Mechanisms on the size partitioning of sodium in particulate matter from pulverized coal combustion

Combustion and Flame 182 (2017) 313–323

Contents lists available at ScienceDirect

Combustion and Flame journal homepage: www.elsevier.com/locate/combustflame

Mechanisms on the size partitioning of sodium in particulate matter from pulverized coal combustion Qian Huang, Shuiqing Li∗, Gengda Li, Qiang Yao Key laboratory for Thermal Science and Power Engineering of Ministry of Education, Department of Thermal Engineering, Tsinghua University, Beijing 100084, China

a r t i c l e

i n f o

Article history: Received 29 December 2016 Revised 16 April 2017 Accepted 18 April 2017

Keywords: Pulverized coal combustion Sodium Particulate matter Size partitioning behaviors Population balance modeling

a b s t r a c t Particulate matter (PM) generated from pulverized coal combustion is influenced by a process of vaporization, nucleation and condensation/reaction of semi-volatile minerals. The objective of the work is to identify the controlling mechanism of the enrichment of volatile sodium (Na). Three representative kinds of coal with different rank and mineral content, Zhundong lignite, Hami lignite and high-ash-fusion (HAF) bituminous, were burned in a 25 kW self-sustained pulverized coal combustor. Zhundong lignite, possessing medium rank among the three coal samples, exhibits the highest formation ability of ultrafine PM0.2 , indicating a prominent effect of mineral content over that of coal rank on PM formation. The size partitioning behavior of sodium is affected by the molar ratio Na2 O/(SiO2 +Al2 O3 ) instead of the absolute Na content. With this ratio increasing from 0.27 (HAF bituminous) to 20 (Zhundong lignite), the gas-to-particle conversion of Na transits from a surface-reaction controlled process to a nucleationcondensation/coagulation dominated pathway. The Na partitioning behavior of Zhundong lignite is then quantitatively interpreted by a population-balance-based theoretical approach. The model reveals the details on the competing processes of homogeneous nucleation and vapor condensation, resulting in that the simulated final particle mass size distribution exhibits reasonable agreement with the experimental result. The model has also successfully reproduced the d0p dependence of Na fraction in the ultrafine size regime from a process of homogeneous nucleation under the decreasing gas temperature from 1420 to 1270 K. © 2017 Published by Elsevier Inc. on behalf of The Combustion Institute.

1. Introduction The airborne particulate matter (PM) arises as an urgent environmental issue which attracts intensive attention in developing countries such as China and India, and pulverized coal combustion is one of the major source contributions [1–4]. The inhalable particulates become even more hazardous because of the toxic metals enriched in PMs [5]. Dozens of works have been dedicated to the formation of carbonaceous/mineral PMs and toxic matter enrichment processes during pulverized coal combustion [5–17]. In the high-temperature combustion environment, a process of vaporization, nucleation, condensation (known as heterogeneous nucleation) and reaction of minerals has been recognized to play a significant role in the formation of ultrafine particulates as well as the concentration of semi-volatile species in PMs [5–14]. The vaporization and nucleation process, which is more prevalent in the submicron range, causes a size-dependent partition-

∗

Correspondence author. E-mail addresses: [email protected], [email protected] (S. Li).

http://dx.doi.org/10.1016/j.combustflame.2017.04.026 0010-2180/© 2017 Published by Elsevier Inc. on behalf of The Combustion Institute.

ing of semi-volatile species across the whole particle size distribution (PSD) of PMs. It has been reported that the mass fraction −2 of semi-volatile species in the fine PM exhibits d−1 p , d p or even 0 d p dependence, respectively, because of the different mechanisms of gas-to-particle conversion [5,18]. For instance, as far as the easily volatilized arsenic (As), selenium (Se) and antimony (Sb) are concerned, the d−1 dependence was usually found in the literap ture [10–12,18], which can be explicable by a surface-reaction controlled mechanism. However, an exception with a d−2 p dependence was also reported for As and Se [12,13], implying a process of direct vapor condensation. These diverse behaviors may be related to coal properties, combustion conditions, etc. Therefore, further fundamental studies are needed to cover different ranks of coal samples. The dynamic behavior of sodium (Na), among the semi-volatile elements from coal, is consistently a special concern, particularly for the reason that the use of low-rank coal is inevitably increased in recent years [19]. During coal combustion sodium is vaporized in devolatilization stage as well as in char burningout stage [20–24]. The subsequent gas-to-particle conversion of the released sodium significantly contributes to ultrafine PM

314

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

Nomenclature

Table 1 Proximate and ultimate analyses of the coal samples. Zhundong lignite

diffusion coefficient of gas monomer, m2 /s particle diameter, m number of monomers in nascent particles, # particle current, #/(m3 •s) homogeneous nucleation rate of nascent particle with critical volume Vp ∗ , #/(m3 •s) Kn Knudsen number Kp chemical equilibrium constant kxs reaction flux at the particle surface, #/(m2 •s) m1 mass of one gas monomer, kg N1 number concentration of gas monomer, #/m3 N1s saturated surface number concentration of gas monomer, #/m3 t residence time in the furnace, s Vp particle volume, m3 xs surface mole fraction of semi-volatile species Y1 mass fraction of Na species in gas monomer Yi mass fraction of species i in particles δ (Vp - Vp ∗ ) delta function ρp particle density, kg/m3 D dp g∗ Icur I(Vp ∗ )

