f

yed

esoscalenjectiones,through

phy/massprovidesO

with assurface

energyoperationte. Dropletof the

ign, side-relying

Combustion and Flame 139 (2004) 77–89www.elsevier.com/locate/jnlabr/cn

Optimization of a catalytic combustor using electrospraliquid hydrocarbons for mesoscale power generation

Dimitrios C. Kyritsisa,1, Bruno Coritona, Fabien Faurea,Subir Roychoudhuryb, Alessandro Gomeza,∗

a Yale Center for Combustion Studies, Department of Mechanical Engineering, Yale University, P.O. Box 208286,New Haven, CT 06520-8286, USA

b Precision Combustion, Inc., 410 Sackett Point Road, North Haven, CT 06473, USA

Received 16 October 2003; received in revised form 15 March 2004; accepted 15 June 2004

Available online 11 September 2004

Abstract

A detailed study on the performance of a combustor to be used as a portable power source for mapplications is presented. The burner operation is based on the combination of liquid fuel electrospray iwith combustion through a stack of catalytically coated grids, for the delivery of≈ 100 W of thermal power. Thmain design challenges relate to emission minimization, versatility for the coupling to power conversion modulethermal management, and miniaturization. Combustion efficiency and emission reduction were pursuedcatalyst optimization. Using two-dimensional infrared temperature measurements and gas chromatograspectrometry/flame ionization detection exhaust gas analysis, we established a catalyst formulation whichin excess of 99% combustion efficiency, based on the conversion of the parent hydrocarbon and air to C2 andH2O. Remarkably, reliable catalyst operation was achieved even using the notoriously polluting JP8,many as 1200 ppm of sulfur naturally present in the fuel. CO emission is undetectable and catalytictemperatures fall in the 900–1500 K range, which is appropriate for coupling with thermal-to-electricconversion systems, such as thermoelectric and Stirling engines. The burner was tested for prolonged(500 h) for catalyst stability and aversion to coking, even under conditions of highair inlet temperature, to simulaconditions of heat recuperation that are indispensable to the design of high efficiency mesoscale devicessizes reveal the need for fuel distributor multiplexing to minimize vaporization time and therefore the sizenecessary preheat chamber. The results of the characterization of a prototypical device led to an improved desutilizing multijet electrospray injection from a single fuel source, an electrospray ring extractor, and whirlport air injection. In addition to reduced emissions and better temperature uniformity, this improved designon conventional fabrication resulted in optimal performance in a volume on the order of 10 cc. 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Microcombustion; Electrospray; Catalytic combustion; Power generation; JP8; Microlith

* Corresponding author.E-mail address: alessandro.gomez@yale.edu(A. Gomez).

1 Current address: M&IE Department, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA.

0010-2180/$ – see front matter 2004 The Combustion Institute. Published by Elsevier Inc. All rights reserved.doi:10.1016/j.combustflame.2004.06.010

78 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

P.thehegy,u-

alesof

ly,t an

ari-ofaleInro-ex-y-

ofasbe

foruchbe-t aus

e-ax-

gy

calente

ale

n-les,ines

rat--

sug-no

ach,er

mo-)om-on

re-e-uldtor-rac-ers

nalaleales

f ahalehasely

e onat-ingde-ards

entle,ed,lytheionr-oide isledforhere-the

ar-ad-the

sedat-

ivera-rat-airof

theines

1. Introduction

In his seminal lecture “for the year 2000” R.Feynman foresaw with nearly prophetic accuracyminiaturization that would be accomplished in tsubsequent 40 years in many fields of technolofrom electronics to information storage, from manfacturing to (micro-)electro–mechanical systems[1].To date it has been demonstrated in the realm of scthat Feynman paradoxically described as “plentyroom at the bottom” in his lecture title. Interestingpower generation was not considered, perhaps noaccidental omission.

Indeed, the power generating component, invably a battery, is the factor determining the sizea significant number of currently portable mesoscdevices and often limiting further miniaturization.this quest for increased power density, liquid hydcarbon combustion is a promising technology toplore. The power density of conventional liquid hdrocarbon fuels (on the order of 40 MJ/kg) exceedsthat of state-of-the-art batteries by a few ordersmagnitude. Even operating with efficiency as low5%, combustion-based devices can in principlefour to five times more power dense than batteries[2].On the other hand, the main technical challengesmall-scale combustion is thermal management. Sdevices inherently present significant heat lossescause of their high surface-to-volume ratio, which asufficiently small scale may quench combustion, thlimiting the minimum size of the combustor. Morover, even if steady combustion is sustained, the mimum efficiency of currently available direct enerconversion modules hardly ever exceeds 10%[3,4].As a result, as much as 90% of the fuel chemienergy is transformed to heat, proper managemof which is crucial for the efficient operation of thburner.

Research activity in the nascent field of mesoscand “micro-” combustion was reviewed in[2]. A firstapproach involves direct miniaturization of convetional engines, using classical thermodynamic cycsuch as those adopted in internal combustion engand gas turbines[2,5–8]. Problems with regard tosealing, balancing rotating components, and sepaing the “hot” from the “cold” side of the thermodynamic cycle and difficulties in microfabrication ofthe complex geometries of the necessary partsgest that a second approach with a minimum of ormoving parts may be preferable. In such an approa first stage of combustion is coupled with eithdirect energy conversion modules, such as therelectric (TEG)[9,10], or thermo-photovoltaic (TPVgenerators, or possibly Stirling engines. Several cbustor designs, operating mainly on light hydrocarband alcohol fuels have been proposed[9–11]. Oper-

ation with light gaseous fuels is convenient for pliminary testing but is not an option for practical dvices, since the high-power-density objective wobe defeated by the requirement for gaseous fuel sage. The thermodynamics and heat transfer chateristics of a particular design of small-scale burnare discussed in detail in[9], although the validity ofsimilarity arguments on the basis of nondimensionumbers and assumptions well justified in large-scsystems may be open to question as the length scdiminish.

