The Oxidation of Soot a Review of Experiments, Mechanisms and Modesl_Stanmore-Etal_2001

Carbon 39 (2001) 2247–2268

Review article

The oxidation of soot: a review of experiments, mechanismsand models

*B.R. Stanmore , J.F. Brilhac, P. Gilot´Laboratoire Gestion des Risques et Environnement, Universite de Haute Alsace, 25, rue de Chemnitz, 68200 Mulhouse, France

Received 8 January 2001; accepted 13 April 2001

Abstract

Soot may be formed when carbonaceous fuels are burned under local reducing conditions. Its subsequent oxidation is ofgreat significance for pollution control in industrial flames, auto engines etc. Oxidation (gasification) can be achieved withoxygen, carbon dioxide, water vapour or nitrogen dioxide. In this review, the experimental techniques which have been usedto study the gasification of soot are described and the methods and results obtained by analysis of the data from them areconsidered. Firstly, the mechanism of soot formation and its structure are briefly discussed. The various scales of particulatewhich comprise it, i.e. spherule, particle and aggregate, influence its properties and behaviour. Next, the experimentalequipment used in the study of its gasification is briefly described. Gasification kinetics at low temperatures are measuredeither in fixed beds or by thermogravimetry. The apparatus may be operated as a thermally programmed desorption system toidentify the species involved. High temperature investigations have been carried out in entrainment burners and shock tubes.The chemistry of soot oxidation is discussed for both non-catalytic and catalytic conditions. The oxidation pathway involvesinteraction between adsorption and desorption processes, which determine the primary products, the order of reaction and theactivation energy. The concensus is that two types of adsorbed surface species are present in uncatalysed combustion. Thecombustion mechanism of individual spherules is considered in terms of basic property changes. During thermogravimetry,the influence of the competition between reaction and oxygen diffusion in soot beds is analysed. The reaction of catalysedsoot displays a different mechanism, as the primary products, the order of reaction and the activation energy all change. Thelower activation energy and higher reactivity lead to lower ignition temperatures. Catalysts may be incorporated into the sootspherules by addition to the fuel, or may be added after formation. Two types of contact between the carbon and addedcatalyst have been identified, ‘loose’ and ‘tight’. Tight catalyst, which has been mechanically ground with the soot, producesmore pronounced effects. Finally, the behaviour of soot during gasification by other oxidants, namely H O, CO and NO is2 2 2

summarised. 2001 Elsevier Science Ltd. All rights reserved.

Keywords: A. Soot; B. Gasification, Oxidation; D. Chemical structure, Reactivity

1. Introduction of diesel exhaust from a catalytic converter have beenreported [1,2]. The finer particle sizes present in ambient

Soot is produced as an unwelcome byproduct in many air, which are formed during combustion, are responsiblepractical combustion systems. With internal combustion for increased mortality [3].engines, particularly diesels, it can foul exhaust systems However, in large flames soot can perform a usefuland generate dark exhaust plumes. More significantly, it function by acting as a radiating agent to promote heatpresents a significant environmental hazard when respired, transfer. Because of these diverse implications, the litera-as it often contains adsorbed PAH. The biological effects ture on soot is scattered. This article reviews those aspects

of the literature which relate to its combustion.Two temperature ranges are of interest, flame tempera-*Corresponding author. Present address: Ecole des Mines

tures (over 8008C) at which its formation competes with itsd’Albi-Carmaux, Route de Teillet, 81013 Albi Cedex 09, France.oxidation, and low temperatures (300–7008C). The latterTel.: 133-5-6349-3000; fax: 133-5-6349-3099.

E-mail address: [email protected] (B.R. Stanmore). are found in diesel exhaust systems, and the controlled

0008-6223/01/$ – see front matter 2001 Elsevier Science Ltd. All rights reserved.PI I : S0008-6223( 01 )00109-9

2248 B.R. Stanmore et al. / Carbon 39 (2001) 2247 –2268

Nomenclatured particle diameter (m)

2 21D diffusion coefficient (m s )e depth of soot bed (m)ERE extended resistance equation

21h mass transfer coefficient for oxidant (m s )m21 21k reactivity on a carbon basis (kg kg s )

23 23 21 21K reactivity on an oxidant basis (kg m (kg m ) s )n order of reaction (–)P partial pressure of oxidant (Pa)RVVR reactor to visualise and video regenerationx displacement into bed from the surface (m)y mass fraction of oxidant (–)u extent of carbon burnout (–)a exponent in ERE equation

21b stoichiometric coefficient (kg kg )ox carbon

e porosity (–)u Thiele modulus (–)

23s density (kg m )

22 21v flux of oxidant (kg m s )

Subscriptsa adsorptionb bedd desorptione effectiveg gaso initialox oxidantp spherules surface

oxidation of particulate emissions using regenerative traps concerning its formation and growth in flames, and only amust be initiated in this range. Where possible, the brief summary is necessary here.discussion will refer to soot behaviour, with some mention Some older reviews of the topic are still useful andof other turbostratic materials if necessary. should be consulted, for example [4–7]. A volume of

Gasification by CO and H O will also be briefly workshop papers is available [8]. The influence of engine2 2

considered. Recent developments in the control of automo- operating conditions on the properties of diesel soot hastive diesel emissions have raised the possibility of using been described [9].nitrogen dioxide NO as a soot oxidant. It is more reactive It is universally accepted that sequential nucleation,2

than oxygen at low temperatures and is the focus of growth and coagulation steps take place. Soot inceptionongoing research. The C–NO reaction will be therefore begins in a laminar diffusion flame for a range of hydro-2

be included in the discussion. carbons at temperatures between 1585 and 1700 K [10]. ItIn order to interpret combustion behaviour it is neces- appears that uncharged radicals and molecules are the

sary to understand the structure and hence the formation flame nucleation precursors, i.e. nucleation units [11].mechanism of soot. A brief description of each of these Growth by addition of gaseous carbonaceous species thenfollows. produces the particles which are comprised of the

spherules described in the next section. The particlescoagulate into aggregates as the combustion gases cool.

Many mechanistic models for describing speciation and2. Formation of soot soot formation in flames are available [12]. Acetylene and

PAHs appear to supply the bulk of the carbon laterThe term soot is given to the particulates formed during incorporated into the soot spherules. Krestinin [13] post-

combustion of carbonaceous fuels under substoichiometric ulates that polyynes (C H , n52, 3, . . . ) are the major2n 2

conditions, either by design e.g. for the production of precursor radicals, partly because their stability increasescarbon black, or in poor mixing conditions (e.g. diesel with temperature.soot). There is a huge and rapidly growing literature With solid fuels, e.g. coal, soot is formed by the

B.R. Stanmore et al. / Carbon 39 (2001) 2247 –2268 2249

pyrolysis of tarry materials ejected by particles undergoing soluble organic fraction (SOF) and inorganic material¨devolatilisation. The droplets dehydrogenate and/or par- (mostly sulphates). Ahlstrom and Odenbrand [19] found

21tially oxidise into chains of soot spherules. Most are that on heating a diesel soot at a rate of 10 K min ,burned before exiting the flame region, but a significant hydrocarbons were lost in the range 150–4008C, with aresidual can be found in the flyash of low NO burners pronounced peak at 2008C. Courcot et al. [20] tested ax

[14]. highly impregnated soot and found three distinct hydro-The soots from various sources, i.e. from different fuels carbon peaks.

and combustion conditions tend to be similar in size and The SOF and other adsorbed species such as sulphatescomposition, which indicates that the same formation and water are captured by the soot in the gas cooling phaseprocesses and growth constraints apply. This has encour- e.g. in the exhaust pipe of a diesel engine [21]. Theaged researchers to produce soot for study in simple quantity of SOF is a function of engine operating con-laboratory combustors [15], or to use commercial carbon ditions, and may be as low as 5% (loaded engine) or asblack as a surrogate [16,17]. high as 60% (engine idle) of the total mass. The sample

analysed above contained little SOF. The mass loss duringsolvent extraction (SOF) is almost the same as the mass

3. Composition and structure of soot lost by thermal treatment under inert gas to 6008C [22].Both lubricating oil and unburned fuel contribute to the

The nature of soot has been discussed at length in SOF of diesel soot [23,24].previous reviews, see for example Ref. [7]. The spherules are joined together by shared carbon

The ultimate analyses of a typical diesel soot and the deposition to form loose particles of 0.1–1 mm size. Thesame material degassed at 13 mPa for 5 h at 1508C [16] nitrogen BET area of a soot was found to be only 40% ofare given in Table 1. the external surface area calculated for spherules whose

The sulphur was apparently present as compounds diameter was measured by electron microscopy [25].adsorbed onto the surface (as sulphates), whereas the Though the porosity of these particles may be of the orderoxygen was strongly bonded. FTIR analysis revealed the of 0.95, they cannot be broken down into the individualpresence of C=O, C–O–C and C–OH bonds, and some spherules, even by strong ultrasound treatment [16].aromatic structures. The unaccounted-for mass in Table 1 Viewing of the spherules by electron microscopy revealsis probably in the form of silicon, aluminium and metals. laminations with surface steps, which are produced by

Soot as sampled, e.g. from a dilution tunnel, is found to numerous crystallites in concentric orientation [7]. Thebe in the form of agglomerates which are around 100 mm form of the BET adsorption isotherms can give infor-in size. These agglomerates are composed of smaller, very mation about the degree of surface defects [26], because ofopen ‘particles’, which are in turn a collection of smaller high energy sites at the edges.carbonaceous spherules. The terms agglomerate (100 mm X-ray diffraction indicates that the structure of thetypical size), particle (0.1–1 mm) and spherule (10–50 nm) crystallites has an inter-layer spacing of 0.36–0.35 nmwill be used for these three scales of particulate. Fig. 1 is a [23,27], i.e. slightly more than in pure graphite. Crystallitemicrograph of some typical particles, showing the in- thickness is reported as 1.2 nm by Heywood [23] and 1.4dividual spherules. nm by Ishiguro [27]. There are two to five platelets of

The agglomerates are readily disaggregated into par- hexagonal face-centred arrays of carbon atoms per crys-ticles by treatment in an ultrasound bath [16]. The mecha- tallite. Dislocations and five- and seven-membered ringsnism of cohesion is surface forces to which adsorbed produce surface wrinkling. Heat treatment above 10008Cmaterials play a part. The method of sampling influences induces crystallite growth and increasing order in thethe morphology, because of different conditions during the structure [28].coagulation process. Otto et al. [29] found similar surface areas when using

The fundamental unit of the soot agglomerates are the either argon or carbon dioxide as adsorbent. The surfacespherules with diameters of 10–50 nm [18]. As the name areas of soots as measured by BET analysis using nitrogenimplies, most of these particles are almost spherical, but a at 77 K as adsorbent give results in the range from 20 [30]

2 21number of less regular shapes may be found. The surface to 230 m g [29]. Since the theoretical surface area ofof the spherules has adhering hydrocarbon material or smooth spheres of diameter 25 nm and density 2000 kg

23 2 21m is 120 m g , the extent of pore surface is not large.The data generally suggest that the pore surface area is less

2 21Table 1 than 100 m g .Typical soot analyses ¨Ahlstrom and Odenbrand [19] found that the BET

surface area of a diesel soot increased from about 35 toElement C H N O S2 21270 m g simply by heating to 6008C under nitrogen.

Virgin soot 83.5 1.04 0.24 10.5 1.13 This implies that the SOF present is plugging someDegassed soot 83.8 0.85 0.22 10.7 0.10 already-existing micropores.


Fig. 1. Micrograph of diesel soot, showing particles consisting of clumps of spherules.