formation and fouling aggravation problems [14,25–27]. Previous studies usually reported a d p−2 dependence of Na [18,28–30], which indicated a heterogeneous-condensation-dominated conversion pathway. Moreover, Takuwa et al. found a d0p dependence in the ultrafine PMs (e.g., smaller than PM0.3 ) from the combustion of a sodium rich coal [30]. More recently, it was further approved that the sodium-rich Zhundong lignite exhibits a much larger formation ability of PM0.1 with highly enriched Na content, in contrast to normal bituminous coals [31]. These experimental findings suggest that, under certain combustion circumstances, the gaseous Na species should go through a pathway of homogeneous nucleation followed by coagulation growth. A justification of the proposed mechanism, which initiates from qualitative observations in the preliminary studies, calls for more extended PSD data on the ultrafine mode PM, as well as a detailed dynamic modeling of this complicated process. When theoretical approaches are concerned, it is noteworthy that the nucleation–coagulation mechanism of ultrafine PM is quite similar to that existing in flame synthesis of nanoparticles [32,33]. Therefore, the population balance model (PBM) can be helpful to quantitatively interpret the size partitioning of Na during the combustion of low-rank coals. This method has been applied to simulate soot formation and char particle evolution during coal combustion [34–36]. Different from nanoaerosol synthesis system, significant amount of micron-sized ash particle is also formed via mineral coalescence and fragmentation during pulverized coal combustion [16,17]. The coarse PMs provide a large number of sites for gas–solid surface reaction and vapor condensation, −2 leading to the aforementioned d−1 p and d p dependences. Hence, a model based on the PBM framework, covering much broader particle size range and multiple compositions, is expected to reveal the crucial details on the Na conversion process as well as its effect on PM evolution. Such evolution systems containing alkali–sulfur– chlorine have been investigated in the PBM model and further validated by bench-scale experiments [37,38]. The applicability of nucleation theory and the formation of sulphate aerosols are drawn from these studies, with further kinetic details revealed in [39,40]. However, it is still challenging to apply the PBM theories to match or interpret the field data of Na partitioning in coal combustion. The objective of the work is to divulge the controlling mechanism of the enrichment of volatile sodium (Na), during the com-

Hami lignite

HAF bituminous

33.00 26.60 40.40 20.74

24.10 20.30 55.50 25.20

Ultimate analysis (wt%, dry and ash free basis) C 71.60 60.28 H 3.16 2.55 N 0.78 1.08 0.52 0.41 Stotal Cl 0.06 0.12 O (by difference) 23.85 35.52

82.50 4.39 0.89 0.93 0.02 11.27

Proximate analysis (wt%, dry basis) Volatile matter 30.58 Ash 5.88 Fixed carbon 63.54 HHV (MJ/kg, dry basis) 28.83

Table 2 Ash compositions of the coal samples. Ash composition (wt%)

Zhundong lignite

Hami lignite

HAF bituminous

SiO2 Al2 O3 Fe2 O3 CaO MgO SO3 K2 O Na2 O P2 O5 TiO2

28.53 3.27 4.01 32.78 2.88 21.47 0.57 6.19 NDa 0.30

57.04 14.12 4.80 5.60 0.91 3.61 1.79 1.92 0.34 1.01

56.80 26.10 6.98 2.68 0.70 2.10 1.02 0.20 0.48 1.20

a

ND – Not Detected. Table 3 The water soluble and organically-bounded portions of mineral elements. wt%

Zhundong lignite

Hami lignite

HAF bituminous

Na Ca Mg Fe K S Si

93.3 69 80.8 0.01 24.2 27.3 0.6

65.2 65.3 77.4 0 6.2 16.3 0.3

18.5 81.2 69.8 0 1.5 9.7 0.1

bustion of coal samples with variations in rank and mineral content. We investigated the high-ash-fusion (HAF) bituminous, the Zhundong lignite, and a lower-rank Hami lignite in the 25 kW self-sustained combustor. We compared the PM formation abilities and Na partitioning behaviors among the several coal samples. We identified the key factor, molar ratio of Na2 O/(SiO2 +Al2 O3 ), in determining the Na conversion mechanisms. Inspired by the experimental results, we further investigated the case of Zhundong lignite, in which homogeneous nucleation of Na was thoroughly described by a multi-compositional PBM model. 2. Experimental 2.1. Coal properties Three coal samples, Zhundong lignite, Hami lignite and HAF bituminous, were investigated. The coal properties are listed in Tables 1 and 2. The chemical fractionation results are listed in Table 3. In Zhundong lignite, more than 90% of Na are either water soluble or organically-bounded, as compared to 18.5% in HAF bituminous. The several coal samples also exhibit distinct differences in coal rank, as shown in the coalification diagram Fig. 1a. Considering the vast differences in mineral contents and occurrence modes, the coal samples are chosen in such a manner aiming at

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

315

Fig 2. Schematic picture of the down-fired pulverized coal combustor and the particulate sampling system.

Fig 1. Coal properties. (a) Coalification diagram of the coal samples. (b) A typical volume PSD of input coal.

clarifying the effects of coal rank and mineral content on PM formation and size partitioning of sodium. The coal particles were milled and a typical volume PSD of input coal is shown in Fig. 1b, as measured by Malvern (Mastersizer 20 0 0). Before each run of combustion experiments, the coals were dried to less than 5% moisture to ensure a steady feeding flow into the furnace. 2.2. Pulverized coal combustor and operation conditions The coals were burned in the 25 kW down-fired pulverized coal combustor, equipped with a two-staged dilution system for fine particulate sampling. The experimental set-up is schematically shown in Fig. 2. More detailed descriptions can be found in our previous works [7,31]. The quasi one-dimensional combustion process was self-sustained in the furnace with inner tube diameter 150 mm and a total length of 3.8 m. The aerodynamic similarity of this pilot-scale furnace to the practical boilers, e.g. in the aspects of both inter-particle and particle–fluid interactions, possesses its advantage over the single particle based reactors such as drop tube furnace in exploring the collective behaviors during pulverized coal combustion, while avoiding additional complexities brought up by the three-dimensional turbulent flow in practical combustion instruments. Therefore, the self-sustained combustor leads to more reliable results of the mass PSDs of coal generated PMs.

Fig 3. The gas temperature profiles of the three coal samples.

Table 4 lists the operation conditions. Coal feeding rate was adjusted to an energy input of ∼25 kW for all the coal samples. The gas temperatures shown in Fig. 3 were measured by the axialpositioned S-thermocouples, and exhibited similar profiles among the three coal samples. In all the three cases, the last point of gas temperature data corresponds to the sampling Port 4 with the particle residence time ∼1.5 s, which can be seen as a simulating condition for the furnace exit in practical boilers. Moreover, the gas components of O2 , CO2 , CO, NO and SO2 at the sampling Port 4 were monitored continuously by a gas analyzer (ABB EL3020) to

316

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323 Table 4 Experimental operation conditions of the combustor.

