Droplet Dynamics Changes in Electrostatic Sprays ... - ResearchGate

9154 J. Phys. Chem. A 1998, 102, 9154-9160Droplet Dynamics Changes in Electrostatic Sprays of Methanol-Water MixturesZohra Olumee, † John H. Callahan, ‡ and Akos Vertes* ,†Department of Chemistry, The George Washington UniVersity, Washington D. C. 20052, andAnalytical Chemistry Section, Code 6113, NaVal Research Laboratory, Washington, D.C. 20375ReceiVed: April 28, 1998; In Final Form: September 22, 1998Two-dimensional phase Doppler anemometry measurements have been carried out to determine the size andvelocity distributions of electrosprayed droplets generated from methanol-water mixtures. We investigatedspraying conditions close to those of electrospray ionization sources. The droplet size and the axial andradial velocity distributions were measured as a function of liquid flow rate, needle-to-counter electrode distance,bias voltage, position of the probe volume, and electrical conductivity of the liquid. In 90:10 (v/v) methanolwatermixtures the droplet size decreased from ∼7to∼1.6 µm as the conductivity increased as a consequenceof a 3 orders of magnitude increase in ionic strength. As the position of the probe volume was moved alongthe spray axis, two different spray dynamics were observed. Solutions of low conductivity (c < 10 -5 M) onthe average produced 5.5 µm droplets at the capillary that gradually decreased to 4.0 µm as the drops movedaway from the tip. Solutions of higher conductivity (c > 10 -3 M), however, resulted in smaller droplets atthe needle (1.6 µm) that increased in size to 4.2 µm as the particles traveled toward the counter electrode.The droplet size reduction can be explained by evaporation and/or Coulomb explosion, whereas the increasein droplet size may be the consequence of droplet segregation or coalescence. Axial velocity distributionsshow compression along the spray axis (e.g., observed at 10 mm ∼50% reduction of the width at 4 mm isseen). Moving downstream, the average velocity of droplets from water-methanol mixtures decreasesmonotonically. Solutions containing KCl exhibit a maximum in axial droplet velocity as the probe volumemoves away from the capillary. These profile changes can be explained by differences in the electric fielddistribution along the spray centerline.IntroductionIn the 1980s Fenn and co-workers initiated a highly successfulapplication of electrostatic spraying as a source of ions for massspectrometry. 1,2 They showed that using electrospray large andnonvolatile molecules can be ionized and transferred into thegas phase without degradation. This discovery revolutionizedthe mass spectrometry of biomolecules. 3,4In the electrospray source, solute ionization is achieved byapplying high voltage to a metal capillary through which a slowflow of the solution is maintained. The potential gradientbetween the tip of the capillary and a counter electrode leadsto the deformation of the meniscus and to the emission ofcharged liquid fragments that often take the form of sphericaldroplets. Depending on the physical properties of the liquid(electric conductivity, dielectric constant, surface tension,viscosity, and density) and on the spraying conditions (fieldstrength at the tip, internal diameter and wettability of thecapillary, and liquid flow rate), a rich variety of spraying modescan be established. According to the phenomenological classificationof Cloupeau and Prunet-Foch, five major sprayingmodes can be distinguished (dripping, cone jet, microdripping,jet, and spindle modes). 5,6 This picture is substantially coloredby variants of these modes (e.g., pulsed cone jet, multicone jet,ramified jet) and by the occurrence of corona discharge atelevated voltages.* Corresponding author. Phone (202) 994-2717, Fax (202) 994-5873,E-mail vertes@gwu.edu.† The George Washington University.‡ Naval Research Laboratory.Systematic investigations on the dispersion of liquids intosmall charged droplets by an electrostatic field originated morethan a century ago when Rayleigh noticed that electrifieddroplets became unstable and underwent fission. 7 On the basisof Rayleigh’s early results describing an isolated droplet, Zelenyalso calculated the criteria for its stability. 