Our objective is to present the detailed design oclean and efficient burner which will exploit the higenergy density of liquid hydrocarbons for mesoscautonomous power generation. The burner sizeto be on the same length scale as that of the widavailable batteries (a few centimeters) and operatcommercially available liquid fuels. It also has to sisfy a series of requirements which relate to couplwith direct power generation modules. Since thisvice is envisioned as a “liquid fuel battery” withwide range of applications, some emission standaare likely to be imposed on its operation. To the extthat indoor applications of the burner are possiba solution to minimize CO emission has to be devisdespite miniaturization efforts that would inevitabresult in short residence times. Furthermore, incase of logistic fuels such as JP8, a jet propulsfuel consisting of a blend of several liquid hydrocabons with a high propensity to soot, the need to avsooting and the attending burner thermal signaturparamount. We discuss in detail the rationale thatus to the choice of a combination of electrosprayliquid fuel dispersion with catalytic combustion as toptimal solution for the power generation problemstricted by these design criteria. We elaborate onprinciple of operation proposed in[12–14], presentresults on the optimization of the performance of vious components of the prototypical burner, anddress problems associated with the transition fromprototype to an engineering device.

2. Design criteria

The two main operational features of the propoburner are catalytic combustion and electrosprayomization. Before proceeding with the quantitatanalysis, it is worth justifying the design considetions that led to these particular choices and elaboing on strategies to enhance mixing through thesupply configuration, which is the third componentthe burner.

2.1. Catalytic combustion

The temperature at the interface betweenburner and the power generating module determ

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 79

er-g)tureod

esm-rlytly

idscell

There-

malldi-at-eae-

all-uchel

thers

ickberto

cil-ient

to

heex-

wsndode,cusidngi-

heof

en-

ane-ap-

that

ingarp

er

edlese--berr.dif-er,pletec-thislex-

-thatec-y asetisc-

inartsionin

islec-in-es,ofratee-

andrat-adelyes

the choice of the power generating strategy. Furthmore, available technologies (TEG, TPV, Stirlintend to be restricted to narrow ranges of temperafor optimal operation. As a consequence, very gouniformity is required on the interface, with valulying in the 900–1500 K range, where catalytic cobustion is a prerequisite for good stability. Particulawell-suited for small scale applications is a recendeveloped catalyst substrate design (Microlith)[15].It consists of a number of catalytically coated gror screens, each with short channel lengths, highdensity, and low thermal mass, stacked serially.resulting reactor is compact, has rapid transientsponse and high energy density, and requires sloadings of precious metal catalysts. Under contions of kinetic control, this design of staggered calytic grids can pack significantly more active arinto a given volume than the classical monolith dsigns, which consist of arrays of uninterrupted smdiameter channels. This means that insertion of scatalysts in a flow can provide more effective fuconversion for a given pressure drop. Also, atprevailing conditions, the thermal boundary layearound the grid wire are expected to be relatively thfor the laminar flows described here (Peclet numbased on the cell opening is of unity order) andaffect the entire flow through the catalyst, thus faitating reactions in the gaseous phase and efficconversion of the fuel.

2.2. Electrospray fuel dispersion

Although spray injection is not the only routeliquid fuel dispersion (see, e.g.,[11]), it is inevitablefor heavy liquid hydrocarbons, since contact of tfuel with hot surfaces of the hardware may causetensive coking[16]. For liquid hydrocarbon fuel flowrates on the order of 1–10 ml/h, which correspond to10–100 W of thermal power, the electrospray allofor atomization with minimal power consumption apressure drop. It can be operated in the cone-jet mcharacterized by the presence of a conical menisat the atomizer outlet, terminating into a fine liquthread that, in turn, breaks into fine, self-repellidroplets (Fig. 1). In this mode the droplets are unform in size (monodisperse)[17], thus providing awell-defined vaporization time to be matched with tresidence time in the “preheat” chamber upstreamthe catalysts.

Another feature of the electrospray is the depdence of droplet diameter size on flow rate. In[17,18],for example, this dependence is studied for heptwith 0.3% by weight of an electric conductivity enhancer. It is shown there that droplet size scalesproximately asD ∝ Q2/3 (Q, volumetric liquid flowrate) in sprays in the cone-jet mode. This means

Fig. 1. JP8 electrospray at the cone-jet mode.

the droplet diameter is decreasing with diminishflow rate and that this decrease is particularly shfor small flow rates (dD/dQ ∝ Q−1/3). So, if a giventotal flow rate, which is dictated by the thermal powrequired from the burner, is partitioned inton electro-sprays (n > 1), the droplet size at injection is reducby (1/n)2/3. Since the droplet evaporation time scawith the square of the droplet diameter, it will be rduced by(1/n)4/3. This, in turn, should result in significant reduction of the dimensions of the chamwhere the liquid fuel will vaporize and mix with aiFuels other than heptane may have a power lawfering from the 2/3 dependence of heptane. Howevthe monotonically increasing dependence of drosize on flow rate is generally applicable to any eltrosprayed liquid. As a result, regardless of fuels,dependence points clearly to electrospray multiping as a route to miniaturization.

In parallel work[13], aimed at visualizing the mixing/evaporation process in the chamber, we foundoptimal performance results from operating the eltrospray in an unsteady regime accompanied bcharacteristic hissing sound, slightly below the onvoltage for corona discharge. Although this modeunlikely to yield monodisperse droplets, the introdution of unsteadiness into the otherwise steady lamflow field from the whipping electrospray ligamenenhances mixing with the oxidizer, before combustis initiated at the catalyst, which ultimately resultsmore uniform combustion temperature.

If the constraint for monodisperse fuel dropletsrelaxed, there are other operating modes of an etrospray that can be appealing. For example, bycreasing the voltage applied to relatively large tubwhile pumping through flow rates within the rangerelevance to the present application, one can opein the so-called multijet mode, in which several conjets are anchored at a relatively large tube outletspread out from the common source while disinteging into fine droplets. Unless special efforts are mto ensure that the flow rate is partitioned uniformamong the various jets, it is likely that droplet siz

80 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

pray

ll

ngais

re-

ar,lti-

ion.p-ce,the

allyeds

mepti-eninre-toignidertsw

tialr-bytheertro-

c-

p-

neiesediond in-nd

ee-ghte).in-

beredin-

gr ofh aa-ereltsays

at

W.

onlsat-

Fig. 2. Electrosprays produced by a 12-grooved electrosatomizer and extracted by a grounded ring.

may vary significantly from jet to jet. This may stibe acceptable in light of the findings in[13,18]. Fig. 2shows a picture of a multijet electrospray originatifrom a 1/8′′ O.D. tube. The simplicity of havingsingle source, unlikely to ever clog up from coking,apparent.

