ILASS-Americas 30th Annual Conference on Liquid Atomization and Spray Systems, Tempe, AZ, May 2019

X-ray Characterization and Spray Measurements of ECN Spray G UsingAlternative Fuels Under Flashing Conditions

B.A. Sforzo∗, A. Tekawade, K.E. Matusik,A.L. Kastengren, J. Ilavsky, K. Fezzaa, and C.F. Powell

Argonne National LaboratoryLemont, IL 60439 USA

AbstractA suite of detailed measurements both inside and outside the injection nozzle for the Engine CombustionNetwork “Spray G” have been collected. The dynamic pintle motion has been measured with high-speed X-ray phase contrast imaging while operating at flashing conditions. Furthermore, time-resolved fuel densitywas measured through tomographic x-ray radiography for neat iso-octane and a blend of 80% iso-octanewith either 20% butanol or 20% ethanol by volume under a flash-boiling condition. Lastly, these same fuelswere used during the collection of Ultra Small Angle X-ray Scattering (USAXS) measurements to quantifythe specific surface area of the spray in the near-nozzle region. Time-resolved imaging through the injectorbody provide the pintle motion profile during heated operation. Orthogonal views consistently show full liftheight of 53 microns, with transverse motions on the order of 5 microns, which represent a non-negligibleportion of the gap between the pintle and seat, a potentially influential blockage to the uniform flow anddistribution of fuel to the orifices. Spray density profiles generated through tomographic reconstructionof projected radiography measurements reveal the high dispersion of the fuel under the flashing condition.This injection behavior also causes the individual plumes to collapse and interact with each other, whichcan lead to additional flow field perturbations. These maps also show density discrepancies between the fuelblends, caused by their differing vaporization profiles. The spray differences between these fuels are furtherhighlighted through the specific surface area profiles of droplets in the plumes. The pintle motion informationserves as influential factors in the performance of the injector, while the combination of radiography andUSAXS measurements provide a valuable quantification of the resulting spray morphology.

∗Corresponding Author: bsforzo@anl.gov

Introduction

Social and environmental concerns continuallyencourage improvements toward combustion effi-ciency and reduced emissions. For light-dutyand passenger automobiles, Gasoline Direct Injec-tion(GDI) systems offer many theoretical benefitsover older gasoline technologies, like port-fuel injec-tion, including improved efficiency and reduction inCO2 output [1]. As a result, GDI has been imple-mented by many manufacturers, but with unantic-ipated drawbacks. For instance, particulate matter(PM) emissions are greater than expected [2] dueto issues associated with poor mixing and wall wet-ting [3], and measurable unburned hydrocarbons andCO are the result of incomplete combustion.

One approach to overcome these emissions issueshas been the adoption of bio-alcohol fuel blends [4].Components such as ethanol, and second generationfuels such as butanol [5], also have the added bene-fits of being a renewable resource and reduce depen-dency on fossil fuels [6].

Notably, ethanol fuel blends, such as E30, havedemonstrated improvements to the operational effi-ciency of GDI engines [7]. Furthermore, use of bu-tanol blends (B33) have also resulted in improvedthermal efficiency and PM emissions reduction [4].Despite these improvements, the predictive capabil-ity of incorporating ethanol or butanol over a rangeof conditions is not well formulated [8, 4].

This system-level performance is influenced bymany parameters and the in-cylinder processes mustbe well understood. Research has already been es-tablished to characterize the combustion processesof blends with bio-derived alcohols [9, 10, 11]. How-ever, these reaction mechanisms are dependent onthe chemical mixture which are affected by the fueland air delivery, atomization, and evaporation pro-cesses. As such, an understanding of the fuel in-jection dynamics is a prerequisite for prediction ofthe combustion and therefore the system-level per-formance.

For these GDI conditions, it is common thatthe fuel introduction occurs at elevated tempera-tures and partial vacuum for many control schemes.With many gasoline distillates and the alcohols ofinterest, this can amount to a flash-boiling condi-tion. In practice, flash boiling occurs when theliquid is delivered at high pressure to the hot in-jector and then undergoes a rapid decompression,leaving it in a metastable, superheated state [12].Although flashing is a common approach to en-hancing atomization[13], historical work has focusedon single-component liquid flashing, which includescharacterization of evaporative cooling associated

with the process [14]. In binary mixtures, super-heated fluids exhibit bubble growth during depres-surization which “explode” upon interaction witheach other [15]. This bubble growth influence alsosupports the observation that small amounts of dis-tributed gas can greatly enhance atomization [16].

These particular conditions, as applied to theinjection process for GDI systems, has resulted inmorphological changes to the fuel atomization andmass distribution. Considerable work has experi-mentally characterized the plume behavior for multi-hole injectors and single-component fuels, where,upon transitioning through a flash-boiling condition,the individual plumes interacted with each other andcollapsed [17, 18]. This plume interaction and sub-sequent spray collapse have a relevant effect on thefuel distribution at the beginning and end of injec-tion [19], and therefore, physical models have beendeveloped to relate the plume spreading and collapsebehavior to the injector, injection, and fuel prop-erties [20]. These recent studies have all centeredon conditions and geometries defined by the EngineCombustion Network (ECN) [21], which serves as ahub for experimental and modeling findings for in-ternal combustion engine research. Following theexperimental findings, high-fidelity computationalfluid dynamics models have been examined underflashing and non-flashing conditions to reveal sim-ilar plume behavior and unique internal flow phe-nomena, which call for additional investigations intothe internal and near-nozzle flow physics [22].

