Comparing Direct and Indirect Thrust Measurements from Passively Fed and Highly Ionic Electrospray Thrusters

Daniel Courtney, Simon Dandavino and Herbert Shea

Microsystems for Space Technologies Laboratory, Institute of Microengineering Neuchâtel campus, School of

Engineering, École Polytechnique Fédérale de Lausanne, Rue de la Maladière 71b, 2000 Neuchâtel, Switzerland

Accepted for publication in the AIAA Journal of Propulsion and Power, 2015

This is an author generated pre-print of the final (accepted) draft manuscript.

Comparing Direct and Indirect Thrust Measurements from Passively Fed and HighlyIonic Electrospray Thrusters

Daniel G. Courtney∗ and Simon Dandavino† and Herbert Shea‡

Microsystems for Space Technologies Laboratory (LMTS),Ecole Polytechnique Federale de Lausanne (EPFL), Neuchatel, CH-2002, Switzerland

Highly ionic beams of several hundred μA per cm2 have been measured from porous glass ionicliquid electrospray sources fabricated using a conventional mill. The thrust output from threeprototype devices, two emitting the ionic liquid EMI-Im and one emitting EMI-BF4, was measureddirectly using a precise balance. Thrusts up to 50 μN were measured when emitting EMI-Im ina bipolar, alternating potential configuration at less than 0.8 W input power and with propellantsupplied from an internal reservoir. Measurements of mass spectra via Time of Flight spectrometry,angle resolved current distributions, ion fragmentation and energy deficits have been applied toaccurately calculate thrust and mass flow rates indirectly from the same devices. For two of threecases, calculated and directly measured thrusts were in agreement to within a few μN at inputpowers from 0.1 W to 0.8 W . Emissions of EMI-BF4 are shown to yield nearly purely ionic beamssupporting high propulsive efficiencies and specific impulses, ∼ 65 % and > 3200 s respectively at 0.5W . Conversely, greater polydispersity was observed in EMI-Im emissions, contributing to reducedspecific performance, ∼ 50 % propulsive efficiency and ∼ 1500 s specific impulse at 0.5 W .

Nomenclature

α = alternation duty cycleγ = liquid surface tension (N/m)ΔV = energy deficit (V)ηprop. = propulsive efficiencyθ = beam angle measuredθeff = effective beam angleθ0 = beam angle offsetκ = ToF thrust factorφ = ToF mass flow factorA = collector plate area (m2)fn = fraction fraction due to species with solvation nh = emitter height (μm)I(t) = ToF collector current (A)Ib = beam current (A)Iem = emitted current (A)Iex = extractor grid current (A)Ic = un-gated collected current (A)Isp = specific impulse (s)FT/m = piecewise fragmentation modifiers to ToF calculations

j(θ) = collected current density (A/m2)Kξ = kinetic energy of species ξL = ToF flight distanceLc = distance from the source to the collector (m)mξ = mass of species ξ (kg)m = mass flow rate (kg/s)q = species charge (C)tξ = flight time of species ξRc = emitter apex radius of curvature (m)s = emitter apex to extractor grid separation (μm)t = flight time (s)T = calculated thrust (N)TDTM = directly measured thrust (N)TITM = calculated, averaged bi-polar thrust(N)TToF = thrust calculated by ToF alone (N)Vem = emitter voltage (V)Va = average beam potential (V)Vstart = estimated starting potential for emission (V)

∗Postdoctoral Researcher, EPFL-IMT-LMTS, AIAA Member,dcourtney@alum.mit.edu .†Deputy Director, EPFL-eSpace.‡Associate Professor, EPFL-IMT-LMTS.

1. INTRODUCTION

Interest in electrospray propulsion has seen a resur-gence over the past decade with the advent of techniquesfor fabricating high current density emission sources[1–5]and the performance benefits identified when employingroom temperature Ionic Liquids (ILs) as propellants[6–8].IL electrosprays can yield highly ionic and monoenergeticbeamlets, leading to high specific impulse and propulsiveefficiency from inherently miniature emission sites (Tay-lor cones[9]). The thrust per IL emission site is small (∼10-100 nN); hence challenges in achieving useful devicesare associated with scaling up the total emission cur-rent while maintaining good specific performance. Thisneed can be compared with, for example, plasma basedpropulsion systems where difficulties in maintaining highpower efficiency quickly arise when attempting to minia-turize (eg. [10]). The technology is therefore well suitedfor small or distributable propulsion systems that requirehigh Delta-v; of which there is a growing need/interestyet relative dearth of fully qualified systems[11, 12].

Electrospray propulsion systems targeting the PurelyIonic Regime (PIR) of emission from ILs, hereafter re-ferred to as Ionic Liquid Ion Sources(ILIS)[7], are a par-ticular niche sharing functional similarities yet numerousbenefits compared with both colloidal [13] and LiquidMetal Ion Source LMIS[14, 15] thrusters. The potentialbenefits of ILIS include: i) the low surface tension of ILsleads to low operating voltages compared with LMIS, ii)ILs are liquid at room temperature and therefore maynot require heating, iii) many ILs have negligible vapourpressure, enabling storage at vacuum, iv) positive andnegative ion emissions can be sustained, enabling chargeneutralization without a dedicated neutralizer, v) highspecific impulses, in excess of 3000 s[16], and propul-sive efficiencies nearing 90 %[8] and finally, vi) the low

2

flow rates inherent can be accommodated through passivefeeding alone[2, 7, 17]. The later feature could greatlyreduce the mass, power and volume requirements of athruster by obviating the need for pumps, valves andpropellant pressurization systems.

In this report we present a simple to fabricate formof ILIS and demonstrate, through both direct and indi-rect thrust measurements, that 10’s of μN thrust levelsand good specific performance can be achieved from asmall (16 g, ∼ 3.7 cm x 3.7 cm x 0.8 cm) device whereinpropellant is supplied to emission sites passively.

The low absolute thrusts (≈ 10’s of μN) yet rela-tively high system masses[18] of electrospray thrusterscan present challenges when attempting to obtain DirectThrust Measurements (DTM) using a dedicated balanceor other force transducer, leading to few examples in theliterature. In Ref. [18] precise measurements and cor-responding indirect calculations are presented for a col-loidal thruster. However, Ref. [16] presents the onlyinstance of DTM from a passively fed ILIS known to theauthors. That study confirmed that thrusts of ∼ 0.1μN/μA can be achieved from arrays of ILIS and thatthese levels are close to those expected by extrapolationof indirect calculations based on individual emitter mea-surements. In section 5 we present both direct thrustmeasurements and a collection of probe measurementsfrom the same devices, enabling the thrust to be calcu-lated indirectly for comparison.

Indirect Thrust Measurements (ITM) by Time ofFlight (ToF) spectrometry[19–22] are often applied tocalculate expected performance. ToF measurements,which account for the spread in species charge to massratio within the beam, are suitable for providing a foun-dation for performance due to the potential diversity ofparticles emitted from electrospray sources. However;other factors such as energy deficits, the angular beamdistribution, electrode impingements and ion fragmen-tation may alter the output to a measurable degree. Insection 2 we arrive at performance expressions derivedfrom a series of measurements targeting each of thesefactors. Subsequently in sections 5 and 6 we presentand discuss a comparison between DTM and ITM ascalculated using this approach.

Electrospray Source Description

Several groups have targeted high current density, pas-sively fed ILIS which can provide 10’s of μN of thrustper cm2, or equivalently ∼ 100’s of μA/cm2. Depend-ing on the type and geometry of ILIS emitter structures,each emitter may support multiple emission sites (pre-sumably stable Taylor cones) and yield from ∼ 0.1[7] to> 5 μA[17, 23] of emission current. In order to generatesufficiently strong electric fields for emission while facili-tating liquid transport[24], structures ∼ 100 μm tall andterminated by an apex radius or capillary orifice of O(10)μm have been targeted. Such dimensions are well withinthe capabilities of numerous microfabrication techniques,leading to a myriad of approaches. These include silicon

Liquid flow

Extractor

grid

Particle beam

Emission site h

Linear emitter strips

Multiple emission

sites per strip

Porous

borosilicate

Vem

s

dg

p

α

FIG. 1: Emitter strips comprising triangular prisms have beenformed by milling commercially available borosilicate glass.Each emitter edge, aligned with a suspended extractor grid,is anticipated to yield a random distribution of emitter sites.

capillaries[3, 4, 25, 26], silicon tips roughened throughetches or CNT growth[1, 17], electrochemical etching ofporous metals[16, 27], use of Rosensweig instabilities inionic liquid ferrofluids[28] and laser ablation of porousglass[5].

Arrays of externally wetted[17] and porous [2, 5] ILIShave, encouragingly, been demonstrated to support a fewhundred μA per cm2 of highly ionic emissions. Poroussubstrates are particularly advantageous in that they in-herently provide a simple and integral flow path froma rear mounted and hydraulically coupled reservoir; en-abling simple capillary driven and zero-g compatible pas-sive feeding. In contrast, externally wetted arrays have,to date, been operated from a drop of liquid placed onthe surface; with a continuous passive flow supply yet tobe demonstrated.

We have identified and exploited two critical traits ofrecently demonstrated porous emitter arrays in develop-ing a unique, low cost fabrication technique:

First, compared with precise siliconmicrofabrication[26], the tendency for porous emit-ters to achieve high and ionic currents is particularlyremarkable considering the relative imprecision andrange in apex radii typical of processes which havesuccessfully been applied (eg. few to 10’s of μm in [27]).In porous materials, the Taylor cone base diameterlikely scales with the characteristic length scale of thereservoir pore’s Laplace pressure; leading to observationsof multiple, small emission sites from each emittingstructure[24]. Although these trends require furtherunderstanding, the behavior is indicative of an apparentdecoupling between the fluidic transport geometry (thepores) and that determining the electrostatic field (theemitter) when using porous structures. If a shape can

3

be fabricated to yield sufficiently high fields to establishemission sites localized to pores, highly ionic and highcurrent beams have tended to follow; including atmultiple points on each structure.

Second, the high conductivities and moderate dielec-tric constants of ILs enable emission from non-conductiveglass structures[5, 29]. Furthermore, the brittle natureof porous glass enables abrasive machining with reducedrisk of sealing surface pores, as could result with rela-tively malleable porous metals.

Figure 1 presents an overview of the emitter geome-try and functional configuration applied here. In lieuof individual microfabricated emitters, a small conven-tional CNC mill was used to create emitting structures oncommercial 1 cm diameter borosilicate filter discs (DuranGroup Inc. P5 grade). This method is not well suited forfabricating 2D arrays of small emitters as the pitch wouldlimited to the end-mill diameter. Instead we have tar-geted linear emitter strips comprising triangular prisms∼ 350 μm tall, 7 mm long and having a half-angle of 30o.The apex radius of curvature is not controlled but must,at least, be larger than the particles within the porousglass, typically a few μm. In the demonstration devicespresented here, 9 emitter strips are fabricated per disc,spaced 0.8 mm apart. The milling process is detailed insection 3 and has been developed exclusively for formingporous glass structures. However, porous metal struc-tures of similar geometry could warrant investigation ifthe risk of pore sealing during milling was well managedor suppressed.

As indicated in the figure we anticipate numerous emis-sion sites to form along the sharp edge of each emitterstrip. These emission sites are expected to form some-what randomly, wherever local electric fields, pore lociand hydraulic interactions permit. A first, empirical, ap-proximation as to the number of anticipated emissionsites can be made through considering Ref. [24]. There,porous nickel emitters were positioned roughly 50 to 210μm below a solid stainless steel plate which served asboth an extractor electrode and a simple target. Afteroperating for 10’s of minutes, numerous distinct impres-sions were observed on the plate and clustered aroundthe positions of individual emitters. The impressionswere assumed to be indicative of individual emission sites.Of particular relevance, when using axisymmetric emit-ter structures terminated by a flat plateau rather thana sharp point, numerous impressions were recorded inan often circular pattern. That observation particularlysupported the notion of multiple emission sites formingfrom pores near the strongest electric field. In that ex-periment, plateau perimeters were on the order of 100μm. Inferring that a few emission sites can thereforebe sustained per 100 μm of length, we anticipate that∼ 500 or more emission sites may be feasible from thepresented devices considering the total edge length of 63mm; although the materials and electrostatic configura-tions differ significantly.