23The sizes of the pores in spherules of some carbon black ranges from 60 to 400 kg m , which corresponds toblacks have been measured by Stoeckli et al. [31]. They a wide range of consolidation. Mechanical compaction of

23have a slit shape typical of carbon and are of the order of 1 soot beds can increase the density to over 500 kg mnm in width, compatible with the half-width of 0.4–0.5 nm [33]. The porosity of the bed e is given from the bedb

measured by Buczek et al. [32] in activated carbon. The density s and the spherule density s by:b p

specific internal surface area (i.e. excluding the external2 21 ssurface) was 70–140 m g depending on the measure- b

]e 5 1 2b sment technique. The void space in the spherules was p3 21between 0.04 and 0.075 cm g , which indicates

23porosities in the range of 8–15%. Negligible SOF is If the spherule density can be taken as 2000 kg m , thepresent in carbon black. equivalent porosities for the bed densities mentioned above

The density s of uncompacted beds of soot and carbon are therefore 0.97–0.80. These porosities represent theb


overall bed, and will be a combination of particle porosity where d is the particle diameter (m), s is its density (kgp23 21 21and the void space in between. The intra-particle voidage m ), k is the carbon reaction rate (kg kg s ), b is the

of some carbon black particles was measured by Lahaye stoichiometric mass ratio of oxidant to carbon, D is thee2 21and Ehrburger-Dolle using thermoporometry [34]. They effective diffusion coefficient (m s ) and s is the bulkox

23conclude that single particles of commercial carbon blacks gas concentration of oxidant (kg m ). The modulus isare formed by strings which are a few (2–3) spherule applied by means of the effectiveness factor which repre-diameters in thickness. sents the fraction of internal surface which is able to react

Many workers investigating diesel trap regeneration as if it were exposed to the surface concentration ofhave substituted a commercial carbon black for diesel soot reactant gas [37].because it has very similar properties to the soot, and is When the Thiele modulus is large (u .1), the reactiongenerated in a controlled, reproducible manner [16,17]. rate outstrips the ability of diffusion to supply reactant andThe spherule size, surface area and overall reactivities tend the effectiveness factor is small. When u is small, diffusionto be very similar. However, some differences between the dominates reaction and the effectiveness is high. Suchtwo in the progress of combustion can be identified [17]. factors as particle size, pore structure and particle tempera-

ture determine the combustion mode. For carbon oxidation,the coefficient b depends on the primary product ofoxidation, being 4/3 when carbon monoxide is the product

4. Gasification and 8/3 for carbon dioxide.Because the nature of the combustion is strongly

4.1. Gasification processes influenced by the penetration of oxygen into soot struc-tures, the evaluation of diffusion within the material

Although the combustion of carbon is an ancient becomes of fundamental importance. This applies topractice, scientific understanding of the phenomenon is diffusion both within an individual spherule and within astill the subject of extensive research. The theory of the bed of particles. Both of these aspects are discussed belowoxidation of a solid carbon by oxygen, steam, carbon (Sections 5.1.2. and 5.1.3.).dioxide or nitrogen dioxide involves both chemical kineticswith the consequent heat transfer, and mass transport at a 4.2. Experimental equipmentnumber of levels. Gaseous oxidant must move to thesurface and be adsorbed to form surface intermediates, Because of the experimental similarities, the samewhich then rearrange, desorb and escape into the gas equipment is used for all four oxidants, O , CO , H O and2 2 2phase. The carbon surface may be the wall of a pore deep NO .2inside the solid, so that diffusion of oxidant from, and ofproduct to, the bulk gas phase is also involved. 4.2.1. Low temperature oxidation

In order to systematise the discussion, combustion will The low temperature (,8008C) oxidation properties ofbe considered at low (,8008C) and high temperatures soot which are of interest are the intrinsic kinetic rates, the(.8008C). Low temperature conditions apply to many ignition behaviour and the spread of an oxidation frontlaboratory systems and to the regeneration of diesel traps through a bed. Most of this work is driven by the need towhere the material is present as a bed, i.e. aggregated. The understand the regeneration characteristics of particulateshigh temperatures tend to apply to flame situations where in regenerative traps attached to diesel exhausts [21]. Inentrainment flow applies, i.e. isolated particles are present. doing so, it is important to identify the gasification

When oxidation rates are specified on a surface basis it behaviour of carbon both from the individual spherules,is usual to consider three regimes (see for example Smith and from the bed as a whole.[36]). Regime I applies when the oxidant fully penetrates Much effort has been expended in studying low tem-the solid and all the internal surface is reacting, regime II perature soot oxidation, under catalysed and uncatalysedwhen oxidant penetration is partial and diffusion into the conditions. Because of the slow reaction rates at lowparticle is insufficient to supply all the reaction, and regime temperatures, all these studies were carried out on sootIII when reaction takes place at the outer surface only. beds. To obtain accurate kinetics, the beds may be diluted

Fundamental to the analysis of the combustion of porous with a large excess of inerts such as quartz [38], kaolinsolids is the effectiveness approach pioneered by Thiele. It [39], silicon carbide [17] or aluminium oxide, with orwill be applied here both to reaction within an individual without catalyst [15]. This diminishes the temperature risespherule and to the whole of a soot bed. When applied to which results from reaction and facilitates the transfer ofporous spherical particles with the reaction in oxidant oxidant within the bed.taken as first order, the Thiele modulus u [37] is: Both pseudo steady-state and transient experiments have

been carried out. The equipment used at low temperatures1s kbd ]p is generally a thermogravimetric analyser or a fixed bed.2] ]]u 5 S D2 D se ox Some specialised equipment has been developed to simu-


late diesel soot deposits in exhaust filters (see Section diffusion rates, the available oxygen then is rapidly4.2.1.3.). removed by reaction. This results in combustion being

limited to the surface of the bed, with the active layer4.2.1.1. Fixed beds. The beds are kept comparatively being less than 1 mm in thickness [43].thin and therefore operate in differential mode with Reported rates which do not take account of thesenegligible change in concentration of oxidant, except in effects may be too low [44]. This will especially be true atsome cases for the exceptionally reactive NO . For kinetic the higher temperature range of a test series. Even if a soot2

measurements, the systems are generally run isothermally, sample is diluted with an inert solid, the supply of oxygenwith an inert gas passing through the system during heatup can be limited. These aspects are developed later inuntil the desired combustion temperature is reached. A Section 5.1.3. concerning burning in a bed.complementary mode is temperature programmed oxida-tion (TPO), in which the sample temperature is raised 4.2.1.3. Other systems. A simple fused silica flow reac-slowly under oxidising conditions and the gases evolved tor containing 150 mg of soot in a stainless steel pan wasare continuously analysed. This approach is useful in developed by Murphy et al. [46] to measure ignitionidentifying reaction mechanisms and the temperatures at temperature. The temperature of the reactor was slowlywhich they occur. raised in a furnace and ignition was determined by a rapid

The carbon beds studied by Neeft et al. [17] were rise in bed temperature as measured by a thermocouple.diluted 10:1 by mass with silicon carbide. A commercial Purpose-built equipment has been developed to identifycarbon black Printex U and a diesel soot were used. ignition behaviour under realistic diesel trap conditions. InCiambelli et al. [39] used a 100:1 excess of quartz beads to each case a circular soot bed was formed on a porous

¨dilute the sample. Ahlstrom and Odenbrand [19] did not ceramic substrate and simulated exhaust gases were passeddilute the bed of uncatalysed soot but used a small fixed through it. In the apparatus of Herz and Sinkovitch [47] the

21bed at the high space velocity of 70 000 h , which should sample was irradiated with an IR heating lamp through abe sufficient both to control temperature and to ensure an calcium fluoride window. An IR pyrometer and a videoadequate oxygen supply. camera recorded the surface temperature and the condition

Ma et al. [40] promoted oxidation by irradiating a heated of the bed.bed with microwaves. Diesel soot was found to respond This type of apparatus was copied [48] and designatedbetter than a commercial carbon black. an RVVR (reactor to visualise and video regeneration). The

ignition behaviour in this version was found to be similar4.2.1.2. Thermogravimetric analysis. There have been a to that found by Herz and Sinkovitch [47] (see Sectionmultitude of studies, some with temperature ramping and 5.1.1.3.). Various bed masses, thicknesses, heating regimesothers isothermal. Ramping tests are useful for identifying and oxygen concentrations were trialled.the quantity of adsorbed material (SOF) and the tempera- The beds used by Petersen [49] were collected on a filterture of initiation (see Section 5.1.1.3.). Fig. 3 is typical of substrate directly from the exhaust of an operating engine.the thermograms obtained. The loss of SOF appears as the The substrate was transferred to a flow reactor where itinitial decline in mass. The burnoff of catalysed soot was ignited by preheated gas. Three thermocouples mea-begins at a lower temperature than uncatalysed. TPO can sured bed temperature across a radius.be readily carried out in a thermobalance, but isothermal There are many studies reported for operating enginestests are better for measuring kinetic rates. with in situ regeneration, mostly for catalysed systems

There is a range of thermobalances on the market, with (Section 5.2.). Since they provide little fundamental data,various configurations of sample containment /gas flow they will not be described here.contact configurations. If care is taken to ensure that thetemperatures recorded are those of the sample, reliable

4.2.2. High temperature oxidationreaction rates can be obtained for particles in the pulver-High temperature combustion takes place inherently inised fuel size range [41]. A new design of thermobalance

flames after the formation of soot. For instance, it ispermits flow-through operation of the reactant gas, whichestimated that more than 90% of the soot formed in dieselshould minimise mass transfer limitations.engines is combusted before the exhaust gases leave theHowever, this may not be the case with soot burned in acylinder [23].crucible. Though the effectiveness factor may indicate

regime I combustion within the bed, transport limitationscan apply between the mouth of the crucible and the bed 4.2.2.1. Laminar burners. Studies at flame temperaturessurface. In a series of DTG studies, Gilot and co-workers have been carried out in laminar flow burners [25,50,51],[42–45] studied the interaction between kinetics and mass generally directly coupled to a sub-stoichiometric soot-transfer. The transport of oxygen to the surface of a sample generating burner. The system used by Lee et al. [50]in a DTG crucible is by diffusion. Because of the large incorporated both steps into the one burner by forming ansurface area of carbon per unit volume of sample and poor annular column of soot conveyed by a vertical gas flow


which was attacked by oxygen supplied around the sorption /desorption experiments on a lignite char andperiphery of the column. Spherocarb, Lear et al. [57] postulate the formation of a

In the investigation of Neoh et al. [25] freshly formed metastable oxide to explain the initial increase in weight.soot was burned in entrainment flow at temperatures of The metastable complex involves dioxygen which after1580–1860 K, using light scattering and absorption tech- adsorption desorbs only after a time delay or at higherniques to measure particle number populations and size. A temperatures, giving CO as the dominant product.flat flame secondary burner was fed with the products of a A good introduction to carbon oxidation is provided bysubstoichiometric primary flame burning the fuel. Marsh [58]. Moulijn and Kapteijn [59] argue that at

gasification temperatures the carbon surface will be rela-4.2.2.2. Shock tubes. The high temperatures produced by tively simple. They identify three types of carbon–oxygenthe compression wave in a shock tube have traditionally complex of increasing stability: carbonyl, semiquinone andbeen used to investigate gas phase reactions over very pyrone groups. Only a fraction of the oxygen-containingshort time intervals. Soot can also be studied, because the complexes take part in gasification reactions [60].small spherule size (high surface area) allows significant Ahmed et al. [61] examined thin films of pyrolyticburning to occur during the milliseconds of elevated carbon deposited from a sub-stoichiometric methane flame.temperature [52,53]. The soot is suspended in the gas near From temperature-programmed desorption (TPD) theythe reflecting wall before the shock is initiated. It is identified two types of adsorbed oxygen complex, andclaimed that the shock assists in dispersing any aggregates propose different sites for CO and CO release.2

[52]. Temperature is calculated from the initial gas con- Du and co-workers [62,63] examined the adsorption /ditions and wave velocity; temperatures up to 3500 K have oxidation steps of oxygen with a soot produced in abeen achieved. laboratory premixed ethylene flame. This method of prepa-

Roth et al. [53] burned soot in suspension using a shock ration ensured that no metals were included. The meantube to initiate combustion. They measured CO/CO spherule size was 43 nm. By TPD they identified a critical2

concentrations using IR-diode laser and the particle prop- activation energy for desorption, below which the oxygenerties with in situ laser light extinction. complex is rapidly desorbed and above which it is persis-

tent. This critical value varies exponentially with tempera-ture but only slightly with time.