This work will focus on the near-nozzle spraybehavior for alcohol fuel blends at flashing condi-tions. The ECN “Spray G” injector, which has beenwell-characterized previously at non-flashing condi-tions [23], will be studied here. X-ray radiography,collected at the Advanced Photon Source (APS) atArgonne National Laboratory, will be exploited forfuel density measurements in the near-nozzle regiondue to the minimal influence of scattering [24] and itsability to generate highly quantitative measurementsof fuel mass distribution. Also, specific surface areavalues will be computed from ultra-small angle X-ray scattering (USAXS) measurements. The resultsof these two techniques will be combined to yield aSauter mean diameter (SMD) quantification in thesprays of interest.

Methodology

Experiments were carried out at the 7-BM, 9-ID, and 32-ID beamlines of the APS to collect sprayradiography, USAXS, and imaging measurements,respectively [25, 26]. For the radiography and US-AXS techniques, the ECN Spray G #016 injector

2

was mounted in a test vessel fitted with polymidewindows for X-ray optical access. The injector wassupplied with fuel at 20 MPa and issued into thechamber which was held at partial vacuum (50 kPa),with a purging flow of nitrogen at 4 and 9.5 stan-dard L·min−1 for radiography and USAXS measure-ments, respectively. The operating conditions are aderivative of the ECN Spray G2, “flash-boiling” con-ditions, outlined in Table 1 as “G2-cold,” since theambient gas is unheated.

Table 1. Target conditions for radiography and US-AXS experiments.

Spray G2-coldFuel 100% iso-octaneInjector temperature, C 90Ambient temperature, K 298Injection pressure, MPa 20Ambient pressure, kPa 50Injection command, ms 0.680

Three fuel blends were tested at this flashingcondition: iso-octane, and blends of 80% iso-octane+ 20% n-butanol by volume, or 80% iso-octane +20% ethanol by volume. Properties of single compo-nents were sourced from reference tables [27, 28, 29].Densities of blends were measured, while viscosityand surface tension were calculated based on a linearblending of the components, weighted by the molefraction of each. These properties of the fuels areprovided in Table 2 for reference, where, ρ is theliquid density taken at 20C, Tb is the boiling tem-perature, µ75 is the viscosity at 75C, and σ50 is thesurface tension at 50C.

High speed imaging

High speed X-ray imaging was performed at the32-ID beamline at the APS [30]. As illustrated inFig. 1, the high intensity beam, originating from theundulator source, was sized to a 2.00 × 1.80 mm(H×V) rectangular cross-section with beam definingslits. A mechanical chopper was synchronized withthe imaging and injection systems to temporally gatethe incoming beam and limit thermal loading onthe hardware. As the beam passed through the op-erating fuel injector, in this case, Spray G #028,the intensity was attenuated and also slightly re-fracted, giving rise to phase-contrast intensity vari-ations. The transmitted beam was then incident ona scintillator, which fluoresced in the visible spec-trum. This light was then imaged using a 5x micro-scope objective, coupled to a Photron SA-Z high-speed camera, resulting in an imaging resolution

of 3.95 µm·px−1. The camera gathered frames at60 kHz over the duration of a given injection, and40 injections were collected at each viewing angle.Two orientations were collected in order to gatherpintle motions in two transverse planes. Further-more, motion was collected for the Spray “G2-cold”condition listed in Table 1, as well as a higher ambi-ent pressure of 600 kPa, which is commonly referredto as the Spray “G-cold” condition since the ambienttemperature is not elevated.

x-raybeam spray chamber

profiling slits

X Y

Z

mechanical chopper

scintillator

camera

Figure 1. Layout of imaging configuration.

Following image collection, a cross-correlationalgorithm was used to track the axial and trans-verse motions of the pintle, to provide mean andstandard deviation histories for each viewing per-spective. The experimental configuration and post-processing methodologies are also detailed in priorworks [31, 32]. An exmaple projection image of theinjector is depicted in Fig. 2 with a box outliningthe region used for the cross-correlation tracking al-gorithm. A location on the side of the check-ball wasused to improve the accuracy of the lateral motiontracking.

Figure 2. Projection image of the Spray G #28injector in the 90 orientation at full lift. The boxindicates the region used for cross-correlation.

Radiography

The diagnostic layout for X-ray radiographymeasurements collected at the 7-BM beamline are

3

Table 2. Properties of tested fuel blends (bold) and their components.