2. INDIRECT PERFORMANCECALCULATIONS

Indirect thrust measurements (ITM) have been calcu-lated with an emphasis on ToF spectrometry, comple-mented and refined with beam angle and Retarding Po-tential Analyzer (RPA) measurements. We first reviewthe interpretation of the data acquired from each of thesemeasurements in order to define corresponding propulsiveparameters which enable functional thrust and mass flowrate calculations.

ToF measurements, where current is recorded at a de-tector some known distance from a fast electrostatic gate,have been applied in numerous propulsion studies[5, 19,21, 22, 30, 31] throughout the development of electro-spray thrusters. Delays between the gate signal andchanges in the beam current detected indicate the flighttime, and thereby speed, of constituent particles withinthe beam. The flight time tξ of particle ξ, relates to itsmass mξ and energy qVa through equation 1.

tξ = L

√mξ

2q|Va| (1)

In so far as the beam is assumed to be monoenergetic,with accelerating potential Va, the distribution of currentover flight time thus equivalently provides a distributionover mass to charge ratio mξ/q. The thrust and massflow rate due to the collected species can be calculatedthrough appropriate integrals of the ToF trace[19],

TToF = −2|Va|L

∫ ∞

0

FT (t)tdI(t)

dtdt (2)

m = −2Va

L2

∫ ∞

0

Fm(t)t2dI(t)

dtdt (3)

Here TToF refers to the thrust as derived from ToFdata and serves as the foundation for ITM in this paper.L is the ToF flight tube length and I(t) is the collectedcurrent as a function of flight time. The factors FT (t)and Fm(t) are non-dimensional modifiers to the ToF inte-grand added to account for solvated ions which fragmentafter emission but prior to complete acceleration[32].

We are focused on IL electrospray sources operating ator near the PIR (ILIS[7]). Emitted beams are thereforeprimarily comprised of ions of the form [AB]nA

+ and[AB]nB

− in the positive and negative emission modesrespectively. Here A+ and B− are the positive and nega-tive constituent ions of the IL respectively and n indicatesthe degree of solvation. The PIR has typically referred toemissions comprising primarily n = 0 and n = 1 ions, asmall population of n = 2 ions and no significant currentwith n > ∼ 4[7, 16, 33].

The ion population may fragment significantly in thefield-free regions between the source and ToF detector.

4

However these events to not alter performance and thoseions are not expected to possess a significantly differ-ent speed than the parent. Hence, in the context of theFaraday cup-type detector used in this study, the speedof those detected ions alone can be interpreted to cal-culate the original parent mass. Conversely, ions whichfragment between the emitter and extractor electrode ac-celerate to a greater speed than the parent ion, and there-fore have a lower flight time. Without adjustment, ToFbased calculations attribute intermediate masses to theseions assuming the complete energy qVa whereas, in actu-ality, signals at inter-ionic flight times are indicative oftwo emitted particles each with energy less than qVa: anion of massmi and a neutral particle of massmN which isotherwise undetected. We have shown in Ref. [32] that,if specific fragmentation transitions are assumed, FT (t)and Fm(t) can be simply calculated using equations 4and 5 respectively.

FT (t) =Ki(t)

qVa+

(mN

mi

Kn(t)Ki(t)

q2V 2a

)1/2

. (4)

Fm(t) =

(1 +

mN

mi

)Ki(t)

qVa. (5)

Energy measurements of ILIS beams have indicatedthat fragmentation events from solvation n to n − 1 aredominant[34]. Neglecting other transitions, the ion andneutral fractional kinetic energies are then simply relatedto the flight time as the functionsKi(t)/Va = t2n−1/t

2 andKn(t)/qVa = 1 − t2n−1/t

2 respectively. Here tn−1 refersto the ion flight time after breakup and the expressionsare piecewise functions for each region tn−1 < t < tn. Insection 5, we find that the effects of these adjustments onaggregate calculations are small, a few percent, yet notnegligible.

It is convenient to isolate the influence of dis-tributed mi/q, the primary output of ToF, fromthe emission current and accelerating voltage inequations 2 and 3. In this study, we con-sider two ionic liquids: 1-ethyl-3-methylimidazolium-bis(trifluoromethylsulfonyl)imide (EMI-Im) and 1-ethyl-3-methylimidazolium-tetrafluoroborate (EMI-BF4). Asthe EMI+ ion is common to both liquids, we define anon-dimensional flight time t such that t = tEMIt wheretEMI is the flight time expected by equation 1 for an ionof mass mEMI = 111.2 Da. Furthermore, we define anon-dimensional current I such that I(t) = IcI(t) whereIc is the current collected in the absence of a gate. Itthen follows that,

TToF = κ

√2mEMI

qIc√|Va|, (6)

m = φmEMI

qIc, (7)

κ = −∫ ∞

0

FT (t)tdI(t)

dtdt, (8)

φ = −∫ ∞

0

Fm(t)t2dI(t)

dtdt. (9)

The parameters κ and φ are non-dimensional coeffi-cients relating thrust and mass flow rate, respectively, tothe the same metrics if the beam were attributed solely toparticles of mass mEMI at the same current and voltage.Recognizing that −dI(t)/dt is a normalized distributionof current versus flight time and, by equation 1, that tis proportional to

√mi/q, we may alternatively view κ

and φ as providing relative measures of the mean√mi/q

and mi/q respectively. Ultimately these non-dimensionalparameters enable a quantitative view to the consistencyin beam composition, as relevant to propulsive perfor-mance, decoupled from voltage and current.

In applying equation 7 to estimate total mass flow wehave assumed an average of φ over beam angle and emit-ter voltage and therefore replace the ToF detector currentIc with the total emitted current Iem.

To calculate total thrust, consideration must be madefor the spread of beam angles with respect to the thrusternormal. Depending on the source type and measure-ment definition, ILIS beam half-angles may range from∼ 15o[7] to 40o[35]. To maintain relevance to propul-sive parameters we define a beam angle parameter basedon thrust rather than current inclusion. Consider probemeasurements at an angle θ to the thruster normal. Re-ferring to equation 6 the thrust contribution due to anelement of collected current dIc = jdA is,

dT (θ) = κ

√2mEMI

q

√|Va|j(θ) cos θdA. (10)

Here j is the current density at the sample angle θ, anddA refers to an elemental area at the probe collectionpoint some Lc from the source. κ and Va may vary overangle; however (see section 6.4) we have observed thesequantities to remain reasonably stable from our sourcesand hence consider them to be constant. The total thrustcan then be estimated through integration of equation 10over a surface S enclosing the beam. Normalizing by thetotal beam current Ib such that j = jIb, and integratingwe obtain,

T = κ

√2mEMI

qIb√|Va| cos θeff, (11)

cos θeff =

∫S

j (θ) cos θdA ≈ πL2c

∫ θmax

0

j (θ) sin 2θdθ.(12)

Intrinsically, in eqn. 11, it has been assumed that thecomplete beam current (Ib) has been sampled; such that

5

the surface integral of j is unity. The angle θeff is theangle of a hollow emission cone yielding the same thrustas the device under test. The approximation on the rightof 12 assumes an axisymmetric beam of extent θmax andis the form ultimately applied here. In section 5, an off-set angle θ0 has also been calculated permitting axisym-metric beam approximations about an axis tilted withrespect to the thruster.The accelerating voltage Va will generally differ from

the applied emitter voltage Vem by some deficit ΔV . Re-ported values of ΔV range from a few volts[34, 36] tomore than 100 V [4] depending on the source and oper-ating conditions. Energy deficits may include intrinsicphysical mechanisms, such as the ion solvation energyand Ohmic losses at the Taylor cone apex, or externalinfluences such as series resistances upstream electricalpath. Some such losses may not be constant over speciesmass[37]. In this study the impact of such variability wassmall and average fractional energy deficits were assumedfor the entire spectrum. These deficits have been deter-mined via both ToF and RPA measurements, see section5.Finally, it is important to acknowledge that the beam

current contributing to thrust Ib may be lower than themeasured emitter current Iem; due to interception at theextraction grid Iex, such that Ib = Iem − Iex.Considering the accumulated factors described, in

equation 13, we relate the expected thrust to the read-ily measurable extractor voltage and beam current alongand with the known mass mEMI. The specific impulseand propulsive efficiency in equations 14 and 15 respec-tively then follow, with the mass flow determined byequation 7 with Ic set to Iem.

T = κ

√Va

Vemcos θeff

[√2mEMI

qIb√|Vem|

](13)

Isp =T

gm=

κ

φ

IbIem

√Va

Vemcos θeff

⎡⎣1

g

√2q|Vem|mEMI

⎤⎦ (14)

ηprop. =T 2

2mIemVem=

(IbIem

)2Va

Vem

κ2

φcos2 θeff (15)

The specific impulse Isp is expressed as that which wouldresult from pure EMI emission at the emitter voltagemagnitude (in square brackets) modified by parametersdue to the mass distribution, beam interception at thegrid, energy deficits and angular spreading. Typical highcurrent passively fed IL electrospray sources operate withVem ∼ 1-2 kV [2, 5, 22], corresponding to an idealized spe-cific impulse of ∼ 4000-6000 s for particles of mass mEMI.However, we demonstrate in this report that lower values,∼ 1000-3000 s, may result once all the listed parametersand both emission polarities are considered.The constituent terms contributing to propulsive effi-

ciency ηprop. in equation 15, are largely analogous with

those considered in other electrostatic propulsion sys-tems (eg. [38]) and are discussed in greater detail in[39] and [34]. The terms Va/Vem and cos2 θeff are eas-ily recognized as extraction energy and angular lossesrespectively while (Ib/Iem)

2, a form of utilization effi-

ciency, arises due to the loss of propellant and expendedenergy associated with accelerated particles striking theextractor grid. The quantity κ2/φ, is of particular im-portance for electrospray thrusters[19]. Referred to hasthe “polydispersive efficiency”[8], it provides a measureof the apparent power loss associated with the emissionof a disperse beam, with respect to charge to mass ra-tios. Low polydispersive efficiencies (<50%) can resultfrom only a few percent of large droplets within an oth-erwise ionic beam[40]; greatly penalizing the power ef-ficiency when targeting a high specific impulse throughthe PIR. This relatively severe loss mechanism has ledto significant focus on achieving the PIR[26, 41], or morerecently and with promising results, obtaining extremelysmall (∼ few kDa) monodisperse droplet beams[20]. Inthis regard, ILISs which achieve the PIR through exter-nally wetted[7] or porous[16] electrospray emitters are apromising avenue for high efficiency, high specific impulsedevices. Indeed, in [8] a relation similar to equation 15was derived and used to motivate that total propulsiveefficiencies approaching 90 % may be achievable throughILISs emitting EMI-BF4. In section 6 we compare thethrust calculated as per equation 11 with direct thrustmeasurements and apply equations 14 and 15 to esti-mate the specific impulse and propulsive efficiencies ofthree demonstration devices.

3. SOURCE PREPARATION

Three devices have been fabricated and tested, twowet with the IL EMI-Im and a third wet with EMI-BF4.These liquids were selected due to existing heritage withILISs in the literature (for example see [34, 37] and [7] forstudies of EMI-Im and EMI-BF4 respectively). Further-more the heavier negative ion mass of Im− (280.2 Da)versus BF−

4 (86.8 Da) and possibility of droplet (or atleast very large n) emissions[37] were expected to yieldhigher thrusts at the expense of polydispersive efficiency.The two EMI-Im and single EMI-BF4 sources are here-after referred to as SRC-Im-1, SRC-Im-2 and SRC-BF4

respectively.

3.1. Fabrication Process

Duran Group P5 grade 1 cm diameter, 3 mm thickporous filter discs have been machined using a small ta-ble top CNC mill (Step-Four Basic 540 ). The formingprocesses is outlined in Figure 2 where dark (colored on-line) regions indicate material to be removed at eachstep. The process is comprised of the following steps:i) A square profile is defined to facilitate mounting, ii)

6

i

ii

iii

iv

v

10mm

2mm Ø

0.5mm Ø

FIG. 2: Process used to machine the emitter strips used in this research from a commercially available borosilicate filter disc(Duran Group P5 grade). Coloured segments are removed at each step with a 2 or 0.5 mm end-mill as shown.

