Mild Combustion

Progress in Energy and Combustion Science 30 (2004) 329–366 www.elsevier.com/locate/pecs

Mild Combustion Antonio Cavalierea,*, Mara de Joannonb a

Dipartimento di Ingegneria Chimica, Universita` Federico II, Piazzale Tecchio no 80, 80125 Napoli, Italy b Istituto di Ricerche sulla Combustione, C.N.R., Napoli, Italy Received 12 March 2002; accepted 27 February 2004

Abstract The Mild Combustion is characterized by both an elevated temperature of reactants and low temperature increase in the combustion process. These features are the results of several technological demands coming from different application fields. This review paper aims to collect information which could be useful in understanding the fundamentals and applications of Mild Combustion. The information in this field are still sparse, because of the recent identification of the process, so that many speculative considerations have been presented in order to make the whole framework more consistent and rich with potential new applications. A rigorous definition of Mild Combustion is preliminarily given in order to fix the input variables of the process. Under these constraints the influence of the physical, thermodynamic and chemical variables on the most relevant outlet parameters are analyzed. The physical aspects taken into account are atomization, evaporation, mixing and radiative heat transfer. In particular, the evolution of the mixing layer for high temperature diluted oxidant is analyzed. It is shown that mass fluxes through the stoichiometric isosurfaces are lower than those in not diluted conditions and that the annihilation of these isosurfaces is enhanced in the Mild Combustion conditions. Both effects infer low rates of heat release according to the experimental results reported in the literature. The thermodynamic aspects are dealt through the comparative analysis of the minimum, maximum and equilibrium temperature profiles versus the mixture fraction in the whole allowable range for the diluted and not-diluted cases. The chemical aspects have been analyzed in relation to the chemical kinetics rates for different oxidative routes and the temporal evolution of the self-ignition process. The molecular oxygen addition, the hydroperoxide dissociation and atomic hydrogen oxidation are evaluated in wide pressure and temperature ranges. In such a way self-ignition regimes which rely on different preferential chemical kinetics routes are identified and comparison between diluted/not diluted conditions are performed for a fixed evolution time. In this case it is shown that Mild Combustion conditions extend the pressure – temperature range, in which the oxidation is depressed, at relatively low pressure, whereas the ‘ceiling temperature’ is shifted to lower temperature for Mild Combustion condition at higher pressure. The second part of the review shows the potentialities of the diluted high temperature air combustion in applications related both to efficiency and pollution of thermal generation as well as to abatement of the pollutants along the flue gas stream of a primary combustion system. Some selected examples in these fields as land-base gas-turbines, boiler combustion chamber and domestic heating systems are presented. In these, the emphasis, is put preliminarily on aspects related more to efficiency than to pollution reduction, even though this target is implicitly taken into consideration. Then environmental benefits are dealt in relation to the major and minor species, either organic or inorganic, which can be produced in gas/liquid combustion. They include carbonaceous material, unburned hydrocarbons, nitrogen oxides and sulphur oxides. Finally, a classification of the possible processes relevant along the whole fuel transformation in Mild Combustion is given. In particular ‘clean’, ‘cleaning’, ‘clearing’ combustion processes are identified as a convenient categorization in relation to

* Corresponding author. Tel.: þ 39-81-768-32-79; fax: þ39-81-593-6936. E-mail address: [email protected] (A. Cavaliere). 0360-1285/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.pecs.2004.02.003

330

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

the incorporation of pre-combustion or post-combustion units in the main combustion systems. q 2004 Elsevier Ltd. All rights reserved. Keywords: Combustion; Hydrocarbons; Reactor

Contents 1. Introduction and definitions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2. Fundamentals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Physical aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.1. Atomization and evaporation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1.2. Mixing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Fluid-dynamic aspects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3. Thermodynamic and chemical aspects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4. Kinetic aspects. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.4.1. Controlling reaction steps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.5. Features of kinetic controlled processes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.6. Luminous emission and radiative and convective heat exchange. . . . . . . . . . . . . . . . . . . . . . . . . 3. Basic aspects related to technological applications with the efficiency, reliability, economic benefit . . . 3.1. High pressure devices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.1. High pressure diluted self-ignition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.2. Alternative engines (HCCI) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.1.3. Gas turbine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2. Low-pressure devices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.1. Heat preheating classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.2. Heat mode (recirculation and staging) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.2.3. Heat and mass mode (recirculation and staging). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. Basic aspects related to technological application with environmental benefits. Clean – cleaning– clearing combustion. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Soot depression, destruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. NOx depression, destruction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Appendix A. Clearing combustion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.1. SOx depression and removal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . A.2. Minor inorganic elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1. Introduction and definitions Combustion processes are controlled by fluid-dynamic, thermodynamic and composition variables. Among them the temperature is the most representative in characterizing the process. It is common to distinguish the combustion processes in dependence of their temperature: in this sense a rough classification of combustion processes divides them as occurring at either low or high or intermediate temperature. This is, indeed, a loose way of classifying such complex processes, and usually needs at least additional specification of the stage of the process considered. There is further complexity when there is more than one temperature relevant to the combustion process. For instance, processes designed to control both the minimum and the maximum temperatures are hard to be described when the two temperatures are changed in opposite directions. It could seem an incongruity to refer to

330 335 335 335 336 337 340 341 341 344 345 346 347 347 348 348 350 350 351 352 354 354 357 357 358 359 359 360 362

a process as developing at low temperature when the reactants are preheated at relatively high value. Nevertheless there are processes which satisfy such conditions [1–3]. From a historical point of view systems with so a high reactant temperature were introduced by researchers who focused their interest on preheating regenerative systems [4– 6] applied to air system. Therefore, the most used acronym is related to air. It is High Temperature Air Combustion (HiTAC) and it appears in several review lectures [5,7,8] as well as in the title of several symposia [1– 3]. HiTAC, previously known as highly preheated air combustion, refers to a rigorous definition reported the first time by Katsuki et al. [5], originated from the concept of the ‘large excess enthalpy combustion’ introduced by Weinberg [9]. It is a process in which air temperature is such high that the inlet temperature of reactants is higher than autoignition temperature of the mixture. High temperature combustion

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

Nomenclature Tin Tsi TWSR E R cp Wf Q Yf Yox Z Zs Zsund Zsdil Zo ; Z1 dm dstretch m

inlet reactor temperature self-ignition temperature working temperature of a well stirred reactor activation energy universal gas constant specific heat at constant pressure fuel molecular weight heat of combustion fuel mass fraction oxidizer mass fraction mass mixture fraction stoichiometric mass mixture fraction stoichiometric mass mixture fraction in undiluted conditions stoichiometric mass mixture fraction in diluted conditions mass mixture fraction boundary conditions mixing layer thickness mixing layer thickness in stretched conditions

technology (HiCOT) refers in wider sense to all the technologies which exploit high temperature reactants. In this sense it is not limited to the use of air. Other definitions have also been used during last years. Mild Combustion is a subset of the HiTAC or HiCOT domain. Mild Combustion is an unfamiliar term and an unfamiliar subject. In order to clarify the conditions at which Mild Combustion refers let us start by using the following example of a well stirred reactor (WSR) where a stoichiometric mixture of methane/oxygen/nitrogen is fed with a residence time of 1 s. In Fig. 1 working temperature ðTwsr Þ computed in adiabatic conditions for three values of oxygen molar fraction ðXO2 Þ are reported as a function of inlet temperature ðTin Þ: The results of the calculation at XO2 ¼ 0:2; represented in Fig. 1 with the dashed line, show the typical trend of Twsr versus Tin expected for such a system, in fact an S-shape curve is partially visible. According to the classical literature on reactor behavior [10,11] self-ignition temperature ðTsi Þ of a WSR is the inlet temperature at which any differential temperature increase makes the system reach the higher branch of the S-shape curve and the chemical process self-sustains. Therefore, for Tin equal or higher than Tsi mixture ignites and burns increasing the temperature of the system. The maximum temperature increase ðDTÞ is the difference between the maximum temperature, which occur in the reactor, and the temperature of inlet reactants Tin : In this case for Tin ¼ 1100 K the resulting DT is about 1600 K and much of the hydrocarbon fuel is consumed during the considered residence time. Now if we change the operating conditions by increasing the dilution level at constant methane/oxygen ratio, the system responds by reducing the temperature

erfðjÞ Csat Dn Yox1 ; Yf 1 Yox2 ; Yf 2

ns XO2 D t g SR DT DTeq DTaf D D0 t C=O

331

error function saturation factor distance between intermaterial surfaces mass fraction of oxidizer and fuel in fuel feed mass fraction of oxidizer and fuel in oxidizer feed stoichiometric mass ratio oxygen molar fraction mass diffusivity time stretching factor stretch ratio temperature increase during combustion temperature increase for equilibrium condition temperature increase when adiabatic flame temperature is reached fuel droplet diameter initial fuel droplet diameter residence time in a well stirred reactor carbon/oxygen molar ratio

increase during oxidation. For instance, at XO2 ¼ 0:1 and Tin ¼ 1100 K DT is about 1000 K and becomes 550 K for XO2 ¼ 0:05 which is a very low temperature increase compared with the conventional combustion processes. The latter condition can be considered to belong to Mild Combustion category. On the basis of these indication we define Mild Combustion in the following way: “A combustion process is named Mild when the inlet temperature of the reactant mixture is higher than mixture self-ignition temperature whereas the maximum allowable temperature increase with respect to inlet temperature during combustion is lower than mixture self-ignition temperature (in Kelvin).” This means that process evolves in a rather narrow temperature range, which could be placed in an intermediate region between the very fast kinetics of the oxidative

Fig. 1. Working temperature of a well stirred reactor as a function of inlet temperature for stoichiometric mixture of methane, oxygen and nitrogen.

332

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

undiluted conditions and the relatively slow kinetics linked to low temperature self-ignition regimes. Parameters relevant for the process are inlet flow composition (i.e. percentage of fuel, oxidizer and diluent), pressure and the minimum residence time of reactants. These are the conditions which determine the self-ignition temperature of the homogeneous fuel/air/diluent mixture. DT varies according to system configuration. For instance, if reactants are not premixed, the maximum temperature corresponds to the adiabatic flame temperature referred to the stoichiometric condition which may also occur when feed ratio is different from the stoichiometric one. On the other hand, the maximum temperature of a premixed stream is fixed by the feed ratio between fuel, oxidizer and diluent. It is noteworthy that the maximum temperature is related to the maximum oxidation level, which may be different from both the equilibrium and real temperature reached in the reactor. For instance a CH4/O2/N2 system evolving in rich, diluted conditions may attain different temperatures according to different kinetic routes followed in the chemical process that, in turn, do not necessary lead to the maximum oxidation level. The diagrams shown in Figs. 2 and 3 are examples for supporting this statement. Fig. 2 pertains to a WSR temperature increase DT plotted versus inlet temperature for XO2 ¼ 0:05 and XCH4 ¼ 0:1 (carbon/oxygen ratio ¼ 1) with a dilution level corresponding to XN2 ¼ 0:85; with a residence time of 1 s (solid line). Details related to this analysis are fully described by de Joannon et al. [12]. The dotted and the dashed lines are DT computed for the equilibrium ðDTeq Þ; and theoretical adiabatic flame temperature ðDTaf Þ of the system respectively. In the case here considered DTaf is always higher than DT because in rich diluted condition the main oxidation product is CO. The shape of solid line is due to different product distributions obtained by changing inlet temperature. The three arrows related to the curve refer to different outlet compositions according to the different oxidation channels.

Fig. 3. Tin-DT locus of different combustion modes for a methane/oxygen/nitrogen mixture.

The usefulness of the Mild Combustion definition may be appreciated by some considerations in relation to the map shown in Fig. 3 obtained for the same chemical system presented before, i.e. CH4/O2/N2 with 0.1/0.05/0.85 molar fractions. It defines all possible inlet temperature (abscissa) and temperature increase (ordinate) for a residence time of 1 s and atmospheric pressure. In this case the self-ignition temperature of reactant mixture is 1000 K according to an evaluation based on a numerical computation [13] and shown in Fig. 2. The map of Fig. 3 is divided in three regions by the straight lines intercepting the self-ignition temperature on both axes. These regions are named Feedback, High Temperature and Mild Combustion respectively. According to the definition given previously (i.e. DT , Tin , Tsi ) Mild Combustion is placed in the lower-right quadrant. The other two combustion modes are placed in the upper part of the map where the condition DT . Tsi is satisfied. Although it could be appear pleonastic, the conditions corresponding to the three combustion modes have been also schematically reported in Table 1 in order to enhance the graphical representation of Fig. 3. The meaning of Mild Combustion in comparison to the other two fields is quite straightforward. It differs from the other two regimes because in Mild Combustion the process cannot be sustained without preheating the reactants. In contrast, Feedback and High Temperature Combustion Table 1 Summary of conditions identifying the different combustion modes reported in Fig. 3

Fig. 2. Temperature increment ðDTÞ computed in a well stirred reactor versus inlet temperature for working, equilibrium and adiabatic flame conditions for a methane/oxygen/nitrogen mixture.

Combustion mode

Inlet conditions

Working conditions

Feedback combustion High temperature air combustion Mild combustion

Tin , Tsi Tin . Tsi

DT . Tsi DT . Tsi

Tin . Tsi

DT , Tsi

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

satisfy the necessary condition for which a traditional combustion process may occur, namely the heat release is sufficient to sustain the minimum temperature required for process evolution. It is worth stressing that self-ignition temperature, to which the definition refers, is that of fuel/air/diluent mixture. Different self-ignition temperatures may be obtained for different compositions, so that each map such as Fig. 3 corresponds to a fixed composition. In other words, different points on the map are working conditions of a single reactor with the same pressure, the same types of fuel, oxidizer and diluent and the same residence time. Another definition of Mild Combustion is given by Oberlack et al. [14] and Peters [15]. In relation to the WSR behavior they identify as Mild the conditions at which the ignition and extinction points no longer exist and a monotonic shift from unburned to burned conditions occurs. The curve reported in Fig. 1 at XO2 ¼ 0:05 is representative of mild condition also according to such definition. It is clear that also in this case the inlet temperature has to be higher than self-ignition temperature of the mixture. By imposing the condition to the steady state mass and heat balances, according to the authors Mild conditions corresponds to the following relation:
 cp Wf Tin E=R #4 1þ ð1Þ Tin QYf where R is the universal gas constant, cp the specific heat at constant pressure, Wf the molecular weight of the fuel, Q the heat of combustion of the mixture, Yf the inlet fuel mass fraction and E the activation energy of an overall one-step reaction. The term ððQYf Þ=ðcp Wf Tin ÞÞ can be approximate to ððT 2 Tin Þ=Tin Þ ¼ ðDT=Tin Þ where DT is the temperature increase during combustion. By substituting this relation in Eq. (1) it is possible to obtain a relation between Tin and DT that has been plotted in Fig. 4 where ðDT=Tin Þ has been reported as a function of Tin for an activation energy of 40 kcal/mol. The region delimited by the curve represents

Fig. 4. Identification of Mild Combustion conditions according to Oberlack et al. [14].

333

the conditions where Mild Combustion occurs. In this region temperature increase is always lower than inlet temperature. For instance, for Tin ¼ 1000 K the temperature increase due to heat release should not exceed 25% of the inlet temperature if Mild Combustion mode is to be maintained. The definition of Mild Combustion given in the present review paper is unambiguous because criteria which should be fulfilled to include a process in Mild Combustion are well defined in univocal way. The need and usefulness to define this process separately from the others and why its name is appropriate deserve further specifications. In other words why is important to consider Mild Combustion regime, is there any practical use of this mode of combustion? The need of such a definition is related to the simultaneous occurrence both of ‘higher temperature than self-ignition one’ and of ‘temperature increase lower than a prefixed value’ that is not a trivial coincidence and it seems apparently contradicting. Reactants pre-heating can be obtained in different way according to the choice to have a premixed system or a fuel/oxidizer separated system. In the last case the high temperature of reactant mixture can be ensured by heating or fuel or oxidizer or both. The reasons why we have chosen the name Mild are two. The first one is that this word is in contrast with the characteristic of all the other combustion processes. These latter evolve in a very wide temperature range in which whatever temperaturedependent process may chaotically (related to turbulence) pass through several regimes. Kinetic can change during the completion of the process from low to intermediate or high temperature regimes, the physical parameter, like diffusion, surface tension, can also change abruptly from one to the other. In contrast, Mild Combustion mode is characterized by ‘mild’ changes and ensures a more gradual evolution during the process. The second reason is that mild is the acronym of ‘moderate or intense low-oxygen dilution’ which is exactly one of the most typical conditions for which the process can be obtained. The relevance of this condition is due to its relatively simple feasibility and that it may be tuned in such a way that it prevents from soot and NOx formation. This point, which will be discussed and supported by literature in Section 4, is also related to the last question on Mild Combustion. Is there any practical use of this mode of combustion? The first answer is that any new concept in combustion has to be explored independently on its immediate application. Usefulness of a combustion process has to be shown along the years and economic constrains sometime obscure long-time convenience. The second answer is that this mode has great potentials. This is linked to the fact that combustion process can be restricted to relatively low maximum temperature and temperature increase when Mild Combustion is adopted. The limitation of the maximum temperature can be exploited to limit soot and NOx production as it has been just mentioned. Furthermore, the maximum temperature can be adjusted in such a way that it is lower than that a high temperature metallic material can resist. In the field of combustion

334

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

engines this leads to an extremely useful freedom degree in the design of a combustion chamber. In fact, wall confinement can be used without external or internal cooling and any kind of fluid-dynamic configuration can be used. Moreover, the requirement of inlet temperature higher than the self-ignition one adds a further freedom degree, because flame stabilization based on internal recirculation is not needed and great variety of fluiddynamic configurations can be envisaged in the combustion chamber design. Finally, homogeneous temperature field can be yield in Mild Combustion regimes. This is ensured both by the constrain on temperature increase, which is intrinsic in the Mild Combustion definition, and on some preliminary results that will be later presented [16,17]. This makes the process very unique in the material treatment field because this ensures temperature homogeneity and control on the material surface. Some applications in the steel treatment testifies the feasibility in exploiting such characteristic of the process [18,19]. The homogeneity makes possible also to control both the combustion process itself and the addition of any chemical which can be beneficial in oxidation process or in its application. In other words the Mild Combustion makes the combustion chamber more similar to other chemical reactors, which are temperature controlled with the consequent benefit to adjust and tune the temperature in a ‘convenient’ window. An example of such characteristic will be given in Section 4 where selective noncatalytic reduction (SNCR) is described. This process is used for the reduction of nitrogen oxides and needs a very narrow temperature window to be exploited. All these advantages should not obscure potential counter indications and fundamental problems associated with this regime of operation. The relatively low temperature interval can favor partial oxidation processes which can yield oxygenated compounds as possible pollutants. The start of oxidation process with the self-ignition of the mixture formed by the two reactants without combustion product recirculation may cause some problems in the stabilization of reaction zone in well-defined position. In fact self-ignition is very sensitive to local temperature and composition. The ideal condition can be obtained in different part of the reactor and this can generate too intense oscillation of the first oxidative region. These two disadvantage are reported [20,21] for one application which can be included in Mild Combustion category, namely the homogeneous charge compression ignition (HCCI) engines. It is known that such engines have problem in oxygenated species and CO production and in the control of the autoignition time. The reason why HCCI exploits a Mild Combustion process and characteristics of the engine operated in such a mode will be discussed in Section 3.1. Related definitions to Mild Combustion are high temperature reactant combustion, flameless combustion and colorless combustion. Flameless [7,22] or colorless [7,18]. Combustion are definitions which pertain to properties of the reactors rather than to inlet conditions.