Coal feeding rate (dry basis), kg/h Air, Nm3 /h Carbon fraction in PM10+ , wt%

Zhundong lignite

Hami lignite

HAF bituminous

2.8 26.4 4.5

3.9 26.9 1.0

3.5 22.4 7.8

ensure a steady operation condition. With O2 concentration 6–8% and CO concentration <300 ppm, it is reasonable to neglect the carbonaceous species in the fine PM10 mode (see also the detailed discussion in the supplementary material). The particulate sampling was achieved by sucking the hightemperature particle-laden flue gas followed by two-stage dilutions to eliminate any unwanted nucleation or coagulation process in the sampling probe. Isokinetic sampling was maintained in the meantime [7,31]. The total dilution ratio was kept as 70–100 to get asymptotic PSD curves, as detailed in the supplementary material. The Dekati Electrical Low Pressure Impactor (ELPI) was used to measure the mass PSDs of PM10 collected on greased Al foils after the flue gas flowing through the cyclone PM10 cutter. An updated version ELPI+ with the finest stage of 0.017 μm in the cut-off diameter was later used to measure the mass PSD of PM10 produced by Hami lignite. The Scanning Mobility Particle Spectrometer (SMPS) from TSI Inc. was also used to measure the nano-sized PSD generated by Zhundong lignite. Other groups of particulate samples were collected on teflon filters for compositional analyses by Inductively coupled Plasma – Atomic Emission Spectrometry (ICP-AES) and on nylon filters for morphology and elemental analyses by SEM-EDS (JSM-6301F and Zeiss Merlin). 3. Theoretical 3.1. Population balance modeling framework As far as the Na enrichment onto the ash particle during coal combustion is concerned, we apply the population balance model (PBM) to interpret the dynamic behaviors of Na vapor. In principle, the model solves the evolution of PSDs through the vaporparticulate interactions including homogeneous nucleation, heterogeneous nucleation, surface reaction and the particle coagulation [41], which is shown as follows:

∂n = ∂t



   ∂n ∂n + ∂ t nucleation ∂ t condensation + surface reaction   ∂n + ∂ t coagulation

(1)

In view of the complex environment encountered during coal combustion, we treat the dynamic behaviors of Na vapor in the postflame region of the 25 kW down-fired furnace as a quasi-onedimensional process. The detailed governing equations used in this work are presented in Table 5, which solve the temporal evolutions of the number PSD of particulate matter, n(Vp , t)dVp , and the volume fractions of Na in the size-segregated ash particles Y (Vp , t). In Eq. (T5.1) for particulate phase, the second term on the LHS accounts for the effects of heterogeneous condensation and surface reaction. For sufficiently large particles, we have Icurr = nF (d p )vm , where vm is the volume of gas monomer and F is the deposition rate of gas monomers (#/s) [41]. F is determined by both heterogeneous condensation and the surface reaction [18,41]. In the diffusion-controlled limit case, the deposition rate F is

F (d p ) = 2π d p D(N1 − N1s ) × f (Kn ).

(2) D = ATB .

The diffusion coefficient of the gas monomer We take A = 1.425 × 10−10 m2 /s, B = 1.88 for NaCl, and A = 1.584

× 10−10 m2 /s, B = 1.50 for Na2 SO4 [41]. The correction term f for the condensation rate beyond the continuum regime is f (Kn ) = (1 + Kn )/(1 + 1.71Kn + 1.333K n2 ) [41]. The Knudsen number Kn = 2λ/dp decides the regime in which the gas–particle interaction lies, where λ is the mean free path of the gas molecules. Kelvin effect is included to take account of the curvature-induced elevation of N1s , especially for ultrafine particles. In the other surface-reaction-controlled limit case, we have

F (d p ) = kxs π d2p ,

(3)

where k is related to the reaction rate constant and xs is the surface concentration of the gaseous species. The third term on the LHS determines the formation rate of nascent particles from homogeneous nucleation. The nascent particle size Vp ∗ and the formation rate I (with unit #/(m3 •s)) are strongly dependent on the local degree of supersaturation S. The detailed forms to referred to Ref. [42]. The terms on the right-hand side (RHS) of Eq. (T5.1) account for the effect of particle coagulation. As for the coagulation kernel function β (m3 /s), we use a summation of the Brownian kernel applicable for the transition size regime (the Fuchs interpolation form [43]) and the laminar-shear-induced coagulation kernel [41]. The terms on the RHS of Eq. (T5.2) are the consumption rates of gas monomers by either homogeneous or heterogeneous nucleation. As an analogy to Eq. (T5.1), we develop Eq. (T5.3) to solve the temporal evolution of the total volume distribution function of Na, nVp Y dVp . The equation reduces to the volume based population balance equations Eq. (T5.1) when assuming Y to be constant. Similarly, the second and third terms on the LHS represent the effects of heterogeneous condensation and surface reaction. The last term on the LHS accounts for the contribution of homogeneous nucleation. The terms on the RHS describe the particle coagulation effect. Faced with the problem of unaffordable equation numbers after full discretization for each particle size, a “discrete-nodal” method (DNM) is developed to meet the need of practical computation with adequate accuracy. As detailed in Prakash et al.’s work [44], the size-splitting method is used to convert the PSDs in the continuous particle volume space to a series of discretized volumes. Both Eq. (T5.1) and Eq. (T5.3) are discretized in the DNM framework. The surface growth term is numerically dealt with in the form of coagulation growth. If only the condensation is accounted for, the corresponding kernel function β1 j = 2π d p, j D × f (Kn ). 3.2. Size partitioning laws of semi-volatile species Eq. (T5.3) can be used to derive rigorously the size partitioning laws of semi-volatile species in ash particles. Based on the thin-film assumption, the particle current is neglected, i.e. ∂ (IcurrVpY )/∂ Vp = 0 and ∂ n/∂ t = 0. After further neglecting the nucleation and coagulation terms, Eq. (T5.3) is reduced to dY /dt = IcurrY1 /(nVp ) = vmY1 F /Vp . If no pre-existing semi-volatile species in the ash particle is assumed, we have Y (d p ) ∼ F /Vp ∼ F /d3p . This simplified relation was usually taken as the first step in previous studies [5,18,28,30]. In the diffusion-controlled limit case, as can be seen from Eq. (2), we have F ∼ d2p in the free molecular regime with Kn  1 and F ∼ dp in the continuum regime with

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

317

Table 5 Population balance and component conservation equations. Population balance equations Particle

