeniscus becomes stressed by the increasing electric potential.Recalling the data plotted in Fig. 8, the average charge ejectedincreases with the size of the nozzle. In general the size of themeniscus may be presumed to be dependent on the size of thecapillary tip. Thus, if we assume the dependence in Fig. 8 is onthe size of the liquid meniscus, then the decrease in the chargeejected shown in Table 2 may be due to the reduction in the conedimensions. If this is correct then the axial mode IIB could beexpected to occur only in situations where increasing the volt-age causes the liquid cone to retract. This does not always occurduring the pulsation regimes, although it often occurs during thestable VMES cone-jet mode and always precedes the multijetmode.

4. Discussion

Many new features of stable pulsating nanoelectrosprayprocess have been observed. Not all pulsation modes are ob-served in all liquids in all capillary systems, and thus we caninfer that the combination of fluidic properties and geometricparameters that have been varied are such that their interac-tion leads to the differing observations. The results presenteddo however demonstrate definable characteristics.

Thus it is apparent that from Fig. 8, the amount of chargereleased during a pulse in axial mode II, increases as the tip

Fig. 11. Voltage effect on the liquid meniscus on E5 sprayed on a 150 µm tip: (a) 1.6, (b) 2.5, (c) 2.7, (d) 2.9 kV.


Fig. 12. Plot of (Qpulse · Ipeak)/(K · Dt) as a function of tip diameter Dt .

diameter increases. The data also indicates that this releaseis dependent, for a given liquid, upon the liquid conductivity.Since the pulsation is a quasi-static process one can infer thatthe collapse of the apex meniscus volume arises principally dueto a removal of charge from the apex more rapidly than thecombined effects of surface advection and bulk conduction cansupply charge to the meniscus. The rate at which charge is re-moved is described by the current waveform of the individualpulses as demonstrated in Fig. 2. We have also seen in Fig. 6,how the peak current during a pulse is dependent both on thefluid conductivity and the dimensions of the capillary tip. Fur-ther, the gradients of the best-fit linear regression of the datadisplayed in Fig. 6, show a distinct trend with the liquid con-ductivity: the high conductivity liquid having a steeper gradientthan the low conductivity data. These observations suggest thatthe combination of charge loss Qpulse and the ratio of peak cur-rent Ipeak to conductivity, K should also be a function of the tipdiameter.

A plot of such data does indeed reveal a broad correlationbetween the value obtained for Qpulse · Ipeak/K for a given liq-uid and the diameter of the tip. We may also regard this andprovide a physical context for this observation from a ratherdifferent starting point. Consider the electrical power requiredto drive the charge flux through the cone and meniscus intothe fluid jet. If the charge flux were to be dominated by bulkconduction, thus neglecting surface advection and bulk convec-tion of charge, during a pulse the total energy required may beapproximated over the pulse on-time, Ton, by

∫ Ton I 2Rcone dt

where Rcone is an electrical resistance associated with the fluidcone. This value for Rcone may be simply derived for a rightcircular cone, with base diameter Dt, of a solution whose con-ductivity is K . It is found to be ∝ (1/K) · Dt. Thus the energyrequired to drive the charge may be approximated to Epulse ∝(Qpulse · I )/(K · Dt) Thus a potentially revealing parameter toevaluate is the value of (Qpulse · I )/(K · Dt) to provide an ex-pression of the amount of electrical energy associated with the

pulsations in a given liquid. This energy value, derived fromdata for the three TEG solutions alone is plotted in Fig. 12.

As can be seen, there appears to be separation between theindividual solutions. The data seem well characterized by alinear dependence of energy with tip diameter, wherein thegradient of the best fitting trend is a function of the solutionconductivity. High conductivity solutions reveal a lower energyper pulse, and the rate at which energy increases with tip size isalso lower for higher conductivity TEG.