Flameless combustion is defined as a mode in which two conditions must be satisfied where the reaction is going to take place: (1) the reactants must exceed self-ignition temperature; (2) the reactants must have entrained enough inert combustion products to reduce the final reaction temperature well below adiabatic flame temperature, so much that a flame front can not be stabilized [23]. The terms flameless and colorless refer to the outstanding characteristic that no visible emission is detectable in oxidation regions. The first one is more restrictive because some other properties are linked to the colorless features, such as essentially uniform distribution of chemical and thermodynamic variables. Based on inlet conditions, a type of definition and classification of high temperature air combustion was given by Hasegawa et al. [24] in order to simplify the identification of HiTAC conditions useful for a specific application. In this sense the authors [24] identified an HiTAC index associated with the level of air preheating and oxygen concentration in the system. More recently, one more definition of Mild Combustion was given by Kumar et al. [25] based once again on parameter inside the reactor. Their definition is based on quantification of the temperature variation inside the reactor. According to the authors a combustion process is mild when the normalized spatial temperature variation is around 15%. It is hard to say at the moment whether these definitions are coincident with Mild Combustion. All these processes rely on highly preheated and diluted systems, but it is difficult to fix the quantitative extent to which the different definitions apply. If these definitions are shown to coincide, the Mild Combustion conditions could indicate the criterion to satisfy at inlet section for the recognition of flameless or colorless regime. The only suitable choice is to emphasize on the identification of inlet conditions also for the precision required in the text. The review is divided in two main parts. The first one is presented in the following section and it deals with fundamental aspects of Mild Combustion. Physical, fluiddynamic, thermodynamic and chemical are considered as separated from each other even though they can interact in complex way. Chemical kinetics is described with more details than other aspects because it presents more numerous peculiarities with respect to the other ones. The second part is reported in Sections 3 and 4. It deals with applications of Mild Combustion which have shown their feasibility as well as with possible applications which still need to prove their effectiveness. In order to make less speculative the presentation of potentials, applications have been described in the framework of basic aspects which are dealt, differently from Section 2, taking into account the complexity of the interactions of the effects described in Section 2. Applications are classified in two sub-sections, according to their use in the field of energy transformation or of pollutants production/destruction. In these sections

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

the review quotes what is known in the literature and deals either with the expected results or with the features of the Mild Combustion when it is thought to be applied for a specific purpose under specific operative condition. For instance, Mild Combustion at high pressure as it is common in engine conditions is dealt in Section 3 either to analyze the influence of the pressure on kinetics or to present some present or future applications. Similarly, Section 4 includes topics which analyses how formation of pollutants is affected by Mild Combustion conditions as well as possible applications, which have been shown to be feasible in other field. A methodological aspect of this paper, which is worthwhile to stress preliminarily, consists in pointing out process potentials even when such potentials have not yet been realized experimentally. In this respect this review is purposely more suggestive rather than propositive. It stresses the aspects related to new combustion concepts rather than partial technical improvements, and it is addressed to unsolved problems rather than to deep understanding of second-order quantitative aspects. Therefore, for this reason, and for lack of room, some interesting ‘classical’ aspects are necessarily neglected and are only briefly quoted by means of mention of other reviews, even though they deserve extended analysis.

2. Fundamentals 2.1. Physical aspects 2.1.1. Atomization and evaporation The relatively high temperatures of oxidizer stream in Mild Combustion processes leads to a faster heating of the liquid fuel. The increase of liquid– gas interface temperature makes surface tension decrease leading to a better atomization with respect to the process without preheating [26]. This effect is stronger when pressure increases. In this case surface tension decreases, approaching zero when the critical pressure is reached [26]. In this pressure range fuel droplet fragments are formed with no resistance and no further coalescence occurs, as the minimization of surface free energy should suggest. The high temperature of the reactants also influences the droplet evaporation that occurs in series and/or in parallel with the atomization process, whereas the pressure effects are negligible when the pressure values are lower than the critical one. On the other hand, when the pressure approaches the critical pressure, the latent heat of vaporization decreases to zero. Therefore liquid gasification occurs without additional heat exchange for phase change [27]. This process evolution is similar to fuel evaporation or gasification in gas turbine (GT) or diesel engine. Critical conditions at the droplet interface are reached only if

335

the pressure and temperature in the environment are about twice the fuel critical pressure and temperature [27]. The relatively high characteristic residence times of Mild Combustion processes are the only significant differences with respect to diesel and GT processes. In relation to vaporization, this means that the liquid fuel crosses an environment where the temperature does not increase with time, in distinction with the traditional combustion processes where the temperatures reach the flame values of 1800–2000 K. The rate of droplet evaporation increases with temperature such as shown in the diagram of Fig. 5 where normalized square droplet diameter ðD2 =D2o Þ of a gasoline droplet is reported as a function of time for several ambient temperatures [26]. The evaporation time passes from 10 ms at T ¼ 2000 K to 20 ms at T ¼ 1200 K. As a consequence in Mild conditions, droplet lifetime can increase and become comparable with mass diffusion time in liquid phase. The thickness of diffusive layer inside the droplet enlarges with time. Therefore, in the case of hydrocarbon mixtures, the evolution of evaporation process shifts from the ‘onion peeling’ toward a ‘distillative’ mechanism [28,29]. In traditional combustion processes this could represent a significant problem due to not-uniform vapor fuel composition that significantly influences the characteristic of ignition, flame stability and pollutant formations. For Mild Combustion processes at least with very high inlet temperature and provided the absence of a flame front, such not-uniform local conditions do not affect the process evolution. Moreover, in the case of very different volatilities among droplet constituents, the liquid phase enriches in the heavier species that, in turn, can solidify leading to the formation of cenospheres that progressively or rapidly lose internal liquid components [30,31]. The high radiative heat transfer that characterizes Mild Combustion systems, discussed in Section 2.6, can also affect the evaporation time by improving the droplet heating [28].

Fig. 5. Normalized square diameter of a gasoline droplet in a hot environment as a function of time.

336

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

2.1.2. Mixing The presence of high diluent concentration also influences the mixing between fuel and oxidizer in the combustion chamber due to very low concentration level that represents the driving force of the process. In order to describe the evolution of the mixing, the mixture fraction of any conservative variable is generally used and here we consider the mass mixture fraction ðZÞ for the evaluation of the dilution effect on the process. Under the hypothesis of equi-diffusion, Z is defined as: Z¼

Yf 2 Yf 2 Yox 2 Yox2 ¼ Yf 1 2 Yf 2 Yox1 2 Yox2

ð2Þ

where Yf and Yox are, respectively, the fuel and the oxidizer mass fraction in any position inside the mixing layer. The subscripts 1 and 2 refer to mass fractions in the fuel and oxidizer feeds. The stoichiometric mixture fraction Zs represents the mixture fraction in the region of the mixing layer where the stoichiometric mass ratio ns is reached. It is given by: Zs ¼ 1 þ ns

Yf 1 Yox2

!21 ð3Þ

where ns is the stoichiometric mass ratio (about 15 for paraffins) [32]. On the basis of the Eq. (3) Zs decreases when the oxidizer is diluted, as occurs in the case of Mild Combustion. It changes from 0.06 to about 0.01 by changing the oxygen molar fraction ðXO2 Þ at the inlet from 0.21 to 0.05 [33]. Following the presentation of mixing layer combustion reported in a review paper [34], the spatial position where the stoichiometric condition occurs can be easily identified in the case of a unidimensional, stretched or unstretched mixing layer that represents all categories of mixing layer. In this case, it is possible to obtain the mixture fraction

distribution Z; in terms of the error function:
 Z 2 Z0 x ¼ erf j ¼ Z1 2 Zo dm

ð4Þ

where Z0 and Z1 are the mixture fraction boundary conditions, dm is the thickness of the mixing layer in unstretched case and x the spatial coordinate. dm and the error function can be expressed as: pffiffiffiffiffi dm ¼ 4Dt ð5Þ ð j 2 2 erfðjÞ ¼ pffiffi e2x dx ð6Þ p 0 where D is the diffusivity and t the time. In the case of the stretched mixing layer it is possible to define a stretch ratio (SR) in relation to a surface area as the ratio of the material surface area at time t and the area at to : In this condition it was shown that dm becomes: pffiffiffiffiffi SR2 stretch ð7Þ dm ¼ dm g ¼ dm SR where SR is the stretch of the intermaterial line perpenstretch dicular to the mixing layer. Therefore, differs pffiffiffiffiffithe dm

from dm by a stretching factor g ¼ ð SR2 =SRÞ: The spatial distribution of the mixing layer mixture fraction is shown in Fig. 6a for four consecutive times and for boundary conditions of x ¼ 21; Z ¼ 1 and x ¼ þ1; Z ¼ 0: In diluted conditions the stoichiometric surface lies in a more peripheral region of diffusive layer with respect to the undiluted case. The blow up of Z profiles for three selected time ðt1 ; t2 ; t3 Þ is reported in Fig. 6b in order to better show this point. In particular the black line pertaining t ¼ t2 shows that when the mixture fraction passes from undiluted conditions Zsund ¼ 0:06 to diluted one, Zsdil ¼ 0:01; the dimensionless position of the stoichiometric isosurface with respect to the intermaterial line jstoich ; passes from 1.1 to 1.8 and the x position passes from 0.026 to 0.038 in the arbitrary unit scale of the plot.

Fig. 6. (a) Spatial distribution of mixture fraction Z in a mixing layer at different time; (b) blow up of the spatial region where the stoichiometric mixture fraction in diluted ðZsdil Þ and undiluted ðZsund Þ condition occur.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

337

The effects of dilution on the mixture fraction gradient can be also evaluated. The mixture fraction gradient is given by:
    2 ›Z  ›Z  ›j 1 2 1 p ffiffi ¼ ¼ ð8Þ e2j Z¼Z   s d ›x Z¼Z ›j ›x p 2 Z¼Zs

s

The derivative of the combined variable and of the error function are given by: 2 d 2 erfðjÞ ¼ pffiffi e2j dj p
 dj d x 1 ¼ ¼ dx d dx d

ð9Þ ð10Þ

Therefore, the ratio between the gradients in diluted and undiluted conditions is:   ›Z  2 e2j Z¼Z und 2j2 þj2 ›x Z¼Zsund  s ¼ e Z¼Zsund Z¼Zsdil ¼ ð11Þ 2 ›Z  e2j Z¼Z dil s ›x Z¼Z dil s

This means that the diffusive fluxes are lower in diluted conditions with respect to the undiluted ones since Zsund . Zsdil : For instance, for diluted ðXO2 ¼ 0:05Þ and undiluted ðXO2 ¼ 0:2Þ conditions, the mixture fraction gradient ratio is of the order of 8. Then, when reaction rates are high enough that the fuel oxidation becomes diffusion-controlled, fuel consumption rate is lower in diluted conditions. Moreover, the diffusive structures have a longer lifetime in diluted conditions because the gradient decrease makes the stochiometric isosurface further shift from the aerodynamic extinction limit. Since the Mild Combustion process is based on fuel self-ignition due to the high temperature of oxidizer, aerodynamic extinction can be neglected at very high temperature because the oxidation process is not dependent on heat release but only on oxidizer temperature. The shift of the stoichiometric mixture fraction zone toward the periphery of the mixing layer also leads to an increase of annihilation effects. This can be analyzed in unidimensional conditions by referring to the double diffusive layer scheme shown in Fig. 7. At t0 the function is a double step. The distance between the two intermaterial surfaces is Dn : The mixture fraction distribution depends on the time and space according to the following equation [34]:



1 x x 2 Dn erf Z¼ 2 erf ð12Þ 2 dm dm The isosurfaces of the two layers interfere only after a certain time leading to a continuous decrease of Z: In the case of the double layer, the presence of an adjacent diffusive layer can be taken into account by means of a saturation factor [34]: "
 # Dn 2 ð13Þ Csat ¼ 1 2 exp 2 dm

Fig. 7. Mixture fraction distribution in a double mixing layer at different times.

that represents the ratio between the diffusive flow in a double diffusive layer and the diffusive flow in a isolated mixing layer and gives an indication on the interaction between the two diffusive layer. Csat is 1 when the diffusive layer evolves as it were isolated ðDn ! 1Þ: It is interesting to note that ðDn =dm ¼ 2Þ corresponds to the distance at which the two stoichiometric isosurfaces converge and disappear, leading to the annihilation of the diffusion flame. For ðDn =dm Þ ¼ 2; Csat is about 0.86. This means that for Dn $ 2dm it is possible to consider the diffusive layer as isolated. In contrast, for ðDn =dm Þ ¼ 0:1 the saturation factor is around 0.1 and the two layer interact intensely. For ðDn =dm Þ , 0:1 the layer is completely saturated by the second layer, whereas for intermediate values the two layers interact. It was noted that in a isolated diffusive layer the stoichiometric isosurface lies in a more peripheral region of diffusive layer with respect to the undiluted case therefore the annihilation of stoichiometric isosurface in a double diffusive layer begins earlier in time in diluted case with respect to the undiluted one. The same considerations can be performed in more general terms for multidimensional flames. For instance, in the case of a vortex structure it can be seen that the same amount of oxidizer entrained inside the structure in the case of diluted conditions leads to a faster oxygen consumption in the structure. Therefore, under the same fluid-dynamic and heat release conditions, Mild Combustion processes can be initially richer than undiluted ones. 2.2. Fluid-dynamic aspects The use of mild conditions leads to a straight modification of the fluid-dynamic configuration and behavior due

338

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

to the peculiar features of such processes. The main dissimilarities of fluid-dynamic characteristics with respect to ‘traditional’ combustion processes are the possibility of a high level of dry-wall confinement, no need of back-mixing for flame stabilization and disproportion of fuel and oxidizer stream due to the high content of diluent. The first aspect is related to the relatively low maximum temperature of the Mild Combustion processes. In this case it usually ranges up to 1500 K, which is compatible with metallic advanced materials. Therefore, even the combustion products can be guided by means of direct confinement, if corrosion problems are not severe. The consequent advantage is the increase of the mixing among fuel/ oxidizer/products streams. On the other hand, the direct confinement can be useful for tailoring both convective heat exchange and gas expansion. The second characteristic (i.e. no back-mixing for flame stabilization) is peculiar because it makes Mild Combustion very different from the majority of combustion processes. The latter have been designed with the main imperative to anchor the flame in a defined temporal-spatial location. This is mainly achieved by means of internal recirculation of the combustion products, which mix with fresh mixture and supply so much enthalpy that ignition temperature is reached. In contrast, in Mild Combustion the enthalpy content of the reactants is so high that self-ignition is independently assured. Therefore, high-velocity, not swirled inlet jets can be used. The use of such fluid-dynamic configurations in some advanced furnaces [22,26,35–38], where air preheating is coupled with a high level of jet entrainment, can increase recirculation. In this case the recirculation is not exploited directly for the stabilization, but it is the contrivance, through which the oxygen content is decreased and the fuel concentration is lowered when it is dispersed in the recirculation flow. However, this condition can be exploited to add a degree of freedom to the process design. The third fluid-dynamic feature also derives from the limit on the maximum temperature, which is usually achieved by means the oxidizer dilution with flue gas. Typical mass oxidizer/fuel flow ratios, which are around 15 for stoichiometric air/paraffins, may reach values on the order of 100. In these conditions mixing between the two components may result in some problems. One way to face with this difficulty is inject fuel directly in the flue recirculated stream, allowing it to mix with the main air flow. Their two step mixing maintains the maximum temperature lower than that required for an efficient (and detrimental) pyrolysis according to what will be described in the following. The previous considerations are based on a rational analysis of the chemical and physical conditions under which Mild Combustion occurs. Their relative relevance has to be assessed on the basis of experimental and numerical modeling, which has yet to be performed. Only few papers have been devoted to the description of fluid-dynamic

patterns in high temperature diluted feeding, and they deserve explicit mention. Velocity measurements have been performed by Weber and coauthors [39] in order to assess quantitatively the recirculation level entrained in the inlet high-velocity jet in a refractory-lined furnace of square cross-section of 2 m in side and a length of 6 m. Fig. 8 from [39] shows the increase of mass flow rate of the inlet vitiated air jet ðXO2 ¼ 0:19Þ due to the entrainment of combustion products of low oxygen concentration ðXO2 ¼ 0:02 – 0:03Þ; along the axial coordinate, normalized with respect to the inlet jet diameter (about 12 cm). Fig. 8 points out that at a distance equal to about 10 jet diameters ðx=d ¼ 10Þ the jet doubles its original mass and continuously increases up to x=d ¼ 25 where the profile attains a maximum value of about 4. This is evidence of the apparent contradiction between no need of exploiting recirculation for flame stabilization and its practical use in order to decrease the oxygen concentration in the main reactant zone. The same authors have also mentioned in several review contributions [39,40] the need for a more extensive investigation of high temperature jets in high temperature environments. An extensive comparison between measured and modeled patterns of velocity, temperature and species in Mild Combustion has been presented by Coelho et al. in a recent paper [41]. The most important result is the satisfactory agreement between the two sets of data and the consequent suitability of the Flamelet Model (for the main patterns) and Eulerian Particle Flamelet Model (for slow chemistry minor species concentration) in these peculiar conditions. It is noteworthy that the velocity patterns and the recirculation flows are well modeled also. Other papers refer to High Temperature Air Combustion and are more focused on targets different from the fluiddynamics itself, but they still have some information content related to this field which has to be evaluated on the basis of

Fig. 8. Mass flow entrainment ðMe Þ in the inlet vitiated air mass flow ðMo Þ as a function of normalized axial coordinate x=d; after Weber et al. [39].