FIG. 3: Side profiles of the three devices used in this test afterCNC machining. Alignment errors between steps iv and v inFig. 2 lead to the asymmetric shapes visible in SRC-Im-1 andSRC-Im-2.

a clear out region and slight chamfer are defined to pro-vide a buffer between the mounting flange end emitterstrips iii) The sample is rotated by 30o with respect tovertical and 9 cuts are made, iv) The sample is rotatedto -30o with respect to vertical and offset strips are cutto fully define the emission edges, v) Finally, materialbetween and beyond each emitter strip is removed to ex-pose triangular prisms. Step i) and ii) through v) aremade with 2 mm and 0.5 mm diameter TiAlN coatedend-mills respectively. Due to the abrasive nature ofborosilicates, a new 0.5 mm bit was required for eachdevice. Figures 3 and 4 present photographs and a mul-tifocal microscope image of the source emitters used inthis study. The multifocal image in Figure 4 demon-strates that while low radii of curvature, a few 10’s ofμm, are possible with the method, an undulated ridgewith a range of apex radii was typical. The asymmet-ric profiles in Fig. 3 for devices SRC-Im-1 and SRC-Im-2

are evidence of inaccuracies in the milling reference pointbetween rotated configurations. However, relative to theapex radius, these asymmetries lie far from the antici-pated emission site loci and the influence is not expectedto be large. Generally, these inconsistencies in fabrica-tion may be regarded as a clear shortfall when comparedwith the precision achievable from silicon microfabrica-tion [17, 21, 26]; however, the tip curvature ranges arecomparable with other porous sources[5, 16, 27].

3.2. Final Assembly and Wetting

The demonstration thruster assembly used in thisstudy is summarized in Figure 5. Each emitter layer wasstacked with a second porous borosilicate reservoir layerof larger, ∼ 16-40 μm, pores (Duran Group P3 grade).These two layers are interfaced using a laser cut disc ofWhatman Qualitative # 1 filter paper; ensuring a con-tinuous porous path between both layers. The emitterlayer is supported within a PEEK mount by means ofan aluminum guard ring. This flange serves two addi-tional purposes: i) it is used as the electrical contact pointwith the IL and, ii) the guard ring suppresses the electricfield strength at the edges of the disc, preventing spuri-ous emissions into the extractor grid. The PEEK upperand lower mounts, and thereby the two porous discs, arethen retained in contact with a mild pressure via springs.External electrical contacts to the emitter layer, via theguard ring, and the extractor grid are achieved throughPogopins soldered to a small PCB mounted on the topof the assembly. Each device was wet with IL in a twopart process prior to alignment and bonding with an ex-tractor grid. First, 90 μL of ionic liquid, which had beenpreviously degassed for at least 1 hour under a <10−5

mbar vacuum, was added to the the front of the device tocompletely fill the emitter layer. The assembly was thenplaced under vacuum for a further two hours to promote

7

FIG. 4: Multifocal microscopic image of SRC-Im-2 indicat-ing the typical degree of apex undulation resulting from themilling process.

filling. Finally, an additional 20 to 30 μL was added tothe emitter surface in order to ensure complete satura-tion of the emitting layer with all excess liquid retainedby the reservoir and interface layers. Liquid transportduring operation was entirely passive, no external flowconnections or feeding mechanisms were present.The extractor grids comprise 350 μm wide, 9 mm long

channels laser cut from sheets of 100 μm thick molybde-num. The grids were aligned using an optical microscopeand held in place using layers of double sided kaptontape. This grid placement method resulted in a degreeof variability in the grid position between devices. Hori-zontally, the alignment was typically within a few 10s ofμm of the strip center along the length of emitter strip.However, the spacing between emitters and the grid, arelatively pertinent dimension in the context of the elec-tric field at a given voltage, varied by 10’s of μm, seetable I. In the table s is the distance between the emit-ter edge and the base of the extractor, as indicated inFig. 1. This preliminary design targeted a separation sof zero, such that the emitting edge and extractor platewere inline; in an effort to ensure moderate starting po-tentials and low grid interception. However; referring tothe table, the emitting edges tended to protrude into, butnot through, the extractor grid for both sources emittingEMI-Im and were 105 μm below the extractor on averagefor SRC-BF4.The minimum voltage to start emission can be esti-

mated through recognizing that the applied electrostaticpressure must overcome the liquid surface tension[38, 42]and, particularly relevant in porous devices[2], the neg-ative Laplace pressure due to the porous reservoir. Thecited relations indicate a minimum voltage which scaleswith the liquid surface tension γ and length scales fand rp for the emitter field enhancement and capillarity

length scales respectively; Vstart ∼ f√

πγ2rp

[42]. Given the

transverse undulations and range of curvature radii re-sulting from this process and the potential for augmentedfield enhancement due to the liquid shape within the di-electric structure[29], accurate predictions by this rela-tion were not expected. Nonetheless, 2D (planar) FEM

1) Lower mount

2) Upper mount

3) Reservoir layer

4) Interface layer

5) Emitter layer

6) Extractor grid

7) Guard ring

8) Connection PCB

9) Retention springs

Section A-A

A

A

1

2

3

4

5

6

6

7

7

8

9

8

FIG. 5: Constituent components of the demonstrationthrusters tested here.

TABLE I: Device Dimensions and Ionic Liquid Properties

Device h (μm) s (μm) γ (mN/m) V ∗start (V )

± ∼20 μm [43] @Rc = 20μm

SRC-Im-1 320 -20 ± 20 34.9 ∼1550

SRC-Im-2 315 -50 ± 35 34.9 ∼1550

SRC-BF4-1 285 105 ± 30 45.2 ∼2100

*Predicted

simulations of the electric field were used to predict ffor a 20 μm representative radius of curvature and thegrid seperations/emitter heights listed in table I. As in-dicated, emissions were broadly anticipated to initiate inthe vicinity of 1550 V and 2100 V for the devices wetwith EMI-Im and EMI-BF4 respectively.

The assembled devices measured 3.7 cm x 3.7 cm x 0.8cm and, once wet, had a typical mass of ∼ 16 g. With ILretained by the porous reservoir, the devices can be eas-ily handled and transported once wet; permitting testsat multiple facilities without disturbing the liquid state.The outer device dimensions could be reduced in the fu-ture to more efficiently enclose the ∼ 1 cm2 of emittingarea. In particular, it should be noted that the PEEKmounting structure represented 10 g of the total massand the design favoured accommodation within the testfacilities over size.

4. EXPERIMENTAL APPARATUS ANDMETHODS

Measurements were acquired at both the ESA Propul-sion Laboratory (EPL) within the European Space Re-search and Technology Centre (ESTEC), Noordwijk, theNetherlands and the Microsystems for Space Technolo-

8

gies Laboratory (LMTS) at EPFL. Direct thrust mea-surements were made first, at ESTEC, followed by probemeasurements at the LMTS.

4.1. Direct Thrust Measurement

The low mass and lack of fluidic connections made thedemonstration thrusters suitable for DTM via a preci-sion commercial mass balance modified for use within avacuum facility. At the ESTEC EPL, A Mettler ToledoXP2004-S balance was installed within a 0.8 m diameter,0.5 m tall cylindrical facility pumped by Pfeiffer TMH261 turbo molecular pump with a Pfeiffer MVP 055-3diaphragm forepump. With all components installed, in-cluding the wetted thruster, the base pressure was ∼ 3x 10−6 mbar; however, during operation at the highestcurrents observed for SRC-Im-1, the recorded pressuresurged to up to 9 x 10−5 mbar. Prior to testing SRC-Im-2 and SRC-BF4 an additional TMH 261 turbo pumpstation was added in parallel to improve the extractionrate. The peak pressure was subsequently < 6 x 10−5

mbar when emitting.

Balance

Beam

Thruster

To turbomolecular

pump stations

V+em

V-emI

em

Iex

DAQ

Force

actuator

Vacuum chamber

Strain

relief

Computer

FIG. 6: Overview of the test configuration at ESTEC. Thethruster was positioned to emit upwards. No fluidic connec-tions were made to the pre-wetted and self-contained device.

The thrusters were positioned as indicated in Figure6; mounted to emit upwards. Propellant wetting wasperformed, as described in section 3.2, prior to instal-lation in the facility. Hence only the emitter and ex-tractor electrical connections were required, no fluid lineswere present. The connections to high voltage vacuumfeedthroughs were made via a strain relief block to reducedisturbances. The emitter voltage was supplied via FUGHCN 140-12500 and FUG HCN 35M-20000 high voltagepower supplies for the positive and negative states re-spectively. The emitter polarity was controlled via a cus-tom switching box controlled via a National InstrumentsUSB-DAQ board.The emitted and extractor (intercepted) currents were

measured with optically-isolated current monitors, accu-rate to within ∼ 3 and 2 μA for the emitter and extractor

monitors respectively. As shown explicitly in Fig. 6 eachmonitor included an inline high-power 100 kΩ resistoralong the current path to protect the monitors againstshort circuits. The voltage drop along these resistors hasbeen calculated and accounted for in all data presented.That is, reported energy deficits ΔV are in excess of thisknown Ohmic drop.

The balance specified repeatability was ∼ 1 μN (0.1mg). In practice, see section 5, the standard deviationin balance output was typically a ∼ 2-3 μN with verylow frequency (10’s to 100’s of mHz) random undula-tions of similar order. The balance stability was thereforea significant source of measurement uncertainty duringdirect thrust measurements. Each thrust measurementconsisted of the difference between an average over, atleast, 30 s of thruster operation and the average of 30 smeasurements preceding and following the measurement.The first 7 s of thruster activity were discarded to allowthe balance and beam emissions to settle.

The balance consistency was evaluated periodicallythroughout the test using an electromagnetic forceactuator. Here, a Keithley 2440 sourcemeter was usedto drive prescribed currents through a coil to impart aconsistent force, from 10 to 1000 μN , while monitoringthe balance output. Beginning with the unloadedbalance in atmosphere, the slope of the balance outputversus driving current and the drift over a ∼ 20 minuteperiod were recorded sequentially as a dry thruster wasinstalled, the chamber was evacuated and, finally highvoltage (1500 V ) was applied. Once a stable configura-tion was reached, the sensitivity curve remained constantto within the standard error (∼ 2 %) while the change inaveraged (over 30 s) output was typically < 5 μN overthe ∼ 20 min process. This feedback consistency wasconfirmed to persist before and after each thruster test.

Potential Alternation Scheme

Emissions from ILIS are known to decay over time[44],particulary at constant emitter polarity. These reduc-tions can, in part, be attributed to electrochemical break-down of IL at the liquid-conductor interface[45]. To sup-press these reactions the potential is typically alternatedto prevent sufficient charge accumulation at the inter-face for charge transfer to occur[33, 44]. Further workhas indicated the efficacy of this method may be aug-mented when using a distal electrode[46], a scheme in-trinsically implemented when using porous glass emittersas done here. The efficacy of potential alternation alonehas been debated indicating that, at least, efforts shouldbe made to balance the total charge emissions over eachperiod[33].

In the context of our goal of comparing direct and in-direct thrust measurements, potential alternation, repre-sents an added complexity due to the possible imbalanceof thrust levels between polarities. Preliminary attemptsto measure thrust in a unipolar (no alternation) config-uration were unreliable as the emission current tendeddecayed dramatically, by over 40 % at times, and did

9

TABLE II: Direct Thrust Measurement Setpoints

Device ID ∼ κ+

κ− α(%)

SRC-Im-1 0.53 34-36

SRC-Im-2 0.77 43-45

SRC-BF4 1.0 50

not tend to fully recover even after subsequent operationwith alternation (in contrast to Ref. [37]). Hence, a po-tential alternation scheme has been adopted. In keepingwith the literature, and in lieu of accurate knowledge ofthe liquid-metal contact area of our device, the poten-tial was alternated at 1 Hz during thrust measurements.Considering the few s response time and high amplitudenoise inherent with the balance, resolution of the thrustwithin each period was not possible. To reduce effectsdue to the unknown balance response when subjected toa pulsed thrust, the emitter voltage in each polarity wasselected in attempt to achieve equal thrusts at any mo-ment. The alternation duty cycle, α was then selected totarget balanced total charge emission over each period.Referring to equation 13, in so far as beam spread-

ing and fractional energy deficit may be approximated asconstant, the thrust from each polarity will be propor-tional to κ|Ib|

√|Vem|. Despite decays, a few (∼ 10) con-stant voltage thrust measurements were made for eachdevice so as to estimate the ratio of proportionality fac-tors κ between polarities, κ+/κ−. These ratios are indi-cated in table II and are discussed, with the hindsight ofthe subsequent indirect thrust measurements in section 6.Measurements of current versus voltage, while alternat-ing the potential, were then used to predict the expectedcurrent imbalance at equal thrusts; thereby enabling se-lection of the duty cycle such that αI+em+(1−α)I−em ≈ 0.For device SRC-BF4, the predicted ratio of thrusts andmeasured current-voltage curves corresponded to wellmatched positive and negative outputs, leading to se-lection of simply matched voltages and a 50 % duty cy-cle. This result is in keeping with previous studies wherevoltage alternation alone, without other means to ensurebalanced charge, was shown to suppress electrochemicalreactions with EMI-BF4[44].