5. Gasification with oxygen (combustion) Du and co-workers [62,63] tested the hypothesis ofAhmed [61] and others that two types of adsorbate are

5.1. Non-catalytic combustion present. They found a range of activation energies withtwo maxima, rather than two distinct activation energies

The oxidation of carbon is catalysed by a range of corresponding to desorption at each site. The activationmetals, which can introduce different oxidation pathways. energies do not change with extent of burnout. TheFor this reason catalytic activity will be reviewed under a desorption energy for the step which releases CO was 285

21separate heading. and 335 kJ mol for CO .2

The more active sites filled (forming CO) during the5.1.1. The chemistry of carbon oxidation adsorption phase and only the slower (CO ) sites were2

Detailed investigations of carbon oxidation show that active during oxidation. The activation energy during21molecular O and the O and OH radicals all participate in oxidation was approximately 145 kJ mol . The occur-2

soot oxidation [54]. OH is particularly effective [50,53]. rence of adjacent active sites with surface migration and aRoth et al. [55] showed that hydrogen peroxide will assist Gaussian range of activation energies was incorporatedsoot oxidation at low temperatures, due to the presence of into an extensive model of the system by Du and co-high concentrations of OH radical. The activation energy is workers [62,63] which predicts most of their experimental

21very low with this reagent, namely 11 kJ mol . data.During oxidation at 5008C Ishiguro [27] found that the The action of the oxidants O , H O, CO , NO, N O and2 2 2 2

IR spectra showed increased C=C stretching vibration with NO involve at least two steps: (i) an O-atom is transferred2

increasing burnout, indicative of the presence of polycyclic from the gas to form a solid complex, and (ii) the complexcompounds. At the same time the intensity of the C=O decomposes and a C-atom is lost from the surface. Thepeak decreased. two-step sequence involves oxidant at each step and

Ismail and Walker [56] found that there are ‘superactive’ produces carbon monoxide. The dominant parameter issites in Saran char which generate measurable reactivity to oxygen occupancy, which is a function of the oxidant.oxygen at temperatures around 2008C. The downwards Molecular oxygen O produces a heavily occupied and2

extrapolation of higher temperature results predicts that no hence reactive surface at low temperatures.reactivity should exist. The sites are rapidly destroyed at The reaction scheme proposed by Marsh and Kuo [58]higher temperatures. involves free carbon sites C , chemisorbed localised mo-f

As a result of low temperature (150–4508C) O ad- lecular oxygen C(O ), chemisorbed mobile molecular2 2


oxygen C(O ) , chemisorbed localised atoms of oxygen2 m

C(O) and chemisorbed mobile atoms of oxygen C(O) :m

C 1 O → C(O ) or C(O )f 2 2 2 m

C(O ) → C(O) 1 C(O) or /and → C(O) 1 C(O)2 m m m m

or /and → C(O) 1 C(O)

C(O) → CO

C(O) → COm

C(O) 1 C(O) → C 1 COm m f 2

C(O) 1 C(O) → C 1 COm f 2

CO 1 C(O) → C 1 COf 2

CO 1 C(O) → C 1 COm f 2

O 1 2C(O) → 2CO2 2

Transient operation of a fixed bed at temperatures between350 and 7008C was carried out by de Soete [38] to identifygasification mechanisms. A bed of soot diluted with quartzbeads was brought to pseudo steady-state conditions under

Fig. 2. Influence of temperature on the rates of adsorption ofa flow of oxidant, and then the gas flow was switched tooxygen and desorption of oxidation products from a soot surfacenitrogen. This permitted the rate constants for the overall[38]; k P 5adsorption of oxygen, k 5desorption of CO, k 51 O 2 3gasification, oxidiser adsorption, CO desorption and CO 22desorption of CO .2desorption reactions to be determined. The oxidants used

were oxygen, steam and carbon dioxide.The apparent combustion rates of soots generated from

pure hydrocarbons, namely an aromatic (a-methylnaph- peaks for SOF and carbon are common in TPO outputs,thalene, a-MN), and an aliphatic (n-hexadecane, n-HD) e.g. Querini et al. [65].were measured by this technique. Fig. 2 shows the relative Ishiguro et al. [27] found that the size of the layer planesmagnitudes of the kinetic rates of the O adsorption (k of crystallites was 0.7–1.0 nm when unburned, but rose to2 1

P for 21 kPa partial pressure), CO desorption (k ) and 2–4 nm at 75% burnoff. At the same time the crystalliteO 22

CO desorption (k ) reactions as defined by de Soete [38]. thickness increased to 1.6 nm. These authors report that the2 3

The rate of adsorption of the oxidiser k P is lower than turbostratic outer layers become more crinkled as burnout1 O2

the desorption steps in this temperature range, especially as proceeds, no doubt as a result of the expansion. They alsothe surface coverage is low. Adsorption completely con- identified small flakes of crystallite spalling of the outsidetrols the reaction processes and the adsorption kinetics are layers and conclude that this is the result of strain energyalways close to the overall carbon gasification rate. The in the layer plane.coverage of active sites was low, and the reaction order A number of studies have shown that the initial stages ofwas virtually 1.0. The ratio of CO to CO in the offgas combustion display anomalously high combustion rates2

rose with temperature, but was generally of the order of [17,38,39], which has been attributed to the removal ofunity. SOF. However it is generally accompanied by a rapid

Most experimental studies of soot combustion are initial rise in pore area, which indicates that blocked porescarried out with the SOF still in place, and its presence are being opened up to adsorbent. It may involve chaoticcould influence combustion behaviour. De Soete [38] edge carbons as well as SOF.found a difference in the kinetics of soot with SOF present De Soete [38] found that oxygen adsorbates occupiedand with it removed by treatment with nitrogen at 11008C. only a small fraction (,5%) of the surface sites. ThisThe thermally-treated a-MN and n-HD materials (called result was confirmed by Ma and Haynes [66] forchar) showed a diminished reactivity at temperatures below spherocarb oxidation at low temperatures. They concludeabout 6008C (900 K), but similar behaviour above this that the carbon surface displays heterogeneity with regardtemperature. to adsorption processes. The adsorption step results in the

There is little indication that the adsorbed hydrocarbons release of a molecule of CO or CO for every molecule of2

facilitate ignition during slow heatup tests [49,64]. They O adsorbed:2

tend to be vaporised and therefore lost as the solids arebrought up to ignition temperature. Distinct conversion –C 1 O → –C(O) 1 CO, CO2 2


5.1.1.1. Primary products. The ratio of carbon monoxide fundamental principles it can be shown that changes into dioxide in the product gases is dependent on the supply density and particle diameter during burnout are related by:of oxygen, e.g. the concentration in the gas [15]. With an

as dexcess of oxygen at low temperatures, the ratio CO/CO2 ] ]5S Ds dhas been reported as ,0.12 [67], but mostly is of the order o o

of unity [16,17,38,42]. At temperatures up to 6008C, itwhere the subscript o denotes initial conditions beforerises with temperature, but falls with the addition of watercombustion. As a guide, a is zero for a true shrinking corevapour as a result of the water gas shift reaction [17].situation, of the order of 1–6 for regime II conditions and¨Ahlstrom and Odenbrand [19] showed that in the presencegreater than 6 for regime I.of water vapour, the selectivity for the production of CO is

It follows from consideration of the ERE that reactiondictated by the water gas shift equilibrium.order is a function of temperature [70]. Essenhigh’sAt higher temperatures, CO is the dominant primaryanalysis [68] for pulverised particles (|100 mm) in experi-product. However, above 7008C, CO begins to be homoge-ments gives the effective reaction order as zero at 1000 K,neously oxidised to CO and conversion is virtually2 rising to unity at 2000 K. This implies that the desorptioncomplete at 8008C [16].rate k controls at low temperatures, but the adsorption rated

at high temperatures, which is contrary to the findings ofde Soete. As will be seen later, some aspects of Es-5.1.1.2. Order of reaction. Order of reaction has tradi-senhigh’s analysis may not apply to soot spherules.tionally been expressed as an exponent on the reaction gas

At high temperatures, the reaction order is reported aspartial pressure P or concentration:ox

unity by Lee et al. [50]. The correlation of Nagle andE n 21 Strickland-Constable [71] is first order at low oxygen]k 5 A expS2 D P soxRT partial pressures, but approaches zero order at higher

pressures. Park and Appleton confirmed this trend atHowever, the values for n fitted to experimental results

oxygen pressures greater than one atmosphere [52]. For airhave varied widely and have been difficult to interpret. The

at temperatures below 1500 K, the value is practically 1.0.bulk of the studies at low temperature report the order of

Roth et al. [53] found the system was roughly first orderreaction in oxygen to be unity [16,29,33,38,42]. De Soete

with respect to O in the range 1600–2500 K.2[38] found that adsorption of oxygen controlled the rate(see Fig. 2), which implies a first order system.

5.1.1.3. Initiation and ignition. The processes of ignitionFollowing Essenhigh [68] for the case where there is noare defined differently by different authors and for differ-boundary layer resistance to mass transfer, we can writeent equipment. In DTG temperature ramping tests, initia-the rate in terms of the adsorption and desorption steps:tion is defined as the temperature at which the mass loss

1 1 1 trace under an oxidant flow first deviates from that under] ]] ]5 1 s an inert gas. Fig. 3 depicts such a test with the initiationk k P ka ox d

temperature marked T [64]. Lahaye et al. [15] use thein

If the desorption rate as given by k is much greater than term acceleration temperature (or onset of effective oxida-d

the adsorption rate term k P as was found by de Soete tion) to describe the intercept on the temperature axis ofa ox

[38], then k~P , i.e. the reaction is first order. In contrast, the linear part of the burning curve (weight loss v.ox

Neeft et al. [17] found orders of 0.76–0.94 at temperatures temperature). The term ignition in a DTG is generally usedaround 5008C, depending on burnout, and Petersen found to indicate a temperature runaway.0.5 at the same temperature [49]. Ciambelli et al. [69] Similarly, ignition in a fixed bed test signifies a thermalconsistently found n to be 0.8 for carbon black burned at runaway, with a sharp peak in temperature and rapid593 K. consumption of the bed [64]. Generally ignition occurs

The experimental order of reaction found by Du et al. locally and then spreads rapidly through the bed.[63] was 0.77 for CO release and 1.0 for CO release. The In an RVVR, ignition temperature is that at which the2

combination of these two reactions reactions gives an irradiated part of the bed sustains a rapid temperature riseoverall order of reaction around 0.83 in the temperature and begins to glow [47,48]. Herz and Sinkevitch [47] andrange 750–850 K. The CO/CO ratio was near unity, in McCabe and Sinkevitch [72] found that ignition was2

keeping with other results. indicated by a sharp, easily discernible spike in theThe empirical approach to reaction order represented by temperature. The progress of burning subsequent to igni-

the above expression has been criticised by Essenhigh tion was erratic, showing snake-like patterns which left[68], who insists that a theoretically-based expression such large areas unburned. The ignition temperature T wasi

as a Langmuir isotherm should be employed. He also sensitive to the oxygen content of the gas, but only slightly21produced the extended resistance equation or ERE to to its velocity, provided that it exceeded 1 cm s . Metals

describe the progress of solid particle combustion. From added to catalyse the combustion lowered T substantially.i


where T are their ignition temperatures. The experimen-i1,2

tally measured and calculated ignition temperatures in theDTG were found to be in satisfactory agreement.

5.1.1.4. Reaction kinetics. Low temperature activationenergy is found to cover a range, namely between 102 and

21210 kJ mol [15,17,19,38,73]. The low value of 102 kJ21mol was ascribed to the presence of a high concentration

of metals in the sample [19], which could have acted as21catalysts. Values between 140 and 170 kJ mol appear

21frequently. A value of 142 kJ mol was found by Otto etal. [29] with 10 mg of undiluted soot in a DTG, but inview of the effect of diffusion described by Marcucilli etal. [16], this figure is probably low.

De Soete [38] found that the oxidation rate of soot froman aromatic source was higher than from an aliphatic. Inboth cases there were changes in activation energy atmoderate temperatures, around 800 K (Fig. 4). Since theactivation energy increased at the higher temperatures, this

Fig. 3. DTG thermograms of diesel soot [64]: (—) under nitro- cannot be interpreted as a movement from regime I togen, (- - -) under 10% oxygen uncatalysed, (? ? ?) under 10% regime II combustion.oxygen catalysed with cerium. Fig. 4 plots reaction rates based on carbon consumption

from a number of investigators, normalised to 10 kPa21A mathematical model of the RVVR system was partial pressure of oxygen. The rate constant k (s ) based

developed by McCabe and Sinkevitch [72], based on a heatbalance over the soot layer. The input term is the heat ofreaction and the output is only the convective transport tothe gas. The bed soot is assumed isothermal and thesensible heat of the gas phase is assumed to be negligible,compared to the solid. The model gives an adequatedescription of the ignition phenomena. An extension ofthis model using measured kinetics for catalysed anduncatalysed soot successfully predicted the differences intheir ignition temperatures [64].

Lahaye et al. [15] found constant surface area onburnout, ascribed to layer combustion. Above 6008C theyalso found what was termed ‘autoignition’, where theinitial reaction rate showed an almost instantaneous stepbefore proceeding at a more controlled rate.

Application of the Semenov analysis to soot ignitionallows values of comparative ignition temperatures to bederived. Following Lahaye [15], for equilibrium at theignition point:

dT dQ dQi 1 2] ]] ]]mC 5 Q 2 Q 5 0 and .1 2dt dT dTi i

where m, C and T are the mass, specific heat and ignitioni

temperature of the solid, respectively, and Q are the1,2

heat production and loss rates. Under the same testconditions the loss by heat transfer will be of the sameform so that two different soots with different activationenergies E and pre-exponential factors A will be1,2 1,2

related by:

2E A T E T 1 E T1 1 i2 1 i2 2 i1 Fig. 4. Arrhenius plot of some kinetic rates of soot gasification]]] ]]]]exp . 1S D2 RT TE A T i1 i22 2 i1 using oxygen, steam and carbon dioxide as reactant.


on carbon removal is related to that based on oxygen 5.1.2. Analysis of burning data for spherules21consumption K (s ) by: At all size scales, it is necessary to take account of mass

transfer effects. This applies to individual spherules as wellbsb 21 as to assemblages of soot particles in beds.]K 5 k ss There is uncertainty as to the mode of burnout of theox

individual soot spherules. In view of the elongated struc-where s is the density of oxygen in the gas phase. Careox ture of aggregated soot particles, it is difficult to isolate theis required in the definition of reactivity in the Thiele behaviour of individual spherules. They consist of denselymodulus, which may be in terms of k as noted in Section packed carbon crystallites with a low porosity, which4.1. or in terms of K (Section 5.1.3.). implies that the diffusion rate of oxidant into the interior

There is good agreement between the different authors would be low. It could therefore be anticipated that oxygenwho used a number of different techniques, and with other penetration is restricted, and that regime II or III conditionsresults not shown [74]. The low temperature data of de would apply.Soete [38] deviate from the other sets. Note however that In order to calculate the Thiele modulus it is necessaryall these rates will be influenced by the combustion mode to know the pore dimensions. With the 1-nm poresand the subsequent formation of additional surface area. estimated earlier, full penetration of oxidant is predicted at

Also shown on Fig. 4 is the generalised correlation for all temperatures below 10008C. This comes about becausecarbon oxidation deduced by Smith [75] for a wide range of the short diffusion path (|10 nm), which more thanof materials and oxidation conditions, corrected to 10 kPa compensates for any diminution of the diffusion coefficient2 21oxygen and 120 m g surface area of carbon. It falls in in the narrow pores. Thus, regime I burning, rather thanthe middle of all the data, so that in the absence of other regime II or III is predicted.information, it can be used to estimate the combustion rate The participation of micropores (,2 nm) in gasificationof uncatalysed soot. reactions has been called into question on a number of

The semi-empirical relationship of Nagle and Stricklan- occasions, for example Gavalas [77], Radovic [78]. Aarnad-Constable [71] is most-commonly introduced to describe and Suuberg [79] present evidence that this is in fact thehigh temperature combustion. Two types of active site are case for a coal char. Since the pores in soot spherules areproposed: a more reactive type A site and a less reactive B. undoubtedly in the micropore range, there is justificationThermal rearrangement of type A sites into B takes place for suggesting that the spherules will burn in a regime IIIaccording to: mode.

Experimental resolution of this uncertainty has been21kT]]x 5 1 1 difficult. Ishiguro et al. [27] report that an early inves-S DP kO B2 tigation into the combustion of carbon black found that the

centres of the spherules were preferentially burned andwhere x is the fraction occupied by type A sites and KTbecame void. The only recent work to confirm this result isand k are rate constants. The rate equation is:Bby Gilot et al. [42], who studied Regal 600 carbon black at

k P 60% burnout. Lahaye et al. [15] report ‘channeling’, i.e.A O2 22 21]]]r 5 120 x 1 k P (1 2 x) kg m sFS D GB O2 internal slices of carbon were removed from spherules1 1 k PZ O2containing 3% cerium which were burned under kinetic

where k and k are further rate constants. At temperatures control.A Z

below 1700 K, the value of x is |1 and the activation Based on their experimental results on burning in air,21energy is effectively 143 kJ mol . Neeft et al. [17] and de Soete [38] conclude that a

The Nagle /Strickland-Constable approach was tested shrinking core situation applies. De Soete offers noextensively by Park and Appleton [52]. They found it to explanation for this conclusion. Neeft et al. base theiraccurately describe the high temperature behaviour, but to statement on the reaction order for carbon as given by thesomewhat underestimate the rates. Similar reservations relation for isothermal reaction:about the rates have been expressed by other authors

m 21 21rate 5 constant(1 2 u) k g g s[51,76].For high temperatures, an expression by Lee et al. [50]

21and a more recent one from Leung et al. [76] are also where u is the burnout and k is the reaction rate (s ). Theavailable. The correlations of Nagle and Strickland-Const- value of m varied between 0.65 and 0.8 for the carbonable and Leung et al. [76] for higher temperatures are also black Printex U, but was difficult to fit for diesel soot.

2 21shown on Fig. 4, assuming a surface area of 120 m g . Because the shrinking core model requires that m be 2/3,The results are not in good agreement and both are higher the conclusion was drawn that shrinking core behaviourthan the correlation of Smith [75]. They are however in the applies.vicinity of an extrapolation of some of the low temperature However, when one examines the rate versus time anddata. rate versus conversion (burnout) curves for Printex U


presented by Neeft et al. [80], the former is ‘exponential-like’ and the latter is almost linear from u50.2–0.9. Thus,the rate of reaction is almost directly proportional to themass of carbon remaining. This is consistent with exten-sive oxygen penetration into the spherule, leading todecreasing density, i.e. regime I conditions.

A good set of data on the change in soot propertiesduring reaction was collected by Ishiguro et al. [27] whooxidised samples of diesel soot at 5008C to 25, 50 and 75%burnout. They examined the products by electron micro-scope and reported the mean diameters of 2000 spherulesat each burnout. Their results are internally consistent andthe following relations for diameter and density in terms ofburnout can be fitted to their data with good accuracy:

d s1 / 6 1 / 2] ]5 (1 2 u) and 5 (1 2 u)d so o

From this data, the value of a in the ERE of Essenhigh is3, which is typical of combustion in regime II, close toregime I conditions. A value ,1 would be required for the Fig. 5. BET surface area as a function of burnout with oxygen atshrinking core situation to apply. The second effectiveness different temperatures [42].factor h as defined by Essenhigh [70] is the ratio of theburning rate to the external burning rate. Since h 5 1 1 3a,under the conditions used by Ishiguro et al. [27] the total The same explanation for differences in BET surfacerate of carbon removal is 10 times that on the exterior area can be made when different gasification media aresurface; an interior:exterior ratio of 9:1. used. At the same extent of conversion, there are greater

When applied to coal particles, the analysis of Es- increases in area with gasification by steam and carbonsenhigh [70] predicts that the initial particle density is an dioxide at 4008C than with oxygen [29]. These lessimportant variable in determining the value of a. Dense reactive oxidants should penetrate the spherules to a

23particles of initial density r |2000 kg m should burn greater extent. It should be noted that the effectivenesspo

in regime I (a 50) and those with low initial density factors calculated for all the conditions above were unity,should have a higher value of a (|3). The very small but significant differences in area result.diameter of the soot particles appears to invalidate this At 50% burnoff under air, the increase in area is 60%aspect of the analysis, as the Thiele modulus is always for Petersen (4808C) [49], 100% for Otto et al. (4508C)much less than unity. [29] and Marcucilli et al. (6008C) [16], but 600% for Gilot

The changes in surface area brought about by mass et al. (6008C) [42]. These results are consistent with theremoval during gasification have been extensively studied. behaviour of a highly purified carbon black (Graphon),Generally, there is a diminishing rate of increase as which showed a 40% increase in area after 35% burnoutcombustion proceeds, with the value continuing to rise up [81]. If the change in spherule diameter follows the patternto 50% burnout [16,27,29,49]. At higher burnoffs, the found by Ishiguro et al. [27], the increase in theoreticalvalue may remain roughly constant [27,29], or fall [49]. outside surface on a mass basis should be 12% at 50%

The change in area depends on the gasification medium burnoff, simply as a result of particle shrinkage. In contrastand the temperature. Gilot et al. [42] and Marcucilli et al. Ishiguro et al. [27] found an increase in surface area of

2[16] found that the increase in area was greater during around 400% at 50% burnoff, but the initial value of 52 m21 2 21burning in air at 600 and 7008C than at 8008C (Fig. 5). g is less than the theoretical external area of 87 m g

This is consistent with there being greater penetration of [25].oxygen (a higher effectiveness factor) at the lower tem- On balance, the experimental data suggest that signifi-perature due to slower consumption by reaction. By cant oxygen penetration occurs at low temperatures.extrapolation of these data, the original effective area can Turning to high temperatures, the increased reactivity

2 21 2be considered to be 160 m g , rather than the 100 m will decrease the extent of oxidant penetration. As reported21g which was measured (see Section 3.). The increase above, Neoh et al. [25] believe that internal burning

predicted by burning under shrinking core conditions is occurs, based mainly on theoretical considerations. For2 21also shown in Fig. 5, based on 160 m g initially. spherules, they predict an effectiveness of 0.74 for oxygen

Burning approaches shrinking core conditions (regime III) in 0.5-nm pores in 20-nm particles (T51775 K). Thisat 8008C. suggests extensive penetration of oxidant.


They found a difference in behaviour between rich and qualitative manner. They emphasise, like Gilot et al. [83],lean flames. There was a rapid increase in the number that sample size should be limited.concentration of particles in lean flames after about 80% The interaction between mass transfer and reaction in aburnout, which they attributed to fragmentation. It appears DTG test is well illustrated by a series of papers by Gilotthat the common material fusing the spherules into par- and co-workers [16,42–45]. Using samples of 5–40 mg, it

21ticles is removed, and the remains of the individual was found that the absolute rate of mass loss (g s ) inspherules are released. isothermal tests was almost constant in each for most of

In rich flames, the overall structure (chain) remains the burnout (Note the rapid nearly-linear fall in mass inintact as the particle concentration remains unchanged. In Fig. 3 for a temperature ramping situation). This suggestsboth fuel-rich and lean flames the particle sizes decreased that a thin reaction zone was always present at the bedwith burnout, rapidly in lean flames but only slightly in surface. The conclusion was supported by a unidimension-rich. The initial diameters were only about 80 nm, which al model of oxygen penetration into the bed, whichthe authors note is typical of the particle dimensions. demonstrated that diffusion was significant [43].

They explain the difference in terms of the reactivity of The effect of mass transfer in both the flow directionthe OH radical compared to molecular oxygen. OH is a and normal to this was first demonstrated by means of amore active oxidant than O , and should produce less finite difference model of the gas space above the crucible2

internal burning They estimated the effectiveness factors and the bed within it [16]. In a typical DTG situation,for pores of radius between 0.5 and 4 nm, the former to convective mass transfer coefficients to the surface of therepresent the pores inside a spherule, and the latter to bed were found from:represent inter-spherule gaps. They conclude that O will2

v 21penetrate the particles and disrupt them, while OH is ]]]h 5 (m s )m s y 2 ys dunable to do this. Since OH is the main oxidant in rich g o s

flames, fragmentation is not observed in that case.22 21There are some difficulties with this interpretation of the where v is the surface mass flux of oxygen (kg m s ),

23data. The chains are long and the concept of size (diam- s is the gas density (kg m ) and y and y are the bulkg o s

eter) is difficult to apply. The particles are small, isolated and bed surface mass fractions of oxygen. The values of hand of the same size as the mean free path of oxidant. The increase with temperature for the same flowrate, and alsoelectron micrograph included in their paper shows that with sample mass because the bed surface was closer totheir structure is virtually linear (Section 3.), so that it is the mouth of the crucible. The coefficients were of the

21difficult to imagine there being any diffusion restrictions. order of 0.5 cm s , which is capable of furnishing a24 22The assumptions of the Thiele approach as applied to a surface mass flux of oxygen of perhaps 2310 kg m

21particle do not appear to be relevant. s . Under the same conditions, molecular diffusion canThe differences in behaviour between lean and rich supply approximately the same amount, which suggests

flames may be the result of differences in reactivity that it is in fact the dominant mechanism.between OH and O , but this reactivity may relate to That diffusion is supplying most of the oxygen to the2

differences in carbon structure between the bulk material bed surface in a DTG test was confirmed by modelling theof the spherules and the interparticle deposits which bind space above the bed with computational fluid dynamicsthe spherules together. CFD [44,45]. The velocity vectors inside a crucible are

virtually zero, even if it is a shallow crucible and facing5.1.3. Analysis of burning data for soot in beds the gas flow. The oxygen mass transfer contours are

The fixed bed systems mentioned above can be expected uniform and parallel to the bed surface.to give reliable kinetic data, provided that the convective Within the bed, diffusion is also the mechanism ofgas flow is adequate to overcome mass transfer limitations, oxygen transport. Although the beds have high porosities,and the bed is diluted with inerts. However when one the small dimensions of the component solids introduceconsiders beds of soot contained in crucibles undergoing significant diffusion constraints. Knudsen diffusion isthermogravimetric analysis (DTG), mass transfer consid- therefore thought to operate. The Thiele modulus in theerations must be addressed. Flows of oxidant gas are bed is high, so that the depth of penetration is shallow.

21modest (perhaps 100 ml min ) and generally directed This leads to surface layer burning as mentioned above.across the mouth of the crucible. A unidimensional model of the gas concentration within

Neeft et al. [82] produced a model of heat and mass the bed was found to be adequate for the extraction oftransfer between a crucible of soot and the surroundings. kinetic data from isothermal DTG tests [64]. The surfaceThe mass transfer of oxygen is limited to that between the oxygen flux v is known from the experimental rate ofbulk gas and the sample surface. The oxygen concentration carbon loss, and the CFD model gives the mean surfaceinside the bed is assumed to be uniform, which has been mass fraction y . This can then used as a boundarys

shown to be false [35]. However it satisfactorily explains condition for the unidimensional model. The expressionthermal runaways during DTG analysis, but only in a for oxygen flux to the the bed surface in terms of


displacement from the surface x (m) in an isothermal bed bed of low density results in a rapid fall in coefficient fromof depth e (m) [43] is: the bulk value.

When estimating effective diffusion coefficients insidee

such beds, the usual random pore or ‘swiss cheese’22 21

v 5 Ks E y dx kg m sg approach [84] is unsuitable. The aggregated structure of0 soot makes it an unusual carbon material, unlike chars,

cokes and activated carbons, which can be regarded as21where K is the rate of consumption of oxygen (s ). Thesolids with holes. Better results are obtained by regardingoxygen concentration at position x within the bed is:it as a collection of solid spheres with extremely high

] porosity, i.e. as a ‘cannonball’ or ‘caviar’ solid.y Ks]]]] ] A simulation of the diffusion process based on they 5 exp(22u ) exp xF S DD1 1 exp(22u ) œ e caviar model was developed by Brilhac et al. [33]. A

] Monte Carlo random walk approach to the passage ofK]1 exp 2 xS DG oxygen molecules through an array of randomly spacedDœ e

spheres was used. This procedure tended to overestimatethe measured effective diffusion coefficient, but waswhere u 5 eœ(K /D ), the bed Thiele modulus. Then:e

improved by considering the bed to be locally non-v 1 2 exp(22u ) homogeneous (Fig. 6). A simpler kinetic theory model]]] ]]]]5 KD F Gœ e based on a mean free path and the mean kinetic velocitys y 1 1 exp(22u )g s

also produced acceptable results. In both cases, betterSince K appears in u, an iterative data fitting is required to estimates are produced if the local disposition of spherulesevaluate it from the surface oxygen flux and concentration. is taken into account, that is, that both the particles and theWith these techniques reliable kinetic data can be found, void space in between are physically simulated.provided that appropriate values of D can be supplied.e

Brilhac et al. [33] measured the effective diffusion 5.2. Catalytic combustioncoefficients of oxygen in soot beds of various porositiesusing a type of Wicke–Kallenbach diffusion cell. An The pursuit of a technique for diesel soot control has

26 2 21experimental coefficient of 2.0310 m s at e 50.82 prompted a plethora of research into the application ofand 298 K is typical. The effect of porosity on the effective catalysts to lower ignition temperatures and so promotediffusion coefficient is shown in Fig. 6. Even a very open combustion [19,39,48]. For example, the ignition tempera-

ture of a soot in a thermobalance fell by 300 K in one caseof catalyst addition [39].

Catalyst can be added either through the fuel, in whichcase it is incorporated into the soot spherules as they form,or physically mixed in after collection of the soot. Thelatter would correspond to the case of a soot filter whosesurface is impregnated with catalyst via a wash coat. Theauto industry favours the addition of catalyst to the fuel, asit ensures that the active components are intimately mixedwith the carbon. Small concentrations of around 50 ppmare sufficient. However, most of the combustion studiesreported in the literature use physically-mixed catalysis.

5.2.1. Catalyst-impregnated sootA multitude of studies consist of empirical tests on

operating engines fed with doped fuel [85–101] to list buta few. There are far fewer studies which focus exclusivelyon the oxidation of soot collected from flames or enginesrunning on doped fuel [15,64,102,103]. The catalystsadded to fuel take the form of soluble compounds (oc-tanoates, naphthanates) of metals such as copper, iron,cerium, lead, manganese etc.

A comparison of the effect on reactivity of iron andFig. 6. The influence of bed porosity on the effective diffusioncopper additions to the fuel showed that copper produced acoefficient of oxygen in beds of carbon black — measured andmodest increase by a factor of around 5 at 6008C, and ironpredicted [33]: (- ♦ -) experimental; (- - -) Monte Carlo; (—)

kinetic theory. by a factor of 3 [102]. Although copper has been found to


be very efficient by all investigators, it is environmentally chanically milled for an hour. In both cases the samplesunacceptable [104]. Cerium has now emerged as the diluted with SiC were placed into a crucible and oxidationpreferred fuel additive of automotive constructors and a tests were carried out in a DTG/DSC. A wide range ofsystem using it has been commercialised [105]. metals was tested for catalytic properties [108,109], using

21Lahaye et al. [15] produced soot by the pyrolysis of a heating rate of 10 K min in 21% oxygen.hydrocarbons, with and without cerium addition. The As an example, the differences between the oxidationcerium is present as oxide [15]. Increasing cerium con- behaviours for the two contact systems using Fe O and2 3

centration diminishes the mean particle size of the soot MoO as catalysts are shown in Fig. 7. With tight contact,3

during formation and lowers the ignition temperature under there is a considerable fall in the temperatures at which theoxidation. Above 6008C very rapid oxidation (called reaction proceeded for both metals. However, MoO is3

autoignition by the authors) occurs. At low temperatures active in the loose condition but Fe O is not. The2 3

(,5208C), CO is the only product but above this the ratio activities of the catalysts were then assessed in terms of the2

CO/CO remains at about 0.5. The activation energy of ‘combustion’ temperature, i.e. the maximum temperature221combustion fell from 170 to 120 kJ mol with the recorded in the DSC trace.

addition of cerium. The action of catalysts can be determined from the formAfter running scores of tests on soots containing varying of the burnout trace [80]. Extremely active catalysts

amounts of cerium oxide as obtained from engines oper- produce an initial increase in reaction rate with burnout upated under three different conditions, Stanmore et al. [64] to u.0.5, followed by a decline. Less active catalysts

2found a significant (¯10 ) increase in reactivity with show a declining or constant rate burnout. Neeft et al.cerium present, but no change in activation energy from [110] showed that if the monolithic trap in a diesel exhaust

21210 kJ mol (Fig. 4). Note the change in activation is coated with catalyst, the oxidation behaviour is similarenergy in their data above about 900 K, where reactivity to the loose contact condition. The full potential of ano longer increases with temperature. This is similar to the catalyst can be achieved only by the addition of a solubleauto-acceleration phenomenon described by Lahaye et al.[15], which has not been explained. Petersen [49] alsofound no change in activation energy with the addition oflead.

Some papers report only on the oxidation of SOF, whichis composed mostly of lubricating oil [106,107]. In thesetests the catalyst was impregnated with oil.

5.2.2. Physically-mixed catalystTurning to externally added catalyst, Murphy et al. [46]

added metal chlorides dissolved in methanol to diesel sootand determined ignition temperatures. Copper was foundto give superior performance compared to Mn and Co,lowering T by about 100 K.i

Much attention has been focussed on the quality ofcontact between the carbonaceous material and the grainsof a solid catalyst. This concept is developed particularlyby Neeft and co-workers [80,108–110] and Ciambelli andco-workers [69,111–113]. If the catalyst grains are not inintimate contact with the carbon surface, their effect willbe diminished. For modelling purposes, Ciambelli et al.[69] postulate two reactivity relationships, one forspherules within the field of action of the catalyst, andanother for those initially outside.

Neeft et al. [80] developed two methods of mixingcatalyst with soot, which they term ‘loose’ and ‘tight’. Themethods produced reproducible combustion behaviour, andpermitted a more reliable classification of catalyst activity.The catalysts were prepared as particles of less than 125mm size. In the loose mode, a 2:1 ratio of catalyst tocarbon black was simply mixed together with a spatula. In Fig. 7. Burning rates in a DTG/DSC of carbon black mixed withthe tight mode, a mixture at the same ratio was me- Fe O and MoO catalysts: loose and tight contact [109].2 3 3


compound to the diesel fuel to ensure that it is incorpo- complexes with relatively low activation energies fordesorption.rated into the soot spherules.

The catalyst generally lowers activation energy, typical-Of the straight metal oxides, PbO is the most active21ly by 30–80 kJ mol , that is from 170 to 120 [38] and[80,114], but CuO and MnO are also efficient. Courcot et2

2121 151 to 100 kJ mol [69]. Since the uncatalysed activational. [20] concluded that Cu ions are the active sites withenergies are generally higher, the rates will overtake thecopper reagents. Copper is more active in the presence ofcatalysed rate as temperature rises. Above the crossoverchloride than other anions and Mul et al. [115] identifytemperature, the rates will be identical.Cu OCl as an oxidation intermediate.2 2

The order of the oxidation reaction with CuO catalyst isThe oxides of metals which readily change valency havereported by de Soete [38] as unity. The measurement of thebeen identified by Setzer et al. [116] as superior catalysts.rates of the reaction stages indicates that catalytic combus-Thus Cu, Fe and Co are far more active than Ni and Zn.tion, like the uncatalysed system, is controlled by theNeeft et al. [109] found that the oxides Co O , V O ,3 4 2 5adsorption of the gaseous oxidiser. This should be con-Fe O La O , MnO and NiO exhibit high to moderate2 3 2 3 2trasted with Ciambelli et al. [112] who identify desorptionactivities in tight contact but hardly any activity in looseas the controlling step. However, higher desorption tem-contact. Other oxides such as PbO, Cr O , MoO , CuO2 3 3 peratures (up to 10008C) were studied [69], whereas deand Ag O are still active in loose mode, although at a2 Soete’s data refer to lower temperatures (Fig. 2).reduced level. Neri et al. [67] found that Fe O was a good2 3 Neri et al. [67] discuss oxygen spillover for catalystscatalyst for SOF oxidation but that CuO and V O were2 5 containing platinum group metals. They found a goodbetter for the carbon component.correlation between catalyst reactivity and the enthalpy of¨The catalysts tested by Ahlstrom et al. [117,118] wereformation of the metal oxides. They proposes that lattice

loaded onto g-alumina which was used at a 10:1 massoxygen acts in a redox manner, consistent with the findings

dilution, while the reference (uncatalysed) tests wereof McKee for graphite [121]. This implies an oxygen

carried out with pure alumina. The arrangement must be‘spillover’ effect from the noble metal promoter, and

considered as ‘loose’. As a result, the gains in reactivity requires good contact between the carbon and catalyst.were modest, with V O being best. It also resisted2 5 During catalyst studies with Cu–V–K, Ciambelli et al. [39]poisoning by the sulphur present in soot. report that significant oxidation of carbon black took place

Ciambelli et al. [111] produced a version of tight contact even in the absence of oxygen, indicating that a redoxby pounding the components together with a mortar and mechanism was active.pestle. Like Neeft et al. [80], they found that the reaction Ciambelli et al. [69] found that during the combustion ofrates were strongly influenced by the mass ratio of soot to uncatalysed soot the order was 0.8 at 593 K, but 0.5 whencatalyst R , finding significant falls in burnoff tempera-m catalysed with Cu–V–K. This is explained by proposing atures as R decreased.m redox or oxygen ‘shuttle’ mechanism. The reaction mecha-

Badini and co-workers [119,120] used a ratio of 1:1 soot nism consists of dissociative adsorption of oxygen onto theto catalyst, which is probably too low to exercise good catalyst surface (oxidative step), followed by carboncontrol of the thermal conditions. oxidation by the oxygen of the catalyst (reductive step). If

When a mixture of soot and one of the following metal the oxidative step is in equilibrium and the reductive stepoxides is heated in the absence of gaseous oxidisers, CO is rate-limiting, the order should be 0.5.and CO are found as reaction products [38]. Good For loose contact with metal oxides, a correlation2

catalysts (CuO, MnO ) produce CO exclusively, while between melting temperature (and hence partial pressure)2 2

poor catalysts (CeO ) produce both, and inactive oxides and catalyst efficiency was noted by Neeft [109], Ahlstrom2

(Al O , SiO , La O ) produce CO only in negligible [117], Neri [67] and McKee [121]. Migration across the2 3 2 2 3

quantities. This suggests an oxygen donor mechanism, carbon surface is considered to be the mechanism in-rather than a mere electronic mechanism. The catalytic volved. Mul et al. [115] note that CuCl is able to ‘wet’2

effect is due to the fact that carbon adsorbs metal-bound alumina and speculates that it will spread across carbon asoxygen atoms (–MO) faster than molecular-bound oxygen. well.

Catalysed reaction mechanisms are evidently different Metals such as copper can be promoted by the additionfrom the uncatalysed condition — as the order of reaction, of a promoter such as potassium [69], molybdenum [115],the activation energy and the primary products all change. niobium [122], iron as spinel [123], vanadium [67,119] orDu et al. [62] found that the activation energy fell from chlorine [104].

21145 to 120 kJ mol with addition of 2% calcium as McKee addressed the mechanism of graphite oxidationcatalyst. At the same time the ratio of CO to CO fell from when catalysed by alkali metals [121,124]. The most2

unity to 0.01 at 663 K. Ciambelli et al. [112] also found effective species were the carbonates, oxides and hy-that the action of a potassium–copper–vanadium catalyst droxides; other active salts tend to convert to these underpromoted CO formation (CO/CO |0.05). They propose gasification conditions. There is a threshold concentration2 2

that the catalyst promotes the formation of surface oxygen of alkali carbonate catalyst below which it exhibits no


activity [19]. Neeft et al. [80] found increases in reactivity Otto and Shelef [127] showed that metals catalyse theof 3–5 orders of magnitude with alkali metals. The carbon–steam reaction, using graphite and metals depos-effectiveness is inversely dependent on the atomic mass of ited from aqueous solutions. The action of water vapour onthe alkali, being highest for caesium and lowest for graphite when catalysed by alkaline earth metals is re-lithium. The increases are attributed to the redistribution of ported by McKee [128]. Ba and Sr were more active thancatalyst. Ca and Mg. A carbonate–oxide oxidation–reduction cycle

The mechanisms of attack of catalyst particles on is proposed.carbon, such as pitting of the graphite basal plane, edge Kapteijn et al. [129] investigated the reactivity of carbonrecession and channeling are demonstrated in Marsh and dioxide with graphite using alkali metal carbonates asKuo [58]. catalysts. The results were similar to those cited above for

oxygen gasification with soot, namely Cs.Rb.K.Na.

Li. They identify five distinguishable temperature regionsduring outgassing under TPD. As concluded earlier, two6. Gasification with steam, carbon dioxide or nitrogentypes of surface oxide species of differing stability aredioxidepresent.

The reactivity of nitrogen dioxide NO towards soot is2Gasification of soots produced from aromatic (a-MN)far greater than that of oxygen at low temperatures. This isand aliphatic (n-HD) hydrocarbons with water vapour anddemonstrated by Fig. 8 [130], which depicts the specificcarbon dioxide was studied by de Soete [38] using therates of oxidation by each at 5008C, at the same con-same techniques as for oxygen. There was negligiblecentration (0.1%). Water acts as a catalyst for the C–NO2reaction with H O below 800 K and the major product at2reaction, and oxygen is also active in promoting oxidationhigher temperatures was CO (CO/CO ratio |9). The2at temperatures below its usual range of activity. A kineticorder of reaction with respect to H O was close to one.2expression involving temperature and the concentrations ofThe gasification rates with carbon dioxide were lower thanNO , O and H O has been derived [131].2 2 2those with water vapour. The reaction order for CO was2

Early interest in the C–NO reaction focussed on2again one, but the quantities of product CO and CO were2atmospheric reduction of NO by carbonaceous particles in2generally of the same order.the air, see Ref. [132] for examples. The temperaturesOtto et al. [29] report on the increase in soot surfaceemployed in these investigations were mostly ambient butareas during gasification with H O and CO , but give no2 2ranged to 2008C. Chemisorption takes place with surfaceother information. Typical Arrhenius data for the gasifica-species like C–NO , C–ONO and C–N–NO being2 2tion of n-HD soot with H O and CO [38] are shown for2 2formed. Chugtai and co-workers [132,133] found that atcomparison on Fig. 4. The data are similar to thosetemperatures below 2008C NO reacts with n-hexane soot2reported for petroleum coke reacted with steam by Harristo form NO, which is then disproportionates to N O and2and Smith [125]. In general, the trends in the reaction ofNO . At the same time small amounts of CO and CO are2 2soot with H O and CO are similar to those with oxygen,2 2released. A dual-site mechanism was proposed for thebut at a reduced rate of reaction.reaction of NO .2A study with isotopic CO by Kapteijn et al. [126]2

For automotive applications, the continuously regenerat-confirmed the presence of two types of desorbable oxygening trap (CRT) [134] is based on the oxidation by NO of2complexes on the surface of activated carbon. The morethe soot collected in a particulate filter in the temperaturereactive species is present in smaller quantities but contri-

butes more to the CO desorption. The desorption transientsare exponential in form and exhibit a range of activationenergies.

Moulijn and Kapteijn [59] propose a unified theory,involving adjacent active sites. The reaction sequencesuggested for CO oxidation involves semi-quinone C(O)2

and carbonyl CC(O) species, the former involving slowreactions and the latter more active:

CO 1 C → CO 1 C(O)2 f

CO 1 CC(O) → CO 1 C(O)(CO)2

C(CO) → CO 1 Cf

C(O) → CO 1 Cf

They imply that other oxidants will follow similar paths, Fig. 8. Comparison of the specific oxidation rates of carbon blackboth involving slow and fast reaction steps. Additional in fixed beds at 5008C for (—) 0.1% O and (- - -) 0.1% NO2 2

steps are involved with different oxidants. [130].


range 200–5008C. In this system NO is produced by the 7. Conclusion2

oxidation of the NO in the engine exhaust by a catalyst inan upstream monolithic converter. Even though the rate of Soots from various sources are structurally similar, andcarbon consumption by NO is small, it is enough to similar to commercial carbon blacks. The fundamental2

ensure an equilibrium in the particulate filter between the spherules are 20–30 nm in diameter, with little porosity.mass of soot produced by the engine and that consumed by The four reactant gases O , CO , H O and NO all oxidise2 2 2 2

the reaction. and gasify soot, with NO the most reactive at low2

The range of temperatures used by Lur’e and Mikhno temperatures. Two kinds of reactive site where adsorption[135] to investigate the gasification kinetics of graphitized intermediates are formed can be identified for the O and2

soot by NO was 100–3508C. The equipment employed NO reactions. When the spherules are gasified, the extent2 2

was a fixed bed with a recirculating flow of reactant. The of reactant penetration depends on the reactant and thedominant reaction was: temperature. In the absence of a flow-through gas supply,

oxidation in beds depends on the supply of oxidant byC 1 2NO → CO 1 2NO diffusion to the interior. The reactivity towards oxygen is2 2

similar to that of other carbons on an area basis.but the presence of condensed surface products compli- With catalyst present, the gasification mechanisms arecates the system. Small amounts of CO and N are formed modified, so that the reaction products, the activation2

at higher temperatures. The order of reaction changes energy and the apparent reaction order change. Two formsdrastically with temperature, being unity above 3008C and of catalyst addition have been identified, depending on therising to more than 2 as the temperature falls. A mecha- intimacy of contact. ‘Tight’ contact results in enhancementnism for the reaction based on the formation and sub- of the reaction rate for many metals, which may be lesssequent decomposition of an intermediate species was effective in ‘loose’ contact. Cerium and the noble metalsproposed. The activation energy was found to be 50 kJ are used for the control of diesel soot emissions. Alkali and

21mol in the temperature range 180–3508C. alkaline earth metals are especially effective with H O and2

Gray and Do [136,137] investigated gravimetrically the CO .2

adsorption and reaction of NO on an activated carbon The control of soot emissions from automotive engines2

(Norit RB) over a similar temperature range (150–3508C). can now be achieved by installing filter monoliths in theAt low temperatures (,1008C) the NO is simply ad- exhaust line. These traps must be periodically regenerated2

sorbed but it undergoes reaction at higher temperatures. by burning the deposited soot. This filtration device can beThe surface reaction controls the overall kinetics, which associated with the use of fuel additives to decrease theare first order with respect to NO at 3008C. When the ignition temperature of the soot material by a catalytic2

random pore model was used to describe the evolution of effect.surface area, the activation energy was found to be 86.2 kJ Another technology, based on the continuous oxidation

21mol . This is lower than those for O , H O and CO of soot by NO , is also a possible solution to meet the2 2 2 2

gasification. requirements regarding soot emissions by diesel engines.The extensive work of Tabor et al. [138] concerns the In that case NO must be produced by a catalytic oxidation2

chemistry of the low temperature (ambient) interaction of of NO before its reaction with the soot material trapped inNO with three kinds of carbon, including carbon black. a monolithic filter.2

The major gaseous product in each case was NO, but the These technologies seem to be able to solve the problemcarbon product remained on the surface of the solid. At a of the soot emissions from automobile engines.temperature of 1508C, Yagud et al. [139] found that anactivated charcoal produced CO and NO according to the2

above equation. ReferencesThe combustion behavior of the soot in a CRT system

was investigated by Cooper and Thoss [140] on a test [1] McClure BT, Bagley ST, Gratz LD. The influence of anoxidation catalytic converter and fuel composition on thebench. The orders of reaction with respect to the NO inlet2chemical and biological characteristics of diesel exhaustconcentration and to soot weight were found equal to 1.emissions, 1992, SAE paper 920854.The benefit of the presence of water in the feed gas for

[2] Christensen R, Hansen MB, Schramm J, Binderup M-L,soot oxidation was noted.Jorgensen V. Mutagenic activity of the soluble organicEnhanced activity of nitric oxide NO in the presence offraction of exhaust gas particulate from a direct injection

water vapour as an oxidant for soot over Ce/Pt, Pt /SiO2 diesel engine, 1996, SAE paper 961977.and Cu/V/K/Cl /Ti catalysts has been demonstrated [141– [3] Muhle H. Toxic and carcinogenic effects of fine particles —143]. In each case the enhancement has been explained by observations and hypotheses. In: International ETH work-the catalysed formation of NO from the NO. The presence shop on nanoparticle measurement, Zurich: ETH, 1999,2

of SO was found to further increase the rate [132,142]. Paper 2.2


[4] Haynes BS, Wagner HG. Soot formation. Prog Energy [25] Neoh KG, Howard JB, Sarofim AF. Effect of oxidation onCombust Sci 1981;7:229–73. the physical structure of soot. In: Twentieth international

[5] Lahaye J, Prado G. Morphology and internal structure of symposium on combustion, Pittsburgh: The Combustionsoot and carbon blacks. In: Siegla DC, Smith GW, editors, Institute, 1984, pp. 951–7.Particulate carbon formation during combustion, New York: [26] Darmstadt H, Roy C, Donnelly PJ. Comparative inves-Plenum Press, 1981. tigation of defects on carbon black surfaces by nitrogen

[6] Smith OI. Fundamentals of soot formation in flames with adsorption and SIMS. In: Third international conference onapplication to diesel engine emissions. Prog Energy Combust carbon black (Mulhouse), 2000, pp. 77–9.Sci 1981;7:275–91. [27] Ishiguro T, Suzuki N, Fujitani Y, Morimoto H. Micro-

