Combustion and Flame 148 (2007) 76–87www.elsevier.com/locate/combustflame

Combustion of multiphase reactants for the synthesis ofnanocomposite materials

S.D. Bakrania ∗, T.A. Miller, C. Perez, M.S. Wooldridge

Department of Mechanical Engineering, University of Michigan, 2350 Hayward St., Ann Arbor,MI 48109-2125, USA

Received 17 May 2006; received in revised form 20 August 2006; accepted 29 August 2006

Available online 8 December 2006

Abstract

Controlling nanocomposite composition and morphology is a vital step toward designing new and advancedmaterials. The current work presents the results of an experimental investigation of the use of mixed-phase re-actants for the synthesis of nanocomposite materials. Gas-phase tetramethyltin was used as a precursor for tindioxide (SnO2). Metal additives were introduced to the SnO2 synthesis system using solid-phase metal acetates asthe precursors. The physical and chemical properties of the metal acetate reactants, the combustion environmentand the nanocomposite materials were characterized in order to clarify the reaction processes important duringsynthesis of nanocomposite materials from these categories of mixed-phase reactants. X-ray diffractometry wasused to determine composition and the crystalline structure of reactant and product materials. Scanning and trans-mission electron microscopy were used to examine particle morphology of reactant and product materials, andX-ray energy dispersive spectroscopy was used for elemental speciation. The results indicate the metal acetatesare an excellent source of metal and metal oxide additives in a flame reactor. The metal acetates rapidly decomposeand experience considerable restructuring, leading to metal additives with a range of microstructures in the SnO2nanocomposites: from metal encapsulation in SnO2 to mixed aggregates.© 2006 The Combustion Institute. Published by Elsevier Inc. All rights reserved.

Keywords: Combustion synthesis; Nanocomposites; Metal acetate; Tin dioxide

1. Introduction

Many engineering applications can be improvedthrough the use of nanocomposite materials. Forexample, gas-sensor sensitivity has been increasedby orders of magnitude by incorporating dopantswithin nanocrystalline gas-sensing materials [1,2].

* Corresponding author. Fax: +1 734 647 3170.E-mail address: bakrania@umich.edu (S.D. Bakrania).

0010-2180/$ – see front matter © 2006 The Combustion Institute.doi:10.1016/j.combustflame.2006.08.008

Core–shell nanostructures of silica and gold havebeen used to create particles with incredible plas-mon resonance tunability, spanning from 650 nm to6 µm [3]. Metal shells surrounding core nanoma-terials have been demonstrated as having excellentpotential for developing advanced bioengineering di-agnostics and treatments using electro-optical inter-actions [4]. Room-temperature blue laser emissionhas been demonstrated using core–shell structuresof CdS/ZnS nanocomposites [5]. Although these ex-amples represent diverse material systems and ap-

Published by Elsevier Inc. All rights reserved.

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 77

plications, a common feature of the nanocompositesystems is the strong dependence of the performancecharacteristics on the nanocomposite properties, in-cluding the relative morphology, phase, composi-tion, size, loading, and distribution of the componentspecies.

A novel method for the synthesis of nanocom-posite materials has recently been demonstrated byMiller et al. [6]. The synthesis method uses solid- andgas-phase precursor reactants to create nanocompos-ite materials, and the demonstration studies showedgood control of nanocomposite properties using thesynthesis conditions for production of tin dioxide(SnO2) doped with various materials [6]. Tin diox-ide nanocomposites are of substantial interest as cat-alysts, as optical materials, and as materials in semi-conductor gas sensors.

The mixed solid- and gas-phase combustion syn-thesis approach demonstrated at the University ofMichigan has considerable potential for materials fab-rication. The use of solid-phase precursor reactantscombined with gas-phase precursor reactants expandsthe parametric range of microstructural properties thatcan be achieved beyond most gas-phase combustionsynthesis methods. For example, staged decomposi-tion of the reactant precursors can be used to controlthe relative growth rates of the condensed phase prod-ucts, which dictate (in part) the morphology of theproduct materials. In addition, there are a large vari-ety of nontoxic, relatively inexpensive, metal organicsolid-phase materials that can be considered for de-sign of the microstructural properties in combustionsynthesis. A challenge associated with this approachis the lack of understanding of the physical and chem-ical mechanisms important in mixed-phase combus-tion synthesis systems.