7,8 They found thatthe potential at which instability began was proportional to thesquare root of the liquid surface tension.Hendricks and co-workers proposed a somewhat arbitrarymathematical model that related the solution properties, suchas surface tension, density, and conductivity, to the size andcharge of a droplet. 9-12 This group varied the solutionconductivity, the flow rate, and the field strength and measuredthe specific charge (charge-to-mass ratio) of the sprayed droplets.Both their theoretical and empirical models suggested that anincrease in solution conductivity and in field strength or adecrease in flow rate leads to an increase in the specific chargeaccording to a power law. 12Tang and Gomez reported results on atmospheric-pressuresize, charge, and velocity distributions of heptane and deionizedwater 13-15 obtained by two optical diagnostic techniques: phaseDoppler anemometry (PDA) and flash shadowgraphy. Theyprovided clear photographic evidence of the Coulomb fissionprocess of charged droplets under intense electric field stress. 13It was found that the droplet charge-to-volume ratio wasmonotonically decreasing with increasing droplet size. Accordingto their observations, larger droplets were closer to thelimiting charge at which they underwent fission. 13Recently, the charge and size distributions of electrosprayeddrops have been studied by Fernandez de la Mora and10.1021/jp982027z CCC: $15.00 © 1998 American Chemical SocietyPublished on Web 10/28/1998

Electrostatic Sprays of Methanol-Water Mixtures J. Phys. Chem. A, Vol. 102, No. 46, 1998 9155TABLE 1: Comparative Chart of Some Electrospray Characterization Studiesmeasured properties spraying mode spraying conditions a methods refcharge and size distributions cone jet dibutyl sebacate and benzyl alcohol differential mobility 16i.d. ) 200 µm, U ) “several kV” analyzeraerodynamic sizespectrometerspecific charge not reported glycerin with NaCl time-of-flight spectrometer 12size distribution, specificchargei.d. ) 500 µm, U < 30 kVcone jet water with NaCl at 10 kV < U < 20 kV PDA 14, 18corona-assisted cone jet heptane at U ) “few kV”, i.d. ) 120 µm flash shadowgraphycurrent measurement1519spray current1-octanol, water, benzyl alcohol,formamide with LiCl or H 2SO 4ethylene glycol, triethylene glycol,ion currentdripping cone jet water, CH 3OH, CH 3CN, water + mass spectrometer 1multicone jetCH 3OH (50:50) with LiCl or NaCli.d. ) 100 µm, 2.4 kV < U < 9.4 kVion current not specified water + CH 3OH with NaCl or HCl mass spectrometer 21, 22i.d. ) 100 µm,2kV< U < 3kVion current, spray currentspray currentcone jetnanospraypulsating cone jetcone jetacetone with rhodamine Gi.d. ) 250 µm at 2.5 kV < U < 3.6 kVi.d. ) 1-3 µmatU ) 600 Vwater + CH 3OH (50:50)2.2 kV < U < 4.0 kVai.d. and U denote the internal diameter and the voltage of the capillary, respectively.mass spectrometer,microscopecurrent measurement,microscope2320co-workers utilizing a differential mobility analyzer and anaerodynamic size spectrometer. 16,17 The sprayed drops wereselected within a narrow electrical mobility range, and subsequentlythe associated current or size was measured. Theirfindings indicated that the charge distribution of electrospraydrops generated in cone-jet mode was 2.5 times broader thanthe corresponding distribution of diameter. They related therandom process of jet breakup to the wavelength of axialmodulation of the jet. 16The complexity of spraying modes made the electrosprayphenomenon an elusive target for rigorous investigations andseriously limited the ability to compare results from differentlaboratories. Table 1 lists the measured spray properties andspraying conditions in some recent studies. It is clear fromcomparing the fundamental studies of the spraying process (firstthree rows of Table 1) with the studies on electrospray ionization(ESI) sources for mass spectrometry (rows four to seven of Table1) that there is very limited overlap between these groups inspraying conditions. While well-controlled sprays were characterizedin detail using powerful spray diagnostics, 12,14-16,18,19the sprays typically used in ESI are mostly studied by followingthe generated ion current or the capillary current. 