∂ n ∂ (Icurr ) + − I (Vp∗ )δ (Vp − Vp∗ ) ∂t ∂ Vp =

1 2



Vp

Vp∗

β (Vp − V˜p , V˜p )n(Vp − V˜p , t )n(V˜p , t )dV˜p − n(Vp , t )



∞

Vp∗

β (Vp , V˜p )n(V˜p )dV˜p

(T5.1)

Vapor

  dN1 = −g∗ I Vp∗ − dt



∞

Vp∗

   

F V˜p n V˜ p dV˜ p

(T5.2)

Component conservation equations Particle

∂ (nVpY ) ∂ (IcurrVpY ) + − IcurrY1 − I (Vp∗ )Vp∗Y1 δ (Vp − Vp∗ ) ∂t ∂ Vp 

=

Vp

Vp∗

β (Vp − V˜p , V˜p )n(Vp − V˜p , t )n(V˜p , t )V˜pY (V˜p , t )dV˜p − nVpY



∞

Vp∗

β (Vp , V˜p )n(V˜p )dV˜p

(T5.3)

Vapor

dY1 =0 dt

Kn  1. Hence, we have Y (d p ) ∼ d p−1 for the ultrafine PMs and Y (d p ) ∼ d p−2 for the coarse mode PMs. In contrast, the surfacereaction-controlled limit case possesses F ∼ d p2 , as seen from

Eq. (3), resulting in that Y (d p ) ∼ d p−1 in the whole size range of particle distribution. However, there are cases where homogeneous nucleation, which is neglected in the above analysis, may play a prominent role in affecting the mass and composition of ultrafine mode PMs. Eq. (T5.3) predicts a size dependence of Y (d p ) ∼ d p0 by merely considering the nucleation term I (Vp∗ )Vp∗Y1 . Nevertheless, this phenomenon becomes important during the combustion of Zhundong lignite and some other coals from the literature [30]. Therefore, a detailed modeling in the PBM framework is of necessity to the interpretation of the complex Na behaviors under these circumstances. 4. Results and discussion 4.1. PM10 formation of various coal samples Figure 4a shows the PSDs of PM10 , on the unit ash basis, generated by the three kinds of coal at combustion exit (Port 4). In particular, Fig. 4b presents the mass PSD of ultrafine PM0.1 formed by Zhundong lignite, transformed from the number PSD measured through SMPS. A reasonable consistence is achieved between the measurements of ELPI and SMPS in the overlapped size range. It is noteworthy that Zhundong lignite, possessing medium rank, produces exceptionally high PM0.1 (or PM0.2 ) concentrations per gram of input ash. Sodium (Na) and sulfur (S) are highly enriched as the major constituents in PM0.2 samples detected by SEM-EDS analysis, as shown in Fig. 5a. This finding agrees well with the chemical fractionation result in Table 3 that the sodium in the raw Zhundong lignite is present in easily volatilized forms. Furthermore, Fig. 5b shows the interactions among semi-volatile species and bulk ash particles, including scavenging phenomenon and the irregular condensed layer. Semi-volatile elements Na, Mg and S are enriched in deposition layer 3 and the inter-particle “glue” site 4. Comparatively, both HAF bituminous and Hami lignite show apparently lower ability of PM0.2 formation than that of Zhundong lignite. For HAF bituminous, the contribution of Na to total ultrafine

(T5.4)

PMs is negligible due to its low concentration in raw coal. The ultrafine PM is mainly composed of refractory elements Si and Fe, as shown in Fig. S5 in the supplementary material. Hami lignite, however, possesses even higher Na content on the unit coal basis than Zhundong lignite (see Table 2). The reason for the lower formation ability of ultrafine PM is inferred to be the higher Si and Al contents and thus lower ratio of Na2 O/(SiO2 +Al2 O3 ), leading to an enhanced portion of Na to be converted to silicate form rather than freely evaporated and condensed [14,28]. It is thus concluded that the mineral content and mode, instead of the coal rank, are more prominent in determining the formation ability of ultrafine particle PM0.1 or PM0.2 .

4.2. Size partitioning of Na for various coal samples Figure 6a–e shows the weight percentage (wt%) of Na in sizesegregated particulate matter, as measured by ICP-AES. Data for the Spring Creek (SC) coal and the Black Thunder (BT) coal are adopted here from Ref. [30] for an insightful comparison. Figure 6f further shows the molar ratio Na2 O/(SiO2 +Al2 O3 ) for the five coal samples. There is a clear transition in the mode of size partitioning of Na from the SC coal (Fig. 6a) to HAF bituminous (Fig. 6e), with continuously decreasing in the value of the molar ratio Na2 O/(SiO2 +Al2 O3 ). Both the SC coal and Zhundong lignite exhibit an apparent trend of Y (d p ) ∼ d p0 in the ultrafine PM0.2 regime, suggesting a process of homogeneous nucleation based on our previous discussion. Besides, the rough d p−1 and d p−2 dependences are identified for coarser PM size ranges of these two coals, consistent with the heterogeneous condensation mechanism caused by the highly supersaturated vapor. Comparatively, the BT coal and Hami lignite, with much lower values of Na2 O/(SiO2 +Al2 O3 ), have no typical d0p dependence in the ultrafine regime, indicating that the Na vapor is not supersaturated enough for the occurrence of homogeneous nucleation. Also, both d p−1 and d p−2 dependences emerge in the coarser size range. It is then speculated that both condensation and surface reaction play a key role in the gas-to-particle conversion pathways of Na. Finally, HAF bituminous possesses much lower values of the ratio Na2 O/(SiO2 +Al2 O3 ). With much lower Na content in PM samples and a rough d p−1

318

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

Fig 4. The PM10 formation abilities: (a) measured by ELPI/ELPI+ for the three coal samples; (b) measured by SMPS/ELPI for Zhundong lignite. Some data are adopted from Ref. [31].