Based on this need, the primary objective ofthe current work was to identify the combustionprocesses and reactant characteristics affecting themicrostructural properties of nanocomposite materi-als fabricated using a combination of solid- and gas-phase precursor reactant materials. Links between theproperties of the solid-phase reactants (such as parti-cle size and chemical structure) and the condensed-phase products are critical for designing compositionand microstructure of a nanocomposite. Improvedunderstanding of these links will greatly facilitateexpansion of the combustion synthesis method to ad-ditional material systems and applications. In order tomeet the objective of the study, the synthesis environ-ment was characterized, and analyses of the reactantand product particles were conducted to identify andquantify the properties as a function of the reactorconditions. The results are discussed in terms of po-tential reaction pathways.

2. Experimental

2.1. Combustion synthesis facility

All experiments were conducted using the com-bustion synthesis facility developed at the Universityof Michigan [6–10]. The facility consists of three ma-jor components: the burner (a multi-element diffusionor Hencken burner, Research Technologies, RD1X1)used to create the high-temperature synthesis environ-ment, the bubbler system used to provide the tetram-ethyltin (TMT) precursor for the tin dioxide particles,and the particle feed system (PFS) used to provide thesolid-phase precursor reactants.

The burner consists of an array of hypodermicfuel tubes (i.d. 0.51 mm) and oxidizer channels (i.d.0.81 mm) arranged in a 2.54 cm square hastalloy hon-eycomb matrix. The matrix is surrounded by an addi-tional 1.0 cm wide nitrogen (N2, 28.3 lpm, CryogenicGases, 99.998%) shroud region. The H2 (2.78 lpm,Cryogenic Gases, 99.99%), O2 (1.46 lpm, CryogenicGases, 99.993%), and Ar (17.1 lpm, Cryogenic Gases,99.998%) rapidly mix above the surface of the burner,creating a primary flame sheet made of diffusionflamelets. At heights >5 mm and regions >4 mmradially from the center of the burner (where the sec-ondary fuel tube is located; see Fig. 1), the primaryflame system is nearly uniform and one-dimensionalin temperature, pressure, and composition. The high-temperature region, with temperatures ranging from1400 K near the burner surface to 400 K 30 cm abovethe burner surface, forms the reaction environmentfor the particle synthesis. All flow rates are moni-tored using a mass flow meter (O2: TSI, model 4100,±2% accuracy) and calibrated rotameters (H2, Ar,N2: Omega, ±5% accuracy). A detailed descriptionof the primary flame system is available [7].

For the current work, the gas- and solid-phase pre-cursor reactants are introduced simultaneously intothe synthesis environment through the secondary fueltube located at the center of the burner (see Fig. 1).Vapor-phase TMT (Sn(CH3)4, Alfa Aesar, 98%) isdirected to the burner by bubbling Ar (63.5 mlpm)through a reservoir of liquid TMT maintained at roomtemperature. Previous characterization studies indi-cate the high vapor pressure of the TMT (∼11 kPaat room temperature) yields saturated mixtures of 21–23% TMT dilute in Ar (mole basis) [9,10]. The argonflow rate is monitored using a calibrated rotameter(Omega, ±5% accuracy).

The PFS introduces solid-phase reactants as pre-cursors for the metal additives to the nanocrystallineSnO2. Similar particle-feed systems have been usedin combustion studies of coal particles [11,12]. ThePFS system presented here is based upon those de-signs. A detailed schematic of the PFS, which consists

78 S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87

Fig. 1. Schematic of the combustion synthesis facility used to introduce the solid- and gas-phase precursor reactants into theflame reactor.

of an enclosed glass entrainment column, syringe, andsyringe pump, is presented in Fig. 1. The glass col-umn has a gas inlet diameter of 3.70 mm and an outletdiameter of 1.07 mm. Argon is used as the carrier gasthrough the column and is regulated by a calibratedrotameter (Omega, ±5% accuracy). Precursors are in-jected into the column via an open-ended, cylindricalsyringe (BD 1 ml U-100), where the tip of the syringeis placed at the centerline of the column. The injectionfeed rate is controlled by a syringe pump (Medfusion2001). A mechanical agitator contacts the syringe toreduce the effects of particle accumulation and to pro-vide a steadier feed rate. During a typical experiment,the argon flow rate to the PFS is set at 390 mlpmand the syringe plunger rate is set at 1.0 ml/h (corre-sponding to ∼1.4 g/h for gold acetate). For all experi-ments, the solid phase precursor reactants were sievedto <45 µm to facilitate the particle flow through thePFS.