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

general individual experience. For instance, the paper of Bee´r et al. [42] underlines the role of jet-pumping in the high temperature environment and supports through this fluiddynamic effect the reduction of NOx. Finally it is of great relevance to consider the chance that Mild Combustion might evolve as in a WSR. So far this has been considered feasible only for very small [43] and research-oriented [44] reactors. Only the fluid-dynamic aspect is here analyzed in detail, whereas the kinetic aspect is only mentioned, when details have not been presented in Section 2.4. The main practical feasibility of a WSR is related to the feasibility of high levels of back mixing. A general rule that yields such a condition is to create reverse flows, which are obtained by impinging jets either on walls or on other jets. This allows for splitting and recirculating part of the main streams, which are comparable to each other, whereas shedding vortices or swirled flows would ensure only fractional recirculation of the main stream. Only partial increase of the kinetic energy of the recirculated stream can be obtained with the injection of the fuel inside the reverse flow. In contrast, the creation of impinging jets, which is difficult in traditional combustion systems due to the high temperature of undiluted combustion products, is feasible in Mild Combustion because the product temperature is kept relatively low and dry-wall confinement can guide the flow in whatever direction, even in reverse flow. The second aspect, which pertains to the specific conditions of two step ignition in paraffin Mild Combustion, is that this regime entails an increase of the ignition delay, but it keeps the reaction time of the oxidation process order of magnitudes lower than the ignition delay times. This statement is supported by the analysis presented in relation of the batch conditions for the methane [13] and for longer paraffins [45]. The chance that slow combustion can be sustained in some particular intermediate temperature range for other types of fuels, has to be still considered an open question, but it is a challenging task to show that this can occur with a suitable rate for practical application. The great difference between the pre-combustion and combustion times entails that the WSR has to be designed with a very long average residence time either inside the reactor itself or in the inlet mixing zones. A high value inside the reactor is desirable, to tolerate variable mixing time and, consequently, fuel type, dilution level and initial temperature. The first and second criteria needed for ensuring WSR conditions seem to be conflicting, because the high level of mixing is favored by high level of kinetic energy and this entails low residence times, whereas chemical kinetics seem to be related to characteristic time longer than such fluiddynamic times. Therefore, configurations that alternate slow and fast flows seem the only possible solutions. One of these is shown in the axial-symmetric scheme of Fig. 9a. The flow pattern, generated by such configuration, is here described because it satisfies the two criteria and it can be a reference for a more complex design.

339

Fig. 9. (a) Example of a fluid-dynamic configuration for a Mild Combustion system; (b) Schematization of a real Mild Combustion reactor as a series of ideal vessels.

The oxidizer flow proceeds from the left through an annular confinement toward a transverse wall on which it impinges. Then it converges toward the centerline of the configuration, where its radial impingement splits the flow in a reverse annular vortex and an axial-symmetric forward jet. The transverse wall is tilted with respect to the lateral wall with such wide angle u that the radial impingement occurs with a negative axial velocity with a consequent enhancement of the recirculated flow. This angle has to be considered an adjustable variable which has to be fixed according to actual realization and the needed back mixing. The main controlling variable is the inner diameter D1 of the transverse wall, which determines, together with the external diameter D2 ; the velocity in the impinging region. The last relevant design variable is the fuel injection position. The choice in this case is to use a radial peripheral injection which is sketched in Fig. 9a at a distance L from the transverse wall, i.e. the reactor inlet. This distance has to be fixed according to the characteristic pre-combustion time, in such a way that this is equal to the sum of the mean time to reach the reactor inlet and the average residence time inside the rector itself. The precise evaluation of this position should be performed on the basis of physical or numerical models and it is the very crucial point which makes feasible a WSR in the Mild Combustion regimes. In fact, this is a conservation and safety criterion because it states that, whenever the internal real back mixing is not sufficient to increase the reactant temperature to the reactor temperature, autoignition takes place in any case. The condition, in which a very slight increase of the product temperature occurs, the ideal and practically feasible reactors are coincident.

340

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

This happens when the small difference of the product and reactant temperature of Mild Combustion is made even smaller for a low level of conversion and possible heat extraction. In conclusion the real reactor can be modeled as a series of three ideal vessels as in Fig. 9b. A WSR preceded and followed by a PFR. The first one ensures that great part of the combustion process occurs inside the WSR, the second one ensures that combustion and post-combustion is completed. 2.3. Thermodynamic and chemical aspects It is unambiguous that temperature plays a key role in the evolution of all combustion processes. Three temperatures are generally used for the evaluation of the condition at which the oxidation proceeds, which are the initial frozen temperature, the maximum temperature and the equilibrium temperature. As will be discussed in the following, they identify the boundaries of temperature range that the system can cover during combustion and can be obtained either by heat balance (initial frozen temperature and maximum temperature) or equilibrium evaluation (equilibrium temperature). The frozen temperature, that is the inlet temperature ðTin Þ of the mixture, in correspondence with the mixture fraction Z is: Tin ¼

c
p l2 c
p l1 Y l ð1 2 ZÞT2 þ Y l ZT c
p lmix ox 2 c
p lmix f 1 1

ð14Þ

where the subscript 1 is related to the stream where fuel is present whereas 2 refers to the oxidizer stream, cp is the specific heat at constant pressure, Yox and Yf are the mass fraction of the oxidizer and of the fuel, respectively. The maximum temperature ðTmax Þ is defined as the temperature reached when the inlet species react to form CO2 and H2O (i.e. when complete oxidation occurs). Therefore: for Z , Zs: Tmax ¼ Tin þ

Y f l1 ZQ c
p lmix

ð15Þ

Yox l2 ð1 2 ZÞQ c
p lmix nstec

ð16Þ

Fig. 10. Initial, equilibrium and maximum temperature profiles as function of the mixture fraction for an initial oxygen molar fraction of 0.21.

been computed by means of the EQUIL routine [46] of the CHEMKIN package [47]. The solid lines refer to inlet temperature, whereas the dashed line and dashed/dotted lines refer to the equilibrium and maximum temperature, respectively. In the first condition, the oxidizer stream is characterized by a temperature of 400 K and an oxygen molar fraction of 0.21 whereas the fuel temperature is 300 K. In this case the stoichiometric mixture fraction Zs is equal to 0.05. As it is shown in Fig. 10, Tin decreases with Z increase. In contrast, Tmax increases with Z for Z , Zs and reaches its maximum value of 2600 K at Z ¼ Zs : For Z . Zs ; Tmax decreases with Z increase. Teq shows the same general trend of Tmax even though it is lower than Tmax until Z ¼ 0:6: In fact, Teq increases with Z reaching its maximum value of 1900 K for Z ¼ Zs : For Z . Zs ; Teq decreases with Z increase, sharper than Tmax ; down to 900 K for Z ¼ 0:2: Then Teq becomes nearly constant up to Z ¼ 0:6: For Z . 0:6Teq and Tmax coincide. The second condition represents a typical feed of Mild Combustion process. In fact, the oxidizer stream with an XO2 of 0.05 is pre-heated up to 1200 K. In this case, Zs is about 0.014.

for Z . Zs: Tmax ¼ Tin þ

where Q is the heat of combustion and ns is the stoichiometric oxidizer/fuel mass ratio. The equilibrium temperature ðTeq Þ is the temperature that the system reaches, evolving from the initial conditions without constraints. All three temperatures have been computed for a CH4/O2/N2 system as function of the mixture fraction for two initial conditions and have been shown in Figs. 10 and 11. The equilibrium temperatures here considered have

Fig. 11. Initial, equilibrium and maximum temperature profiles as function of the mixture fraction for an initial oxygen molar fraction of 0.05.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

The trends of the three temperatures are generally the same even though the temperature ranges involved are different. The maxima of Tmax and Teq are about 1800 and 1600 K, respectively. Also in this case, the curve of the equilibrium temperature lays under the Tmax curve up to Z ¼ 0:45: For Z higher than 0.45 the two curves overlap. The temperature profiles shown in Figs. 10 and 11, highlight some features of diluted combustion with respect to the undiluted case. In the first condition, relative to undiluted combustion, the oxidation process occurs according to deflagrative or diffusive structures. Therefore, the temperature is the one obtained from the fuel conversion to CO2 and H2O. This conversion preferably occurs in stoichiometric conditions and only a small amount of combustion products is mixed with fresh reactants. In other words, the working temperature is Tmax or Tin : However, any temperature value is possible because Tmax can decay toward Teq : In contrast, it is not possible that a mixture in the frozen condition can evolve toward equilibrium because ignition does not occur. This behavior is even more important in the case of highly stretched diffusion flames, where the temperatures are relatively low due to the cooling related to the diffusive field. In the case of diluted combustion, equilibrium condition can be reached starting from either the diffusive structure or frozen conditions. The high temperature of the oxidizer and high residence times make it possible for the system to evolve toward the equilibrium condition starting from the frozen one. However, it is significant that the frozen and the equilibrium conditions are nearly the same. Summarizing, in undiluted conditions temperature range between the maximum oxidation and equilibrium temperature whereas in the diluted case they range from the maximum oxidation to the frozen temperature. The comparison of these two cases can be easily obtained from Figs. 10 and 11. The temperature variation is higher in the first case than the second. This is even more true in slightly rich conditions. This consideration is not valid for a small domain around Z ¼ 0:2: The intersection of the two temperature ranges can be extended by increasing the pre-heating temperature, but the temperatures in the diluted case will be always lower than those of the undiluted one. In contrast, in the case of lean conditions, the temperatures in the diluted case are always higher than those obtained in the undiluted condition for the same Z: In this case the lowering of oxidation rates due to the low oxygen concentrations in diluted conditions is balanced by pre-heating. Frozen conditions in the diluted case correspond to a temperature range related to low temperature oxidation kinetics, where an oxygen addition mechanism characteristic of self-ignition occurs with relatively long induction time. In the case here considered, for Z between 0.2 and 0.5 the inlet temperature varies between 800 and 500 K.

341

2.4. Kinetic aspects 2.4.1. Controlling reaction steps The oxidation kinetics of paraffinic hydrocarbon is well described in some reference works [48 – 50]. Its complexity makes difficult the identification of the pathway of fuel consumption in Mild Combustion conditions. In this review paper the attention has been focused on self-ignition mechanisms and their dependences on the environmental parameters, such as temperature, pressure and reactant concentrations. The main reactions affecting self-ignition characteristic times are summarized in Table 2. In the low temperature range fuel self-ignition occurs through a double oxygen addition on the alkyl radical. A sequence of isomerization and oxidation reactions leads to the formation of alkyl-peroxy and alkyl-hydroperoxy radicals that represent the branching agent of the mechanism. The oxygen addition to the alkyl radical is: Rz þ O2 , Rz O2

ð17Þ

Fuel oxidation through the low temperature mechanism depends on the equilibrium conditions of reaction (17). The formation of the alkyl-peroxy radical Rz O2 is favored at low temperatures. When temperature increases, the Rz O2 production decreases until reaction (17) is shifted toward the reactants. The temperature at which the ½Rz O2 =½Rz  ratio is equal to unity marks the limits of the temperature ranges where either the forward or backward reaction is favored. This temperature is indicated as ‘ceiling’ temperature [51,52]. It depends on the temperature and on the oxygen concentration in the system (Fig. 12). In order to have an indication of the pressure and temperature range where the direct or inverse reaction is favored, curves representing the ceiling temperature can be reported on a ðP; TÞ diagram. For reaction (17): ½Rz O2  ¼ Kc ½Rz ½O2 

ð18Þ

By replacing ½O2  ¼ ðpXO2 =RTÞ and for ½Rz O2 =½Rz  ¼ 1 : P¼

RT XO2 Kc

ð19Þ

Table 2 Controlling reaction steps of self-ignition mechanism of paraffinic hydrocarbon in the different temperature ranges Text Ref. number Hydrocarbons [17] [19] [21] [24] [28] [29]

Rz þ O2 , Rz O2 Rz þ O2 ) R0 þ HO2 Rz ) R0 þ R00 H2/O2 system H2 O2 þ M ) 2OH þ M H þ O2 þ M ) HO2 þ M H þ O2 ) OH þ O

Temperature range Low Medium High Medium Medium High

342

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

can be obtained:
 17; 000 RT P ¼ 500 exp 2 atm T XO2

ð23Þ

On a ðP; TÞ plot, Eq. (23) shows the same trend as the ceiling temperature. For T and P at the left side of the curves the reaction (20) is favored whereas for P and T at the right side of the curves fuel consumption occurs preferentially through reaction (21). When reaction (20) prevails, the branching mechanism, i.e. fuel self-ignition, is due to the production and destruction of H2 O2 that leads to the formation of OH radicals [53,54]: H2 O2 þ M ) 2OH þ M

Fig. 12. Ceiling temperature curves (Eq. (19)) on a ðP; TÞ plot at different XO2 :

The curves representative of Eq. (19) computed for Kc ¼ 9:8 £ 1024 expð17; 080=TÞ cm3 mol21 [52] (at different values of oxygen molar fraction) appear in Fig. 12. For each XO2 value, the conditions represented at the left side of the ceiling curve favor the alkyl-peroxyde formation. The pressure and temperature values to the right of the ceiling curve shift the reaction (17) toward the reactants. A decrease of XO2 makes the curves shift toward left, reducing the temperature and pressure range where the low temperature mechanism occurs. For P ¼ 1 atm, the ceiling temperature decreases by 100 K as XO2 decreases from 0.2 to 0.02, passing from 860 to 760 K. The same consideration applies for the pressure. Tceiling increases with pressure as expected from Le Chatelier’s principle, because a pressure increase encourages reactions occurring with a mole number reduction. In the temperature and pressure domain to the right of the ceiling curve the process evolution is due to the middle and high temperature mechanism. For an intermediate range of values the fuel oxidation occurs through the dehydrogenation reaction of the alkyl radical: Rz þ O2 ) R0 þ HO2

ð24Þ

The H2O2 formation is very slow and the dissociation reaction needs an induction time t that depends on temperature and pressure. The relation among t; T and P can be obtained from the rate of the reaction (24) [53]: d½H2 O2  ¼ 2k24 ½H2 O2 ½M dt

ð25Þ

where ½M is the total molar concentration. The dissociation characteristic time t can be expressed as [H2O2]/(d[H2O2]/ dt) and can be obtained from relation (25):

t ¼ ½H2 O2 =

d½H2 O2  expðE=RTÞ ¼ 1=½Mk ¼ ½M21 dt A

ð26Þ

For k24 ¼ 1:2 £ 1017 expð245; 500=RTÞ cm3 mol21 s21 and ½M ¼ P=RT : RT

P¼t 8:3 £

10218

exp




22; 750 T

 atm

ð27Þ

This expression has been shown in Fig. 13 for different induction times. Also in this case the curves show the temperature and pressure ranges where reaction (24) occurs for a fixed induction time. The dissociation reaction is

ð20Þ

At higher temperatures, this reaction is in competition with the unimolecular scission of the alkyl radical: Rz ) R0 þ R00

ð21Þ

The range of T and P where either reaction (20) or reaction (21) prevails can be individuated from the following relation: k20 XO2 r20 k ½O  ¼1 ¼ 20 2 ¼ k21 r21 k21 RT

ð22Þ

By considering k20 ¼ 1£1014 expð225;000=TÞ cm3 mol21 s21 and k21 ¼ 5 £ 1016 expð242; 000=TÞ s21 [52] from Eq. (22)

Fig. 13. Temperature and pressure range for H2 O2 dissociation for different residence times.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

343

This expression, along with the others obtained in this section can be used to build a map on the ðP; TÞ plane that allows for the identification of the prevailing reaction pathway for any fixed initial condition. In Fig. 14 the map obtained for XO2 ¼ 0:21 and t ¼ 0:1 s has been reported. In this diagram the reactions representative of the prevailing kinetic mechanism have been reported at the left and right side of curves to which they refer. In this case four different regions can be recognized. In the first zone, at the left of the ceiling curve, fuel self-ignition occurs through the low temperature mechanism. The passage from low to high temperature kinetics is represented by the zone among

the ceiling curve, the curve representative of H2O2 dissociation and the dashed curve related to Eq. (31). In this region branching occurs through the formation and the destruction of H2O2. In the area on the right of the diagram, below the dashed line the high temperature branching occurs, with the formation of O and OH. Moreover, on the same diagram a region (the grey zone) where no reaction branching occurs can be identified. In this zone fuel selfignition does not occur. In order to evaluate the influence of XO2 and t on the extension of these regions, same diagrams of Fig. 14 have been obtained for different values of these two parameters and have been shown in Figs. 15 and 16. By comparing the maps obtained for the different initial conditions it can be pointed out that the oxidizer dilution makes the grey zone, where no self-ignition occurs, extend. However, this extension of low reactivity zone can be counterbalanced by the effect of the residence times. Mild Combustion relies on fuel self-ignition that allows longer residence time with respect to traditional flame structures. Because of these longer residence times, dissociation of hydroperoxyde and low reactivity zone are reduced also in diluted condition. This is more evident at high pressure where low reactivity zone can completely disappear and the change in self-ignition mechanism can occur without discontinuity. It is also interesting to note that in this temperature range the reaction network can be quite complex. On the left of the dashed curves hydrogen abstraction occurs on fuel molecule whereas on the right (i.e. at higher temperature) hydrogen abstraction involves small pyrolytic fragments of fuel molecules. This means that fuel evolution is characterized by carbon enrichment. In the second case, a certain fuel amount is available for oxygen addition before that dehydrogenation is completed. Therefore, dilution allows for the extension of this range. In fact, the dashed line shifts toward lower temperature with dilution increase.

