A study on initiation of ash agglomeration in fluidized bed gasification systems

A study on initiation of ash agglomeration in fluidized bed gasification systems

Fuel 152 (2015) 48–57 Contents lists available at ScienceDirect Fuel journal homepage: www.elsevier.com/locate/fuel A study on initiation of ash ag...

2MB Sizes 147 Downloads 396 Views

Fuel 152 (2015) 48–57

Contents lists available at ScienceDirect

Fuel journal homepage: www.elsevier.com/locate/fuel

A study on initiation of ash agglomeration in fluidized bed gasification systems Aditi B. Khadilkar a,b,d, Peter L. Rozelle c, Sarma V. Pisupati a,b,d,⇑ a

John and Willie Leone Family Department of Energy and Mineral Engineering, The Pennsylvania State University, University Park, PA 16802, United States The EMS Energy Institute, The Pennsylvania State University, University Park, PA 16802, United States c United States Department of Energy, Office of Fossil Energy, FE-221/Germantown Building, 1000 Independence Avenue, S.W., Washington, DC 20585-1209, United States d National Energy Technology Laboratory, Morgantown, 3610 Collins Ferry Road, Morgantown, WV 26507-088, United States b

h i g h l i g h t s  Gravity separated fractions analyses reveal local initiation of agglomeration.  FactSage calculations help to predict slag formation in fluidized bed gasification.  Slag formation occurs even at fluid bed temperatures in high rank coal gasification.  Hematite increases alumino-silicate slag formation temperature, until it reduces.  Integrated ash agglomeration model was developed to predict agglomeration kinetics.

a r t i c l e

i n f o

Article history: Received 16 September 2014 Received in revised form 25 November 2014 Accepted 15 January 2015 Available online 29 January 2015 Keywords: Reducing Mineral matter transformation Particle Ash High rank coal FactSage™

a b s t r a c t Agglomeration in fluidized beds begins locally by the sticking of slag–liquid-covered particles. Analysis of the composite fuel is not adequate in predicting agglomeration problems. Separation of the high rank Pittsburgh seam coal into particle classes based on specific gravity (SG1: <1.3 g/cm3; SG2: 1.3–1.6 g/cm3; SG3: 1.6–2.6 g/cm3 and SG4: >2.6 g/cm3) and particle size (PS1 through PS7) helped to identify important particle-level slag–liquid formation tendencies. Slag–liquid formation tendencies under fluidized bed operating temperatures were determined both computationally and experimentally. Particles rich in certain iron and calcium phases melt at very low temperatures that are well within fluidized bed operating conditions. The iron rich particle classes (SG3 and SG4) showed the presence of several phases containing iron in different oxidation states. The presence of these iron phases was not detected in the composite bulk fuel. The possibilities of equilibrium liquid phase formation in the presence of different ratios of these iron and calcium oxides to alumino-silicates were determined. Presence of hematite was found to delay slag–liquid formation. Each of the particle classes showed distinct slag–liquid formation tendencies that indicate initiation of agglomeration around SG3 and SG4 particles. The study revealed the importance of particle class-level differences in mineral matter composition for the prediction of agglomeration during fluidized bed gasification. A novel integrated ash agglomeration model that accounts for particle hydrodynamics as well as particle class level ash chemistry has been outlined to predict agglomeration kinetics. Ó 2015 Elsevier Ltd. All rights reserved.

1. Introduction Fluidized bed gasification is gaining importance due to the advantages of fuel-flexibility and low temperature sulfur capture [1]. Gasification of a wide-variety of fuels such as coals of different ⇑ Corresponding author at: John and Willie Leone Family Department of Energy and Mineral Engineering, The Pennsylvania State University, University Park, PA 16802, United States. Tel.: +1 814 865 0874. E-mail address: [email protected] (S.V. Pisupati). http://dx.doi.org/10.1016/j.fuel.2015.01.039 0016-2361/Ó 2015 Elsevier Ltd. All rights reserved.

ranks and biomass is becoming an important, environmentally friendly means of generating power. Gasification in fluidized beds helps to avoid slag flow issues that can occur in entrained flow systems. However, problems of agglomeration of bed material can restrict the efficient utilization of this technology. Agglomeration problems have been previously experienced in fluidized bed boilers. Similar problems have been witnessed even with fluidized bed gasification systems [2–4], such as the problems with the Pinon Pine gasifier [5]. Several studies have shown higher slag formation tendency in a reducing atmosphere [6–8], thereby