4.2. Indirect Thrust Measurement

The LMTS facility consists of a 6-way, ISO160 crosscombined with an 0.8 m ISO100 extension serving as theflight tube and evacuated by a Varian V70 turbomolecu-lar pump. The base pressure was < 3x10−6 mbar, risingto ∼ 6x10−6 mbar when the emission spray was active.Figure 7 provides an overview of facility as configured forToF and beam angle measurements.Emitter and extractor currents were monitored using

the same optically-isolated transducers used at ESTEC.Similarly the emission polarity was alternated using the

ThrusterGate

Detector

Turbomolecular

pump system

Beam

74.6 cm

Rotation

stage

40VVacuum

chamber

V+em

V-em

Vg

DSO

Pulse

generator

Ic

Iem

Iex

DAQ

FIG. 7: Schematic of the test configuration used at LMTS asconfigured for ToF measurements. An aperture and baffle im-mediately upstream of the gate permitted a ± 1.4o half-anglebeamlet to reach the detector assembly. The RPA assembly(not shown) was inserted and positioned 71 mm from thesource when utilized.

aforementioned switching box now controlled via an Ag-ilent 33220a arbitrary function generator. Stanford Re-search Systems P350 high voltage power supplies wereused to control the electrostatic gate potential along withthe positive and negative emitter currents.

The ToF spectrometer was positioned 13 cm from thesource and comprised a gate assembly, flight tube anddetector assembly. The flight distance from the gate elec-trode to the detector assembly was 74.6 cm. The detectorassembly comprised two 81 % transparency stainless steelmeshes upstream of a stainless steel plate, with the cen-tral mesh biased to -40 V to suppress secondary electronemission from the detector plate. The detection area was11.3 cm2.

The electrostatic gate consisted of three 81 % trans-parency stainless steel planar meshes. The outer meshelectrodes were held at ground while the inner electrodewas switched from ± 3000 V to ground at 25 Hz using afast PVX-4140 high voltage switch. The detector currentwas converted to a voltage using a FEMTO DLPCA-100high current amplifier, set to a gain of 106 V/A and 1MHz bandwidth. The amplified output was recorded onan LeCroy WaveSurfer 424 digital oscilloscope with ∼ 30trace averaging enabled to improve the recorded signalto noise ratio.

The ToF flight tube was accessed through a 6.3 mmdiameter aperture, corresponding to a narrow ± 1.4o ac-ceptance angle. Despite the clear drawback of reducedsignal intensity, the narrow acceptance angle had twobenefits. Primarily, we have observed that at the 100’s ofμA beam currents typical from these devices, secondaryelectron emission from impinged surfaces within the flighttube can become a major source of error. Spurious cur-rent signals, at times including offset currents of compa-

10

rable order to the primary signal were typical until accessto the flight tube was restricted. The aperture geometrywas selected such that, geometrically, all particles passinginto the ToF flight tube will reach the detector; that is,none are anticipated to strike the inner walls of the ToFtube upstream nor beside the collector assembly. As partof a larger baffle structure, the aperture also inhibits sec-ondary electrons born near the flight tube entrance fromreaching the high voltage gate electrode. With this sys-tem, and the aforementioned -40 V screen grid upstreamthe detector in place, relatively clean ToF traces with low(few nA) offsets were achieved.

The thrusters were mounted on a manual rotationstage enabling angle resolved beam current and ToF mea-surements over ± 50o about the central emitter strip.Beam profiles over angle were made using the ToF as-sembly detector with the gate electrode held to ground.The narrow acceptance angle of the ToF flight tube re-duced corruption of the data due to the finite extent ofthe sources. Specifically, referring to equation 12, weseek an effective angle representative of all beamlets re-gardless of their spatial distribution within the plane ofemitting structures. The spatial spread due to long slitscould present system challenges in managing the exhaustplume, but would not alter the thrust output in so faras all beamlets are considered equal. Using the apertureconfiguration described, the detected current contribu-tions due to a beamlet located at the maximum extentfrom the center of the thruster, ∼4.7 mm, were within∼1o of the stage angle. However; the collection solid an-gle for such a beamlet, as estimated using simulationswith the commercial software Simion 8.1, was signifi-cantly reduced compared to emissions from the center ofthe device. Referring to section 2, integration of the mea-sured current density versus angle should correspond tocomplete collection of the emitted beam. However; dueto this reduction in transmission for distributed beamletswe have calculated, through the aforementioned simula-tions, that no more than 60 % and as little as 40 % of theemitted beam may be detected. It is important to clarifythat this reduction is geometric and should not, in so faras emission characteristics are uniform over the source,correspond to a possible ambiguity in interpreting theToF spectra.

ToF and beam angle measurements have been compli-mented by limited Retarding Potential Analyzer (RPA)measurements. The RPA primarily consisted of 7,stainless steel mesh grid electrodes inline with a Kim-ball Physics FC-72 Faraday cup detector. In keepingwith previous high resolution energy measurements fromILIS[34, 36] the RPA design was based on the recom-mendations of Ref. [47]. Specifically, to ensure smoothpotential gradients the retarding potential is distributedover 5 grids wherein the 3 central grids are at the retard-ing potential Vr while the outer grids are maintained at96 % of Vr. A 9.5 mm diameter limiting aperture on theRPA’s Faraday cup collector was positioned 71 mm fromthe source.

5. RESULTS

5.1. Direct Thrust Measurements

Emitted currents (Iem) versus the applied voltage|Vem| (I-V data) for each of the tested devices are shownin Figures 8 and 9. All three devices were shown toemit several hundred μA at potentials consistent with therough starting potential estimates in table I. These datawere acquired with the potential polarity alternating at 1Hz and with a 50 % duty cycle. Each datum representsthe mean and standard deviation (errorbars) of data col-lected at each polarity over a 3 s period. The figureinsets provide representative snapshots of the raw emit-ted current measured at one setpoint. Short transientswhen switching were discarded. For both devices oper-ated with EMI-Im, negative emission currents were lowerthan positive currents at a particular |Vem| while emis-sion from EMI-BF4 was relatively balanced. As shownin the Figure insets, negative polarity emitted currentstended to increase gradually across each half period forall devices, contributing to the larger standard deviationsindicated in the primary figures. These I-V relationshipswere applied to establish the duty cycle setpoints in tableII necessary to approach balanced total charge emissionin tandem with equal positive and negative thrusts.

Typically a few percent of the beam was interceptedby the extraction grid, as indicated by Iex in the figures.Interception was greatest for SRC-Im-2, Fig. 8(b) whereup to 10 % of the beam was intercepted. These data arerepresentative of the current-voltage behaviour immedi-ately prior to DTMs. However; irreversible changes in theemission and interception currents did occur. In partic-ular, the emission current tended to decrease graduallyover the test campaign, particularly after returning toLMTS; while the current intercepted by the grid tendedto decrease. The decays are discussed in more detail insection 6.3.

Raw current data and the accompanying thrust outputacquired during a typical thrust measurement are shownin Figure 10, from SRC-Im-1. Figure 10(a) demonstrateshow lower current, yet longer pulses per period in the neg-ative mode were applied as per the open loop thrust andcurrent balancing scheme. The emission current clearlydecayed throughout each thrust measurement, predom-inantly over the first few seconds. Horizontal lines in-dicate the region, after a 7 s delay, sampled to gener-ate representative average positive and negative currents(solid lines) and standard deviation (dashed lines) for themeasurement.

In Figure 10(b) the emitter supply was off prior to andafter measuring thrust, providing reference zero levels.This example demonstrates both the few μN instabili-ties typical at short (few s) timescales along with a largerjump in the mean off state output before and after emis-sion. These pre- and post-measurement zero levels wereoften in agreement however, as in the presented example,a discrepancy approaching ∼ 5 μN was not uncommon.

11

1000 1200 1400 1600 1800 2000 2200

−400

−300

−200

−100

0

100

200

300

400

500

|Vem

| (V)

Cu

rre

nt (μ

A)

0 1 2−500

0

500

Time (s)

I em (

μA

)

Vem

=±2200 V

+ Iem

+ Iex

− Iem

− Iex

(a)SRC-Im-1

600 800 1000 1200 1400 1600 1800 2000 2200

−400

−300

−200

−100

0

100

200

300

400

500

|Vem

| (V)

Curr

ent (μ

A)

0 1 2−500

0

500

Time (s)

I em (

μA

)

V =±2100 V

+ Iem

+ Iex

− Iem

− Iex

em

(b)SRC-Im-2

FIG. 8: Current versus voltage (I-V) characteristics for thetwo sources emitting the IL EMI-Im, immediately prior todirect thrust measurements.

To arrive at a single measurement, the average balanceoutput, after a 7 s settling period, was compared with theaverage balance output for 30 s prior to and 30 s afterthe pulse, as indicated in the figures. Solid and dashedlines again represent the calculated means and standarddeviations respectively.In Figure 11, the thrust output from SRC-Im-2 has

been plotted along with the quantity⟨Ib√|Vem|

⟩=

αI+b

√V +em+(1− α) I−b

√|V −

em| recorded during each of a

sequence of 10 repeated measurements at Vem = +1997V / -2007 V . Despite the consistent emitter voltages,

the beam current, and therefore⟨Ib√|Vem|

⟩, differed be-

tween repeated trials; tending to decrease. Accordinglyand consistent with equation 13, the thrust also tendedto decrease supporting correlation with this parameter,

800 1000 1200 1400 1600 1800 2000 2200 2400 2600−300

−200

−100

0

100

200

300

|Vem

| (V)

Cu

rre

nt (μ

A)

0 1 2

−300

0

300

Time (s)

I em (

μA

)

Vem

=±2500 V

+ Iem

+ Iex

− Iem

− Iex

FIG. 9: Current versus voltage (I-V) characteristic for SRC-BF4, immediately prior to direct thrust measurements.

see the figure inset. In section 6, we therefore compareall direct and indirect thrust measurements as a functionof

⟨Ib√|Vem|

⟩.

Scans of thrust measurements were made at set pointsinterpolated, based on the I-V measurements, to yield100 to 400 μA of positive emitter current. Measurementsat these set points were repeated several (3-10) times foreach device, with the negative voltage and duty cycleestablished through table II. The summarized results,demonstrating 10’s of μN at less than 1 W of power forall three devices, are presented later (Figures 16 and 17)in comparison with ITM.

5.2. Indirect Thrust Measurements

The three devices were transported from theESA/ESTEC EPL facilities back to the LMTS at EPFLfor beam diagnostics. No modifications were made toany device, including to the liquid state. Specifically, noliquid was added or removed between tests. The deviceswere stored under vacuum whenever possible.

5.2.1. Time of Flight Spectrometry

ToF traces were acquired from each of the three testdevices forming a basis for propulsive metric calculations.As described previously, the measurement setup did notpermit ToF measurements while simultaneously alternat-ing the potential. Instead all ToF measurements weremade while emitting at constant polarity for several sec-onds.

Each trace was post-processed prior to determining theparameters κ and φ by equations 8 and 9. First the raw

12

440 460 480 500 520 540 560 580−200

−100

0

100

200

300

400

500

Time (s)

I em (

μA

)

Thruster OFF ON OFF

517 518 519−200

0

200

400

Time(s)

I em (

μA

)

(a)Current

440 460 480 500 520 540 560 580

−5

0

5

10

15

20

25

Time (s)

Th

rust (μ

N)

20 ± 3 μN

OFF OFFONThruster

(b)Thrust

FIG. 10: Example of a typical current and correspondingthrust pulse from SRC-Im-1. The thrust measurement hasbeen taken relative to the average of two zero level measure-ments recorded before and after the pulse. Solid, horizontallines indicate the mean and standard deviation of data re-tained as representative of each region.

data was filtered using a Savitzky-Golay filter, as hasbeen previously applied to ToF processing[20]. Any off-set in output from the current amplifier, typically 1-2nA, was then subtracted. Although modified by activat-ing the source, this residual current was deemed spuriousas it was confirmed to persist over the full 20 ms half-period of the gate signal. Finally the current trace wasnormalized by an average inlet current prior to the thefirst species flight time.