Fig. 14. ðP; TÞ map of paraffinic fuel self-ignition mechanism for undiluted conditions ðXO2 ¼ 0:21Þ and t ¼ 0:1 s:

Fig. 15. ðP; TÞ map of paraffinic fuel self-ignition mechanism for diluted conditions ðXO2 ¼ 0:05Þ and t ¼ 0:1 s:

favored for T and P to the right of the curves. Moreover, the curves shift to the right when the induction time decreases. This means that the lower the induction time, the higher the temperature and pressure needed for fuel self-ignition through H2 O2 dissociation. The H2 O2 formation passes through HO2 production that, in the high temperature range (T . 900 K), occurs according to the reaction: H þ O2 þ M ! HO2 þ M

ð28Þ

This reaction is in competition with the branching reaction: H þ O2 ! OH þ O

ð29Þ

Following the same approach used for the ceiling temperature curves, it is possible to identify on the ðP; TÞ plane the zones where prevails either reaction (28) or reaction (29): r29 k29 k RT ¼1 ¼ 29 ¼ r28 k28 ½M k28 p

ð30Þ

For k28 ¼ 2:3 £ 1018 T 0:8 cm6 mol22 s 21 and k29 ¼ 2:00 £ 1014 expð28455=TÞ cm3 mol21 s21 the Eq. (29) becomes [52]:
 28455 RT atm ð31Þ P ¼ 8:7 £ 1025 T 0:8 exp T

344

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

Fig. 16. (a) ðP; TÞ map of paraffinic fuel self-ignition mechanism for undiluted conditions ðXO2 ¼ 0:21Þ and ðt ¼ 0:01 s; (b) ðP; TÞ map of paraffinic fuel self-ignition mechanism for diluted conditions ðXO2 ¼ 0:05Þ and t ¼ 0:01 s:

2.5. Features of kinetic controlled processes In Section 2.4 the general features of the fuel selfignition controlling steps have been analyzed by considering the different pathways active in the different temperature range. A general discussion on the evolution of the oxidation process after ignition is even more complex because of the lack of experimental and numerical data in mild conditions. Very few detailed kinetic studies have been devoted to diluted combustion processes [12,13]. de Joannon et al. [12,13] showed by means of numerical analysis of a WSR that the oxidation of methane can develop in mild conditions when rich and diluted feed composition are considered. Only these conditions allow for a unique steady state solution, avoiding problems related to ignition and extinction phenomena due to steady-state multiplicity [12,13,41]. Therefore, the authors considered an XO2 ¼ 0:05 and a C/O equal to 1 as representative of diluted combustion in order to study the influence of working temperature ðTWSR Þ on product distribution. They identified three reaction regimes on the basis of product distribution. It was shown that in the ‘oxidation’ regime, occurring for TWSR lower than 1300 K, methane is principally converted to CO and H2O. The temperature range between 1300 and 1700 K delimitates the ‘recombination’ regime where C2H2 and H2 are the main reaction products along with H2O and CO. The last regime, named ‘pyrolytic’, is placed at TWSR higher than 1700 K and yields the complete carbon and hydrogen conversion to CO and H2. A similar product distribution was experimentally confirmed by Weber et al. [16] in the intermediate range of temperature. They measured high concentration of CO and H2 in High Temperature Air Combustion facility with natural gas. The net rate of production and yield of each species containing carbon atoms, computed in the same boundary condition, were considered by the same authors [55] in order

to search for the reaction pathways of methane for different working temperatures. For TWSR ¼ 1013 K the carbon conversion is very low (in the order of 0.1) and only one of the possible reaction routes is active. This channel passes through the formation of methyl radical that, in turn, oxidizes to methoxy radical and formaldehyde. Then, dehydrogenation reactions lead to the formation of both CO2 and CO, which is the main carbon product obtained in these conditions. An increase of TWSR up to about 1400 K sets in a second kinetic path that becomes the main reaction net. Also in this case the methane conversion passes through the methyl radicals that, in turn, produce ethane by means of a recombination route. A sequence of dehydrogenation and oxidation reactions transforms most of the converted carbon in CO, without any substantial accumulation of intermediate species, like ethylene or acetylene. A TWSR of about 1700 K corresponds to a sizeable increase of carbon conversion. As for the previous condition, this temperature lies in a range that promotes the formation of C(2) species but, in the present case, the production of unsaturated hydrocarbons is favored with respect to recombination products, like C2H6 and C2H5. Moreover, the carbon yield of C2H2 increases up to values comparable with carbon yield of CO. It is worthwhile stressing that CO is mainly produced through this route because the direct oxidation channel, is very slow. This kinetic pathway is very active at higher temperature too, as shown by data computed for TWSR ¼ 2068 K. The carbon yield to C2H2 decreases because the oxidation of C2H2 to CO via HCCO is favored at this temperature. The very complex network of reactions and the great number of species involved in oxidation processes make the choice of diluent composition crucial. It can alter the process not only because it can affect the heat capacity of the system but also because it can contribute in the chemical

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

reaction, thus modifying the product distribution. This occurs when the diluent contains any species that is at the same time a reaction product. This is the case in most diluted combustion applications where internal or external gas recirculation are adopted in order to achieve the required dilution level. In this case, it can be supposed that CO2 and H2O are the main species in the recirculated flow. The effect of CO2-content was also analyzed by Cavaliere and de Joannon [55] for the WSR configuration and the results are summarized in Fig. 17. These data have been computed for XO2 ¼ 0:05 and C=O ¼ 1 by changing XCO2 from zero up to 0.85, the remaining part of the diluent being nitrogen. In Fig. 17a TWSR has been plotted versus CO2 molar fraction ðXCO2 Þ for three inlet temperatures, of 1400, 1800 and 1900 K, respectively. For all the cases, the working reactor temperature decreases by increasing the CO2-content. In fact, for Tin ¼ 1400 K the TWSR changes from 1600 K at XCO2 ¼ 0 to 1520 K for XCO2 ¼ 0:85: Similar behavior is shown at higher Tin . TWSR profiles computed for Tin ¼ 1800 and 1900 K decrease while XCO2 increases. In this temperature range this effect is stressed by

345

the fact that the overall oxidation process from exothermic becomes endothermic. In both cases the increase of CO2 content reduces the TWSR down to values lower than inlet temperature. This feature is evidenced in Fig. 17a by the intersection of TWSR curves with isothermal lines at Tin : The XCO2 at which the skip from exothermic to endothermic behavior occurs, decreases by increasing the inlet temperature. Although CO2 addition can influence the process from different points of view, the working temperature decrease has also to be ascribed to a change in the heat capacity of the reacting mixture. A more significant effect of CO2 enrichment can be pointed out by analyzing the carbon yield of C2H2 computed in the same conditions and reported as a function of XCO2 in Fig. 17b for three values of inlet temperature. The C2H2 profiles evidence an intermediate temperature range for which the C2H2 formation is favored. For lower temperatures, C2H2 is produced in small quantities whereas for higher temperatures it is formed but rapidly oxidized to CO. In all the cases, carbon yield to C2H2 significantly decreases by adding CO2 in the diluent and a temperature increase augments this behavior. In fact, at Tin ¼ 1400 K C2H2 yield is reduced by about 40% changing XCO2 from 0 to 0.85, whereas at Tin ¼ 1900 K, it is reduced by about 100% for the same XCO2 variation. Such a strong influence of CO2 concentration should not be related to the temperature variation. If this were the case, the C2H2 yield computed for Tin ¼ 1800 K and XCO2 ¼ 0 would be equal to the one obtained for Tin ¼ 1900 K and XCO2 ¼ 0:2: In contrast, the change of product distribution is due to the influence of CO2 concentration on chemical kinetics involved in the process. The introduction of CO2 as diluting species improves the formation of CO via CO2 thermal dissociation. This leads to the production of O atoms that, in turn, are involved in the reaction network. In the temperature range where recombination or pyrolytic reaction channels set in, the O atoms are principally consumed by C2H2 in order to form HCCO. Therefore, the presence of CO2, enhances the C2H2 consumption, by producing O atoms in the endothermic thermal CO2 scission. Very similar considerations apply for the effects of H2O as diluent. 2.6. Luminous emission and radiative and convective heat exchange

Fig. 17. (a) Working temperature of a WSR and (b) carbon yield to C2H2 as a function of CO2 content in diluent.

One of the most directly evident features of Mild Combustion is the change of luminous emission characteristics in the visible region of reaction zones as compared to traditional combustion processes [56– 58]. By decreasing oxygen concentration and increasing preheating temperature the flame volume increases whereas the flame luminous emission decreases, both for concentration and chemical effects.

346

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

In high temperature diluted methane flames the flame color changes from yellow to blue by changing the oxygen concentration from 21 to 8% whereas the flame becomes colorless for oxygen concentration lower than 5% [7]. The change of flame color in the case of methane was discussed by de Joannon et al. [12]. By means of numerical analysis, they showed that the ‘kinetic’ fluxes of methane oxidation yield to the emitting radicals, like CH and C2, only at relatively high temperature. In diluted rich conditions, typical of first stage of Mild Combustion, the maximum temperature is relatively low and this is particularly suitable for pyrolytic or reforming stages leading to H2 and CO. Even very high temperatures in the oxidative region are not capable of producing CH and C2, which are among the main emitting species in traditional flames. Different comments apply for different fuels. It has been shown that very peculiar greenish flames are obtained in high temperature diluted combustion regimes when propane is used as fuel [7]. In this case a blue-greenish or green flame can be detected for oxygen concentration lower than 5%. This incurs a further analysis of fuel property influences on these processes with particular emphasis on the chemical kinetics starting from the well documented presence of the green emitting C2 itself. An example of such behavior was shown in Fig. 18 where photographs of a propane flame at 1400 K were reported for different oxygen levels [7]. They were obtained in a regenerative burner where the fuel was injected in a direction normal to the heated air flow as shown in Fig. 18. Some studies show that emission features also depend on the dilution agent [59]. For example, propane becomes less green emitting when nitrogen is substituted with CO2, and C2 bands seem negligible even with an oxygen molar fraction around 0.1. As discussed in the previous section, the increase of CO2 concentration enhances acetylene oxidation, which competes with its dehydrogenation; this is the main route for C2 radical formation. In synthesis, the different behavior of methane and propane under mild

Fig. 18. Photographs and green/yellow luminous emission intensity profiles of a propane flame at 1400 K for different oxygen levels (after Gupta [7]).

conditions with different dilution species should be exploited to study the characteristics of heavier paraffin fuels because part of this fuel is expected to undergo consecutive scissions, which reduces the chemistry to that of species like methane, ethane and propane. In Mild Combustion processes radiative heat transfer can be significantly different from that occurring during traditional combustion processes when the dilution is due to inert streams containing carbon dioxide and water [57,60– 62]. In fact, these species increase the infrared radiative flux [60– 64]. Therefore, in diluted conditions radiative fluxes are higher in the first oxidation zone with respect to the undiluted case because both the emitter concentration and temperature are higher. In contrast, in the post-combustion zone, the temperatures are relatively low and the diluent concentration is nearly the same as that obtained in undiluted conditions. This condition occurs when the dilution is obtained by means of flue gas recirculation because the CO2 and H2O produced are in the same concentration as in the recirculated flow. For the same chemical composition, the lower residence times in the case of Mild Combustion lead to a decrease of total radiative fluxes. Summarizing, in the case of Mild Combustion radiative fluxes are higher in the first zone of the combustion chamber and lower in the second with respect to traditional combustion processes. This effect contributes to the homogeneity of the thermal field in Mild Combustion [24,39,60– 62,65].

3. Basic aspects related to technological applications with the efficiency, reliability, economic benefit It is not easy to separate the technological applications related to efficiency, reliability and economics from those related to environmental benefits. The categorization, here reported, is used only to show some of the potentials of Mild Combustion. It is straightforward to note that benefits from both sides are expected for each example. It is worthwhile to stress that the conveniences mentioned in this section are mainly based on two characteristics. As discussed in Section 2.3 the first is that the maximum temperature cannot be higher than that calculated for the maximum level of the oxidation. The second characteristic is that Mild Combustion is a nearly homogeneous process and the variance of any thermodynamic variable in the process is much lower than would occur in traditional ones [22,25,39]. The highest minimum temperature and the lowest maximum temperature are fixed in the process by definition. Furthermore, the longer autoignition delay time relative to that in systems with internally recirculated flame stabilization allows for a more effective mixing between inlet and outlet species, because

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

the mixing time becomes comparable or lower that the selfignition delay time. It is also of interest to stress that, in the evaluation of Mild Combustion application feasibility, economics is tightly linked to efficiency and reliability. It is mentioned explicitly because it is possible that a process may become convenient because of the combination of several factors. For instance, the required conditions for such a process could result in a not advantageous economic analysis that, however, could be overcome by the positive effects on efficiency. 3.1. High pressure devices The classification of Mild Combustion devices in high and low pressure systems is not only academic, but relies on the fact that at high pressure self-ignition of many fossil fuels is easier [66 – 69]. Ignition delay is shorter than in atmospheric reactors for different oxygen concentrations and ignition can be achieved for much lower temperature. This, in turn, makes it possible to design metal combustion chambers that can contain high pressure. The classical example is diesel combustion, which is made feasible because of high compression ratio. This process cannot be ascribed to Mild Combustion, however, because it involves undiluted oxidation, which produces temperatures higher than 2000 K. Similar comments can be made regarding the so called HCCI systems because the self-ignition relies on the same mechanism as the diesel, but the maximum temperature is kept low by the high level of air excess. This system will be dealt with in the Section 3.1.2. GT combustion has some affinity with Mild Combustion as well, because it has a maximum temperature limitation due to material resistance, but it differs in that self-ignition is purposely avoided and hot spots at very high temperature cannot be disregarded because of stratification of the air fuel mixture even in lean premixed processes. This latter situation is an issue, which occurs more from the step by step evolution of classical GTs rather than from a conceptual limitation. GT combustion is an ideal candidate for Mild Combustion due to its temperature range. The main reason for this apparent contradiction is the fact that self-ignition is elusive in respect to oxidizer dilution and it is fuel dependent, making such combustion in GTs neither reliable nor flexible. Therefore, evaluating possible future applications of Mild Combustion must consider the main features of the autoignition process in diluted conditions. This is done in the following section. In particular the experimental and modeled characteristics of the self-ignition are analyzed with the help of some plots reported in the literature for such conditions. 3.1.1. High pressure diluted self-ignition The study of autoignition is usually performed by means of shock tube or rapid compression machine. For heavy

347

molecular mass fuel the papers that report such data are very few. For instance, some works have been devoted to heptane/air [66,68] and decane/air [69] mixtures. Unfortunately, these works do not deal with systematic investigation on all the parameters in wide value ranges. In particular the oxygen dependence has not been considered explicitly, so that direct application to Mild Combustion is not possible. On the other side, interesting assessment of dilution can be derived from some more practical studies, which have been directly related to diesel [70 – 72] or GT engines [73]. Unfortunately, also in this case the data are elusive or report different oxygen dependences, which entail controlling chemical steps ranging from unimolecular dissociation to termolecular recombination. The advantage of these studies are that they are relative to conditions in which the reactants (fuel and oxidizer) are separated at the inlet boundary and are mixed in the control volume itself. Therefore, they are stratified mixtures which cover the whole mixture fraction spectrum and include the mixture with the shortest autoignition delay. In other words, they mimic the first phase of real Mild Combustion. Of course, self-ignition in stoichiometric conditions is the reference around which fuel rich or lean conditions may play a tuning role. Finally, autoignition in diffusion controlled mixing conditions is easy to create, because it can rely on fast liquid injection in a high temperature, high pressure, diluted environment. The time base of the experiment is the start of the injection time and the physical delay of the ignition delay can be evaluated in single experiment. In particular, it is easy to investigate the oxygen dependence, because this parameter does not significantly affect many physical parameters involved in the ignition delay (such as atomization or evaporation) and it affects only weakly fluid-dynamic quenching. In Fig. 19 the ignition time ðtÞ of a tetradecane pulsed spray measured by de Joannon et al. [45] from 13 up to 40 atm at Tin ¼ 900 K were reported as a function of oxygen molar fraction ðXO2 Þ: On the same diagram are results which were computed from a detailed oxidation mechanism of decane [45,74]. The comparison between experimental and numerical data shows that the model prediction fits quite well the measured ignition delay for pressures of 13 and 20 atm, thus confirming that, in these conditions, fuel oxidation kinetic is the controlling step of the process [67]. At higher pressure the model reproduces the experimentally observed dependence on XO2 but fails in estimating the correct values. The authors attribute this difference to the significant contribution to the measured ignition delay of the time linked to the evolution of physical processes that gain relevance when the pressure is about twice the fuel critical pressure [67]. However, both experimental and numerical data show that the ignition delay related to the chemical kinetics depends on the second power of XO2 at 900 K, thus testifying to an overall reaction order of 2. This reflects a first order reaction with respect both to the oxygen and the fuel.

