betaFeedback

WorksPhysics and Astronomy

A Quantitative Analysis of the Ignition Characteristics of Fine Iron Particles

Abstract

Ignition of iron particles in an oxidizing environment marks the onset of a self-sustained combustion. The objective of the current study is to quantitatively examine the ignition characteristics of fine iron particles governed by the kinetics of solid-phase iron oxidation. The oxidation rates are inversely proportional to the thickness of the oxide layer and calibrated using the experimentally measured growth of iron-oxide layers over time. Steady-state and unsteady analysis have been performed to probe the dependence of the critical gas temperature required to trigger a thermal runaway (namely, the ignition temperature TignT_\mathrm{ign}) on particle size, initial thickness of oxide layer, inert gas species, radiative heat loss, and the collective heating effect in a suspension of particles. Both analyses indicate that TignT_\mathrm{ign} depends on δ0δ_0, i.e., the ratio between the initial oxide layer thickness and particle size, regardless of the absolute size of the particle. The unsteady analysis predicts that, for δ0≲0.003δ_0 \lesssim 0.003, TignT_\mathrm{ign} becomes independent of δ0δ_0. Under standard conditions in air, TignT_\mathrm{ign} is approximately 1080 K for any particle size greater than 5 microns. Radiative heat loss has a minor effect on TignT_\mathrm{ign}. The collective effect of a suspension of iron particles in reducing TignT_\mathrm{ign} is demonstrated. The transition behavior between kinetic-controlled and external-diffusion-controlled combustion regimes of an ignited iron particle is systematically examined. The influences of initial oxide-layer thickness and particle temperature on the ignition delay time, τignτ_\mathrm{ign}, of iron particles are parametrically probed. A d2d^2-law scaling between τignτ_\mathrm{ign} and particle size is identified.

Introduction

Iron is an excellent fuel for long-term storage and long-distance transport of clean energy owing to its carbon-free nature, high energy density, and potential for non-volatile combustion in air [1, 2, 3]. For developing practical energy-conversion technologies based on iron fuel, a better understanding of the fundamentals underlying the combustion process of fine (i.e., micron- to hundreds-of-micron-sized) iron particles at elevated temperatures is required. The full oxidation process of an iron particle in a combustion system consists of three major stages: (1) Preheating and ignition leading to a thermal runaway; (2) a rapid oxidation process of a molten droplet at elevated temperatures until iron is completely oxidized; (3) a slow further oxidation process of the lower oxidized products, i.e., from FeO or Fe3_3O4_4 to Fe2_2O3_3, upon cooling. To design a real-world combustion device of iron particles, quantitative answers are required for the following questions: (A) Under what conditions can iron particles be ignited? (B) How to ensure a non-volatile combustion of molten iron droplets? (C) What are the rates of energy release during the rapid and slow iron oxidation processes? To answer these questions, we need to better identify the rate-controlling mechanisms among a complex of physicochemical processes in the different stages of iron combustion, including oxygen diffusion in the gas, oxygen absorption at the particle surface, phase change and dissociative evaporation of iron and its oxides, solid- and liquid-phase kinetics of iron oxidation, and heat exchange between particles and the ambient gas. Both experimental and theoretical endeavors are required for this exploration.

Prior to the research campaigns on fine iron particles, knowledge of iron combustion has mostly been acquired from experimental investigations on the growth of solid iron-oxide scales in air [4, 5, 6, 7, 8] and iron rod combustion in high pressure oxygen [9, 10, 11, 12, 13, 14, 15]. Metallurgical studies [4, 5] dated in the 1950s demonstrated that the growth of solid-phase iron oxides on a specimen of pure iron at a steady temperature follows a parabolic rate law, namely, with a rate inversely proportional to the thickness of the oxide scale. This finding agrees well with the prediction of Wagner’s theory [16] asserting that the rate of solid-phase metal oxidation is controlled by the lattice diffusion of ions across the oxide scale [17]. Quantitatively accurate kinetics for solid-phase iron oxidation can be determined from these experimental results.

The combustion complex of a mm-diameter iron rod consists of a molten droplet of iron and iron oxides attached to a reaction front that melts and propagates into the solid rod [9]. Dreizin [18] elucidated that the burning rate of an iron rod can be either limited by the iron oxidation kinetics at sufficiently high gas temperatures [11, 12] or the incorporation rate of gaseous oxygen at low gas temperatures [9, 19]. Recent studies by Muller et al. [14, 15] reveal more details of the molten mixture formed on the top of a laser-ignited iron rod. A spatially non-uniform distribution of molten iron, oxides, and gas bubbles inside the droplet was identified [15]. A subsequent modeling work [20] further describes the complex thermo-fluid processes within this molten-phase mixture controlling the burning rate of an iron rod. Alas, quantitatively useful information to answer the aforeposed questions for iron-particle combustion is mostly masked by the intrinsic complexity of iron-rod combustion.

Over the past two decades, the state-of-the-art understanding in the combustion of fine iron particles has been advanced by researchers around the globe. This collective effort has, however, been mainly focused on experimentally examining [21, 22, 23, 24, 25, 26, 27, 28] and qualitatively modeling [29, 30, 31, 32, 33, 34] the macroscopic observables of flame propagation in a dispersion of iron particles in an oxidizing gas. Although a few recent experimental studies [35, 36, 37, 38] attempt to monitor the combustion process of individual iron particles, in situ measurements of particle temperature and composition have not yet been extensively obtained. Thus, the detailed rate-controlling mechanisms of iron-particle combustion, especially for a molten droplet, remain elusive. Without a physics-based, quantitatively reliable model for calculating the energy release rate of an iron particle, the development of high-fidelity simulations of large-scale combustion systems is baseless despite the sophistication achieved in modeling other components of the problem.

This paper reports an effort in seeking a quantitative answer to the aforementioned Question (A) with regard to the ignition characteristics of fine iron particles. To ignite an iron particle means that a set of critical conditions (e.g., gas or particle temperature) is satisfied to trigger a thermal runaway—the rate of energy release due to iron oxidation exceeds the rate of heat loss to the surrounding. There is yet no experimental measurement of the critical particle or gas temperature required for triggering thermal runaway (referred to as “ignition temperature” in the remainder of this paper) of an isolated particle, or a suspension of iron particles, in a flame. The experimental data of iron ignition temperature in the literature were obtained for suspensions and precipitated beds of fine iron powders [39, 40] and mm- to cm-sized iron or steel specimens [41, 42, 43]. These data, as partly summarized by Breiter et al. [44], are widely scattered owing to the difference in experimental conditions, specimen properties and morphologies, and definition of “ignition temperature”. The experimental measurement reported by Grosse and Conway [41] and Bolobov [43] are perhaps the closest to the herein defined ignition temperature. In Grosse and Conway’s study [41], a 10-g-weighted specimen of iron was placed in an Alundum crucible and heated to a designated temperature in an argon atmosphere by an electric furnace, and then, a flow of oxygen was fed to the top of the crucible. This procedure was repeated with an incremented temperature until an event of thermal runaway was detected by visual observation. An ignition temperature of 1203±10 K1203 \pm10\,\mathrm{K} was obtained. A similar procedure was followed by Bolobov [43] to determine the ignition temperature of a 5×5×0.5 mm5 \times5 \times0.5\,\mathrm{mm} low-carbon steel foil. The specimens were heated in vacuum to a designated temperature before supplying oxygen. The event of a thermal runaway was detected as an abrupt increase in the temporal record of specimen temperature. An ignition temperature of 1233±20 K1233 \pm20\,\mathrm{K}, independent of ambient oxygen concentration, was found. Both of these experimental results of ignition temperature are significantly below the melting point of FeO\mathrm{FeO}, i.e., 1650 K1650\,\mathrm{K}, indicating that the ignition process occurs when a particle remains completely in solid phase.1

Khaikin [45] proposed a generic ignition model for reactive metal particles considering the obstructing effect of an oxide layer on the reaction rate, but did not apply it for quantitatively estimating the ignition temperature of iron particles. A quantitative model was developed by Bolobov and Podlesnikh [46] to explain the ignition characteristics of a low-carbon steel foil as experimentally probed in Ref. [43]. Despite some ambiguity in the kinetic parameters for iron oxidation, the ignition temperatures predicted by Bolobov and Podlevskikh’s model [46] are fairly close to the experimental results with a discrepancy less than 50 K. The current study aims to develop a model to quantitatively capture the ignition characteristics of fine iron particles. To this end, a physics-based kinetic model of solid-phase iron oxidation—a parabolic rate law reflecting that the oxidation rate is controlled by the lattice diffusion of Fe cations through oxide layers—is considered in the current ignition model. The values of the kinetic parameters were calibrated against the experimental measurement of growth rates of iron-oxide scales over a temperature range from 973 K to 1523 K obtained by Païdassi [4]. Using this model, the effects of initial particle size, initial oxide layer thickness, inert gas species, radiative heat loss, and collective heating in a particulate suspension on the ignition characteristics of iron particles are examined. The model prediction of ignition temperature is compared with the experimental results of Grosse and Conway [41] and Bolobov [43]. The transition behavior between kinetic-controlled and external-diffusion-controlled combustion regimes and the ignition delay times of iron particles are further examined. The possible sources of error of this model are also discussed.

Iron oxidation kinetic model

Physics underlying solid-phase iron oxidation

Microscopic cross-section views of a solid-phase oxide scale grown on iron under isothermal conditions (over a range in temperature from 973 K to 1523 K) in air were first obtained by Païdassi in the 1950s [4]. Figure 1(a) is a sample image from Ref. [4] showing that an iron-oxide scale consists of three compact layers of hematite (Fe2O3\mathrm{Fe_2O_3}), magnetite (Fe3O4\mathrm{Fe_3O_4}), and wüstite (FeO\mathrm{FeO}) stacked from the gas-oxide interface to the oxide-iron interface. The relative thicknesses of the Fe2O3\mathrm{Fe_2O_3}, Fe3O4\mathrm{Fe_3O_4}, and FeO\mathrm{FeO} layers with respect to the total thickness of the oxide scale are 1%, 4%, and 95%, respectively. As conceptually illustrated in Figure 1(b), for a compact oxide film (i.e., without cracks or grain boundaries to facilitate oxygen transport to the iron core), solid-phase iron oxidation consists of the following processes:

Multilayered oxide scale and schematic illustration of solid-phase iron oxidation

Figure 1. (a) A sample cross-section view of a multilayered oxide scale grown on iron as shown by Païdassi [4]; (b) a schematic illustration showing the key processes underlying solid-phase iron oxidation.

  1. Diffusion of oxygen (O2\mathrm{O_2}) molecules from the bulk gas to the gas-oxide interface.

  2. Incorporation of oxygen into the oxide scale (i.e., a dissociative absorption of O2\mathrm{O_2} molecules as oxygen anions O2−\mathrm{O^{2-}}).

  3. Absorption of iron into the oxide film as iron cations.

  4. Diffusion of ions and electrons through the oxide scale.

  5. Reactions at FeO-Fe3_3O4_4, Fe3_3O4_4-Fe2_2O3_3, and Fe2_2O3_3-O2_2 interfaces.

At relatively low temperatures, the diffusion of oxygen in the gas phase (Process 1) is much more rapid than the kinetic processes (2 to 5) in the solid phase. As the temperature of the particle increases, the solid-phase kinetics become increasingly rapid, and thus, the external diffusion of oxygen becomes the rate-limiting process. In this analysis, the oxidation rate is considered to be only controlled by the solid-phase kinetics for determining the critical conditions for particle ignition, i.e., the critical temperature required to trigger thermal runaway, in Sects. 5.1-5.5; a switch-type reaction-rate model, considering the transition from kinetic- to external-diffusion-controlled combustion, is used describe the ignition behavior of iron particles beyond the critical temperature for thermal runaway, in Sects. 5.6 and 5.7.

For a sufficiently thick oxide scale, the growth rate of each layer is controlled by the diffusion of ions subjected to the equilibrium activities of Fe and O at the interfaces [47]. Considering the fact that the transport of electrons and the establishment of local equilibria at the interfaces are significantly more rapid than the diffusion of ions across the oxide layers, Wagner’s theory [16] relates the oxide growth rate to the diffusivity of ions. Since Fe cations are significantly more mobile in FeO and Fe3_3O4_4 layers than O anions due to a large size of O anion (1.4 Å) [48], the growth of oxide film is predominantly governed by the outward diffusion of Fe cations [7, 49]. The growth of the FeO layer is the most rapid due to the fact that the diffusion coefficient of Fe cations in FeO is greater than that in Fe3_3O4_4 or Fe2_2O3_3. Such a diffusion-controlled growth follows a parabolic rate law,

dXidt=kp,iXi(1)\frac{\mathrm{d}X_i}{\mathrm{d}t} = \frac{k_{p,i}}{X_i} \tag*{(1)}

where XiX_i is the thickness of an oxide layer, kp,ik_{p,i} is the parabolic rate constant, and ii is the index of each oxide layer. Yurek et al. [6] and Garnaud and Rapp [8] extended Wagner’s theory to show that the growth of a multilayered oxide film on iron follows a parabolic rate law governed by the diffusion coefficients of Fe in FeO and Fe3_3O4_4. It is of importance to note that, as revealed by Goursat and Smeltzer [50], the oxidation kinetics of FeO is independent of ambient oxygen concentration over a temperature range (at least) from 1073 K to 1273 K for a partial pressure of O2_2 (pO2p_{\mathrm{O_2}}) above 4×10−44 \times10^{-4} atm. As the activities of Fe and O at the Fe-FeO, FeO-Fe3_3O4_4, and Fe3_3O4_4-Fe2_2O3_3 interfaces are fixed by the phase equilibria, the growth of FeO and Fe3_3O4_4 are barely affected by the ambient O2_2 concentration [51]. In this analysis, the kinetic rates of FeO and Fe3_3O4_4 formation are assumed to be independent of ambient O2_2 concentration as long as pO2p_{\mathrm{O}_2} is greater than zero. The implication of this assumption on the ignition process—the transition from a kinetic-controlled combustion to an O2_2-diffusion-controlled (also referred to as external-diffusion-controlled) combustion—of an iron particle is discussed in Sect. 5.6.

Since the growth of the Fe2_2O3_3 layer is significantly slower in comparison to the growth of the FeO and Fe3_3O4_4 layers, the energy release due to the formation Fe2_2O3_3 is negligible. The formation of this thin Fe2_2O3_3 layer is thus neglected in the current analysis. It is, however, of importance to understand the role played by the hematite layer on the formation of the inner oxidized layers. There is experimental evidence [52, 53] indicating that the formation of Fe2_2O3_3 is due to both an outward diffusion of Fe cations and an inward diffusion of O anions at comparable rates (as illustrated in Figure 1(b)), possibly along cracks in the hematite layer. Goursat and Smeltzer [50] showed that the Fe2_2O3_3 layer formed upon the Fe3_3O4_4 layer is a dense “forest” of whiskers and platelets. The growth rate of this irregular hematite layer adapts to the ambient O2_2 concentration, establishing an equilibrium at the Fe3_3O4_4-Fe3_3O2_2 interface, and thus, making the growth rates of FeO and Fe3_3O4_4 independent of O2_2 concentration.