ρ, kg·m3 Tb,C µ75, mPa·s σ50, mN·m

100% iso-octane 692.7 99.3 0.289 18.77100% n-butanol 810.0 117.3 0.833 22.69100% ethanol 789.0 78.4 0.476 19.8980% iso-octane, 20% n-butanol 714.5 0.458 19.7080% iso-octane, 20% ethanol 710.8 0.366 19.05

illustrated in Figure 3. The application of theradiography methodology to fuel sprays is thor-oughly described in previous works [24, 33]. Specifi-cally here, the synchrotron-produced X-ray light wasconditioned to 8.0±0.01 keV mean photon energy(4% ∆E/E) using a double-multilayer monochroma-tor. The incoming beam was sized to a 1.3 × 1.3 mmsquare cross-section with beam-defining slits and theincident beam intensity, I0, was measured using a di-amond monitor. The beam was then focused usinga pair of Kirkpatrick-Baez [34] mirrors to a waist of4 × 6 µm (V × H) located at the tip of the injec-tor at the experimental origin. After transmissionthrough the spray region of interest, the beam in-tensity, I, was measured with a PIN diode. Thisintensity was measured in synchronization with therepetitive injection timing and time averaged overthe storage-ring orbit period of 3.68µs, defining thetemporal resolution of the measurement. At eachspatial location, this time history was recorded andensemble averaged for 32 injection events to improvethe signal-to-noise.

x-raybeam

PINdiode

K-B focusing mirrors

spraychamber

guard slitsI0

XY

Z

θ

Figure 3. Layout of the radiography experiment.

Following the Beer-Lambert law, the projectedmass of the spray (M) within the X-ray path wascomputed from the beam intensity measurementsas described by:

M (x, y, θ, t) = − 1

µlog

[I (x, y, θ, t)

I0 (t)

], (1)

where µ is the mass attenuation coefficient, whichwas empirically collected. The parameters I0 (t) and

I (x, y, θ, t) are the measured beam intensities of theincident beam and the transmitted beam, respec-tively. These measurements were collected for a se-quence of locations, transverse to the injector axis,at a fixed distance. Furthermore, transverse scanswere collected at a set of rotation angles, θ of the in-jector to capture projected mass profiles at multipleviewing angles.

Ultra-small-angle X-ray scattering

USAXS measurements were carried out at the9-ID beamline [26] of the APS, using a Bonse-Hart instrument as schematized in Figure 4. Inthis configuration, the incoming X-ray light hasbeen conditioned to 21 keV photon energy (λ =0.59 A) using a monochromator, and profiled to50 × 500 µm (V × H) with beam-defining slits. TheX-rays then interact with a pair of Si (220) crys-tals arranged as a collimating optic that subjects thelight to multiple Bragg reflections, thus resulting in abeam with minimal angular divergence. The beamintensity is then measured with a windowless ionchamber for high accuracy without introduction ofparasitic scattering. The beam then passes into thetest chamber, and interacts with the pulsed spray,scattering the photons at small angles.

X-ray beam

beam defining

slits

Si (220)collimating

crystals

ion chamber

spraychamber

Si (220)analyzing crystals

Photo-diode

X

Y

Z

Θ

I0

Figure 4. Layout of the USAXS experiment.

During the spray characterization measure-ments, the injector was commanded for a 1.3 msinjection duration and acquisition was synchronizedto occur during 1 ms of the steady-state portion ofinjection. A total integration time of 40 ms was usedfor each measurement. Background signal measure-

4

ments were collected 100 ms prior to injection eventsto account for the influence of lingering spray in thechamber between events.

The next pair of Si (220) analyzing crystals canprecisely rotate to reject light that does not meet theBragg condition at a range of angles, or scatteringvectors, q defined by,

q =4π

λsin Θ, (2)

where, λ is the X-ray wavelength, and Θ is halfof the angle between the incident and scatteredbeam. The intensity of this angularly-selected lightis then measured with a photodiode. To observe thescattering profile, the intensity was measured forscattering vectors -6.45 × 10−5 ≤ q ≤ 0.050 A−1.The scattering intensity, I (q) can be measured andrelated to the differential cross-section through thefollowing relation [35],

I (q) = I0 τ T εΩdΣ

dΩ(q) , (3)

where, I0 is the incident X-ray intensity, τ is thesample thickness, T is the sample transmission, ε isthe efficiency of the detector, Ω is the detector solidangle, and dΣ/dΩ is the differential scattering cross-section per unit volume per unit solid angle. All butthe differential cross-section are characterized for theexperiment, resulting in a direct comparison of thisquantity to the scattering vector, q.

Signal collected for q < 1× 10−4 A−1 wasindistinguishable from the background signal, as aresult of the angular precision of the instrument,the incident beam divergence, and the bandwidth.Furthermore, for q > 1× 10−2 A−1, weak signalprohibited usable measurements above this range.Of the two limiting cases [36] of correlations be-tween the scattering vector and the differentialcross-section, Porod’s Law [37], applicable to largeq, was exploited for the present study. In thisregime, the X-ray wavelength is small comparedto the particle size, and the differential scatter-ing cross-section follows a q−4 relation, as defined by,

dΣ

dΩ(q) = 2π∆ρ2 S q−4, (4)

where, S is the surface area per beam area, and∆ρ is the scattering length contrast of the parti-cles compared to the surrounding medium,is deter-mined analytically. As applied here, this relation-ship assumes that the particles are homogeneous,

relatively dilute, and locally non-interacting. Fur-thermore, it is assumed that the particles are ran-domly oriented which is expected to hold for thisgasoline-type spray. Thus, using the Irena softwarefor analyzing the USAXS measurements [38], thespray surface area can be deduced for each probevolume.