348

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

At Tin ¼ 1200 K, ignition delay depends on XO212 ; as high temperature kinetics requires. This relation could be explained by considering that the bottleneck of the high temperature mechanism is represented by the initiation reactions, i.e. unimolecular fuel scission leading to C1 –C2 radicals. This behavior is common to most paraffins starting from methane. The even slighter dependence of t on XO2 obtained at 700 K is hardly justified by a simple analysis. It could be supposed that the chain ramification due to dihydroperoxy scission is slow with respect to the oxygen addition to the fuel molecules. The detailed analysis of the complex dependence is reported in Ref. [45].

Fig. 19. Measured and computed ignition delay time of a tetradecane spray versus XO2 at different pressure (after de Joannon et al. [45]).

By using the kinetic model tested at 900 K the authors [45] extend their analysis to different range of temperatures, where different routes of the kinetic mechanism, already discussed in Section 2.4, are active. Therefore, considering a general expression of the type t / XO2n2 ; they computed n for different values of the initial temperature and pressure. The results have been summarized in Fig. 20 where n has been reported versus T and has been compared with few data reported in literature [70,73,75]. In the temperature range between 600 and 700 K, n is constant at about 0.3 whereas it steeply increases at 800 K, reaching a value of about 0.8. It gains its maximum of 2 at 900 K, then it continuously decreases to the value of about 1.

Fig. 20. Exponent n of the power dependence of ignition delay time on XO2 ðt / XO2n2 ) (after de Joannon et al. [45]).

3.1.2. Alternative engines (HCCI) The homogeneous charge, compression ignition (HCCI) engine is based on a concept which can be inserted in the category of Mild Combustion. It aims to yield a homogeneous charge in the combustion chamber at such high compression ratio, that self-ignition should occur homogeneously in the entire chamber. In this respect, as has been mentioned the mixture temperature is higher than the selfignition temperature of the mixture. At the same time, however, the maximum temperature of the charge is kept low by means of super-lean mixture and/or inert dilution. With such condition, very low NOx is produced and relatively high efficiency cycles can be designed. There are also some drawbacks, however. In particular the performance dependence of this process on fuel type [20,76] is high because it is kinetically controlled. In addition, it is possible to increase emission of partial oxidation products [77] because of the influence of the thermal boundary layer on the process. Great advances are expected for this application by the analysis of numerical models [78– 82] and of advanced multidimensional optical techniques [21,22]. Nevertheless, the aforementioned drawbacks and some technological constraints should be overcome, and the complexity of HCCI should be compensated for by much better performance and lower emissions. In this respect it is important to consider the type of diluent in order to adjust the maximum achievable temperature. For example, flue gas recirculation can be a more effective superlean solution because it does not imply perfect homogeneity and it can have some beneficial effect in the soot formation (and oxidation) process. At the same time the influence of the flue gas composition has be evaluated with respect to the autoignition delay time [83 – 85]. Finally, it has be stressed that fine adjustment of oxygen concentration and initial temperature (compression ratio and heat exchange) should be ensured, because, according to the results shown in the previous section, both affect severely the autoignition delay. 3.1.3. Gas turbine The application described here is exploratory since few examples of GT can be reported in which partial fulfillment of the Mild condition occurs. In this case, part of the dilution

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

air in the ground-based turbine is substituted with inert gas (flue gas or water) in such a way that the maximum allowable temperature does not exceed the maximum allowable temperature for mechanical and corrosion resistance of the engine materials. It is well known that hot spots in the higher temperature fluctuation band affect significantly the material characteristics over time [86]. Therefore, the occurrence of narrow variance allows the combustion process in a GT to be designed with higher average outlet temperature and lower maximum fluctuating temperature. This means, in turn, higher thermodynamic efficiency coupled with higher reliability of the turbine material (and minor nitrogen oxides production, as will be presented in Section 3.2). The combination of these factors can be very convenient from an economic point of view, even given more expensive units in the power plant. A further, but not secondary, benefit of Mild Combustion is its noiseless characteristic [22,24]. This makes it a very promising candidate for abatement of the destructive pulsations related to lean premixed systems. Also in this case, it is not completely clear which phenomenon eliminate pulsations, though the phenomenon is well documented, at least at atmospheric pressure [87]. A possible explanation consists in the wide residence time distribution created by Mild Combustion, which produces very distributed oxidation process. Therefore, no resonant zone is occupied, preferentially, by reacting species yielding concentrated heat release. The conceptual feasibility of Mild Combustion in GT can be evaluated by means of the plot of Fig. 21. The autoignition delay time of a stoichiometric mixture of Decane/Air diluted with flue gas is reported versus the mixture temperature. The composition is defined univocally by the mixture temperature. The plot is an example of a possible GT condition where the maximum temperature of the mixture in burning conditions is fixed lower than the maximum allowable temperature at the turbine inlet. In this case a maximum temperature of 1700 K is chosen. The sketch over the plot shows how the oxygen concentration is calculated. For each inlet temperature only a fixed oxygen concentration can give the needed outlet temperature. These values are reported as the gray dashed line in the figure. It is easily seen that the autoignition delay time is a complex function of the inlet temperature, as is consistent with data reported by several authors [66,68,69] showing a negative temperature coefficient in the intermediate temperature range. Therefore, well structured minima and maxima are present at temperatures appropriate for compressor outlet. This means that a well-defined strategy can be exploited for the design of the GT combustor, taking into account that near the peak a temperature fluctuation in either direction induces the same effect in terms of autoignition delay. It is also interesting to note the indifference of extremes (minimum and maximum) positions of the autoignition delay time to the pressure, whereas the excursion between the minimum and the maximum

349

Fig. 21. Computed ignition delay of decane versus dilution.

decreases with the pressure. Therefore higher pressure, around 30 bar, seems to be more appropriate for GT applications. Furthermore, the absolute values of autoignition delay are in this case similar to chemical characteristic times accepted in the present GT design. Very few GT systems have been built with the purpose to satisfy Mild Combustion conditions. Nevertheless some of them deserve mention, because they partially fulfill Mild Combustion conditions and they are of interest to foresee possible potentials and problems. In particular, the four following examples show applications in which external gas recirculation has not been used. Therefore they are also an extension of the application proposed in the first part of this section. The first example is a configuration which has been used in GT for power generation. It is based on Sequential Combustion, which is a multiple stage process where combustion alternates with power extraction [42,88]. A schematic lay out of one such configuration is reported in the web site of Alstom [88] and it is shown in Fig. 22. The first combustion chamber on the left of the figure confines an air diluted system designed for low NOx emission. The central part of the system is constituted by a turbine, which extracts part of the enthalpy content, so that the stream outlet is formed by partially exhausted gas at temperatures comparable to that of standard turbines and with composition with lower oxygen content than these. This stream feeds the last section of the plant depicted in the right side of

350

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

3.2. Low-pressure devices 3.2.1. Heat preheating classification The self-ignition temperature at atmospheric pressure is higher than that at high pressure, so a high level of reliable preheating is needed in order to yield Mild Combustion conditions. Preheating can increase the sensible enthalpy of air and/or fuel by means of exchange of energy in: Fig. 22. Layout of sequential combustion application in Alstom gas turbine [88].

Fig. 22. The aforementioned advantage in terms of high temperature containment easily evaluated. Other characteristics related to noise suppression and performance enhancement can be assessed only from deeper analysis of the related literature. The second example consists in an experimental GT configuration with very high levels of internal recirculation [89] and possible air preheating. This system is created in order to have very low NOx emission, safety and reliability operations, a high level of uniform temperature for low thermal stress on the wall and circumferential pattern factors as well as low heat value fuel combustion. The velocity patterns relative to this configuration show the relevance of the fuel injection location and momentum on the whole pattern itself. In this case, the low oxygen concentration is obtained in the internal recirculation zone just before or after mixing with the main oxidizer stream. In the first case this is another example of Fuel Direct Injection systems [90,91], which will be extensively described in Section 3.2. The third example [92,93] is the system in which GTs are fed by highly preheated air and low heat value fuel. The first condition ensures that the temperature is higher than the autoignition one of the fuel, and the second condition allows reduction in both the film cooling on the wall and the dilution air flow rate, since the low heat value is due to high concentration of diluent in the fuel. Finally, the fourth example are GTs where significant amounts of water are injected in order to lower both the maximum combustion temperatures and the compression work of the working fluid [94 – 96]. Limitations of the maximum allowable water flow rates are reported, because the injection has been performed in combustion processes with recirculation stabilizing mechanisms, which can not be effective for high levels of dilution. This is the very contradiction of the ‘traditional’ combustion, which does not exploit preheating or larger compression ratio in order to overcome the autoignition temperature [22]. Therefore, these systems cannot be included in the Mild Combustion category; they are here reported because they reduce the maximum temperature analogously to Mild Combustion, stressing the fact that it is not sufficient to have simple dilution to reach Mild Combustion. The second requirement related to inlet temperature is needed as well.

a. b. c.

‘heat’ mode ‘heat and mass’ mode ‘work’ mode

The origin of this energy may be recirculation either by the same principal combustion unit or generated by a companion unit in a staged combustion process or by an external unit. The distinction among the three origins of the energy is purely conventional, either because of possible ambiguity in the definition of the terms ‘principal’, ‘companion’ or ‘external’, or because of possible overlapping of different units in complex combined cycles. Fig. 23 is an example of the possible difficulty in the evaluation of the energy origin. It shows, in a very schematic way that the fuel and the oxidizer are introduced in the sub-unit 1p, where they are mixed and heated. The combustion develops in the sub-unit 2p, after autoignition of the mixture, and, finally, the flue gas may undergo partial cooling and expansion in the sub-unit 3p. The heating in the sub-unit 1p may be accomplished according to the three logical – physical aforementioned channels which are labeled with letters a, b and c. These originate from a ‘recirculation’ channel or ‘staging’ channel, which, in turn, consists in part of the energy produced by means of the sub-units 1c, 2c, 3c of the companion unit. A more specific example of ambiguity may be the compression heating in alternative internal combustion engines. In a single-cylinder engine the compression

Fig. 23. Possible scheme of principal (‘p’) and companion (‘c’) unit involved in preheating of reactants.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

351

work derives from the previous cycles of the same principal unit, whereas in multi-cylinder engines, it comes from other cylinders which are arbitrary considered ‘principal’ or ‘companion’ units. 3.2.2. Heat mode (recirculation and staging) The simple heating without mass exchange is usually performed by means of ‘traditional’ heat exchangers or through an intermediate heating of a solid support. The second procedure is preferable at conditions in which high reactant temperature and/or high efficiency recovery of waste heat [97,98] have to be reached. In this case porous inert material (PIM) [99 – 102] are used, which are temperature resistant due to the use of special alloys or ceramic [103] and which have high efficiency due to the high interfacial surface density of porous materials. The energy recovery can be obtained in alternative flowing of flue gas and reactants through one of the two regenerative units as it is shown in Fig. 24. It is difficult to classify this type of reactant heating as due to recirculation or to staging because the whole process develops on two combustion units, with a couple of separated regenerators [104] or in one compact device [22]. These types of regenerative burners are of particular interest in industrial furnaces because of the high enthalpy content of their exhaust gases. Another use of PIM is in the system, in which premixed combustion occurs inside the material itself [5,99]. It has been shown that part of the heat released in the oxidative region (Fig. 25) is transferred to the unreacted mixture passing through a conduction channel inside the solid material or a radiative one inside the pores. The process is outlined by the scheme and the temperature space plot of Fig. 25. It is very similar to

Fig. 24. Lay-out of regenerative systems for energy recovery from flue gases.

Fig. 25. Lay-out of systems for energy recovery by means of porous media.

the homogeneous deflagration but it differs mainly from this latter because it reaches a temperature higher than the maximum one evaluated by maximum release of reactive enthalpy. For this reason the process is also named ‘superadiabatic combustion’ [99] and it is useful in enlarging flammability limits of gaseous [105] and liquid fuels [99]. The applications shown in this section are only those where the maximum temperature constraint is ensured by internal recirculation of flue gas. This section is devoted only to the heat recirculation without external mass recirculation. In this case, the need of cooling the internally recirculated flue gas has to be stressed. In fact, only for this application is possible to envisage a decrease of the maximum temperature. It is trivial to add that adiabatic furnaces, fed with preheated stoichiometric reactants, reach higher temperature than that occur without preheating. The only possible exception consists in very lean processes analogous to those created in rich-quench-lean (RQL) GTs. The difference is that in many atmospheric practical devices the fuel is directly injected in the recirculated flue gas so that in this first stage the process can develop in vitiated-rich conditions. The temperature does not increase significantly or decreases because the conditions favor reforming or pyrolysis of the fuel. Only in the second stage is the oxidation completed when the superstoichiometric oxidizer stream is fast-mixed with the partially oxidized fuel. The two-stage process can be named vitiated-rich/quench lean combustion analogously to what it has been done in the RQL process. As it has been also mentioned before, this process is also named direct fuel injection in the first experiments made by Tokyo Gas [106]. Its first applications have been reported to be in the field of steel treating. In fact, its development has been guided by the constraints imposed by the boundary conditions involved in this technology. For instance, the most reliable applications exploit recuperative

352

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

devices, which fit properly with material heating and treating. A list of the applications are easily found in the specialized literature on the subject [1– 3] so that it is not shown here. In particular, the very leading references to be consulted as introduction to the field are [3– 5,22,39]. In principle, the main advantage is the relatively uniform temperature which flows on the material to be treated. This results in better quality materials and surface properties. Even though higher efficiency has been shown in several processes [107], the quality is the premium target of this process. The economics associated with Mild Combustion are more controversial for power generation boilers because there are no straightforward correlations between efficiency, reliability, maintenance of the plants, and the use of high temperature combustion. For example, sometimes vitiated air has to be used for pollutant abatement (as it will be discussed in Section 3.2.3) or for heat recovery in a more efficient cycle, which exploits diluted air only because it is the only feasible option. Recovery boilers, hot wind boxes, sequential combustion are a few examples of this use. One example of flameless combustion is here shown for completeness, but these applications still must show their advantage. The example is a 48 MWth facility [12], used to run tests of natural gas combustion with different fluid-dynamic configurations. The fuel volume flow rate was constant at 3000 Nm3/h. The air-flow, at Tin ¼ 600 K, was fed in such a way that the oxygen in the flue gas was 0.8%. The main burner configuration is outlined schematically on the left side of Fig. 24a. It confines three coaxial fluxes with tailored swirl levels. The fuel is injected through 43 outlet holes, placed on one central spud and grouped in cylindrical rows tilted in the symmetrical plane at 308, 608, 908 respect the cylindrical axis. The dashed lines in the figure represent these orientations. This configuration yields a diffusioncontrolled flame with shape and visible emission of traditional characteristics. The yellow compact region, is placed around the axis and extends over the entire chamber, as the white region in Fig. 26a shows. The second configuration is outlined on the left side of Fig. 26b. The natural gas, depicted with blue color, is injected through eight spuds placed on the cylindrical surface. The distance of the spud head from the symmetry axis and from the burner outlet have been determined by cold fluid-dynamic studies on the external side of the most external air jets on the border of the outward reverse flow. Four nozzle holes on each spud were oriented in such a way that they take into account the enlargement and the rotation of the external swirl. The most striking characteristic of the new combustion regime is revealed by the visual observation of the combustion chamber. The picture of Fig. 26b shows details of the internal part of the combustion chamber. It is possible to observe the hopper, which is on the bottom of the chamber and the walls, which are slightly emitting due to

Fig. 26. (a) Traditional natural gas flame; (b) natural gas flameless combustion (after de Joannon et al. [12]).

the refractory coating. Visible emission is negligible and, according to the narrowest dynamic range of the detectiondigitization device, it is smaller than any part of the previous flames by two order of magnitudes. This means that not only soot, but also the bluish emission due to CH and to CO2 decaying from the triplet to the singlet electronic state chemiluminescence are completely absent. Fuel injection in intermediate positions did not result in this dramatic change, whereas air vitiation with exhaust gas recirculation (EGR) did not affect the second colorless combustion. Heat-recirculating microscale burners for miniaturized power-generating device is an innovative field where heat recovery from flue gases is used [108 – 111]. The characteristic dimension of such systems is comparable or lower than the quenching distance at normal combustion conditions thus requiring the preheating of inlet flow at a temperature higher than the autoignition temperature in order to make the combustion proceed. This is obtained using a counter current heat-recirculating combustors [109,110] where the reactants are pre-heated using the sensible enthalpy of flue gases, typically choosing a ‘Swiss roll’ burner configuration [108,109,111], schematically reported in Fig. 27. 3.2.3. Heat and mass mode (recirculation and staging) The reactants temperatures can be increased by mixing the reactant with a partially cooled flue gas stream. As it has been quoted in the comment of Fig. 23, this flue gas can originate from the same or other units. In the first case

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

Fig. 27. Swiss roll burner configuration for microscale burner applications [109].

the energy transfer is named EGR. It is of interest to note that the mass recirculation can be used both to increase the reactant temperature and to decrease the outlet temperature. Unfortunately, examples can be given only for applications where the mass recirculation has been exploited to lower the adiabatic flame temperature, while not increasing the reactant one. In this case, combustion with EGR cannot be considered a very Mild Combustion, but it suggests potential benefits and problems related to full-scale applications. In particular, the convenience of EGR Combustion in domestic heating systems stresses the fact that these systems can be made completely free from soot. What is relevant here is that in this case very narrow

353

passages can be used in the heat exchanger since no fouling is produced due to the combustion process and the walls do not need to be cooled in the most convoluted parts. Even though the recirculation level is not higher than 30% of the total air diluted stream, this process demonstrates the main features of Mild Combustion since it limits the maximum allowable temperature. The main difference is that dilution limits the stabilization range of the burner, whereas mild oxidation is similar to high temperature combustion since it overcomes the autoignition temperature. An example of such technology is shown in Fig. 28, which shows a schematic layout of the system designed during an Italian-Swiss cooperation in the early 1970s. In the higher part of the figure the scheme is composed by two feeding lines, one for the direct breathed air and the fuel oil. A third line is drawn for the recirculated flue gas, which is aspirated by the stack of the systems, after they have left their thermal enthalpy to the water loop [112]. The recirculation level affects the oxygen concentration and the temperature of the main oxidizer stream. Differently from the high temperature systems, problems related to acid condensation may occur starting from the mixer of the recirculated and the fresh stream. In the lower part of the figure the cross-sections of the heat exchanger are shown in order to show the complex flue gas passages. The flue gas flows also in the section, shown in the figure, so that the residence time is made longer with relatively short combustor depth. The flow pattern is made

Fig. 28. Example of domestic heating system with flue gas recirculation [112].