The mean shift in beam energy between the mea-sured and expected location of ion peaks was determinedthrough a gaussian fit to the derivative of the trace nearthe ion flight times anticipated by equation 1 for then = 0 and n = 1 species. The edge was then taken asbeing 0.22 μs from each peak, corresponding to half theexpected decay time of the amplifier from 95 to 5 % at 1

0 2 4 6 8 1030

35

40

45

50

Trial #

Th

rust (μ

N)

0 2 4 6 8 10

10

11

12

13

11 12 1330

35

40

45

〈IbV

em 〉1/2 (AV1/2x 10-3)

Thru

st (μ

N)

〈I b

Vem

〉

1/2

(A

V1/2x 1

0-3)

FIG. 11: Repeated thrust measurements at +1997 V and−2001 V and a 43 % duty cycle from device EPFL4. The

thruster output as reflected by both the quantity⟨Ib√|Vem|

⟩

and the measured thrust tended to decrease over the sequenceof measurements.

MHz bandwidth. Voltage deficits ΔVToF for n = 0 andn = 1 species were taken as the difference between Vem,after correction for the inline sensing circuit resistance,and the equivalent voltage consistent with each edge byequation 1. The average of these two deficits was thentaken to estimate Va and used to calculate the flight timeexpected for ions of mass mEMI, enabling a normalized tto be calculated and applied in equations 8 and 9 for κand φ respectively.

ToF measurements were made at ± 10o to 30o of ro-tation in addition to aligned 0o measurements. However;as detailed in section 6.4, trends versus neither angle norvoltage were determined to a level of confidence clearlybeyond the repeatability at consistent setpoints. Insteadthe derived metrics presented here are an average over allmeasurement conditions (voltage, beam angle) per polar-ity.

Examples of ToF measurements for each device, at zerodegree rotation, are given in Figures 12(a) though 12(b).The traces are presented as a function of the calculatedspecies mass after correcting for energy deficits and as-suming singularly charged ions. Vertical lines indicate lo-cations of the known n = 0, n = 1 and n = 2 ion masses;showing good agreement in the first two cases. Highlyionic emissions were achieved in all cases with most cur-rent corresponding to n = 0 and n = 1 species; however,a high n tail extending to several kDa was particularlyevident at negative Vem from EMI-Im in Figs 12(a) and12(b). The current contributions due to each ion species,taken as the change in fractional current between eachcalculated drop location, are given in table III.

Average voltage deficits for the n = 0 and n = 1 peaksare also provided in table III. In all instances, a higherΔV was found for the n = 0 ion compared with n = 1,

13

102 103 104

−60

−40

−20

0

20

40

60

Mass (Da)

I c (

nA

)[EMI−Im]

n EMI+

n=2

n=0

n=0

[EMI−Im]n Im-

n=2

n=1

n=1|Vem

|Vem

|

|

(a)SRC-Im-1: |Vem| increasing from ± 2000 V to 2300 V in100 V steps.

102 103 104

−60

−40

−20

0

20

40

Mass (Da)

I c (

nA

)

[EMI−Im]n EMI+

n=2

n=0

n=0

[EMI−Im]n Im-

n=2

n=1

n=1

|Vem

|

|Vem

|

(b)SRC-Im-2: |Vem| increasing from ± 1800 V to 2200 V in100 V steps.

FIG. 12: ToF traces when emitting EMI-Im, a tail of largemass to charge ratio species was particularly evident for nega-tive polarity emissions. Sloped signals between primary dropsare indications of fragmentation within the emitter to extrac-tor gap.

and lower deficits were observed at negative polarities forall devices. The fractional deficit of EMI-Im devices was,on average, less than 3 % while for SRC-BF4 larger re-ductions, in excess of 7 % were observed. These lossesare relatively large compared with published results in-dicating only a few V of deficit[34, 36]. In table IV theaverage fractional n = 0 and n = 1 ΔV are listed for eachpolarity facilitating inclusion in equations 13 through 15.

The non-zero slopes between discrete ion masses can beattributed to fragmentation while accelerating betweenthe emitter to extractor grid, where solvated (n ≥ 1)ions breakup into an ion and neutral component prior tocomplete acceleration[34, 48]. In full beam ToF where alarge detector collects the entire beam, similar drops just

102 103 104

−60

−40

−20

0

20

40

60

Mass (Da)

I c (

nA

)

n=0

[EMI−BF4]n EMI+

n=2

n=0

[EMI−BF4]n BF

4-

n=2

n=1

n=1

|Vem

|

|Vem

|

FIG. 13: SRC-BF4: |Vem| increasing from ± 2100 V to 2400V in 100 V steps. Compared with Figure 12, a relatively smallpopulation of high mass to charge ratio species was observed.

after each nominal flight time may manifest due to thespread in axial velocity[48]; however, here we only samplea small portion of the beam passing within the 1.4o halfangle acceptance cone of the limiting aperture. Moleculardynamics simulations of fragmentation events[49] haveindicated that breakup can be induced by both the ion’sthermal energy and an applied electric field; and thatthe degree to which each process is dominant can differbetween ion solvation number. Slopes in the ToF tracesbetween the n = 1 to n = 2 expected masses, partic-ularly for cations in Fig. 12, are relatively consistentdespite increasing emitter voltages. If ion breakup weredictated solely by a fixed time constant, a reduction inthe degree of fragmentation would be expected at highervoltages as ions are more quickly swept out of the ac-celeration region. However; a corresponding increase infragmentation due to higher external fields may counterthis effect. Meanwhile; the relatively shallow slopes im-mediately after discrete steps in the trace, particularlyn = 0, indicate ions have moved beyond the immediatevicinity of the emission site, where fields are strongest,before breaking up.

While these observations thus imply further investiga-tion is needed to describe ion fragmentation more com-pletely, RPA measurements presented in section 5.2.3confirm kinetic energy deficits over a range consistentwith this phenomenon. We have therefore applied frag-mentation correction factors, equations 4 and 5, at-tributing non-zero slopes to fragmentation events alone;n = 2 → 1 for t1 < t < t2 and n = 1 → 0 for t0 < t < t1.The aggregate influence of these corrections (increases)on the metrics κ and φ have tended to be small, up to∼ 3 % and 5 % for the thrust and mass flow rate respec-tively. However; the presented values have nonethelessbeen adjusted to include compensation for fragmentationfor completeness.

14

TABLE III: Beam Composition and Energy Deficits*

Device Ion n fn (%) ΔVToF(V ) ΔVToFVem

∗(%)

SRC-Im-1 EMI+ [EMI-Im]n 0 40 ± 3 55 ± 20 2.7 ± 0.8

1 47 ± 3 30 ± 15 1.6 ± 0.7

2 9.5 ± 2 - -

≥3 4 ± 1.5 - -

SRC-Im-1 Im− [EMI-Im]n 0 40 ± 3 26 ± 4 1.2 ± 0.1

1 40 ± 2 13 ± 3 0.6 ± 0.1

2 10 ± 2 - -

≥3 10 ± 2 - -

SRC-Im-2 EMI+ [EMI-Im]n 0 44 ± 3 70 ± 45 2.3± 0.6

1 43 ± 3 55 ± 40 1.3 ± 0.5

2 9.5 ± 3 - -

≥3 3.5 ± 2 - -

SRC-Im-2 Im− [EMI-Im]n 0 40 ± 4 23 ± 4 1.2± 0.2

1 40 ± 4 14 ± 4 0.7± 0.1

2 9 ± 3 - -

≥3 10 ± 3 - -

SRC-BF4 EMI+ [EMI-BF4]n 0 42 ± 2 160 ± 25 7.2 ± 1.1

1 47 ± 3 140 ± 15 6.2 ± 0.7

2 8.5 ± 2 - -

≥3 3.5 ± 2 - -

SRC-BF4 BF−4 [EMI-BF4]n 0 44 ± 2 97 ± 9 4.2 ± 0.4

1 44 ± 3 61 ± 6 2.7 ± 0.3

2 8.5 ± 2 - -

≥3 4.5 ± 1.5 - -

*Uncertainties indicate the standard deviation over 19 or more measurements

Average propulsive parameters κ, φ, and the resultantpolydispersive efficiency (κ2/φ) are given in table IV. Ofparticular relevance to our goal of comparing direct andindirect thrusts, the standard deviation of the param-eter κ was, at most, ∼ 7 % despite the inelegance ofaveraging values over all voltage and beam angles mea-surement conditions. Relatively large fluctuations in φcontributed to the large propagated errors in polydisper-sive efficiency reported. Regardless, the comparativelylarge polydispersive efficiencies measured when emittingEMI-BF4 are consistent with emissions within or closeto the PIR[7, 16, 41]. Conversely the larger spread inconstituent masses and possible droplet tail, after re-moving offsets, observed from both EMI-Im sources con-tributed to lower efficiency. For those devices, the neg-ative polarity mass flow rate, proportional to the factorφ, was roughly double that at positive potential. Withinthe standard deviations quoted, all parameters measuredfrom both devices emitting EMI-Im were consistent.

5.2.2. Angular Current Distribution

Figure 14 provides examples of the collected current toemitted current fraction versus thruster angle with theaxis of rotation orientated parallel and perpendicular tothe emission strips. Parallel rotations were recorded from

all devices, while perpendicular rotations were only mea-sured from SRC-Im-1 and SRC-Im-2. The prevalence ofmultiple Taylor cones (emission sites) per emitter struc-ture in porous ILIS in [24] has driven the development ofthe linear emission strip geometry. However; that samework found that emission sites are expected wherever thefield is strong enough and may, as a result, not be wellaligned axially. In this context, the reasonably well col-limated beams shown are a welcome confirmation thatconventional machining can produce sufficiently sharpstructures to avoid wide beam angles.

For all devices, more current relative to that emittedwas collected in the negative polarity versus positive. Asshown by the indicated insets in Figures 14(a) and 14(b),the two polarities nevertheless are in agreement whenscaled by their respective maxima. Considering the five81 % geometric transparency grids between the sourceand detector and the finite source dimensions, the neg-ative current was typically consistent with sampling theentire beam with a source misalignment of a few mm.The reduction in positive current may be attributed toneutralization by secondary electrons born from the im-pact of ions on the baffles and aperture plate surroundingthe narrow ToF tube inlet aperture.

An axisymmetric approximation was made from eachset of data bi-directional rotation data, facilitating useof the right hand side of equation 12. This process in-

15

TABLE IV: Summary of Indirect Propulsive Parameters*

Device Polarity κ φ κ2

φ(%)

⟨ΔVToFVem

⟩(%)

⟨ΔVRPAVem

⟩(%) θ0 θc

SRC-Im-1 + 1.90 ± 0.06 5.1 ± 0.7 70 ± 11 2.1 ± 1.0 2.2 ± 0.9 ‖ 0.1o ‖ 21.3o

⊥ 2.0o ⊥ 20.0o

− 2.60 ± 0.18 9.8 ± 2.4 69 ± 19 0.9 ± 0.2 - ‖ 0.2o ‖ 22.4o

⊥ 2.8o ⊥ 21.2o

SRC-Im-2 + 1.83 ± 0.06 4.8 ± 0.7 70 ± 12 1.8 ± 0.8 3.0 ± 0.8 ‖ 4.5o ‖ 21.3o

⊥ 0.5o ⊥ 21.7o

− 2.71 ± 0.19 11.7 ± 2.7 63 ± 17 0.9 ± 0.2 - ‖ 5.7o ‖ 21.6o

⊥ 1.7o ⊥ 22.0o

SRC-BF4 + 1.48 ± 0.02 2.6 ± 0.3 83 ± 9 6.7 ± 1.3 - ‖ 5.5o ‖ 20.8o

− 1.45 ± 0.03 2.7 ± 0.3 77 ± 9 3.5 ± 0.4 - ‖ 3.4o ‖ 20.2o

*Uncertainties indicate the standard deviation over 19 or more measurements

cluded determining an offset angle (θ0), taken as that an-gle about which axisymmetric surface integrations usingrelatively positive or negative rotation angles are equal.These two halves were then averaged to yield a single rep-resentative profile for each orientation and/or voltage.Given the inconsistency between the integrated and ex-

pected beam currents, axisymmetric approximations foreach device/polarity have been normalized such that theintegration of j(θ) over S is unity. The effective beamangle has then been calculated by equation 12, see ta-ble IV. In the table, ‖ and ⊥ indicated rotations paral-lel and perpendicular to the emitter strips respectively.Measurements repeated at differing emitter voltages wereconsistent to within ∼ 2o. The effective cone angle θeffwas close to 21o for all cases, corresponding to a lossof ∼ 7 % in thrust output compared with a perfectlycollimated beam. The consistency in θeff between rota-tion orientations, despite a more narrow profile in Fig-ure 14(b) versus 14(a), reflects the diminishing returns,in terms of propulsive performance, achieved when colli-mating the central portion of the beam. Rather the smallcurrent contributions beyond ∼ 35o, in both orientations,are a relatively significant inefficiency driver.