Results and discussion

Needle motion

Tracking of the check-ball of the Spray G #28injector yielded the lift and lateral motions plottedin Fig. 5, for two ambient pressure conditions and fu-eled with iso-octane. The Zlift plot shows the steady-state needle lift of -53.0 ± 1.2 µmfor the 50 kPa con-dition and -51.2 ± 1.7 µmfor the 600 kPa condition.The axial motion appears to be relatively repeat-able for this injector and over this range of condi-tions. However, the wobble plots illustrate some os-cillatory lateral motion. Though the guide geometryfor the check-ball constrains the transverse motion,there appears to be up to 5 µmmotion in all lateraldirections from the original axial position. Thoughthis accounts for less than 10% of the lift, due to thegeometry of the seat region, the narrowest distancebetween the ball and seat may be less than 40 µm,suggesting that a 5 µmwobble may alter the flow-field around the valve. Furthermore, although theneedle motions with different fuels is not presentedhere, these wobble effects may be enhanced throughthe viscosity variability of the fuels, or the dischargecharacteristics [39].

Fuel distribution

The projected mass profiles, collected at in-dividual rotation angles θ, were applied to gen-erating X-ray tomographic radiography, as de-scribed in previous studies for characterizing asym-metric sprays [40]. Similarly, a penalized maxi-mum likelihood algorithm was used to reconstructthe ensemble-averaged density fields at each timestep [41]. The resulting tomography slices are pre-sented in Figure 6 for the three fuels, time-averagedover the steady portion of the spray (0.6 - 0.8 ms af-ter command). The profiles are depicted in the ECN“Primary” view [21], and the ECN plume numberingis provided in the figure.

These profiles provide a transverse slice of thefuel mass per volume for each condition. Compar-ing each case, there is consistency in the asymmetryof fuel distribution, where higher peak density val-ues are measured for holes 5 and 6. The distributionof mass for these peak values is more concentratedfor the neat iso-octane, potentially indicating a con-tinued breakup and distribution of a more cohesive

5

60

50

40

30

20

10

0

Z lift

, m

50 kPa600 kPa

5

0

5

X wob

ble,

m

0.0 0.2 0.4 0.6 0.8Time, ms

5

0

5

Y wob

ble,

m

Figure 5. Pintle lift and lateral motion plots forSpray G#28 operating at two pressures with neatiso-octane.

liquid jet at this downstream location.Furthermore, the separation between the plumes

is more distinct for the neat iso-octane and the blendwith n-butanol than in the ethanol blend. This is anindication of a more promoted expansion and there-fore plume collapse phenomena as observed at flash-boiling conditions [19].

Referring to Table 2, the property that standsout within these fuel blends is the boiling tempera-ture for the constituents, with ethanol (Tb=78.4 C)being well below the 90 C operating temperatureof the injector. The iso-octane and n-butanol con-stituents have boiling temperatures above the op-erating temperature, though the vaporization canstill be enhanced by the sub-atmospheric chamberpressure. Since the hot and pressurized conditionwithin the injector would cause the ethanol to be ina superheated state as it emerges from the nozzle, arapid expansion would take place during injection,which could enhance the jet velocity above that ofthe other two blends. This hypothesis is also sup-ported by the collapsing plume and homogeneousmass distribution within the plumes visible for the

42024YECN, mm

4

2

0

2

4

X ECN

, mm

0.001.332.674.005.336.678.009.3310.6712.00

(g

mm

3 )

42024YECN, mm

4

2

0

2

4

X ECN

, mm

0.001.332.674.005.336.678.009.3310.6712.00

(g

mm

3 )

42024YECN, mm

4

2

0

2

4

X ECN

, mm

0.001.332.674.005.336.678.009.3310.6712.00

(g

mm

3 )

Figure 6. Tomographic reconstructions of X-rayspray radiography, time-averaged between 0.6 and0.8 ms after command, and plotted in the ECN “Pri-mary” orientation.

6

0.0 0.5 1.0 1.5Time, ms

0

20

40

60

80

100

120TI

M,

g/m

m100% Iso-Octane80% Iso-Octane, 20% Ethanol80% Iso-Octane, 20% Butanol

Figure 7. Transverse integrated mass of X-rayspray radiography for each fuel blend.

ethanol case in Figure 6.The spray behavior differences can also be quan-

tified by integrating these transverse slices over thespatial domain. This transverse integrated mass(TIM) is derived as,

TIM (t) =

∫ ∫ρfuel (x, y, t) dy dx, (5)

where, ρfuel is the fuel mass per volume as obtainedthrough the tomographic reconstructions at a givenaxial slice. This time-resolved, integrated quantityis plotted for each of the blends in Figure 7. Here,the histories for the neat iso-octane and the blendwith n-butanol are similar, while the ethanol blendhas a consistently lower TIM.