As a result there are essentially two ways toalize multiplexing: the first is by merely duplicatingan individual capillary again and again to achievegiven level of multiplexing;the second, far simpleis by using a single source operated in the mujet mode. Both were used in the present applicatHowever, the risk of clogging of the multiplexed caillaries, with attending time-consuming maintenanprompted us to resort to the second approach forbulk of the experiments described below, especithose requiring burner operation over tens or hundrof hours.

2.3. Air supply

Determination of an appropriate air supply scheis an indispensable component of the burner omization. Preliminary experiments showed that eva 10 m/s stream of air at atmospheric densitycross flow would not perturb the electrosprays. Thefore, rather than providing the oxidizer coaxiallyeach individual electrospray, as in the original desin [12], one or two lateral ports can be used for sair intake. If air is introduced from these side powith a tangential velocity component, a whirling flofield is established in the burner. Yetter et al.[19] dis-cuss the potential of whirl combustion for substanreduction of pollutants in the range of very low oveall equivalence ratio values (0.05–0.4). Motivatedthese results, we discuss in a following sectionapplication of whirl combustion in the burner undconsideration. This entailed a single-source elecspray atomizer operated in the multijet mode coupledwith a whirled air injection which “shaped” the eletrosprays.

3. Experimental apparatus

A schematic representation of the principle of oeration of one burner design is shown inFig. 3.The flow was distributed through a polyetherketomanifold into a number of stainless steel capillar(1.59 mm O.D., 127 µm I.D., 10 cm length) arrangin a hexagonal pattern. The principle of operatwas demonstrated with seven capillaries arrangea hexagonal ring[12] and the effect of further multiplexing was studied here with the addition of a secoring with twice the radius and the same pitch (19 ndles in total). The capillaries were mounted throua machinable ceramic flange (zirconium phosphaFor experiments that required optical access forchamber observation, a cylindrical pyrex cham38.1 mm in I.D. and 38.1 mm in height was mounton the flange. Air delivery through side ports wastroduced as opposed to the coaxial delivery of[12].To achieve overallϕ = 0.30–0.70, air flow rates onthe order of 5 g/min were required, correspondinto a speed through the supply ports on the orde1 m/s. The chamber was capped at the top witmetal holder in which a number of Microlith catlyst screens were housed. Once the capillaries wcharged to a voltage on the order of several kilovorelative to the grounded catalyst holder, electrosprwere established at the capillary tips.

The fuel was metered with a syringe pumpa flow rate on the order of 10 g/h which corre-sponded to a thermal power of approximately 100

Fig. 3. The principle of operation of a burner operatingthe combination of electrospray injection of logistic fuewith catalytic combustion on appropriately formulated calytic screens.

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 81

edr-

ic6),ls

nt-to

ason

i-izen-

fortanceillar-terthewasedforssathantor.ize

hetedleIn-the

ealt-

ngedridsra-

rdi-

tic

ndm-ideono

, iny

into

erro-

d by

nd

etaltom-ns,e-to

tict re-

lystMa-

-ereato-ysh ax-ed asestsustgi-amentedep--1

ll-ol-

elyies.netem

The burner was operated with three fuels, all dopwith an antistatic additive (Stadis 450; Octel Ameica Inc.), a dilute solution of dinonylnaphthylsulfonacid in a mixture of aromatic solvents (C9–C1which provided some electric conductivity. The fuewere JP8,n-dodecane which is the single componefuel the physical properties of which are closestJP8 [20], and 1,3 di-isopropyl-benzene which hstructure and properties similar to the most commaromatics in JP8[21]. The concentration of addtive had to be kept as low as possible to minimsolid deposition during prolonged operation. A cocentration of 0.05% yielded good spray qualityJP8, whereas 0.3% was necessary for pure-subsfuel surrogates. The pressure loss across the capies was minimal, i.e., approximately 20 cm of wa(∼ 2 kPa), and so was the power consumption ofelectrosprays. The current carried by the spraysmeasured to be on the order of 100 nA, which yielda parasitic loss of less than 1 mW. Even accountingadditional losses in the voltage multiplication proceof high-voltage power supply, it is anticipated ththe electrospray power requirements will be less t1% of the thermal power generated by the combusThis burner design was used in the tests to optimthe catalyst formulation.

It was observed that after multiple turnoffs of tburner, the capillary lines often became obstrucby solid deposits in the capillary tubes. To tackthis problem, the fuel distributor was redesigned.stead of emerging from several capillary tubes,electrospray emerged from a single large tube (1/8′′O.D.) and was operated in the multijet mode[22],as discussed above. This configuration proved idfor tests of very long duration (100–500 h) for caalyst evaluation, requiring minimal periodic cleanievery 75 h of continuous operation. Air was admittthrough the chimney side walls and the catalyst gwere used as ground electrode. With this configution, the durability tests were performed.

The third and last configuration on the path towaburner optimization incorporated the whirl air addtion, with a tangential injection of the air. A schemaof the optimized burner is shown inFig. 4. In this casea 38.5-mm-diameter ring acting as electric grou(ring extractor) was mounted on the Teflon atoizer base and a circular Teflon flap in the inner sof the ring was used to improve electric insulatiand prevent sparking. To establish a whirl flow, tw6-mm-I.D. Pyrex tubes were mounted tangentiallydiametrically opposed positions, on a Pyrex chimnemeasuring 38.5 mm in inner diameter and 20 mmheight. The chamber height was adjustable downa minimum of 11.4 mm, by sliding the chimney ovthe bottom Teflon flange. Teflon tape was used to pvide leak-tight tolerance. The chamber was cappe

Fig. 4. Multijet electrospray burner with ring extractor awhirl air intake.

several layers of catalytic meshes, housed in a mholder. Interposed between the catalyst and the aizer were a few uncoated high-density cell screethe purpose of which was to prevent direct impingment of the droplets on the catalyst and contributethe mixture uniformity at the entrance of the catalyreactor. The air intake was heated to simulate heacuperation in an operating device.