Consider now the other solutions tested. If we assume fromthe foregoing that conductivity influences the rate at which thepulse energy increases with tip diameter, it is most appropri-ate to compare solvent solutions having similar conductivity.Unfortunately, solutions with identical conductivity in differentsolvents are not available at this time. However two solutionshaving similar conductivity are the TEG solution T6 and thewater solution W70. The data for pulse energy for these is col-lected in Fig. 13. Again we see in the water data a similar trendof increasing energy with tip diameter.

For these two data sets presented, although of rather lim-ited scope, it is quite clear that the higher viscosity solution hasa higher energy requirement per pulse. It is interesting to notealso that the gradients of the best fit trend lines have very similarvalues, although at this stage it would be premature to concludethat this gradient is solely dependent upon the solution conduc-tivity.

In conclusion these results suggest that for liquids havinghigher viscosity more energy is required to drive the pulse, rel-ative to those of lower viscosity, in order to extract liquid in apulsatile jet. Additionally for a given tip diameter greater en-ergy is required to extract a liquid having lower conductivity.These observations suggest that any model developed to cap-ture the key features of the nanoelectrospray pulsation modemust necessarily include the defining role of bulk conduction ofcharge flow within the cone structure itself, as well as the roleof surface advected charge in defining the shape of the menis-cus itself and its deformation.


Fig. 13. Plot of (Qpulse · Ipeak)/(K · Dt) as a function of tip diameter Dt .

5. Summary

This work has investigated the characteristics of unforcedVMES for two very similar liquids, ethylene glycol and tri-ethylene glycol, as well as water. When spraying TEG solutionswe found the frequency of the pulsations was larger for higherconductivity liquids and smaller tip diameters. The peak heightsof the current pulses increased with both conductivity and tipdiameter pulse duration increases with tip diameter. We esti-mated the total charge ejected during a single pulse and foundthis to be smaller for smaller tip diameters. This may result fromthe charge ejected being related to the dimensions of the liq-uid meniscus, and so is fixed for a certain tip size for a range ofconductivities. Higher conductivity liquids result in larger pulsecurrents so the total charge is ejected more quickly, resulting ina shorter pulse duration.

The results from the water solutions showed a trend, similarto the TEG solutions, of higher frequencies for higher conduc-tivity and smaller tip diameters but the results were less con-clusive. However the maximum frequency obtained, 635 kHz,was 31 times higher than the maximum frequency obtained forTEG. Even for liquids of similar conductivities, W700 and T6,the water frequencies are considerably higher. In contrast, thelowest charge ejected by a water solution pulsation was an or-der of magnitude lower than from the TEG solutions.

A new VMES mode was reported in water, which was sim-ilar to the axial mode I described for forced flow [26] but ob-served here for an unforced flow. Water solutions were sprayedin stable cone-jets in the unforced VMES mode over wide volt-age ranges. This is the first report that uses the tools of fastcurrent measurement and fast microscopy imaging to verify thatthe stable cone-jet mode for water solutions in unforced electro-sprays is stable and free of current oscillations.

In the pulsation mode a fixed amount of charge and pre-sumably fixed liquid volume is ejected from each pulse. It isbelieved that the inability of the system to replenish the liquid

cone with either charge or liquid causes the pulse to stop. Theelectrical field then draws both charge and liquid to the apexregion until the surface charge and radius of curvature is suchthat the electrical stress overcomes the surface tension and thejet forms. As the field increases with voltage the time taken toreplenish the charge and liquid decreases and therefore the pul-sation frequency increases.

The analysis of the electrical energy required to drive thepulsations suggests that bulk conduction has a role in the chargetransport process. The pulsation energy is dependent on boththe fluid conductivity and viscosity.

Acknowledgments

Royal Society Research Award, EPSRC Basic TechnologyGrant.