Theoretically, within the ∼4% variability in liq-uid densities (ρ), the TIM histories should match forthe same discharge behavior and also be conservedas the spray develops downstream (+Z). Consider-ing the proximity to the nozzle, a discrepancy in theTIM would indicate a difference in the supply massflow rate of liquid, or a change in jet velocity. Sincethe inflow conditions and the injection commandswere controlled between cases, the latter case is morelikely.

It is observed in diesel sprays, at low ambientdensity, that the TIM is consistent at different axialdistances, most likely due to the minimal amount

of entrainment of the ambient gas [24]. In contrast,the GDI design for enhanced entrainment causesa decrease in TIM with axial distance [42], whichcan be used to compute the relative change in themass-averaged velocity, vma between distances [43].Similarly, for a given axial location and accountingfor liquid density differences, this comparison canbe made between conditions to characterize therelative average velocity as defined by:

vma

vref=

TIMref/ρrefTIM/ρ

. (6)

The time-averaged TIM during the steady-state por-tion of injection for the fuel blends was compared tothe neat iso-octane as the reference condition, ac-commodating for different liquid densities. As illus-trated in Figure 7, the n-butanol blend differs onlyby 1.7% from the iso-octane, whereas, the mass-averaged velocity for the ethanol blend is 11.8%greater.

Specific surface area

To further understand the nature of the spraysand the breakup process, droplet sizing measure-ments for the same fuels and conditions were col-lected. Following previous work aimed at simplify-ing droplet surface area measurements [44, 45], thedetailed measurements of complete rocking curves ateach spatial location were forgone for a subset withina particular condition. It was realized that the loga-rithm of Equation 4 results in a linear relationship,

log

[dΣ

dΩ(q)

]= −4 log(q) + log

(2π∆ρ2 S

), (7)

where, the slope of the relationship is fixed, and thespecific surface area, S, represents an offset in theintensity. At a given condition, generation of onlya single rocking curve is necessary [44, 45], and atother spatial locations along a transverse scan, thescattered intensity was measured at a fixed scatter-ing vector, q = 3 × 10−4 A−1. Therefore, this profileof measurements could be scaled by the specific sur-face area determined for the detailed measurementthrough the q-curve fitting process.

With this simplification, a detailed USAXSrocking curve was collected at only one to threetransverse (Y) locations for each axial (Z) distance,while a complete profile of intensities was collectedat a fixed q for 200 transverse locations. Profileswere collected and scaled for the three fuel blendsat several axial distances from the injector, as illus-trated in Figure 8.

Considering that the profiles were scanned inthe Y-dimension, the expected projection for this

7

4 2 0 2 4y [mm]

0

50

100

150

200

250

300

350

400Sp

ecifi

c Su

rface

are

a [m

m2 /m

m2 ] Z = 1.00

Z = 2.00Z = 5.00Z = 10.00

4 2 0 2 4y [mm]

0

50

100

150

200

250

300

350

400

Spec

ific

Surfa

ce a

rea

[mm

2 /mm

2 ] Z = 1.00Z = 2.00Z = 5.00

4 2 0 2 4y [mm]

0

50

100

150

200

250

300

350

400

Spec

ific

Surfa

ce a

rea

[mm

2 /mm

2 ] Z = 1.00Z = 2.00Z = 5.00

Figure 8. Transverse profiles of the specific sur-face area for each of the fuel blends at multiple axialdistances.

ECN “Primary” view is to resolve three plumes,which can be seen in some of the cases and axialdistances. However, the absolute values for specificsurface area vary greatly between the different fuels,and as a function of axial distance within a givenfuel case. It is noted that the mean value for sur-face area remains the most consistent for iso-octaneas the scans progress in axial distance. Whereas,for the n-butanol blend, the near-nozzle measure-ments are very low, suggesting that there are likelylarge, unbroken ligaments of liquid that do not con-tribute to the scattering signal as much as smallerdroplets. Progressing downstream, though, the n-butanol blend has a marked rise in surface area at2 mm, and then a lower absolute signal at 5 mm,where the spray has also widened. In contrast, thesurface area values for the ethanol blend are one totwo orders of magnitude higher than the other fuelcases at Z=1 mm. This could be caused by signifi-cant disintegration of the liquid into many particleswith interfacial boundaries contributing to the scat-tering signal. At the 2 mm and 5 mm locations, thissignal has decreased, suggesting that those dropletshave either coalesced or evaporated.

Particle sizing

Assuming that there are only spherical particlesin our measurement volume, specific surface areameasurements can be combined with the measure-ment of liquid volume in the probe region by X-ray radiography to yield the Sauter mean diameter(SMD, d32) of the droplets through the following re-lation:

d32 = 6V

S, (8)

where, S was collected through the USAXSmeasurements, and V obtained from radiography.This computation was made using transverse US-AXS scans taken at Z = 1 mm and the correspond-ing radiography projections for each fuel condition.These SMD profiles are plotted in Figure 9 on asemi-log scale to highlight the variability betweenthe conditions.

