Flame suppression by aerosols derived from aqueous solutions

e

s

atives toctive flamendensed-in rate ofof the

fficiencyed with aand flowo evap-re solidsnt solu-e flame.shonate(OPA),

were in-y more/waterlivered,t waterpounds

mpounds

Combustion and Flame 141 (2005) 308–321www.elsevier.com/locate/combustflam

Flame suppression by aerosols derived from aqueousolutions containing phosphorus

T.M. Jayaweeraa,1, E.M. Fishera,∗, James W. Flemingb

a Sibley School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY 14853, USAb Combustion Dynamics Section, Naval Research Laboratory, Washington, DC 20375, USA

Received 5 September 2003; received in revised form 28 October 2004; accepted 28 October 2004

Available online 3 March 2005

Abstract

Aerosols derived from aqueous solutions containing phosphorus are investigated as possible alternhalon-based flame suppressants. Phosphorus-containing compounds (PCCs) have been shown to be effesuppressants in the gas phase; a water-based solution would provide a practical means of delivering a cophase PCC to the flame. Flame suppression was characterized by measuring the global extinction straa counterflow, nonpremixed methane–air flame with and without a PCC additive. An aqueous solutionadditive was introduced as an aerosol into a heated chamber upstream of the air flow tube. A high-enebulizer produced a polydisperse spray of droplets with a Sauter mean diameter of 8 µm, as measurphase-Doppler particle analyzer. The droplet size distribution was nearly independent of the compositionrate of the liquid. Externally maintaining the reactant streams at 360 K allowed the water in the aerosol torate prior to reaching the flame. Evaporation of water leaves behind residual particles for solutes that aat 360 K, and thus solid particulates, not droplets of solution, enter the flame zone. Comparisons of differetions with different phase solutes were used to provide an estimate of the physical effect of particles on thThree neat phosphorus-containing compounds, dimethylmethylphosphonate (DMMP), diethylmethylpho(DEMP) and dimethylphosphonate (DMP) and five 1.6% molar aqueous solutions, orthophosphoric acidphosphorus acid, phosphonic acid, methylphosphonic acid, and dimethylmethylphosphonate (DMMP),vestigated. The acid solutions (with solid solutes) all displayed similar effectiveness, and were all slightleffective, on a per mole basis, than the DMMP solution (a liquid solute). The effectiveness of a DMMPsolution is found to be a weighted average of the effectiveness of its constituents. Per mole of water dea 1.6% molar solution of a PCC is approximately twice as effective in fire suppression compared to neavapor. Flame modeling calculations for the extinction condition with added gas-phase phosphorus comusing a published phosphorus reaction mechanism grossly underpredict the total effectiveness of the coinvestigated. 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

* Corresponding author. Fax: +1(607)255-9410.E-mail address: [email protected](E.M. Fisher).

1 Present address: FM Global, 1151 Boston-Providence Tpke, Norwood, MA, USA.

0010-2180/$ – see front matter 2005 The Combustion Institute. Published by Elsevier Inc. All rights reserved.doi:10.1016/j.combustflame.2004.10.013

http://www.elsevier.com/locate/combustflame

mailto:[email protected]

T.M. Jayaweera et al. / Combustion and Flame 141 (2005) 308–321 309

sent,

hastuteyetlid-

iver-f po-iffi-ideich

entse ofcondiativege-

tye in

re-nt ina-

cedaireposp-

n isolu-

thetedizeing

as

ef-e-les,ane iselyec-beent

in-l

n a-sria-d”

ndthe

ateris

reessg ittionic-

ithof

ofam-asshanne–edeat

ddi-lyo arti-ctsup-

aterex-nd

revi-ef-

ho-ase,-d-ofthec-

byrtedPO

ofO

ondiesrad-ut-theofnse

pos-e.

1. Introduction

With the restrictions on the manufacture and uof CF3Br (halon 1301), a common fire suppressadue to its high ozone-depletion potential, therebeen widespread interest in finding a viable substi[1–4]. A search for a drop-in replacement has notbeen successful, and as an alternative, liquid- or sophase agents are currently being considered. Deling the agent as an aerosol broadens the range otential agents available, but also makes it more dcult to predict flame suppression effectiveness. Asfrom gas-phase chemical and physical effects, whoccur after vaporization of the aerosol, these agmay exhibit increased flame suppression becausvarious phenomena related to the presence of a sephase. These phenomena include enhanced radheat transfer, enthalpy of vaporization, and heteroneous reactions involving flame radicals[5–7]. Par-ticle or droplet delivery also introduces uncertainin the amount of agent present in the gas phasthe flame. Gas-phase loading in the flame can beduced from the amount of condensed-phase agethe air stream because of insufficient time for vporization in the flame, and can be either reduor increased when aerosol trajectories differ fromstreamlines[8]. Differences in the location in thflame where the agent reaches the gas phase andsible nonuniform loading can also affect flame supression effectiveness. In this work, a comparisomade between additives introduced as aqueous stions leaving residual particles which then enterflame and those that completely vaporize in the heaair stream. The experiments are designed to minimseveral of the sources of uncertainty in agent loaddescribed above.

Most of the work with solid-phase additives hfocused on NaHCO3 [9–12] or KHCO3 [13]. Thesestudies have all shown these powders to be morefective than CF3Br at fire suppression. The effectivness is highly dependent on the size of the particwith smaller particles having a greater effect thlarger ones on a per mass basis. This differencattributed to the large particles not being completconsumed in the flame zone. The high level of efftiveness from these small particles is believed todue, at least in part, to a chemical effect of the agin the gas phase[5,12].

In the liquid phase, water mists have been gaing in popularity, due to their relatively high leveof effectiveness and zero toxicity[8,14–17]. A wa-ter mist has been shown to be more effective, oper mass basis, than CF3Br in nonpremixed counterflow flames[17]. However, water acts primarily aa physical agent in suppressing the flame. A vation of the traditional water mist is an “enhance

-

water mist: a water solution containing a compouthat, through effects on flame chemistry, increasesfire suppression effectiveness over that of pure w[18–20]. A concern with adding a chemical agentthat toxicity of the solution will be higher than of puwater. However, as the chemical agent’s effectivenis supplemented by that of the water accompanyinin an enhanced water mist, its required concentramay be acceptably low from the standpoint of toxity.

A few experimental studies have been done wenhanced water mists, namely water solutionsNaOH or NaCl[19–21]. These show enhancementsflame suppression compared to pure water. For exple, a water mist enhanced with 18% NaOH by mwas shown to be several times more effective tpure water in an opposed-jet, nonpremixed methaair flame[21]. The mechanism by which an enhancwater mist results in improved suppression over nwater is believed to be due to several factors. In ation to the physical effect of the water, there is likea chemical effect by the solute, and possibly alsphysical effect from the droplets or residual pacles[19,21]. The relative magnitudes of these effehave not been determined. In our study, the flame spression effectiveness of aerosols derived from wsolutions containing phosphorus was investigatedperimentally, and contributions of the chemical aphysical effects were evaluated.