5.2.3. Energy Spectra

RPA data were limited to constant positive Vem andzero rotation angle. The data, see Figure 15, confirma primary population of particles had energies close tothat of the emitters. However, a significant portion ofthe collected beam was, apparently, repelled at a lowerenergy. In Ref. [34] and [36] low energy beam repulsionwas attributed to downstream ion fragmentation. How-ever, downstream fragmentation manifests as a relativelydiscrete step/peak at a known energy corresponding tobreakup in a field free region. The expected peak loca-tions due to such events, are readily determined by scal-ing the initial particle energy qVem by the ratio of postto pre breakup ion mass[34]. Specifically, if a solvatedion of mass mn has kinetic energy qVem and breaks into

a charged particle of mass mn−1 and a neutral of massmn − mn−1 the charged particle (ion) will have energyKn−1 given by equation 16.

Kn−1 =mn−1

mnqVem (16)

Emitting positive EMI-Im ions, the post-fragmentationion energies will then be only 22% and 56% of the initialkinetic energy after n = 1 → 0 and n = 2 → 1 tran-sitions respectively. These energy levels are indicatedin Figure 15(a). Although the n = 1 → 0 marker didtend to indicate the lower limit of species energy, promi-nent peaks were not observed. In addition to breakupwithin the emitter to extractor accelerator region as dis-cussed in the context of ToF measurements above, thesemeasurements could therefore indicate that significantfragmentation occurred within the RPA’s decelerationfield. Compared with[34], the RPA setup here did notinclude an upstream Einzel lens and was therefore po-sitioned closer to the source. Ultimately, the physicalextent of the multiple retarding grids was roughly equalto the flight distance prior to the RPA. Hence, relativelylittle flight distance/time was available for breakup to oc-cur downstream of the extractor grid yet prior to enteringthe probe.

The peak energy deficit was found to be typicallywithin 20 to 80 V of Vem after correction for the cur-rent monitor resistance. These small deficits, a few per-cent of the emitter voltage, are in reasonable agreementwith the average deficits estimated by ToF, see table IV.However; they are, again, large compared with previousstudies of ILIS emitting EMI-Im. Specifically, Lozano[34]measured only a few (∼ 5-7 V ) volt deficits and FWHMof similar order. Here the FWHM was typically ∼ 10-40 V . The same instructions, Ref [47], were applied toour RPA and that of Lozano. Hence we anticipated asimilar high resolution probe however, it should be notedthat the resolution of our RPA was not rigorously verifiedwith a monoenergetic reference source.

16

−60 −40 −20 0 20 40 600

50

100

150

200

250

300

350

Stage Angle (o)

I c/I

em

(x

10

-6)

+2200V

+ Spline

−2200V

− Spline

−50 0 500

0.5

1

Stage Angle (o)

A.U

.

(a)Rotation parallel to the strip edges.

−60 −40 −20 0 20 40 600

50

100

150

200

250

300

350

Stage Angle (o)

I c/Ie

m (x

10

-6)

+2200V

+ Spline

−2200V

− Spline

−50 0 500

0.5

1

Stage Angle (o)

A.U

.

(b)Rotation perpendicular to the strip edges.

FIG. 14: Beam current fractions versus angle when rotatingon the indicated axes. The inset figures demonstrate that,when normalized to the peak collection ratio, the positiveand negative data are consistent.

6. DISCUSSION

6.1. Comparison of Direct and Indirect ThrustMeasurements

Direct (DTM) and indirect (ITM) thrust measure-ments are plotted for each of the three tested devicesin Figures 16 and 17. The ITM correspond to TITM =αT++(1−α)T− where the thrusts T± for each polaritywere determined through equation 13 using the indirectpropulsive parameters listed in table IV and the factorIb√|Vem| measured at each direct thrust measurement.

Averaged input powers are indicated in the figure insets.All devices generated 10’s of μN with less than 0.8 Wof input power. Maximum thrust levels of ∼ 50 μN and∼ 30 μN were measured when emitting the ILs EMI-Imand EMI-BF4 respectively. Higher thrusts from EMI-Imwere anticipated due to the higher anion mass.The directly measured and indirectly calculated

thrusts from SRC-Im-2 and SRC-BF4 were in good agree-ment for all but the highest emission levels. Specifically,

0 500 1000 1500 20000

0.2

0.4

0.6

0.8

1

Retarding Potential, Vr (V)

I C (

A.U

.)

Vem

=2000V

n=1 n=0

n=2 n=1[EMI−Im]

n EMI+

0 500 1000 1500 2000

0

0.5

1

Vr (V)

dI c

/dV

r (A

U)

ΔV ≈ 60 V

FWHM ≈ 35 V

(a)SRC-Im-1

1500 1700 1900 2100 2300

0

0.2

0.4

0.6

0.8

1

Retarding Potential, Vr (V)

dI c

/dV

r (A

.U.)

20 V 60 V 60 V 60 V 80 V

10 V 20 V 25 V 20 V 40 V FWHM ≈

ΔV ≈

(b)SRC-Im-2

FIG. 15: (a) Smoothed current data measured against theretarding potential from SRC-Im-1. Vertical lines indicatethe anticipated location of peaks associated with downstream(of the extractor) solvated ion breakup. (b) Distributions ofcurrent over retarding potential from SRC-Im-2. Peak en-ergy deficits, with respect to Vem and FWHMs were typicallyseveral 10’s of V .

the values agreed to within a few μN up to ∼ 40 μNand ∼ 20 μN for SRC-Im-2 and SRC-BF4 respectively,with deviations closer to 5 μN at higher thrust levels.Less consistent agreement was observed from SRC-Im-1where disparities approached 15 μN at high power.

The predicted thrusts tended to exceed that emitted athigh current. The composition, angle and energy thrustcoefficients have been approximated as constant averagesover all voltages and angles. We therefore only considerthe bulk standard deviations in κ listed in table IV, whichcorrespond to up to ∼ 3 μN , as providing bounds to thecalculations. Were more consistent ToF measurementsobtained, the divergence at high voltage may have beenreduced through parameterizing κ versus voltage. Re-gardless, lower than expected thrusts are supportive ofhighly ionic emissions since an undetected droplet popu-lation would have lead to a higher thrust.

Inadequate thrust balancing may have contributed to

17

2 4 6 8 10 12 14 160

5

10

15

20

25

30

35

40

45

50

Thru

st, T

(μ

N)

Direct (TDTM

)

Iindirect (TITM

)

4 8 12 160

0.2

0.4

0.6

Avg

. P

ow

er

(W)

〈IbV

em 〉1/2 (AV1/2x 10-3)

〈IbV

em 〉1/2

(AV1/2x 10-3)

(a)SRC-Im-1

2 4 6 8 10 12 14 16 180

10

20

30

40

50

60

Thru

st, T

(μ

N)

Direct (TDTM

)

Indirect (TITM

)

4 8 12 160

0.25

0.5

0.75

1

Avg

. P

ow

er

(W)

〈IbV

em 〉1/2 (AV1/2x 10-3)

〈IbV

em 〉1/2

(AV1/2x 10-3)

(b)SRC-Im-2

FIG. 16: Summary of the directly (DTM) and indirectly(ITM) measured thrusts from devices wet with EMI-Im. Theindirect thrust measurement standard error was ∼ ± 2-3 μN .Excellent agreement between direct and indirect measure-ments was obtained for SRC-Im-2 at most setpoints.

the disagreement in Figure 16(a) for SRC-Im-1. Refer-ring to table IV, the ratio of positive to negative κ param-eters calculated by ToF was approximately ∼ 0.7 for bothSRC-Im-1 and SRC-Im-2, and ∼ 1.0 for device SRC-BF4.The ratios applied to establish DTM setpoints, table II,were similar for SRC-Im-2 and SRC-BF4 yet SRC-Im-1may have been relatively poorly balanced. In all cases,force imbalances have been accounted for in TITM to a de-gree through the duty cycle weighted average described;however, the response of the balance to pulsations, asmay have been present particularly with SRC-Im-1, wasunknown.

2 4 6 8 10 12 14 160

5

10

15

20

25

30

35

Thru

st, T

(μ

N)

Direct (TDTM

)

Indirect (TITM

)

4 8 12 160

0.2

0.4

0.6

Avg

. P

ow

er

(W)

〈IbV

em 〉1/2

(AV1/2x 10-3)

〈IbV

em 〉1/2 (AV1/2x 10-3)

FIG. 17: Agreement between directly (DTM) and indirectly(ITM) measured thrusts from SRC-BF4. The indirect thrustmeasurement standard error was ∼ ± 1 μN .

6.2. Compiled Thruster Performance

The agreements between direct and indirect thrustmeasurements from the same devices, a unique scenariofor a passively fed and ionic electrospray thruster, per-mit total performance estimates. In table V we have in-terpolated the propulsive efficiency, specific impulse andthrust at 0.5 W of input power. The specific impulseand propulsive efficiency have been calculated by equa-tions 14 and 15, modified to account for potential alter-nation. For all devices, values in parentheses were ob-tained when using an interpolation of the directly mea-sured thrust; although the calculated mass flow was usedfor all values. In this form the data re-affirm the im-proved specific performance of EMI-BF4 over EMI-Im, atthe expense of reduced thrust, due to greater ion contentand lower anion mass associated with EMI-BF4 emis-sions. Given the comprehensive nature of the diagnosticmeasurements presented, and good agreement with di-rect measurements for two of three device, we are confi-dent that these performance metrics present an accurateportrayal of the capabilities available from such a device.

In interpreting these data, the implicit scalability ofILIS thrusters should be considered. The active area ofthese prototype devices was roughly 1 cm2; hence nu-merous sources could be accommodated within a smallarea thereby increasing the thrust. Indeed, to obtaincharge neutrality, the intended application would includeat least two sources operated simultaneously at oppositepolarities.

Referring to equation 15 and table IV it is clear thatpolydispersive efficiency represents the greatest ineffi-ciency within the system; consistent with the effortsto increase this parameter discussed in the introduc-tion. Beam spreading and interception by the extrac-

18

TABLE V: Thruster Performance Summary at 0.5 W

Device T (μN) Isp (s) ηprop. (%)

SRC-Im-1 33 (23) 1660 (1120) 53 (25)

SRC-Im-2 35 (34) 1440 (1410) 49 (47)

SRC-BF4 20 (20) 3260 (3290) 65 (66)

() indicate values calculated by interpolation of DTM.

tion grid were the next most significant contributors re-spectively; although for SRC-BF4 energy deficit lossesalso approached 7 %. Grid interception must be reducedthrough design improvements, not only to improve per-formance but to eliminate erosion as a lifetime limitation.Meanwhile beam spread over angle would be improved toa degree if a downstream secondary accelerating electrodewere included[26]; although the efficiency benefits wouldlikely be small. The higher energy deficits compared withthe literature should be further explored. The ill definedIL to aluminum contact made through the guard ringmay have contributed to this loss, although the deficitswere only weakly correlated with Vem (and therefore Iem)indicating loss mechanisms were not entirely Ohmic. De-spite these losses, when compared with other miniatur-ized ion propulsion systems[11] and considering the sys-tem benefits of completely passive (un-powered) propel-lant feeding, the high specific impulses and ∼ 50 % to 65% power efficiencies achieved with these prototypes areencouraging.