The flow from the PFS and the flow from theTMT bubbler mix in a glass L-shaped connector (seeFig. 1). The vapor- and solid-phase precursors arethen directed through the secondary fuel tube into

the reactive environment above the surface of theburner. The flow of particle precursor reactants formsa slightly lifted laminar diffusion flame. For the flowrates considered in this work, the Reynolds numberbased on the total argon flow rate and the i.d. of thesecondary fuel tube is Re = 720. The height of thesecondary flame is typically between 12 and 25 cm,based on the luminous region of the flame.

2.2. Flow-field diagnostics

Temperature profiles in the primary and secondaryflame regions were determined using uncoated ther-mocouples (type S, Pt/10%Rh–Pt, with bead diame-ters of 0.050 and 0.125 mm). The measurements werecorrected for radiation effects and for changes in beaddiameter and emissivity due to deposition of mate-rial onto the thermocouple bead (which only affectmeasurements made within 4 mm of the secondaryflame region). Uncertainties in the temperature mea-surements are estimated as ±50 K.

Particle velocities were examined for the alu-minum acetate system using a high-speed digital cam-

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 79

era (Vision Research, Phantom v. 7.1 color, Nicor50 mm lens f/0.95). The argon flow rate to the primaryflame was lowered to 11.4 lpm (as opposed to the typ-ical operating condition of 17.1 lpm), to improve theimaging and TMT was not used with the secondaryfuel tube. Sequences were acquired for 3 s with aframe rate of 1 kHz and an exposure time of 938 µs.Particles were tracked and counted using Vision Re-search software (v. 6.0.6). Residence times along theaxis of the burner were estimated by integrating thevelocity profiles obtained from the particle trackingmeasurements to the appropriate distance from theburner surface.

2.3. Particle sampling, preparation and analysis

Scanning electron microscope (SEM, PhilipsXL30 Series) imaging of the morphology and size ofthe unburned metal acetate particles was achieved bydepositing ∼10–20 mg of the powders onto conduc-tive copper tape. A 15-nm-thick layer of Au/Pd wasthen sputtered onto the samples to improve imagingquality. Samples of the as-received metal acetates andfinal product materials (sampled using a cold plateheight at 27 cm) were analyzed using X-ray diffrac-tion (XRD), with a typical sample size of ∼50 mg.The product materials were removed from the samplecoupon and then dispersed in methanol or isopropylalcohol (∼0.02 ml). The paste (0.5 ml) was spreadon a glass slide and dried at room temperature for∼10 min. Scans for phase identification and for aver-age additive particle size were obtained using an au-tomated Scintag Theta–Theta XRD with incrementsof 0.02◦ 2θ and CuKα radiation (λ = 1.5405 Å). Thescans were obtained over a 2θ range of 15◦–85◦ ata scan rate of 5◦ 2θ/min. Spectral scans for aver-age crystallite size for SnO2 were measured over a2θ range of 22◦–31◦ at a scan rate of 0.5◦ 2θ/min.Peak positions and relative intensities of the powderpatterns were identified by comparison with referencespectra [13].

The average crystallite size was determined fromthe XRD spectra using the Scherrer equation,

(1)dXRD = 0.9λ

β1/2 cos θ,

where dXRD is the average crystallite size, λ is thesource wavelength, β1/2 is the full width at half-maximum of the peak used for the analysis, and θ isthe XRD scattering angle of the peak.

Transmission electron microscopy images (TEM,Philips CM12 and JEOL 3011 high-resolution elec-tron microscopes) were obtained of the final productpowders and as a function of location above the sur-face of the burner. The samples were obtained bydirect deposition onto TEM grids (3 mm diameter,

Electron Microscopy Sciences, carbon film, 300 meshcopper) rapidly inserted into the secondary flame orthe exhaust region. X-ray energy dispersive spec-troscopy (XEDS) was used for elemental speciation.

2.4. Analysis of micrograph images

Automated image analysis of combustion-gener-ated particles can be challenging due to highly aggre-gated, overlapping, and necked features. Often man-ual counting is employed for better feature identifi-cation. In the current work, a semiautomated methodwas developed to assist the manual counting andgreatly decrease the time required to generate sta-tistically significant particle size distributions. Themethod automates the accounting of feature sizes,leaving feature identification to be performed manu-ally.

The first step of the process is to duplicate theparticle features in a vector graphics package (e.g.,CorelDraw or Adobe Illustrator). The features of in-terest (e.g., primary spherical particles) are identifiedin the image by overlaying equivalent outlines on thefeatures. The overlays are then aligned in a columnwith no overlap between features and extracted to abinary (black and white) bitmap file—with either a1 or 0 value for each pixel. Such a file can be ana-lyzed using automated software (e.g., Matlab) to ex-tract the particle size distribution by counting the ze-ros in the central column of each feature and applyingthe appropriate scale. In the current work, the parti-cle size distributions were extracted assuming spher-ical particles; however, the technique can be easilyadapted to more complex shapes. As a demonstrationof the method, Fig. 2 presents an image of overlap-ping spherical gold particles. Image analysis resultedin an average particle diameter of 0.86 ± 0.35 µm,which is within the manufacturer specifications (AlfaAesar, 99.96% Au) of 0.5–0.8 µm.