Fig 5. SEM-EDS analyses of the PM samples formed by Zhundong lignite: (a) PM0.1 aggregates; (b) Supermicron particle samples with the phenomena of surface scavenging and condensation.

dependence, it is inferred that the surface reactions dominate the size partitioning behavior of volatile Na for this bituminous. In general, instead of the absolute Na content, the molar ratio of Na2 O/(SiO2 +Al2 O3 ) is demonstrated to govern the gas-to-particle conversion of Na to a more substantial extent. As further explored in Fig. S6 in the supplementary material, this ratio, actually equivalent to another indicator (acid soluble Na2 O)/(SiO2 +Al2 O3 ) in the literature [45] but simpler, can be used as an index to characterize the size partitioning of volatilized Na. The assumption that this indicator affects the quantitative partitioning of sodium in submicron PM mode has been examined by several studies in the literature, as shown in Fig. S7 in the supplementary material, with our data included [46]. A closer investigation on the correlation between this index and the Na partitioning in the ultrafine mode PM0.2 is shown in Fig. 7. The reasonable correlation in Fig. 7 indicates that the fraction of Na accumulated within the ultrafine regime is inversely proportional to the existence of Si-Al components. In the cases with low and moderate values of molar Na2 O/(SiO2 +Al2 O3 ), the Na-coal-mineral reactions should be considered. Here we evaluate the reaction fluxes for HAF, Hami and BT coals. The mass fraction of Na in the particle, Y(dp ), has the

form

Y (% ) =

6 × 108 (kx∞ m1 )tr

ρp

1 + y0 , d p ( μm )

(4)

where tr is the reaction time and y0 is the initial mass fraction. The inset figure in Fig. 8 shows the linear fitting of Y as the function of 1/dp for the three coal samples in the micron size range. The average reaction fluxes kx∞ m1 are determined from the fitted slopes when assuming tr = 1.0 s, as shown in Fig. 8. The results demonstrate that the Na enrichment flux kx∞ m1 increases with the Na–Si indicator. As discussed earlier, the larger Na–Si indicator implies a more extensive vaporization of Na, resulting in a larger vapor concentration x∞ . Moreover, k is the combined coefficient determined by reaction and condensation. With larger Na–Si indicator, the coefficient k is affected by the condensation process to a greater extent. In contrast, during the combustion of Zhundong lignite with large value of Na2 O/(SiO2 +Al2 O3 ), the nucleation of gaseous Na significantly affects the ultrafine PM fraction. The vapor condensation onto the newly formed and pre-existing particles occurs simultaneously, leading to a complexly dynamic process of PM evolution and Na partitioning. Then, a detailed modeling in the PBM

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

319

Fig 6. The wt% of Na in size-segregated PMs for various coal samples. Data for SC and BT coals are adopted from ref. [30].

Fig 7. The mass fraction of Na in ultrafine mode PM0.2 as a function of the molar ratio of Na2 O/(SiO2 +Al2 O3 ) for the various coal samples. The data for HB lignite and semi-char are adopted from [46].

framework is introduced below to quantitatively interpret this process. 4.3. Model interpretations on Na partitioning in Zhundong lignite As mentioned above, the gas-to-particle conversion of Na during the combustion of Zhundong lignite is governed by nucleation and heterogeneous condensation. The surface reaction gradually becomes insignificant in this case after the condensed layer covers most of the reaction sites. The condensation flux F is thus determined by Eq. (2) in the model.

Fig 8. The rates of Na enrichment for the three coal samples. The left bars correspond to the molar ratio of Na2 O/(SiO2 +Al2 O3 ), and the right bars denote the flux kx∞ m1 . The inset figure shows the data fitting of wt% of Na v.s. 1/dp for the three coal samples in the micron size range.

4.3.1. Model inputs The PBM simulation starts at the moment when the Na vapor becomes saturated as the temperature drops in the postflame region of the furnace. The amount of the vaporized Na is inversely estimated from the measured Na retention (2.0 wt% by ICP-AES) in bulk ash PM10+ collected at Port 4. The retention of Na in bulk ash comes from originally nonvolatilized part (mainly acid insoluble), the part trapped by Si–Al compounds during vaporization and the gas–solid reaction captured part [14,28]. It is assumed that the supermicron modes

320

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

Fig 9. The input number PSD and the modal lognormal distributions. The experimental result transited from ELPI measurement is also included. Table 6 Parameters for the input modal lognormal distributions.

1 2 3 4

Total number concentration Ni 0 (#/Nm3 )

Median diameter CMDi (μm)

Width lnσ i (μm)

1.0 × 1014 5.0 × 1012 6.4 × 1010 1.25 × 108

0.002 0.08 1.30 30

0.20 0.90 0.54 0.25

PM1 – 10 and PM10+ possess this 2.0% of Na as the non-volatilized “base-line” part. The vaporized Na is thus taken as the rest portion (50.4%). The gaseous Na species has to be thermodynamic stable under the experimental conditions. A thermodynamic calculation of the Na–S–Cl–H–O system is detailed in the supplementary material. The results in Fig. S8 indicate that, under the temperature range of 110 0–170 0 K in the postflame region, the dominant occurrence forms of Na are Na2 SO4 and NaCl. This is consistent with the observed Na–S enriched ultrafine mode PM0.2 in the d p0 dependent size range. It has been pointed out that the sulfate found in the particulate phase comes from vapor phase sulfation reactions because of the relatively short residence time [38–40]. Hence, the Na2 SO4 vapor is formed from the sulfation of Na atom and NaOH which are stable at higher temperatures (see Fig. S8). With these thermodynamic considerations, the gaseous Na is thus assumed to be the mixture of NaCl and Na2 SO4 . The Na2 SO4 vapor saturates at 1505 K, and the dynamic modeling lasts for 1.0 s to the sampling temperature 1100 K. The time step used in the simulation is ∼4 μs. Moreover, there is a potential for sulfation of NaCl vapor in the lower temperature range through the reaction [38],

2NaCl+SO2 +H2 O+0.5O2 =Na2 SO4 +2HCl.

(5)

If the sulfation reaction of NaCl vapor is incorporated, the reaction Eq. (5) is assumed to be in equilibrium at all times. The initial PSD of the ash particle, with no accurate data available, is assumed to be the adding of four modal lognormal distributions, as shown in Fig. 9. Each of the modal distribution has the form



dn0i (d p ) N0 (ln d p − ln CMDi )2 = √ i exp − 2 d ln d p 2π ln σi 2ln σi



.