6.3. Emission Decay and Longevity

This initial study has been intended to confirm the effi-cacy of indirect thrust measurements and to demonstratethat a simple fabrication process can be used to formhigh current porous emission sources. Many design im-provements are required to transition from these demon-stration thrusters to engineering breadboards suitable forlong duration tests and precise thrust resolution mea-surements. Decays in emission current over the course of

600 800 1000 1200 1400 1600 1800 2000 22000

100

200

300

400

500

Vem

(V)

I em (

μA

)

Initial

Post−DTM

Post−ToF

FIG. 18: Emission currents tended to decay over the courseof measurements, particularly after ToF measurements whichnecessitated prolonged constant polarity emissions.

the test campaign were a particular concern. Figure 18provides an example of the decrease in emitted currentbefore and after direct thrust measurements at ESTECand subsequently after ToF measurements at EPFL. It istempting to broadly attribute this behavior to degrada-tion due to electrochemical reactions at the ionic liquid toguard ring contact interface[2, 44]. We have followed themost recent recommendations in the literature wheneverpossible: use of a the glass emitter substrate enforces adistal electrical contact[46] and attempts were made tobalance positive and negative emission charge transferover each period of potential alternation during thrustmeasurements[33, 50]. However; long (>10 s) periods ofsustained unipolar operation were unavoidable with theToF and RPA systems as configured, and the efficacy ofpotential alternation to suppress reactions was not con-firmed within our specific application. If, for example,reactions were occurring at the ill-defined aluminum toIL interface the effective current path resistance couldhave been increasing, despite the absence of breakdownsat the glass emission sites. To gain further insight into

0 1 2 3 4

−2000

0

2000

Time (hours)

Ve

m

(V) +2100 V

−2030 V

+2450 V

−2500 V

0 1 2 3 4

−200

−100

0

100

200

Time (hours)

I em

(µA

)

FIG. 19: One device was run for 4.5 hours with a closed loopfeedback control set to maintain 15 μN of predicted thrust

by maintaining a prescribed⟨Ib√|Vem|

⟩. As the current de-

cayed, the voltage increased accordingly.

emission decays we operated SRC-Im-1 continuously for> 4.5 hours at 1 Hz and with a closed loop feedbackmechanism to ensure constant charge balance while alsoattempting to maintain 20 μN of predicted thrust outputin each polarity through constant ± Ib

√|Vem| setpoints.The current and voltage in each polarity are presentedin Figure 19 over the course of the test. The voltagerequired to sustain the commanded thrust rose steadilyfrom +2100/-2030 V to +2450/−2500 V while the cur-rent decayed accordingly. It is notable that the averagepower rose less dramatically, from ∼0.29 W to ∼0.32 W .The current spikes visible were, in part, indications of de-ficiencies in the simple controller employed which could

19

not rapidly suppress surges.The test was ended intentionally, no specific or dra-

matic failure mode was encountered and liquid remainedin the device. The total time required to deplete theavailable propellant was calculated using equation 7 withthe coefficients φ listed in table IV and the average cur-rents measured during this test. If the device were freshlyloaded with the initial 110 to 120 μL of EMI-Im, all liquidwould have been consumed after approximately 40 hours.In actuality, the device had already been operated forapproximately 2 hours during the thrust and diagnosticdata acquisitions prior to beginning this test. However;even considering this time, many more hours would havebeen required to exhaust all available propellant.At the culmination of the test we inspected the device.

Most evidently, the extractor grids were eroded near theemitter strip ends and the previously white/clear sur-face was distinctly yellow. Erosion was not unexpectedat these positions since the emitting strips terminateabruptly at each end, resulting in sharp features per-pendicular to the desired orientation. Emissions fromthese ends had struck the extractor grid and, after sev-eral hours, caused significant erosion. The accompanyingsputtered material may also be responsible for some ofthe observed discoloration and, possibly, current decaythrough contamination of the emitting surfaces.

6.4. Triangular Prisms as Emission Sources

The conventional machining approach applied was suc-cessful in obtaining up to 500 μA of highly ionic emis-sion from 1 cm diameter active areas; current densitiessimilar to reported measurements from 100’s of axisym-metric porous emitters multiplexed in arrays[2, 5]. Thenumber of active emission sites was not confirmed di-rectly. Maximum reported currents from individual ax-isymmetric porous emitters have ranged from 100’s ofnA[5] to more than 5 μA[23]. The hundreds of μA re-ported here are therefore broadly indicative of activatingseveral hundred emission sites over the 9 emitting edges.However, the number of sites per structure in the litera-ture has not been reported with certainty. Furthermore,a definitive or phenomenological maximum current levelwas not reached here nor has one been typically reportedfrom previously reported devices using structures with aporous bulk material. Correspondingly, detailed quan-titative comparisons between current levels and emitterdensities in the literature as means to calculate emissionsite distributions are imprudent at present.As local field strengths will have exceeded the mini-

mum threshold for emission at differing voltages alongthe undulated edges, the number of active sites likelyincreased with voltage. Further study is required to es-tablish criteria for emission site spacing; accounting forpotential hydrodynamic interactions and space charge ef-fects along with the local electrostatics. Such an un-derstanding will be important in developing the design.

1600 1700 1800 1900 2000 2100 2200 23000

0.5

1

1.5

2

2.5

3

3.5

|Vem

| (V)

κ

|Vem

|>0

|Vem

|<0

+ Sweep 1

+ Sweep 2

− Sweep 1

− Sweep 2

(a)

−30 −20 −10 0 10 20 300

0.5

1

1.5

2

2.5

3

3.5

Stage Angle (o)

κ

|Vem

|<0

|Vem

|>0

+ Rot. ┴+ Rot. ||

− Rot. ┴− Rot. ||

(b)

FIG. 20: Small and inconsistent trends were observed whenmeasuring the ToF paramater κ at differing voltages (a) andbeam angles (b). In (a) sweeps indicate repeated data acqui-sition under the same conditions.

For example, the emitting edges protruded into the gridthickness for both EMI-Im devices. Some emissions fromsites off the symmetry axis of those devices may havebeen lost due to strong electric fields oriented towardsthe grid walls. As with misguided emissions from theends of each strip, resultant sputtered material or sec-ondary electrons could have interacted with the nearbyIL and may have contributed to the degrading emissionover time. Enhanced knowledge of emission site localiza-tion may also be beneficial to interpreting fragmentationevents through providing an improved understanding ofthe electric field each particle experiences after emission.

ToF measurements were made at differing emitter volt-ages and beam angles. However; all measurements wereaveraged to provide single propulsive metrics for eachdevice and polarity. This decision was made for two rea-sons. Firstly, the beam current decays over the course ofToF measurements resulted in ambiguity when attempt-ing to attribute voltage resolved ToF parameters with

20

the conditions present during DTM. Second, any changesin propulsive metrics which were observed tended to berelatively small and comparable to the deviations inher-ent when repeating measurements at consistent inputs.Specifically, no clear trends were visible in beam com-position versus extraction voltage nor angle, via eitherthe thrust metric κ nor the mass flow rate factor φ. Forexample, figures 20(a) and 20(b) demonstrate the calcu-lated parameter κ versus voltage (at 0o) and versus anglerespectively from SRC-Im-2.The lack of trend versus beam angle result would be

particularly noteworthy for an individual emitter or anarray of geometrically identical emitters. Previous stud-ies have, for example, found significant trends in massflow rate (droplet content) versus angle from individualILIS[37]. However; measurements versus angle here com-prise a summation of currents due to beamlets originatingat different surface angles and with differing intensitiesdue to the lack of enforced emission site localization onthe porous surface.

7. CONCLUSION

We have presented a simple to fabricate ionic electro-spray thruster with fully passive propellant feeding. Asmall yet conventional CNC mill was used to machinetriangular prisms, ∼ 300 μm tall and with an apex ra-dius of curvature of roughly 10-30 μm, from commerciallyavailable porous borosilicate filter discs. Three prototypethrusters have been tested, two emitting the ionic liquidEMI-Im and one using EMI-BF4. For all sources, severalhundred μA were emitted from less than 1 cm2 of activearea at voltages near 2000 V . Despite limited controlover the emitting apex curvature, the emission spreadover angle was equivalent to a ∼ 21o hollow cone, repre-senting a small ∼ 7 % reduction in thrust over a perfectlycollimated beam.The utility of such a source as a compact thruster has

been demonstrated through direct thrust measurements.Controllable thrusts from 5 μN to 50 μN , correspondingto 0.1 W to 0.8 W , were recorded when emitting EMI-Im and from 7 μN to 25 μN at 0.2 W to 0.7 W whenemitting EMI-BF4. These data were measured while al-ternating the emission polarity at 1 Hz while attemptingto balance total charge emission through appropriate se-lection of the duty cycle; in an effort to suppress electro-chemical degradation. The thrusters were controlled byan applied emitter to extractor grid voltage alone, pro-pellant IL was passively drawn, by capillarity, from anintegral porous reservoir.Accurate calculation of thrust using indirect probe

measurements has been discussed and demonstrated.Through defining non-dimensional modifiers describingthe average mass to charge ratio of constituent beam par-ticles, an effective beam angle and energy deficits com-pared with the applied voltage we have made calcula-

tions of thrust output as a function of beam current andemission voltage. For two of three devices, these calcu-lations matched the directly measured thrust to withina few μN . Specifically, we have shown that while Timeof Flight spectrometry alone can provide a thrust esti-mate close to that directly measured, modifications dueto beam angle and energy deficits should be includedto obtain good accuracy. The effects of fragmentationwithin the emitter to extractor grid were also consid-ered yet were found to have a small (∼3 %) influence onthrust.

The mass flow rate has been similarly calculated us-ing ToF measurements and the results have been ap-plied, along with thrust, to calculate the expected spe-cific impulse and total propulsive efficiencies of the proto-type thrusters. Emissions from EMI-Im included a smalldroplet, or highly solvated ion, content; representing upto 10 % of the total current during negative polarity op-eration. This population contributed to relatively lowcalculated propulsive efficiencies and specific impulses,roughly 50 % and 1500 s at 0.5 W from devices emittingEMI-Im. Conversely, spectra from EMI-BF4 emissionsrevealed a lower intensity (< 4.5 %) and lower mass sol-vated ion tail; consistent with this liquid approachingthe PIR. Consequently, higher propulsive efficiency (65%) and specific impulse (3260 s) were calculated.

These novel, conventionally machined thrusters can en-able low-cost and accessible study of the operation ofporous IL electrospray sources without an emphasis ondevice fabrication; beginning with this study of thrustand propulsive performance. We intend to continue theseinvestigations through i) developing the device design, inorder to increase the operational lifetime and improveemission consistency, and ii) continued efforts to seek ex-perimental evidence of the predicted performance bene-fits of ILIS thrusters. The latter will include measure-ments of thrust output while simultaneously emittingbalanced currents, of opposite polarity, in order to con-firm charge neutralization and measurements of true spe-cific impulse by directly recording the propellant mass re-duction over very long durations while monitoring thrust.

Acknowledgments

This work has been supported through ESA NPI con-tract #4000109063/13/NL/PA. The authors would liketo thank Alexandra Bulit, Davina Di Cara, Kathe Dan-nenmayer and Eduard Bosch Borras of the ESA Propul-sion Laboratory for their assistance with the direct thrustmeasurements.

References

21

[1] Gassend, B., Velasquez-Garcia, L. G., M., A. I. A.,and Martinez-Sanchez, “A microfabricated planar elec-trospray array ionic liquid ion source with integratedextractor,” Microelectromechanical Systems, Journal of,Vol. 18, No. 3, 2009, pp. 679–694.

[2] Courtney, D. G., Li, H. Q., and Lozano, P., “EmissionMeasurements from Planar Arrays of Porous Ionic LiquidIon Sources,” IOP Journal of Physics D, Vol. 45, No. 48,2012, p. 485203.

[3] Stark, J., Stevens, B., Alexander, M., and Kent, B.,“Fabrication and Operation of Microfabricated Emittersas Components for a Colloid Thruster,” Journal of Space-craft and Rockets, Vol. 42, No. 4, 2005, pp. 628–639.