Calibration of the kinetic model

The experimental data of Païdassi [4] reveal that, at a steady temperature, the thickness of each oxide layer XiX_i increases with square root of time t\sqrt{t} at a constant rate, indeed following a parabolic rate law. The parabolic rate constant kik_i can be measured as the slope of the fitted line to the experimental data of XiX_i as a function of t\sqrt{t} at each temperature, i.e., Xi=kit+Xi,0X_i = k_i\sqrt{t} + X_{i,0}, as reported in Table 1 of Ref. [4]. Note that kik_i can be converted to the rate constant kp,ik_{p,i} in Eq. (1) via the relation as follows:

k0k_0 (m2^2 s−1^{-1})TaT_a (K)
FeO2.670×10−42.670 \times10^{-4}20319
Fe3_3O4_41.027×10−61.027 \times10^{-6}21310

Table 1. Calibrated parameters for the parabolic growth rate of FeO and Fe3_3O4_4.

kp,i=ki22(2)k_{p,i} = \frac{k_i^2}{2} \tag*{(2)}

The natural logarithmic values of the experimental data of kp,ik_{p,i} for the growth of total oxide film thickness (blue diamond), FeO layer thickness (red square), and Fe3_3O4_4 layer thickness (black circle) are plotted in Figure 2 as a function of the reciprocal of temperature, 1/T1/T. These data points demonstrate a linear regression, i.e., the parabolic rate constant increases exponentially with temperature, indicating that the diffusion of Fe cations in the oxide layers is an activated process. The temperature dependence of kp,ik_{p,i} can be described by an Arrhenius function as follows,

Natural logarithmic values of experimental data and fitting lines

Figure 2. The natural logarithmic values of the experimental data of kp,ik_{p,i} for the growth of total oxide scale thickness (blue diamond), FeO layer thickness (red square), and Fe3_3O4_4 layer thickness (black circle) plotted as a function of the reciprocal of temperature, 1/T1/T, and a fitting line showing a linear regression of the corresponding data. Note that the FeO and total oxide growth rates are effectively equal due to the much slower growth rate of Fe3_3O4_4.

kp,i(T)=k0,iExp⁡(−Ta,iT)(3)k_{p,i}(T) = k_{0,i}\operatorname{Exp}\left(-\frac{T_{a,i}}{T}\right) \tag*{(3)}

wherein the values of pre-exponential factor k0,ik_{0,i} and activation temperature Ta,iT_{a,i} for the growth of FeO and Fe3_3O4_4 layers can be calibrated via line fitting to the plotted data points as reported in Table 1.

Model of iron particle ignition

In the current analysis, a thermophysical model based on the mass and energy balance equations with an empirically calibrated kinetic model of iron oxidation is used to describe the ignition process of an isolated iron particle and a suspension of iron particles. The detailed formulation and major assumptions made in this model are presented this section.

Model formulation for an isolated iron particle

This model considers a spherical particle consisting of a core of iron, an inner FeO layer, and an outer Fe3_3O4_4 layer as illustrated in Figure 3(a). For the temperature range over which the oxidation kinetic model is calibrated, i.e., 973 K to 1523 K, the thickness of Fe2_2O3_3 is about 1 % of the total thickness of the oxide film. The formation of Fe2_2O3_3 is thus neglected in this model. Since the thermal conductivities of Fe and FeO are greater than that of the surrounding gas by two to three orders of magnitude, a typical Biot number of a fine iron particle is on the order of 0.001–0.01. Hence, a uniform particle temperature, TpT_p, can be assumed. The particle exchanges heat with the ambient gas at a temperature TgT_g via convective and radiative heat transfer. Other major assumptions on which the current model is based are summarized in the list below:

Schematic of the thermophysical model of (a) an isolated particle and (b) a spherical suspension of reacting iron particles

Figure 3. Schematic of the thermophysical model of (a) an isolated particle consisting of a core of pure Fe, an inner oxide layer of FeO, and an outer oxide layer of Fe3_3O4_4 and (b) a spherical suspension of reacting iron particles.

  1. The densities of solid iron, FeO, Fe3_3O4_4 are considered as constant values, i.e., the effect of thermal expansion is neglected. Given the fact that, in the temperature range between 900 K and 1200 K, which is relevant to the current analysis, the coefficients of thermal expansion of Fe, FeO, and Fe3_3O4_4 vary over approximately the same range on the order of 1×10−61 \times10^{-6} K−1^{-1} [54, 55, 56], oxide layer cracking due the mismatch in thermal expansion between the oxide shell and the metal core is unlikely to core for iron particles.

  1. The phase changes from α\alpha-Fe to γ\gamma-Fe and from γ\gamma-Fe to δ\delta-Fe are not considered since the associated latent heats are much less than the specific energy release of FeO and Fe3_3O4_4 formation. The thermodynamic properties of iron used in the calculation are those of α\alpha-Fe.

  1. Curvature effect on the oxide layer growth is neglected as the oxide layer thickness is much less than the particle radius during the ignition process.

  1. The flow velocity of the ambient gas relative to the particle is negligible, i.e., the corresponding Reynolds number is zero.

The mass balance equations for Fe, FeO, and Fe3_3O4_4 can be related to the parabolic kinetic model (Eqs. (1) and (3)) as follows,

dmFedt=−νFe/FeOdmFeOdt−νFe/Fe3O4dmFe3O4dt(4)\frac{\mathrm{d}m_{\mathrm{Fe}}}{\mathrm{d}t}=-\nu_{\mathrm{Fe}/\mathrm{FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}-\nu_{\mathrm{Fe}/\mathrm{Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t} \tag*{(4)}
dmFeOdt=ρFeOAFeOdXFeOdt=ρFeOAFeOk0,FeOXFeOExp⁡(−Ta,FeOTp)(5)\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}=\rho_{\mathrm{FeO}}A_{\mathrm{FeO}}\frac{\mathrm{d}X_{\mathrm{FeO}}}{\mathrm{d}t}=\frac{\rho_{\mathrm{FeO}}A_{\mathrm{FeO}}k_{0,\mathrm{FeO}}}{X_{\mathrm{FeO}}}\operatorname{Exp}\left(\frac{-T_{\mathrm{a,FeO}}}{T_{\mathrm{p}}}\right) \tag*{(5)}
dmFe3O4dt=ρFe3O4ApdXFe3O4dt=ρFe3O4Apk0,Fe3O4XFe3O4Exp⁡(−Ta,Fe3O4Tp)(6)\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}=\rho_{\mathrm{Fe_3O_4}}A_{\mathrm{p}}\frac{\mathrm{d}X_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}=\frac{\rho_{\mathrm{Fe_3O_4}}A_{\mathrm{p}}k_{0,\mathrm{Fe_3O_4}}}{X_{\mathrm{Fe_3O_4}}}\operatorname{Exp}\left(\frac{-T_{\mathrm{a,Fe_3O_4}}}{T_{\mathrm{p}}}\right) \tag*{(6)}

where νFe/FeO\nu_{\mathrm{Fe}/\mathrm{FeO}} and νFe/Fe3O4\nu_{\mathrm{Fe}/\mathrm{Fe_3O_4}} are the stoichiometric mass ratios. Note that the formation of FeO takes place at the outer surface of the FeO layer (with an area AFeO=4π(rFe+XFeO)2A_{\mathrm{FeO}}=4\pi(r_{\mathrm{Fe}}+X_{\mathrm{FeO}})^2) and the formation of Fe3_3O4_4 at the outer surface of the particle (with an area Ap=4π(rFe+XFeO+XFe3O4)2A_{\mathrm{p}}=4\pi(r_{\mathrm{Fe}}+X_{\mathrm{FeO}}+X_{\mathrm{Fe_3O_4}})^2). The ratio between the total oxide layer thickness and particle radius, δ\delta, and the ratio between the Fe3_3O4_4 layer thickness and the total oxide layer thickness, δFe3O4\delta_{\mathrm{Fe_3O_4}}, are defined as follows:

δ=Xrp=XFeO+XFe3O4rp,δFe3O4=XFe3O4X(7)\delta=\frac{X}{r_{\mathrm{p}}}=\frac{X_{\mathrm{FeO}}+X_{\mathrm{Fe_3O_4}}}{r_{\mathrm{p}}},\qquad\delta_{\mathrm{Fe_3O_4}}=\frac{X_{\mathrm{Fe_3O_4}}}{X} \tag*{(7)}

In this analysis, the initial value of δFe3O4\delta_{\mathrm{Fe_3O_4}} is set to be 0.05 as inferred from experimental measurements [4, 5]. Given the fact that the total thickness of the oxide layer is much thinner than the radius of the iron core, i.e., X0≪rFe,0X_0\ll r_{\mathrm{Fe},0} and rFe,0≈rp,0r_{\mathrm{Fe},0}\approx r_{\mathrm{p},0}, the initial ratio between the oxide layer thickness and the particle radius, δ0\delta_0, is approximately proportional to the initial mass fraction of the oxide layer via the follow relation:

moxide,0mFe,0≈4πrFe,02X0ρoxide43πrFe,03ρFe=3ρoxideX0ρFerp,0=3ρoxideδ0ρFe(8)\frac{m_{\mathrm{oxide},0}}{m_{\mathrm{Fe},0}}\approx\frac{4\pi r_{\mathrm{Fe},0}^{2}X_0\rho_{\mathrm{oxide}}}{\frac{4}{3}\pi r_{\mathrm{Fe},0}^{3}\rho_{\mathrm{Fe}}}=\frac{3\rho_{\mathrm{oxide}}X_0}{\rho_{\mathrm{Fe}}r_{\mathrm{p},0}}=\frac{3\rho_{\mathrm{oxide}}\delta_0}{\rho_{\mathrm{Fe}}} \tag*{(8)}

where ρoxide\rho_{\mathrm{oxide}} is the average density of the oxide layer. The initial mass fraction of iron oxides is an experimentally measurable quantity.

The total enthalpy (or internal energy) of the particle HpH_{\mathrm{p}} is the sum of the enthalpy of each solid-phase species,

Hp=∑iNsmihi(Tp)Wi(9)H_{\mathrm{p}}=\sum_{i}^{N_{\mathrm{s}}}\frac{m_i h_i(T_{\mathrm{p}})}{W_i} \tag*{(9)}

where NsN_{\mathrm{s}} is the number of solid-phase species, ii is the index of each species, and WiW_i is the molar weight of the ithi^{\mathrm{th}} species, and hih_i is the corresponding molar enthalpy as a function of temperature described by the Shomate equation based on the NIST Standard Reference Data [57]. Knowing the value of HpH_{\mathrm{p}} and mim_i, particle temperature TpT_{\mathrm{p}} can be determined via Eq. (9) following an iterative root-finding procedure. The energy balance equation of the particle tracks the rate of change in HpH_{\mathrm{p}} as follows,

dHpdt=qFeOdmFeOdt+qFe3O4dmFe3O4dt−hO2(Tg)WO2dmO2dt−Ap[hp(Tp−Tg)+σϵ(Tp4−Tg4)](10)\frac{\mathrm{d}H_{\mathrm{p}}}{\mathrm{d}t}=q_{\mathrm{FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}+q_{\mathrm{Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}-\frac{h_{\mathrm{O_2}}(T_{\mathrm{g}})}{W_{\mathrm{O_2}}}\frac{\mathrm{d}m_{\mathrm{O_2}}}{\mathrm{d}t}-A_{\mathrm{p}}\left[h_{\mathrm{p}}(T_{\mathrm{p}}-T_{\mathrm{g}})+\sigma\epsilon(T_{\mathrm{p}}^4-T_{\mathrm{g}}^4)\right] \tag*{(10)}

The first two terms on the right-hand side of Eq. (10) are the energy release rates due to the formation of FeO and Fe3_3O4_4, respectively, where qFeOq_{\mathrm{FeO}} and qFe3O4q_{\mathrm{Fe_3O_4}} are values of the specific energy release that are related to the enthalpy of formation Hf∘H^\circ_{\mathrm{f}} of the oxide products as qFeO=−ΔHf,FeO(S)∘/WFeOq_{\mathrm{FeO}}=-\Delta H^\circ_{\mathrm{f,FeO(S)}}/W_{\mathrm{FeO}} and qFe3O4=−ΔHf,Fe3O4(S)∘/WFe3O4q_{\mathrm{Fe_3O_4}}=-\Delta H^\circ_{\mathrm{f,Fe_3O_4(S)}}/W_{\mathrm{Fe_3O_4}}. The third term in Eq. (10) accounts for the increase in the enthalpy of the particle due to the incorporation of the mass of gaseous oxygen2^{2}, wherein hO2h_{\mathrm{O_2}} is the molar enthalpy of O2_2 given by the NASA Thermodynamics Database [58]. The mass consumption rate of O2_2 is calculated as follows,

dmO2dt=−νO2/FeOdmFeOdt−νO2/Fe3O4dmFe3O4dt(11)\frac{\mathrm{d}m_{\mathrm{O_2}}}{\mathrm{d}t}=-\nu_{\mathrm{O_2/FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}-\nu_{\mathrm{O_2/Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t} \tag*{(11)}

where νO2/FeO\nu_{\mathrm{O_2/FeO}} and νO2/Fe3O4\nu_{\mathrm{O_2/Fe_3O_4}} are the stoichiometric mass ratios.

The fourth and fifth terms in Eq. (10) are the rates of convective and radiative heat transfer between the particle and the ambient gas, respectively. The coefficient of convective heat transfer of the particle hph_{\mathrm{p}} can be determined as follows,

hp=Nuλgdp(12)h_{\mathrm{p}}=\frac{\mathrm{Nu}\lambda_{\mathrm{g}}}{d_{\mathrm{p}}} \tag*{(12)}

where Nu\mathrm{Nu} and dpd_{\mathrm{p}} are the Nusselt number and diameter of the particle, respectively, and λg\lambda_{\mathrm{g}} is the thermal conductivity of the ambient gas mixture of O2_2 and an inert species (i.e., N2_2, Ar, He, and Xe considered in this analysis). The thermal conductivity of each gaseous species λi\lambda_i as a function of temperature can be determined using the NASA polynomials [58]. The transport properties of the gaseous species are evaluated in the particle-gas boundary layer, TsfT_{\mathrm{sf}}. A “two-third law” [59, 60] is used to estimate TsfT_{\mathrm{sf}}, i.e., Tsf=(2Tp+Tg)/3T_{\mathrm{sf}}=(2T_{\mathrm{p}}+T_{\mathrm{g}})/3, in this paper.2 The mixture-averaged thermal conductivity λg\lambda_{\mathrm{g}} can then be calculated as follows,

λg=12(∑iNgXiλi+1∑iNgXi/λi)(13)\lambda_{\mathrm{g}}=\frac{1}{2}\left(\sum_i^{N_{\mathrm{g}}}X_i\lambda_i+\frac{1}{\sum_i^{N_{\mathrm{g}}}X_i/\lambda_i}\right) \tag*{(13)}

where NgN_{\mathrm{g}} is the number of species in the gaseous mixture and XiX_i is the mole fraction of the ithi^{\mathrm{th}} species. The Nusselt number can be estimated using the Frössling (or Ranz-Marshal) correlation:

Nu=2+0.552Re1/2Pr1/3.(14)\mathrm{Nu}=2+0.552\mathrm{Re}^{1/2}\mathrm{Pr}^{1/3}. \tag*{(14)}

As a result of Assumption 4, the Nusselt number equals to two (i.e., Nu=2\mathrm{Nu}=2) for a spherical object in a quiescent medium, and thus, hp=λg/rph_{\mathrm{p}}=\lambda_{\mathrm{g}}/r_{\mathrm{p}}. Note that, although the term hp(Tp−Tg)h_{\mathrm{p}}(T_{\mathrm{p}}-T_{\mathrm{g}}) describes the rate of convective heat transfer, it is referred to as “conductive heat loss” when discussing the results of the current analysis given the assumption of a quiescent gas medium. In the radiative heat transfer term, σ\sigma and ϵ\epsilon denote the Stefan-Boltzmann constant and the emissivity of the oxide layer (Fe3_3O4_4) covering the particle surface, respectively. A constant value of ϵ=0.88\epsilon=0.88 is considered in this analysis as the total emissivity of magnetite varies slightly from 0.85 to 0.89 as temperature increases from 773 K to 1473 K [61]. Note that, since the total emissivity of hematite varies from 0.75 to 0.85 over a range in temperature from 850 K to 1500 K [62, 63], the current model may slightly overestimate the radiative heat transfer rate of an iron particle covered by a thin layer of Fe2_2O3_3. The values of the key properties of the solid species are provided in Table 2.