When plotted as SMD, the characteristic plumeseparations are suppressed, which may be causedby the relatively large cross section of the beam,or that the measurements are line-of-sight projec-tions through the spray. The transverse plot of SMDalso highlights the differences in the atomization ofthe different fuel blends. In particular, the diame-ter of iso-octane droplets at this near-nozzle locationappear to be on the order of 10 µm, which agreeswith other works at these elevated temperatures and

8

3 2 1 0 1 2 3Y-Location, mm

10 1

100

101

102SM

D,

m

100% Iso-Octane80% Iso-Octane, 20% EtOH80% Iso-Octane, 20% BuOH

Figure 9. Transverse profiles of SMD for each of the fuels at Z = 1 mm.

partial-vacuum conditions [46]. Additionally, theobservation of larger droplets from the n-butanolblend are supported by the higher viscosity and sur-face tension, which cause resistance to breakup. Thehigher boiling temperature for n-butanol may alsoprovide contrast to the other two cases, which maylimit the effect of flashing in this blend. Converselyin the ethanol blend, the previously-discussed massdistribution and spray surface area, and here, thederived SMD, point to a flash-boiling-promoted dis-charge and severe fragmentation of the liquid iso-octane.

Summary and Conclusions

This work has investigated the near-nozzle spraybehavior for a GDI injector, operating at elevatedinjection temperatures, into a sub-atmospheric en-vironment. Three fuel blends were tested: neat iso-octane and blends of 80% iso-octane with either 20%n-butanol, or 20% ethanol, by volume.

Internal needle motion characterization was car-ried out for the Spray G #28 injector at the flash-ing (G2-cold) and non-flashing (G-cold) conditions.Though the fuel effects were not examined in thismetric, the injector mechanism provides a repeatablelift profile for the tested conditions. Furthermore,the lateral motion of the check-ball was tracked toreveal motions that could account for over 10% ofthe fluid passage height, potentially influencing theinternal flowfield. Other studies have demonstratedan influence of fluid properties on needle motion, andmay be an important topic for future investigations.

Quantitative spray radiography measurementswere made and processed using computed tomog-raphy to deduce the fuel mass per volume map-

ping in the transverse plane. The density planefor iso-octane showed the most concentrated peakintensities, while the ethanol blend had the mostdispersed plumes. Furthermore, the interaction be-tween plumes is more apparent for the ethanol blendat the measured axial distance of 1 mm, indicatingthat the tendency for spray collapse is greater forthis case. The transverse integrated mass historiesshowed good agreement between the neat iso-octaneand n-butanol blend, while the ethanol blend exhib-ited an 11% greater mass-averaged velocity over theneat iso-octane case.

Additionally, ultra-small-angle X-ray scatteringmeasurements were collected following a recently pi-oneered efficient methodology. This allowed for thecollection of specific surface area profiles across thespray at multiple axial distances. At 1 mm, then-butanol blend showed a markedly lower surfacearea than the neat iso-octane, with increased val-ues farther downstream, indicating a progressive at-omization. On the other hand, the ethanol blendhad much higher surface area at 1 mm, indicating athoroughly atomized droplet field.

The combined radiography and USAXS mea-surements were used to compute the Sauter meandiameter profile for the common axial position of1 mm. The range of values for each fuel conditionfurther illustrates the difference in atomization be-havior in reference to neat iso-octane, where the n-butanol blend has larger droplets, which correlateto the higher viscosity, surface tension, and boilingpoint of the added alcohol. By contrast, the vastlysmaller near-nozzle droplets from the ethanol blendwith the correlation to much lower boiling temper-ature of ethanol points to a significantly different

9

atomization process of the liquid.Although the general topic of flash-boiling

sprays has been thoroughly studied, much workis yet to be done on the impact to the atom-ization mechanism in multi-component fuels. Asevidenced here, the properties of different distil-lates influences the break-up mechanics of the spray.This work serves to provide quantitative metrics forcomparison in model development of blended fuelsprays at hot injection and partial vacuum condi-tions. Through this, improved predictions of atom-ization characteristics, and thereby combustion per-formance, can be leveraged in the design for appro-priate fuel blending and injector design.

Acknowledgements

This research was conducted as part of theCo-Optimization of Fuels & Engines (Co-Optima)project sponsored by the U.S. Department of En-ergy (DOE) Office of Energy Efficiency and Renew-able Energy (EERE), Bioenergy Technologies andVehicle Technologies Offices. Research presented inthis paper was performed at the 7-BM, 9-ID, and32-ID beamlines at the Advanced Photon Source atArgonne National Laboratory. Use of the APS issupported by the U.S. Department of Energy (DOE)under Contract No. DEAC0206CH11357. Argonnefuel spray research is sponsored by the DOE VehicleTechnologies Program under the direction of Gur-preet Singh and Kevin Stork.