4. Diagnostic techniques

Exhaust gas sampling downstream of the catawas used to determine exhaust gas composition.jor combustion products and light species (N2, O2,H2, CO2, CO, CH4, C2H6, and in general hydrocarbons with three or fewer atoms of carbon) wmeasured using a two-channel Micro-Gas Chromgraph of Agilent Technologies. The device emplotwo channels, one with a molesieve and one witPoraPLOT backflush injector. Acquisition of an ehaust gas sample lasted 60 s and argon was usa carrier gas for both injectors. In some targeted tto identify the unburned hydrocarbons in the exhagas, we sampled gas and analyzed it with an Alent gas chromatograph/mass selective detector/flionization detector (GC/MSD/FID) system (Agile5973N MSD, 6890 GC). Three columns were usin the GC: a Supelco Carboxen column, able to sarate CO2 and light gaseous hydrocarbons, an HPwell-suited to separating hydrocarbons up to C12 andeven higher; and a third column, Molesieve, wesuited to the separation of CO. The Supelco cumn, in fact, could separate CO from N2 only witha small sampling loop, which would have adversaffected the instrument sensitivity to trace specTwo columns were connected to the MSD and owas connected to the FID through an elaborate sys

82 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

fed

rner-ey,hethe

ledthe

ea-eosig-d al onthecali-the

theer-ik)fu-

-tar-utar-edncethelyst

re-orta-try

na

andB

lere-d Cesason

tiont

ols

ridthusdif-hesameo-

lyst

m-

ax-unc-ghsedture,

with two 10-port valves, allowing for backflushing oone of the columns. The GC oven was programmfor 35◦C, hold 10 min, ramp at 5◦C/min to 230◦C,hold 40 min. Sampling the exhaust gases of the buwas done using a quartz probe positioned approximately 1 cm above the catalytic grid into a chimnlocated on top of the burner, to avoid dilution of tgases with the atmosphere. The tubing connectingprobe to the sampling rig was heated at 200◦C toavoid condensation. The sampling loops were filat a pressure of 400 mm Hg before injection intocapillary columns.

The temperature at the catalyst surface was msured using a PV-320 Electrophysics infrared vidcamera and a germanium objective lens. To avoidnal saturation, a glass filter of 0.3 optical density an2.5-µm cutoff was used, so that the recorded signathe camera chip was from the 1.5–2.5-µm part ofspectrum. The temperature measurements werebrated using a K-type thermocouple coated withcatalyst material.

To characterize the droplet size behavior forfuels of interest as a function of flow rate, a commcial Phase Doppler Anemometer (Dantec Electronwas used and operated at Brewster’s angle for theels of interest.

5. Results and discussion

5.1. Catalyst optimization

In [12] we demonstrated the principle of combustor operation using a catalyst that was notgeted specifically for liquid hydrocarbon fuels bwas rather optimized for light gaseous hydrocbon combustion. The 97% efficiency thus achievwas certainly encouraging. However, the preseof amounts of CO as high as 2% per volume influe gases necessitated optimization of the cataso that more efficient conversion to CO2 could beachieved, since CO emissions at that level wouldstrict significantly the use of such burners in indoapplications. During catalyst optimization, both calyst and support formulation and substrate geomewere considered, as shown inTable 1. The catalyston sample A was palladium on stabilized alumiand was the same formulation as that used in[12].The catalyst precious metal ratios on samples BC were 8/2 Pd/Pt. The catalyst support used onand C was stabilized alumina. Additionally, sampC was doped with Ce and Ni for enhanced sulfursistance. The meshes used for samples A, B, anwere approximately orthogonal grids of 125-µm wirwith 1 mm pitch. For sample D, the formulation wkept the same as that of sample B, but the pitch

Table 1Evaluated catalysts

Catalyst Pd/Pt ratio Normalized geometricsurface area

A 10/0 1B 8/2 1C 8/2 (doped for

S resistance)1

D 8/2 1.25

(a)

(b)

Fig. 5. CO mole fraction in the flue gases for JP8 combus(a) as a function of overall equivalence ratio for a constanfuel flow rate of 12.6 ml/h and (b) as a function of fuel flowrate for an overall equivalence ratio of 0.48. The symbindicate catalyst formulations fromTable 1.

the mesh was reduced to 0.8 mm; the catalytic gtherefore was denser and the geometric area wasincreased. In summary, samples A, B, and C hadferent catalyst and/or support formulations with tsame substrate, whereas sample D had the exactcatalyst formulation as sample B but different gemetric characteristics. In all cases the total cataloading was significantly less than 1 g.

The results of catalyst optimization for JP8 cobustion are shown inFigs. 5 and 6where the ratio ofCO to CO2 concentration in the flue gases and mimum catalyst temperature are presented as a ftion of equivalence ratio and mass flow rate throuthe burner. The maximum temperature can be uas a reasonable estimate of the average tempera

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 83

foratio-8.

as-. Ifflue

theageis

ofuel

thethe

te,

ns--ndsofhanver, ox-theuiv-

heationli-nceatetotoof

ro-theom-ieldof

a-esinimitherr-geo-s in

effi-di-ithus-for

ostop-tiosletery

rgylyst

ly,

x-

om

theare.n%le

ledeat

(a)

(b)

Fig. 6. Maximum temperature on the catalytic screensJP8 combustion (a) as a function of overall equivalence rfor a constant fuel flow rate of 12.6 ml/h and (b) as a function of fuel flow rate for an overall equivalence ratio of 0.4The symbols indicate catalyst formulations fromTable 1.

since it was shown in[12] that a±5% uniformity isachieved. It was preferred only because it was eier to acquire from infrared thermography imageswe ignore other carbon-containing species in thegas, the ordinates inFigs. 5a and 5bcan be interpretedas approximately the per-volume compositions ofburner exhaust in CO or, equivalently, the percentof carbon which was not oxidized completely. ThCO/(CO + CO2) ratio is presented as a functionthe equivalence ratio for a constant volumetric fflow rate of 12.6 ml/h in Fig. 5a and as a functionof mass flow rate forϕ = 0.48 in Fig. 5b. In the for-mer case the equivalence ratio varies by varyingmass flow rate of air, whereas in the latter casetotal mass flow rate is proportional to fuel flow rasinceϕ is kept constant.