3. Results

3.1. Characterization of unburned solid-phasereactants

Table 1 provides a summary of the solid-phaseprecursors investigated and their thermophysical pro-perties. The most critical data for this study are the de-composition temperature and the density of the solid-phase reactants. Estimates were available for all themetal acetates considered, except the gold acetate. Asa consequence, efforts were made to determine theseproperties using simplified yet quantitative methods.The as-received gold acetate was analyzed using a

80 S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87

Fig. 2. Schematic indicating the technical approach for semiautomated particle dimension analysis. On the left is an SEM imageof gold particles illustrating the identification of particle features and the corresponding scale bar. The center section shows thefeatures aligned in a single column. On the right is the binary matrix generated from the black and white particle array used toobtain particle statistics.

displaced gas volume method (Micromeritics, Ac-cuPyc 1330) to determine the particle mass densitylisted in Table 1. The decomposition temperature ofthe gold acetate was estimated using a visual melt-ing temperature tool (MEL-TEMP 3.0). A sample ofthe unreacted gold acetate (∼20 mg) was placed ina capillary tube (i.d. 0.9 mm). The tube was heated(5 ◦C/min) and the sample observed. The decompo-sition temperature was approximated as the temper-ature at which a dramatic change in the color of thepowder took place, from light brown to a dark brown.The decomposition temperature determined for goldacetate was 150–160 ◦C, which is consistent withother metal acetate data (see Table 1). Additionally,this method reproduced a palladium acetate decom-position temperature of 222–229 ◦C, which is in goodagreement with the literature value of ∼240 ◦C [14].

Fig. 3 presents SEM images of as-received metalacetates powders. As seen in the images, gold and alu-minum acetates appear as porous and irregular struc-tures, while copper and palladium acetates appear ascondensed and well-defined crystalline structures.

Typical XRD patterns of the as-received unreactedgold, aluminum, copper, and palladium acetate pow-ders and corresponding reference patterns are pre-sented in Fig. 4. There are no standards available forgold acetate in the JCPDS database [13]. However,Kritl and Drofenik [15] presented an XRD scan ofgold acetate in their article on sonochemical prepa-ration of Au2S3, which is reproduced in Fig. 4a forreference. The spectra agree well with respect to loca-tion of the peak features and relative intensities. Notethe copper acetate spectra (see Fig. 4c) indicated thepresence of both copper acetate (24-1126) and copperacetate hydrate (27-145) in the precursor powder.

Table 2 provides a summary of the average crys-tallite sizes determined from the metal acetate XRD

spectra (dXRD). Note that an average crystallite sizefor palladium acetate could not be accurately deter-mined due to interference from multiple peak fea-tures. The crystallite sizes obtained from the metalacetate XRD spectra suggest that the particles pre-sented in Fig. 3 consist of crystallites approximately12 nm in size for the aluminum acetate, 26 nm in sizefor the copper acetate, and 160 nm in size for the goldacetate.

3.2. The combustion synthesis environment

Temperature profiles for the combustion synthe-sis system with and without the presence of the sec-ondary flame are presented in Fig. 5 as a function ofheight above the surface of the burner. No solid-phaseprecursors were used in these characterization stud-ies. The thermocouple was located slightly off-centerof the burner (see inset in Fig. 5) for both data sets.The presence of the TMT flame leads to higher tem-peratures throughout the synthesis region encounteredby the particles. Particle residence times based on theparticle tracking are presented in the inserts.