PropertyValueUnit
ρFe\rho_{\mathrm{Fe}}7874kg m−3\mathrm{kg}\,\mathrm{m}^{-3}
ρFeO\rho_{\mathrm{FeO}}5745kg m−3\mathrm{kg}\,\mathrm{m}^{-3}
ρFe3O4\rho_{\mathrm{Fe_3O_4}}5170kg m−3\mathrm{kg}\,\mathrm{m}^{-3}
qFeOq_{\mathrm{FeO}}3.787MJ kg−1\mathrm{MJ}\,\mathrm{kg}^{-1}
qFe3O4q_{\mathrm{Fe_3O_4}}4.841MJ kg−1\mathrm{MJ}\,\mathrm{kg}^{-1}
ϵ\epsilon0.88-

Table 2. Properties of Fe, FeO, and Fe3_3O4_4

Model formulation for a suspension of iron particles

This model considers a spherical cloud of iron particles suspended in a gas mixture of O2_2 and an inert species at constant pressure. The radius of this spherical suspension (denoted as rsr_{\mathrm{s}} shown in Figure 3) will increase as a result of thermal expansion of the gas mixture. The model monitors the mass and heat transfer between the particles and the gas mixture (as described by the model for an isolated particle in Sect. 3.1) and the heat transfer between the gas mixture and the surrounding gas at a temperature T∞T_{\infty} outside the suspension. Several simplifying assumptions are made in order to better focus this study on the key parameters controlling the ignition characteristics of an iron-particle suspension:

  1. A monodisperse suspension, i.e., all of the particles are of the same initial size.

  2. The total volume occupied by the particles is negligible compared to the total volume of the suspension.

  3. Radiative heat transfer is neglected.

  4. The suspension expands freely at a constant pressure of 1 atm.

  5. No slip between the expanding gas mixture and particles, i.e., the spatial distribution of particles expands with the gas mixture so that the volume of the entire suspension increases due to thermal expansion of the gas.

  6. Flow velocity of the surrounding gas relative to the suspension is zero.

Knowing the initial radius of each iron particle, rFe,0r_{\mathrm{Fe},0}, the initial, total mass of iron particles in a suspension, mFe,tot,0m_{\mathrm{Fe,tot},0}, can be related to the number of particles NpN_{\mathrm{p}} as follows:

mFe,tot,0=4πNpρFerFe,033.(15)m_{\mathrm{Fe,tot},0} = \frac{4\pi N_{\mathrm{p}}\rho_{\mathrm{Fe}}r_{\mathrm{Fe},0}^{3}}{3}. \tag*{(15)}

Although the overall stoichiometry of an iron-oxygen mixture is not uniquely defined, due to multiple possible oxidation states, an equivalence ratio considering FeO as the only product can be defined as follows:

ϕFeO=mFe,tot,0/mO2,02WFe/WO2.(16)\phi_{\mathrm{FeO}} = \frac{m_{\mathrm{Fe,tot},0}/m_{\mathrm{O_2},0}}{2W_{\mathrm{Fe}}/W_{\mathrm{O_2}}}. \tag*{(16)}

The initial mass of O2_2 in the suspension mO2,0m_{\mathrm{O_2},0} is related to the initial temperature, Tg,0T_{\mathrm{g},0}, pressure, p0p_0, and O2_2 mass fraction YO2,0Y_{\mathrm{O_2},0}:

mO2,0=Vs,0ρg,0YO2,0=p0Vs,0YO2,0RgTg,0,(17)m_{\mathrm{O_2},0} = V_{\mathrm{s},0}\rho_{\mathrm{g},0}Y_{\mathrm{O_2},0} = \frac{p_0V_{\mathrm{s},0}Y_{\mathrm{O_2},0}}{R_{\mathrm{g}}T_{\mathrm{g},0}}, \tag*{(17)}

where RgR_{\mathrm{g}} is the specific gas constant of the gas mixture, ρg,0\rho_{\mathrm{g},0} is the initial gas density, and Vs,0V_{\mathrm{s},0} is the initial suspension volume. Thus, by specifying Tg,0T_{\mathrm{g},0}, p0p_0, YO2,0Y_{\mathrm{O_2},0}, rFe,0r_{\mathrm{Fe},0}, rs,0r_{\mathrm{s},0}, and ϕFeO\phi_{\mathrm{FeO}}, one can calculate NpN_{\mathrm{p}}, mFe,tot,0m_{\mathrm{Fe,tot},0}, and mO2,0m_{\mathrm{O_2},0} via Eqs. 15-17 as input parameters to this suspension model.

Based on the above-listed assumptions, the gas-phase energy balance equation of a suspension is formulated as follows:

dHgdt=NpAphp(Tp−Tg)+Ashs(T∞−Tg).(18)\frac{\mathrm{d}H_{\mathrm{g}}}{\mathrm{d}t} = N_{\mathrm{p}}A_{\mathrm{p}}h_{\mathrm{p}}(T_{\mathrm{p}}-T_{\mathrm{g}}) + A_{\mathrm{s}}h_{\mathrm{s}}(T_{\infty}-T_{\mathrm{g}}). \tag*{(18)}

The total enthalpy of the gas mixture HgH_{\mathrm{g}} is the sum of the enthalpies of O2_2 and an inert species is,

Hg=mO2hO2(Tg)WO2+mIhI(Tg)WI,(19)H_{\mathrm{g}} = \frac{m_{\mathrm{O_2}}h_{\mathrm{O_2}}(T_{\mathrm{g}})}{W_{\mathrm{O_2}}} + \frac{m_{\mathrm{I}}h_{\mathrm{I}}(T_{\mathrm{g}})}{W_{\mathrm{I}}}, \tag*{(19)}

where the subscript “I” denotes the inert species (i.e., N2_2, Ar, He, and Xe considered in this study). The mass balance equation of O2_2 in the suspension is formulated as:

dmO2dt=−Np(νO2/FeOdmFeOdt+νO2/Fe3O4dmFe3O4dt).(20)\frac{\mathrm{d}m_{\mathrm{O_2}}}{\mathrm{d}t} = -N_{\mathrm{p}}\left(\nu_{\mathrm{O_2/FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t} + \nu_{\mathrm{O_2/Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}\right). \tag*{(20)}

As the gas temperature, TgT_{\mathrm{g}}, and the mass of O2_2 change over time, the volume of the suspension changes according to the ideal gas law. Note that an isobaric condition at 1 atm is considered in all of the calculations with the suspension model.

Sample results

An isolated particle

Sample results of the calculations considering an isolated iron particle with an initial diameter dp,0=20 μmd_{\mathrm{p},0} = 20\,\mu\mathrm{m} and an initial oxide layer thickness X0=XFeO,0+XFe3O4,0=10 nmX_0 = X_{\mathrm{FeO},0} + X_{\mathrm{Fe_3O_4},0} = 10\,\mathrm{nm} (δ0=10−3\delta_0 = 10^{-3}) in air (XO2:XN2=0.21:0.79X_{\mathrm{O_2}}:X_{\mathrm{N_2}}=0.21:0.79) are plotted in Figure 4. The subfigures (a) and (b) show the time histories of particle temperature TpT_{\mathrm{p}} and the thicknesses of the FeO layer (solid curves) and Fe3_3O4_4 layer (dashed curves), respectively. Each curve corresponds to a case with a different constant gas temperature TgT_{\mathrm{g}}. Particle temperature is initialized to be equal to TgT_{\mathrm{g}} for all of the cases. The initial thickness of the FeO layer is 95% of X0X_0. For the cases with Tg≤1081 KT_{\mathrm{g}} \le1081\,\mathrm{K}, as shown in Figure 4(a), TpT_{\mathrm{p}} non-monotonically varies over time, exhibiting an increase to a peak value followed by an asymptotic decrease to TgT_{\mathrm{g}}. Figure. Figure 4(b) shows that the oxide layer thicknesses for Tg≤1081 KT_{\mathrm{g}} \le1081\,\mathrm{K} reach a quasi-plateau after an initial increase until approximately 0.01 s. For the case with Tg=1082 KT_{\mathrm{g}} = 1082\,\mathrm{K}, TpT_{\mathrm{p}} increases monotonically with time as shown in Figure 4(a). This increase in temperature becomes abrupt after approximately 0.01 s. The calculation was stopped once TpT_{\mathrm{p}} reaches the melting point of FeO (i.e., 1650 K). At this stopping point, the slopes of the XFeOX_{\mathrm{FeO}} and XFe3O4X_{\mathrm{Fe_3O_4}} curves (shown in Figure 4(b)) indicate that oxide layer thicknesses continue to increase instead of asymptotically approaching a plateau value.

Time histories of particle temperature and oxide layer thicknesses

Figure 4. Sample results showing the time histories of (a) particle temperature TpT_{\mathrm{p}} and (b) the thicknesses of FeO layer (solid curves) and Fe3_3O4_4 layer (dashed curves) at four different gas temperatures (i.e., from bottom to top, Tg=1060 KT_{\mathrm{g}} = 1060\,\mathrm{K}, 1075 K, 1081 K, and 1082 K) for an initial particle size dp,0=20 μmd_{\mathrm{p},0} = 20\,\mu\mathrm{m} with an initial oxide layer thickness of X0=XFeO,0+XFe3O4,0=10 nmX_0 = X_{\mathrm{FeO},0} + X_{\mathrm{Fe_3O_4},0} = 10\,\mathrm{nm} in air (XO2:XN2=0.21:0.79X_{\mathrm{O_2}}:X_{\mathrm{N_2}}=0.21:0.79).

Suspension of particles

Sample suspension model results shown in Figure 5 are for the cases with an initial suspension radius rs,0=1r_{\mathrm{s},0} = 1 cm, particle size dp,0=20d_{\mathrm{p},0} = 20 µm, and oxide layer thickness of X0=10X_0 = 10 nm in air at equivalence ratios ϕFeO=1\phi_{\mathrm{FeO}} = 1, 0.25, and 0.1. The particle and gas temperatures of a suspension are initialized to be equal to the surrounding gas temperature T∞T_{\infty}, and are plotted as dashed and solid curves in Figure 5, respectively. Each pair of TpT_{\mathrm{p}} and TgT_{\mathrm{g}} curves correspond to a case with a constant T∞T_{\infty}. The curve of TpT_{\mathrm{p}} is slightly above that of TgT_{\mathrm{g}} for all of the cases. For the cases with ϕFeO=1\phi_{\mathrm{FeO}} = 1 as shown in Figure 5(a), T∞≤904T_{\infty} \le904 K, the resulting temperatures vary non-monotonically over time; for the cases with T∞≥905T_{\infty} \ge905 K, TpT_{\mathrm{p}} and TgT_{\mathrm{g}} increase abruptly after a gradual increase over early times. For the cases with an increasingly smaller ϕFeO\phi_{\mathrm{FeO}}, the separation between TpT_{\mathrm{p}} and TgT_{\mathrm{g}} increases. For the cases with ϕFeO=0.1\phi_{\mathrm{FeO}} = 0.1 as shown in Figure 5(c), TpT_{\mathrm{p}} increases much more rapidly than TgT_{\mathrm{g}} does. In the case with T∞=1079T_{\infty} = 1079 K, while TpT_{\mathrm{p}} promptly increases, TgT_{\mathrm{g}} slightly increases from the surrounding temperature.

Sample results showing the time histories of gas and particle temperatures

Figure 5. Sample results showing the time histories of gas temperature TgT_g (solid curves) and particle temperature TpT_p (dashed curves) of a suspension of iron particles with an initial suspension radius rs,0=1r_{s,0}=1 cm, particle size dp,0=20d_{p,0}=20 μm, and oxide layer thickness of X0=10X_0=10 nm in air (XO2:XN2=0.21:0.79X_{\mathrm{O}_2}:X_{\mathrm{N}_2}=0.21:0.79) at equivalence ratios (a) ϕFeO=1\phi_{\mathrm{FeO}}=1, (b) ϕFeO=0.25\phi_{\mathrm{FeO}}=0.25, and (c) ϕFeO=0.1\phi_{\mathrm{FeO}}=0.1 subjected to different surrounding gas temperatures.

Analysis and discussion

Critical condition of thermal runaway

A gradual increase in temperature followed by a rapid upsurge (as identified in some cases shown in Sect. 4) indicates a particle or suspension undergoing a thermal runaway. The bifurcation between the “go” and “no-go” cases is identified as a critical phenomenon of ignition. Based on the time histories resulting from the isolated particle and suspension models, critical values of TgT_{\mathrm{g}} and T∞T_{\infty} above which an isolated particle or a suspension undergoes a thermal runaway, respectively, can be determined. This critical temperature for ignition is referred to as the ignition temperature and denoted as TignT_{\mathrm{ign}}. In this study, TignT_{\mathrm{ign}} is used as a metric to probe the effects of various factors contributing to the ignition process of iron particles. Note that the radiative heat transfer term in Eq. (10) is neglected in most of the analysis shown in this section, with an exception of Sect. 5.4 wherein the effect of radiation on iron particle ignition is examined.

Effect of initial particle size and oxide layer thickness

The results of TignT_{\mathrm{ign}} are plotted in Figure 6 as a function of initial particle diameter, dp,0d_{\mathrm{p},0}, with three different initial oxide layer thicknesses for an isolated particle in air. The smallest particle size considered in this set of results is dp,0=1d_{\mathrm{p},0} = 1 µm. An ignition temperature significantly below the melting point of FeO is found for all particle sizes above 1 µm over a broad range of initial oxide layer thickness from 1 nm to 0.1 µm. This result suggests that micron-sized or larger iron particles, which were used in most of the experimental studies on iron-fuel flames [22, 24, 25, 26, 27, 28, 35], very likely undergo an ignition process—transitioning from a kinetic-controlled reaction to an external-diffusion-controlled combustion—rather than a slow, kinetic-controlled burnout at the ambient gas temperature.

Critical temperature for ignition as a function of initial particle diameter

Figure 6. Critical temperature for ignition, TignT_{\mathrm{ign}}, as a function of initial particle diameter, dp,0d_{\mathrm{p},0}, with three different initial oxide layer thicknesses, i.e., X0=0.1 μmX_0 = 0.1\ \mu\mathrm{m} (solid red curve), X0=10 nmX_0 = 10\ \mathrm{nm} (dashed maroon curve), and X0=1 nmX_0 = 1\ \mathrm{nm} (dash-dotted black curve). Results from the current model for an isolated particle in air are compared to the result of a model [60] neglecting the obstructing effect of oxide layer on the oxidation rate (plotted as the dotted blue curve).