To interpret the results ofFig. 5, one should keepin mind that the operation of the burner is heat trafer limited as explained in[12]. As combustion becomes leaner, heat loss (the rate of which depemuch more strongly on the physical propertiesthe burner hardware and the environment rather ton the burner operation parameters) prevails ocombustion-related heat release. In such a caseidation is not completed; i.e., CO concentration influe gases increases. If, on the other hand, the eq

alence ratio approaches the stoichiometric value,losses cannot hinder full oxidation and the conversto CO2 is more complete. The scenario is compcated somewhat for the case when the equivaleratio is kept constant and the fuel mass flow rvaries. In that case, fuel flow rate is proportionaltotal flow rate and therefore inversely proportionalresidence time in the burner and in the vicinitythe catalytic grid. At increased flow rates, we pvide increased heat input to the system; however,residence time in the catalytic burner decreases. Cpetition between these two effects is expected to yan optimum flow rate with regard to completenesscombustion.

Fig. 5 shows that under all conditions of opertion, catalyst D, i.e., the one with the finest grid, givsignificantly more complete conversion. Actuallymost cases, CO is close to or below the detection lof the employed GC technique (10 ppm). On the othand, it is difficult to decipher any substantial diffeence between samples A–C which have the samemetric surface areas, despite significant differencetheir formulation. Using the process outlined in[13](carbon atom balance) to estimate combustionciency from such data, we evaluate that in all contions of operation it exceeds 99% for operation wcatalyst D. The stoichiometry of complete combtion was calculated using the average C–H ratioJP8, which is close to that ofn-dodecane.

Increased combustion efficiency caused an alm5% increase of maximum catalyst temperature foreration with catalyst D at elevated equivalence ra(Fig. 6a). Since the combustion was almost comp(approximate efficiency of 97%) even for preliminaresults with a nontargeted catalyst[13], we did notexpect a drastic increase in temperature. An enebalance in a control volume surrounding the catabed yields

dmf

dtQLHV = dm

dtcp(T − To)

(1)+ εσA(T 4 − T 4

o) + Q′

cond,

wherem andmf are total and fuel mass, respectiveQLHV is the lower heating value of the fuel,ε is theemissivity of the catalyst material which was approimated as that of the steel material of the wires,A isthe radiating surface which can be estimated frthe dimensions of the grid wires for catalystD, andQcond is the heat loss due to conduction throughcatalyst material and the rest of the burner hardwQcond is difficult to measure directly but substitutiointo Eq.(1) yields that it must be on the order of 12of the total energy input, which is entirely conceivabfor the current experimental setup. Without detaiknowledge of what happens with 12% of the fuel hrelease, the results ofFig. 6b that show practically no

84 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

nsr is

tion

attect-ly

ofs area-oge-

lsad-ion

theod-heliq-edofer-

atedthe-

K,ra-r-, itsur-ok-

ano-sferdriftape

r-ns

ar-

enonhedanttic

kledngthem-

tion

ed

xi-

de-ap-mce.

ra-ale,theturehein

theis

keb-id

wnis-n oftedlyde-ve,noasop-f aslyheu-onacecov-eter

ivegas

re-

difference between the various catalyst formulatioare not surprising. Heat management in the burnediscussed further in the context of heat recuperain the following section.

The fact that complete combustion is achievedthe low temperatures indicated inFig. 6suggests thaheterogeneous reactions play significant roles affing the bulk of fuel conversion and do not simpinitiate gas phase chemistry. However, the rangetemperatures is such that homogeneous reactionlikely to be also significant. The study of the reltive importance between homogeneous and heterneous chemistry will be addressed in the future.

5.2. Prolonged operation of the burner

For continuous operation during time intervafrom a few hours to hundreds of hours, there areditional design concerns pertaining to coke formatand catalyst durability.

An efficient use of the heat rejected by eitherburner or any potential direct energy conversion mule coupled to it would be to heat the air fed to tburner, which could then also be used to warm theuid fuel to facilitate vaporization. Since, as explainabove, at a minimum 85% of the chemical energythe fuel is expected to convert to heat, the tempature of the preheated air stream could be elevto levels close to the combustion temperature ofburner, in something akin to a regime of mild combustion[23]. For a maximum temperature of 1200it is not unreasonable to expect air intake tempetures on the order of 700 K. Although this is a themodynamically advantageous mode of operationhas been established that, when heated on solidfaces, liquid hydrocarbons present a very strong cing tendency[16,21]. Any form of solid deposition isunwanted for the operation of the burner. Coke cobstruct capillary fuel delivery lines and solid depsitions at the burner walls may alter the heat trancharacteristics and cause the operational point tofrom a desired steady state. Moreover, since the shof the tip of the fuel injector is important for the fomation of the electrospray, even minimal depositioon the tip may alter drastically the atomization chacteristics.

Early indications of limited solid deposition on thburner walls prompted us to examine the phenomein a more systematic manner. First, it was establisthat the amount of antistatic additive had a significeffect on solid deposition. Decreasing the antistaadditive content to 0.05%, as described above, tacthe problem successfully for approximately 1-h-loexperimental runs. To test possible obstruction ofcapillaries by coking, we increased the air intake teperature to 750 K to simulate extensive recupera

Fig. 7. Unburnt hydrocarbon and CO emission for prolongburner operation.

in an actual device and ran the burner for appromately 3½ h at a JP8 flow rate of 11 g/h and overallequivalence ratio of 0.48.