1,20-23 Morespecifically, there is no detailed droplet size diagnostics availablefor the very important case of methanol-water mixtures sprayedat relatively low voltages (2.0 kV < U < 5.0 kV).Studying the capillary current as a function of capillarypotential in 50:50 methanol-water mixtures, Rollgen and coworkersestablished three spraying modes based on the presenceand frequency of current oscillations. 20 Mode I (2.2 kV < U< 2.5 kV) and mode II (2.5 kV < U < 2.8 kV) are characterizedby 30 Hz and low kilohertz current pulsation, respectively. Inmode III (2.8 kV < U < 4.0 kV) no current pulsation occurs.Optical investigation of the meniscus revealed that modes I andII corresponded to the pulsed cone-jet mode, whereas mode IIIcoincided with the cone-jet mode.The formation of charged droplets and their fate in the sprayare the basis of ion production in an ESI source. In ESI sourcesused in mass spectrometry, the solution properties can changedramatically from one sample to the other. For example, theionic strength and thus the conductivity of the sprayed solutionchange significantly with the salt content of the sample. Thus,Figure 1. Constant flow-rate electrospray setup consisting of a syringepump, syringe, stainless steel capillary, and planar ground electrode.Intersection of the PDA laser beams marks the probe volume.the main objective of our present study was to obtain dropletsize and velocity distributions for sprayed solutions typical ofESI under a broad range of spraying conditions. We alsowanted to characterize the generated droplets as a function offluid delivery system and liquid properties such as liquid flowrate, bias voltage, position of the probe volume, and electricalconductivity.Experimental SectionThe schematic representation of the electrospray setup isshown in Figure 1. A steady flow of the sprayed liquid wasfed by a syringe pump (Sage Instruments, Boston, MA) into ani.d. ) 150 µm and o.d. ) 510 µm stainless steel capillary. Tofacilitate the desolvation of the droplets and to prevent electricbreakdown, the capillary was surrounded by a tapered glass tubethat provided a coaxial flow of dry nitrogen (see inset in Figure1). The flow rate of the gas was selected by a needle valveand kept constant throughout the measurements. An X-Y-Zstage afforded the translation of the syringe pump and the

9156 J. Phys. Chem. A, Vol. 102, No. 46, 1998 Olumee et al.spraying assembly with an accuracy of 0.5 mm. Perpendicularto the capillary, a flat ground electrode was positioned at adistance of 30-35 mm. Stable spraying was achieved byapplying high voltage (3.0-4.8 kV) to the needle. These valueswere similar to the ones used in ESI sources for massspectrometry. 21,22 Lowering the voltage to 2-3 kV invariablyresulted in pulsating sprays. Most of the systematic studies wereconducted at U ) 4.0 kV capillary voltage. The flow rate ofthe liquid was maintained at 24 µL/min, which was typical ofESI systems 21,22 but significantly higher than the values in someof the fundamental investigations. 16 As shown in Figure 1, theprobe volume was defined by the intersection of four laserbeams.A two-color PDA system (Aerometrics, Sunnyvale, CA) wasused to simultaneously sample the size and the axial and radialvelocity distributions of the electrospray-generated droplets. Thescattered light from the particle traveling through the probevolume was picked up by a receiving lens of 250 mm focallength positioned at 30° angle in forward scattering mode. Thesize of the probe volume depends on the collection angle andon the focal length of the focusing lens. Using a 250 mm focallength lens and a 100 µm slit resulted in 200 µm width for theprobe volume. As has been pointed out by Sankar et al., areceiving lens with short focal length improves both the accuracyand the resolution in the measurements of small sized droplets. 