3.3. Characterization of particle evolution

The evolution of particle microstructure through-out the combustion synthesis environment was exam-ined for the gold acetate/TMT system using particlesampling and ex situ imaging at increasing samplingheights. The series of TEM images presented in Fig. 5(bottom panel) show the progressive formation andgrowth of the Au/SnO2 nanocomposites. Gold parti-cles are formed early in the system (5 ms) with rela-tively few discrete, small SnO2 particles present. Asthe residence time increases, the frequency of appear-ance of the gold particles increases slightly; however,

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 81

Table 1Thermophysical properties of bulk condensed-phase materials relevant to the material synthesis systems studied

Chemicalformula

Name Molecularweight(CAS #)

Reactantmanufac-turer,purity

Meltingpoint(K)a

Boilingpoint(K)

Vaporpressureb

(atm)

Density(g/cm3)

Refe-rence

ReactantsAl(OH)(C2H3O2)2 Aluminum 162.163 Sigma–Aldrich, 327 N/A – 1.74 [13,19]

diacetate [142-03-0] N/AAu(C2H3O2)3 Gold(III) 374.101 Alfa Aesar, Decomposes – 3.26d [19]

acetate [not established] 99.9% at 423–433 Kc

Cu(C2H3O2)2·H2O Copper(II) 199.650 Alfa Aesar, 388 Decomposes – 1.88 [19]acetate [6046-93-1] 99.999% at 513 Kmonohydrate

Pd(C2H3O2)2 Palladium 224.507 Sigma–Aldrich, 478 Decomposes – 3.00e [14]acetate [3375-31-3] 98% at 480–570 K

Intermediate and product speciesAl Aluminum 26.982 933 2792 – 2.70 [19]

[7429-90-5]Al2O3 Alumina 101.961 2326 3253 – 3.97 [19]

[1344-28-1]Au Gold 196.967 1337 3129 6.61 × 10−4 19.3 [19]

[7440-57-5]Au2O3 Gold oxide 441.931 Decomposes at 423 K – N/A [19]

[1303-58-8]Cu Copper 63.546 1358 2868 4.38 × 10−3 8.96 [19]

[7440-50-8]Cu2O Cuprous 143.091 1508 Decomposes – 6.00 [19]

oxide [1317-39-1] at 2073 KCuO Cupric oxide, 79.545 1719 N/A – 6.31 [19]

copper oxide [1317-38-0]PdO Palladium(II) 122.42 Decomposes at 1023 K – 8.3 [19]

oxide [1314-08-5]Sn Tin 118.71 504 2875 3.94 × 10−3 7.27 [19]

[7440-31-5]SnO Tin 134.709 Decomposes at 1353 K – 6.45 [19]

monoxide [21651-19-4]SnO2 Tin dioxide, 150.709 1903 N/A – 6.85 [19]

tin(IV) oxide [18282-10-5]

a Metal acetate melting points are too close to the decomposition temperature to differentiate.b At 2000 K.c Estimated using a visual melting tool (MEL-TEMP 3.0).d As determined using a density analyzer (Micromeritics, AccuPyc 1330).e Estimated density [14].

the gold particles are always spherical and sparsely lo-cated. The tin oxide is initially present in infrequent,small clusters; the primary particles develop as theheight increases; and a larger degree of agglomera-tion begins approximately 3.5 cm above the burner.

The results of the semiautomated image analysisof the particle size distribution of the SnO2 particlesand typical SnO2 particle morphologies are presentedin the top panel images of Fig. 5. The average SnO2particle size and the lognormal distribution are pro-vided for each residence time and sampling height. Asthe residence time increases, the average particle sizegrows from ∼7 to 13 nm and the lognormal distrib-

ution becomes more representative of the population,as indicated by an increased quality of fit to the exper-imental data.

3.4. Characterization of final nanocompositematerials

The effects of the operating conditions and solid-phase reactants on the final product morphology, av-erage SnO2 crystallite size, and metal loading arereported in detail in Miller et al. [6]. The results ofthat study are briefly summarized here for reference.The average SnO2 crystallite size of the product pow-

82 S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87

Fig. 3. SEM images of as-received unreacted (a) gold acetate, (b) aluminum acetate, (c) copper acetate, and (d) palladium acetateparticles.

ders could be varied compared to a baseline undopedSnO2 synthesis condition using the additives and theburner operating conditions. Metal or metal oxide ad-ditives were created (as determined via XRD spectra)depending on the precursor used. For example, onlymetallic gold was identified from the XRD spectra ofthe nanocomposite materials made using gold acetateand TMT, whereas copper oxides were formed whencopper acetate and TMT were used. A range of mi-crostructures was observed from partial encapsulationof the metal additive within a layer of SnO2 particlesto mixed material systems. The metal acetates typi-cally produced metal additive particles that were read-ily identifiable by the size, crystalline structure, andcontrast in the TEM images (see [6] for examples).

XRD spectra of the Au/SnO2 nanocompositesproduced using gold acetate and tetramethyltin arepresented in Fig. 6. Typical morphology of thesenanocomposites is presented as the inset in Fig. 6. TheXRD data demonstrate that the SnO2 is in the cassi-terite phase and the gold is present in metallic form.XRD analyses of the average crystallite size for theSnO2 particles and how the SnO2 is affected by thepresence of the metal additives are presented in detailin Miller et al. [6]. XRD analyses of the metallic goldpeaks conducted as part of the current work yieldedan average particle size of 80 nm for the gold in the

nanocomposites, which is in excellent agreement withthe TEM imaging results conducted previously, wherean average value of 83 nm was determined [2].

The aluminum acetate system was unusual com-pared to the other metal acetates, in that no metalsor metal oxides (besides SnO2) were identified in theXRD spectra. Additionally, it was challenging to de-tect aluminum additives in the TEM images usingcontrast or crystalline structure. After considerableanalysis, some particles were identified as containingaluminum using XEDS. Such an example is presentedin Fig. 7. The particles are noticeably smaller than theSnO2 particles and appear highly fused.

Although the TEM data indicate the presence ofdiscrete metal additive particles, trace quantities ofthe metal additives may also be present within theSnO2 particles, but below the detectable limits ofTEM XEDS. The relatively high vapor pressures ofthe metals associated with the metal acetates supportthis possibility. Table 1 presents the melting and boil-ing points of the metal and metal oxide compounds.At the higher temperatures close to the burner sur-face, most of the materials are above the bulk meltingpoints. Table 1 also provides the vapor pressures at2000 K of the pure metal species associated with themetal acetates. The metal vapors in the reactor canreach high concentrations with mole fractions on the

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 83

Fig. 4. XRD spectra of as-received unreacted (a) gold acetate, (b) aluminum acetate, (c) copper acetate, and (d) palladium acetate.

Table 2Comparison of average crystallite size based on XRD analyses of as-received metal acetates, predicted metal additive size basedon Eq. (2), and TEM-observed metal additive particle dimensions

Precursor 2θ feature usedwith Eq. (1) (◦)

dXRD(nm)

Metal additiveproduct

dMe(nm)

dp,TEM(nm)

Gold acetate 15.4 160 Gold, Au 63.4 83Au(C2H3O2)3

Aluminum acetate 19.6 12 Aluminum oxide 9.8 12Al(OH)(C2H3O2)2 Al2O3

Copper acetate 11.4 (Cu(C2H3O2)2) 26a Copper oxide 12.8 (CuO) 5012.9 (Cu(C2H3O2)2·H2O) 19.9 (Cu2O)

a Average based on the anhydrate and hydrated copper acetate.

order of hundreds to thousands of parts per million.These gas-phase species are highly mobile and likelyto condense on existing nanoparticles as the tempera-tures in the reactor decrease.

4. Discussion

Solid-phase precursors can react in a variety ofmanners in a high-temperature environment. For ex-

ample, the particles can decompose or explode, yield-ing reactive intermediates in the gas, solid, or liquidphases; or the particles can burn in a fashion simi-lar to coal particles where volatiles are emitted, fol-lowed by reaction with the remaining particle “char.”For metals, the burning process can entail gas-phaseoxidation of evaporated metal, surface oxidation lead-ing to a volatile or suboxide species, or surface oxi-dation leading to a nonvolatile oxide deposit, whichmay dissolve into the molten metal [16]. Combustion

84 S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87

Fig. 5. Temperature and particle imaging results as a function of height (h) above the burner surface (i.e., increasing residencetime). The flame temperature data were obtained at a location approximately 2 mm from the centerline of the burner or 26 mmfrom the outer edge of the shroud region. The residence times are based on average velocities measured using high-speedimaging. The top panel of figures presents the tin dioxide particle morphology and primary particle size distributions in thepresence of metal additives. The bottom panel of figures presents the evolution of the gold–SnO2 nanocomposites.

of pure metals has also been observed to be explo-sive, resulting in particle fragmentation and disper-sion [17].

The experimental data obtained in the currentwork indicate the solid-phase metal acetates rapidlydecompose at the temperatures found in the com-bustion synthesis environment. Specifically, the pres-ence of spherical gold particles at low samplingheights/short residence times (see Fig. 5) indicatesthe conversion from solid-phase metal acetate to goldnanoparticles occurs in less than 5 ms. In addition,no gold aggregates or agglomerates were observedat any sampling location, demonstrating the sinteringprocess between gold particles must occur virtuallyinstantaneously. Indeed, calculations for the char-acteristic time for gold nanoparticle sintering [18]yielded values below 100 µs for regions near the sur-face of the burner.

The Au particles produced in the nanocompositesystem can be compared with the gold acetate precur-sor as a means to determine if there is a relationshipbetween the parent and child materials. Such a con-

nection, if it exists, would be valuable for predictingnanocomposite morphology. Assuming spherical par-ticles, the size of the gold particle (the child particle)is related to the gold acetate (the parent particle) via

(2)dMe =(

NρMeAc

ρMe

MWMe

MWMeAc

)1/3dMeAc,

where N is the molar conversion factor, ρ, is thedensity and MW is the molecular weight. The sub-scripts Me correspond to metal and MeAc correspondto metal acetate. Using the densities and molecularweights listed in Table 1 for gold and gold acetateand assuming a one-to-one molar conversion rate,Eq. (2) yields dAu = 0.45dAuAc. Recall that an aver-age gold particle size of dAu = 83 nm was determinedbased on the Au size distribution measured from theTEM images [2]. When Eq. (2) is used to estimatethe characteristic dimension required for gold acetateto directly form 83-nm-diameter particles, a value ofdAuAc = 186 nm is determined. This value is veryclose to the average crystallite size of 160 nm deter-mined using XRD of the gold acetate.

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 85

Fig. 6. XRD spectra of Au-doped SnO2 powder produced using gold acetate as the additive precursor. The features and latticeparameters associated with the cassiterite phase of SnO2 (21-1250) [13] and metallic gold (4-784) [13] are indicated in thefigure. The inset presents a typical TEM image of the nanocomposites. The larger high-contrast particle was identified as Auusing TEM-XEDS.

Fig. 7. TEM image of Al-doped SnO2 powder produced us-ing aluminum acetate as the additive precursor. The smallerparticles were identified as containing Al using TEM-XEDSand are assumed to be in an oxide form.

Using the average crystallite size from the XRDanalysis of the solid-phase reactants and the appro-priate data from Table 1, Eq. (2) can be applied todetermine the size of metal particles in nanocompos-ite materials produced using the aluminum acetateand copper acetate precursors. Table 2 provides themetal particle dimensions for the aluminum and cop-per systems for the appropriate molar conversions

from the precursor crystallite sizes (dXRD). TEM im-ages of aluminum and copper acetate products indi-cate typical particle sizes (dp,TEM) of approximately12 and 50 nm for alumina and copper oxide, respec-tively. Note, the TEM values are biased to larger sizes(smaller particles are more convolved with SnO2 par-ticles and therefore more difficult to identify), andbecause the additives are sparsely located in the TEMimages, the size estimates have high uncertainties.There is good correlation between the metal particlesizes (dMe) and the measured dimensions (dp,TEM)for the gold and aluminum acetates, while there isover a factor of 2 difference for the copper acetate.The discrepancy may be due to different physicaldecomposition mechanisms for the metal acetates.Specifically, the porous precursor materials may un-dergo additional particle fragmentation which maynot occur for the solid-phase reactants with limitedporosity.

The SEM images of the metal acetates show re-markable differences in the structures of the materi-als. Combined with the data of Table 2, the resultsindicate that morphological differences in the pre-cursor materials may affect the final metal additivedimensions, particularly for gold and aluminum ac-etate which exhibit appreciable porosity. The porousstructure supports the theory that the acetate particlescan fragment, potentially along grain boundaries, due

86 S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87

Fig. 8. Schematic of the decomposition of gold acetate and subsequent formation of gold nanoparticles.

Fig. 9. Schematic of the formation of nanocomposite Au/SnO2 materials from mixed solid-phase gold acetate and gas-phasetetramethyltin reactants.

to increased pressure within the pores as the metalacetate vaporizes and decomposes. The metal atomsnucleate locally, and these nanoparticles undergo col-lision to form aggregates and agglomerates. If the sin-tering rates are high, as is the case for gold nanoparti-cles, spherical particles are formed with dimensionson the order of the grain size of the gold acetate.Fig. 8 presents a schematic representing the steps ofmetal acetate decomposition and metal particle for-

mation suggested by the results for the porous mate-rials. When the copper and palladium acetates vapor-ize and decompose, these denser structures may notfragment along grain dimensions. Without the addi-tional fragmentation, neighboring grains can coalesceto form larger particles as observed in the TEM im-ages [6].

A schematic integrating the gold acetate and TMTsystems during the formation of nanocomposites is

S.D. Bakrania et al. / Combustion and Flame 148 (2007) 76–87 87

presented in Fig. 9, where the early gold acetate de-composition processes are omitted for clarity. Thepresence of the metal additives can affect the SnO2formation and growth processes in a variety of com-plex ways. The metals can induce new reaction path-ways via homogeneous and heterogeneous catalysis.The temperature and flow fields can be altered whenthe additive precursors are introduced. The metal ad-ditives can modify the SnO2 grain growth kinetics andprovide sites for heterogeneous deposition and SnO2particle growth. Based on the results of the currentstudy, the spherical gold nanoparticles are presumedto interact with SnO2 primarily through interparti-cle collision and aggregation and as sites for surfacegrowth. The coalescence of SnO2 is slower than thatof gold, resulting in the agglomerated structures.