To exacerbate possible coking formation, weliberately increased the protrusion of the steel cillaries by a factor of 10 to a total length of 5 mto increase the contact time of fuel and hot surfaThis affected adversely the uniformity of tempeture on the catalyst. However, even on this time scno effect on the macroscopic characteristics ofburner operation (maximum temperature, temperanonuniformities of the catalyst) was observed. Tmain reasons for this are that the capillary lengthcontact with the hot air stream is short and thatfuel residence time in contact with the hot surfacetoo brief, on the order of 1 ms, for coking to taplace in the tube. Minor solid deposition was only oserved on the spray tip, where the “cone” of liqufrom which the spray forms sat.

The results of a 500-h test of the catalyst are shoin Fig. 7, where CO and unburnt hydrocarbon emsion measurements are presented as a functiotime during the test. The experiment was interrup10 times for fuel distributor cleaning, since a highviscous deposit, the apparent quantity of whichpended directly on the amount of antistatic additiformed at the exit of fuel supply tubes, althoughobstruction of the supply lines due to coking wobserved. The results indicate practically stableeration of the catalyst, despite the presence omuch as 1200 ppm of sulfur in the fuel. It is likethat the observed cyclic variability is related to tperiodic interruptions of operation for the distribtor cleaning. After 500 h of aging, scanning electrmicroscopy analysis of the washcoat support surfshowed a well-dispersed Pd phase (bright spots)ering the entire surface with average particle diamof up to∼ 1 µm (Fig. 8). This indicates relatively lowcatalyst sintering. Comparison of energy dispersspectroscopy signals in catalyst exposed to theflow and catalyst sheltered from the flow did not

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 85

ter

tes

n-m-for-ger

-jet

se.zesre-an

testhatowi-owoplet

Therays

s.

siertailldpos-ray.tillofifi-

antion

r--theheyus

ch-

edi-metemat-tedec-

asoletattheeres-

ofeedrox-pes

a-u-s.toan-div-

of

Fig. 8. Particle size distribution of catalyst surface af500 h aging.

Fig. 9. Droplet diameter as a function of JP8 flow rathrough 127-µm (0.005′′ ) I.D. capillaries for electrosprayat the cone-jet mode.

veal meaningful differences, suggesting little changein morphology. The washcoat integrity remained uaffected with no indications of distress, e.g., delaination. These results suggest that catalyst permance should be durable for periods much lonthan 500 h.

5.3. Droplet sizing

JP8 droplet size measurements in the conemode are reported inFig. 9. In this mode of oper-ation, the electrospray is practically monodisperThe relative standard deviation in the reported siis less than 0.10. The exponent of a power lawgression through the data is significantly smaller ththat for the heptane case (n ≈ 0.4), which indicates amilder dependence on flow rate at higher flow raand a drastic one for smaller flows. If we assumethis relationship can be extrapolated to very low flrates, as shown inFig. 9, the need for extensive multplexing is exacerbated, since only for very small flrates per spray can a substantial decrease in dr

size and therefore vaporization time be observed.necessary increase of the number of fuel electrospcan be pursued for the designs ofFigs. 3 and 4onlythrough microfabrication of extensive arrays of tipHowever, the ring extractor configurations ofFigs. 2and 4offer the perspective of geometries much eato manufacture and will be discussed in more dein the following sections. A word of caution shoube spent on the use of droplet size data for other,sibly unstable, operating modes of the electrospThe monotonic behavior of size versus flow rate sholds but only qualitatively, and the distributiondroplet sizes for a given flow rate broadens signcantly. Thus, the functional dependence ofFig. 9canbe used only for qualitative guidance.

5.4. Optimized combustor: multijet electrospray withring extractor and whirl air injection

Our analysis so far has demonstrated significadvantages of pursuing multijet electrospray injectfrom a relatively large (1/16–1/8′′ O.D.) tube (Fig. 2)over distributing the flow through an array of capillaies (Figs. 3 and 4). In summary, multijet configurations are significantly easier to manufacture forsame degree of electrospray multiplexation and tare much less vulnerable to solid or highly viscodeposits or to fuel line obstruction after burner switoff.

An important additional modification toward thoptimization of the particular geometry is the adtion of an electrospray ring extractor, which, to soextent, decouples the electric ground of the sysfrom the catalytic screens. To get a more “radial”omization, a metal ring was grounded and mounat the base of the chimney, coaxially with the eltrospray nozzle. Individual cone-jets are extracteda “fountain” (seeFig. 2). The catalytic screens dnot attract the droplets electrostatically. The dropstreams move toward the ring, thus “lingering”the bottom of the preheat chamber. This extendsresidence time of the fuel into the mixing chambfor a given chamber size. Equivalently, the necsary preheat chamber height for the vaporizationa given flow rate of fuel can be decreased. Inda decrease in chamber height by a factor of appimately 3 was observed compared to the prototyof [12,13].

Whirl air intake is expected to improve temperture uniformity on the catalytic surface, which is crcial for efficient coupling with direct energy moduleIn fact, uniformity of temperature relates directlythe homogeneity of the local equivalence ratio. Qutification of the combustion uniformity is performewith the IR camera over a range of overall equalence ratios, using JP8 at a constant flow rate

86 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

yticnce

facefor

in

er-ngheis

an

JP8m touiv-

ob-

omanyeatare-Thepe-to

n,torticthisby

heurera-

rein-air

ereer-

elyture

i-s.

ed

be-nedu-

uni-se

t ara-eane inac-ra-t of

are-m

Fig. 10. False color temperature distribution on the catalscreens for whirl air intake, JP8 fuel, and overall equivaleratio of 0.6.

Fig. 11. Temperature distribution across the catalytic surdiameter as a function of the overall equivalence ratiowhirl air intake;φ = 0.4, 0.5, 0.6, and 0.7 (top to bottom).