Nomenclature

SMD = Sauter mean diameter = d32TIM = Transverse integrated mass∆ρ = Scattering length contrast compared to

the mediumε = Detector efficiencyλ = Incident light wavelengthµ = Mass absorption coefficientµ75 = Viscosity at 75CΩ = Detector solid angleρ = Liquid density at 20Cρfuel = Fuel mass per volumeσ50 = Surface tension at 50Cτ = Sample thicknessΘ = Scattering half-angled32 = Sauter mean diameterI = Measured beam intensityI0 = Reference beam intensityM = Projected mass of the fuelq = USAXS scattering vectorS = Total surface area per volume of sampleT = Sample transmissionTb = Boiling temperaturevma = Mass-averaged velocity

vref = Reference velocity

References

[1] F. Zhao, M-C Lai, and D. L. Harrington.Progress in Energy and Combustion Science,25(5):437 – 562, 1999.

[2] H. Zhao. Advanced Direct Injection Combus-tion Engine Technologies and Development, vol-ume 1, chapter 1 - Overview of gasoline directinjection engines, pp. 1 – 19. Woodhead Pub-lishing, 2010.

[3] Georgios Karavalakis, Daniel Short, Diep Vu,Robert L. Russell, Akua Asa-Awuku, Hee-jung Jung, Kent C. Johnson, and Thomas D.Durbin. Energy, 82:168 – 179, 2015.

[4] C. Hergueta, M. Bogarra, A. Tsolakis, K. Essa,and J.M. Herreros. Fuel, 208:662 – 672, 2017.

[5] Stefan Heyne and Simon Harvey. Applied En-ergy, 101:203 – 212, 2013. Sustainable Develop-ment of Energy, Water and Environment Sys-tems.

[6] Ayhan Demirbas. Applied Energy, 86:S108 –S117, 2009. Bio-fuels in Asia.

[7] Christopher P Kolodziej and Thomas Wallner.Combustion Engines, 56, 2017.

[8] Felix C.P. Leach, Richard Stone, DavidRichardson, James W.G. Turner, AndrewLewis, Sam Akehurst, Sarah Remmert, StevenCampbell, and Roger Cracknell. Fuel, 225:277– 286, 2018.

[9] S.M. Sarathy, M.J. Thomson, C. Togb, P. Da-gaut, F. Halter, and C. Mounaim-Rousselle.Combustion and Flame, 156(4):852 – 864, 2009.

[10] Liming Cai and Heinz Pitsch. Combustion andFlame, 162(5):1623 – 1637, 2015.

[11] Katharina Kohse-Hoinghaus, Patrick Oßwald,Terrill A. Cool, Tina Kasper, Nils Hansen, FeiQi, Charles K. Westbrook, and Phillip R. West-moreland. Angewandte Chemie InternationalEdition, 49(21):3572–3597, 2010.

[12] A. Solomon, S. Rupprecht, L.-D. Chen, andG. Faeth. Atomization and combustion prop-erties of flashing injectors. 20th Aerospace Sci-ences Meeting, Aerospace Sciences Meetings.American Institute of Aeronautics and Astro-nautics, January 1982.

10

[13] Tali Bar-Kohany and Moti Levy. Atomizationand Sprays, 26(12):1259–1305, 2016.

[14] Ralph Brown and J. Louis York. AIChE Jour-nal, 8(2):149–153, 1962.

[15] Eran Sher and Chaim Elata. Industrial & En-gineering Chemistry Process Design and Devel-opment, 16(2):237–242, 1977.

[16] Arthur H Lefebvre and Vincent G McDonell.Atomization and sprays. CRC press, 2017.

[17] G Zhang, M Xu, T Li, RO Grover, T Kuo, andY He. ILASS Americas 26th Annual Confer-ence on Liquid Atomization and Spray Systems.ILASS Americas, 2014.

[18] Matthew Blessinger, Julien Manin, Scott ASkeen, Maarten Meijer, Scott Parrish, andLyle M Pickett. International Journal of En-gine Research, 16(2):238–252, 2015.

[19] Panos Sphicas, Lyle M Pickett, Scott A Skeen,and Jonathan H Frank. International Journalof Engine Research, 19(10):1048–1067, 2018.

[20] J. Lacey, F. Poursadegh, M.J. Brear, R. Gor-don, P. Petersen, C. Lakey, B. Butcher, andS. Ryan. Fuel, 200:345 – 356, 2017.

[21] Sandia National Laboratories. Engine Com-bustion Network gasoline spray combus-tion. https://ecn.sandia.gov/gasoline-spray-combustion/, 2019. Accessed March 1, 2019.

[22] E.T. Baldwin, R.O. Grover, S.E. Parrish, D.J.Duke, K.E. Matusik, C.F. Powell, A.L. Kasten-gren, and D.P. Schmidt. International Journalof Multiphase Flow, 87:90 – 101, 2016.

[23] Daniel J. Duke, Alan L. Kastengren,Katarzyna E. Matusik, Andrew B. Swan-tek, Christopher F. Powell, Raul Payri, DanielVaquerizo, Lama Itani, Gilles Bruneaux,Ronald O. Grover, Scott Parrish, Lee Markle,David Schmidt, Julien Manin, Scott A. Skeen,and Lyle M. Pickett. Experimental Thermaland Fluid Science, 88(Supplement C):608 –621, 2017.

[24] Alan Kastengren and Christopher F. Powell.Experiments in Fluids, 55(3):1686, Feb 2014.