[7] Donnet J-B. Fifty years of research and progress on carbon structural changes of diesel soot during oxidation. Combustblack. In: Second international conference on carbon black Flame 1991;85:1–6.(Mulhouse), 1993, pp. 1–10. ¨[28] Goritz D, Weigert A, Raab H. Investigation of the localisa-

[8] Bockhorn H, editor, Soot formation in combustion: mecha- tion of active sites on the surface of carbon black bynisms and models, Berlin: Springer, 1994. scanning tunneling microscopy. In: Third international con-

[9] Murphy MJ, Hillenbrand LJ, Trayser DA, Wasser JH. ference on carbon black (Mulhouse), 2000, pp. 43–50.Assessment of diesel particulate control — direct and [29] Otto K, Sieg MH, Zinbo M, Bartosiewicz L. The oxidationcatalytic oxidation, 1981, SAE paper 810112. of soot deposits from diesel engines, 1981, SAE paper

[10] Glassman I, Nishida O, Sidebotham G. Critical temperatures 800336.of soot formation. In: Bockhorn H, editor, Soot formation in ´[30] Yuan S, Meriaudeau P, Perrichon V. Catalytic combustion ofcombustion: mechanisms and models, Berlin: Springer, diesel soot particles on copper catalysts supported on TiO .21994, pp. 316–24. Effect of potassium promoter on the activity. Appl Catal B

[11] Hall-Roberts VJ, Hayhurst AN, Knight DE, Taylor SG. The 1994;3:319–33.origin of soot in flames: Is the nucleus an ion? Combust [31] Stoeckli F, Guillot A, Slasli A, Hugi-Cleary D. Microporosi-Flame 2000;120:578. ty in carbon blacks. In: Third international conference on

[12] Frenklach M. Surface growth mechanism of soot particles. carbon black (Mulhouse), 2000, pp. 67–75.In: Twenty-sixth international symposium on combustion, [32] Buczek B, Swiatkowski A, Zietek P, Trznadel BJ. Ad-Philadelphia: The Combustion Institute, 1996, pp. 2285–93. sorption properties and porous structure within granules of

[13] Krestinin AV. Detailed modeling of soot formation in hydro- activated carbons with different burnoff. Fuelcarbon pyrolysis. Combust Flame 2000;121:513–24. 2000;79(10):1247–53.

[14] Veranth JM, Fletcher TH, Pershing DW, Sarofim AF. Mea- [33] Brilhac J-F, Bensouda F, Gilot P, Brillard A, Stanmore B.surement of soot and char in pulverised coal fly ash. Fuel Experimental and theoretical study of oxygen diffusion2000;79:1067–75. within packed beds of carbon black. Carbon 2000;38:1011–

[15] Lahaye J, Boehm P, Chambrion P, Ehrburger P. Influence of 9.cerium oxide on the formation and oxidation of soot. [34] Lahaye J, Ehrburger-Dolle F. Mechanisms of carbon blackCombust Flame 1996;104:199–207. formation. Correlation with morphology of aggregates. In:

[16] Marcucilli F, Gilot P, Stanmore BR, Prado G. Experimental Second international conference on carbon blackand theoretical study of diesel soot reactivity. In: Twenty- (Mulhouse), 1993, pp. 11–22.fifth international symposium on combustion, Philadelphia: [35] Noirot R, Gilot P, Gadiou R, Prado G. Control of sootThe Combustion Institute, 1994, pp. 619–26. emission by filtration and post-combustion. A laboratory

[17] Neeft JPA, Nijhuis TX, Smakman E, Makkee M, Moulijn study of the regeneration of an organic particulate trapJA. Kinetics of the oxidation of diesel soot. Fuel assisted by hydrocarbon injection. Combust Sci Technol1997;76(12):1129–36. 1994;95:139–60.

[18] Walker PL, Austin LG, Nandi SP. In: Walker PL, editor, [36] Smith IW. The combustion rate of coal chars: a review. In:Chemistry and physics of carbon, vols. 1 and 2, New York: Nineteenth international symposium on combustion, Phila-Marcel Dekker, 1966. delphia: The Combustion Institute, 1982, pp. 1045–65.

¨[19] Ahlstrom AF, Odenbrand CUI. Combustion characteristics of [37] Walker PL, Rusinko Jr F, Austin LG. Gas reactions ofsoot deposits from diesel engines. Carbon 1989;27:475–83. carbon. Adv Catal 1959;11:133.

¨[20] Courcot D, Abi-Aad E, Capelle S, Aboukaıs A. Preparation [38] de Soete G. Catalysis of soot combustion by metal oxides.and characterisation of copper–cerium oxide catalysts and In: Western States section meeting, Salt Lake City, 21–22their application to the combustion of diesel soot. In: CaPOC March, The Combustion Institute, 1988.4, Brussels, 1997, pp. 303–12. [39] Ciambelli P, Corbo P, Parrella P, Scialo M, Vaccaro S.

[21] Neeft JPA, Makkee M, Moulijn JA. Diesel particulate Catalytic oxidation of soot from diesel exhaust gases. 1.emission control. Fuel Proc Technol 1996;47:1–69. Screening of metal oxide catalysts by TG-DTG-DTA analy-

[22] Zinbo M, Skewes LM, Hunter CE, Schuetzle D. Thermo- sis. Thermochim Acta 1990;162:83–9.gravimetry of filter-borne diesel particulates. Thermochim [40] Ma J, Fang M, Li P, Zhu B, Lu X, Lau NT. Microwave-Acta 1990;166:267–75. assisted catalytic combustion of diesel soot. Appl Catal A

[23] Heywood JB. Internal combustion engine fundamentals, New 1997;159:211–28.York: McGraw Hill, 1990. [41] Stanmore BR. Use of the differential thermal gravimetric

[24] Farrauto RJ, Voss KE. Monolithic diesel oxidation catalysts. analyser for carbon/oxygen reactivity measurement. FuelAppl Catal B 1996;10:29–51. 1991;70:1485–7.


[42] Gilot P, Bonnefoy F, Marcucilli F, Prado G. Determination temperature-programmed desorption: Modeling and applica-of kinetic data for soot oxidation. Modeling of competition tion to the soot-oxygen system. Energy Fuels 1990;4:296–between oxygen diffusion and reaction during thermogravi- 302.metric analysis. Combust Flame 1993;95:87–100. [63] Du Z, Sarofim AF, Longwell JP, Mims CA. Kinetic measure-

[43] Stanmore B, Gilot P, Prado G. The influence of mass transfer ment and modelling of carbon oxidation. Energy Fuelsin DTG combustion tests. Thermochim Acta 1994;240:79– 1991;5:214–21.89. [64] Stanmore B, Brilhac J-F, Gilot P. The ignition and combus-

[44] Stanmore BR, Gilot P. The influence of sample containment tion of cerium-doped diesel soot, 1999, SAE paper 1999-01-on the thermogravimetric measurement of carbon black 0115.reactivity. Thermochim Acta 1995;261:151–64. ´[65] Querini CA, Ulla MA, Requejo F, Soria J, Sedran UA, Miro

[45] Gilot P, Brillard A, Stanmore BR. Geometric effects on massEE. Catalytic combustion of diesel soot particles. Activity

transfer during thermogravimetric analysis: application to theand characterisation of Co/MgO and Co, K/MgO catalysts.reactivity of diesel soot. Combust Flame 1995;102:471–80.Appl Catal B 1998;15:5–19.[46] Murphy MJ, Hillenbrand LJ, Trayser DA, Wasser JH.

[66] Ma MC, Haynes BS. Surface heterogeneity in the formationAssessment of diesel particulate control — direct andand decomposition of carbon surface sites. In: Twenty-sixthcatalytic oxidation, 1981, SAE paper 810112.international symposium on combustion, Philadelphia: The[47] Herz RK, Sinkevitch RM. Reactors for investigation of sootCombustion Institute, 1996, pp. 3119–25.combustion on filter surfaces. Carbon 1986;4:457–62.

[67] Neri G, Bonaccorsi L, Donato A, Milone C, Musolino MG,´[48] Vonarb R. PhD thesis, Universite de Haute Alsace, 1999.Visco AM. Catalytic combustion of diesel soot over metal[49] Petersen RC. The oxidation rate of diesel particles whichoxide catalysts. Appl Catal B 1997;11:217–31.contain lead, 1987, SAE paper 870628.

[68] Essenhigh RH. An integration path for the carbon–oxygen´[50] Lee KB, Thring MW, Beer JM. On the rate of combustion ofreaction with internal reaction. In: Twenty-second interna-soot in a laminar soot flame. Combust Flame 1962;6:137–45.tional symposium on combustion, Pittsburgh: The Combus-[51] Garo A, Prado G, Lahaye J. Chemical aspects of soottion Institute, 1988, pp. 89–96.particles radiation in a laminar methane–air diffusion flame.

Combust Flame 1990;79:226–33. [69] Ciambelli P, d’Amore M, Palma V, Vaccaro S. Catalytic[52] Park C, Appleton JP. Shock-tube measurements of soot combustion of an amorphous carbon black. Combust Flame

oxidation rates. Combust Flame 1973;20:369–79. 1994;99:413–21.[53] Roth P, Brandt O, von Gersum S. High temperature oxida- [70] Essenhigh RH. Influence of the initial particle density on the

tion of suspended soot particles verified by CO and CO reaction mode of porous carbon particles. Combust Flame2

measurements. In: Twenty-third international symposium on 1994;99:269–79.combustion, Pittsburgh: The Combustion Institute, 1990, pp. [71] Nagle J, Strickland-Constable RF, editors, Fifth conference1485–91. on carbon, London: Pergamon Press, 1962, p. 154.

[54] Cavaliere A, Barbella R, Ciajolo A, D’Anna A, Ragucci R. [72] McCabe RW, Sinkevitch RM. A laboratory combustion studyFuel and soot oxidation in diesel-like conditions. In: Twenty- of diesel particulates containing metal additives, 1986, SAEfifth international symposium on combustion, Philadelphia: technical paper no. 860011.The Combustion Institute, 1994, pp. 167–74. [73] Miyamoto N, Hou Z, Ogawa H. SAE paper 881224, 1988.

[55] Roth P, Eckhardt T, Franz B, Patschull J. H O -assisted [74] Ciambelli P, Palma V, Russo P, Vaccaro S. Catalytic ceramic2 2

regeneration of diesel particulate traps at typical exhaust gas filter for diesel soot removal: preliminary investigations. In:temperatures. Combust Flame 1998;115:28–37. CaPOC 4, Brussels, 1997, pp. 313–21.

[56] Ismail IMK, Walker Jr. PL. Detection of low temperature [75] Smith IW. The intrinsic reactivity of carbons to oxygen. Fuelcarbon gasification using DSC and TGA. Carbon 1978;57:409.1989;27:549–59. [76] Leung KM, Lindsedt RP, Jones WP. A simplified reaction

[57] Lear AE, Brown TC, Haynes BS. Formation of metastable mechanism for soot formation in non-premixed flames.oxide complexes during the oxidation of carbons at low Combust Flame 1991;87:289–305.temperatures. In: Twenty-third international symposium on [77] Gavalas GR. A random pore model with application to charcombustion, Pittsburgh: The Combustion Institute, 1990, pp. gasification at chemically controlled rates. Am Inst Chem1191–7. Eng J 1980;30:967–73.

[58] Marsh H, Kuo K. Kinetics and catalysis of carbon gasifica- [78] Radovic LR, Walker Jr PL, Jenkins RG. Importance oftion. In: Marsh H, editor, Introduction to carbon science, carbon active sites in the gasification of coal chars. FuelLondon: Butterworths, 1989, p. 107. 1983;62:849–56.