Phosphorus compounds have been studied pously in the gas phase and shown to be highlyfective flame suppressants[22–25]. Since low vaporpressure hinders the introduction of many phosprus-containing compounds (PCCs) in the gas phtheir use in practical fire fighting situations is limited. However, experimental and numerical stuies [22,26–28]have shown that the effectivenessdifferent gas-phase PCCs is similar, expandingrange of PCCs available. This similarity of effetiveness is supported by the proposed mechanismwhich PCCs suppress the flame: they are conveto phosphorus-containing radicals, such as HOand HOPO2, which catalyze the recombinationthe important combustion radicals, H, OH, and[23,25,29]. This process slows the overall reactirate and thus suppresses the flame. Earlier stuused laser-induced fluorescence to measure OHical levels in a nonpremixed flame with and withodimethylmethylphosphonate[24,30]. These measurements show a decrease in the OH level withaddition of this PCC, supporting the hypothesiscatalytic radical recombination. With the inhibitioresulting primarily from the phosphorus atom, the uof PCC solutions, as enhanced water mists, is asible means of getting the phosphorus into the flam

310 T.M. Jayaweera et al. / Combustion and Flame 141 (2005) 308–321

airsev-e airuid,ndthex-

herso-leters-up-asecan

ed-nnd

e of-ad-si-

rate

ingk-os-

hasends.dif-d be-e in

eraltheousted

in-diedte,s-yl-os-)house-ouso-

ed%arateterreso-oris

osenam-

ivesobe-be

se,ixedingzerure-

si-ture-c-cesell-exitad-. In

In the current study, a nonpremixed methane–flame was used to investigate the effectiveness oferal PCCs. These compounds are delivered into thstream in the vapor phase, as a fine mist of neat liqor as a fine mist of aqueous solution. Neat liquids aaqueous solutions involving liquid solutes, such asPCC dimethylmethylphosphonate (DMMP), are epected to evaporate fully before the flame. For otcompounds introduced as droplets of the aqueouslutions with solid solutes, a small residual particis expected to remain upon evaporation of the wafrom the solution droplet. Assuming all of the phophorus compounds have similar chemical fire spression effectiveness once they are in the gas phfire suppression effects due to the particle phasebe investigated.

Experiments were performed using an opposjet burner apparatus[31]. This configuration has beeused extensively with experimental, numerical, aanalytical techniques to evaluate the performancflame suppressants[4,32,33]. The counterflow configuration is useful in studying flame-suppressingditives as the flame is thermally isolated and quaone-dimensional along the centerline[34,35]. Flamestrength can be characterized by the global strainat extinction[36,37].

Extinction calculations were also performed usOPPDIF [38], part of the Chemkin software pacage[39]. These calculations used a proposed phphorus mechanism and investigated the gas-psuppression effect of three phosphorus compouThese calculations were done to further elucidateferences in the suppression effectiveness observetween PCCs delivered in the gas phase and thosthe particle phase.

2. Experimental method

The flame suppression effectiveness of sevPCCs, introduced into the burner system either invapor phase or as droplets of neat liquid or aquesolution, was investigated. The compounds are lis

,

with their CAS number and chemical structureTable 1, along with the form in which they were introduced into the burner system. The compounds stuwere: distilled water, dimethylmethylphosphonadiethylmethylphosphonate (DEMP), dimethylphophonate (DMP), orthophosphoric acid (OPA), methphosphonic acid (MPA), phosphonic acid, and phphorous acid. DMMP, DEMP, DMP (all 97% pureand MPA (98% pure) were supplied by AldricChemical Company, phosphonic and phosphoracids (each 97% pure) were supplied by Alfa Asar, and OPA was supplied as a premixed aquesolution from Aqua Solutions. Each of the phosphrus acids was in a 1.6% molar solution in distillwater, and DMMP was introduced both in a 1.6aqueous solution and as a neat compound, in sepexperiments. As shown below, essentially all waand DMMP evaporate before the oxidizer mixtuleaves the burner nozzle. In the case of the acidlutions, a solid residual particle remains, while fDMMP and water, the mixture entering the flameentirely gas phase. The acid compounds were chbecause they have been seen, through in situ spling and GC/MS analysis[40], to be decompositionproducts of DMMP, a PCC shown to be an effectflame suppressant[22,23]. These compounds are alattractive because of their low heating value andcause, unlike many PCCs, they are not believed toneurotoxic.

For compounds introduced in the liquid phathe experimental apparatus consists of a nonpremcounterflow burner equipped with a system for addthe liquid agent into the oxidizer stream: a nebulimounted in a large chamber, leading into temperatcontrolled tubing. The apparatus is shown inFig. 1.Its key features are small droplet size, large redence time of the reactant stream under temperacontrolled conditions, and the availability of an acurate procedure for measuring losses to surfawithin the feed system. These features lead to a wcharacterized state of the reactant mixture at theof the burner tube, and well-defined gas-phase loing of agent at the flame, as documented below

Table 1Compounds used

Compound Form CAS number Molecular formula

Orthophosphoric acid (OPA) Aqueous solution 7664-38-2 P(=O)(OH)3Phosphorous acid Aqueous solution 10294-56-1 P(OH)3Phosphonic acid Aqueous solution 13598-36-2 P(=O)(H)(OH)2Methylphosphonic acid Aqueous solution 993-13-5 P(=O)(CH3)(OH)2Distilled water Neat liquid, vapor 7732-18-5 H2ODimethylmethylphosphonate (DMMP) Neat liquid, vapor, 756-79-6 P(=O)(CH3)(OCH3)2

aqueous solutionDiethylmethylphosphonate (DEMP) Neat vapor 683-08-9 P(=O)(CH3)(OCH2CH3)2Dimethylphosphate (DMP) Neat vapor 813-78-5 P(=O)(OCH)2(OH)


Fig. 1. Schematic of experimental apparatus with opposed-jet burner, droplet generator, and chamber.

se,ingevi-

onter-jet

aretionwsry

up-re-the

thethe

e-tyb-ly.reandns ofaining.

de-

ari-inor-wo

the case of additives introduced in the vapor phathe nebulizer and chamber are replaced by a syrpump and further length of heated tubing, as preously described[22].

2.1. Burner

The flame is produced approximately 2 mmthe oxidizer side of the stagnation plane of counflowing streams of fuel and oxidizer. This opposed-burner geometry has been detailed previously[22].As in previous experiments, the reactant nozzlesstraight glass tubes (0.98 cm i.d.) with a separadistance of 0.95 cm. Methane (99% pure) fuel flofrom the lower nozzle and the oxidizer, a primastandard mixture of 21± 0.2% O2 with N2 balance,flows through the upper nozzle. All gases were splied by MG industries. The temperature of theactant streams 10 cm upstream from the exit ofnozzles are actively maintained at 360± 1 K.

The relative flame strength is characterized byglobal strain rate at extinction, calculated using

following expression:

(1)aq = 2VO

L

(1+ VF

VO

(ρF

ρO

)1/2).