For the case with X0=0.1X_0 = 0.1 µm (the solid red curve in Figure 6), there is a significant decrease in TignT_{\mathrm{ign}} as dp,0d_{\mathrm{p},0} increases from 1 µm to approximately 20 µm; TignT_{\mathrm{ign}} nearly plateaus for further larger particles. Although qualitatively the same trend is identified for the case with a much thinner initial oxide layer of X0=10X_0 = 10 nm (dashed maroon curve), TignT_{\mathrm{ign}} decreases less significantly with an increase in dp,0d_{\mathrm{p},0} and reaches nearly the same plateau value (≈1082\approx1082 K) as that resulting from the case with X0=0.1X_0 = 0.1 µm. For even thinner initial oxide layer of X0=1X_0 = 1 nm (dash-dotted black curve), the entire curve of TignT_{\mathrm{ign}} plateaus at 1082 K, independent of initial particle size. This independence of TignT_{\mathrm{ign}} on particle size (for sufficiently large particles) has been identified as a result of a parabolic oxidation rate law in Khaikin et al. as [45] generic analysis of metal particle ignition, which is qualitatively explained as follows: For increasingly large particles, the increase in energy release rate (proportional to particle surface area, i.e., ∝dp,02\propto d_{\mathrm{p},0}^{2}) surpasses the increase in the rate of conductive heat removal (proportional to particle size, i.e., ∝dp,0\propto d_{\mathrm{p},0}). As a result, larger particles are more prone to ignition. If a parabolic oxidation rate law is considered, however, to heat up larger particles requires a greater amount of energy release associated with a thicker oxide layer grown, in return, impeding the oxidation rate. Hence, a large particle size has both enhancing and hindering effects on the ignition propensity of an iron particle. These mutually compensating effects result in the independence of TignT_{\mathrm{ign}} on particle size for sufficiently large particles governed by a parabolic oxidation rate law. As a comparison, the result of a model (recently proposed by Hazenberg and Oijen [60]) neglecting the impeding effect of an oxide layer on the oxidation rate is plotted as the dotted blue curve in Figure 6. The resulting TignT_{\mathrm{ign}} persistently decreases as dp,0d_{\mathrm{p},0} increases, manifesting only the enhancing effect of an increase in particle size. Therefore, prudence must be practiced when interpreting the flame simulation results using this likely oversimplified reaction model [60].

Further analysis is performed to understand the effect of initial oxide layer thickness. The criterion for thermal runaway is that the rate of energy release from iron oxidation, q˙R\dot{q}_{\mathrm{R}}, must exceed the rate of heat loss to the surrounding gas, q˙L\dot{q}_L, and is formulated as follows (neglecting radiative heat transfer):

q˙R>q˙L\dot{q}_R > \dot{q}_L
qFeOdmFeOdt+qFe3O4dmFe3O4dt>Aphp(Tp−Tg).(21)q_{\mathrm{FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}+q_{\mathrm{Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}>A_p h_p(T_p-T_g). \tag*{(21)}

Note that the rate of energy increase due to the incorporation of gaseous oxygen (i.e., the third term on the right-hand side of Eq. (10)) is also neglected in this simplified criterion. This simplification is based on the fact that the specific enthalpy associated with the incorporated O2\mathrm{O}_2 is insignificant compared to the specific energy release of FeO\mathrm{FeO} and Fe3O4\mathrm{Fe}_3\mathrm{O}_4 formation. Considering the kinetic rate laws (Eqs. (5) and (6)) and the definition of δ\delta and δFe3O4\delta_{\mathrm{Fe_3O_4}} (Eq. (7)), the thermal runaway criterion can be written as:

q~FeOExp⁡(−Ta,FeOTp)(1−δδFe3O4)2rpδ−δδFe3O4+q~Fe3O4Exp⁡(−Ta,Fe3O4Tp)rpδδFe3O4>λg(Tp−Tg)rp,(22)\tilde{q}_{\mathrm{FeO}} \operatorname{Exp}\left(\frac{-T_{\mathrm{a,FeO}}}{T_{\mathrm{p}}}\right)\frac{\left(1-\delta\delta_{\mathrm{Fe_3O_4}}\right)^2 r_{\mathrm{p}}}{\delta-\delta\delta_{\mathrm{Fe_3O_4}}}+\tilde{q}_{\mathrm{Fe_3O_4}} \operatorname{Exp}\left(\frac{-T_{\mathrm{a,Fe_3O_4}}}{T_{\mathrm{p}}}\right)\frac{r_{\mathrm{p}}}{\delta\delta_{\mathrm{Fe_3O_4}}}>\lambda_{\mathrm{g}}(T_{\mathrm{p}}-T_{\mathrm{g}})r_{\mathrm{p}}, \tag*{(22)}

where q~FeO=ρFeOqFeOk0,FeO\tilde{q}_{\mathrm{FeO}}=\rho_{\mathrm{FeO}}q_{\mathrm{FeO}}k_{0,\mathrm{FeO}} and q~Fe3O4=ρFe3O4qFe3O4k0,Fe3O4\tilde{q}_{\mathrm{Fe_3O_4}}=\rho_{\mathrm{Fe_3O_4}}q_{\mathrm{Fe_3O_4}}k_{0,\mathrm{Fe_3O_4}}. Equation 22 shows that the energy release rates (left-hand side) and conductive heat loss rate (right-hand side) both scale linearly with particle radius, rpr_{\mathrm{p}}. Thus, the thermal runaway criterion is independent of particle size when only conductive heat transfer is considered in the loss rate. As a result, TignT_{\mathrm{ign}} exclusively depends on the initial oxide layer to particle radius ratio, δ0\delta_0, for a given gas mixture (while assuming a constant value of δFe3O4,0=0.05\delta_{\mathrm{Fe_3O_4},0}=0.05) as shown in Figure 7(a).

Ignition temperature and steady-state curves for an isolated iron particle

Figure 7. (a) Ignition temperature, TignT_{\mathrm{ign}}, as a function of δ0\delta_0, the ratio between initial oxide layer thickness and particle radius for an isolated iron particle in air. The results of the steady-state analysis based on the Semenov ignition criterion and the unsteady model are plotted as the dashed red curve and solid black curve, respectively. (b) The curves of TpT_p and TgT_g satisfying q˙R=q˙L\dot{q}_R = \dot{q}_L in the steady-state analysis with various δ0\delta_0. The turning point of each curve marks the critical condition for ignition.

A simple steady-state analysis—neglecting the growth of the oxide layers over time—was first performed to estimate TignT_{\mathrm{ign}} based on the Semenov ignition criterion (i.e., solving for TpT_{\mathrm{p}} and TgT_{\mathrm{g}} satisfying q˙R=q˙L\dot{q}_{\mathrm{R}}=\dot{q}_{\mathrm{L}} and dq˙R/dTp=dq˙L/dTpd\dot{q}_{\mathrm{R}}/dT_{\mathrm{p}}=d\dot{q}_{\mathrm{L}}/dT_{\mathrm{p}}). The relation between TpT_{\mathrm{p}} and TgT_{\mathrm{g}} satisfying q˙R=q˙L\dot{q}_{\mathrm{R}}=\dot{q}_{\mathrm{L}} is plotted as the curves in Figure 7(b) for various δ0\delta_0. Each curve exhibits a turning-point behavior. The TpT_{\mathrm{p}} and TgT_{\mathrm{g}} at the turning point are found as the critical condition for ignition (i.e., satisfying dq˙R/dTp=dq˙L/dTpd\dot{q}_{\mathrm{R}}/dT_{\mathrm{p}}=d\dot{q}_{\mathrm{L}}/dT_{\mathrm{p}}). The branches below and above the turning point correspond to stable and unstable solutions for q˙R=q˙L\dot{q}_{\mathrm{R}}=\dot{q}_{\mathrm{L}}, respectively. As shown in Figure 7(b), the turning point or TignT_{\mathrm{ign}} increases with an increase in δ0\delta_0. This steady-state analysis prediction is also plotted in Figure 7(a) as the dashed red curve. This trend is due to the fact that reaction rate is greater for an increasing oxide layer, as described by the parabolic rate law. A noteworthy finding of the steady-state analysis is that, as shown in Figure 7(b), the ignition point can be identified for all values of δ0\delta_0 regardless of the particle size. In other words, there is no minimum size of an iron particle below which the particle cannot be ignited, i.e., ignition degenerates and the particle burnout is limited to the kinetic-controlled regime without undergoing a thermal runaway or transitioning to an external-diffusion-controlled regime. This finding is in contrast to the previous work by Soo et al. [64], who found a minimum particle size for ignition, using a generic combustion model of metal particles that did not consider the effect of the oxide layer described by a parabolic rate law.

As shown in Figure 7(a), TignT_{\mathrm{ign}} resulting from the unsteady model, which considers the thickening of the oxide layer over time, is significantly greater than the steady-state estimation and plateaus at approximately 1080 K for δ0≤0.003\delta_0\leq0.003, as indicated by the gray vertical line in Figure 7(a). To understand this plateau behavior, the transient process of oxide-layer growth must be considered.

The time histories of particle temperature and oxide-layer growth resulting from the unsteady model for an isolated particle in air with dp,0=20 μmd_{p,0} = 20\ \mu\mathrm{m} and different initial oxide-thickness ratios subjected to two gas temperatures (Tg=1070 KT_g = 1070\ \mathrm{K} and Tg=1082 KT_g = 1082\ \mathrm{K}) are shown in Figure 8. The subfigures (a) and (b) correspond to the cases with a gas temperature Tg=1070 KT_g = 1070\ \mathrm{K} that is slightly below the plateau value of Tign=1080 KT_{\mathrm{ign}} = 1080\ \mathrm{K} marked in Figure 7. None of these cases, with various δ0\delta_0, results in a thermal runaway. As shown in Figure 8(b), δ\delta increases the most rapidly in the case with δ0=10−4\delta_0 = 10^{-4} and collapses onto the curve of δ0=10−3\delta_0 = 10^{-3} upon reaching δ≈0.003\delta\approx0.003 (indicated by the gray horizontal line) around t=1×10−4 st = 1 \times10^{-4}\ \mathrm{s}. Figures 8(c) and (d) are for the cases with Tg=1082 KT_g = 1082\ \mathrm{K}, which is slightly greater than the plateau TignT_{\mathrm{ign}}. The cases with δ0=10−4\delta_0 = 10^{-4} and δ0=10−3\delta_0 = 10^{-3} result in a thermal runaway while the other two cases, with greater values of δ0\delta_0, do not. Note that the curves of δ\delta for the cases with δ0=10−4\delta_0 = 10^{-4} and δ0=10−3\delta_0 = 10^{-3} shown in Figure 8(d) are cut off because the calculations were stopped once the melting point of FeO is reached. Again, the δ\delta growth curves for the cases with δ0=10−4\delta_0 = 10^{-4} and δ0=10−3\delta_0 = 10^{-3} collapse upon reaching δ≈0.003\delta\approx0.003 around t=1×10−4 st = 1 \times10^{-4}\ \mathrm{s}. By comparing the cases with Tg=1070 KT_g = 1070\ \mathrm{K} and Tg=1082 KT_g = 1082\ \mathrm{K}, the characteristic time scale of thermal runaway is identified to be greater than 1×10−3 s1 \times10^{-3}\ \mathrm{s}. The order of magnitude of this time scale suggests that, for increasingly thin initial oxide layers below δ0≈0.003\delta_0 \approx0.003, a rapid oxide-layer growth up to a time scale of 1×10−4 s1 \times10^{-4}\ \mathrm{s} has no effect on the later processes determining whether a thermal runaway occurs. For sufficiently thin initial oxide layers, TignT_{\mathrm{ign}} is thus independent of δ0\delta_0.

Time histories of particle temperature and oxide layer thickness ratio

Figure 8. The time histories of particle temperature, TpT_{\mathrm{p}} [(a) and (c)] and the ratio between oxide layer thickness and particle radius, δ\delta [(b) and (d)], for an isolated iron particle in air with dp,0=20 μmd_{\mathrm{p},0}=20\,\mu\mathrm{m} and four different initial oxide thickness ratios, i.e., δ0=10−4\delta_0=10^{-4} (dotted), δ0=10−3\delta_0=10^{-3} (dash-dotted), δ0=10−2\delta_0=10^{-2} (dashed), and δ0=10−1\delta_0=10^{-1} (solid), subjected to two gas temperatures, i.e., Tg=1070 KT_{\mathrm{g}}=1070\,\mathrm{K} for (a) and (b), and Tg=1082 KT_{\mathrm{g}}=1082\,\mathrm{K} for (c) and (d).

Effect of inert gas species

The current model considers the fact that the kinetics of iron oxidation is controlled by solid-phase diffusion and is independent of gas-phase O2\mathrm{O}_2 concentration (at least, for pO2>4×10−4 atmp_{\mathrm{O}_2} > 4 \times10^{-4}\ \mathrm{atm}). This means that, as the gaseous composition is varied, the ignition characteristics of an iron particle are only influenced by the change in the thermal conductivity of the ambient gas. The TignT_{\mathrm{ign}} resulting from the unsteady model for an isolated particle with δ0=10−3\delta_0 = 10^{-3}, in a gas mixture with O2\mathrm{O}_2 and different inert species, is plotted as a function of the mole fraction of O2\mathrm{O}_2 in Figure 9. The result of TignT_{\mathrm{ign}} for the case with N2\mathrm{N}_2 (plotted as the solid curve) does not change significantly as XO2X_{\mathrm{O}_2} varies, owing to the fact that the thermal conductivities of diatomic gases O2\mathrm{O}_2 and N2\mathrm{N}_2 are very close. For the case with He, the resulting TignT_{\mathrm{ign}} increases with a decrease in XO2X_{\mathrm{O}_2}. This trend is due to the fact that He has a higher thermal conductivity than that of O2\mathrm{O}_2. A greater mole fraction of He thus enhances the rate of conductive heat loss from the particle to the gas mixture, resulting in a greater TignT_{\mathrm{ign}}. On the other hand, the TignT_{\mathrm{ign}} for the cases with Ar and Xe decrease significantly as oxygen mole fraction decreases, since Ar and Xe have an increasingly low thermal conductivity. A reduced thermal conductivity of the ambient gas has a heat preserving effect on a reacting particle, thereby facilitating thermal runaway.

Ignition temperature as a function of oxygen mole fraction

Figure 9. The ignition temperature TignT_{\mathrm{ign}} of an isolated iron particle with δ0=10−3\delta_0=10^{-3} as a function of mole fraction of oxygen XO2X_{\mathrm{O}_2} in a bulk gas mixture with different inert species, i.e., He (dotted), N2_2 (solid), Ar (dashed), and Xe (dash-dotted).

Effect of radiative heat transfer

If the heat loss from a particle to the surrounding gas via radiative heat transfer is considered, the criterion of thermal runaway (formulated as Eqs. (21) and (22)) is augmented as follows:

qFeOdmFeOdt+qFe3O4dmFe3O4dt>Aphp(Tp−Tg)+Apσ(Tp4−Tg4)q_{\mathrm{FeO}}\frac{\mathrm{d}m_{\mathrm{FeO}}}{\mathrm{d}t}+q_{\mathrm{Fe_3O_4}}\frac{\mathrm{d}m_{\mathrm{Fe_3O_4}}}{\mathrm{d}t}>A_{\mathrm{p}}h_{\mathrm{p}}(T_{\mathrm{p}}-T_{\mathrm{g}})+A_{\mathrm{p}}\sigma(T_{\mathrm{p}}^4-T_{\mathrm{g}}^4)
q~FeOexp⁡(−Ta,FeOTp)(1−δδFe3O4)rp2δ−δδFe3O4+q~Fe3O4exp⁡(−Ta,Fe3O4Tp)rp2δδFe3O4>λg(Tp−Tg)rp+σ(Tp4−Tg4)rp2.(23)\tilde{q}_{\mathrm{FeO}}\exp\left(-\frac{T_{\mathrm{a,FeO}}}{T_{\mathrm{p}}}\right)\frac{(1-\delta\delta_{\mathrm{Fe_3O_4}})r_{\mathrm{p}}^2}{\delta-\delta\delta_{\mathrm{Fe_3O_4}}}+\tilde{q}_{\mathrm{Fe_3O_4}}\exp\left(-\frac{T_{\mathrm{a,Fe_3O_4}}}{T_{\mathrm{p}}}\right)\frac{r_{\mathrm{p}}^2}{\delta\delta_{\mathrm{Fe_3O_4}}}>\lambda_{\mathrm{g}}(T_{\mathrm{p}}-T_{\mathrm{g}})r_{\mathrm{p}}+\sigma(T_{\mathrm{p}}^4-T_{\mathrm{g}}^4)r_{\mathrm{p}}^2. \tag*{(23)}

Note that the radiative heat loss rate scales quadratically with particle size while the energy release rates and conductive heat loss rate scale linearly with rpr_{\mathrm{p}}. The heat removal effect due to radiation is thus enhanced for increasingly large particles. This effect is reflected by the results of TignT_{\mathrm{ign}} as a function of δ0\delta_0, for the cases with various initial particle sizes, and TignT_{\mathrm{ign}} as a function of dp,0d_{\mathrm{p},0}, for the cases with different initial oxide layer thicknesses, as shown in Figure 10(a) and (Figure 10b), respectively. In Figure 10(a), the resulting TignT_{\mathrm{ign}} for the cases with radiative heat loss are greater than the TignT_{\mathrm{ign}} for the case neglecting radiation (the solid curve), and increase with an increasingly large dp,0d_{\mathrm{p},0} for all values of δ0\delta_0. For a fixed initial oxide layer thickness X0X_0 and relatively small particle sizes, as shown in Figure 10(b), TignT_{\mathrm{ign}} decreases as dp,0d_{\mathrm{p},0} increases. The results for the cases with (thick curves in Figure 10(b)) and without (thin curves) radiative heat loss are indistinguishable below approximately dp,0=10d_{\mathrm{p},0}=10 μm. For relatively large particles, e.g., dp,0≥60d_{\mathrm{p},0}\geq60 μm as shown in the inset of Figure 10(b), TignT_{\mathrm{ign}} resulting from the cases with radiative heat loss slightly increases with an increase in dp,0d_{\mathrm{p},0} while the TignT_{\mathrm{ign}} without radiation effect slightly decrease or plateau. As demonstrated by the results in Figure 10, the difference in TignT_{\mathrm{ign}} between the cases with and without radiative heat loss is less than 20 K, i.e., a relative discrepancy less than 2%, suggesting that radiative heat loss has a minor effect on the ignition characteristics of fine iron particles.