References

[1] M. Yamamshita, J.B. Fenn, J. Phys. Chem. 88 (1984) 4451.[2] R.P.A. Hartmann, D.J. Brunner, D.M.A. Camelot, J.C.M. Marijnissen,

B. Scarlett, J. Aerosol Sci. 31 (2000) 65.[3] G.J. Deng, G. Sanyal, J. Pharm. Biomed. Anal. 40 (2006) 528.[4] M. Karas, U. Bahr, T. Dulcks, Fresenius J. Anal. Chem. 366 (2000) 669.[5] F.W. McLafferty, E.K. Fridriksson, D.M. Horn, M.A. Lewis, R.A. Zu-

barev, Science 284 (1999) 1289.[6] Z. Takáts, J.M. Wiseman, B. Gologan, R.G. Cooks, Science 306 (2004)

471.[7] M.S. Wilm, M. Mann, Int. J. Mass Spectrom. Ion Processes 136 (1994)

167.[8] L. De Juan, J. Fernandez de la Mora, J. Colloid Interface Sci. 186 (1997)

280.[9] A.M. Ganán-Calvo, J. Dávila, A. Barrero, J. Aerosol Sci. 28 (1997) 249.

[10] A. El-Faramawy, K.W.M. Siu, B.A. Thomson, J. Am. Soc. Mass Spec-trom. 16 (2005) 1702.

[11] A. Schmidt, J. Am. Soc. Mass Spectrom. 14 (2003) 492.[12] M. Wilm, M. Mann, Anal. Chem. 68 (1996) 1.[13] R. Juraschek, T. Dülcks, M. Karas, J. Am. Soc. Mass Spectrom. 10 (1999)

300.


[14] J.C. Hannis, D.C. Muddiman, Rapid Commun. Mass Spectrom. 12 (1998)443.

[15] M. Wetterhall, S. Nilsson, K.E. Markides, J. Bergquist, Anal. Chem. 74(2002) 239.

[16] Y. Li, R.B. Cole, Anal. Chem. 75 (2003) 5739.[17] G.A. Valaskovic, N.L. Kelleher, D.P. Little, D.J. Aaserud, F.W. McLaf-

ferty, Anal. Chem. 67 (1995) 3802.[18] S. Geromanos, J. Philip, G. Freckleton, P. Tempst, Rapid Commun. Mass

Spectrom. 12 (1998) 551.[19] S. Geromanos, G. Freckleton, P. Tempst, Rapid Commun. Mass Spec-

trom. 72 (2000) 777.[20] K.L. Smith, M.S. Alexander, J.P.W. Stark, Phys. Fluids, in press.[21] M.S. Alexander, M.D. Paine, J.P.W. Stark, Anal. Chem. 78 (2006) 2658.[22] G.I. Taylor, Proc. Royal Soc. 280 (1964) 383.[23] J. Fernández de la Mora, I.G. Loscertales, J. Fluid Mech. 260 (1994) 155.

[24] M. Cloupeau, B. Prunet-Foch, J. Aerosol Sci. 25 (1994) 1021.[25] A. Jaworek, A. Krupa, J. Aerosol Sci. 30 (1999) 873.[26] R. Juraschek, F.W. Rollgen, Int. J. Mass Spectrom. 177 (1998) 1.[27] I. Marginean, L. Parvin, L. Heffernan, A. Vertes, Anal. Chem. 76 (2004)

4202.[28] L. Parvin, M.C. Galicia, J.M. Gauntt, L.M. Carney, A.B. Nguyen, E. Park,

L. Heffernan, A. Vertes, Anal. Chem. 77 (2005) 3908.[29] J. Wei, W. Shui, F. Zhou, Y. Lu, K. Chen, G. Xu, P. Yang, Mass Spectrom.

Rev. 21 (2002) 148.[30] G.A. Valaskovic, J.P. Murphy III, M.S. Lee, J. Am. Soc. Mass Spec-

trom. 15 (2004) 1201.[31] J.M. López-Herrera, A. Barrero, A. Boucard, I.G. Loscertales, M. Már-

quez, J. Am. Soc. Mass Spectrom. 15 (2004) 253.[32] J. Wu, J.P.W. Stark, Meas. Sci. Technol. 16 (2005) 1234.[33] D.P.H. Smith, IEEE Trans. Industrial Applications 22 (1986) 527.

nozzle and liquid effects on the spray modes in nanoelectrospray

Documents