354

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

more complex by the injection of part of the burnt gas perpendicularly to the main streams. This induces higher level of stretching in the main flow and along the solid walls, so that an increase of the convective coefficient is obtained. It is worthwhile to stress again that the relatively complex pattern can be used because of the lack of any solid particulate or obstructing deposit.

4. Basic aspects related to technological application with environmental benefits. Clean – cleaning – clearing combustion Most of the combustion systems consist of three main sections, which may be integrated in a single plant or may be divided in separated processes. They are devoted to fuel treatment, the combustion process itself, and pollutant abatement. The first of them is usually spatially separated from the other two, since the fuel may be carried in condensed form, solid or liquid, with particular chemical and physical properties suitable for different users. On the opposite end, the flue gas undergoes processes to convert pollutants into less noxious or more separable species in units or plants, which are nearly always located in the neighborhood of the combustion unit. The abatement may also be completed in not-local units when it consists also of transformation of the condensed or condensable part of the flue stream. It is interesting to analyze the entire cycle of fuel oxidation with the parallel evolution of the minor pollutant species in relation to the possible coupling of the three main processes. In particular, this is crucial when the central combustion process proceeds under new thermo-chemical conditions, because these may determine useful interactions with the preceding and following processes. The analysis is extended to three categories of lowpollutant combustion strategies which are defined in this paper as ‘clean combustion’, ‘cleaning combustion’ and clearing combustion. The first one refers to a ‘classic’ condition in which the three aforementioned units ideally process the same fuel with the only final product being CO2 and H2O. Mild Combustion is expected to be a clean combustion process. One objective of the present review is to support such a statement and to show how in this case part of the fuel treatment and pollutant abatement required in some ‘traditional’ processes is reduced. It is worthwhile to stress that ‘clean’ refers only to systems fed by pure organic fuel and pure oxidizer (oxygen and nitrogen). Other inorganic species are disregarded. The same is true for the term cleaning. It is used in this review with the meaning of abatement of all possible organic pollutants (solid, liquid as well as gaseous) and the pollutant species which include also nitrogen atoms, like nitrogen oxides and cyanidric acid. More specifically ‘cleaning combustion’ refers to a process (schematized in Fig. 29) inside the solid perimeter.

Fig. 29. Logical layout of cleaning combustion.

The pollutants, present in the oxidizer or the oxidizer diluent streams, are destroyed rather than abated in separated units. These are represented in the sketch with isolated blocks on the left of the clean combustion unit. The convenience of such a process consists in the possibility of getting more power from a heavily polluting combustion unit. Finally clearing combustion refers to a process which is capable of reducing the need for an abatement unit of inorganic pollutants. In Fig. 30 this system is sketched with a cleaning combustion unit in which the last abatement unit is partially included in the external solid thick line. This potential has been discussed in the appendix section and is only mentioned here because of its very speculative nature. It is based on the possibility of combining the stage of fuel purification with oxidation in Mild conditions taking advantage from the partial overlapping of temperature range of such a combustion process with the temperature useful for inorganic metal capturing. For instance, CaCO3 to CaO calcination can take place during Mild Combustion leading to the formation of sorbent particles that chemically or physically adsorb sulphur downstream to the oxidation process. Moreover, sulfur as well as minor inorganic species present in the fuel can be trapped in the carbonaceous particles (such cenospheres and pleosphere) that form from asphaltene fraction during fuel droplet evaporation. 4.1. Soot depression, destruction Plenty of experimental and numerical works have been carried out over the years in order to evaluate the influence of environmental parameters on the different phases of soot production, i.e. inception, surface growth, coagulation and oxidation. Very interesting conclusions concern the effect of temperature and air dilution on the fuel fractional conversion to soot. It is well-known that soot formation is very sensitive to the temperature achieved during fuel conversion. In particular, it has been assessed both in pyrolytic and oxidative conditions that there is an intermediate temperature range where soot production is promoted. This feature has been exemplified by the bell-shaped curves shown in Fig. 31. They represent the soot yield and soot volume fraction as a function of temperature, measured during the pyrolysis of n-heptane (W) and n-hexadecane (O) in a shock

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

355

Fig. 30. Logical layout of clearing combustion.

tube [113,114]. In the case of n-hexadecane pyrolysis, soot is present between about 1700 and 2500 K, reaching a maximum concentration at about 2200 K. This behavior is generally found for most hydrocarbons even though the temperature range of soot presence can slightly shift or change in its extent, as is shown in Fig. 31 by the curve relative to n-heptane. Moreover, the comparison between the latter curve with the one obtained from nhepatne with a small amount of oxygen (X) shows that the curve shape does not substantially change even though the maximum of soot concentration can occur at lower temperature values [114]. This sensitivity of soot formation to the temperature is not related to the system configuration. In fact, the same trend has been obtained from premixed flames, as shown in Fig. 32 where normalized soot volume fractions for C2H4/ air, C2H2/air and C6H6/air flames with a C/O ratios of about 0.7 have been shown as a function of temperature [115]. It is

noteworthy that an increase of C/O makes the temperature range of soot formation extends due to the shift of upper limit toward higher values whereas the low temperature limit remains nearly constant [115]. A reduction of soot production with temperature decrease has been also demonstrated in diffusion flames [116– 120]. As a consideration on flame stabilization, in this case, the extinction velocity sharply decreases with oxygen concentration until it becomes negligible for XO2 ¼ 0:16 (Fig. 33) [121]. A traditional diffusion flame can not develop in mild conditions and therefore, soot cannot be a product of diffusive flame structure. However, it was pointed out that soot formation is very sensitive to the flame temperature. Such an effect is shown in Fig. 34 where axial profiles of soot volume fraction obtained in not-diluted (X) and diluted (O) conditions from an acetylene laminar co-flow diffusion flame have been shown [119]. The latter conditions have

Fig. 31. Soot volume fraction measured in pyrolytic and notpyrolytic condition in a shock tube (adapted from Refs. [113,114]).

Fig. 32. Soot volume fraction measured in premixed flames (adapted from Ref. [115]).

356

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

heights above the burner because the temperature decrease depresses the reactions leading to soot formation. Consequently, longer residence times are required for soot inception and less soot is formed. The maximum of soot volume fraction is sensibly lower in the case of diluted conditions with respect to the not-diluted ones, at least in correspondence of heights where soot formation is predominant and soot oxidation is negligible [116 – 120]. When soot oxidation begins, i.e. after the maximum, the situation changes completely. In this case, the soot volume fraction under diluted condition is higher than those measured in not-diluted conditions. This behavior results from the joint effects of temperature and oxygen concentration. Their decrease not only directly depress soot oxidation but also slows the formation of OH radicals that are the main species responsible for soot oxidation in flame conditions. In practical systems, however, the dilution is performed by means of flue gases that principally contain CO2 and H2O. These species, as products of hydrocarbon oxidation, can be involved in the reaction network thus modifying the species distributions. For instance, an increase of CO2 concentration due to flue gas shifts the equilibrium of the reaction CO þ OH $ CO2 þ H

Fig. 33. Extinction velocity of diffusion flames versus oxygen concentration (after Sjo¨gren [121]).

been achieved by changing the composition of co-flow. In particular, oxygen was partially substituted by N2 thus lowering the O2 concentration from 20 to 15%. By comparing the two curves, two main considerations arise. In diluted conditions the soot inception occurs at higher

towards the reactants [119,120] thus leading to a twofold positive effect with respect to soot reduction. Firstly, the reduction of H radicals depresses soot formation due to the key-role of these radicals in the polymerization reactions yielding to soot nucleation and surface growth. On the other hand, the H concentration decrease is related to an OH concentration increase that yields an increase in the soot oxidation rate. Therefore, the synergistic effect of such kinetic modifications strongly reduces the soot production. The influence of CO2 addition on soot formation in diffusion flames is clearly shown by the soot volume fraction axial profile shown in Fig. 34 relative to a not-diluted condition, where some N2 in the oxidizer has been substituted with CO2 (O). In this case, CO2 is responsible for both the temperature and chemical effects on soot formation. Its higher heat capacity with respect to N2 reduces the flame temperature. This reduction, along with the reduction of the H radical concentration, discussed above, leads to a lower soot formation with respect to the not-diluted diffusion flame with air. After the maximum, when soot oxidation sets in, the enhancing effect of OH increase hinders the slowing effect of the temperature decrease, thus leaving the axial profile unchanged. All these effects are augmented when H2O is considered as diluent. In this case a reduction of H and an increase of OH radicals occur due to a shift of the reaction H2 O þ H $ OH þ H2

Fig. 34. Soot volume fraction measured in diffusion flames in different experimental conditions (adapted from Ref. [119]).

ð32Þ

ð33Þ

toward the products that takes place faster than reaction (32) [119,120]. Moreover, the H2O can also depress soot surface

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

357

growth by deactivating radical sites on particle surfaces [119,120]. An additional effect of temperature decrease is the decrease of soot mean particle radius which is related to reduction of soot inception velocities [119]. This means that oxidation of soot can be favored. In light of the phenomenologies above described, both temperature and oxygen decrease have a positive effect with respect to soot depression and destruction. In this sense, Mild Combustion is a clean and cleaning process. 4.2. NOx depression, destruction Several works exploit the influence of mild conditions on formation of NOx through the different reaction paths [22,121– 123]. It is well known that nitrogen oxides form along three possible paths, namely thermal NOx, fuel-NOx and prompt NOx. The first two mechanisms, which are more efficient in the NOx formation, are depressed by the lowtemperature rich conditions in Mild Combustion. The first mechanism is most effective for high levels of temperature and oxygen, according to the Zel’dovich mechanism, whereas the second one is maximized for high levels of fuel richness and fuel nitrogen concentration. The last mechanism relies on high reactant temperature also and can proceed in very rich conditions only if the temperature is high enough to sustain the process. The same reaction paths, which depress NOx production in fuel-NOx formation are behind the mechanisms which destroy NOx in reburning and selective non-catalytic reduction (SNCR) techniques. These paths can work also in synergistic and non-synergistic advanced reburning techniques [124,125]. Therefore, Mild Combustion is an optimum condition for NOx destruction because this type of reactor requires a very rich condition in the first combustion stage, with a strong reductive environment. The use of fuel with high bound-nitrogen content should also guarantee a rich presence of nitrogen atoms and or cyanidric compounds in the first stage of Mild Combustion. Furthermore SNCR and synergistic advanced reburning requires a relatively narrow temperature range, which is around 1300 K. For instance, in Fig. 35, the complement to the conversion of NO by SNCR in a plug flow reactor (PFR), is reported as a function of the inlet temperature by Rota et al. [126]. The experimental data of Østberg et al [127] follow the model prediction of Rota et al [126]. In the same figure the experimental data from Caton et al. [128], as reported by Xu et al. [124], are also plotted. All of these experimental and predicted data show that a temperature window for conversion optimization is near the aforementioned temperature (1300 K). This means that flue gases have to be kept at this temperature for a relatively long residence time (tenths of second) to make this process efficient. It could be convenient to exploit the device which creates these conditions for additional heat generation possibly linked to reburning conditions too. In this respect Mild Combustion

Fig. 35. NO yield versus temperature obtained by means of selective non catalytic reduction technique.

may be a suitable process because it has a narrow temperature range near the ideal SNCR temperature. 4.3. Synthesis Fig. 36 lists favorableness in the reduction and abatement of soot, NOx and SOx as a function of working temperature. The bar related to each class of process (for instance soot reduction, etc.) is filled according to the legend shown at the top of the figure. In this legend the term ‘favorable’ refers to a temperature range in which the reduction or abatement of the pollutant is improved. Moreover, in the temperature range corresponding to a ‘favorable, rich’ condition the pollutant abatement or reduction needs a C/O ratio higher than the stoichiometric one. In contrast, the term ‘unfavorable’ identifies ranges in which the process conditions enhance pollutant production. The analysis of the first three bars, related to the soot and NOx, shows that low working temperature does not allow the formation of these species and at the same time it improves NOx abatement by means of reburning when rich conditions are considered. In contrast, high temperature is not recommended for both soot and thermal NOx whereas it does not affect the fuel NOx. The abatement of NOx by means of SNCR is applicable in a very narrow range, as it has been underlined in Section 4.2. The fifth line refers to the SO2 removal, discussed in Section A.1 of the appendix. It shows the suitable temperature range for the formation of the sorbent material with high surface density. The temperature refers to the process for CaO preparation and not for the adsorption process itself. This has to take place at much lower temperature, in the exhausted flue stream. At the same time it is clear that Mild Combustion is the only process which could be compatible with CaO formation from calcium acetate [129,130] because the sorbent material does

358

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

Fig. 36. Summary of temperature range favorable to formation/reduction and abatement of soot, NOx and SOx.

not undergo very high temperature heating which is detrimental for the surface annealing and sorption efficiency. The last two lines refer to large carbonaceous particle formation. The fuel evaporation, when it is conditioned by an environmental temperature evolution consistent with a quasi distillative model [131] yields empty carbonaceous spheres, known in the literature with the name of cenosphere [131] or solid spheres, named very recently as pleosphere [132]. Both of them recall that rich conditions increase the efficiency of pyrolytic process, but they evidence too that the more compact material, pleospheres, are formed preferentially in the low temperature range. The driving process for large particle formation is distillation, which is favored by Mild Combustion conditions. Fig. 37 reports a possible simplified lay out of a combustor, which takes into account the aforementioned considerations. In particular the first section consists in a spray pyrolysis unit, in which the maximum temperature is fixed at values considered high in the pollutant abatement technology and low in the combustion field. This temperature can be adjusted on the ground of empirical evidence and according to different fuels. A first tentative value, which can be considered a reference one for heavy fuel oil is around 1500 K. This is lower than the soot formation threshold. It is also suitable for the calcination process for CaO formation as part of sulfur removal. This maximum temperature is consistent with an average value around 1300 K, which is

favorable in NOx abatement through SNRC. The minimum or inlet temperature should be fixed on the grounds of reliability of autoignition, which is fuel dependent. A wide literature is available for the determination of ignition delay at high pressure. In contrast few and sparse papers are available for precise assessment of this quantity [133– 135].

5. Conclusions This review paper collects information which could be useful in understanding the fundamentals and applications of Mild Combustion. The pieces of information in this field are still sparse, because of the recent identification of the process, so that many speculative considerations have been presented in order to make the whole framework more consistent and rich with potential new applications. The main points to be stressed as concluding remarks pertain to two main considerations. The first is that Mild Combustion has to be considered a new combustion regime. It is neither a deflagration nor a detonation nor a diffusion flame. It is a combustion process which is a superdiluted explosion or a continuous auto-ignition/explosion. The fluid-dynamic local conditions and thermodynamic constraints under which Mild Combustion develops are quite straightforward. They have been described in Sections 2.2 and 2.3. The overall fluid-dynamic configurations and the chemical kinetics which are part of this regime are not

Fig. 37. Lay out of the clean-cleaning combustor.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

completely explored. Section 2.4 reports examples of specific conditions, which need to be enlarged to broader categories. In particular, one apparent contradiction deserves further comment. The continuous explosive condition, which Mild Combustion undergoes, allows process design in such a way that internal recirculation is not needed. Nevertheless the most noteworthy applications of this process have been created with so much internal recirculation that highly diluted local conditions are obtained. To eliminate this constraint could be one of the most innovative steps in Mild Combustion system design. The second consideration is that the narrow temperature range, under which the process proceeds, allows design, optimization and adjustment in the process by fine tuning external parameters. These parameters provide controllable shifts of internal parameters in the reactor. In contrast ‘traditional combustion processes’ are difficult to control because they proceed along temperature excursions of thousands of degrees, which allows only broad adjustment of the residence time inside a fixed temperature range. Furthermore, moderate temperature increases keep the maximum allowable temperature under any well-defined value. Combustion processes in mild conditions can be considered similar to a chemical conversion process, in which selectivity can be varied in a controllable way. In this respect the broad literature in fields such as oxidative pyrolysis, oxidative vaporization/gasification, fuel reforming, organic compounds production may suggest chemical physical understanding of the basic processes and further extension of the technological applications presented in this review. We discussed only two application categories, which were kept separated for the sake of classification. They were clean energy conversion and pollutant abatement. It is easy to consider also systems in which these two applications merge in a process which could be named a clean/cleaning/clearing combustion. The peculiarity of this approach consists either in unifying different sections of power production plants into a single unit or in producing such features, that contiguous units can take advantage of conditions established by each other. This is a very important area for probing the feasibility of reliable synergistic plants for polygeneration of different forms of energy and materials with the contemporaneous destruction of pollutants in the diluent streams as well as transformation of the inorganic inclusion in the fuel into proper storable or reusable materials.

Appendix A. Clearing combustion A.1. SOx depression and removal Sulfur is bonded to carbon atoms in the liquid fuel in two categories according to the C– S cleavage resistance.