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57

suggesting greater operational difficulties in the case of fluidized bed gasification. These issues call for more in-depth studies on mineral matter transformations under such operating conditions to understand their implications on plant operability. There are several studies of ash slagging tendencies at high temperatures that are applicable to entrained flow gasifiers [6]. However, slag–liquid formation tendencies have not been extensively studied under the lower temperatures at which fluidized bed gasifiers operate. Small amounts of slag–liquid that form under fluidized bed conditions may be adequate to cause operational difficulties such as agglomeration. In order to estimate the rate of agglomeration for a given fuel, the slag–liquid formation tendencies of the ash particles under fluidized bed operating temperatures must be studied. This study aims at understanding some of these slag–liquid formation tendencies at fluidized bed temperatures under a reducing atmosphere. Hitherto, techniques such as Initial Deformation Temperature (IDT) that involve visual detection of change in the shape of an ash cone to detect onset of slag formation and indices such as acid to base ratio have been used to predict the slag–liquid formation and agglomeration tendencies. However, several studies have shown that these methods, which depend entirely on the bulk fuel chemical composition, do not allow accurate predictions [9–11]. In order to understand the physics and chemistry at particle level, particle classes of the composite fuel, based on differences in density and size, were used in this study. This facilitates the understanding of slag–liquid formation tendencies at the particle level. The particle fractions were analyzed both computationally and experimentally. FactSage™ thermodynamic simulation software was used and the results were validated with high temperature X-ray diffraction and thermo-mechanical analyses. 2. Materials and methods 2.1. Division of composite coal into fractions based on size and density Pittsburgh no. 8 coal was separated into fractions using different specific gravity liquids and float–sink gravity separation techniques utilized in an earlier study [12]. Coal was mixed with organic liquids having the specific gravities of 1.3, 1.6 and 2.6. The fraction heavier than the liquid was allowed to sink and thus separated from the lower specific gravity particles that remained afloat. The specific gravities of the organic liquids were chosen based on the desired mineral rich fractions being generated. For example, the need for the separation of a distinct particle class dominated by the heavier iron rich particles led to the choice of 2.6 as the highest specific gravity. At the same time, the attempt was made to keep reasonable ash yields for each of the fractions generated. The whole coal is termed as SG0. The four fractions generated were denoted as SG1, SG2, SG3 and SG4 in the order of increasing specific gravity. The chemical composition of ash from each of these fractions was obtained using XRF (Table 1). Each gravity-separated fraction was then separated into seven size classes using sieves of different mesh sizes. Experimental results in this paper are reported from the finer size class only (75–106 lm – PS6). 2.2. Generation of ash for experimental analyses The coal fractions were heated in a proximate analyzer up to 650 °C. A moisture removal step was done at 110 °C for an hour, which was followed by ashing. The ashing was done at 650 °C until a constant mass of the residue was achieved. In this way, ash, which had not been subjected to relatively high temperatures, was generated from the coal fractions.

Table 1 XRF ash analyses of particle classes (75–106 lm) of Pittsburgh seam coal [2]. Species

Wt. percent in fuel ash (%)



SG1 SG2 SG3 SG4 [1.3 float] [1.3 sink, 1.6 float] [1.6 sink, 2.6 float] [2.6 sink]

SiO2 Al2O3 Fe2O3 CaO TiO2 K2O MgO Na2O SrO BaO MnO

46.4 22.0 24.5 3.41 1.05 1.56 0.73 0.42 0.11 0.08 0.03

49.9 28.7 9.41 5.05 2.47 1.85 1.32 0.76 0.34 0.17 0.02

55.3 25.2 11.3 2.78 1.22 1.98 1.08 0.91 0.10 0.07 0.03

51.8 20.8 18.9 4.87 0.74 1.48 0.76 0.52 0.04 0.03 0.02

12.5 4.67 76.0 5.79 0.21 0.24 0.29 0.13 0.29 0.03 0.07

2.3. Thermodynamic equilibrium phase calculations using FactSage™ [13] Simulations were run using FactSage™ version 5.2 for each of the four gravity-separated fractions (75–106 lm). Ash from each gravity fraction was allowed to equilibrate individually with a reducing gasification medium. The reducing gasification medium consisted of 32.6% carbon monoxide, 27.6% hydrogen, 14.3% carbon dioxide and balance water vapor at atmospheric pressure. The mass ratio of the total amount of reducing gas to the carbon content of the fraction was taken as 0.2 based on an average value reported in literature from commercial gasifiers [14–16]. Since these gasifiers use mainly lignite, sub-bituminous coals and a few bituminous coals, the carbon content was considered as 55 wt.% for calculation of the ratio. FactSage™ computes the Gibbs free energy for various possible compounds at a given temperature and pressure using thermodynamic properties from FACT databases while maintaining the elemental balances. The FACT Slag A database was used for the slag phase thermodynamic properties. All the phases available in this Slag A database as well as phases available in the FACT database for pure solids, liquids and gases were used unless the number of phases exceeded the maximum allowed in a single calculation. If the number of possible phases exceeded this limitation, the hydrocarbon phases and the phases containing water of hydration from the solid and liquid phase were not included in the calculation, since they would not exist at the high temperatures studied. The ash compositional analysis as oxides was used as input. Shannon et al. shows that the results are not significantly affected by the forms of sulfur [12]. Therefore only sulfur, in its elemental form, was used as input in this study. The FactSage™ calculations were used to obtain the stable solid, liquid, gaseous and slag–liquid phases at equilibrium. The temperature at which slag–liquid formation was initiated was also estimated. In order to obtain the initial slag formation temperature, the FactSage™ calculations were first performed at 900 °C and then performed by gradually increasing the temperature in increments of 10 °C until slag–liquid presence was seen. Similarly, if slag–liquid was present at 900 °C, the temperature was lowered in increments of 10 °C until no more slag–liquid was seen. Once the initial slag–liquid formation temperature had been determined, the FactSage calculations were done using step intervals of 50 °C. The weight percentage of slag for each of the gravity separated fractions was calculated using the equilibrium amounts of solid and liquid phases. The percentage of slag in the bed depends on the slag formation tendency of each gravity fraction and also the ash content of that fraction and the contribution of that fraction to the whole coal. The weight percent of total slag in the bed is