12.5 cc/h, and whirl air injection at different flowrates to vary the equivalence ratio.Fig. 10 shows atypical infrared photograph of the catalytic screenfalse color for burner operation of 10 g/h, an overallequivalence ratio ofϕ = 0.6, and whirl air injection ina 25.4-mm-high burner. InFig. 11, radial temperaturedistributions are presented as a function of the ovall equivalence ratio, with values of the latter rangifrom 0.4 (top) to 0.7 (bottom). The area where ttemperature is within 5% of the peak value, whicha more stringent criterion than that used in[12], isindicated by a horizontal line in the plots. A me

temperature of 1170 K is reached when burningat an equivalence ratio of 0.6. There does not seebe a substantial increase of temperature as the eqalence ratio increases from 0.6 to 0.7. This wasserved also in some of the catalyst tests ofFig. 6and is attributed to details of the heat transfer frthe burner. The temperature can be increased forvalue of the equivalence ratio by reducing the hloss due to conduction through the burner hardwas indicated by Eq.(1) with a selection of appropriate insulating material, such as a ceramic blanket.presence of a low-temperature area at the outerriphery of the catalyst screens is mainly attributedthe whirl flow. Because of the lateral air injectiothe mixture is leaner on the side of the combusand the air injection can locally cool the catalyscreens. Consequently, combustion is weak inarea. This low-temperature area can be minimizedinsulating the burner sidewall and by blocking toutlet of the combustor with a suitable disk to ensthat the only exposed surface is at uniform tempeture.

The extent to which the combination of mocompact size provided by the ring extractor withcreased temperature uniformity provided by whirlintake can be sustained is illustrated inFig. 12. At afixed overall equivalence ratio of 0.5, attempts wmade to minimize the chamber height and tempature distributions were measured at heights ofh =16.5 mm (Fig. 12a) and 11.5 mm (Fig. 12b). In thefirst case, temperature was uniform over a relativlarge area of the catalyst, although the temperafluctuations were larger than those inFig. 11. A tem-perature within 10% of the maximum value is indcated with a horizontal line in the related profileThis corresponds to a variation of±5% around amean value of 1088 K, which was the criterion usin [12]. In the case of the shortest chimney (Fig. 12b),the electrosprays impinged on the catalytic gridfore evaporating and the reactor bed was lengtheto prevent fuel loss. During operation, fuel accumlated on the screens before evaporating. Goodformity in temperature is achieved also in this ca(Fig. 12b), i.e., with a variation of±5% around amean value of 958 K, when air was injected atemperature of 523 K. In this last case, tempeture was uniform over a smaller area and the mcombustion temperature was lower. This decreasmean temperature indicated that the minimumceptable burner height for the particular configution was between 11.5 and 16.5 mm. The effecair whirl is evidenced by a comparison ofFigs. 12aand 12c. In both cases experimental parametersidentical, but forFig. 12c, whirl injection was substituted with coaxial air injection from the burner botto

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 87

er.r a

Fig. 12. (a) Two-dimensional (2D) temperature map and radial temperature distribution for a 16.5-mm-high optimized burn(b) 2D temperature map and radial temperature distribution for a 11.5-mm-high optimized burner. (c) 2D temperature fo16.5-mm-high burner without whirl air intake.

ni-

che-r,for

sediledon-

sesed.7).incess

na-ederalOredon-r of

which was obviously detrimental to temperature uformity.

5.5. Exhaust gas analysis by GC/MSD/FID

The micro-GC results ofFig. 5 indicate a min-imal amount of CO in the burner flue gas whifor most operational conditions was below the dtection limit of the micro-GC technique. Howevegiven the strict emission requirements expecteddevices that may be required to operate in clospaces of human activity, we pursued a detacharacterization of emissions for the optimized c

figuration determined above. Exhaust gas analyfrom the optimized whirl combustor were performover a range of overall equivalence ratios (0.4–0For this test dodecane was used as a fuel, sits combustion products can be identified with leambiguity than those of JP8 using chemical alytical techniques. The catalytic reactor was sizto ensure complete conversion by stacking sevgrids. For the optimized whirl air intake burner, Ccould not be detected by GC/MS over the explorange of equivalence ratios, with a threshold ccentration detectable by the GC/MS on the orde1 ppm.

88 D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89

alwith

(a)

(b)

Fig. 13. (a) Sample MSD chromatogram, obtained with the burner operated at an equivalence ratio of 0.4, that is, under marginconditions with regard to emissions (seeFig. 5). (b) FID spectrum, showing only peaks of alkenes and dodecane, obtainedthe burner operated at an equivalence ratio of 0.5, that is, under optimal emission conditions.

heichs),esthere-

ell

r-5–

ith

the

Ofos-

thedi-ns,rnsblethzed

tro-ri-ngn inavyith

d aJP8

maleron

storri-s.

itetheas

per-rnerlb-t of

A sample MSD chromatogram, obtained with tburner operated at equivalence ratio of 0.4 (whFig. 5indicates is marginal with regard to emissionis shown inFig. 13a. Alkenes, aldehydes, and ketonwith carbon atom numbers smaller or equal thanparent fuel were detected. The peaks located fortention times larger than 35 min were neither wseparated nor identified. The results ofFig. 13a arenot typical for all conditions of operation. For opeation at overall equivalence ratios closer to the 0.50.65 optimal region suggested byFig. 5 the structureof the chromatograms is drastically simplified wonly the peaks of alkenes and dodecane (C12H26)being present. FID was used for quantitation ofspecies present in the chromatogram.Fig. 13b showsthe FID spectrum for these optimized conditions.the species that could be detected with our diagntics, the dominant emission is unburnt vapor ofparent fuel at a level on the order of 10 ppm. This incates clean combustion of heavy liquid hydrocarbowhich are notorious for the environmental concethat their combustion entails, except for the inevitasulfur emission, which could be curtailed only wiscrubbing approaches unsuitable for a miniaturicombustor.