Ignition temperature of isolated iron particles with and without radiative heat transfer

Figure 10. The ignition temperature, TignT_{\mathrm{ign}}, of an isolated iron particle of (a) different initial sizes, i.e., dp,0=50d_{\mathrm{p},0}=50 μm (dotted), dp,0=25d_{\mathrm{p},0}=25 μm (dash-dotted), and dp,0=10d_{\mathrm{p},0}=10 μm (dashed), with radiative heat transfer to the ambient air and without radiative heat transfer (solid) plotted as a function of δ0\delta_0, and (b) different initial oxide layer thicknesses, i.e., X0=0.1X_0=0.1 μm (solid), X0=10X_0=10 nm (dashed), and X0=1X_0=1 nm (dash-dotted), with (thick curves) and without (thin curves) radiative heat transfer plotted as a function of dp,0d_{\mathrm{p},0}. The inset in (b) is a zoom-in view showing the resulting trends of TignT_{\mathrm{ign}} for relatively large particles.

Collective heating effect in a suspension of particles

The results of TignT_{\mathrm{ign}} for a spherical suspension of iron particles (with δ0=10−3\delta_0=10^{-3} and a fixed initial particle size of dp,0=20d_{\mathrm{p},0}=20 μm) in air are plotted as a function of ϕFeO\phi_{\mathrm{FeO}}, i.e., the fuel equivalence ratio considering FeO as the only oxide product in Figure 11(a). The calculations were performed for a constant gas composition XO2:XN2=0.21:0.79X_{\mathrm{O_2}}:X_{\mathrm{N_2}}=0.21:0.79 while ϕFeO\phi_{\mathrm{FeO}} is varied by changing the number of particles inside the suspension according to Eqs. (15) and (16). As the initial radius of a spherical suspension increases from rs,0=1r_{\mathrm{s},0}=1 mm (dashed curve) to 5 cm (dotted curve), the resulting TignT_{\mathrm{ign}} decreases from approximately 1075 K, which is slightly below the TignT_{\mathrm{ign}} for an isolated particle (indicated by the solid horizontal line), to values below 800 K. This reduction in TignT_{\mathrm{ign}} is due to the collective effect of energy released by multiple particles suspended within a finite-sized volume that is losing heat to the surrounding via its external surface [64]. Analogous to an isolated particle, the conductive heat loss rate of a spherical suspension scales linearly with suspension radius as As∝rsA_{\mathrm{s}} \propto r_{\mathrm{s}} and hs∝1/rsh_{\mathrm{s}} \propto1/r_{\mathrm{s}}. The rate of energy release from the reacting particles in a suspension scales with NpN_{\mathrm{p}} that is proportional to rs3r_{\mathrm{s}}^{3}. The increase in energy release rate thus surpasses the increase in heat loss rate for an increasingly large suspension. This analysis suggests that, as a result of the collective effect, the critical ignition temperature of iron particles within a flame may be lower than that for an isolated particle. Future experimental effort is required to quantitatively probe the collective heating effect on flame propagation in suspensions of iron particles.

Ignition temperature of iron-particle suspensions

Figure 11. The ignition temperature, TignT_{\mathrm{ign}}, of a suspension of iron particles in air with different initial suspension radii, i.e., rs,0=1r_{\mathrm{s},0} = 1 mm (dashed), rs,0=1r_{\mathrm{s},0} = 1 cm (dash-dotted), and rs,0=5r_{\mathrm{s},0} = 5 cm (dotted), and δ0=10−3\delta_{0} = 10^{-3} (a) as a function of fuel equivalence ratio (assuming FeO as the only product) with dp,0=20d_{\mathrm{p},0} = 20 μ\mum and (b) as a function of initial particle diameter with ϕFeO=1\phi_{\mathrm{FeO}} = 1. The value of TignT_{\mathrm{ign}} for an isolated particle (with δ0=10−3\delta_{0} = 10^{-3}) in air is marked by the horizontal solid line.

Another remark on the result shown in Figure 11(a) is that TignT_{\mathrm{ign}} monotonically decreases as ϕFeO\phi_{\mathrm{FeO}} increases. As the kinetic oxidation rate is independent of the O2\mathrm{O}_{2} concentration of the suspension, the overall energy release rate is not affected by a slight depletion of O2\mathrm{O}_{2} prior to transitioning to an external-diffusion-controlled combustion. Thus, the ignition propensity of a suspension is enhanced by an increase in volumetric concentration of iron particles, regardless of the corresponding equivalence ratio. As ϕFeO\phi_{\mathrm{FeO}} decreases to very small values, the volumetric number density of iron particles approaches zero and, thus, the resulting TignT_{\mathrm{ign}} of a suspension reverts back to the value of an isolated particle.

Although TignT_{\mathrm{ign}} of an isolated iron particle is independent of initial particle size for a fixed value of δ0\delta_{0} (as discussed in Sect. 5.2), Figure 11(b) shows that TignT_{\mathrm{ign}} of a suspension increases with dp,0d_{\mathrm{p},0} of individual particles for a fixed equivalence ratio ϕFeO=1\phi_{\mathrm{FeO}} = 1. For a fixed iron concentration in a suspension, the smaller the particles are, the larger internal surface area is available to reaction, while the external surface area of the suspension, for heat loss into the surrounding, remains the same. Thus, smaller particles give rise to a more pronounced collective effect on enhancing the ignition propensity of a suspension. A similar enhancement due to the collective effect for smaller particles has also been identified in the ignition analysis for agglomerates of metal particles [64, 65].

Transition from kinetic- to diffusion-controlled combustion

Although the current model considering a completely kinetic-controlled reaction rate is sufficient to predict the ignition temperature, for temperatures greater than TignT_{\mathrm{ign}}, it can be further developed to estimate the conditions under which an iron particle transitions from a kinetic-controlled combustion to an external-diffusion-controlled combustion. This further development is based on the generic kk-β\beta paradigm [64] of particle reaction that considers the interplay between the condensed-phase (or heterogeneous) oxidation kinetics and the transport of oxidizer from the ambient gas to the particle surface. The kinetic consumption rate of O2\mathrm{O}_{2}, denoted by m˙R\dot{m}_{\mathrm{R}}, is governed by the parabolic rate laws of iron oxidation as described by Eq. (11). This kinetic rate is approximated to be independent of the gaseous O2\mathrm{O}_{2} concentration at the particle surface, CO2,sfC_{\mathrm{O}_{2},\mathrm{sf}}, for any CO2,sf>0C_{\mathrm{O}_{2},\mathrm{sf}} > 0, as justified in Sect. 2.1. The rate of O2\mathrm{O}_{2} mass transport, m˙D\dot{m}_{\mathrm{D}}, from the ambient gas to the particle surface (as illustrated in Figure 12) can be estimated as follows,

Schematic illustrating the diffusion of O2 from the bulk gas to the particle surface

Figure 12. Schematic illustrating the diffusion of O2_2 from the bulk gas to the particle surface. The dashed curve corresponds to the case of the maximum diffusion rate at a given O2_2 concentration in the bulk gas.

m˙D=Apβp(CO2,g−CO2,sf)(24)\dot{m}_{\mathrm{D}} = A_{\mathrm{p}}\beta_{\mathrm{p}}(C_{\mathrm{O}_{2},\mathrm{g}} - C_{\mathrm{O}_{2},\mathrm{sf}}) \tag*{(24)}

where CO2,gC_{\mathrm{O}_2,\mathrm{g}} and CO2,sfC_{\mathrm{O}_2,\mathrm{sf}} are the concentrations of O2\mathrm{O}_2 in the bulk gas and at the particle surface, respectively. The diffusive mass transfer coefficient βp\beta_{\mathrm{p}} is expressed as:

βp=ShDO2dp.(25)\beta_{\mathrm{p}} = \frac{\mathrm{Sh}D_{\mathrm{O}_2}}{d_{\mathrm{p}}}. \tag*{(25)}

The Sherwood number (Sh) for a spherical particle can also be calculated using the Frössling correlation:

Sh=2+0.552Re12Sc13.(26)\mathrm{Sh} = 2 + 0.552\mathrm{Re}^{\frac{1}{2}}\mathrm{Sc}^{\frac{1}{3}}. \tag*{(26)}

Under the assumption of a quiescent gas medium, Sherwood number equals to two, thus, βp=DO2/rp\beta_{\mathrm{p}} = D_{\mathrm{O}_2}/r_{\mathrm{p}}. The diffusivity, DO2D_{\mathrm{O}_2}, through this diffusion boundary layer around a particle is estimated based on the mixture-averaged thermal conductivity calculated via Eq. 13 assuming a unity Lewis number. The maximum O2\mathrm{O}_2 diffusion rate corresponds to the case wherein CO2,sf=0C_{\mathrm{O}_2,\mathrm{sf}} = 0 as illustrated by the dashed curve in Figure 12 and can be calculated as:

m˙D,max=4πrpDO2CO2,g,(27)\dot{m}_{\mathrm{D,max}} = 4\pi r_{\mathrm{p}}D_{\mathrm{O}_2}C_{\mathrm{O}_2,\mathrm{g}}, \tag*{(27)}

where DO2D_{\mathrm{O}_2} is a function of gas composition and particle surface temperature, TsfT_{\mathrm{sf}}, estimated via the two-third law based on TpT_{\mathrm{p}} and TgT_{\mathrm{g}}.

For CO2,sf>0C_{\mathrm{O}_2,\mathrm{sf}} > 0, the kinetic rate is less than the maximum rate of O2\mathrm{O}_2 diffusion, i.e., m˙R<m˙D,max\dot{m}_{\mathrm{R}} < \dot{m}_{\mathrm{D,max}}, and the O2\mathrm{O}_2 consumption rate of the particle, m˙O2\dot{m}_{\mathrm{O}_2}, is completely controlled by the solid-phase kinetics. When CO2,sf=0C_{\mathrm{O}_2,\mathrm{sf}} = 0, the maximum O2\mathrm{O}_2 diffusion rate is reached such that m˙O2\dot{m}_{\mathrm{O}_2} must be equal to m˙D,max\dot{m}_{\mathrm{D,max}}. Considering the nature of solid-phase iron oxidation, the generic kk-β\beta model can thus be specialized to a switch-type model between the kinetic- and external-diffusion-controlled oxidation rates as:

{m˙O2=m˙Rif m˙R<m˙D,maxm˙O2=m˙D,maxotherwise(28)\begin{cases} \dot{m}_{\mathrm{O}_2} = \dot{m}_{\mathrm{R}} & \text{if } \dot{m}_{\mathrm{R}} < \dot{m}_{\mathrm{D,max}} \\ \dot{m}_{\mathrm{O}_2} = \dot{m}_{\mathrm{D,max}} & \text{otherwise} \tag*{(28)} \end{cases}

to more accurately describe the transitions between the particle combustion regimes at temperatures greater than TignT_{\mathrm{ign}}. Once a particle transitions into external-diffusion-limited combustion, the total reaction rate of O2\mathrm{O}_2, m˙O2\dot{m}_{\mathrm{O}_2}, is partitioned into the formation rates of FeO and Fe3_3O4_4 based on the ratio between the kinetic rates (as shown in Figure 2) of FeO and Fe3_3O4_4 formation at the corresponding TpT_{\mathrm{p}}, XFeOX_{\mathrm{FeO}}, and XFe3O4X_{\mathrm{Fe_3O_4}}. Figure 13 shows the comparison between m˙R\dot{m}_{\mathrm{R}} (dash-dotted curve) and m˙D,max\dot{m}_{\mathrm{D,max}} (solid curve) resulting from the unsteady model for an isolated particle of dp,0=20 μmd_{\mathrm{p},0} = 20\,\mu\mathrm{m} in air. When m˙R\dot{m}_{\mathrm{R}} exceeds m˙D,max\dot{m}_{\mathrm{D,max}}, it indicates the transition of a particle from a kinetic-controlled to an external-diffusion-controlled combustion. The results of m˙R\dot{m}_{\mathrm{R}} and m˙D,max\dot{m}_{\mathrm{D,max}} stop when TpT_{\mathrm{p}} reaches the melting point (m.p.) of FeO.

Comparison of kinetic consumption rate and maximum external diffusion rate for six cases

Figure 13. Comparison between the kinetic consumption rate, m˙R\dot{m}_R, and the maximum external diffusion rate, m˙D,max⁡\dot{m}_{D,\max}, of O2\mathrm{O}_2 in air for the cases of an initially 20 μm-sized particle with (a-c) a fixed gas temperature of Tg=1250T_g = 1250 K and increasingly large initial oxide thickness ratios and (d-f) a fixed δ0=10−3\delta_0 = 10^{-3} and increasingly high gas temperatures.

The results for the cases with a fixed gas temperature of Tg=1250 KT_{\mathrm{g}} = 1250\,\mathrm{K} and various initial oxide thicknesses are plotted in Figure 13(a)–(Figure 13c), and for the cases with a fixed initial oxide thickness ratio of δ0=10−3\delta_0 = 10^{-3} and various gas temperatures are plotted in Figure 13(d)–(Figure 13f). For a relatively thin oxide layer of δ0=10−3\delta_0 = 10^{-3} and a moderately high gas temperature of 1250 K, as shown in Figure 13(a) and (e), m˙R\dot{m}_{\mathrm{R}} is initially greater than m˙D,max\dot{m}_{\mathrm{D,max}} for a very short period of time and then decreases below m˙D,max\dot{m}_{\mathrm{D,max}} around t=10 μst = 10\,\mu\mathrm{s} due to the growth of the oxide layer slowing down the reaction rate. Over this short period of m˙R>m˙D,max⁡\dot{m}_R > \dot{m}_{D,\max}, the particle reaction rate is external-diffusion controlled. After a short oxidation phase, the oxide layer grows sufficiently that the reaction rate slows and a transition to kinetic control is observed. At a later time around t=2×10−4 st = 2 \times10^{-4}\,\mathrm{s}, m˙R\dot{m}_R again becomes greater than m˙D,max⁡\dot{m}_{D,\max}. This crossover marks the transition of the particle from a kinetic-controlled combustion to an external-diffusion-controlled combustion prior to reaching the melting point of FeO. This double transition behavior is identified in the region of thin initial oxide layer and moderately high initial particle temperature (which equals to gas temperature) as marked by the orange circles ○ in Figure 14.