In Eq. (1), L refers to the separation distance btween the nozzles,V is the average stream velociat extinction,ρ is the stream density, and the suscripts O and F refer to oxidizer and fuel, respectivePlug flow boundary conditions at the nozzle exit aassumed. This expression, derived by SeshadriWilliams [41], is referred to as the global extinctiostrain rate. To determine suppression effectivenesthe added compounds, the global extinction strrate was measured as a function of dopant loadThe resulting extinction strain ratesaq are normal-ized by the undoped value,aq0, which was found tobe 340 s−1 in the absence of the heated chamberscribed in the next section and 322 s−1 when thechamber was present (each with a day-to-day vation of less than 4%). Although this differenceabsolute extinction strain rate is significant, the nmalized extinction strain rates obtained with the tconfigurations were not significantly different.


ed-nd

ro-neusess-owan-s).-

(ox-edionwnas

er-

hefine),

t ofgasr

amf atedor-owese

cmthe

if-ant

N).

imi-thatheper-tersPAsed

tersandin-,per-00o

. Asare

s ofEN

The extinction strain rate is determined at a fixnozzle separation,L. The reactant velocities are increased until the critical strain rate is achieved athe flame is extinguished. In most studies[4,42], boththe fuel and the oxidizer velocities are increased pportionally such that the flame or stagnation plais maintained near the center of the burner. Becathe addition of a liquid-phase dopant is via a maflow syringe pump, changing the doped reactant flrate would change the dopant loading, leading to trsients of a fairly long time scale (several minuteThis difficulty is circumvented by performing extinction measurements holding the doped reactantidizer) stream fixed, and only varying the undop(fuel) stream flow rate. The flame and stagnatplane move in this method, but for a range of knoflame positions, the global extinction strain rate wfound to agree within±2% [22]. A validation ofthis technique with residual particles present was pformed for the current study.

2.2. Generation and characterization of droplet mist

A small glass nebulizer is used to deliver tliquid-phase dopant to the oxidizer stream as amist of droplets. A high-efficiency nebulizer (HENa Meinhard® nebulizer, is detailed inFig. 2. Liquid isintroduced via a programmable syringe pump. Parthe oxidizer stream is supplied as the nebulizingat a flow rate of 1.00 SLM. This flow rate allows fosufficient aerodynamic breakup of the liquid streinto a fine mist. The HEN is mounted at the top olarge heated chamber (15 cm i.d., 18 cm long) locaapproximately 75 cm upstream of the flame. Unftunately, the presence of the chamber induced flfluctuations that perturbed the flame. To dampen thfluctuations, a small diameter (3.2 mm i.d., 7.6long) tube was placed between the chamber andreactant flow tube.

In order to make valid comparisons between dferent compounds at several loadings, it is import

Fig. 2. Schematic of the high-efficiency nebulizer (HEdroplet generator, provided by Meinhard and Associates

for the droplets produced in each case to be slar. The droplets should also be small enough sothey will completely evaporate prior to reaching tflame. Measurements of droplet diameters wereformed with a phase-Doppler particle anemome(PDPA, Dantec FlowLite Fiber PDA). The HEN wanot mounted in the burner apparatus for the PDmeasurements; rather, it was mounted in an enclochamber with good optical access. Droplet diamewere measured for a variety of concentrationsflow rates of phosphorous acid, as well as for a sgle set of conditions for water, DMMP, OPA, MPAand phosphonic acid. PDPA measurements wereformed 1.3 cm downstream of the HEN tip, with 1.SLM of nebulizing gas. Typical histograms for twdifferent cases are shown inFigs. 3a and 3b; the un-certainty of the PDPA measurement is±2 µm for theoptical configuration and processor settings usedcan be seen in the histogram, all of the droplets

(a)

(b)

Fig. 3. PDPA measured size histogram for 2000 droplettwo different PCCs, measured 1.3 cm downstream of Hexit. The liquid flow rate for both compounds was 25 µl/minwith 1.00 SLM of nebulizing air. (a) Neat DMMP (b) 1.6%(molar) aqueous solution of OPA.


own-erso-.

acid,.5%eatso-the

bu-is-g

ares

fer-ntina-

o-xitra-atedthe

Thesec-

ctsate-sert be-heforob-lete

ichora-l-that

formesieveandndsingat

dofdi-theo.l ofoxi-m-ofer-ameesece azlev-

aryor

lityizeva-

s ofpar-ase

dualuredpletrti-C.foreter

Table 2The droplet diameter size distributions at 25 µl/min obtainedfrom PDPA measurements fit to a log-normal distribution

Compound A (dimensionless) x0 (µm) α (µm)

Neat H2O 162.9 1.625 0.6887.5% P(OH)3 216.8 1.844 0.5147.5% MPA 235.0 1.833 0.4697.5% H3PO3 216.8 1.844 0.5144.3% DMMP 169.4 1.611 0.6621.6% OPA 170.6 1.742 0.662Neat DMMP 213.9 1.566 0.589

Note. The parameters given fit the formAexp(−(x − x0)2/

α2), and were evaluated using a least-squares method.

Fig. 4. PDPA measured Sauter mean diameter, 1.3 cm dstream of HEN exit, as a function of liquid flow rate. Undconditions in this paper for the 7.5% phosphorous acidlution, 25 µl/min corresponds to a total loading of 1.49%The various compounds tested were 7.5% phosphorous1.6% orthophosphoric acid, 7.5% phosphonic acid, 7methylphosphonic acid, neat DMMP, 4.3% DMMP, and nH2O. All concentrations are molar based in an aqueouslution. Error bars represent one standard deviation indiameter measurement of 2000 droplets.

less than 20 µm in diameter and the size distritions for the two solutions are very similar. Size dtributions were fit to a log-normal distribution usinan equation of the formAexp(−(x − x0)2/α2). Thefit parameters were determined using a least-squmethod and are given inTable 2for all the compoundstested. The Sauter mean diameters from all the difent cases are plotted inFig. 4. The error bars represeone standard deviation of the droplet size determtion.

2.3. Evolution of droplet mist

For droplets of neat liquids and DMMP/water slutions, complete evaporation upstream of the eof the oxidizer nozzle is ensured by high tempetures and long residence times. The chamber is heelectrically, while the temperature of the gas in

burner tube is actively controlled to 360± 1 K byelectric heating of the surrounding sheath gas.average residence time of the droplets is severalonds, while a simpled2 analysis[43] of the evap-oration rate for the maximum droplet size predithat pure DMMP or water droplets should evaporentirely within a few milliseconds within the chamber. To confirm complete evaporation, a He–Ne labeam was passed through the reactant stream juslow the exit plane of the oxidizer nozzle, and toff-axis forward direction was scanned by eyescattered light. For neat solutions, there was noservable scattered light, consistent with the compevaporation of the liquid at this point.

For droplets of phosphorus acid solutions, whshould produce residual solid particles upon evaption of water, little quantitative information is avaiable on evaporation rates. However, it is expectedthese evaporation times are comparable to thoseneat liquids, and that the much longer residence tiupstream of the burner nozzle are sufficient to achphase equilibrium between the residual particlethe surrounding gas stream. Whether the compouunder consideration exhibit deliquescence (producdry residual particles) or retain some water evenlow relative humidity is not known. As calculatefrom the reactant flow rates, the relative humiditythe air stream is only 2.4% under burner exit contions, and thus the amount of water retained inparticle at equilibrium is likely to be small if not zer

For all compounds, active temperature controthe reactant streams is crucial. The agent-dopeddizer stream is heated externally to maintain its teperature at 360 K. The enthalpy of vaporizationthe water is not supplied by the flame but from extnal sources, and thus does not contribute to the flsuppression process for the liquid droplets in thexperiments. The apparatus is designed to produwell-defined state for the oxidizer stream at the nozexit: phase equilibrium at 360 K, with all phases traeling at the same velocity. This well-defined boundcondition at the nozzle exit is highly desirable fcomputational modeling.