6. Conclusions

We demonstrated that the combination of elecsprays of liquid fuels with combustion on appropately formulated fine catalytic grids is a promisiroute toward combustion-based power generatiothe mesoscale, even in the case of fuel blends of hehydrocarbons such as JP8 ( jet propulsion fuel) wtheir notorious emission challenges. We optimizeprototype burner based on this combination usingat flow rates on the order of 10 ml/h with overallequivalence ratios of 0.35–0.70 to produce therpower on the order of 100 W in a volume on the ordof 15 cc. Uniform temperatures were demonstratedthe incandescing catalytic screens at the combuoutput in the 900–1500 K range, which are appropate for coupling with TEG, TPV, and Stirling systemThe stability of the catalyst formulation for prolongedburner operation (> 500 h) was demonstrated, despthe presence of as many as 1200 ppm of sulfur infuel. Even when tested with air inlet temperatureshigh as 700 K, to simulate the extensive heat recuation expected in small-scale combustors, the buexhibited durability over several hour-long tests. Fuesupply line obstruction due to coking was not oserved, because of the limited surface of contac

D.C. Kyritsis et al. / Combustion and Flame 139 (2004) 77–89 89

i-

ris-etoof

net

nsemes

is-

-l-

tionsys-udyith

ab-wertorhiscesson

byra-stichein-

r-t-Ints,er-

llyon

,

02)

97)

20

7–

asis,

c.

.2)

.

A.

.51u-

rdion

ch,

1)

6)

st.

.ch-

s 1

94)

ust.

the fuel capillaries with the hot air flow. Key ingredents to the optimization are the following:

(a) catalyst formulation and geometric charactetics, with a Pt/Pd formulation used to achievcombustion efficiency in excess of 99% andreduce CO emission below the detection limitanalytical instrumentation;

(b) a modification of the initial prototypical desigwhich involved the establishment of a multijelectrospray at the exit of a relatively large (1/8′′O.D.) tube, which eliminated general concerabout the robustness of a fuel delivery systbased on easily cloggable capillary supply linand simplified the manufacturing of the fuel dtributor system; and

(c) the introduction of a whirl air supply, which resulted in enhanced mixing in relatively small voumes.

These results provide the basis for consideraof the proposed combustors from an engineeringtems analysis perspective. This can include the stof manufacturing techniques for such prototypes wa particular emphasis on the application of microfrication techniques and the study of integrated poproducing devices which will combine the combuswith direct energy conversion modules. Although tstudy was motivated by the search for power deviat the level of tens of watts, there is no a priori reathat the approach could not be scaled up (down)a few orders (one order) of magnitude, without dmatic departures from the proposed design. Draminiaturization will necessitate a careful study of tmicrofabrication and thermal management issuesvolved.

Acknowledgments

We acknowledge Dr. R.S. Tranter of the Univesity of Illinois at Chicago for his consulting in seting up the GC/MSD/FID, Mr. R.W. Dean of PCfor his assistance with the micro-GC measuremeMr. N. Bernardo of Yale University for machining thhardware, and Mr. P.A. Dobrowolski of Yale Univesity for acquiring the photograph ofFig. 1. The sup-port of DARPA under Grant No. DAAD19-01-1-0664(Dr. Richard J. Paur, Contract Monitor) is gratefuacknowledged. Microlith is a trademark of PrecisiCombustion Inc.

References

[1] R.P. Feyman, in: H.D. Gilbert (Ed.), MiniaturizationReinhold, New York, 1961, pp. 282–296.

[2] A.C. Fernandez-Pello, Proc. Combust. Inst. 29 (20883–889.

[3] G.D. Mahan, Solid State Phys. 51 (1998) 81–120.[4] G.D. Mahan, B. Sales, J. Sharp, Phys. Today 3 (19

42–47.[5] I.A. Waitz, G. Gauba, Y. Tzeng, J. Fluids Eng. 1

(1998) 109–117.[6] A. Mehra, X. Zhang, A.A. Ayon, I.A. Waitz, M.A.

Schmidt, C.M. Spadaccini, J. MEMS 9 (2000) 51527.

[7] K. Mijit, Design, analysis and experimentation ofmicro internal combustion swing engine, Ph.D. theUniversity of Michigan, Ann Arbor, MI, 2000.

[8] http://www.me.berkeley.edu/mrcl/why.html.[9] F.J. Weinberg, D.M. Rowe, G. Min, P.D. Ronney, Pro

Combust. Inst. 29 (2003) 941–947.[10] J. Vican, B.F. Gajdeczko, F.L. Dryer, D.L. Milius, I.A

Aksay, R.A. Yetter, Proc. Combust. Inst. 29 (200909–916.

[11] W.A. Sirignano, T.K. Pham, D. Dunn-Rankin, ProcCombust. Inst. 29 (2002) 925–931.

[12] D.C. Kyritsis, I. Guerrero-Arias, S. Roychoudhury,Gomez, Proc. Combust. Inst. 29 (2002) 965–972.

[13] S.A. Kaiser, D.C. Kyritsis, P.A. Dobrowolski, M.BLong, A. Gomez, J. Mass Spectrom. Soc. Jpn.(2003) 42–49, Special Issue in Honor of Nobel Lareate J.B. Fenn.

[14] D.C. Kyritsis, S. Roychoudhury, A. Gomez, in: The 3Joint Meeting of the U.S. Sections of the CombustInstitute, Chicago, 2003.

[15] R.N. Carter, P. Menacherry, W.C. Pfefferle, G. MuenS. Roychoudhury, SAE Paper 980672, 1998.

[16] T. Edwards, L.Q. Maurice, J. Prop. Power 17 (200461–471.

[17] K. Tang, A. Gomez, J. Colloid Interface Sci. 184 (199500–511.

[18] K. Tang, A. Gomez, Phys. Fluids A 6 (1994) 2317.[19] R.A. Yetter, I. Glassman, H.C. Gabler, Proc. Combu

Inst. 28 (2000) 1265–1272.[20] A. Violi, S. Yan, E.G. Eddings, A.F. Sarofim, S

Granata, T. Faravelli, E. Ranzi, Combust. Sci. Tenol. 174 (2002) 399–417.

[21] D.J. Cookson, C.P. Lloyd, B.E. Smith, Energy Fuel(1987) 438–447.

[22] M. Cloupeau, B. Prunet-Foch, J. Aerosol Sci. 25 (191021–1036.

[23] S. Kumar, P.J. Paul, H.S. Mukunda, Proc. CombInst. 29 (2002) 1131–1137.