[59] Moulijn JA, Kapteijn F. Towards a unified theory of [79] Aarna I, Suuberg EM. Changes in reactive surface area andreactions of carbon with oxygen-containing molecules. Car- porosity during char oxidation. In: Twenty-seventh interna-bon 1995;33:1155–65. tional symposium on combustion, Philadelphia: The Com-

´[60] Kapteijn F, Meijer R, Moulijn JA, Cazorla-Amoros D. On bustion Institute, 1998, pp. 2933–9.why do different carbons show different gasification rates: a [80] Neeft JPA, Makkee M, Moulijn JA. Catalytic oxidation oftransient isotopic CO gasification study. Carbon carbon black. I. Activity of catalysts and classification of2

1994;32:1223–31. oxidation profiles. Fuel 1998;77(3):111–9.[61] Ahmed S, Back MH, Roscoe JM. A kinetic model for the [81] Walker PL, Taylor RL, Ranish JM. An update on the

low temperature oxidation of carbon. Combust Flame carbon–oxygen reaction. Carbon 1991;29:411–21.1987;70:1–16. [82] Neeft JPA, Hoonaert F, Makkee M, Moulijn JA. The effects

[62] Du Z, Sarofim AF, Longwell JP. Activation energy in of heat and mass transfer in thermogravimetrical analysis. A


case study towards the catalytic oxidation of soot. Ther- at low temperatures in conjunction with a cerium basedmochim Acta 1996;287:261–78. additive, 1998, SAE paper 980543.

[83] Gilot P, Marcucilli F, Prado G. Porous media and the use of [102] Bonnefoy F, Gilot P, Stanmore BR, Prado G. A compara-thermobalances: the kinetics of combustion processes. In: tive study of carbon black and diesel soot reactivity in theNickel KG, editor, Corrosion of advanced ceramics, Am- temperature range 500–6008C; effect of additives. Carbonsterdam: Kluwer, 1994, p. 329. 1994;32:1333–40.

[84] Bhatia SK, Perlmutter DD. A random pore model for fluid– [103] Lahaye J, Boehm S, Ehrburger P. Metallic additives in sootsolid reactions. I. Isothermal, kinetic control. Am Inst Chem formation and post-combustion. In: Bockhorn H, editor,Eng J 1980;26:379–86. Soot formation in combustion: mechanisms and models,

[85] Simon GM, Stark TL. Diesel particulate trap regeneration Berlin: Springer, 1994, pp. 307–15.using ceramic wall-flow traps, fuel additives, and supple- [104] Mul G, Neeft JPA, Makkee Kapteijn F, Moulijn JA.mental electric lighters, 1985, SAE paper 850016. Catalytic oxidation of model soot by chlorine based cata-

[86] Tan JC, Opris CN, Baumgard KJ, Johnson JH. A study of lyst. In: CaPOC 4, Brussels, 1997, pp. 323–32.the regeneration process in diesel particulate traps using a [105] Solvat O, Marez P, Belot G. Passenger car serial applicationcopper fuel additive, 1996, SAE paper 960136. of a particulate filter system on a common rail direct

¨[87] Lepperhof G, Luders H, Barthe P, Lemaire J. Quasi-continu- injection diesel engine, 2000, SAE paper 2000-01-0473.ous particle trap regeneration by cerium additives, 1995, [106] Voss KE, Lampert JK, Farrauto RJ, Rice GW, Punke A,SAE paper 950369. Krohn R. Catalytic oxidation of diesel particulates with

[88] Lepperhof G. Soot formation and oxidation in IDI and DI base metal oxides. In: CaPOC 3, Brussels (April), 1994, pp.diesel engines, SIA, 1990, pp. 55–61. 227–41.

[89] Golothan DW. Diesel engine exhaust smoke: the influence of [107] Farrauto RJ, Voss KE. Monolithic diesel oxidation catalysts.fuel properties and the effects of using barium-containing Appl Catal B 1996;10:29–51.fuel additive, 1967, SAE paper 670092. [108] Neeft JPA, van Pruissen OP, Makkee M, Moulijn JA.

[90] Wiedemann B, Neumann KH. Vehicular experience with Catalytic oxidation of diesel soot: catalyst development. In:additives for regeneration of ceramic diesel filters, 1985, CaPOC 3, Brussels (April), 1994, pp. 355–64.SAE paper 850017. [109] Neeft JPA, Makkee M, Moulijn JA. Catalysts for the

[91] Ise H, Saitoh K, Kawagoe M, Nakayama O. Combustion oxidation of soot from diesel exhaust gases. I. An explorat-modes of light duty diesel particulates in ceramic filters with ory study. Appl Catal B 1996;8:57–78.fuel additives, 1986, SAE paper 860292. [110] Neeft JPA, van Pruissen OP, Makkee M, Moulijn JA.

[92] Lawson A, Vergeer HC, Drummond W, Mogan JP, Dainty Catalysts for the oxidation of soot from diesel exhaustED. Performance of a ceramic diesel particulate trap over gases. II. Contact between soot and catalyst under practicaltypical mining duty cycles using fuel additives, 1985, SAE conditions. Appl Catal B 1997;12:21–31.paper 850150. [111] Ciambelli P, D’Amore M, Palma V. Catalytic combustion of

[93] Gantawar AK, Opris CN, Johnson JH. A study of the carbon particulate at high values of the carbon/catalystregeneration characteristics of silicon carbide and cordierite mass ratio. In: Twenty-sixth international symposium ondiesel particulate filters using a copper fuel additive, 1997, combustion, Philadelphia: The Combustion Institute, 1996,SAE paper 970187. pp. 1789–96.

[94] Konstandopoulos AG, Gratz LD, Johnson JH, Bagley ST, [112] Ciambelli P, Di Pietro A, Palma V, Vaccaro S. Characterisa-Leddy DG. Ceramic particulate traps for diesel emissions tion of the catalytic oxidation of carbon black by TPD.control: effects of a manganese–copper fuel additive, 1988, Thermochim Acta 1993;227:19–26.SAE paper 880009. [113] Ciambelli P, Palma V, Vaccaro S. Low temperature carbon

[95] Dementhon JB, Martin B, Richards P, Rush M, Williams D, particulate oxidation on a supported Cu/V/K catalyst. CatalBergonzini L, Morelli P. Novel additive for particulate trap Today 1993;17:71–8.regeneration, 1995, SAE paper 952355. [114] McKee DW. Metal oxides as catalysts for the oxidation of

[96] Wade WR, White JE, Florek JJ, Cikanek HA. Thermal and graphite. Carbon 1970;8:623–35.catalytic regeneration of diesel particulate traps, 1983, SAE [115] Mul G, Neeft JPA, Kapteijn F, Makkee M, Moulijn JA.paper 830083. Soot oxidation catalysed by a Cu/K/Mo/Cl catalyst:

[97] Levin MD, Koehler DE, Saile JA. Copper fuel additives as a Evaluation of the chemistry and performance of the cata-part of a particulate emission control strategy, 1990, SAE lyst. Appl Catal B 1995;6:339–52.

¨ ¨paper 901619. [116] Setzer C, Schutz W, Schuth F. Transition metal compound[98] Saito T, Nabetani M. Surveying tests of diesel smoke oxide catalysts for lowering the light-off temperature of

suppression with fuel additives, 1973, SAE paper 730170. particles from diesel exhaust. In: Proceedings 10th interna-[99] Pattas KN, Michalopoulou CC. Catalytic activity in the tional congress on catalysis, Budapest, Hungary, Amster-

regeneration of the ceramic diesel particulate trap, 1992, dam: Elsevier, 1993, pp. 2629–32.¨SAE paper 920362. [117] Ahlstrom AF, Odenbrand CUI. Catalytic combustion of

[100] Pattas KN, Samaras Z, Roumbos A, Lemaire J, Mustel W, soot deposits from diesel engines. Appl Catal 1990;60:143–Rouveirolles P. Regeneration of DPF at low temperatures 56.

¨with the use of a cerium based fuel additive, 1996, SAE [118] Ahlstrom AF, Odenbrand CUI. Combustion of soot depositspaper 960135. from diesel engines on mixed oxides of vanadium pentox-

[101] Pattas KN, Kyriakis N, Samaras Z, Manikas T, Mihailidis ide and cupric oxide. Appl Catal 1990;60:157–72.A, Mustel W, Rouveirolles P. The behaviour of metal DPFs [119] Badini C, Saracco G, Serra V. Combustion of carbonaceous


materials by Cu–K–V based catalysts. I. Role of copper oxidation of carbon black by NO : influence of the pres-2

and potassium vanadates. Appl Catal B 1997;11:307–28. ence of water and oxygen. Carbon 2001: in press.[120] Badini C, Serra V, Saracco G, Montorsi M. Thermal [132] Chugtai AR, Welch WF, Akhter MS, Smith DM. A spectro-

stability of Cu–K–V catalyst for diesel soot combustion. scopic study of gaseous products of soot — oxides ofCatal Lett 1996;37:247–54. nitrogen/water reactions. Appl Spectosc 1990;44:294–8.

[121] McKee DW. Mechanisms of the alkali metal catalysed [133] Chugtai AR, Gordon SA, Smith DM. Kinetics of thegasification of carbon. Fuel 1983;62:170–5. hexane soot reaction with NO /N O at low concentration.2 2 4

´[122] Bellaloui A, Varloud J, Meriaudeau P, Perrichon V, Lox E, Carbon 1994;32:405–16.Chevrier M, Gauthier C, Mathis F. Low temperature diesel [134] Cooper BJ, Radnor HJ, Jung W, Thoss JE. Treatment ofsoots combustion using copper based catalysts modified by diesel exhaust gases. US patent 4,902,487 (1990).niobium and potassium promoters. Catal Today [135] Lur’e BA, Mikhno AV. Interaction of NO with soot. Kinet2

1996;29:421–5. Catal 1997;38:490–7.[123] Shangguan WF, Teraoka Y, Kagawa S. Simultaneous [136] Gray PG, Do DD. Modelling the interaction of nitrogen

removal of NO and diesel soot particulates over ternary dioxide with activated carbon. I. Adsorption dynamics at ax

AB O spinel-type oxides. Appl Catal B 1996;8:217–27. single particle. Chem Eng Commun 1992;117:219–40.2 4

[124] McKee DW, Chatterji D. The catalytic behaviour of alkali [137] Gray PG, Do DD. Modelling the interaction of nitrogenmetal carbonates and oxides in graphite oxidation reactions. dioxide with activated carbon. II. Kinetics of reaction withCarbon 1975;13:381–90. pore evolution. Chem Eng Commun 1993;118:333–42.

[125] Harris DJ, Smith IW. In: Twenty-third international sym- [138] Tabor K, Gutzwiller L, Rossi MJ. Heterogeneous chemicalposium on combustion, Philadelphia: The Combustion kinetics of NO on amorphous carbon at ambient tempera-2

Institute, 1990, p. 1185. ture. J Phys Chem 1994;98:6172–86.´[126] Kapteijn F, Meijer R, Moulijn JA, Cazorla-Amoros D. On [139] Yagud BY, Amirova SA, Kefer RG. The reaction of

why do different carbons show different gasification rates: a nitrogen dioxide and nitrosyl chloride with activated char-transient isotopic CO gasification study. Carbon coal. Russ J Inorg Chem 1971;16:792–4.2

1994;32:1223–31. [140] Cooper BJ, Thoss JE. SAE paper 890404, 1989.[127] Otto K, Shelef M. Effects of selected catalysts on kinetic [141] Jelles SJ, Makkee M, Moulijn JA, Acres GJK, Peter-

parameters of the carbon–steam reaction. Chem Eng Com- Hoblyn JD. Diesel particulate control. Application of anmun 1980;5:223–32. activated particulate trap in combination with fuel additives

[128] McKee DW. Catalysis of the graphite–water vapour re- at an ultra low dose rate, 1999, SAE paper 1999-01-0113.action by alkaline earth salts. Carbon 1979;17:419–25. [142] Oi-Uchisawa J, Obuchi A, Ogata A, Enomoto R,

[129] Kapteijn F, Abbel G, Moulijn JA. CO gasification of Kushiyama S. Effect of feed gas composition on the rate of2

carbon catalysed by alkali metals. Fuel 1994;63:1036–42. carbon oxidation with Pt /SiO and the oxidation mecha-2

[130] Jacquot F, Logie V, Brilhac JF, Gilot P. Comparison of soot nism. Appl Catal B 1999;21:9–17.reactivity in the presence of oxygen or NO . In: Nanoparti- [143] Ciambelli P, Palma V, Russo P, Vaccaro S. The role of No in2

cle Congress of ETH, Zurich, 2000, August. the regeneration of catalytic ceramic filters for soot removal[131] Jacquot F, Logie V, Brilhac JF, Gilot P. Kinetics of the from exhaust gases. Catal Today 2000;60:43–9.

Documents

The Oxidation of Soot a Review of Experiments, Mechanisms and Modesl_Stanmore-Etal_2001