24Since the transmitter and receiver were positioned on oppositesides of the flow, system alignment was crucial and difficult tomaintain. However, this orientation offered the best overallperformance, high S/N, and excellent size sensitivity for thePDA system. The scattered light was projected onto threephotomultiplier tubes (A, B, and C). Each detector produced aDoppler burst signal with a frequency proportional to the particlevelocity. The phase shifts between the Doppler burst signalsfrom the different detectors were proportional to the size of thespherical particles. Scattering obtained from nonsphericalparticles was rejected via a validation method based oncomparing phase shifts on different pairs of detectors (e.g., A-Bwith B-C). If the phase shift difference exceeded 5%, thereading was discarded by the signal-processing unit.Special care was taken to minimize the effect of scatteringby multiple particles. Regular direct observation of the Dopplerburst signal on an oscilloscope (TDS 320, Tektronix, Beaverton,OR) provided evidence that the probe volume was small enoughto avoid significant contribution to the scattering signal by morethan one particle. Temporal variations in the spray conditionswere monitored by repeating the PDA measurements at differentstages after establishing stationary spray conditions. Nosignificant variations were detected. Simultaneous size andvelocity measurements on the same droplet confirmed that therewas no size-dependent segregation between the droplets in termsof their velocities (see details in the Results and Discussionsection).Analytical reagent grade methanol (

Electrostatic Sprays of Methanol-Water Mixtures J. Phys. Chem. A, Vol. 102, No. 46, 1998 9157Figure 3. Effect of conductivity (ionic strength) on the average dropletdiameter, D 10, probe volume position, d pr, relationship. Ionic strengthof a 90:10 methanol-water mixture was adjusted by addition of KCl([, no KCl; 1, 5.0 × 10 -6 M; 2, 5.0 × 10 -5 M; b, 5.0 × 10 -4 M; 9,5.0 × 10 -3 M KCl). The grayed box on the right shows the cross sectionof the ground electrode.in the initial droplet size, the observed variations at larger probevolume distance showed different trends for low-conductivity(no salt, 5.0 × 10 -6 and 5.0 × 10 -5 M) and for highconductivity(5.0 × 10 -4 and 5.0 × 10 -3 M) solutions. Inagreement with our preliminary observations, increasing themeasurement distance from the capillary led to a decrease indroplet size for low-conductivity solutions, whereas highconductivity resulted in an increase in droplet size with distance.To compare our results on the conductivity dependence ofthe average diameter with previous studies, we used the scalinglaw introduced by Fernandez de la Mora for solutions withelectrical conductivities κ > 10 -7 Ω -1 cm -1 . 16 For a varietyof conditions the droplet diameter, D, can be approximated asD ) G(ɛ,Π µ )( ɛɛ 0 Qκ) 1/3 (1)where Q and κ are the flow rate and the electrical conductivityof the liquid, respectively. G(ɛ,Π µ ) is a slowly varying functionof the dielectric constant, ɛ, and the viscosity variable, Π µ .Itisassumed in the derivation of eq 1 that the liquid is conductiveenough so that the diameter of the jet is much smaller than thei.d. of the needle. In addition, the liquid has to satisfy specificconditions on polarity or viscosity. 16 Since (1) has also beenverified on deionized water, 14 the solutions used in the presentstudy also satisfy the conditions for validity. For strongelectrolytes κ can be written asκ ) (Λ 0 m - Kc)c (2)where Λ 0 m is the molar conductivity at infinite dilution and Kis the Kohlrausch constant. Assuming that for dilute solutionsthe variations in dielectric constant and viscosity are negligibleand keeping the flow rate constant, the concentration dependenceof the average droplet diameter is expressed asD′D ( c′)1/3( ) c Λ0 m - Kc 1/3Λ 0 m - Kc′ (3))where D′ is the average droplet diameter at c′ concentration.Figure 4. Variation of relative droplet diameter with the inverse ofthe relative concentration for electrosprayed dilute KCl solutions. Thesolid line represents eq 4, whereas the dashed line is a linear fit to theexperimental data (R ) 0.99).For the lowest KCl concentration in this study c ) 5.0 ×10 -6 M and Λ m ) 104 Ω -1 cm 2 mol -1 for the methanol-basedsolutions; 25 thus, κ ) 5.2 × 10 -7 Ω -1 cm 2 and eq 1 applies. Inthe concentration range of our interest variations in the molarconductivity of KCl are less than 6%; therefore, eq 3 simplifiestoD′D ≈ ( c c′) 1/3 (4)Figure 4 shows the measured relative diameter as a functionof relative concentration at d pr ) 2 mm probe volume positionfor KCl solutions spanning 4 orders of magnitude in concentration.The solid straight line with a slope of one in Figure 4corresponds to eq 4. Although there was clear deviation fromeq 4 at elevated concentrations, a linear fit of the data pointsyielded a slope of 0.82 and a regression coefficient of 0.99.Some of the assumptions, most notably about the concentrationindependentnature of G(ɛ,Π µ ), are likely to break down at highsalt content and may cause the discrepancies observed in Figure4.Shrinking droplet size with increasing distance from thespraying tip has been observed repeatedly for high vaporpressure liquids such as heptane 13,18 and for deionized water. 14The droplet size reduction in these systems is initiated byevaporation. Further dispersion of the liquid takes place as aconsequence of Coulomb fission leading to smaller droplets.The driving force behind Coulomb fission is that the evaporationincreases the charge-to-mass ratio of the droplets leading tosurface instabilities and eventually to their rupture. On the basisof this physical picture, we had no difficulty explaining theresults observed on low-conductivity liquids. It is not feasible,however, to explain the observed average droplet size increasein high-conductivity systems with the fission mechanism.The most plausible explanation for the apparent growth inthe observed droplet size is a fractionation effect. Cone-jet typesprays are known to have two major directional components.The first one is composed of the larger primary dropletstraveling close to the spray axis leaving under a small solidangle. The second component is the “skirt” of the spray madeof smaller satellite droplets departing under a larger solid

9158 J. Phys. Chem. A, Vol. 102, No. 46, 1998 Olumee et al.Figure 5. Charge and potential distributions in two homopolar dropletsat large distance (top) and at close proximity (bottom). As the dropletsapproach each other, excess charges congregate at the opposite endsof the spheroids, and due to the presence of the electrolyte shieldingtakes place. Depending on the conductivity (ionic strength), shieldingmay be sufficient to minimize electrostatic repulsion and promotecoalescence.angle. 13 As the spray diverges along its axis, the two componentsincreasingly segregate. 16 While the smaller satellitedroplets move off axis, the larger primary droplets remain closeto the center. This process leads to the enrichment of thesampled downstream axis locations in larger droplets. Increasingsegregation with increasing conductivity can be explainedby assuming enhanced space-charge repulsion due to theformation of smaller highly charged droplets.Another possible interpretation of the growing droplet sizeis the increasing incidence of droplet coalescence at higherconductivities, i.e., higher ionic strengths. Given the substantialamount of charges associated with the droplets, their collisionand coalescence are considered unlikely even at large particledensities. There are two factors, however, that might help toexplain the collision of homopolar droplets. The first one istheir high kinetic energy. Based on the measured values of theaverage diameter (∼2 µm) and the velocity (∼10 m/s), thekinetic energy of a single droplet is E kin ≈ 2 × 10 -6 erg or∼10 6 eV. More importantly, when two droplets of similarcharge approach each other, they undergo deformation andcharge rearrangement (see Figure 5). As the droplets approacheach other, their shape changes from spherical to elongatedspheroid, and the free charges originally located at the surfacecongregate at the opposite ends of the spheroid. Within thedroplet the ions of the strong electrolyte rearrange, so as theymodify the original Coulomb potential around the sphere intoa shielded Coulomb potential with a shielding length that isinversely proportional to the square root of the ionic strengthFigure 6. Axial velocity distribution of droplets electrosprayed from90% CH 3OH at different probe volume positions, d pr. Moving awayfrom the capillary tip, the average velocity drops and the velocity spreadis compressed. Corresponding radial velocity distributions are shownin the insets for d pr ) 4mmandd pr ) 22 mm.(see insets in Figure 5). Increasing the ionic strength shortensthe shielding length and at sufficiently high conductivity; i.e.,high electrolyte concentration may eliminate the repulsionbetween the two approaching droplets. For example, theshielding length, r D , for a 5.0 × 10 -6 M KCl aqueous solutionis 140 nm, whereas at 5.0 × 10 -3 M r D ) 4 nm. This sharpincrease in shielding in combination with the charge rearrangementmay lead to coalescence upon collision. We need tomention thatsas was shown by Rayleigh and Taylors too muchdeformation of a droplet leads to fission. 7,26 When the lengthof a droplet exceeds the equatorial diameter by a factor of 1.9,the droplet becomes unstable. Thus, collision of droplets canlead to coalescence, fission, or a combination of the two.Axial and radial velocity distributions of droplets sprayedfrom 90% CH 3 OH are shown in Figure 6. There was aremarkable variation in the axial velocity distributions as afunction of probe volume position. When the measurementswere conducted closer to the capillary tip (e.g., d pr ) 4 mm),broader distributions and larger mean velocities were observed.As the distance increased from d pr ) 4 to 10 and to 22 mm, theaxial mean velocity decreased from 34.5 to 11.4 and to 9.6 m/s,respectively. At the same time, the width of the velocitydistributions became significantly narrower. Comparing thedistributions revealed that the fwhm at d pr ) 10 and 22 mmdownstream were about one-half and one-third of the fwhm atd pr ) 4 mm, respectively. While slowing of the droplets canbe a natural consequence of the interplay between the electricfield and the drag force, the velocity compression can berationalized by looking at the actual form of the latter. Forflows with Re < 1000, the explanation can be based on thevelocity-dependent nature of the corrected Stokes drag force,F dr : 18F dr ) 3πµD(1 + 0.158Re 2/3 )(v - v g ) (5)where µ is the viscosity of the medium, Re is the Reynoldsnumber, and v and v g are the velocities of the particle and thegas, respectively. Assuming that the velocity of the gas isconstant in the region of interest, higher droplet velocitycorresponds to higher decelerating force. Thus, the fasterparticles are slowed more efficiently than the slow ones leadingto the bunching of velocities.

Electrostatic Sprays of Methanol-Water Mixtures J. Phys. Chem. A, Vol. 102, No. 46, 1998 9159Figure 7. Probe volume position dependence of average axial velocity,V ax, of droplets generated from a 90:10 methanol-water mixture (b)and from the same solvent doped with 5.0 × 10 -6 M(2) and with 5.0× 10 -3 M(9) KCl.Radial velocity distributions are shown in the insets of Figure6. These distributions were substantially narrower than theiraxial counterparts. They were centered around zero and showedno dependence on the probe volume position. This observationhad some implications on the forces controlling the motion inthe radial direction. Since the most significant forces actingon a droplet at these velocities are related to the external electricfield, E ext , the electric field associated with space charge, E sc ,and the drag force, one can writeF ) zeE ext + zeE sc - F dr (6)where z is the number of elementary charges, e, on the droplet.No change in the radial velocity distribution means that theensemble average of the radial component of F is close to zero.Thus, in radial direction either the electrostatic forces balanceout the drag force or all of them are negligible.For three solutions of different conductivity, the probe volumeposition dependence of the average axial velocity is displayedin Figure 7. Observing the data at d pr ) 2 mm probe positionindicated that increasing the conductivity led to lowering of theinitial velocity. Indeed, the droplets from a 90:10 methanolwatermixture departed with an average velocity >37 m/s,whereas from 5.0 × 10 -6 and 5.0 × 10 -3 M KCl solutions theyleft with a speed of 27 and 4 m/s, respectively. One mightconclude that this behaviorsin combination with the