359

In the first category there is sulfur in fuels, produced by mature crude oil with low sulfur content or transformed in desulfurization processes. This sulfur is characterized by its hindered accessibility because cleavage has affected the original crude oil in the geological transformation or in petroleum treatment [136]. In contrast, immature or highlevel-sulphur content crude oil is characterized by very efficient thermal scission for the weakest C– S bond in sulphides, which are quite dominant in many asphaltene structures, as it is reported by Straus et al. [137]. The authors report that 40% of the sulphur is bound in sulphide for an Athabasca Asphaltene, 60% of which is in the sulphur bridge, holding together core segments. This means that the asphaltene is thermolabile and is characterized by high reactivity. The sulphur counteracts a tendency for growth in molecular size and coking as long as it is present in the asphaltene [137]. As soon as sulphur extraction is attained, a strong tendency toward condensation and solidification is shown by the residual asphaltenic striker. A sulphur removal strategy can be envisaged according to the C– S cleavage resistance and its concentration in the asphaltene/polar fraction. When these quantities are very high, the operative conditions should be optimized for sulphur enrichment in cenospheres. This combustion fraction is relatively easy to separate and therefore it is ideal for carrying the sulphur which is originally present in the fraction which generates cenosphere itself. The same is appropriate in the case where the carbonaceous residue formed by liquid fuel pyrolisis is concentrated in a single compact sphere, which can be defined as a pleosphere [132]. In the case where the sulphur escapes in the gas phase during the pyrolitic-ossidative evolution of the fuel, the only feasible process which can keep the sulphur in a separable condensed phase is its adsorption, and possibly chemoadsorption, on suitable sorbents [129,130,138]. One of the most recent and interesting desulfurization processes is reported by Nimmo et al. [129] and it is schematically sketched in Fig. A1. It is possible to follow the consecutive phase of calcium evolution starting from calcium acetate in water solution. This species is concentrated, saturated and precipitated when the water vaporizes. It forms cenospheres when the solution is spray-pyrolized because the precipitation in a droplet is placed just under the liquid– gas interphase in very fast evaporation processes. Then the acetate is transformed, through acetone release, into carbonate, which is then calcinated to calcium oxide. This process can evolve in a temperature range comparable to that of Mild Combustion. It is straightforward, therefore, to design spray pyrolysis of oil-in-water fuel which partial oxidized the hydrocarbons and produces carbonation/ calcination of the water solution. This combination can produce a significant synergistic interaction because liquid stratification of the organic and water fraction may help in the creation of spherical CaO structure. A cooling stage, also, receives benefits for SO2 adsorption, because the CaO cenospheres are suitably distributed where the oxidation has

360

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

Fig. A1. Evolution for calcium acetate solution to calcinated calcium oxide [129].

taken place and consequently also the SO2 concentration is high. The different processes for sulfur removal, described so far, have been summarized in Fig. A2 where possible solid particles, which can be generated by combustion systems, are shown. The first two, labeled (a) and (b), show in white cenospheres and pleospheres described before. The sulfur inclusions are shown with black dots. The sketch (c) refers to a CaO cenosphere obtained through the pyrolysis of calcium acetate water solutions. It can be injected as spray into Mild Combustion region or it can be injected together with the fuel in oil/water or water/oil emulsion. The left part of the figure is a schematic representation of a single oil in water droplet, which is here shown only as a possible innovative solution. It has never been used but it could be a way of creating CaO cenospheres containing carbonaceous pleospheres. The advantage should consist in dispersing the fuel and the absorbent material in the same region without problem of mixing. As it has been discussed before, both

CaO cenosphere and carbonaceous pleosphere can be produced in mild conditions. A.2. Minor inorganic elements Minor inorganic elements are dispersed in many fossil fuels. The most abundant are vanadium, nickel, magnesium, copper, iron, zinc, lead, arsenic, beryllium, cadmium, chromium, mercury and selenium are present with concentrations orders of magnitude lower than those of the first category. Inorganic compound concentrations correlate well with the boiling temperature of the fuel. For instance, in the higher part of Fig. A3 an ensemble of fuels is represented with a circular set in which sectors are both fraction of fuels with different boiling temperature or distillation fraction of a generic crude oil. A thick solid line is marked at 340 8C because it represents the lowest boiling temperature of residue from an atmospheric distillative process.

Fig. A2. Scheme of possible sulfur removal techniques.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

361

Fig. A3. Fate of some minor organic species during fuel treatment and combustion.

Inside the sectors at higher temperature the mass fraction (mg/g) of the most abundant elements are reported with annular sectors following the schematization of Whitehead [139] and Molero de Blas [140]. For instance vanadium, nickel, sodium and Iron mass fractions, for fuel fraction with boiling ranges between 700 and 800 8C, are on the average

of 160, 50, 25 and 13, respectively. Fuels or crude oil fraction with higher boiling temperatures contain higher concentration of metallic elements. The present paper narrows the analysis of inorganic mild oxidation/pyrolysis to the Vanadium element, because it is often the most abundant, and because it is quite similar to all

362

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

of the earth and transition metals. Inside the sectors at high temperatures Fig. A3 also reports some characteristic features. The central dark part represents an asphalteneagglomerate and empty circles represent the resin-polar fraction, according to the model proposed by Yen [141]. An asphaltene mycelle is dispersed in the polar fraction, which is responsible for the ‘solubilization’ of the heavier part. The evaporation of the lighter resins may yield flocculation and precipitation of the mycelles. Each single asphaltene unit consists of a planar cyclic/aromatic with several inorganic substitutions, which are linked together by bridges of organic and sulphur-based complexes (Fig. A3, inset (a). In this picture it is possible to identify the molecular bonding of Vanadium. It is mainly a ‘metallo-porphyrin’ type, as it is depicted in the inset b. In other words Vanadium is placed in the center of the four pyrrole rings, which form the core of the porphyrin [141]. The very peculiar abundance of the Vanadium in heavy crude oil and its very peculiar bonding type are consistent with its diagenic origin. It comes mainly from the green pigment of plants and algae, i.e. the chlorophyll. “This porphyrin structure loses its Mg atom during deposition. The major change, that porphyrins undergo following burial and diagenesis, with its increasing temperature, is gradual aromatization of green chlorins to red porphyrins. As the porphyrins form, they first chelate nickel, which is replaced later at higher temperature by vanadium [142]. Hydrolysis of the phytol side chain may occur with re-chelation of porphyrins by Ni or V, which may serve to stabilize the structure [143]. Porphyrin content parallels that of resin and asphaltene content of the oils. Highly paraffinic crude oils with low density and a low resin and asphaltene content contain only traces of porphyrins, and hence low V contents.” [144]. This in turn means that the evolution of the liquid fuel is again important in any ‘clearing strategy’, because it involves the metal element too (in this case the vanadium). As already said in Section 4, the fuel evaporation, when it is conditioned by an environmental temperature evolution consistent with a quasi distillative model [131] yields empty carbonaceous spheres, known in the literature with the name of cenosphere [131] or solid spheres, named very recently as pleosphere [132]. The evolution toward one of these carbonaceous forms and their asphaltene content depends mainly on the environmental temperature as well as on the droplet size and the residence time. The scanned pictures shown in Fig. 35 refer to different conditions [132,145,146]. It is reasonable to envisage that relatively low environmental temperatures with low oxygen concentration can favor pleosphere formation (Fig. A3, inset (c) [145]. Envelope flames cannot be stabilized because of the low oxygen concentration and the oxidation itself cannot raise significantly the final temperature [121]. The droplet size will be small enough that the extent over which mass diffusion of the lighter fuel fraction could be effective, will be comparable with drop size itself. The mass diffusion length increases, of course, with the residence time.

When batch distillation occurs, the fuel undergoes oxidation and pyrolysis in the gaseous phase. In this case, the tetraporphyrin should be opened by thermal decomposition. Unfortunately, very little information may be collected in this field. For instance, hydro-thermal atmospheric or pressure treatments up to 500 8C are not able to change the porphynic structure [139]. In any case, it is a matter of fact that vanadium forms its oxides in the presence of oxygen [140]. Also in this case mild condition with its low oxygen concentration may favor vanadium tetraoxide rather than pentaoxide, which is more noxious and less convenient because of its relatively low melting point (T ¼ 690 8C) [140]. Pleospheres and/or cenospheres may undergo partial mass loss, which could be due to physical (e.g. melting/vaporization/gasification) or chemical (e.g. dehydrogenation/oxidation) processes. In both cases, it is more likely that the organic part leaves the condensed material in such a way that the metallic part will be enriched. Sparse evidence of this effect have been collected by some authors. For instance, [147] shows that a great part of the metallic species produced by the combustion of the residual fuel oil are present in the PM 2.5 particulate as sulfate. In particular, vanadium seems to be present as vanadyl sulphate (VO*SO4*xH2O). Indirect evidence of this trend is the ternary sulfur/vanadium/carbon plot generated from measurements based on computer-controlled scanning electron microscopy (CCSEM) and shown in the inset f of Fig. A3. The same metal enrichment effect has been detected for the combustion of heavy oil in water emulsion [132]. Also the smaller is the condensed particle the higher is the level of the metal enrichment. Similar phenomena have been reported for material sampled in the stack of coal fired power plants [146]. They show the presence of spherical metallic particles inside empty carbonaceous spheres. The structure, named plerosphere by the authors, is shown in Fig. A3 (inset e), as a possible condensed phase particle, which may be generated straightway from the quasi-distillative process or after a partial metal enrichment of the organic/inorganic particle. The condensed particles, whatever process they undergo, in some conditions are more easily filtered the higher are their dimension. Larger particles can be achieved if the temperatures are kept as low as possible, compatibly with the minimum level for a reasonable oxidation activity. In other words Mild condition seems to be favorable to a high yield of metallic micronic particles.

References [1] Proceedings of Second International Seminar on High Temperature Combustion in Industrial Furnaces. Jernkontoret-KTH, Stockholm, Sweden; January 17 –18, 2000.

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366 [2] Crest—Third International Symposium on High Temperature Air Combustion and Gasification. Yokohama, Japan; March 7–9, 2000. [3] Symposium of High Temperature Air Combustion and Applications. Hsinchu, Taiwan; May 16– 17, 2000. [4] Yoshikawa K. Recent progress and future prospect of the CREST MEET Project. Crest—Third International Symposium on High Temperature Air Combustion and Gasification, Yokohama, Japan; 2000. A1/1. [5] Katsuki M, Hasegawa T. The science and technology of combustion in highly preheated air. Proc Combust Inst 1998; 27:3135–46. [6] Bolz S, Gupta AK. Effect of air preheated temperature and oxygen concentration on flame structure and emission. Int Joint Power Generat Conf, ASME 1998;1:193–205. [7] Gupta AK. Flame characteristic and challenges with high temperature air combustion. Proceedings of Second International Seminar on High Temperature Combustion in Industrial Furnaces, Jernkontoret-KTH, Stockholm, Sweden; 2000. p. 1. [8] Niioka T. Fundamentals and applications of high temperature air combustion. Proceedings of the Fifth ASME/JSME Joint Thermal Engineering Conference, San Diego; 1999. [9] Weinberg F. Combustion in heat-recirculating burners. In: Weinberg FJ, editor. Advanced combustion methods. Florida: Academic Press; 1986. [10] Aris R. Elementary chemical reactor analysis. Englewood Cliffs: Prentice Hall; 1969. [11] Fogler HS. Elementents of chemical reaction engineering. Englewood Cliffs: Prentice Hall; 1986. [12] de Joannon M, Saponaro A, Cavaliere A. Zero-dimensional analysis of methane diluted oxidation in rich conditions. Proc Combust Inst 2000;28:1639–46. [13] de Joannon M, Langella G, Beretta F, Cavaliere A, Noviello C. Reactor characteristics related to moderate or intense lowoxygen dilution for clean/cleaning combustion plants. Clean Air: Int J Energy Clean Environ 2003;4(1):1 –20. [14] Oberlack M, Arlitt R, Peters N. On stocastic Damko¨hler number variations in a homogeneous flow reactor. Combust Theory Modell 2000;4:495–509. [15] Peters N. Principles and potential of HiCOT combustion. Proceedings of the Forum on High-Temperature Air Combustion Technology; 2001. Josui Kaikan, p. 109. [16] Weber R, Orsino S, Lalleman N, Verlaan A. Combustion of natural gas with high-temperature air and large quantities of flue gas. Proc Combust Inst 2000;28:1315 –21. [17] Plessing T, Peters N, Wu¨nning G. Laser optical investigation of highly preheated combustion with strong exhaust gas recirculation. Proc Combust Inst 1998;27:3197–204. [18] Yasuda T. Dissemination project of high performance industrial furnace with use of high temperature air combustion technology. Proc of Second International High Temperature Air Combustion Symposium, Taiwan; 1999. p. B3. [19] Masunaga A., 2001. Application of regenerative burner for forging furnace. Proceedings of the Forum on HighTemperature Air Combustion Technology, p. 109. [20] Kelly-Zion PL, Dec JE. A computational study of the effect of fuel type on ignition time in homogeneous charge compression ignition engines. Proc Combust Inst 2000;28: 1187–94.

363

[21] Nygren J, Hult J, Richter M, Alde´n M, Christensen M, Hultqvist A, Johansson B. Three-dimensional laser induced fluorescence of fuel distribution in an HCCI engine. Proc Combust Inst 2002;29:679– 85. [22] Wu¨nning JA, Wu¨nning JG. Flameless oxidation to reduce thermal NO formation. Prog Energy Combust Sci 1997; 23:81. [23] Milani A, Wu¨nning J., 2002. What are the stability limits of flameless combustion? IFRF Online Combustion Handbook. ISSN 1607-9116, Combustion File No: 173 http://www. exchange.ifrf.net/coop/imp.htm, 2002. [24] Hasegawa T, Mochida S, Gupta AK. Development of advanced industrial furnace using highly preheated combustion air. J Propulsion Power 2002;18(2):233. [25] Kumar S, Paul PJ, Mukunda HS. Studies on a new high intensity-low emission burner. Proc Combust Inst 2002;29: 1131–7. [26] Lefebvre A. Atomization and sprays. New York: Hemisphere Publishing Corp; 1989. [27] Givler SD, Abraham J. Supercritical droplet vaporization and combustion studies. Prog Energy Combust Sci 1996;22: 1– 28. [28] Law CK. Recent advances in droplet vaporization combustion. Prog Energy Combust Sci 1982;8(3):171–201. [29] Sirignano WA. Fuel droplet vaporization and spray combustion theory. Prog Energy Combust Sci 1983;9(4):291– 322. [30] Moszkowicz P, Witzel L, Claus G. Modelling of very fast pyrolysis of heavy fuel oil droplets. Chem Engng Sci 1996; 51(17):4075–86. [31] Lawn CJ, Cunningham ATS, Street PJ, Matthews KJ, Sarjeant M, Godridge AM. The combustion of heavy fueloil. Principles of combustion engineering for boilers, Florida: Academic Press; 1987. [32] Peters N. Turbulent combustion. Cambridge: Cambridge University Press; 2000. [33] Peters N. Laminar diffusion flamelet models in non-premixed turbulent combustion. Prog Energy Combust Sci 1984;10: 319 –39. [34] Cavaliere A, Ragucci R. Gaseous diffusion flames: simple structures and their interaction. Prog Energy Combust Sci 2001;27:547–85. ¨ zdemir IB, Peters N. Characteristics of the reaction zone in [35] O a combustor operating at Mild Combustion. Exp Fluids 2001; 30(6):683–95. [36] Yasuda T. Development of super advanced regenerative furnace with HITAC. Crest—Fifth International Symposium on High Temperature Air Combustion and Gasification, Japan: Tokyo Institute of Technology; 2002. [37] Akinyemi O, Toqan MA, Bee´r JM, Syska A, Thijsson J, Benson Ch, Moreland D. Development of a high air preheated low NOx burner; experimental studies aided by computational modeling. Proceedings of the Second International High Temperature Air Combustion Symposium, Taiwan; 1999. p. C1. [38] Rørtveit J, Zepter K, Skreiberg Ø, Fossum M, Hustad JE. A comparison of low-NOx burners for combustion of methane and hydrogen mixtures. Proc Combust Inst 2002;29:1123–9. [39] Weber R, Verlaan AL, Orsino S, Lallemant N. On emerging furnace design methodology that provides substantial energy savings and drastic reductions in CO2, CO and NOx emissions. J Inst Energy 1999;72(492):77–83.