[25] Alan Kastengren, Christopher F. Powell, DohnArms, Eric M. Dufresne, Harold Gibson, andJin Wang. Journal of Synchrotron Radiation,19(4):654–657, Jul 2012.

[26] Jan Ilavsky, Pete R. Jemian, Andrew J. Allen,Fan Zhang, Lyle E. Levine, and Gabrielle G.Long. Journal of Applied Crystallography,42(3):469–479, Jun 2009.

[27] Robert L. Brown and Stephen E. Stein. NISTChemistry WebBook, NIST Standard ReferenceDatabase Number 69, volume 69, chapter Boil-ing point data. NIST, 2001.

[28] A. A. H. Pdua, J. M. N. A. Fareleira, J. C. G.Calado, and W. A. Wakeham. Journal ofChemical & Engineering Data, 41(6):1488–1494, 1996.

[29] David R Lide. CRC handbook of chemistry andphysics. CRC press, 2005.

[30] Qun Shen, Wah-Keat Lee, Kamel Fezzaa,Yong S. Chu, Francesco De Carlo, PeterJemian, Jan Ilavsky, Mark Erdmann, andGabrielle G. Long. Nuclear Instruments andMethods in Physics Research Section A: Accel-erators, Spectrometers, Detectors and Associ-ated Equipment, 582(1):77 – 79, 2007. Proceed-ings of the 14th National Conference on Syn-chrotron Radiation Research.

[31] Juan P. Viera, Raul Payri, Andrew B. Swantek,Daniel J. Duke, Nicolas Sovis, Alan L. Kasten-gren, and Christopher F. Powell. Energy Con-version and Management, 112:350 – 358, 2016.

[32] Alan L. Kastengren, F. Zak Tilocco, Christo-pher F. Powell, Julien Manin, Lyle M. Pickett,Raul Payri, and Tim Bazyn. Atomization andSprays, 22(12):1011–1052, 2012.

[33] Alan L. Kastengren, F. Zak Tilocco, DanielDuke, Christopher F. Powell, Xusheng Zhang,and Seoksu Moon. Atomization and Sprays,24(3):251–272, 2014.

[34] Paul Kirkpatrick and A. V. Baez. J. Opt. Soc.Am., 38(9):766–774, Sep 1948.

[35] G. G. Long, P. R. Jemian, J. R. Weertman,D. R. Black, H. E. Burdette, and R. Spal.Journal of Applied Crystallography, 24(1):30–37, Feb 1991.

[36] KC Lin, Michael Ryan, Campbell Carter, AlecSandy, Suresh Narayanan, Jan Ilavsky, and JinWang. ILASS Americas, 21st Annual Confer-ence on Liquid Atomization and Spray Systems,pp. 18–21, 2008.

11

[37] Jens Als-Nielsen and Des McMorrow. Elementsof modern X-ray physics. John Wiley & Sons,2011.

[38] Jan Ilavsky and Peter R. Jemian. Journalof Applied Crystallography, 42(2):347–353, Apr2009.

[39] R Torelli, BA Sforzo, KE Matusik, AL Kas-tengren, K Fezzaa, CP Powell, S Som, Y Pei,T Tzanetakis, Y Zhang, et al. Investigation ofshot-to-shot variability during short injections.ICLASS 2018. ILASS Americas, 2018.

[40] Daniel J. Duke, Andrew B. Swantek, Nicolas M.Sovis, F. Zak Tilocco, Christopher F. Pow-ell, Alan L. Kastengren, Doa Grsoy, and TekinBier. SAE International Journal of Engines,9(1):143–153, 2016.

[41] Doga Gursoy, Francesco De Carlo, XianghuiXiao, and Chris Jacobsen. Journal of Syn-chrotron Radiation, 21(5):1188–1193, Sep 2014.

[42] Zihan Wang, Andrew Swantek, Riccardo Scar-celli, Daniel Duke, Alan Kastengren, Christo-pher F. Powell, Sibendu Som, Ronald Reese,Kevin Freeman, and York Zhu. SAE In-ternational Journal of Fuels and Lubricants,8(1):147–159, 2015.

[43] Alan L. Kastengren, Christopher F. Powell, Yu-jie Wang, Kyoung-Su Im, and Jin Wang. Atom-ization and Sprays, 19(11):1031–1044, 2009.

[44] K. E. Matusik, B. A. Sforzo, H. J. Seong, D. J.Duke, A. L. Kastengren, J. Ilavsky, and C. F.Powell. X-ray measurements of fuel spray spe-cific surface area and sauter mean diameterfor cavitating and non-cavitating diesel sprays.ICLASS 2018. ILASS Americas, 2018.

[45] K. E. Matusik, B. A. Sforzo, H. J. Seong, D. J.Duke, A. L. Kastengren, J. Ilavsky, and C. F.Powell. Atomization and Sprays, 2019. UnderReview.

[46] P.G. Aleiferis, J. Serras-Pereira, Z. van Ro-munde, J. Caine, and M. Wirth. Combustionand Flame, 157(4):735 – 756, 2010.

12