5. Conclusions

Controlling nanocomposite morphology is a vi-tal step toward designing new and advanced optical,semiconductor, catalyst, and sensor materials. In par-ticular, encapsulation of metals has yielded remark-able photonic properties. Combustion synthesis meth-ods offer the potential to produce such high-value-per-gram materials at high rates with good controlof the nanocomposite properties. The results of thecurrent work indicate that metal acetates provide anexcellent source for metal additives in flame synthesisenvironments. Metal acetates can also provide accessto nanosized additives, where the approximate dimen-sion of the metal additive can be predicted using thegrain size of the parent material. The formation ofnanodimensioned metals and the ease of handling,low toxicity, and relatively low cost of metal acetatespoint toward a promising new scope for the synthesisof nanocomposite materials.

Acknowledgments

The authors acknowledge the generous support ofthe National Science Foundation (NSF Grant DMI0329631 and NSF Instrumentation for Materials Re-search Grant DMR-0315633).

References

[1] N. Yamazoe, Sens. Actuat. B 5 (1–4) (1991) 7–19.[2] T.A. Miller, S.D. Bakrania, C. Perez, M.S. Wooldridge,

Tin dioxide/metal/metal oxide nanocomposites for gassensor applications, in: E. Rosenberg, K.E. Geckeler(Eds.), Functional Nanomaterials, American ScientificPublishers, CA, 2006, pp. 1–24, chap. 30.

[3] E. Prodan, C. Radloff, N.J. Halas, P. Nordlander, Sci-ence 302 (5644) (2003) 419–422.

[4] C. Radloff, R.A. Vaia, J. Brunton, G.T. Bouwer, V.K.Ward, Nano Lett. 5 (6) (2005) 1187–1191.

[5] Y. Chan, J.S. Steckel, P.T. Snee, J.-M. Caruge, J.M.Hodgkiss, D.G. Nocera, M.G. Bawendi, Appl. Phys.Lett. 86 (7) (2005) 73102-1-3.

[6] T.A. Miller, S.D. Bakrania, C. Perez, M.S. Wooldridge,J. Mater. Res. 20 (2005) 2977–2987.

[7] M.S. Wooldridge, P.V. Torek, M.T. Donovan, D.L.Hall, T.A. Miller, T.R. Palmer, C.R. Schrock, Combust.Flame 131 (1–2) (2002) 98–109.

[8] D.L. Hall, P.V. Torek, C.R. Schrock, T.R. Palmer, M.S.Wooldridge, Mater. Sci. Forum 386–388 (2002) 347–352.

[9] D.L. Hall, A.A. Wang, K.T. Joy, T.A. Miller, M.S.Wooldridge, J. Am. Ceram. Soc. 87 (11) (2004) 2033–2041.

[10] D.L. Hall, Gas-Phase Combustion Synthesis of Nano-crystalline Tin and Tin Oxide Particles, M.S. thesis,Mechanical Engineering, University of Michigan, De-cember 2001.

[11] J. Ma, Soot Formation During Coal Pyrolysis, Ph.D.dissertation, Chemical Engineering Department, Brig-ham Young University, August 1996.

[12] S. Niksa, R.E. Mitchell, K.R. Hencken, D.A. Tichenor,Combust. Flame 60 (2) (1985) 183–193.

[13] Powder Diffraction File, compiled by the JCPDS, Inter-national Centre for Diffraction Data, Swarthmore, PA,1990.

[14] P.K. Gallagher, M.E. Gross, J. Therm. Anal. 31 (6)(1986) 1231–1241.

[15] M. Kristl, M. Drofenik, Inorg. Chem. Comm. 6 (2003)1419–1422.

[16] E.L. Dreizin, Prog. Energy Combust. Sci. 26 (2000)57–78.

[17] E.L. Dreizin, Combust. Explos. Shock Waves 39 (6)(2003) 92–96.

[18] S.D. Bakrania, C. Perez, M.S. Wooldridge, Proc. Com-bust. Inst. (2006), doi:10.1016/j.proci.2006.08.020.

[19] CRC Handbook of Chemistry and Physics, 85th ed.,CRC Press, New York, 2004.