When residual particles are present, their abito follow gas streamlines in a decelerating flow is sdependent. Size also determines whether in-flameporization is fast enough to release the entire masthe added agent into the gas phase. Clearly, smallticles are desirable to produce well-defined gas-phagent loadings for both of these reasons. The resiparticle diameter can be estimated from the measdroplet size distribution by assuming that each droloses all of its water and becomes a spherical pacle with density equal to that of the pure solid PCWith this assumption, evaporation of the dropsthese aerosols yields a Sauter mean particle diam


m.enrti-esionot

ncelet

re-c-the6%re-arem-erelloriong

l-ar-ithticlt-istye toar-

ve-ther-allainteticonde-m

toc-nt,erynlyof

fectcalcu-ofm.

tedv-

glyn-

re-essts

edx-the

anee-vedre-toasthe

ob-tes,

toNaeamdid

ns,to

st,n

n,ndi-m-

th ashthe

is-wassesred

tes

a-nsib-less is

forthat.roions.MPtsthat

of ∼3 µm and a maximum particle diameter of 7 µAlthough rapid heating of solution droplets has beshown to produce highly nonspherical residual pacles[44], the assumption of a spherical particle givconservative results for both drag and vaporizatrates. The current estimate of particle size will nbe conservative, however, if significant coalesceof droplets occurs downstream of where the dropsize distribution was measured.

Stokes numbers represent the ratio of particlesponse time to flow residence time. At typical extintion strain rates in this burner, Stokes numbers forestimated mean and maximum particle size for 1.OPA solution are calculated to be 0.02 and 0.1,spectively. These low values indicate that particlesable to follow the gas streamlines, avoiding the coplications of nonuniform loading described by othinvestigators[8]. The estimated particle sizes are wbelow 30 µm, the diameter for critical damping four flow conditions, determined using the expressdeveloped by Li et al. for estimating critical dampinin this flow field configuration[45]. Thus the particlesshould not oscillate in the flowfield.

Another indication of how well the particles folow the gas streamlines is obtained from a compison of the drag force (assuming Stokes flow) weither the gravitational force or the thermophoreforce. The terminal velocity for 8-µm particles (resuing from 20-µm droplets), relative to the gas flow,of the order of 1.5 mm/s, and for 3-µm particles iis 0.2 mm/s. A thermophoretic velocity can similarlbe calculated, by equating the thermophoretic forcthe drag force. This velocity is constant for small pticles with a Knudsen number greater than∼1 [46].At smaller Knudsen numbers, the thermophoreticlocity decreases and becomes dependent on themal conductivity of the particle. Thus, using a smparticle that satisfies the Knudsen number constrgives a conservative estimate of the thermophorvelocity. The thermophoretic velocity at the locatiof the maximum temperature gradient can berived using temperature and velocity profiles frocalculations performed under similar conditions[30].This velocity, for a 0.1-µm particle, was foundbe 0.145 m/s, or about 25% of the local gas veloity. Although this relative value appears significaits effect on particle trajectories and dopant delivis small because thermophoresis is significant owhen particles are very small, i.e., near the endthe vaporization process. The magnitude of this efcan be seen for NaCl particles (for which the physiproperties are well known). Thermophoresis is callated to shift the location of complete evaporationa 0.1-µm NaCl particle upstream by only about 2 µThis shift is negligible when compared to a calculaflame FWHM of around 1.7 mm. Thus, neither gra

itational nor thermophoretic forces should stroninfluence the particle trajectories for the flame coditions reported here.

Viewing of scattered laser light, as described pviously, provides information about the completenof particle vaporization in the flame. Experimenwere performed with a mist of 1.6% OPA introducby the HEN, with a strain rate 10% below the etinction value. Scattered light was observed whenlaser beam was positioned just below the exit plof the oxidizer nozzle confirming the ability to dtect residual particles. No scattered light was obserwhen the laser was positioned below the flamegion, indicating that the particles are small enoughbe consumed in the flame. At this position, care wtaken to ensure that the laser beam was betweenstagnation plane and the flame: scattered light wasserved in this region at the same reactant flow rabut with no flame present.

2.4. Wall losses

The chamber housing the HEN was designedminimize wall losses. The liquid spray exits the HEin a small-angle cone (∼15◦ half-angle), and enterslarge chamber that has a supplemental oxidizer strentering near the edge. The use of this chambernot prevent all wall losses, however. For the solutiosome droplets or residual particles were still lostthe walls. To quantify the amount of acid PCC lothe following rinsing procedure was used. A knowamount of OPA, in the form of a 1.6% water solutiowas sprayed into the apparatus under the same cotions used for the extinction experiments. The chaber and all tubing downstream were then rinsed wiknown amount of distilled water. The pH of the wawas measured using an Accumet pH probe, andconcentration of OPA was determined from its dsociation constant. The amount of acid recoveredcompared to the amount entering the HEN, and loswere calculated to determine the net amount deliveto the flame. Several different air and liquid flow rawere tested. PCC losses were found to be 13.8±0.6%of the initial amount delivered. This correction, mesured with OPA, is applied to all aqueous solutioof phosphorus acids. Since the initial droplet distrutions for all compounds producing residual particare very similar, the assumption of the same losreasonable. It is assumed that losses occur onlythe acid PCC, and that any accompanying waterimpinges on the walls is subsequently evaporated

For liquids not producing residual particles, zelosses are assumed on the basis of two observatFirst, as observed previously for vapor-phase DMdelivery [22], no trend in extinction measuremenwas observed as a function of time, suggesting


ligi-the

lyon-e inted,f athe

pre-nodP

re-t

ng

re-is noe tount-ci-

s todif-onlym-

gardent.la--ded

ar-ci-ted.e

er-ary

re-

a

re-(ofrenedve-e-atech apm

for

edquidtersses,For, viaamfor

iththered

r-t ofbe-odsten-an.

lueser-ef-edntsdc-

K

ater inad-

izer

wall adsorption/desorption processes were negble or had reached steady state. Secondly, whenDMMP flow was abruptly shut off, the flame rapidreturned to its undoped blue color on a time scale csistent with the average gas-phase residence timthe heated chamber. Wall deposits, if they had exiswould be expected to vaporize in the presence olower gas-stream dopant concentration and delayreturn of the undoped appearance of the flame. Invious experiments with vapor-phase DMMP andheated chamber[22] the return of the blue color habeen virtually instantaneous. Consistency of DMMextinction results obtained with the two systems,ported under Section3, provides further evidence thawall losses are negligible in the HEN system.

2.5. Calculations

Extinction calculations were performed usithe OPPDIF code, from the Chemkin II suite[38].OPPDIF is not designed to incorporate multiphaseactants, such as particle-phase dopants. Theremeans of calculating radiative losses or effects dusurface chemistry. Nor is there a means of accoing for differences between particle and gas veloties. Therefore, OPPDIF is used in these studiecalculate the differences in effectiveness betweenferent phosphorus-containing compounds basedon their gas-phase effects. Computationally, all copounds are treated as being in the gas phase, reless of the phase in which they exist in the experim

The phosphorus mechanism is employed by Gude et al.[47] and was developed for DMMP combustion. OPA and phosphorous acid are also incluin this mechanism. The GRI-Mech 3.0[48], exclud-ing nitrogen chemistry, was used for the hydrocbon combustion. Mixture-averaged diffusion veloties were used, and thermal diffusion was neglecA potential flow boundary condition with a centerlinaxial velocity of twice the measured volume avaged velocity was used. This set of velocity boundconditions was determined by MacDonald et al.[30]to best match experimental results of [OH] measuments with this particular burner.