observedshrinkage in initial droplet size with increasing conductivitysis related to the morphological changes of the cone-jet structureand to the profound differences in field penetration due toshielding.Another interesting feature in Figure 7 is the difference inthe position dependence of droplet velocities from differentsolutions. The droplets from pure solvent (90% methanol)exhibited monotonically decreasing average velocities withincreasing distance from the probe tip, whereas the KCl solutionsshowed maximum velocity at a finite distance from the capillary.Tang and Gomez obtained the spray momentum equation onthe centerline in the following form:V z∂V z∂z = 1 m F z (7)Figure 8. Velocity-size correlation of droplets electrosprayed from90:10 CH 3OH-water mixture and from 5.0 × 10 -5 and 5.0 × 10 -3 MKCl solutions. In the lowest-conductivity case (no salt) there are somevery large droplets, whereas in the high-conductivity case (5.0 × 10 -3M KCl) the velocity distribution is substantially broadened. As theconductivity increases, the size distribution compresses and the averagevelocity as well as its spread grows.where the z subscript signifies the axial component of thevelocity and the force determined by eqs 5 and 6. The overbarindicates average quantities defined using the f(x,v) dropletdistribution function; for example, the average velocity iscalculated as v ≡ (1/N)∫f(x,v)v dv, where N ≡∫f(x,v) dv.Equation 7 is based on a number of specific assumptionsregarding the spray (no source or sink terms, no evaporationeffects, steady-state conditions, monodispersity, etc.). Some ofthese are clearly not fulfilled in our spray (changing dropletsize, departure from monodispersity), and some others are notverified. Thus, quantitative evaluation of the results based oneq 7 is unfounded. Already in this simplified form, however,eq 7 shows a relationship between the position dependence ofthe axial velocity and the distribution of electric field. Assumingthat there is no first-order effect by the conductivity on the dragforce, one can argue that the significant changes in the shapeof V ax (d pr ) with conductivity are primarily due to changes inthe space-charge component of the electric field. More specifically,one can assume that the accelerating effect on highconductivitydroplets near the capillary is the consequence of asharper potential drop in that region corresponding to “shielding”by the high-ionic-strength droplets.Axial velocity-size correlation charts for methanol-watermixtures at different salt content are shown in Figure 8. Inthis data acquisition mode velocity and diameter data aregathered on the same particle and rendered on a scatter chart.Statistical analysis of correlated measurements helps to findinterdependence between variables and to detect segregationeffects. The data were collected on the centerline, 10 mm awayfrom the capillary tip where the accelerating effect of the electricfield was negligible. One of the interesting features, observedwith 90% methanol, is the presence of an extended tail in the

9160 J. Phys. Chem. A, Vol. 102, No. 46, 1998 Olumee et al.size distribution leading to some very large droplets (D > 40µm). In accordance with Figures 3 and 7, at 5.0 × 10 -6 MKCl concentration the center of the distribution shifts to smallersizes and higher velocities. Somewhat unexpectedly, the spreadof axial droplet velocities also increases significantly. At 5.0× 10 -3 M KCl concentration the center of the distributionshifted to smaller diameters, and the average velocity (as wellas its spread) decreased.It was worth noting that there was no correlation betweensize and velocity for any of the solutions. Similarly, nosegregation of particles was observed. These findings supportthe mathematical decoupling of the droplet distribution function:f(z,V ax ) ) f 1 (z) f 2 (V ax ). In other words, the size and velocitydistributions measured separately can be combined to providethe droplet distribution function. It is known, however, thatthe radial component of the velocity shows correlation with thesize of the particles. 18 The larger primary droplets exhibit lowerradial velocities, whereas the smaller satellite droplets havehigher radial velocity components. This effect leads to thesegregation of the primary and satellite droplets under differentsolid angles. Although we measured radial droplet velocitydistributions, the corresponding size-velocity correlations werenot recorded. Moreover, due to the small size cutoff of ourPDA system (0.5 µm), most of the satellite droplets probablywent undetected.ConclusionsBased on PDA, observations of ESI type sprays used in iongeneration led to the recognition of significant changes in spraydynamics as the conductivity of the liquid was increased. Asthe conductivity grew within the boundaries of normal operatingconditions, we detected a reversal in trend from shrinkingaverage droplet size along the spray axis to a growing one. Atthe same time, the average axial velocity profile changed froma monotonic drop to maximum behavior. The samplinguniformity of these sprays improved downstream in terms ofboth their average diameter and axial velocity. For example,at 22 mm probe volume position V ax was ∼10 m/s independentof the velocity at the tip of the capillary and independent of theconductivity (Figure 7). Similarly, the D 10 values convergedas the droplets approached this region (Figure 3). This behaviorhas beneficial implications for the design of ESI sources.Prevailing source designs feature a comparable distance betweenthe spraying needle and the sampling capillary. To rationalizethe growing average droplet size in high-conductivity solutions,we offered two alternative explanations. One was based on sizedependentsegregation along the spray axis, whereas the otherevoked a coalescence model based on droplet-droplet polarizationand internal shielding.Acknowledgment. The authors thank the financial assistanceof the National Science Foundation (CHE-9512441) for purchasingthe PDA instrument. Z.O. is grateful for partial supportby the George Washington University. Discussions with KeqiTang of Finnigan Co. were valuable in providing alternativeexplanations for some of the data. The fabrication of the sprayingassembly and helpful advice on its operation by M. Shahgholiare appreciated.References and Notes(1) Yamashita, M.; Fenn, J. B. J. Phys. Chem. 1984, 88, 4451-4459.(2) Meng, C. K.; Mann, M.; Fenn, J. B. J. Phys. D: At., Mol. Clusters1988, 10, 361-368.(3) Fenn, J. B. J. Am. Soc. Mass Spectrom. 1993, 4, 524-535.(4) Fenn, J. B.; Mann, M.; Meng, C. K.; Wong, S. F.; Whitehouse, C.M. Mass Spectrom. ReV. 1990, 9, 37-70.(5) Cloupeau, M.; Prunet-Foch, B. J. Aerosol Sci. 1994, 25, 1021-1036.(6) Cloupeau, M.; Prunet-Foch, B. J. Electrost. 1990, 25, 165-184.(7) Rayleigh F. R. S. Philos. Mag. 1882, 14, 184-186.(8) Zeleny, B. A. Proc. Cambridge Philos. Soc. 1916, 18, 71-83.(9) Hendricks, C. D. J. Colloid Interface Sci. 1962, 17, 249-259.(10) Hogan, J. J.; Hendricks, C. D. AIAA J. 1965, 3, 296-301.(11) Swatik, D. S.; Hendricks, C. D. AIAA J. 1968, 6, 1596-1597.(12) Pfeifer, R. J.; Hendricks, C. D. AIAA J. 1968, 6, 496-502.(13) Gomez, A.; Tang, K. Phys. Fluids 1994, 6, 404-414.(14) Tang, K.; Gomez, A. J. Aerosol Sci. 1994, 25, 1237-1249.(15) Tang, K.; Gomez, A. J. Colloid Interface Sci. 1995, 175, 326-332.(16) De Juan, L.; Fernandez de la Mora, J. J. Colloid Interface Sci. 1997,186, 280-293.(17) Loscertales, G.; Fernandez de la Mora, J. J. Chem. Phys. 1995,12, 5041-5060.(18) Tang, K.; Gomez, A. Phys. Fluids 1994, 6, 2317-2332.(19) Fernandez de la Mora, J.; Loscertales, I. G. Fluid Mech. 1994, 260,155-184.(20) Juraschek, R.; Schmidt, A.; Karas, M.; Rollgen, F. W. In Proceedingsof the 45th ASMS Conference on Mass Spectrometry and Allied Topics,Palm Springs, CA; ASMS: Santa Fe, NM, 1997; p 117.(21) Kebarle, P.; Tang, L. Anal. Chem. 1993, 65, 972A-986A.(22) Tang, L.; Kebarle, P. Anal. Chem. 1991, 63, 2709-2715.(23) Wilm, M. S.; Mann, M. Int. J. Mass Spectrom. Ion Processes 1994,136, 167-180.(24) Sankar, S. V.; Weber, B. J.; Kamemoto, D. Y.; Bachalo, W. D.Appl. Opt. 1991, 30, 4914-4920.(25) Kebarle, P.; Ho, Y. In Electrospray Ionization Mass Spectrometry:fundamentals, instrumentation, and applications; Cole, R. B., Ed.;John Wiley & Sons: New York, 1997; pp 3-63.(26) Taylor, G. Proc. R. Soc. London, Ser. A 1964, 280, 383-397.