364

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

[40] Weber R, Verlaan AL, Orsino S, Lallemant N. Proceedings of the Second International Seminar on High Temperature Combustion in Industrial Furnaces, Jernkontoret-KTH, Stockholm, Sweden 2000;. [41] Coelho PJ, Peters N. Numerical simulation of a Mild Combustion burner. Combust Flame 2001;124:503–18. [42] Bee´r JM. Combustion technology developments in power generation in response to environmental challenges. Prog Energy Combust Sci 2000;26:301–27. [43] Longwell JP. In: Bartok W, Sarofim AF, editors. Fossil fuel combustion. New York: Wiley; 1991. p. 39. [44] Tan Y, Dagaut P, Cathonnet M, Boettner JC. Pyrolysis, oxidation and ignition of C1 and C2 hydrocarbons: experiments and modeling. J Chim Phys 1995;92:726–46. [45] de Joannon M, Cavaliere A, Donnarumma R, Ragucci R. Dependence of autoignition delay on oxygen concentration in Mild Combustion of heavy molecular paraffin. Proc Combust Inst 2002;29:1139–46. [46] Lutz AE, Rupley FM, Kee RJ, Reynolds WC, Sandia National Laboratories Report, Livermore, CA; 1998. [47] Kee RJ, Rupley FM, Miller JA, Sandia National Laboratories Report No. SAND 89-8009B, Livermore, CA; 1989. [48] Westbrook CK, Dryer F. Chemical kinetic modeling of hydrocarbon combustion. Combust Sci Technol 1984;10(1): 1–57. [49] Faravelli T, Gaffuri P, Ranzi E, Griffiths JF. Detailed thermokinetic modeling of alkane autoignition as tool for the optimization of performance of internal combustion engines. Fuel 1998;77(3):147–55. [50] Ranzi E, Dente M, Goldaniga A, Bozzano G, Faravelli T. Lumping procedures in detailed kinetic modeling of gasification, pyrolysis, partial oxidation and combustion of hydrocarbon mixture. Prog Energy Combust Sci 2001;27(1): 99–139. [51] Benson SW. Thermochemical kinetics. New York: Wiley; 1976. [52] Griffiths JF, Barnard JA. Flame and combustion. Glasgow: Blackie Academic and Professional; 1995. [53] Westbrook CK. Chemical kinetics of hydrocarbon ignition in practical combustion systems. Proc Combust Inst 2000;28: 1563–77. [54] Westbrook CK, Curran HJ, Pitz WJ, Griffiths JF, Mohamed C, Wo SK. The effects of pressure, temperature and concentration on the reactivity of alkanes: experiments and modeling in a rapid compression machine. Proc Combust Inst 1998;27:371 –8. [55] Cavaliere A, de Joannon M. Detailed chemical kinetics in the reactor design for diluted high temperature combustion of air/ paraffin mixtures. Crest—Third International Symposium on High Temperature Air Combustion and Gasification, Yokohama, Japan; 2000. [56] Ishiguro T, Tsuge S, Furuhata T, Kitigawa K, Arai N, Hasegawa T, Tanaka R, Gupta AK. Homogenization and stabilization during combustion of hydrocarbons with preheated air. Proc Combust Inst 1998;27:3205–13. [57] Gupta AK. High temperature air combustion: from energy conservation to pollution reduction. Boca Raton: CRC Press; 2002. [58] Dally BB, Karpetis AN, Barlow RS. Structure of turbulent nonpremixed jet flames in a diluted hot coflow. Proc Combust Inst 2002;29:1147–54.

[59] Shimo N. Fundamental research of oil combustion with highly preheated air. Proceedings of the Second International Seminar on High Temperature Combustion; 2000. [60] Ruan J, Kobayashi H, Niioka T, Ju Y. Combined effects of nongray radiation and pressure on premixed (CH4/O2)CO2 flames. Combust Flame 2001;124:225 –30. [61] Guo H, Ju Y, Maruta K, Niioka T, Liu F. Radiation extinction limit of counterflow premixed lean methane– air flames. Combust Flame 1997;109:639–46. [62] Wang J, Niioka T. The effect of radiation reabsorption on NO formation in CH4/air counterflow diffusion flames. Combust Theory Modell 2001;5:385–98. [63] Kitigawa K, Konishi N, Arai N, Gupta AK. Two-dimensional distribution of flame fluctuation during highly preheated air combustion. Int Joint Power Generation Conf, ASME 1998; 239:242. [64] Lalleman N, Soyre A, Weber R. Evolution of emissivity correlations for H2O–CO2 –N2/air mixture and coupling with solution methods of the radiative transfer equation. Prog Energy Combust Sci 1996;22:543–74. [65] Maruta K, Muso K, Takeda K, Niioka T. Reaction zone structure in flameless combustion. Proc Combust Inst 2000; 28:2117–23. [66] Fieweger K, Blumenthal R, Adomeit G. Self ignition of S.I. engine model fuels: a shock tube investigation at high pressure. Combust Flame 1997;109(4):599–619. [67] de Joannon M, Cavaliere A, Ragucci R. Air dilution effects on tetradecane spray autoignition in transcritical and supercritical regime. Atomization Spray 1999;9(2):153–72. [68] Minetti R, Carlier M, Ribaucour M, Therssen E, Sochet LR. A rapid compression machine investigation of oxidation and autoingition of n-heptane: measurements and modeling. Combust Flame 1995;102(3):298 –309. [69] Pfahl U, Fieweger K, Adomeit G. Self-ignition of diesel relevant hydrocabon air mixtures under engine conditions. Proc Combust Inst 1996;26:781 –9. [70] Kwon S, Arai M, Hiroyasu H. Ignition delay of a diesel spray injected into a residual gas mixture. SAE Technical Paper Series 911841; 1991. [71] Abd-Alla GH. Using exhaust gas recirculation in internal combustion engines: a review. Energy Conversion Mgmt 2002;43(8):1027 –42. [72] Sasaki S, Sawada D, Ueda T, Sami H. Effects of EGR on direct injection gasoline engine. JSAE Rev 1998;19(3): 223 –8. [73] Davidson DF, Horning DC, Herbon JT, Hanson RK. Shock tube measurements of JP-10 ignition. Proc Combust Inst 2000;28:1687–92. [74] Battin-Leclerc F, Fournet R, Glaude PA, Judenherc B, Warth V, Come GM, Scacchi G. Modeling of the gas-phase oxidation of n-decane from 550 to 1600 K. Proc Combust Inst 2000;28:1597 –605. [75] Mullins BP. The spontaneous ignition of fuel injected into a hot air stream. I Development of a combustion test rig for measuring the ignition delay of fuels. Fuel 1953;32:211–33. [76] Flowers D, Aceves S, Westbrook CK, Smith JR, Dibble R. Detailed chemical kinetic simulation of natural gas HCCI combustion: gas composition effects and investigation of control strategies. J Engng Gas Turbine Power 2001;123:433. [77] Kraft M, Maigaard P, Mauss F, Christensen M, Johansson B. Investigation of combustion emissions in a homogeneous

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366

[78]

[79]

[80]

[81]

[82]

[83]

[84]

[85]

[86] [87]

[88] [89] [90]

[91]

[92]

[93]

[94]

[95]

charge compression injection engine: measurements and a new computational model. Proc Combust Inst 2000;28: 1195–201. Soyan H, Lovas T, Mauss F. A stochastic simulation of an HCCI engine using an automatically reduced mechanism. Proceedings of the Ninth International Conference on Numerical Combustion, Sorrento; 2002. p. 163. Westbrook CK, Pitz WJ. Alternatives for modeling autoignition in homogeneous charge, compression ignition (HCCI) combustion. Proceedings of the Ninth International Conference on Numerical Combustion, Sorrento; 2002. p. 167. Zheng J, Yang W, Miller DL, Cernansky NP. Prediction of pre-ignition reactivity and ignition delay for HCCI using a reduced chemical kinetic model. SAE Paper no 2001-011025; 2001. Kong SC, Reitz RD. Application of detailed chemistry and CFD for predicting direct injection HCCI engine combustion and emissions. Proc Combust Inst 2002;29:663–9. Flowers DL, Aceves SM, Martinez-Frias J, Dibble RW. Prediction of carbon monoxide and hydrocarbon emissions in iso-octane HCCI engine combustion using multi-zone simulations. Proc Combust Inst 2002;29:687 –94. Smith JA, Bartley GJJ. Stoichiometric operation of a gas engine utilizing synthesis gas and EGR for NOx control. J Engng Gas Turbine Power 2000;122:617. Liu Z, Karim GA. An examination of the ignition delay period in gas-fueled diesel engines. J Engng Gas Turbine Power 1998;120(1):225–31. Dryer F. Comments at:Kelly-Zion P. Dec J.E., A computational study of the effect of fuel type on ignition time in homogeneous charge compression ignition engines. Proc. Combust. Inst. 2002;28:1187–94. Lefebvre A. Gas turbine combustion. Washington: Hemisphere Publishing; 1983. de Joannon M, Tregrossi A, Cavaliere A. Temperature oscillation in methane mild oxidation. 29th Symposium on Combustion; 2002. WIP book. www.power.alstom.com. www.floxcom.ippt.gov.pl/. Weber R. Flow and mixing in gas-fired furnaces operated with highly preheated combustion air. Proceeding of Second International High Temperature Air Combustion Symposium, Taiwan; 1999. C2. Robertson T, Newby JN. A review of the development and commercial application of the LNI technique. Proceedings of the Fourth High Temperature Air Combustion and Gasification, Rome; 2001. paper no 11. Kobayashi H, Shioda S, Yoshikawa K. Coal/waste gasification power generation system using high temperature air. Proceeding of Second International High Temperature Air Combustion Symposium, Taiwan; 1999. E1. Yamada M, Onoda A, Maeda F, Furukawa T. Gas turbine combustor for gasified L-BTU fuel. Proceeding of Second International High Temperature Air Combustion Symposium, Taiwan; 1999. B5. Guarinello Jr. F, Cerqueira SAAG, Nebra SA. Thermoeconomic evaluation of a gas turbine cogeneration system. Energy Conversion Mgmt 2000;41(11):1191–200. Gallo WLR. A comparison between the hat cycle and other gas-turbine based cycles: efficiency, specific power and water

[96]

[97]

[98] [99]

[100]

[101] [102]

[103] [104]

[105]

[106]

[107]

[108]

[109]

[110] [111]

[112]

[113]

[114]

[115]

365

consumption. Energy Conversion Mgmt 1997;38(15–17): 1595–604. Traverso A, Massardo A. Thermoeconomic analysis of mixed gas-steam cycles. Appl Thermal Engng 2002;22(1): 1–21. Tanaka R, Kishimoto K, Hasequawa T. Combust Sci Technol 1994;3:257. in japanese, quoted by Yiguang Yu: Proceedings of The First Asia-Pacific Conference on Combustion May 12-15, OSAKA, 460, 1997. Masters J, Webb RJ, Davies RM. J Inst Energy 1979; 196 –204. Howell JR, Hall MJ, Ellzey JL. Combustion of hydrocarbon fuels within porous inert media. Progr Energy Combust Sci 1996;22:121–45. Hoffmann JG, Echigo R, Yoshida H, Tada S. Experimental study on combustion in porous media with a reciprocating flow system. Combust Flame 1997;111:32 –46. Thermatrix: http://www.thermatrix.com. Session on Porous Media of the Twenty-seventh International Symposium on Combustion. Boulder (CO): The Combustion Institute; 1998. Hsu PF, Howell JR. Exp Heat Transfer 1993;5:293–313. Hasegawa T, Tanaka R, Niioka T, Proceedings of The First Asia–Pacific Conference on Combustion, Osaka; 1997. p. 290. Chomiak J, Longwell JP, Sarofim AF. Combustion of low calorific value gases. Problem and prospects. Prog Energy Combust Sci 1989;15:109 –29. Shigeta E, Kanazawa H, Koizumi T, Nagata T. Low-NOx combustion technique for high-temperature furnace. AFRC/ JFRC International Conference on Environmental Control of Combustion Processes, Honolulu, Hawaii; 1991. Milani A. Mild Combustion techniques applied to regenerative firing in industrial furnaces. Proceedings of Second International Seminar on High Temperature Combustion in Industrial Furnaces, Jernkontoret-KTH, Stockholm, Sweden; 2000. Fernandez-Pello CA. Micropower generation using combustion: issues and approaches. Proc Combust Inst 2002;29: 883 –99. Weinberg FJ, Rowe DM, Min G, Ronney PD. On thermoelectric power conversion from heat recirculating combustion systems. Proc Combust Inst 2002;29:241 –947. Ronney P. Analysis of non-adiabatic heat-recirculating combustors. Combust Flame 2003;135:421–39. Vican J, Gajdeczko BF, Dryer FL, Milius DL, Aksay IA, Yetter RA. Development of a microreactor as a thermal source for microelectromechanical systems power generation. Proc Combust Inst 2002;29:909 –16. Meier JG, Vollerin BL. The design of an integrated burnerboiler system using flue-gas recirculation. Proc Combust Inst 1976;16:63–76. Douce F, Djebaı¨li-Chaumeix N, Paillard CE, Clinard C, Rouzaud JN. Soot formation from heavy hydrocarbons behind reflected shock waves. Proc Combust Inst 2000;28: 2523–9. Wang R, Cadman P, Soot AH. Soot and PAH production from spray combustion of different hydrocarbons behind reflected shock waves. Combust Flame 1998;112:359–70. Bo¨hm H, Hesse D, Jander H, Lu¨ers B, Pietscher J, Wagner HGG, Weiss M. The influence of pressure and temperature

366

[116]

[117]

[118]

[119]

[120]

[121] [122]

[123]

[124]

[125]

[126]

[127]

[128]

[129]

[130]

A. Cavaliere, M. de Joannon / Progress in Energy and Combustion Science 30 (2004) 329–366 on soot formation in premixed flames. Proc Combust Inst 1988;22:403 –11. Axelbaum R, Flower W, Law C. Diluition and temperature effects of inert addition on soot formation in counterflow diffusion flames. Combust Sci Technol 1988;61:51–73. Du D X, Axelbaum R, Law C. The influence of carbon dioxide and oxygen as additives on soot formation in diffusion flames. Proc Combust Inst 1990;23:1501– 7. Glassman I. Sooting laminar diffusion flames: effect of dilution, additives, pressure and microgravity. Proc Combust Inst 1998;27:1589–96. Angrill O, Geitlinger H, Streibel T, Suntz R, Bockhorn H. Influence of exhaust gas recirculation on soot formation in diffusion flames. Proc Combust Inst 2000;28:2643– 9. Zhang C, Atreya A, Lee K. Sooting structure of methane counterflow diffusion flames with preheated reactants and dilution by products of combustion. Proc Combust Inst 1992; 24:1049–57. Sjo¨gren A. Soot formation by combustion of an atomized liquid fuel. Proc Combust Inst 1973;14:919–27. Fuse R, Kobayashi H, Ju Y, Maruta K, Niioka T. NOx emission from high-temperature air/methane counterflow diffusion flame. Int J Therm Sci 2002;41:693 –8. Ju Y, Niioka T. Computation of NOx emission of a methaneair diffusion flame in a two-dimensional laminar jet with detailed chemistry. Combust Theory Modell 1997;1:243– 58. Xu H, Smoot LD, Hill SC. Computational modeling for NOx reduction by advanced reburning. Energy Fuel 1999;13(2): 411–20. Zamansky VM, Ho L, Maly PM, Seeker WR. Reburning promoted by nitrogen and sodium containing compounds. Proc Combust Inst 1996;26:2075–82. Rota R, Zanoelo EF, Morbidelli M, Carra` S. Effect of mixing on selective non-catalytic nitric oxide reduction. Fifth International Conference on Technologies and combustion for a Clean Environment, Lisbona; 1999. p. 119– 28. Østberg M, Dam-Johansen K, Johnsson JE. Influence of mixing on the SNCR process. Chem Engng Sci 1997;52(15): 2511–25. Caton JA, Sieber DL. Comparison of nitric oxide removal by cianuric acid and by ammonia. Combust Sci Technol 1989; 65:277–93. Nimmo W, Agnew J, Hampartsoumian E, Jones JM. Removal of H2S by spray-calcined calcium acetate. Indust Engng Chem Res 1999;38(8):2954–62. Adanez J, Garcia-Labiano F, de Diego LF, Fierro V. Utilization of calcium acetate and calcium magnesium acetate for H2S removal in coal gas cleaning at high temperatures. Energy Fuels 1999;13:440–8.

[131] Lawn CJ, Cunningham ATS, Street PJ, Matthews KJ, Sarjeant M, Godridge AM. The combustion of heavy fueloil. In: Lawn CJ, editor. Principles of combustion engineering for boilers. Florida: Academic Press; 1987. [132] Allouis C, Beretta F, D’Alessio A. Structure of inorganic and carbonaceous particles emitted from heavy oil combustion. Chemosphere 2003;51(10):1091– 6. [133] Affens RS, Sheinson WA. Am Inst Chem Engng 1979;13:83. [134] Lucka K, Ko¨hne H. Usage of cold flames for the evaporation of liquid fuels. Fifth International Conference on Technologies and Combustion for a Clean Environment, Lisbona; 1999. p. 207 –13. [135] Zabetakis MG, Furno AL, Jones GW. Minimum spontaneous ignition temperature of combustible in air. Ind Engng Chem 1954;46:2173–8. [136] Ping’an P, Morales-Isquierdo A, Lown EM, Strausz OP. Chemical structure and biomarker content of Jinghan asphaltene and kerogens. Energy Fuel 1999;13(2):248 –65. [137] Strausz OP, Mojelsky TW, Farhad F, Lown EM. Additional structural details on Athabasca asphaltene and their ramifications. Energy Fuel 1999;13(2):207– 27. [138] Iisa K, Lu Y, Salmenoja K. Sulfation of potassium chloride at combustion conditions. Energy Fuels 1999;13(2):1184–90. [139] Whitehead EV. In: Yen TF, Chilingarian GV, editors. Fuel oil chemistry and asphaltenes in asphaltenes and asphalts, vol. 1. Amsterdam: Elsevier; 1994. [140] Molero de Blas L.J., www.chemeng.ucl.ac.uk/research/ combusti/thesis/. PhD Thesis. University of London; 2001. [141] Yen TF, Chilingarian GV. In: Yen TF, Chilingarian GV, editors. Fuel oil chemistry and asphaltenes in asphaltenes and asphalts, vol. 1. Amsterdam: Elsevier; 1994. [142] Hunt J. Petroleum geochemistry and geology. San Francisco: W. H. Freeman and Co; 1979. 617 p. [143] Tissot BP, Welte DH. Petroleum formation and occurrence. Berlin: Springer; 1978. 538 p. Sectiom 1.9. [144] Royle RA. www.vanadium.com.au/ScienceStuff/Chemistry/ geochemistry_of_vanadium.htm; 2001. [145] Massoli P, Beretta F, D’Alessio A, Lazzaro M. Proceedings of Joint Meeting of the Soviet and Italian Section of the Combustion Institute, 9.2, Pisa; 1990. [146] Fisher GL, Chang DPY, Brummer M. Fly ash collected from electrostatic precipitators: microcrystalline structures and the mystery of the spheres. Science 1976;192(4239):553–8. [147] Huffman GP, Huggins FE, Shah N, Huggins R, Linak WP, Miller CA, Pugmire RJ, Meuzelaar HLC, Seehra M, Mannivannan A. Characterization of fine particulate matter produced by combustion of residual fuel oil. J Air Waste Mgmt Assoc 2000;50(7):1106–14.