The extinction calculations were performed inmanner similar to that described by Pitts et al.[49].That is, extinction was approached with small incmental steps in flow rates of the fuel and oxidizerat most 0.01 SLM for the fuel). The flow rates wechanged such that the position of the flame remaiconstant. Subsequent calculations with a higherlocity were restarted from previous solutions to dcrease computational time. The extinction strain rwas determined to be the highest strain rate at whisolution was found. For all the calculations, 250 pof PCC and/or 1.54% H2O was added.

-

3. Results

Extinction measurements were performedH2O and DMMP as neat compounds.Fig. 5presentsthe normalized global extinction strain rate obtainwhen the dopant enters the burner system as a lispray, comparing it to results when the dopant enas a vapor. The loadings are calculated, for all caassuming the dopant is completely evaporated.the vapor-phase tests, the dopants were addedsyringe pump, into a heated reactant line upstreof the oxidizer flow tube. The vapor-phase testsDMMP were performed and reported earlier[22]. Thesame additive loadings were investigated here wthe dopant introduced into the burner system inliquid phase, using the HEN. The absence of scattelaser light indicated complete evaporation of the H2Oand DMMP droplets.Fig. 5demonstrates that the nomalized global extinction strain rate is independenthe initial phase of the additive. The agreementtween extinction measurements for the two methof dopant introduction is consistent with compleevaporation of the neat liquid droplets with the ethalpy of vaporization supplied externally rather thfrom the exothermicity of the combustion reaction

The suppression effectiveness of H2O vapor de-termined here can be compared to literature vaobtained with reactants at a slightly lower tempature. Lentati et al. numerically investigated thefect of various concentrations of water vapor addto a nonpremixed methane–air flame with reactaat 300 K[8]. A roughly linear behavior is predictewith water addition, with a 25% reduction in extintion strain rate by a saturated air stream at 300

Fig. 5. Normalized global extinction strain rate of neDMMP and H2O, as a function of dopant loading. Thdopant was introduced either as a spray via the HEN othe vapor phase upstream of the oxidizer flow tube. Loings are given as the mole fraction of dopant in the oxidstream assuming complete vaporization.


ina-)senture-

een-2%taltdi-hiswith

eyent0 K,

in-ultsx-

i’sed

aren

rusIt

thecer-

tion

ted,of

tion

un-to

atainer-

in aayedt-

m-tontionthetched,

allwasx-r-

thes isod

-isivenant.of

o-ure

theva-i-

d byres-

with

ar-CCsfec-were

P,nss-

ced-temin

Fig. 6. Comparison of normalized global extinction strarate as a function of H2O loading (assuming complete vporization) due to addition of neat H2O and a 1.6% (molaraqueous solution of OPA. Error bars on OPA data repreone standard deviation due to scatter in extinction measments(δyi ) and uncertainty in loading(δxi ).

(water loading 3.51%). Experiments have also bperformed by Lazzarini et al.[21] with saturated water vapor at the same temperature, finding only a 1reduction in extinction strain rate. Our experimenresults, shown inFig. 5 for 4 to 40% of the amounof water in a saturated air stream at 300 K, incate that the extinction strain rate is linear over trange. These results are in reasonable agreementLentati’s calculations over the range in which thoverlap. It should be noted, however, that the presmeasurements were made with the reactants at 36while the numerical predictions assume 300 K. A lear extrapolation of the current experimental resto 3.51% water loading gives a 24% reduction in etinction strain rate, which agrees well with Lentatcalculated value but it is twice Lazzarini’s measurvalue.

Experiments were performed comparing H2O toan aqueous solution of 1.6% OPA. The resultsgiven inFig. 6, with the normalized extinction strairate plotted as a function of molar loading of the H2Oonly. For the case of the PCC, the H2O loading iscalculated by subtracting the amount of phosphocompound (1.6%) from the total dopant loading.also assumes that all of the water is delivered toflame. Error bars in the OPA data set represent untainties of one standard deviation. They-error bars,between 1 and 2%, are an appropriate combinaof uncertainties inaq andaq0, each of which is foundempirically as the standard deviation of eight repeameasurements. Thex-error bars, between 1 and 5%are derived from the uncertainties of flow ratesdopant and oxidant, and from the standard deviaof the measured dopant loss rate.

The absolute value of the slope ofFig. 6 repre-sents the effectiveness of the suppressant[50,51]. Theuncertainty in slope can be calculated from thecertainties in the measured quantities contributingit, using standard methods[52]. Two types of uncer-tainty must be considered: those affecting each dpoint randomly, and those affecting all data pointsa given set in the same way. The truly random unctainty contributions are responsible for the scattergiven data set, and are described above and displas error bars inFig. 6. For the purposes of calculaing a slope, uncertainties in thex coordinate (load-ing) were converted to an equivalenty uncertainty, asrecommended by Bevington[52]. Then the randomuncertainties in the individual data points were cobined, contributing an uncertainty of at most 1.6%the slope. In addition to this uncertainty it is importato include the uncertainty in the solute concentrat(1.8%), which has the same fractional effect onloadings of all data points obtained from a given baof aqueous solution. When these errors are includthe uncertainty in the slope is still under 2.5% forcases presented here. This calculated uncertaintyverified by comparing the slopes for two sets of etinction measurements for OPA obtained with diffeent solute mixtures; the two slopes agreed withincalculated uncertainty. Not included in this analysithe error in extinction strain rate due to the methby which extinction is approached (±2%; see Section 2.1). Because of the choice of conditions, therror is nonrandom and has the same value at a gstrain rate, regardless of the choice of suppressThus it has no effect on the relative effectivenessthe different suppressants.

A linear regression on the slopes inFig. 6 witha fixed y-intercept indicates that 1.6% of phosphrus approximately doubles the effectiveness of pH2O vapor, per mole of H2O delivered. Thus, anenhanced water mist can substantially improvesuppression performance over that of pure waterpor, without introducing large quantities of a chemcal substance. This result has also been observeother researchers who have investigated the suppsion effectiveness of enhanced water mists dopedsodium-containing compounds[18–20,53].

To test whether the chemical structure of the pent phosphorus compound is important, several Pwere introduced to the flame and suppression eftiveness was compared. The compounds testedthree neat phosphorus compounds (DMMP, DEMDMP), and five 1.6% (molar) aqueous solutio(DMMP, OPA, MPA, phosphorous acid, and phophonic acid). The neat compounds were introduusing a method described earlier[22] and the aqueous solutions were introduced into the burner sysvia the HEN. Flame extinction results are given


inrentos-

di-ateeat

and-

inrentcedds,r-

, andsys-so-

eions,er-onzerpor-ule

ndsent

he

lopeted

ingsomsPri-

rusrndds,orustedby

ndse-he

hee

t ofpes

in-ns.nsres-Thisibu-oressin-

vs

Fig. 7. Comparison of normalized global extinction strarate as a function of phosphorus loading for several diffephosphorus-containing compounds, namely trimethylphphonate (TMP), dimethylmethylphosphonate (DMMP),methylphosphate (DMP), and diethylmethylphosphon(DEMP). These compounds are all introduced in the nform, and enter the flame in the vapor phase. TMPDMMP results are from[22] and are included here for comparison.