(6)

The parameters Ni0 , CMDi and σ i denote total number concentration, median diameter and width, respectively, as listed in Table 6. The modal distributions 2 and 3 are chosen to fit the measured final PSD in the diameter range between 0.2 μm and 10 μm,

Fig 10. The temporal evolution of number PSDs during combustion of Zhundong lignite. The measured PSD result at t = 1 s is also shown. : PSD measured by SMPS, and : PSD measured by ELPI.

because of the limited surface area concentration of the coarse mode PMs. The modal distribution 4 is assumed for bulk mode PM10+ . The modal distribution 1 represents the ultrafine particle formed from refractory sub-oxides at higher temperatures before the nucleation of Na compounds. Considering the homogeneous nucleation of SiO2 at 1700 K [9], an estimation by the monodispersed particle coagulation reveals that the mean particle size will not exceed 2 nm when temperature drops to 1500 K in less than 0.1 s. 4.3.2. PBM model prediction of Na size partitioning Figure 10 shows the temporal evolution of number PSD predicted from the model. Two cases are included in the simulation, i.e. one case without NaCl conversion, and the other case where the sulfation of NaCl, Eq. (5), is in equilibrium at all times. Figure 10a presents the input PSD at t = 0, and Fig. 10d includes the measured PSDs transited from SMPS and ELPI (red color symbols) for the comparison. It is noted that, at the sampling point t = 1 s, both numerical cases, with and without NaCl conversion, well predict the number PSD peak at the diameter of ∼30 nm measured by SMPS. The peak initially appears at 1–2 nm at t = 0.4 s and gradually grows up. The PSD evolution is the consequence of the complex vapor– particle interactions. In order for a deeper insight into the dynamic process, Fig. 11 reveals the gaseous Na concentrations and the consumption rates by both homogeneous nucleation and heterogeneous condensation. It indicates that during a period of t = 0.22–0.58 s with decreasing gas temperature from 1420 to 1270 K, the partial pressure of Na2 SO4 exceeds the critical value pcrit,Na2 SO , which is a product of the critical saturation ratio (Scrit ) 4

and the saturation pressure ( psat,Na2 SO ) [9,41]. Correspondingly in 4

Fig. 11b, the homogeneous nucleation emerges, forming the PSD peak at sizes of 1–2 nm which appears in Fig. 10b. The incipient particles undergo an enhanced process of heterogeneous conden-

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

321

Fig 12. Temporal evolution of Na partitioning in vapor and particulate phases. The experimental result at Port 4 (corresponding to t = 1 s for model) is also included. Left column: with NaCl conversion; Right column: no NaCl conversion.

Fig 11. (a) The temporal evolution process of vapor phase NaCl and Na2 SO4 , where pi , psat,i and pcrit,i denote the partial pressure of component i, its saturation partial pressure, and the critical partial pressure for the occurrence of homogeneous nucleation, respectively; (b) the vapor consumption rates of monomer by homogeneous nucleation and heterogeneous condensation.

sation as well as particle coagulation, and grow up to the final size range of 10–100 nm. Both cases give similar results before t = 0.6 s. However, after that moment when gas temperature drops below 1270 K, for the case assuming the sulfation of NaCl Eq. (5) in equilibrium, about 90% of NaCl are eventually converted to Na2 SO4 (see Fig. 10a). As a consequence, Na2 SO4 vapor undergoes a second period of intense homogeneous nucleation from t = 0.58–1 s, which results in the nano-sized peak of predicted PSD in Fig. 10d. On the contrary, for the case assuming no NaCl conversion, homogeneous nucleation does not occur after 0.58 s and the rate of heterogeneous condensation decreases gradually (see Fig. 11b). The predicted final PSD in the case of no NaCl conversion exhibits more similar trend to the SMPS measured results in the size range of 1–10 nm. This lack of nano-sized PSD peak also implies that the nucleation within the probe has been inhibited by the two-stage dilution sampling process. In addition, the model solves the evolution of Na partitioning in both vapor and size-segregated particles, as illustrated in Fig. 12. Experimental results sampled at t = 1 s are also included, in which the portion of gaseous Na retention, denoted as “gas”, is obtained from the mass balance of total Na content. It can be seen that the bulk ash particles PM10+ scavenge only a small portion of gaseous Na. The vapor phase Na is primarily transformed to PM10 particles. In particular, the mass fraction of Na in PM1 – 10 particles increases from the initial 5.5% to final 12.3% due to heterogeneous condensation. However, the two predictions, with and without NaCl conversion, show vast differences in PM1 particles and vapor retentions. In the case where NaCl is converted, the nucleation process after

Fig 13. Simulated and experimental results of wt% of Na in size-segregated PMs at Port 4 formed by Zhundong lignite.

t = 0.6 s results in a final 7.1% of Na remained in the vapor phase and 23.2% in PM0.1 . In contrast, when assuming no NaCl conversion, there is a 30.0% retention in vapor phase, which is closer to the experimental result of 27.1%. This comparison, together with the PSD result in Fig. 10d, implies a much weaker conversion of NaCl sulfation than that predicted by the thermodynamic equilibrium. It may be possibly attributed to the kinetic restriction on the reaction rate. According to a rate-controlling elementary SO2 /SO3 reactions proposed by Glarborg and Marshall [39], it is approximately estimated that the reaction time scale is about 1–2 orders higher than the residence time of 1.0 s. Figure 13 shows the size-dependent partitioning of Na in particulate phase, with both experimental and modeled results. The modeling predictions reproduce the plateau of d0p dependence in the ultrafine regime because of the formation of Na2 SO4 -dominant particles, and agree reasonably well with experimental results in the whole particle size range. Therefore, it is demonstrated that the gas-to-particle conversion of Na during Zhundong lignite combustion is a physically dominant nucleation-condensation/coagulation process, which is prominent in shaping the PSDs shown in Fig. 10. 4.3.3. Model remarks The model integrates the insights from alkali–S–Cl interactions [37,38] with the accurate measurements in the 25 kW furnace sys-

322

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323

tem. The dynamic simulation, which interprets the origination of the dp 0 -dependence for Zhundong lignite, illustrates that the nucleation of gaseous Na2 SO4 occurs within the postflame temperature range of 1420–1270 K. This process is currently beyond the reach of direct experimental detection in this pilot scale furnace. We also remind the readers that Fig. 11a suggests that NaCl vapor stays unsaturated at sampling Port 4. However, as shown in Fig. 5a, the ultrafine PM0.1 has a Cl content of 1–3 wt%. There are two possible reasons. First, during our two-stage dilution sampling process, a slight condensation may still occur during the secondstage quenching, although the nucleation is inhibited by the firststage isothermal dilution. Second, the current model does not consider the surface vapor pressure reduction and the increment in dew point during the binary condensation [37]. However, with the minor contents of Cl in PMs, our modeling can still reveal the dominant postflame process of Na evolution. More insights can be obtained by evolving the model into a predictive tool applicable for a wider variety of coals and combustion conditions. The gas phase reactions [39,40] should be incorporated in manner of Eq. (5). The Na–Si/Al surface reactions are coupled with PBM through Eq. (3), where k is related to the Arrhenius reaction rate. The reactions of gaseous Na2 O / NaOH / NaCl with solid SiO2 /2SiO2 •Al2 O3 are of particular significance [25,28,47].