[4] Krpoun, R. and Shea, H. R., “Integrated out-of-planenanoelectrospray thruster arrays for spacecraft propul-sion,” Journal of Micromechanics and Microengineering,Vol. 19.

[5] Coffman, C., Perna, L., Li, H., and Lozano, P. C., “Onthe Manufacturing and Emission Characteristics of aNovel Borosilicate Electrospray Source,” in “49th AIAAJoint Propulsion Conference and Exhibit,” San Jose, CA,2013.

[6] Romero-Sanz, I., Bocanegra, R., Gamero-Castano, M.,de la Mora, J. F., and Lozano, P., “Source of heavymolecular ions based on Taylor cones of ionic liquids op-erating in the pure ion evaporation regime,” Journal ofApplied Physics, Vol. 94, No. 5, 2003, pp. 3599–3605.

[7] Lozano, P. and Martınez-Sanchez, M., “Ionic liquid ionsources: characterization of externally wetted emitters,”Journal of Colloid and Interface Sciences, Vol. 282, No. 2,2005, p. 415.

[8] Lozano, P. and Martınez-Sanchez, M., “EfficiencyEstimation of EMI-BF4 Ionic Liquid ElectrosprayThrusters,” in “41st Joint Propulsion Conference & Ex-hibit,” Tuscon, Arizona, 2005.

[9] Taylor, G. I., “Disintegration of Water Drops in an Elec-tric Field,” Proceedings of the Royal Society of LondonA, Vol. 280, No. 1382, 1964, pp. 383–397.

[10] Warner, N. Z., “Theoretical and Experimental Investi-gation of Hall Thruster Miniaturization,” Ph.D. Thesis,Massachusetts Institute of Technology, Cambridge, MA,2007.

[11] Mueller, J., Hofer, R., and Ziemer, J., “Survey of propul-sion technologies applicable to cubesats,” NASA Techni-cal Report, , No. 10-1646.

[12] Dandavino, S., “Microfabricated Electrospray Thrustersfor a Modular Spacecraft Propulsion System,” Ph.D.Thesis, Ecole Polytechnique Federale de Lausanne, 2014.

[13] Ziemer, J. K. and Merkowitz, S. M., “MicrothrustPropulsion for the LISA Mission,” in “40th Joint Propul-sion Conference,” Fort Lauderdale, FL, 2004. AlsoAIAA–2004–3439.

[14] Marcuccio, S., Giusti, N., and Pergola, P., “Slit FEEPThruster Performance with Ionic Liquid Propellant,” in“49th AIAA Joint Propulsion Conference and Exhibit,”San Josa, CA, 2013.

[15] Tajmar, M., Vasiljevich, I., Plesescu, F., Grienauer, W.,Buldrini, N., Betto, M., and del Amo, J. G., “Devel-opment of a Porous Tungsten mN-FEEP Thruster,” in“Space Propulsion 2010,” San Sebastian, Spain, 2010.

[16] Legge, R. S. and Lozano, P. C., “Electrospray propul-

sion based on emitters microfabricated in porous met-als,” Journal of Propulsion and Power, Vol. 27, No. 2,2011, pp. 485–495.

[17] Hill, F. A., Leon, P. P. D., and Velasquez-Garcıa,L. F., “High-throughput ionic liquid electrospray sourcesbased on dense monolithic arrays of emitters with inte-grated extractor grid and carbon nanotube flow controlstructures,” in “17th International Conference on Solid-State Sensors, Actuators and Microsystems, Transducers2013,” IEEE, 2013, pp. 2644–2647.

[18] Ziemer, J. K., Randolph, T. M., Franklin, G. W., Hruby,V., Spence, D., Demmons, N., Roy, T., Ehrbar, E.,Zwahlen, J., Martin, R., and Connolly, W., “Colloidmicro-newton thrusters for the space technology 7 mis-sion,” in “IEEE Aerospace Conference 2010,” Big Sky,MT, 2010. Also AERO–2010–5446760.

[19] Gamero-Castano, M. and Hruby, V., “Electrospray as aSource of Nanoparticles for Efficient Colloid Thrusters,”Journal of Propulsion and Power, Vol. 17, No. 5, 2001,pp. 977–987.

[20] Alonso-Matilla, R., Fernandez-Garcıa, J., Congdon, H.,and de la Mora, J. F., “Search for liquids electrosprayingthe smallest possible nanodrops in vacuo,” Journal ofApplied Physics, Vol. 116, 2014, p. 224504.

[21] Lenguito, G. and Gomez, A., “Development of a multi-plexed electrospray micro-thruster with post-accelerationand beam containment,” Journal of Applied Physics, Vol.114, 2013, p. 154901.

[22] Ryan, C., Daykin-Iliopoulos, A., Stark, J. P. W.,Salaverri, A., Vargas, E., Rangsten, P., Dandavino, S.,Ataman, C., Chakraborty, S., Courtney, D., and Shea,H., “Experimental progress towards the MicroThrustMEMS electrospray electric propulsion system,” in “33rd

International Electric Propulsion Conference,” Washing-ton, DC, 2013. Also IEPC-2013-146.

[23] Courtney, D. G. and Lozano, P., “Characterization ofConical Ionic Liquid Ion Sources for 2-D ElectrosprayThruster Arrays on Porous Substrates,” Transactions ofthe Japan Society for Aeronautical and Space Sciences,Aerospace Technology Japan, Vol. 8, 2010, pp. 73–78.

[24] Courtney, D. G., “Ionic Liquid Ion Source Emitter Ar-rays Fabricated on Bulk Porous Substrates for Space-craft Propulsion,” Ph.D Thesis, Massachusetts Instituteof Technology, Cambridge, MA, 2011.

[25] Deng, W., Klemic, J. F., Li, X., Reed, M. A., and Gomez,A., “Increase of electrospray througput using multiplexedmicrofabricated sources for the scalable generation ofmonodisperise droplets,” Aerosol Science, Vol. 37, 2006,pp. 696–714.

[26] Dandavino, S., Ataman, C., Ryan, C. N., Chakraborty,S., Courtney, D. G., Stark, J. P. W., and Shea, H., “Mi-crofabricated electrospray emitter arrays with integratedextractor and accelerator electrodes for the propulsion ofsmall spacecraft,” Jounal of Micromechanics and Micro-engineering, Vol. 24, No. 7, 2014, p. 075011.

[27] Courtney, D. G., Li, H. Q., and Lozano, P., “Electro-chemical Micromachining on Porous Nickel for Arraysof Electrospray Ion Emitters,” Journal Of Microelectro-chemical Systems, Vol. 22, No. 2, 2013, pp. 471–482.

[28] Meyer, I. V., Edmond, J., and King, L. B., “Electrosprayfrom an Ionic Liquid Ferrofluid utilizing the Rosensweig

22

Instability,” in “49th Joint Propulsion Conferene and Ex-hibit,” San Jose, CA, 2013. Also AIAA-2013-3823.

[29] Lozano, P., Martınez-Sanchez, M., and Lopez-Urdiales,J. M., “Electrospray emission from nonwetting flat dielec-tric surfaces,” Journal of Colloid and Interface Sciences,Vol. 276, 2004, pp. 392–399.

[30] Perel, J., Mahoney, J. F., Moore, R. D., and Yahiku,A. Y., “Research and Development of a Charged-ParticleBipolar Thruster,” AIAA Journal, Vol. 7, No. 3, 1969,pp. 507–511.

[31] Legge, R. and Lozano, P., “Performance of HeavyIonic Liquids with Porous Metal Electropsray Emit-ters,” in “44th Joint Propulsion Conference & Exhibit,AIAA2008–5002,” Hartford, CT, 2008.

[32] Courtney, D. G. and Shea, H., “Accounting for Fragmen-tation in Time of Flight Spectrometry Based Calculationsof Ionic Electrospray Thruster Performance,” Journal ofPropulsion and Power. Submitted.

[33] Castro, S. and de la Mora, J. F., “Effect of tip curva-ture on ionic emission from Taylor cones of ionic liquidsfrom externally wetted tungsten tips,” Journal of AppliedPhysics, Vol. 105, 2009, p. 034903.

[34] Lozano, P., “Energy Properties of an EMI-Im ionic liq-uid ion source,” Journal of Physics D: Applied Physics,Vol. 39, 2006, pp. 126–134.

[35] Chakraborty, S., Ataman, C., Dandavino, S., and Shea,H., “Microfabrication of an electrospray thruster forsmall spacecraft,” Proceedings of POWERMEMS, pp. 2–5.

[36] Fedkiw, T. P. and Lozano, P. C., “Development and char-acterization of an iodine field emission ion source for fo-cused ion beam applications,” Journal of Vacuum Sci-ence and Technology B, Vol. 27, No. 6, 2009, pp. 2648–2653.

[37] Chiu, Y.-H., Gaeta, G., Levandier, D. J., Dressler, R. A.,and Boatz, J. A., “Vacuum electrospray ionization studyof the ionic liquid,[Emim][Im],” International Journal ofMass Spectrometry, Vol. 265, No. 2, 2007, pp. 146–158.

[38] Martınez-Sanchez, M., “MIT Course 16.522,Space Propulsion Notes,” MIT Open Coursware,http://ocw.mit.edu.

[39] Ziemer, J. K., “Performance of Electrospray Thrusters,”in “31st International Electric Propulsion Conference,”Ann Arbor, MI, 2009. Also IEPC–2009–242.

[40] Lozano, P. and Martınez-Sanchez, M., “Studies on theIon-Droplet Mixed Regime in Colloid Thrusters,” Ph.DThesis, Massachusetts Institute of Technology, Cam-bridge, MA, 2003.

[41] Krpoun, R., Smith, K. L., Stark, J. P. W., and Shea,H., “Tailoring the hydraulic impedance of out-of-planemicromachined electrospray sources with integrated elec-trodes,” Applied Physics Letters, Vol. 94, No. 16, 2009,p. 163502.

[42] Khayms, V., “Advanced Propulsion for Microsatellites,”Ph.D Thesis, Massachusetts Institute of Technology,Cambridge, MA, 2000.

[43] Larriba, C., Garoz, D., Bueno, C., Romero-Sanz, I., Cas-tro, S., and de la Mora, J. F., “Taylor Cones of IonicLiquids as Ion Sources: The Role of Electrical Conduc-tivity and Surface Tension,” in Brennecke, J. F., Rogers,R. D., and Seddon, K. R., eds., “Ionic Liquids: Not JustSolvents Anymore,” Oxford University Press, New York,2007.

[44] Lozano, P. and Martınez-Sanchez, M., “Ionic liquid ionsources: suppression of electrochemical reactions usingvoltage alternation,” Journal of Colloid and InterfaceSciences, Vol. 280, 2004, pp. 149–154.

[45] de la Mora, J. F., Berkel, G. J. V., Enke, C., Cole, R. B.,Martinez-Sanchez, M., and Fenn, J. B., “Electrochemicalprocesses in electrospray ionization mass spectrometry,”Journal of Mass Spectrometry, Vol. 35, No. 8, 2000, pp.939–952.

[46] Brikner, N. and Lozano, P. C., “The role of upstream dis-tal electrodes in mitigating electrochemical degradationof ionic liquid ion sources,” Applied Physics Letters, Vol.101, No. 19, 2012, p. 193504.

[47] Enloe, C. L. and Shell, J. R., “Optimizing the energyresolution of planar retarding potential analyzers,” Re-view of scientific instruments, Vol. 63, No. 2, 1992, pp.1788–1791.

[48] Fedkiw, T. P., “Characterization of an Iodine-BasedIonic Liquid Ion Source and Studies on Ion Fragmenta-tion,” S.M. Thesis, Massachusetts Institute of Technol-ogy, Cambridge, MA, 2010.

[49] Prince, B. D., Tiruppathi, P., Bemish, R. J., Chiu, Y.-H.,and Maginn, E. J., “Molecular Dynamics Simulations of1-Ethyl-3-methylimidazolium Bis [(trifluoromethyl) sul-fonyl] imide Clusters and Nanodrops,” The Journal ofPhysical Chemistry A, Vol. 119, No. 2, 2015, pp. 352–368.

[50] Coffman, C., Courtney, D. G., Hicks, F., Jamil, S., Li,H., and Lozano, P., “Progress Toward a Variable SpecificImpulse Electrospray Propulsion System,” in “47th JointPropulsion Conference & Exhibit,” San Diego, CA, 2011.AIAA–2011–5591.