Map in the parametric space of initial particle temperature and oxide-layer thickness

Figure 14. A map in the parametric space of initial particle temperature Tp,0T_{\mathrm{p},0} (=TgT_{\mathrm{g}}) and δ0\delta_0 summarizing different behaviors of transition (prior to the melting point of FeO) between particle combustion regimes that are identified via the comparison between m˙R\dot{m}_{\mathrm{R}} and m˙D,max\dot{m}_{\mathrm{D,max}} for the case with dp,0=20d_{\mathrm{p},0}=20 µm. The different transition behaviors are marked as follows: ×\times Completely external O2_2-diffusion-controlled (m˙R>m˙D,max\dot{m}_{\mathrm{R}}>\dot{m}_{\mathrm{D,max}}, e.g., Figure 13(d)); ∘\circ transition from an initially diffusion- to kinetic-controlled regime followed by a transition from kinetic- to diffusion-controlled regime (m˙R>m˙D,max→m˙R<m˙D,max→m˙R>m˙D,max\dot{m}_{\mathrm{R}}>\dot{m}_{\mathrm{D,max}}\to\dot{m}_{\mathrm{R}}<\dot{m}_{\mathrm{D,max}}\to\dot{m}_{\mathrm{R}}>\dot{m}_{\mathrm{D,max}}, e.g., Figure 13(a) and (Figure 13e)); ∗\ast transition from kinetic- to diffusion-controlled regime (m˙R<m˙D,max→m˙R>m˙D,max\dot{m}_{\mathrm{R}}<\dot{m}_{\mathrm{D,max}}\to\dot{m}_{\mathrm{R}}>\dot{m}_{\mathrm{D,max}}, e.g., Figure 13(b) and (Figure 13f)); ++ completely kinetic-controlled, no transition below the melting point of FeO (m˙R<m˙D,max\dot{m}_{\mathrm{R}}<\dot{m}_{\mathrm{D,max}}, e.g., Figure 13(c)).

For a fixed gas temperature at 1250 K with a larger initial oxide thickness ratio δ0=10−2\delta_0 = 10^{-2}, the particle initially burns under a kinetic-controlled regime (m˙R<m˙D,max⁡\dot{m}_R < \dot{m}_{D,\max}) and transitions to a diffusion-controlled combustion shortly prior to reaching the melting point of FeO as shown in Figure 13(b). A similar transition behavior is found for the case with a smaller initial oxide ratio δ0=10−3\delta_0 = 10^{-3} and a lower gas temperature at 1150 K as shown in Figure 13(f). In Figure 14, the green asterisks * mark the cases with a kinetic-to-diffusion transition below the melting point of FeO.

For cases with a sufficiently thin initial oxide layer and high gas temperature, e.g., δ0=10−3\delta_0 = 10^{-3} and Tg=1350T_g = 1350 K as shown in Figure 13(d), m˙R\dot{m}_R remains greater than m˙D,max⁡\dot{m}_{D,\max} throughout the combustion process prior to the FeO melting point, suggesting that the thermal runaway process is likely to be completely controlled by the external diffusion of O2\mathrm{O}_2. This scenario is identified for the cases with a thin initial oxide thickness and high gas temperature as marked by the red crosses × in Figure 14. This finding implies that, if a particle is abruptly heated to a sufficiently high temperature without a significant growth of oxide layer (e.g., by laser heating in Ref. [36]), its ignition characteristics may not be governed by the kinetics of solid-phase iron oxidation at all.

In the case with Tg=1250T_g = 1250 K and an even thicker initial oxide layer of δ0=10−1\delta_0 = 10^{-1}, m˙R\dot{m}_R remains smaller than m˙D,max⁡\dot{m}_{D,\max} throughout the entire thermal runaway process prior to reaching the FeO melting point as shown in Figure 13(c). Thus, no transition from kinetic- to diffusion-controlled combustion occurs before the particle starts to melt. The region in the Tp,0T_{p,0}-δ0\delta_0 parametric space marked by the cyan plus symbols + in Figure 14 corresponds to the cases without a transition below the melting point.

The characteristics of ignition delay time, τign\tau_{\mathrm{ign}}

For isolated particles with heat loss to the surrounding, ignition delay times, τign\tau_{\mathrm{ign}}, can be determined for the cases with a gas temperature (and initial particle temperature) equal to or greater than the ignition temperature, i.e., Tg=Tp,0≥TignT_g = T_{p,0} \ge T_{\mathrm{ign}}, using the augmented reaction-rate model, Eq. (28). For metal particles, since the increase of temperature is interrupted by phase changes, ignition delay time cannot be conveniently measured based on the moment of maximum temperature increase as commonly defined for homogeneous reactive mixtures. In this study, τign\tau_{\mathrm{ign}} is measured as the time elapsed from the beginning of each calculation with Tg=Tp,0T_{\mathrm{g}}=T_{\mathrm{p},0} to the moment when TpT_{\mathrm{p}} reaches the melting point of FeO, i.e., 1640 K. This quantity can be measured and compared consistently in future studies. The effects of initial oxide-layer thickness and particle size on τign\tau_{\mathrm{ign}} of isolated particles are first examined in Sects. 5.7.1 and 5.7.2, respectively. In order to better probe the dependence of τign\tau_{\mathrm{ign}} on the reaction nature of iron particles, τign\tau_{\mathrm{ign}} is also analyzed for adiabatic suspensions of particles in Sect. 5.7.4.

Effect of initial oxide-layer thickness on τign\tau_{\mathrm{ign}} of isolated particles

For a selected initial particle size of dp,0=20d_{\mathrm{p},0}=20 µm, the model prediction of τign\tau_{\mathrm{ign}} for isolated iron particles are plotted as color contours in the parametric space of Tp,0T_{\mathrm{p},0} and δ0\delta_0, as shown in Figure 15(a). The predicted TignT_{\mathrm{ign}} as a function of δ0\delta_0 is plotted as the white curve on the color contours. The dashed curves indicate the iso-contours of τign\tau_{\mathrm{ign}}. The gray curve indicates the boundary of the completely external-diffusion-controlled region marked by ×\times in Figure 14. Figure 15(a) shows that, outside the external-diffusion-controlled region, the iso-contours of τign\tau_{\mathrm{ign}} follow qualitatively the same trend of TignT_{\mathrm{ign}} as a function of δ0\delta_0. For a fixed initial particle size, a particle with a greater δ0\delta_0 requires a higher initial temperature to result in the same ignition delay time, reflecting the temperature dependence of kinetic-controlled combustion. Inside the diffusion-controlled region, τign\tau_{\mathrm{ign}} barely varies with Tp,0T_{\mathrm{p},0} or δ0\delta_0.

Contour plots of ignition delay time in parametric spaces of initial particle temperature, initial particle size, and particle diameter

Figure 15. Contour plots of ignition delay time, τign\tau_{\mathrm{ign}}, in the parametric spaces of (a) initial particle temperature Tp,0T_{\mathrm{p},0} (=TgT_{\mathrm{g}}) and δ0\delta_0 for a constant initial particle size of dp,0=20d_{\mathrm{p},0}=20 μm and (b) initial particle temperature and initial particle diameter for δ0=10−3\delta_0=10^{-3}. The white curve in each plot indicates the ignition temperature, TignT_{\mathrm{ign}}. The dashed curves are the iso-contours of τign\tau_{\mathrm{ign}}. The gray curve in (a) indicates the boundary of the completely external-diffusion-controlled regime marked by × in Figure 14. The gray horizontal lines in (b) indicate the boundaries (in Tp,0T_{\mathrm{p},0}) between the diffusion-controlled, double-transition, and kinetic-to-diffusion-controlled regimes, as identified in Figure 14.

In Figure 15(a), it is also of interest to note that the iso-contour of τign=10−3.2\tau_{\mathrm{ign}}=10^{-3.2} s follows an opposite trend to those outside the diffusion region: For a fixed initial size, a particle with a greater δ0\delta_0 requires a lower initial temperature to result in the same ignition delay time. The turning point of the iso-contour of τign=10−3.2\tau_{\mathrm{ign}}=10^{-3.2} s is located on the boundary of the completely diffusion-controlled region, which is marked by the gray curve. This opposite trend is observed since, for a fixed particle size and temperature with an increasingly large δ0\delta_0, the heat capacity (in J K−1^{-1}) of the particle slightly decreases, resulting in a stronger feedback to thermal runaway.

In Figure 16(a), the predicted τign\tau_{\mathrm{ign}} for the cases with dp,0=20d_{\mathrm{p},0}=20 µm and various values of δ0\delta_0 are plotted as functions of the reciprocal of Tp,0T_{\mathrm{p},0}. The curves for the cases with δ0=10−4\delta_0=10^{-4} and 10−310^{-3} collapse onto a nearly horizontal line at high temperatures, corresponding to the diffusion-controlled regime identified in the Tp,0T_{\mathrm{p},0}-δ0\delta_0 parametric space. As Tp,0T_{\mathrm{p},0} approaches TignT_{\mathrm{ign}}, marked by the circles in Figure 16(a), τign\tau_{\mathrm{ign}} increases more rapidly than exponential due to an increasingly pronounced effect of heat loss from the particle to the surrounding.

Model predictions of ignition delay time as functions of reciprocal initial particle temperature and initial particle size

Figure 16. The model prediction of τign\tau_{\mathrm{ign}} as (a) a function of the reciprocal of initial particle temperature, Tp,0T_{\mathrm{p},0}, with dp,0=20 μmd_{\mathrm{p},0}=20\ \mu\mathrm{m} and various values of δ0\delta_0, and (b) a function of initial particle size, dp,0d_{\mathrm{p},0}, with δ0=10−3\delta_0=10^{-3} and various values of Tp,0T_{\mathrm{p},0}. The inset of (b) shows the plot of τign\tau_{\mathrm{ign}} as a function of dp,02d_{\mathrm{p},0}^{2}.

Effect of particle size on ign of isolated particles

Figure 15(b) shows the contour plot of τign\tau_{\mathrm{ign}} in the parametric space of Tp,0T_{\mathrm{p},0} and initial particle diameter, dp,0d_{\mathrm{p},0}, for a fixed value of δ0=10−3\delta_0=10^{-3}. Although TignT_{\mathrm{ign}} is solely dependent on δ0\delta_0, as indicated by the horizontal white line in Figure 15(b), τign\tau_{\mathrm{ign}} significantly increases with particle size. For the cases with Tp,0T_{\mathrm{p},0} slightly greater than TignT_{\mathrm{ign}}, τign\tau_{\mathrm{ign}} increases over three orders of magnitude, i.e., from 10−410^{-4} s to 0.1 s, as dp,0d_{\mathrm{p},0} increases from 1 μm to 100 μm. The horizontal gray lines mark the boundaries of different regimes of particle combustion in terms of Tp,0T_{\mathrm{p},0}, and show that the transition behavior between kinetic- and diffusion-controlled combustion regimes is independent of particle size. This independence is due to the fact that the kinetic rate of oxidation, m˙R\dot{m}_{\mathrm{R}}, and the maximum diffusion rate of O2_2 from the bulk gas, m˙D,max\dot{m}_{\mathrm{D,max}}, both linearly scale with particle size, as shown below:

m˙R=ν~O2/FeOExp⁡(−Ta,FeOTp)(1−δδFe3O4)2rpδ−δδFe3O4+ν~O2/Fe3O4Exp⁡(−Ta,Fe3O4Tp)rpδδFe3O4\dot{m}_{\mathrm{R}}=\tilde{\nu}_{\mathrm{O_2}/\mathrm{FeO}}\operatorname{Exp}\left(-\frac{T_{\mathrm{a,FeO}}}{T_{\mathrm{p}}}\right)\frac{\left(1-\delta\delta_{\mathrm{Fe_3O_4}}\right)^2r_{\mathrm{p}}}{\delta-\delta\delta_{\mathrm{Fe_3O_4}}}+\tilde{\nu}_{\mathrm{O_2}/\mathrm{Fe_3O_4}}\operatorname{Exp}\left(-\frac{T_{\mathrm{a,Fe_3O_4}}}{T_{\mathrm{p}}}\right)\frac{r_{\mathrm{p}}}{\delta\delta_{\mathrm{Fe_3O_4}}}
m˙D,max=4πrpDO2CO2,g,(29)\dot{m}_{\mathrm{D,max}}=4\pi r_{\mathrm{p}}D_{\mathrm{O_2}}C_{\mathrm{O_2},\mathrm{g}}, \tag*{(29)}

where ν~O2/FeO=νO2/FeOρFeOk0,FeO\tilde{\nu}_{\mathrm{O_2}/\mathrm{FeO}}=\nu_{\mathrm{O_2}/\mathrm{FeO}}\rho_{\mathrm{FeO}}k_{0,\mathrm{FeO}} and ν~O2/Fe3O4=νO2/Fe3O4ρFe3O4k0,Fe3O4\tilde{\nu}_{\mathrm{O_2}/\mathrm{Fe_3O_4}}=\nu_{\mathrm{O_2}/\mathrm{Fe_3O_4}}\rho_{\mathrm{Fe_3O_4}}k_{0,\mathrm{Fe_3O_4}}.

As shown in Figure 15(b), below the boundary of the completely diffusion-controlled regime at 1281 K, the iso-contours of τign\tau_{\mathrm{ign}} (i.e., the dashed curves) show that a larger particle requires a greater initial temperature to result in the same ignition delay time. In the completely diffusion-controlled regime, the iso-contours of τign\tau_{\mathrm{ign}} are nearly vertical, indicating that, for a fixed initial particle size, τign\tau_{\mathrm{ign}} very slightly depends on particle temperature. This slight dependence is owing to the temperature-dependence of the transport properties of the gaseous species near the particle surface. The contour plot in Figure 15(b) shows that τign\tau_{\mathrm{ign}} significantly increases with initial particle size. This trend is further clarified in Figure 16(b) showing τign\tau_{\mathrm{ign}} as functions of dp,0d_{\mathrm{p},0} for δ0=10−3\delta_0=10^{-3} and different values of Tp,0T_{\mathrm{p},0}. The inset of Figure 16(b) demonstrates that τign\tau_{\mathrm{ign}} scales with dp,02d_{\mathrm{p},0}^2 at all of the values of Tp,0T_{\mathrm{p},0} corresponding to both kinetic- and diffusion-controlled regimes. The ignition delay time is proportional to the thermal capacity of the particle and inversely proportional to the reaction rate. As shown in Eq. (29), both kinetic- and diffusion-controlled reaction rates linearly scale with particle size. The thermal capacity of the particle scales with particle volume, i.e., ∝dp,03\propto d_{\mathrm{p},0}^3. Thus, the scaling between τign\tau_{\mathrm{ign}} and particle size follows a d2d^{2}-law4.

Ignition delay time of adiabatic suspensions of particles

To better examine the dependence of τign\tau_{\mathrm{ign}} on the reaction nature of iron particles, the model prediction of τign\tau_{\mathrm{ign}} for adiabatic suspensions is analyzed in this section. As the size of a spherical suspension of particles becomes sufficiently large, i.e., rs→∞r_s \to\infty, the conductive heat transfer from the boundary of the suspension to the inert surrounding becomes negligible. An adiabatic suspension can thus be understood as an infinitely large cloud of suspended particles. The critical ignition temperature for such a suspension goes to the limit of zero. In other words, any initial temperature can trigger a thermal runaway of an adiabatic suspension. Thus, τign\tau_{\mathrm{ign}}, consistently defined as the time at which the particle temperature reaches the melting point of FeO, can be measured over a broad range of initial particle (and gas) temperatures.