5. Conclusions The fine particulate formation and size partitioning of sodium of a typical Zhundong lignite was investigated in a 25 kW downward pulverized coal combustor, by using Hami lignite and HAF bituminous as contrast fuels. The main summaries are drawn as follows. (1) Among the three coal samples, Zhundong lignite exhibits the highest formation ability of ultrafine PM0.2 . It is found that the content of minerals plays a more prominent role than the coal rank in determining fine particulate emissions. (2) The molar ratio Na2 O/(SiO2 +Al2 O3 ) is demonstrated to be the key factor in characterizing the Na partitioning behaviors for several coals involved in the study. As the value increases, the partitioning of Na transits from a surfacereaction controlled process for HAF bituminous, to a nucleation – condensation/coagulation dominated process for Zhundong lignite, which exhibits a remarkable d0p dependence in the ultrafine PM size range. (3) A model on the framework of population balance is established to interpret the dynamic process of Na partitioning, in which the competing processes of homogeneous nucleation and vapor condensation are particularly considered. The simulated results for the case of Zhundong lignite are well consistent with the experimental data of particle mass size distribution at the furnace exit. The PBM model has successfully reproduced the d0p dependence of Na fraction in the ultrafine size regime under the circumstance of decreasing gas temperature from 1420 to 1270 K. The Na partitioning implies a low conversion of NaCl sulfation before the flue gas temperature drops down to 1100 K.

Acknowledgment This work was mainly funded by the National Sciencetechnology Support Plan of China (No. 2015BAA04B03) and the National Natural Science Foundation of China (No. 51390491). The authors are grateful to Dr. Mengmeng Yang for his kind help and constructive discussions on the modeling work.

Supplementary materials Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.combustflame.2017.04. 026. References [1] Q. Yao, S.Q. Li, H. Xu, J.K. Zhuo, Q. Song, Studies on formation and control of combustion particulate matter in China: a review, Energy 34 (2009) 1296–1309. [2] D.A. Chu, Y.J. Kaufman, G. Zibordi, J.D. Chern, J. Mao, C.C. Li, B.N. Holben, Global monitoring of air pollution over land from the Earth Observing System – Terra Moderate Resolution Imaging Spectroradiometer (MODIS), J. Geophys. Res. – Atoms. 108 (2003) 4661–4678. [3] Y.H. Zhang, X.L. Zhu, S. Slanina, M. Shao, L.M. Zeng, M. Hu, M. Bergin, L. Salmon, Aerosol pollution in some Chinese cities (IUPAC Technical Report), Pure Appl. Chem. 76 (2004) 1227–1239. [4] Q. Wang, M. Shao, Y. Zhang, Y. Wei, M. Hu, S. Guo, Source apportionment of fine organic aerosols in Beijing, Atmos. Chem. Phys. 9 (2009) 8573–8585. [5] W.P. Linak, JostO.L. Wendt, Toxic metal emissions from incineration: mechanisms and control, Prog. Energy Combust. Sci. 19 (1993) 145–185. [6] F. Carbone, F. Beretta, A. D’Anna, Multimodal ultrafine particles from pulverized coal combustion in a laboratory scale reactor, Combust. Flame 157 (2010) 1290–1297. [7] J.K. Zhuo, S.Q. Li, Q. Yao, Q. Song, The progressive formation of submicron particulate matter in a quasi one-dimensional pulverized coal combustor, Proc. Combust. Inst. 32 (2009) 2059–2066. [8] W.P. Linak, J.I. Yoo, S.J. Wasson, W. Zhu, JostO.L. Wendt, F.E. Huggins, Y. Chen, N. Shah, G.P. Huffman, M.I. Gilmour, Ultrafine ash aerosols from coal combustion: characterization and health effects, Proc. Combust. Inst. 31 (2007) 1929–1937. [9] M.J. McNallan, The formation of inorganic particulates by homogeneous nucleation in gases produced by the combustion of coal, Combust. Flame 42 (1981) 45–60. [10] R.L. Davison, D.F.S. Natusch, J.R. Wallace, C.A. Evans, Trace elements in fly ash. Dependence of concentration on particle size, Environ. Sci. Technol. 8 (13) (1974) 1107–1113. [11] W.S. Seams, JostO.L. Wendt, Partitioning of arsenic, selenium and cadmium during the combustion of Pittsburgh and Illinois #6 coals in a self-sustained combustor, Fuel Process. Technol. 63 (20 0 0) 179–196. [12] L.E. Bool, J.J. Helble, A laboratory study of the partitioning of trace elements during pulverized coal combustion, Energy Fuels 9 (1995) 880–887. [13] A.H. Biermann, J.M. Ondov, Application of surface-deposition models to size-fractionated coal fly ash, Atmos. Environ. 14 (3) (1980) 289–295. [14] E. Raask, Mineral impurities in coal combustion: behavior, problems and remedial measures, Hemisphere Publishing Corporation, New York, U.S., 1985. [15] O. Senneca, B. Apicella, S. Heuer, M. Schiemann, V. Scherer, F. Stanzione, A. Ciajolo, C. Russo, Effects of CO2 on submicronic carbon particulate (soot) formed during coal pyrolysis in a drop tube reactor, Combust. Flame 172 (2016) 302–308. [16] L. Yan, R.P. Gupta, T.F. Wall, A mathematical model of ash formation during pulverized coal combustion, Fuel 81 (2002) 337–344. [17] AuraC.Davila Latorre, Ash formation under pressurized pulverized coal combustion conditions, University of Connecticut, Mansfield, U.S., 2005. [18] B.S. Haynes, M. Neville, R.J. Quann, A.F. Sarofim, Factors governing the surface enrichment of fly ash in volatile trace species, J. Colloid Interface Sci. 87 (1) (1982) 266–278. [19] C.Z. Li, Some recent advances in the understanding of the pyrolysis and gasification behavior of Victorian brown coal, Fuel 86 (2007) 1664–1683. [20] Y. Yuan, S.Q. Li, Q. Yao, Dynamic behavior of sodium release from pulverized coal combustion by phase-selective laser-induced breakdown spectroscopy, Proc. Combust. Inst. 35 (2015) 2339–2346. [21] P.J. van Eyk, P.J. Ashman, G.J. Nathan, Mechanism and kinetics of sodium release from brown coal char particles during combustion, Combust. Flame 158 (2011) 2512–2523. [22] P.J. van Eyk, P.J. Ashman, Z.T. Alwahabi, G.J. Nathan, The release of water-bound and organic sodium from Loy Yang coal during the combustion of single particles in a flat flame, Combust. Flame 158 (2011) 1181–1192. [23] D.M. Quyn, H. Wu, C.Z. Li, Volatilisation and catalytic effects of alkali and alkaline earth metallic species during the pyrolysis and gasification of Victorian brown coal. Part I. Volatilisation of Na and Cl from a set of NaCl-loaded samples, Fuel 81 (2002) 143–149. [24] C.Z. Li, C. Sathe, J.R. Kershaw, Y. Pang, Fates and roles of alkali and alkaline earth metals during the pyrolysis of a Victorian brown coal, Fuel 79 (20 0 0) 427–438. [25] L.J. Wibberley, T.F. Wall, Alkali-ash reactions and deposit formation in pulverized-coal-fired boilers: experimental aspects of sodium silicate formation and the formation of deposits, Fuel 61 (1982) 93–99. [26] L.J. Wibberley, T.F. Wall, Alkali-ash reactions and deposit formation in pulverized-coal-fired boilers: the thermodynamic aspects involving silica, sodium, sulphur and chlorine, Fuel 61 (1982) 87–92. [27] P.M. Walsh, Fouling of convective heat exchangers by lignitic coal ash, Energy. Fuels 6 (1992) 709–715.