Fig. 8. Comparison of normalized global extinction strarate as a function of phosphorus loading for several diffephosphorus-containing compounds. DMMP was introduas a neat compound via the HEN. All other compounincluding 1.6% (molar) aqueous solutions of DMMP, othophosphoric acid, phosphorous acid, phosphonic acidmethylphosphonic acid, are introduced into the burnertem via the HEN. The phosphorus loadings for the acidlutions have been corrected for wall losses.

Figs. 7 and 8, for the compounds introduced into thburner system as neat liquids and aqueous solutrespectively. The extinction results are plotted vsus the “phosphorus loading,” or the mole fractiof phosphorus-containing molecules in the oxidistream expressed in ppm assuming complete vaization of the additive. Since each PCC molec

Table 3Global extinction strain rate reduction by the compoutested: slope of normalized extinction strain rate vs agloading

Compound Slope(ppm−1)

Corrected slope(ppm−1)

NeatDMMP 223 267TMP 269 311DMP 247 278DEMP 204 255

1.6%aqueous solutionOrthophosphoric acid (OPA) 997 996Phosphorous acid 835 840Phosphonic acid 814 818Methylphosphonic acid 792 808Dimethylmethylphosphonate(DMMP)

654 697

Note. The uncertainty in the slope is 2.5% (see text). TTMP data were reported in an earlier paper[22] and areincluded here for reference. Values for the corrected sare derived by considering the heating value of the indicacompounds (see text).

contains a single phosphorus atom, these loadrepresent the number of moles of phosphorus atper total moles in the oxidant mixture. Neat DMMis included with the aqueous solutions for compason. For the DMMP/water solution, the phospholoading is the mole fraction of DMMP in the oxidizestream, assuming total vaporization of DMMP awater. Here, as well as for all of the neat compounzero wall losses are assumed. For the phosphacid solutions, the phosphorus loading is calculaas for the DMMP/water solution, and then reduced13.8% to correct for wall losses. (See Section2.4.)

Fig. 7shows that the neat phosphorus compouare all similarly effective. There is total scatter btween the different compounds of about 20%. Tresults from this study, indicating that the form of tparent compound is relatively unimportant in flamsuppression, are used when examining the effecaqueous solutions containing phosphorus. The slofor the curves inFig. 7are given inTable 3.

The results inFig. 8 fall into two groupings:the neat substance, DMMP, and the substancestroduced into the burner system as water solutioCompared to the neat DMMP, the water solutiohave steeper slopes, implying higher flame suppsion effectiveness per atom of added phosphorus.difference is due to the flame suppression contrtion from the water in the solution. The results fthe 1.6% solution of DMMP can be used to asswhether DMMP and water have additive effects. Lear regressions of the data inFig. 5 give values ofthe slopes of the normalized extinction strain rate


“ef-ressun-heis

vior-

rease

turea

ofure

em-iontedar-

reef-ve-thearly6%

,im-eraxi-ec-hen

heof

be-ith

cea-nsed

thesti-em-m-ionom-if-

therad-tics

or

up-

n-theinc-th aairult-

elthethaterene,cor-tantoneretheThetedmo-

us-Thesm-reofbe

this

ncidstsAntheob-

ac-to

ar-ter

os-atead-ectres-

ro-thes,ase

clesm-

mole fraction for neat DMMP and H2O. A weightedaverage of these two numbers yields a predictedfectiveness” of a 1.6% solution of DMMP in wateto be only 5% more than the measured effectivenof the solution. Thus, one can assume that withincertainty, additivity of effectiveness is valid under tconditions of the current experiment. However, itexpected that synergistic rather than additive behawill be seen at high H2O loadings, as the effectiveness of chemical agents has been observed to incwith decreasing temperatures[30,54]. In our experi-ments, the change in the adiabatic flame temperaby the amount of water added would result in onlysmall change in the extinction strain rate. Addition1.5% H2O changes the adiabatic flame temperatby only 20 K; according to previous work[50], thischange would increase the effectiveness of the chical compound by roughly 10% compared to additto a flame without an inert dopant. It should be nothat due to considerable scatter in the cited work, pticularly in the range of interest for this work, theis uncertainty in the magnitude of this synergisticfect. Nonetheless, synergy implies that the effectiness of the solution should be more effective thatweighted average of the neat compounds, not neequally as effective as observed here for the 1.DMMP solution. Further work at higher H2O load-ings is needed.

All of the acid solutions studied (OPA, MPAphosphonic acid, and phosphorous acid) exhibit silar reduction of the global extinction strain rate, pmole PCC added to the air stream. There is a mmum difference of about 22% in the measured efftiveness, based on a comparison of the slopes. Wthe data for the DMMP solution are included, tspread is larger: there is a maximum difference42% in slopes in the set of water solution data (tween DMMP and OPA). These slopes, along wcorrected slopes described below, are given inTa-ble 3. The possible significance of the differenbetween results for the DMMP solution (which vporizes entirely) and for phosphorus acid solutio(which produce residual particles) will be discusslater.

The spread in suppression performances forphosphorus acid solutions studied is outside the emated experimental measurement uncertainty. Chical and physical differences in the phosphorus copounds could result in differences in fire suppresseffectiveness. Such differences in the added cpounds, including differences in heating values, dferences in the kinetics leading to the formation ofphosphorus-containing species that participate inical recombination cycles, and specific characterisof any residual particles such as vaporization rate

radiative properties, may also contribute to the spression difference.

Additive species, or portions thereof, that cotain hydrocarbons, will to some extent promoteflame due to their associated heating values. Exttion strain rate measurements were performed wismall amount (400 ppm) of iso-octane added to thestream of a methane/air nonpremixed flame, resing in a promotional effect of 4%[22]. To estimatethe magnitude of the promotional effect of the fucontent of other additives considered in this work,hydrocarbon content of the PCC is compared toof the iso-octane. The additives’ heating values wcompared (on a molar basis) to that of iso-octaallowing the measured PCC effectiveness to berected for the fuel contents effect (assuming conspromotional effect per unit heating value). In additito the contribution of the hydrocarbon groups, this some contribution to the total heating value ofPCC species due to the phosphorus atom itself.heating values of PCCs in this study were calculaassuming that the phosphorus product is the therdynamically preferred species, P4O10 [55]. Correct-ing for heating value effects brings the phosphorbased acids into better agreement with each other.corrected slopes, given inTable 3, represent the slopecorrected for the heating value of the different copounds. After correction, OPA is still some 20% moeffective than the other acid solutions. The sourcethis increased effectiveness is uncertain but mayrelated to the nature of the residual particles fromsolution.