Figure 17(a) shows the natural logarithm of τign\tau_{\mathrm{ign}} predicted for adiabatic suspensions of iron particles with dp,0=20 μmd_{\mathrm{p},0}=20\ \mu\mathrm{m} and various values of δ0\delta_0, plotted as functions of the reciprocal of initial particle temperature, i.e., 1000 K/Tp,01000\ \mathrm{K}/T_{\mathrm{p},0}. Over the entire range of temperature from 973 K to 1523 K, the predicted curve for the case with a relatively thick initial oxide layer of δ0=10−1\delta_0=10^{-1} follows a nearly linear trend with a slope of the activation temperature of FeO formation. This linear trend indicates that the reaction rate of the particles is kinetic-controlled during the ignition process. As δ0\delta_0 decreases from 10−110^{-1} to 10−410^{-4}, the curves of ln⁡(τign)\ln(\tau_{\mathrm{ign}}) vs. 1/Tp,01/T_{\mathrm{p},0} become increasingly nonlinear.

To understand this nonlinear trend, the kinetic rate of oxygen consumption, m˙R\dot{m}_{R}, and the maximum external-diffusion rate, m˙D,max⁡\dot{m}_{D,\max}, for the cases with δ0=10−3\delta_0=10^{-3} and five selected values of Tp,0T_{\mathrm{p},0} (as marked in Figure 17(a)) across the herein studied temperature range, are compared in Figure 17(b). At a sufficiently high temperature, e.g., Case ⑤ with Tp,0=1400 KT_{\mathrm{p},0}=1400\ \mathrm{K}, the ignition process is completely diffusion-controlled, as m˙D,max⁡<m˙R\dot{m}_{D,\max}<\dot{m}_{R} throughout the entire process. In this completely diffusion-controlled regime, τign\tau_{\mathrm{ign}} becomes independent of oxide layer thickness, as shown in Figure 17(a) that the curves of 10−210^{-2}, 10−310^{-3}, and 10−410^{-4} collapse. For Case ④ with Tp,0=1250 KT_{\mathrm{p},0}=1250\ \mathrm{K}, the ignition process is partly diffusion-controlled as indicated by the double-transition behavior identified in Figure 17(b). The gradual increase in the slope of the ln⁡(τign)\ln(\tau_{\mathrm{ign}})-vs.-1/Tp,01/T_{\mathrm{p},0} curve from Case ⑤ to Case ④, as shown in Figure 17(a), can be attributed to this partly diffusion-controlled regime. Cases ①–③ with relatively low initial temperatures undergo a completely kinetic-controlled ignition process, as shown by the comparison between m˙D,max⁡\dot{m}_{D,\max} and m˙R\dot{m}_{R} plotted in Figure 17(b). The resulting ln⁡(τign)\ln(\tau_{\mathrm{ign}})-vs.-1/Tp,01/T_{\mathrm{p},0} curve follows a nearly linear trend with a slope of −Ta,FeO-T_{\mathrm{a,FeO}} over two separated ranges of Tp,0T_{\mathrm{p},0}, which are marked by Cases ① and ③. In the intermediate temperature range marked by Case ②, τign\tau_{\mathrm{ign}} decreases more rapidly with an increase in Tp,0T_{\mathrm{p},0} than the slope of −Ta,FeO-T_{\mathrm{a,FeO}}. This anomalous dependence of ignition delay time on temperature is likely due to the fact that τign\tau_{\mathrm{ign}} is measured as the time when TpT_p reaches a fixed temperature, i.e., the melting point of FeO. Thus, the particles with a higher initial temperature, not only have a greater kinetic reaction rate, but also require a smaller amount of energy release to reach the melting point of FeO.

Natural logarithm of ignition delay time and comparison of reaction rates

Figure 17. (a) The natural logarithm of τign\tau_{\mathrm{ign}} predicted for adiabatic suspensions of iron particles with dp,0=20 μmd_{p,0}=20\,\mu\mathrm{m} and various values of δ0\delta_0, plotted as functions of the reciprocal of initial particle temperature, i.e., 1000 K/Tp,01000\,\mathrm{K}/T_{p,0}. (b) Comparison between the kinetic consumption rate, m˙R\dot{m}_{R}, and the maximum external diffusion rate, m˙D,max⁡\dot{m}_{D,\max}, of O2\mathrm{O}_2 in air for the cases with δ0=10−3\delta_0=10^{-3} and ① Tp,0=1000 KT_{p,0}=1000\,\mathrm{K}, ② Tp,0=1050 KT_{p,0}=1050\,\mathrm{K}, ③ Tp,0=1150 KT_{p,0}=1150\,\mathrm{K}, ④ Tp,0=1250 KT_{p,0}=1250\,\mathrm{K}, and ⑤ Tp,0=1400 KT_{p,0}=1400\,\mathrm{K}, as marked in (a).

Ignition delay time of non-adiabatic suspensions of particles

To more clearly demonstrate the collective heating effect, the model prediction of τign\tau_{\mathrm{ign}} as a function of ϕFeO\phi_{\mathrm{FeO}} for non-adiabatic suspensions at a common surrounding gas temperature, 1087 K, which is slightly above the TignT_{\mathrm{ign}} for an isolated particle, is plotted in Figure 18. The resulting τign\tau_{\mathrm{ign}} for suspensions of different initial sizes are on the same order of magnitude as τign\tau_{\mathrm{ign}} for an isolated particle. For larger values of ϕFeO\phi_{\mathrm{FeO}}, τign\tau_{\mathrm{ign}} of suspensions becomes increasingly lower than that for an isolated particle due to a more pronounced collective heating effect. Furthermore, for a sufficiently large non-adiabatic suspension, i.e., rs,0≥1r_{\mathrm{s},0} \ge1 cm, the resulting τign\tau_{\mathrm{ign}} approaches to the values for adiabatic suspensions as indicated by the cyan curve in Figure 18. It has been demonstrated in Sect. 5.6 that the collective heating mechanism can significantly reduce the ignition temperature of sufficiently large non-adiabatic suspensions with sufficiently high particle number densities. Note that, at a reduced TignT_{\mathrm{ign}}, the ignition delay time of a suspension can be significantly longer than that of an isolated particle at its corresponding, higher, TignT_{\mathrm{ign}}.

Ignition delay time of a non-adiabatic suspension of iron particles

Figure 18. The ignition delay time, τign\tau_{\mathrm{ign}}, of a non-adiabatic suspension of iron particles in air with different initial suspension radii and δ0=10−3\delta_0 = 10^{-3}, as a function of fuel equivalence ratio (assuming FeO as the only product) with dp,0=20 μmd_{\mathrm{p},0} = 20\ \mu\mathrm{m} at a fixed temperature 1087 K, i.e., slightly above TignT_{\mathrm{ign}} for an isolated particle. The value of τign\tau_{\mathrm{ign}} for an isolated particle in air is marked by the horizontal solid line. The τign\tau_{\mathrm{ign}} for an adiabatic suspension is plotted as the thick cyan curve.

Comparison with experimental measurement

The ignition temperatures measured by Grosse and Conway [41] and Bolobov [43] for mm-sized iron specimens are perhaps the closest to the definition of TignT_{\mathrm{ign}} in the current paper. In these authors’ experiments, a specimen was first heated in vacuum or an inert gas to a steady and uniform temperature T0T_0; oxygen or air was then fed into the test chamber; the minimum T0T_0 triggering an abrupt increase in the temperature of the specimen, i.e., a thermal runaway, was determined as the critical temperature for ignition. The ignition temperatures reported by Grosse and Conway [41] and Bolobov [43] are 1203 ±\pm 10K and 1233 ±\pm 20K, respectively. These values are close, but slightly greater, than the range of TignT_{\mathrm{ign}} predicted by the current model for an isolated iron particle in air, i.e., from 1080 K to 1220 K for δ0=10−4\delta_0 = 10^{-4}–10−110^{-1} (as shown in Figure 7). This small discrepancy can be attributed to the following reasons: (1) The iron specimens in these experiments were in direct contact with the supporting device, thus, losing heat to the surrounding via solid-phase thermal conduction; (2) a mm-sized specimen might be subjected to a natural convection with Rayleigh number greater than unity (as considered in the analysis of Bolobov and Podlevskikh [46]); (3) the supplied oxygen flow might be of a non-negligible velocity at a lower temperature than the specimen temperature.

Bolobov [43] also found that the ignition temperature of iron is barely dependent on oxygen concentration over a broad range of pO2p_{\mathrm{O}_2} from 0.2 to 20 MPa. This finding is consistent with the fact that the growth rate of the FeO and Fe3_3O4_4 layers is controlled by the lattice diffusion of Fe ions subjected to the equilibrium Fe activities at the interlayer boundaries, and not significantly affected by the ambient O2_2 concentration. Thus, the consideration of O2_2-concentration-independent kinetic rates in the current analysis is supported by this experimental evidence.

As reviewed by Gorokhov [39], the experimentally measured ignition temperatures for clouds of sub-mm-sized iron particles scatter over a range from 588 K to 1053 K. This significant reduction in ignition temperature from those measured for isolated iron specimens [41, 42, 43] is likely due to the collective effect as captured by the current suspension model and discussed in Sect. 5.5. However, it is of importance to note that, since the values of the kinetic parameters (in Table 1) are only valid for iron oxidation over a temperature range from 973 K to 1523 K, the current model prediction below this range might not be quantitatively accurate.

TignT_{\mathrm{ign}} (K)SpecimenConditions
Grosse & Conway, exp. [41]1203 ±\pm 1010-g-weighted ironFlow of O2\mathrm{O}_{2} at 1 atm
Bolobov, exp. [43]1233 ±\pm 205×5×0.55 \times5 \times0.5 mm low-carbon steel foilO2\mathrm{O}_{2} at 0.14-0.6 MPa
Gorokhov (reviewed), exp. [39]588-1053Cloud of sub-mm Fe particlesAir at 1 atm
Gorokhov (reviewed), exp. [39]553-733Precipitated layer of sub-mm Fe particlesAir at 1 atm
Current analysis1080Isolated Fe particle with δ0≲0.003\delta_{0} \lesssim0.003Air at 1 atm
Current analysis760-902Spherical suspensions of Fe particles of dp,0=20d_{\mathrm{p},0}=20 μ\mum, ϕFeO=1\phi_{\mathrm{FeO}}=1, and rs,0=5r_{\mathrm{s},0}=5 cm−1^{-1} cmAir at 1 atm

Table 3. Comparison between experimental measurements and predictions by the current model of TignT_{\mathrm{ign}} of various iron (or steel) specimens.

Another possible source of inaccuracy is rooted in the assumption of spherical particles made in this study. Pulverized sponge iron (PSI) particles, e.g., those used for combustion experiments by Tóth et al. [35], have irregular shapes and porous structures. Iron particles produced by the reduction of spherical iron-oxide particles are porous because the iron oxides take up more volume than iron due to the fact that the Pilling-Bedworth ratios for Fe →\to Fe2O3_{2}\mathrm{O}_{3}, Fe3O4_{3}\mathrm{O}_{4}, and FeO are all greater than unity. An irregularly shaped, porous iron particle has a greater surface area that has access to and can react with oxygen. Thus, the ignition temperature might be significantly reduced. Future efforts are required to quantitatively measure the porosity of reduced iron particles and estimate how particle porosity enhances the ignition of particles and their suspensions.

It is noteworthy that, in reality, the oxidation rate of iron is governed by different processes when the oxide layer is extremely thin, i.e., under approximately 20 nm. The parabolic rate law considered in this analysis is based upon the assumption that Fe-cation diffusion across a sufficiently thick oxide layer with a uniform electric field is the rate-limiting process. However, for a very thin oxide layer, its growth is governed by ionic transfer driven by a large electric field and space charges in the oxide layer [53]. In this thin-layer regime, the oxidation process is better explained by the theory of Cabrera and Mott [66]. As a quantitatively reliable rate law for thin-layer growth of iron oxidation is not available, the thick-layer parabolic rate law is extrapolated in the current analysis to a thin-layer regime. The question arises as to how this extrapolation affects the accuracy in estimating the ignition characteristics of iron particles. An example of the growth rate of NiO as a function of oxide layer thickness in both thin- and thick-layer regimes is provided in Figure 7 of Atkinson’s review [53]: If one extrapolates the parabolic rate law (Wagner’s model) to nano-scale thicknesses, the growth rate would be underestimated compared to the prediction of Cabrera and Mott’s model. An underestimation of the initial oxide-layer growth rate in the current analysis for iron is unlikely to have a significant effect on TignT_{\mathrm{ign}}. As demonstrated in Figure 8(b) and (Figure 8d) and discussed in Sect. 5.3, for a sufficiently thin initial oxide layer, the initial rapid oxide-layer growth has no effect on the subsequent processes determining whether a thermal runaway occurs. Therefore, the thin layer mechanisms described by the theory of Cabrera and Mott, which result in greater oxidation rates over a time period much shorter than the thermal-runaway time scale of an iron particle, are unlikely to have a significant effect on the ignition characteristics.

A recent study by Senyurt and Dreizin [67] demonstrates that, if the size of a reactive metal particle is comparable to the molecular mean free path of the surrounding gas, the transport processes between the particle and the gas are hindered by the rarefaction effects in the Knudsen transition regime. These authors’ analysis shows that, owing to a reduced rate of heat loss to the surrounding in the Knudsen transition regime, the ignition temperatures of sufficiently small particles are significantly lower than the model prediction based upon a continuum assumption. However, this analysis is limited to a steady-state model and not applied to study iron particles. A detailed, unsteady analysis of how the ignition characteristics of fine iron particles differ in the Knudsen transition regime will be reported in a future publication.

Concluding remarks

A model describing the heat and mass balance equations for heterogeneously burning micrometric iron particles was developed. The solid-phase oxidation kinetics described by a parabolic rate law were empirically calibrated using measured experimental data on the time-evolution of the growth of iron-oxide layers. The model was applied to quantitatively analyze the dependence of the minimum gas temperature required for a thermal runaway (namely, the ignition temperature) on particle size, initial thickness of the oxide layer, gas-phase composition, heat loss by radiation, and the collective heating effect in particle suspensions. Findings showed that the ignition temperature depends on the ratio between the initial oxide layer thickness and particle size, δ0\delta_0, and is independent of particle size. For isolated iron particles in air, the predicted ignition temperature plateaus at approximately 1080 K, and becomes independent of δ0\delta_0, for δ0≫0.003\delta_0 \gg0.003. The lower the thermal conductivity of the gas-phase environment, the lower the ignition temperature. The ignition behavior was found to be largely independent of radiative heat loss effects. The collective heating effect in burning particle suspensions was shown to significantly reduce the ignition temperature. The slight underestimation of the ignition temperature in the current model compared to experimental literature [41, 43], is likely associated with experimental conditions that facilitated further heat loss from the iron specimens in addition to convective and radiative heat loss.

The time histories of the reaction rate of the particles were used to parametrically probe the transition behavior between kinetic-controlled and external-diffusion-controlled combustion regimes. The ignition delay time is significantly influenced by the initial oxide layer thickness and particle temperature under a kinetic-controlled combustion regime; once a particle transitions to an external-diffusion-controlled regime, its ignition delay time becomes nearly independent of δ0\delta_0 and Tp,0T_{p,0}. Furthermore, the scaling of the ignition delay time of iron particles with particle size follows a d2d^2-law. Future experimental and modeling efforts should work to elucidate the rate of heat release in iron particle combustion in liquid-phase, and to explore how to ensure a non-volatile combustion of molten iron droplets.