Q. Huang et al. / Combustion and Flame 182 (2017) 313–323 [28] M. Neville, A.F. Sarofim, The fate of sodium during pulverized coal combustion, Fuel 64 (1985) 384–390. [29] N.B. Gallagher, L.E. Bool, JostO.L. Wendt, T.W. Peterson, Alkali metal partitioning in ash from pulverized coal combustion, Combust. Sci. Technol. 74 (1990) 211–221. [30] T. Takuwa, I.S.N. Mkilaha, I. Naruse, Mechanisms of fine particulates formation with alkali metal compounds during coal combustion, Fuel 85 (2006) 671–678. [31] G. Li, S.Q. Li, Q. Huang, Q. Yao, Fine particulate formation and ash deposition during pulverized coal combustion of high-sodium lignite in a down-fired furnace, Fuel 143 (2015) 430–437. [32] P. Biswas, C.Y. Wu, M.R. Zachariah, B. McMillin, Characterization of iron oxide-silica nanocomposites in flames: part II. Comparison of discrete-sectional model predictions to experimental data, J. Mater. Res. 12 (1997) 714–723. [33] M.S. Wooldridge, Gas-phase combustion synthesis of particles, Prog. Energy Combust. Sci. 24 (1998) 63–87. [34] S. Rigopoulos, Population balance modelling of polydispersed particles in reactive flows, Prog. Energy Combust. Sci. 36 (2010) 412–443. [35] R.E. Mitchell, An intrinsic kinetics-based, particle-population balance model for char oxidation during pulverized coal combustion, Proc. Combust. Inst. 28 (20 0 0) 2261–2270. [36] K.V. Shah, M.K. Cieplik, C.I. Betrand, W.L. van de Kamp, H.B. Vuthaluru, A kinetic-empirical model for particle size distribution evolution during pulverised fuel combustion, Fuel 89 (2010) 2438–2447. [37] K.A. Christensen, H. Livbjerg, A plug flow model for chemical reactions and aerosol nucleation and growth in an alkali-containing flue gas, Aerosol Sci. Technol. 33 (20 0 0) 470–489.

323

[38] J.R. Jensen, L.B. Nielsen, C.S. Moller, S. Wedel, H. Livbjerg, The nucleation of aerosols in flue gases with a high content of alkali – a laboratory study, Aerosol Sci. Technol. 33 (20 0 0) 490–509. [39] P. Glarborg, P. Marshall, Mechanism and modeling of the formation of gaseous alkali sulfates, Combust. Flame 141 (2005) 22–39. [40] L. Hindiyarti, F. Frandsen, H. Livbjerg, P. Glarborg, P. Marshall, An exploratory study of alkali sulfate aerosol formation during biomass combustion, Fuel 87 (20 08) 1591–160 0. [41] S.K. Friedlander, Smoke, dust, and haze: fundamentals of aerosol dynamics, Oxford University Press, New York, U.S., 20 0 0. [42] S.L. Girshick, C. Chiu, Kinetic nucleation theory: a new expression for the rate of homogeneous nucleation from an ideal supersaturated vapor, J. Chem. Phys. 93 (1990) 1273–1277. [43] N.A. Fuchs, The mechanics of aerosols, Pergamon, New York, U.S., 1964. [44] A. Prakash, A.P. Bapat, M.R. Zachariah, A simple numerical algorithm and software for solution of nucleation, surface growth, and coagulation problems, Aerosol Sci. Technol. 37 (2003) 892–898. [45] N.B. Gallagher, T.W. Peterson, J.O.L. Wendt, Sodium partitioning in a pulverized coal combustion environment, Proc. Combust. Inst. 26 (1996) 3197–3204. [46] Q. Huang, S.Q. Li, G. Li, Y. Zhao, Q. Yao, Reduction of fine particulate matter by blending lignite with semi-char in a down-fired pulverized coal combustor, Fuel 181 (2016) 1162–1169. [47] P.O. Mwabe, J.O.L. Wendt, Mechanisms governing trace sodium capture by kaolinite in a downflow combustor, Proc. Combust. Inst. 26 (1996) 2447–2453.