The difference between the DMMP solutio(which vaporizes entirely) and the phosphorus asolutions (which produce residual particles) suggean enhanced effectiveness role for the particles.enhanced radiative heat loss from flames due toaddition of inert particles has been predicted andserved experimentally in premixed flames[7]. Othereffects, such as enthalpy of vaporization and retions on the particle surface, may also contributean increase in effectiveness from the addition of pticles. In the current experiments, particles that enthe flame are consumed in it. Thus, all of the phphorus atoms are eventually available to participin the gas-phase catalytic removal of the flame ricals H, O, and OH. Otherwise, one would expa decrease in effectiveness resulting from the pence of particles. As the data inTable 3 indicate,all of the phosphorus-based acid solutions that pduce residual particles are more effective thanDMMP solution that does not form particles. Thuthese results are consistent with a predicted increin suppression effectiveness for the residual partias long as they are small enough to volatize copletely in the flame.


notthethe1%.om-MPe-ase-ownectstheionticleap-ffectmp-forso-

fectp-r alltheer,in-if

giv-cidP

aseateain

finc-ved

lueo-ess

by

te

CCare:

isivefthal-ha-to

e.g.,be

a-nds,ons

totopededure-r es-iste

orof

timelledro-thevap-ll-exitad-of

tro-is

rmitw-ctof

uc-nlarlts.ofeseiver-o anof aad-theere.ef-

on-

The differences in effectiveness seen here arelarge: the difference in the slopes, corrected forheating values, between the DMMP solution andaverage of the phosphorus acid solutions is 2Gas-phase kinetic differences may be significant cpared to the remaining discrepancy between DMand acid solutions, as this difference in effectivness is similar to that found between the vapor-phcompounds shown inFig. 7. Kinetics for the aqueous phosphorus solutions tested are largely unknso it is not possible to separate gas-phase efffrom those due to the particles. However, taking21% difference between DMMP and the acid solutslopes as an estimate of the magnitude of the pareffects, we conclude that the net particle effectpears to be smaller than the gas-phase chemical eof the phosphorus. Under the conditions and assutions reported here, the particles would accountabout one-quarter of the total suppression of thelution, compared with the gas-phase chemical efthat is responsible for about 40% of the total supression. The same wall losses were assumed fothe acid solutions, and this assumption may affectobserved magnitude of the particle effect. Howevsome positive particle effect on suppression can beferred even without making this assumption. Evenzero losses are assumed for the acid solutions (ing a lower bound on their effectiveness), the asolutions are still all more effective than the DMMsolution.

Some information on differences in the gas-phchemical effect was gained from extinction strain rcalculations. The normalized global extinction strrate obtained from these calculations are given inTa-ble 4. The effect of the H2O is decoupled from that othe PCC. As seen, the calculated reduction in exttion strain rate is significantly less than that obserexperimentally: the PCC alone (no H2O included) re-duces the extinction strain rate from its undoped vaby only about: 1.5% for DMMP; 2.3% for phosphrous acid; and 3% for OPA. The trend in effectiven

Table 4Summary of calculated extinction strain rates, normalizedthe undoped value (aq/aq0)

Dopant Normalized extinction strain ra

DMMP 0.99OPA 0.97P(OH)3 0.98H2O 0.88DMMP + H2O 0.87OPA+ H2O 0.86P(OH)3 + H2O 0.87

Note. The effect of PCCs was found using Glaude’s Pmechanism (see text for details). Quantities of dopants250 ppm of PCC and/or 1.54% H2O (molar).

is similar to that seen experimentally; that is, OPAthe most effective and DMMP is the least effectPCC. However, the effectiveness is about one-fithat seen experimentally for neat DMMP. Thus, cculations using the current chemical kinetic mecnism can be used to gain only qualitative insight inthe gas-phase effect. Newer kinetic mechanisms ([27,28]) may yield better agreement, and shouldinvestigated.

4. Summary

An investigation of the flame suppression for wter, vapor-phase phosphorus-containing compouand residual particles derived from aqueous solutiof PCCs was performed. To do so, a techniqueintroduce an additive as a fine mist of droplets inan opposed-jet nonpremixed burner was develoand validated. A high-efficiency nebulizer was usto produce the droplets, and droplet size measments using a phase-Doppler particle anemometetablished that the size distribution of the dropletsindependent of the compound and liquid flow raused. This allows wide application of the HEN fstudying different potential additives. Key featuresthe system are small droplet size, large residenceof the reactant stream under temperature-controconditions, and the availability of an accurate pcedure for measuring losses to surfaces withinfeed system. These features lead to complete eoration of volatile constituents of the droplets, a wecharacterized state of the reactant mixture at theof the burner tube, and well-defined gas-phase loing of added agent at the flame. A comparisonthe flame suppression effectiveness of PCCs induced into the burner system in the liquid phasemade. This experimental approach does not pea study of the effect of droplets on the flame; hoever, it does allow investigation of the physical effeof residual particles formed from the evaporationdroplets.

Experimental results indicate a significant redtion in global extinction strain rate with the additioof pure water vapor (10% reduction at 1.5% moloading), in good agreement with numerical resuWith the addition of a small amount (1.6% molar)PCC in water solution, this reduction doubles. Thresults support the use of water as a means of deling a chemically active, condensed phase agent tactual fire. They also show that the effectivenessPCC/water solution can be determined by a lineardition of the effectiveness of the components, overrange of concentrations and loadings reported hComparisons of experimental flame suppressionfectiveness are made for several volatile and n


ra-anthe

ela-ss,ef-

russion

b-es,ndC,

,andC,

ust.

3

3–

5

22

st.

64.e

95)

li,

n-98)

W..

n-

h-

0)

-

C.

.is-bys ofce,

-ow

ointti-

n-an-

st-G.st.

rg,c.

ov,itz,30

-

ity

5)

9–

3–

.

l.

y,d-s,

-ofbo-

l-

21

7–

nd.

4

3–

volatile PCCs. Volatile PCCs all showed compable values of effectiveness, all observably lower ththose of the nonvolatile PCCs. Thus it appears thatchemical structure of the parent compound has rtively little impact on flame suppression effectivenewhile the residual particles enhance it. Despite thefect of residual particles, participation of phosphoin the gas-phase chemistry is the primary suppresmechanism.

References

[1] C.C. Grant, NFPA J. 88 (1994) 41–54.[2] R.E. Tapscott, T.A. Moore, J.A. Kaizerman, C.J. Ki

ert, R.A. Tetla, Advanced agent halon substitutin: Proceedings of the 1995 International CFC aHalon Alternatives Conference, Washington, D1995, pp. 644–649.

[3] D. Trees, K. Seshadri, A. Hamins, in: A.W. MiziolekW. Tsang (Eds.), Halon Replacements: TechnologyScience, American Chemical Society, Washington, D1995, pp. 190–203.

[4] P. Papas, J.W. Fleming, R.S. Sheinson, Proc. CombInst. 26 (1997) 1405–1412.

[5] T. Mitani, Combust. Flame 44 (1982) 247–260.[6] M. Rumminger, G. Linteris, Combust. Flame 12

(2000) 82–94.[7] Y. Ju, C.K. Law, Proc. Combust. Inst. 28 (2000) 291

2920.[8] A.M. Lentati, H.K. Chelliah, Combust. Flame 11

(1998) 158–179.[9] D. Trees, K. Seshadri, Combust. Sci. Technol. 1

(1997) 215–230.[10] R.A. Dodding, R.F. Simmons, A. Stephens, Combu

Flame 15 (1970) 313–315.[11] T. Mitani, Combust. Flame 50 (1983) 177–188.[12] A. Hamins, Proc. Combust. Inst. 27 (1998) 2857–28[13] T.A. Milne, C.L. Green, D.K. Benson, Combust. Flam

15 (1970) 255–264.[14] A. Jones, P.F. Nolan, J. Loss Prevent. Proc. 8 (19

17–22.[15] C.C. Ndubizu, R. Ananth, P.A. Tatem, V. Moteval

Fire Saf. J. 31 (1998) 253–276.[16] K. Prasad, C. Li, K. Kailasanath, C. Ndubizu, R. A

anth, P.A. Tatem, Combust. Sci. Technol. 132 (19325–364.

[17] E.J.P. Zegers, B.A. Williams, R.S. Sheinson, J.Fleming, Proc. Combust. Inst. 28 (2000) 2931–2938

[18] J.C. Yang, M.K. Donnelly, N.C. Prive, W.L. Grosshadler, Fire Saf. J. 36 (2001) 55–72.

[19] R. Zheng, K.N.C. Bray, B. Rogg, Combust. Sci. Tecnol. 126 (1997) 389–401.

[20] B. Mesli, I. Gokalp, Combust. Sci. Technol. 153 (200193–211.

[21] A.K. Lazzarini, R.H. Krauss, H.K. Chelliah, G.T. Linteris, Proc. Combust. Inst. 28 (2000) 2939–2945.

[22] M.A. MacDonald, T.M. Jayaweera, E.M. Fisher, F.Gouldin, Combust. Flame 116 (1997) 166–176.

[23] R.T. Wainner, K.L. McNesby, R.G. Daniel, A.WMiziolek, V.I. Babushok, Experimental and mechantic investigation of opposed-flow propane/air flamesphosphorus-containing compounds, in: Proceedingthe 10th Halon Options Technical Working ConferenAlbuquerque, NM, 2000, pp. 141–150.

[24] R.R. Skaggs, R.G. Daniel, A.W. Miziolek, K.L. McNesby, Spectroscopic studies of inhibited opposed flpropane/air flames, in: Proceedings of the First JMeeting of the U.S. Sections of the Combustion Instute, Washington, DC, 1999, pp. 575–579.

[25] J.W. Hastie, D.W. Bonnell, Molecular Chemistry of Ihibited Combustion Systems, National Bureau of Stdards, 1980.

[26] T.M. Jayaweera, C.F. Melius, W.J. Pitz, C.K. Webrook, O.P. Korobeinichev, V.M. Shvartsberg, A.Shmakov, I.V. Rybitskaya, H.J. Curran, CombuFlame 140 (2005) 103–115.

[27] A.G. Shmakov, O.P. Korobeinichev, V.M. ShvartsbeD.A. Knyazkov, T.A. Bolshova, I.V. Rybitskaya, ProCombust. Inst. 30 (2004) 2345–2352.

[28] O.P. Korobeinichev, V.M. Shvartsberg, A.G. ShmakT.A. Bolshova, T.M. Jayaweera, C.F. Melius, W.J. PC.K. Westbrook, H. Curran, Proc. Combust. Inst.(2004) 2350–2357.

[29] A. Twarowski, Combust. Flame 94 (1993) 91–107.[30] M.A. MacDonald, F.C. Gouldin, E.M. Fisher, Com

bust. Flame 124 (2001) 668–683.[31] T.M. Jayaweera, PhD dissertation, Cornell Univers

(2002).[32] K. Seshadri, N. Ilincic, Combust. Flame 101 (199

271–294.[33] A.R. Masri, Combust. Sci. Technol. 96 (1994) 18

212.[34] Y. Otsuka, T. Niioka, Combust. Flame 21 (1973) 16

176.[35] H.K. Chelliah, C.K. Law, T. Ueda, M.D. Smooke, F.A

Williams, Proc. Combust. Inst. 23 (1990) 503–511.[36] G.F. Carrier, F.E. Fendell, F.E. Marble, SIAM J. App

Math. 28 (1975) 463–500.[37] A. Linan, Acta Astronaut. 1 (1974) 1007–1039.[38] A.E. Lutz, R.J. Kee, J.F. Grcar, F.M. Ruple

OPPDIF: A Fortran Program for Computing OpposeFlow Diffusion Flames, Sandia National LaboratorieSAND96-8243, 1996.

[39] R.J. Kee, F.M. Rupley, J.A. Miller, Chemkin-II: A Fortran Chemical Kinetics Package for the AnalysisGas-Phase Chemical Kinetics, Sandia National Laratories, SAND89-8009, 1989.

[40] D.C. Rapp, M.F.M. Nogueira, E.M. Fisher, F.C. Goudin, Environ. Eng. Sci. 14 (1997) 133–140.

[41] K. Seshadri, F. Williams, Int. J. Heat Mass Transfer(1978) 251–253.

[42] I.K. Puri, K. Seshadri, Combust. Flame 65 (1986) 13150.

[43] S. Turns, An Introduction to Combustion: Concepts aApplications, first ed., McGraw–Hill, New York, 1996

[44] K.H. Leong, J. Aerosol Sci. 12 (1981) 417–435.[45] S.C. Li, P.A. Libby, F.A. Williams, Combust. Flame 9

(1993) 161–177.[46] D.E. Rosner, Y.F. Khalil, J. Aerosol Sci. 31 (2000) 27

293.


ok,

i-K.n-/

r-er-rt,

C..

50

for9.–

me

orro-

heute,

[47] P.A. Glaude, H.J. Curran, W.J. Pitz, C.K. WestbroProc. Combust. Inst. 28 (2000) 1749–1756.

[48] G.P. Smith, D.M. Golden, M. Frenklach, N.W. Morarty, B. Eiteneer, M. Goldenberg, C.T. Bowman, R.Hanson, S. Song, C. William, J. Gardiner, V.V. Lissiaski, Z. Qin, available at:http://www.me.berkeley.edugri_mech/.

[49] W.M. Pitts, J.C. Yang, M.L. Huber, L.G. Blevins, Chaacterization and Identification of Super-Effective Thmal Fire Extinguishing Agents—First Annual RepoNISTIR 6414, 1999.

[50] M.A. MacDonald, T.M. Jayaweera, E.M. Fisher, F.Gouldin, Proc. Combust. Inst. 27 (1998) 2749–2756

[51] T. Niioka, T. Mitani, M. Takahashi, Combust. Flame(1983) 89–97.

[52] P.R. Bevington, Data Reduction and Error Analysisthe Physical Sciences, McGraw–Hill, New York, 196

[53] Z. Liu, A.K. Kim, J. Fire Protection Eng. 10 (2000) 3250.

[54] Y. Saso, Y. Ogawa, N. Saito, H. Wang, Combust. Fla118 (1999) 489–499.

[55] E.M. Fisher, Equilibrium composition calculations fcombustion of organophosphorus compounds, in: Pceedings of the 1995 Fall Technical Meeting of tEastern States Section of the Combustion InstitWorcester, MA, 1995, pp. 179–182.

http://www.me.berkeley.edu/gri_mech/



Documents

Flame suppression by aerosols derived from aqueous solutions