Acknowledgement

The authors are grateful to Y. Pyo for his contributions to the early stages of this study and to J. Palečka and S. Goroshin for useful discussion in developing this paper. X.C.M. thanks L. Thijs for pointing out some lacks of clarity in the model description. Funding was provided by the Canadian Space Agency through the Flight and Fieldwork for the Advancement of Science and Technology (FAST) Grant Program. A.F. was supported by a TISFD Award for Summer Undergraduate Research (TASUR).

References

  1. [1]D. Beach, A. Rondinone, B. Sumpter, S. Labinov, and R. Richards, “Solid-State Combustion of Metallic Nanoparticles: New Possibilities for an Alternative Energy Carrier,” Journal of Energy Resources Technology, vol. 129, pp. 29–32, 07 2006.DOI
  2. [2]J. Bergthorson, S. Goroshin, M. Soo, P. Julien, J. Palecka, D. Frost, and D. Jarvis, “Direct combustion of recyclable metal fuels for zero-carbon heat and power,” Applied Energy, vol. 160, pp. 368 – 382, 2015.DOI
  3. [3]J. Bergthorson, “Recyclable metal fuels for clean and compact zero-carbon power,” Progress in Energy and Combustion Science, vol. 68, pp. 169 – 196, 2018.DOI
  4. [4]J. Païdassi, “Sur la cinetique de l’oxydation du fer dans l’air dans l’intervalle 700-1250°C,” Acta Metallurgica, vol. 6, no. 3, pp. 184–194, 1958.DOI
  5. [5]J. Païdassi, “Sur l’oxydation du protoxyde de fer dans l’air dans l’intervalle 600-1350°C,” Acta Metallurgica, vol. 6, no. 3, pp. 219–221, 1958.DOI
  6. [6]G. Yurek, J. Hirth, and R. Rapp, “The formation of two-phase layered scales on pure metals,” Oxidation of Metals, vol. 8, no. 5, pp. 265–281, 1974.DOI
  7. [7]W. Smeltzer and D. Young, “Oxidation properties of transition metals,” Progress in Solid State Chemistry, vol. 10, pp. 17–54, 1975.DOI
  8. [8]G. Garnaud and R. Rapp, “Thickness of the oxide layers formed during the oxidation of iron,” Oxidation of Metals, vol. 11, no. 4, pp. 193–198, 1977.DOI
  9. [9]J. Sato, K. Sato, and T. Hirano, “Fire spread mechanisms along steel cylinders in high pressure oxygen,” Combustion and Flame, vol. 51, pp. 279–287, 1983.DOI
  10. [10]T. Hirano and J. Sato, “Fire spread along structural metal pieces in oxygen,” Journal of Loss Prevention in the Process Industries, vol. 6, no. 3, pp. 151–157, 1993.DOI
  11. [11]T. Steinberg, G. Mulholland, D. Wilson, and F. Benz, “The combustion of iron in high-pressure oxygen,” Combustion and Flame, vol. 89, no. 2, pp. 221–228, 1992.DOI
  12. [12]T. Steinberg, J. Kurtz, and D. Wilson, “The solubility of oxygen in liquid iron oxide during the combustion of iron rods in high-pressure oxygen,” Combustion and Flame, vol. 113, no. 1, pp. 27–37, 1998.DOI
  13. [13]N. Ward and T. Steinberg, “The rate-limiting mechanism for the heterogeneous burning of cylindrical iron rods,” Journal of ASTM International, vol. 6, no. 6, pp. 1–13, 2009.
  14. [14]M. Muller, H. El-Rabii, and R. Fabbro, “Laser ignition of bulk iron, mild steel, and stainless steel in oxygen atmospheres,” Combustion Science and Technology, vol. 186, no. 7, pp. 953–974, 2014.DOI
  15. [15]M. Muller, H. El-Rabii, and R. Fabbro, “Liquid phase combustion of iron in an oxygen atmosphere,” Journal of Materials Science, vol. 50, no. 9, pp. 3337–3350, 2015.DOI
  16. [16]C. Wagner, “Beitrag zur theorie des anlaufvorgangs,” Zeitschrift für Physikalische Chemie, vol. 21B, no. 1, pp. 25–41, 1933.DOI
  17. [17]K. Hauffe, The Mechanism of Oxidation of Metals—Theory, pp. 79–143. Boston, MA: Springer US, 1965.DOI
  18. [18]E. Dreizin, “Phase changes in metal combustion,” Progress in Energy and Combustion Science, vol. 26, no. 1, pp. 57–78, 2000.DOI
  19. [19]T. Hirano, K. Sato, Y. Sato, and J. Sato, “Prediction of metal fire spread in high pressure oxygen,” Combustion Science and Technology, vol. 32, no. 1-4, pp. 137–159, 1983.DOI
  20. [20]H. El-Rabii, K. Kazakov, and M. Muller, “Experimental and theoretical study of iron and mild steel combustion in oxygen flows,” Physics of Fluids, vol. 29, no. 3, p. 037104, 2017.
  21. [21]J. Sun, R. Dobashi, and T. Hirano, “Combustion behavior of iron particles suspended in air,” Combustion Science and Technology, vol. 150, no. 1-6, pp. 99–114, 2000.DOI
  22. [22]F. Tang, S. Goroshin, and A. Higgins, “Modes of particle combustion in iron dust flames,” Proceedings of the Combustion Institute, vol. 33, no. 2, pp. 1975 – 1982, 2011.DOI
  23. [23]P. Julien, S. Whiteley, S. Goroshin, M. Soo, D. Frost, and J. Bergthorson, “Flame structure and particle-combustion regimes in premixed methane-iron-air suspensions,” Proceedings of the Combustion Institute, vol. 35, no. 2, pp. 2431 – 2438, 2015.DOI
  24. [24]A. Wright, A. Higgins, and S. Goroshin, “The discrete regime of flame propagation in metal particulate clouds,” Combustion Science and Technology, vol. 188, no. 11-12, pp. 2178–2199, 2016.DOI
  25. [25]M. McRae, P. Julien, S. Salvo, S. Goroshin, D. Frost, and J. Bergthorson, “Stabilized, flat iron flames on a hot counterflow burner,” Proceedings of the Combustion Institute, 2018.DOI
  26. [26]J. Palečka, J. Sniatowsky, S. Goroshin, A. Higgins, and J. Bergthorson, “A new kind of flame: Observation of the discrete flame propagation regime in iron particle suspensions in microgravity,” Combustion and Flame, vol. 209, pp. 180–186, 2019.DOI
  27. [27]J. Palečka, S. Goroshin, A. Higgins, Y. Shoshin, P. de Goey, J.-R. Angilella, H. Oltmann, A. Stein, B. Schmitz, A. Verga, S. Vincent-Bonnieu, W. Sillekens, and J. Bergthorson, “Percolating reaction–diffusion waves (per-waves)—sounding rocket combustion experiments,” Acta Astronautica, vol. 177, pp. 639–651, 2020.DOI
  28. [28]N. Poletaev and M. Khlebnikova, “Combustion of iron particles suspension in laminar premixed and diffusion flames,” Combustion Science and Technology, vol. 0, no. 0, pp. 1–22, 2020.DOI
  29. [29]F. Tang, A. Higgins, and S. Goroshin, “Effect of discreteness on heterogeneous flames: propagation limits in regular and random particle arrays,” Combustion Theory and Modelling, vol. 13, no. 2, pp. 319–341, 2009.DOI
  30. [30]F. Tang, A. Higgins, and S. Goroshin, “Propagation limits and velocity of reaction-diffusion fronts in a system of discrete random sources,” Physical Review E, vol. 85, p. 036311, Mar 2012.pubmed.ncbi.nlm.nih.gov/22587184
  31. [31]S. Goroshin, F. Tang, and A. Higgins, “Reaction-diffusion fronts in media with spatially discrete sources,” Physical Review E, vol. 84, no. 2, p. 027301, 2011.pubmed.ncbi.nlm.nih.gov/21929144
  32. [32]X. Mi, A. Higgins, S. Goroshin, and J. Berghorson, “The influence of spatial discreteness on the thermo-diffusive instability of flame propagation with infinite leis number,” Proceedings of the Combustion Institute, vol. 36, no. 2, pp. 2359–2366, 2017.arxiv.org/abs/1512.01431
  33. [33]F. Lam, X. Mi, and A. Higgins, “Front roughening of flames in discrete media,” Phys. Rev. E, vol. 96, p. 013107, Jul 2017.DOI
  34. [34]F. Lam, X. Mi, and A. Higgins, “Dimensional scaling of flame propagation in discrete particulate clouds,” Combustion Theory and Modelling, vol. 0, no. 0, pp. 1–24, 2019.
  35. [35]P. Tóth, Y. Ögren, A. Sepman, P. Gren, and H. Wiinikka, “Combustion behavior of pulverized sponge iron as a recyclable electrofuel,” Powder Technology, vol. 373, pp. 210–219, 2020.DOI
  36. [36]D. Ning, Y. Shoshin, J. van Oijen, G. Finotello, and L. de Goey, “Burn time and combustion regime of laser-ignited single iron particle,” Combustion and Flame, vol. 230, p. 111424, 2021.DOI
  37. [37]J. Huang, S. Li, W. Cai, Y. Qian, E. Berrocal, M. Aldén, and Z. Li, “Quantification of the size, 3d location and velocity of burning iron particles in premixed methane flames using high-speed digital in-line holography,” Combustion and Flame, vol. 230, p. 111430, 2021.DOI
  38. [38]D. Ning, Y. Shoshin, M. van Stiphout, J. van Oijen, G. Finotello, and P. de Goey, “Temperature and phase transitions of laser-ignited single iron particle,” Combust. Flame, vol. 236, p. 111801, 2022.DOI
  39. [39]Y. Gorokhov, “On the pyrophoric properties, explosion hazard, and toxicity of powders and dusts of iron and its compounds,” Soviet Powder Metallurgy and Metal Ceramics, vol. 3, no. 1, pp. 82–86, 1964.DOI
  40. [40]V. Leshchevich, O. Penyazkov, A. Fedorov, A. Shul’gin, and J. Rostaing, “Conditions for ignition and delay time of ignition of iron microparticles in oxygen,” Journal of Engineering Physics and Thermophysics, vol. 85, no. 1, pp. 148–154, 2012.DOI
  41. [41]A. Grosse and J. Conway, “Combustion of metals in oxygen,” Industrial & Engineering Chemistry, vol. 50, no. 4, pp. 663–672, 1958.DOI
  42. [42]V. Bolobov, A. Berezin, P. Drozhzhin, and A. Shteinberg, “Ignition of compact stainless steel specimens in high pressure oxygen,” Combustion, Explosion and Shock Waves, vol. 27, no. 3, pp. 263–266, 1991.DOI
  43. [43]V. Bolobov, “Conditions for ignition of iron and carbon steel in oxygen,” Combustion, Explosion and Shock Waves, vol. 37, no. 3, pp. 292–296, 2001.DOI
  44. [44]A. Breiter, V. Mal’tsev, and E. Popov, “Models of metal ignition,” Combustion, Explosion, and Shock Waves, vol. 13, no. 4, pp. 475–485, 1977.DOI
  45. [45]B. Khaikin, V. Bloshenko, and A. Merzhanov, “On the ignition of metal particles,” Combustion, Explosion and Shock Waves, vol. 6, no. 4, pp. 412–422, 1970.DOI
  46. [46]V. Bolobov and N. Podlevskikh, “Numerical analysis of conditions for ignition of compact metal specimens and foil in oxygen,” Combustion, Explosion and Shock Waves, vol. 37, no. 6, pp. 655–663, 2001.DOI
  47. [47]R. Chen and W. Yeun, “Review of the high-temperature oxidation of iron and carbon steels in air or oxygen,” Oxidation of metals, vol. 59, no. 5, pp. 433–468, 2003.DOI
  48. [48]F. Li, Z. Sun, S. Luo, and L. Fan, “Ionic diffusion in the oxidation of iron—effect of support and its implications to chemical looping applications,” Energy Environ. Sci., vol. 4, pp. 876–880, 2011.DOI
  49. [49]L. Himmel, R. Mehl, and C. Birchenall, “Self-diffusion of iron in iron oxides and the Wagner theory of oxidation,” JOM, vol. 5, no. 6, pp. 827–843, 1953.DOI
  50. [50]A. Goursat and W. Smeltzer, “Kinetics and morphological development of the oxide scale on iron at high temperatures in oxygen at low pressure,” Oxidation of Metals, vol. 6, no. 2, pp. 101–116, 1973.DOI
  51. [51]D. Young, High temperature oxidation and corrosion of metals, vol. 1. Elsevier, 2008.
  52. [52]P. Kofstad, “Nonstoichiometry, diffusion, and electrical conductivity in binary metal oxides,” 1972.
  53. [53]A. Atkinson, “Transport processes during the growth of oxide films at elevated temperature,” Rev. Mod. Phys., vol. 57, pp. 437–470, Apr 1985.
  54. [54]R. Carter, “Thermal expansion of mgfe2o4, feo, and mgo: 2feo,” Journal of the American Ceramic Society, vol. 42, no. 7, pp. 324–327, 1959.
  55. [55]G. Holcomb, “A review of the thermal expansion of magnetite,” Materials at High Temperatures, vol. 36, no. 3, pp. 232–239, 2019.DOI
  56. [56]Y. Kozlovskii and S. Stankus, “The linear thermal expansion coefficient of iron in the temperature range of 130–1180 k,” in Journal of Physics: Conference Series, vol. 1382, p. 012181, IOP Publishing, 2019.DOI
  57. [57]M. Chase, NIST-JANAF Thermochemical Tables, 4th Edition. American Institute of Physics, -1, 1998-08-01 1998.
  58. [58]B. McBride, Coefficients for calculating thermodynamic and transport properties of individual species, vol. 4513. NASA Langley Research Center, 1993.
  59. [59]G. Hubbard, V. Denny, and A. Mills, “Droplet evaporation: Effects of transients and variable properties,” International Journal of Heat and Mass Transfer, vol. 18, no. 9, pp. 1003–1008, 1975.DOI
  60. [60]T. Hazenberg and J. van Oijen, “Structures and burning velocities of flames in iron aerosols,” Proceedings of the Combustion Institute, vol. 38, no. 3, pp. 4383–4390, 2021.DOI
  61. [61]G. Burgess and R. Waltenberg, The Emissivity of Metals and Oxides: Measurements with the micropyrоmeter, vol. 11. US Department of Commerce, Bureau of Standards, 1915.
  62. [62]Y. Touloukian and D. DeWitt, “Thermophysical properties of matter - the tprc data series. volume 8. thermal radiative properties - nonmetallic solids,” Tech. Rep. AD-A-951942/2/XAB; CNN: DSA900-73-C-2101, Purdue University, Lafayette, IN, 1972.
  63. [63]J. Jones, P. Mason, and A. Williams, “A compilation of data on the radiant emissivity of some materials at high temperatures,” Journal of the Energy Institute, vol. 92, no. 3, pp. 523–534, 2019.DOI
  64. [64]M. Soo, X. Mi, S. Goroshin, A. Higgins, and J. Berghorson, “Combustion of particles, agglomerates, and suspensions - a basic thermophysical analysis,” Combustion and Flame, vol. 192, pp. 384 – 400, 2018.DOI
  65. [65]X. Mi, S. Goroshin, A. Higgins, R. Stowe, and S. Binguette, “Dual-stage ignition of boron particle agglomerates,” Combustion and Flame, vol. 160, no. 11, pp. 2608–2618, 2013.DOI
  66. [66]N. Cabrera and N. Mott, “Theory of the oxidation of metals,” Reports on Progress in Physics, vol. 12, pp. 163–184, jan 1949.
  67. [67]E. Senyurt and E. Dreizin, “At what ambient temperature can thermal runaway of a burning metal particle occur?,” Combust. Flame, vol. 236, p. 111800, 2022.DOI

Paper details

vlnc:W_s4vnhcqw
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents