Author's personal copy - web.njit.edudreyzin/Papers-pdf/CNF160-10-Oct2013-Al-Mg-H... · The outer...

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

http://www.elsevier.com/authorsrights

Author's personal copy

Combustion of fine aluminum and magnesium powders in water

Amy Corcoran a, Stefano Mercati b, Hongqi Nie a, Massimo Milani b, Luca Montorsi b, Edward L. Dreizin a,⇑a Otto H. York Department of Chemical, Biological, and Pharmaceutical Engineering, New Jersey Institute of Technology, Newark, NJ 07102, United Statesb DISMI, University of Modena and Reggio Emilia, Via Amendola 2, Pad. Morselli, Reggio Emilia 42122, Italy

a r t i c l e i n f o

Article history:Received 4 March 2013Received in revised form 24 April 2013Accepted 25 April 2013Available online 24 May 2013

Keywords:Metal combustionParticle burn timeHydrogen productionUnderwater propulsion

a b s t r a c t

Micron-sized metal powders carried by a nitrogen flow were fed along the axis of a cylindrical hydrogen/oxygen diffusion flame. The particles ignited and burned in the water vapor at approximately 2500 K.Experiments were performed at atmospheric pressure. The environment in which particles burned wascharacterized in detail using computational fluid dynamics. The computations confirmed that the metalpowders burned in water while the effect of oxygen and other oxidizing species could be neglected. Com-bustion was characterized experimentally for micron-sized powders of both aluminum and magnesium.Particle size distributions were measured using low-angle laser light scattering. Optical emission of theburning particles was recorded using filtered photomultiplier tubes. Measured durations of individualparticle emission pulses were assumed to represent their burn times; these data were classified into log-arithmically spaced time bins. The distribution of the particle burn times was correlated with their sizedistributions assuming that larger size particles burned longer. It was observed that correlation betweenthe burn times, t, and particle diameters, D, can be approximately described as t � D0.64 and t � D0.68 foraluminum and magnesium powders, respectively. The results were compared to previous reports andpossible reasons for discrepancies between the present and earlier results were discussed.

� 2013 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

1. Introduction

Metal powders have high combustion enthalpies and thus arewidely used as components for pyrotechnics, explosives, and solidpropellants for both air and underwater propulsion systems [1–6], where they react with such oxidizers as O2, CO2, and H2O.More recently, both low- and high-temperature reactions of met-als with water were explored in systems aimed at producinghydrogen and electric energy, with most studies focused on Aland Mg [5–16]. Compared with the traditional hydrogen produc-tion plants based on natural gas steam reforming, producing CO2

with rates close to 7.5 kgCO2/kgH2, the systems based on metal/water combustion are theoretically free of greenhouse gas emis-sion [8,10,11].

Design of both traditional energetic systems involving combus-tion of Al and Mg and novel hydrogen and energy generatingplants requires detailed understanding and quantitative descrip-tion of dynamics of respective combustion processes. However,there are relatively few studies presenting quantitative data onburn rates of Al and Mg in water. Most of the published reportspresent experimental data and correlations derived for coarsemetal particles, e.g., particles coarser than 20 lm were used in

Refs. [17,18], and 50-lm particles were used in Refs. [18–20]. Apopular correlation proposed by Beckstead [17] for burn timesof aluminum particles in different environments is largely basedon experiments with coarse powders. The correlation might notremain useful for fine particles, although fine particles will likelybe employed in practical systems, where dimensions of thecombustion chambers are limited, and thus shorter reaction timesare desired. Reactions of fine aluminum particles in a heatedsteam were recently addressed using a shock tube [21,22] andlaser ignition experiments [23]. In the shock tube studies, aerosol-ized particle clouds reacted in high temperature water (2650 K) atelevated pressures (4–13 atm). The burn times were recovered forthe entire clouds, based on their emission signatures. In laser igni-tion tests, individual particles burned in a gas composed of 77% ofwater and 23% of nitrogen and heated slightly above the waterboiling point. Additional measurements focused on the effect ofaluminum particle size on its burn rate are desired in environ-ments with different concentrations of water. We were unableto find reports on combustion of fine magnesium particles inwater.

The present research investigates the dynamics of combustionof aluminum and magnesium powders in high temperature waterenvironment (around 2500 K), generated by a H2/O2 flame, atatmospheric pressure.

0010-2180/$ - see front matter � 2013 The Combustion Institute. Published by Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.combustflame.2013.04.019

⇑ Corresponding author.E-mail address: [email protected] (E.L. Dreizin).

Combustion and Flame 160 (2013) 2242–2250

Contents lists available at SciVerse ScienceDirect

Combustion and Flame

journal homepage: www.elsevier .com/locate /combustflame


2. Experimental

2.1. Approach

Experiments were performed with fine powders of aluminumand magnesium. Particles were injected into products of a hydro-gen/oxygen flame as schematically shown in Fig. 1. Particle emis-sion signatures were recorded using an array of fourphotomultiplier tubes (PMT’s) equipped with interference filters[23]. Assuming that larger particles burn longer the measured dis-tribution of particle burn times was correlated with the distribu-tion of particle sizes. A computational fluid dynamics (CFD)model was developed to describe the oxidizing environment forthe present experiments.

The fiber optics bundle connected to the PMT array was placedabout 50 mm away from the burner and about 30 mm above itstop. This placement produced optimal results for capturing emis-sion from entire particle streaks while minimizing the noise pro-duced by the flame. The PMT outputs were recorded and savedat a rate of 100,000 samples per second using a 16-bit PCI-6123data acquisition board by National Instruments and Lab-view soft-ware. The data were processed using a custom Matlab code.

2.2. Powder feeder

A custom screw feeder described in more detail elsewhere [24]was used to feed the metal particles into the burner. To coat thescrew with powder, the dry powder was placed in a plastic dishand weighed. Using a short bristle brush, the powder was depos-ited into the threads of a stainless steel, 3=4

00 (19.05 mm) diameterscrew with 16 threads per inch. A typical powder load was 0.2 g,which was placed over about 8-cm length of the screw. After thepowder had been applied, the remaining weight was recorded todetermine the sample mass to be used in experiments. The screwloaded with powder was then placed into a cylindrical enclosureand attached to a DC-motor. A narrow nitrogen jet blowing acrossa thread of a loaded screw removed powder coating and fed it intothe burner. The feed rate of powder was determined by the rota-tional speed of the screw. It was pre-selected in preliminary exper-iments to ensure production of as many particle streaks as possibleduring the data acquisition time while minimizing overlappingpeaks produced by multiple particles burning simultaneously.

2.3. Burner

The design and dimensions of the burner are shown in Fig. 2. Athin stainless steel tube placed at the center of the burner is usedto inject particles from the powder feeder carried by nitrogenflowing at 3.07 standard cubic feet per hour (SCFH) or 1.45 l/min. The nitrogen flow is surrounded by a flow of hydrogen fedthrough a tube with narrowed internal diameter at 9.04 SCFH(4.27 l/min). Oxygen flow is produced around the hydrogen flowto prevent the particles from reacting with oxygen directly. Oxy-gen is fed at 12.22 SCFH (5.77 l/min) through the outer cylinder ofthe burner. The outer cylinder contains several layers of 3-mmdiameter steel balls (top layer shown at the bottom of Fig. 2) thathelp in producing a laminar oxygen flow. The flame shape is sen-sitive to the position of the central tube carrying the nitrogenflow, which must be carefully centered to achieve the desiredsymmetric flame configuration. The produced flame is laminar.Individual rotameters are used to monitor hydrogen, nitrogen,and oxygen flow rates.

2.4. Materials

Commercially available powders of Al and Mg were used. Bothaluminum powders were spherical, including Al 3–4.5 lm by AlfaAesar (nominal size 3–4.5 lm) and X65 powder by Alcan-ToyoAmerica Inc., (nominal size 6.5 lm). The magnesium powder usedwas 99.8% purity, -325 mesh by Alfa Aesar Stock #10233. Thispowder is coarser in comparison to the aluminum; particles arenon-spherical but generally equiaxial. Scanning Electron Micro-scope (SEM) images for all three powders are shown in Fig. 3. Com-pared to X65 powder, the particle sizes appear to vary in a broaderrange for the Al 3–4.5 lm powder, including many agglomeratedsmall particles.

The particle size distributions for all three powders were mea-sured using a Beckman–Coulter LS230 Enhanced Particle Analyzerand are shown in Fig. 4. In agreement with the SEM image, the sizedistribution for the Al 3–4.5 lm includes a significant amount ofsmall particles. The size distribution for the X65 powder is rela-tively narrow. The magnesium powder is substantially coarserthan aluminum. As noted above, these measured particle size dis-tributions are correlated with the burn time durations for eachpowder.

Fig. 1. Schematic of the experimental apparatus.

Fig. 2. Design of the burner.

A. Corcoran et al. / Combustion and Flame 160 (2013) 2242–2250 2243


3. H2/O2 flame computational fluid dynamics (CFD) simulation

The objective of the experiments was to characterize the burnrates of aluminum and magnesium particles in water. A CFD simu-lation was carried out to describe the gas composition, velocity,and temperature profiles within and around the produced flame,where particle combustion occurred. The model considered onlythe gas-phase flame with no particles following the Eulerian ap-proach. The numerical simulation used a commercial CFD codeSTAR CCM+ 7.06, developed by CD-ADAPCO [25].

3.1. Computational domain

The computational domain is shown in Fig. 5 in relationship tothe burner configuration. The fluid domain around the burner isdescribed with a 2D axisymmetric mesh. The fluid domain repre-sents an open environment that includes the burner; it is chosento be 150 mm along the radial direction (approximately six timesthe radius of the burner) and 500 mm from the top of the burner(about five times the experimental flame height). These dimen-sions enabled us to describe boundary regions sufficiently far fromthe flame and predict stable velocity profiles in the flame vicinity.An inset in Fig. 5 shows a magnified view of the adopted mesh inthe main reaction region.

The final mesh is composed by 140,000 polyhedral elementscharacterized by dimensions that span from 1.0 mm close to theoutlet boundaries (far from the flame region) to 0.03 mm in theflame region. The grid refinement is carried out using different vol-ume shapes enabling one to obtain a smooth grading of the celldimension [25].

3.2. Modeling

The H2/O2 flame was described using a reaction model based onthe adiabatic equilibrium presumed probability distribution func-tion (PPDF) that is specific for non-premixed combustion (assum-ing that the reaction kinetics is fast compared to transportprocesses). In the adiabatic PPDF model, the atomic concentrationsand the temperature determine the value of any scalar at any spa-tial location. Since the number of atoms in the reactor does notchange because of mass conservation, the reaction state spacecan be characterized by a single conserved scalar: the mixture frac-tion. An assumption of this model is that the turbulent diffusivity ismuch greater than the molecular diffusivity, so that the moleculardiffusivity was considered negligible. A detailed discussion of thePPDF model is available in Refs. [26,27]. In the reaction model,all chemical species involved in the hydrogen combustion were ac-counted for: H, H2, H2O, HO2, H2O2, O, OH, and O2. The properties ofspecies were calculated using NASA polynomials [28].

To approach the problem, a multi component ideal gas modelwas selected. Due to the low flame emissivity, the flame radiationheat losses were neglected, and the burner walls were consideredadiabatic. The effect of gravity is accounted for. The gas mixture

Fig. 3. SEM images of the powders used: left to right: Al 3–4.5 lm, 20-lm scale bar; Al-X65, 20-lm scale bar; Mg-325 mesh, 40-lm scale bar.

Fig. 4. Particle size distribution for three powders used in experiments.

Fig. 5. Computational domain for CFD modeling.

2244 A. Corcoran et al. / Combustion and Flame 160 (2013) 2242–2250


dynamic viscosity, l, was calculated using the Sutherland’s law (airas a reference gas):

ll0¼ T

T0

� �3=2 T0 þ ST þ S

� �ð1Þ

where T0 = 273.15 K is a reference temperature, l0 = 1.716e�5 Pa sis the air viscosity at T0, and S = 111 K, is the Sutherland tempera-ture [29].

The simulations were carried out under steady state conditions.The turbulence effects were modeled using the Reynolds-AveragedNavier–Stokes k-omega shear stress transport turbulence approach[30]. Attention was paid when constructing the mesh to obtain thecorrect values for the non-dimensional wall distance for a wall-bounded flow, y+, corresponding to the turbulence model selected.

3.3. Boundary conditions

The air surrounding the burner was modeled as an ideal gasmixture composed by 76.7 mass% of N2 and 23.3 mass% of O2.The oxygen flow was defined by introducing a boundary conditionwith a constant gas velocity at the bottom of the computationaldomain (Fig. 5). The flows of hydrogen and nitrogen were definedat the respective burner outlets, cf. Figs. 2 and 5. Further details onthe boundary conditions set according to the experimental flowrates are given in Table 1.

In all simulations the convergence was achieved when residualsstabilized at values below 10�4 for each independent variable. Theconvergent velocity and temperature profiles were additionally in-spected to identify any unusual patterns.

A qualitative comparison of the flame shape and the predictedconcentration field of H2O is shown in Fig. 6. The general flameshape and dimensions are reproduced reasonably well. More quan-titative validations of the present model are discussed below.

3.4. Validation of the CFD results

The CFD results were validated using two different experi-ments. The first experiment measured the temperature profilewithin the flame. The second experiment quantified the gas veloc-ity in a selected flame region.

To measure the flame temperature profile, a thin ceramic fila-ment was inserted into the flame (with no powder fed into it),and its emission was measured using an EPP2000 High ResolutionSpectrometer by StellarNet Inc. Prior to inserting the filament intothe flame, the spectrum from the flame emission was recorded andtreated as the baseline signal. The filament inserted into the flamewas allowed to equilibrate before its emission spectrum was ac-quired. The measurements were performed at several filamentheights, until the filament emission became weak and comparableto the background flame emission. Each recorded spectrum wasapproximated by a gray body emission spectrum using the emit-ter’s temperature as an adjustable parameter. As a result, the ver-tical temperature profile was obtained. Because the hottest part ofthe heated filament produced the strongest emission signal, themeasured values represent the maximum flame temperaturesachieved at each height. These values were compared to the max-imum temperatures predicted to exist at each height in the CFD

model. The comparison is shown in Fig. 7. The experimental andcomputed temperature profiles are consistent with each other.Both the plateau and the decay of the temperature at increasedheights are predicted well.

The gas velocity was measured using particle image velocime-try. A vertical, 532-nm laser sheet modulated with a frequency of7500 Hz was placed at the flame. Streaks of aluminum-X65 parti-cles injected into the flame and illuminated by the laser sheet werephotographed. A 532-nm interference filter was placed in front ofthe camera lens to minimize the effect of the particle thermalemission. The dashed particle streaks were processed to recoverthe vertical components of their velocities. The results of measure-ments are shown in Fig. 8 along with the predicted velocity pro-files. For most particles, their vertical velocity components agreewell with the predictions. Some of the particles, however, are ob-served to move significantly faster, which is attributed to the effectof aluminum combustion, well known to be associated with accel-eration of the burning particles.

In summary, both temperature and velocity measurements cor-relate well with the predictions of the present CFD model. There-fore, the CFD model-predicted concentration profiles are alsolikely to represent well the actual environments existing in thepresent experiments.

Table 1Boundary conditions used to describe flows in the model.

Gas Temperature (K) Flow rate (l/min) Velocity (m/s)

N2 293.15 1.45 5.33O2 293.15 5.77 0.065H2 293.15 4.27 5.23

Fig. 6. A photograph of the H2/O2 flame (left) and a CFD-predicted water molefraction contour plot.

Fig. 7. Maximum temperature in the H2/O2 flame as a function of height above theburner.



4. Predicted temperature, concentration, and velocity profiles

The CFD simulation results are shown in terms of temperature,velocity, and molar distributions of species involved in thereaction.

Vertical temperature profiles and concentration profiles for themain oxidizers present in the flame, H2O and O2, are shown inFig. 9 for different radial positions. Near the burner, the tempera-ture is high at the radial distances of about 4–8 mm, quickly drop-ping further away from the burner’s axis. For smaller radialdistances, the temperature exceeds 1500 K approximately 40 mmabove the burner.

The water concentration is around 30% right above the burnerat the radial distances of 4–6 mm. Closer to the center of the flame,the water concentration exceeds 10–15% at about 40 mm abovethe burner. At the radial distances less than 8 mm, the water con-centration remains above 20% up to about 100–110 mm above theburner. Oxygen concentrations are high right above the burner atthe radial distances exceeding 8 mm. Close to the center, with axial

distances less than 4 mm, the oxygen concentration remains below1–1.5% up to 60 mm above the burner. Thus, the effect of oxygen asan oxidizer can be neglected for particles burning within 4 mm ra-dius around the burner axis and consumed at the heights lowerthan 60 mm from the top of the burner.

Vertical profiles for mole fractions of all species produced in theH2/O2 flame along the burner’s axis are shown in Fig. 10. In addi-tion to illustrating once again that the water concentration is sub-stantially higher than that of oxygen in the region where theparticles are expected to burn, the plots also show that presenceof other potential oxidizers, such as O and OH, can be neglected.

Vertical profiles for the vertical and horizontal components ofthe gas velocity are shown in Fig. 11. In the flame region, the ver-tical velocity component is about an order of magnitude greaterthan its horizontal component. However, the horizontal velocitycomponent is not entirely negligible. It may cause some ‘‘focusing’’of the particle jet in the region between 4 and 6 mm from the flameaxis, as a result of the flows of oxygen and hydrogen towards theflame sheet. This effect was, indeed observed from close inspectionof the particle trajectories. However, such effects are predicted tooccur (and were observed) in a relatively narrow range of heights;for the most part, the particle trajectories are defined by the verti-cal component of the gas velocity.

Fig. 8. Vertical component of the gas velocity as a function of the radial distance forlocations 60–70 mm above the burner.

Fig. 9. Vertical profiles of temperature and main oxidizing species at different axialdistances.

Fig. 10. Vertical profiles of the species produced in the H2/O2 flame along theburner’s axis.

Fig. 11. Vertical profiles for the vertical and horizontal components of the gasvelocity above the burner for different radial positions.



5. Experimental results

Multiple images of the ignited particle streaks were taken andprocessed to identify locations where they ignited and burned.An example of a characteristic photograph used in the processingas well as locations of particle ignition recovered from multiplesimilar images are shown in Fig. 12. The scales for both parts ofFig. 12 are the same, so that the particle ignition locations can bereadily visualized relative to the flame region. Most particles igniteat the heights from 30 to 60 mm from the top of the burner andwithin ±4 mm from its axis. Some particles ignite lower and higherthan this region, very few particles are observed to ignite at the ax-ial distances greater than 5 mm. Comparing these results with thedata shown in Fig. 9, it is clear that most particles ignite and burnin water with negligible effect of oxygen on their burn rates.

Particle emission was measured with a substantial backgroundemission produced by the flame. Recorded traces were processedto select peaks with sharp rise and clear decline back to the base-line level. Particle emission peaks were also observed to have char-acteristic shapes separating them from the flame emissionfluctuations. Characteristic particle peaks are shown in Fig. 13 forall three powders tested. The peaks for 3–4.5 lm aluminum pow-der were relatively weak, so that the signal to noise ratio was rel-atively low. Aluminum-X65 powder typically produced clean,strong peaks with well-distinguished patterns corresponding todifferent combustion stages. An initial increase in the signal emis-sion was accompanied by an oscillatory pattern, as shown inFig. 13. The signal then declined and stabilized at a lower level be-fore extinguishing, while oscillations in emission intensity contin-ued. Qualitatively, such emission traces are similar to thoseobserved for aluminum burning in different environments earlier

[23,31]. For magnesium, emission peaks typically showed a sharpincrease in intensity followed by gradual decline. To measure burntimes, the signals were processed for more than 250 individualparticle peaks for each powder. The burn times were determinedby the duration the signal stayed above the background flameemission level.

Recorded signals were sufficiently strong to roughly evaluatecolor temperatures for aluminum-X65 and magnesium powders.Traces recorded at 532 and 589 nm were used. Resulting tempera-tures varied in broad ranges, probably illustrating that strongersignals were needed for reliable measurements. Despite the largevariation, it could be concluded that the peak flame temperaturesfor the aluminum-X65 and for magnesium were respectively closeto 2600 and 2300 K.

The burn times measured from individual particle peaks weresorted into logarithmic bins and graphed as a histogram plottingthe number of particles in each time bin. The results are shown

Fig. 12. Locations of particle ignition identified from processing selected flameimages (height above the burner is x and axial position is y). A characteristic flameimage used in the analysis is shown on the right. The scales are the same for boththe image and plot.

Fig. 13. Typical particle emission peaks for aluminum 3–4.5 lm, aluminum-X65,and magnesium -325 mesh.

Fig. 14. Statistical distributions of the measured burn times for different powders.



in Fig. 14. It was assumed that larger size particles burn longer, sothere is a direct correlation between the particle size distributionsshown in Fig. 4 (cast in terms of particle number density) and burntimes shown in Fig. 14. To justify this approach, an additional cor-relation, between particle burn times and emission intensities wasalso considered. It was expected that larger particles, characterizedby longer burn times, would also produce higher amplitude emis-sion pulses. As shown in Fig. 15, this assumption works for mostparticles. Removal of a few outliers from the trends shown inFig. 15 did not result in a noticeable change in the observed effectof particle size on their burn times, shown in Fig. 16.

Details of a procedure used to correlate particle size distribu-tions (Fig. 4) and burn times (Fig. 14) are discussed elsewhere [24].

The burn times generally vary in the range of 1–10 ms. The ef-fect of powder particle sizes on burn times is relatively weak; thestraight lines shown in the log–log plot in Fig. 17 are close tot � d0.65 trends for all powders. Comparing magnesium and X65aluminum powder, it appears that Mg particles burn somewhatlonger than the same size Al.

There is a discrepancy between the trends observed for two alu-minum powders. It appears that particles from the 3–4.5 lm pow-der have longer burn times compared to the same size particlesfrom the X65 aluminum. These longer burn times are thought tobe produced by agglomerated particles, rather than indicative oftrue difference in combustion rates between the two aluminumpowders. Indeed, the particle size distribution of the 3–4.5 lmpowder shows a substantial presence of submicron particles; atthe same time, the SEM images show that such fine particles oftenadhere to the larger particles, as seen in Fig. 3. The agglomeratescan be broken during particle size measurements, when the pow-der in solution is actively agitated by ultrasound; however, theagglomerate might survive passing through the powder feeder inthe present combustion experiments. Particles in X65 aluminumhave a narrower size distribution and fewer fines. Their SEMimages show that they are less likely to agglomerate and skewthe correlations between the measured burn time distributionand particle sizes. Based on the above consideration, further dis-cussion of burn times for aluminum is focused only on the exper-imental trend obtained for X65 powder.

6. Discussion

The results are discussed separately for magnesium and alumi-num powders.

Relatively few reports identified magnesium particle burn timesin water [18–20]; all such reports dealt with powders that werecoarser compared to those used here. Results of the present exper-iments are shown in Fig. 17 together with several data sets foundin the literature. The experiments described here result in longerburn times and a weaker effect of powder size on the burn timecompared to those expected for fine Mg particles based on the lit-erature data. Using a conventional t � Dn law, where t is the burntime and D is particle diameter, present experiments are describedreasonably well when n � 0.68.

A recent study [20] used very large 1.5–5 mm particles ignitedwith a hot wire. Water vapor balanced with argon was suppliedinto the combustor chamber. Partial water pressure was0.05 MPa. A high-speed camera was used to observe the flamestructure and combustion process. The results produce a trend linedescribed by t � D2 relationship.

Interestingly, results from earlier experiments [18,19] fit wellwith the t � D2 trend line reported in Ref. [20]. Prachukho et al.[19], produced water vapor using a flat flame H2/O2 burner withan addition of nitrogen to vary the water concentration. The flamewas placed in an enclosed environment. Water temperature wasassessed to be about 1373 K. Magnesium particles were fed intothe burner using a vibrating feeder and ignited within the flame.The particle sizes of the powder used varied between 1 and250 lm; however, the powder was size classified into six fractions:24–40, 37–64, 66–116, 86–133, 102–152, and 144–205 lm. Burntimes were recorded on a continuously moving film and measuredusing a rotating disk as a shutter. Ozerov and Yurinov [18] used the

Fig. 15. Correlations between the measured burn times and maximum emissionintensities for different powders.

Fig. 16. Correlations of the measured burn times and particle sizes for differentpowders tested.

Fig. 17. Combustion time as a function of particle size for magnesium powders.



same technique for igniting particles; experiments were performedat higher temperatures (up to 2073 K) and water concentrationreaching 100%. Particles with sizes from 66 to 185 lm were used.

Surprisingly, the burn times measured for relatively coarse par-ticles in Refs. [18,19] are in the same range as the burn times mea-sured in the present effort, although for substantially finerparticles. The discrepancy may be explained considering that theentire length of the reaction chamber in experiments [18,19] wasless than 250 mm. It is suggested that larger particles, which burnlonger simply did not have enough time to complete combustionwhile traveling through the chamber. This could significantly skewthe measurement results. Other possible issues may relate to insuf-ficiently well characterized oxidizing environment produced bythe flat flame burner. The experimental details provided in Refs.[18,19] are insufficient to quantify possible presence of oxygenand other oxidizers. Finally, the size classification of the 1–250 lm powder into narrow size fractions is notoriously difficult;the actual size distributions of the obtained size fractions are notreported and could have been skewed by the presence of fine par-ticles, which are difficult to distinguish using optical microscopy.

The combustion behavior of aluminum particles in water hasbeen studied in more details; both experimental data and empiri-cal correlations are available to describe the respective burn rates.Present data as well as some of the earlier results and correlationsare shown in Fig. 18. The present results are reasonably well de-scribed by a t � D0.64 trend.

Using the same equipment as discussed above for Mg powders,Ozerov and Yurinov [18] investigated combustion of aluminumparticles in water vapor at 2073 K at one atmosphere. Aluminumparticles with the average size of 120 lm were used. Ozerov’s re-sult and proposed correlation are shown in Fig. 18.

Recently, fine Al particles were ignited in a shock tube [21,22].Burn times were determined as a function of the pressure and themole fraction of the primary oxidizers, including water, oxygenand carbon dioxide. In the reflected shock, the aerosolized alumi-num particles were subjected to a high temperature (2650 K) andhigh pressure (8.5 atm) environment. The duration of the combus-tion emission pulses at 489 nm was associated with the burntimes. Bazyn et al. [22] used pure spherical aluminum powders se-lected by sieving between 5 and 10 lm; the sample mass distribu-tion showed a strong peak around 10 lm. The water mole fractionvaried approximately between 0.25 and 0.55. Lynch et al. [21] ana-lyzed two different powders: a spherical powder with the massdistribution close to 3 lm and a sample of powder selected by siev-ing between 5 and 10 lm with a mass distribution close to 11 lm.The water mole fraction varied between 0.35 and 0.65. In Fig. 18,results of the shock tube studies are shown with large error barsaccounting for the width of the size distributions for the sievedpowders.

In experiments [23], that are, perhaps, most relevant to thepresent study, a nitrogen gas stream fed aluminum powder intoa 393-K water vapor environment. The particles were ignited usinga focused CO2 laser beam. The mole fraction of water vapor pro-duced was 0.77. The experiments used a commercial sphericalpowder with the nominal particle sizes of 10–14 lm and individ-ual particle burn times were identified. The results appear tomatch with the present measurements for larger particles; how-ever, the effect of particle size on the burn time is weaker andsmaller particles observed in Ref. [23] to burn longer than in thepresent experiments. The longer burn times observed in Ref. [23]are likely associated with a much lower environment temperature.

Beckstead [17] proposed an empirical correlation for the alumi-num particle burn times. Approximately 400 experimental datapoints available in literature and collected from different experi-ments carried out by different authors were used. These data werecollected for particles with sizes spanning from 20 to 750 lm. Thiscorrelation defines the aluminum burning time, t (ms) as a func-tion of the particle size, D (lm), the oxidizer mole fraction, C, theenvironmental pressure, p (atm), and the temperature, T (K):

tb ¼aDn

ðCO2 þ 0:6CH2O þ 0:22CCO2 Þp0:1T0:2 ð2Þ

Beckstead [17] also suggested that the data are best correlatedwith coefficients a = 0.00735 and n = 1.8.

The Beckstead’s correlation was used to produce trend linesshown in Fig. 18; the experimental conditions used in all other ref-erences considered were employed, resulting in several parallellines (for all cases, n = 1.8).

As expected, early experiments with coarse particles are welldescribed by the Beckstead’s correlation. It is difficult to make aquantitative comparison with the shock tube data, some of whichseem to correlate well within the proposed trend, while others ap-pear to deviate from it substantially. The shorter burn times re-ported from the shock tube studies compared to the presentexperiments are likely explained by the effect of elevated pressure.This effect, proposed to be very weak by the Beckstead’s correla-tion, may become stronger for fine particles because of tworeasons:

– the heat and mass transfer processes occur in the transitionregime [32,33] and shift in pressure results in qualitativechanges in the transport mechanisms;

– the pressure may also have a stronger effect for kineticallydriven reactions, which are becoming more likely for fineparticles [21].

The present results as well as results from Ref. [23] deviate fromthe Beckstead’s correlation significantly. The measured burn timesare about an order of magnitude longer than predicted; the effectof particle size is also much weaker than the D1.8 trend expectedbased on the experiments with coarse powders. As mentionedabove, changes in both transport and reaction mechanisms occur-ring for small particles are likely responsible for the observed com-bustion behavior. The weak effect of particle size on burn timeobserved here is consistent with our earlier experimental reportsstudying aluminum combustion in oxygen and other environments[23,31,34,35]. A reaction model accounting for both surface reac-tion rates and transition transport processes is desired to interpretthe present results quantitatively.

7. Conclusions

Combustion of fine magnesium and aluminum powders wascharacterized in a high temperature water vapor. For the selected

Fig. 18. Combustion time as a function of particle size for aluminum powders(results for X65 powder for present experiments are shown).



experimental configuration, effect of other oxidizers on the mea-sured particle burn rates could be neglected. The compositionand temperature profiles for the gaseous environment used inexperiments were validated by detailed CFD calculations. For bothmetals, the particle combustion times could be approximately de-scribed by Dn law; the exponent n was close to 0.64 for aluminumand to 0.68 for magnesium. The results deviate substantially fromearlier experimental data for both metals burning in water; there isalso substantial discrepancy with a popular empirical correlation[17] proposed for aluminum burn times. The measured burn timesare longer than implied by previous work employing coarse pow-ders; this discrepancy is attributed to differences in both transportand chemical reaction mechanisms for fine metal particles. Thepresent results are qualitatively consistent with recently reportedtrends describing burn time as a function of particle size for finealuminum particles burning in oxygen and other oxidizers[23,31,34,35].

Acknowledgment

This work was supported by Defense Threat Reduction Agency.

References

[1] T. Miller, J. Herr, in: 40th AIAA/ASME/SAE/ASEE Joint Propulsion Conferenceand Exhibit, American Institute of Aeronautics and Astronautics, 2004, doi:http://dx.doi.org/10.2514/6.2004-4037.

[2] T.B.R. Foote, J.P., J.T. Lineberry, in: G.D. Roy (Ed.), Advances in ChemicalPropulsion: Science to Technology, vol. 1, 2002, pp. 133–147.

[3] T.R. Sippel, T.L. Pourpoint, S.F. Son, Propellants, Explos., Pyrotech. 38 (1) (2013)56–66.

[4] T.L. Pourpoint, T.D. Wood, M.A. Pfeil, J. Tsohas, S.F. Son, Int. J. Aerospace Eng.2012 (2012) 11 (Art # 874076).

[5] G. Risha, T. Connell, R. Yetter, V. Yang, T. Wood, M. Pfeil, T. Pourpoint, S. Son, in:45th AIAA/ASME/SAE/ASEE Joint Propulsion Conference & Exhibit, AmericanInstitute of Aeronautics and Astronautics: 2009, doi: http://dx.doi.org/10.2514/6.2009-4890.

[6] M. Diwan, D. Hanna, E. Shafirovich, A. Varma, Chem. Eng. Sci. 65 (1) (2010) 80–87.

[7] J. Petrovic, G. Thomas, Reaction of Aluminum with Water to Produce Hydrogen,U.S. Department of Energy, 2008.

[8] H.Z. Wang, D.Y.C. Leung, M.K.H. Leung, M. Ni, Renew. Sust. Energy Rev. 13 (4)(2009) 845–853.

[9] M.S. Vlaskin, E.I. Shkolnikov, A.V. Bersh, A.Z. Zhuk, A.V. Lisicyn, A.I.Sorokovikov, Y.V. Pankina, J. Power Sources 196 (20) (2011) 8828–8835.

[10] F. Franzoni, S. Mercati, M. Milani, L. Montorsi, Int. J. Hydrogen Energy 36 (4)(2011) 2803–2816.

[11] S. Mercati, M. Milani, L. Montorsi, F. Paltrinieri, Appl. Energy 97 (2012) 686–694.

[12] P. Dupiano, D. Stamatis, E.L. Dreizin, Int. J. Hydrogen Energy 36 (8) (2011)4781–4791.

[13] H. Nie, M. Schoenitz, E.L. Dreizin, Int. J. Hydrogen Energy 37 (15) (2012)11035–11045.

[14] W.-Z. Gai, W.-H. Liu, Z.-Y. Deng, J.-G. Zhou, Int. J. Hydrogen Energy 37 (17)(2012) 13132–13140.

[15] W. Yang, Z. Han, J. Zhou, J. Liu, K. Cen, Int. J. Hydrogen Energy 36 (17) (2011)10608–10613.

[16] M.H. Grosjean, M. Zidoune, L. Roué, J.Y. Huot, Int. J. Hydrogen Energy 31 (1)(2006) 109–119.

[17] M.W. Beckstead, Combust. Explo. Shock+ 41 (5) (2005) 533–546.[18] E.S. Ozerov, A.A. Yurinov, Combust. Explo. Shock+ 13 (6) (1977) 778–780.[19] V.P. Prachukho, E.S. Ozerov, A.A. Yurinov, Combust. Explo. Shock+ 7 (2) (1971)

195–198.[20] X. Huang, Z. Xia, L. Huang, J. Hu, Sci. China Technol. Sci. 55 (9) (2012) 2601–

2608.[21] P. Lynch, H. Krier, N. Glumac, Proc. Combust. Inst. 32 (2) (2009) 1887–1893.[22] T. Bazyn, H. Krier, N. Glumac, J. Propul. Power 21 (4) (2005) 577–582.[23] R.J. Gill, C. Badiola, E.L. Dreizin, Combust. Flame 157 (11) (2010) 2015–2023.[24] A.L. Corcoran, V.K. Hoffmann, E.L. Dreizin, Combust. Flame 160 (3) (2013) 718–

724.[25] CD-adapco, USER GUIDE Star CCM+ Version 7.06, CD-adapco, 2012, p. 12179.[26] D. Veynante, L. Vervisch, Prog. Energy Combust. 28 (3) (2002) 193–266.[27] S.B. Pope, Prog. Energy Combust. 11 (2) (1985) 119–192.[28] B.J. McBride, S. Gordon, M.A. Reno, in: National Aeronautics and Space

Administration, vol. NASA Technical Memorandum 4513, Cleveland, OH, 1993,pp. 1–94.

[29] W. Sutherland, Philos. Mag. 36 (S.5) (1893) 507–531.[30] F.R. Menter, M. Kuntz, R. Langtry, in: K. Hanjalic, Y. Nagano, M. Tummers (Eds.),

Ten Years of Industrial Experience with the SST Turbulence Model, Turbulence,Heat and Mass Transfer 4, Antalya, Turkey, 12–17 October 2003, Antalya,Turkey, 2003.

[31] C. Badiola, R.J. Gill, E.L. Dreizin, Combust. Flame 158 (10) (2011) 2064–2070.[32] S. Mohan, L. Furet, E.L. Dreizin, Combust. Flame 157 (7) (2010) 1356–1363.[33] S. Mohan, M.A. Trunov, E.L. Dreizin, Combust. Flame 156 (11) (2009) 2213–

2216.[34] C. Badiola, E.L. Dreizin, Combust. Sci. Technol. 184 (12) (2012) 1993–2007.[35] E.L. Dreizin, C. Badiola, S. Zhang, Y. Aly, Int. J. Energ. Mater. Chem. Propul. 10

(4) (2011) 297–319.


Author's personal copy - web.njit.edudreyzin/Papers-pdf/CNF160-10-Oct2013-Al-Mg-H... · The outer...

Documents

Transcript of Author's personal copy - web.njit.edudreyzin/Papers-pdf/CNF160-10-Oct2013-Al-Mg-H... · The outer...