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57

then calculated using the total amount of slag formed from each of the gravity fractions, at a given temperature. This technique of estimation of slag content of the bed is referred to as attribute tracking. In this calculation, the bed is assumed to consist of the ash from the whole coal which is made up of ash obtained from each of the gravity fractions.

2.4. Thermo-mechanical analyses Ash pellets of 6 mm diameter and 5–6 mm thickness were made using hydraulic pressing. They were then subjected to a constant stress by a vertical alumina rod while the sample was heated. The sample was heated up to about 1200 °C and the change in the thickness of the pellet was continuously recorded. This change in thickness has been correlated to the melting of ash particles in previous studies [17]. The thickness decreases gradually at first and then more rapidly thereby forming a curve with a knee. The initial sintering temperature was determined as the temperature at which the drop from the horizontal tangent to the curve begins. Hence, this technique is useful to understand the slag–liquid formation tendencies of ash as a function of temperature. The measurements were made in a reducing atmosphere similar to that used for the FactSage™ experiments. The gaseous atmosphere consisted of 30.5% carbon dioxide and balance carbon monoxide. Due to the difficulty in experimental handling of hydrogen and possible damage to the TMA furnace, the gaseous atmosphere was simulated using a mixture of only carbon monoxide and carbon dioxide that had the same oxygen partial pressure. Trends in slag–liquid onset and the relative amount of slag–liquid were not affected by the differences in gaseous atmospheres. Thus, as a simplification and due to the practical limitations, hydrogen was not used during the validation of comparative trends using TMA.

2.5. High temperature X-ray Diffraction An XPERT-PRO diffractometer with a PW3050/60 goniometer with copper as the anode material was used to perform X-ray diffraction measurements. The ash powder was uniformly spread over a platinum strip with the sample length of about 10 mm. Xray spectra of the ash samples were recorded in-situ every 100 °C starting from 700 °C. A scan was also taken before starting the heating cycles in order to get a baseline that reflects the components in the high temperature ash that had been previously generated. A reducing atmosphere consisting of 30.5% carbon dioxide and balance carbon monoxide was used for these experiments.

2.6. Uncertainty analysis The gravity separation of the coal into fractions of different specific gravities utilized the float and sink method, which involves experimental errors in the determination of the location of the layer separating the float and sink fractions. Error is also introduced due to the loss of the sample during filtration. The reproducibility was tested by conducting 5 repeated separations and calculating the yield of the 1.6 float and sink fractions. These repetitions gave an average yield of 4.5 wt.% of sinks with a standard deviation of 0.95. The uncertainty of measurement of the oxide content of the fractions was determined through the analysis of a rock standard over time using the same technique and making observations across several previous measurements. Thus, the relative uncertainty was estimated to be about 2–3 wt.% for measured values of greater than 10 wt.%. If the measured weight percent of oxides was between 1 and 10 wt.%, the relative uncertainty was about 5 wt.%. It increased to about 10 wt.% for smaller measured values between 0.1 and 1 and was greater than 10 wt.% for values smaller than 0.1 wt.%. The uncertainty in the thermo-mechanical analyses was determined based on the comparison of results obtained using the SG4 fraction under an oxidizing atmosphere. The uncertainty was assumed to be the same for all the fractions even under reducing atmospheres. It was determined through the calculation of a standard error on the mean of three repeated sets of measurements. Fig. 2 shows the reproducibility of measurements for SG4 with the calculated error bars. 2.7. Development of Penn State integrated ash agglomeration model Due to the importance of differences in properties at the particle-class level, a model is being developed at Penn State based on two-particle collisions using the Stokes’ criterion. The Stokes’ number is the ratio of kinetic energy to the viscous dissipation forces. The higher the viscous dissipation forces the smaller the Stokes’ number and higher the tendency of particles to remain stuck after collision. If the calculated Stokes’ numbers are less than a computed critical Stokes’ number, the particles are said to stick on collision and otherwise they rebound. This model incorporates effects of both the particle-class chemistry (based on FactSage™ calculations) as well as the particle-class physics. The particle-class physics are obtained in terms of the collision frequency from computational fluid dynamics using the Eulerian formulation with the kinetic theory of granular flow. The amount and viscosity of slag–liquid are obtained using the

Inial Parcle Size distribuon

Chemistry FactSage Thermodynamic Calculaons

Physics Hydrodynamics using MFIX Parcle velocity Collision frequency

Amount and composion of slag Probability of wet Slag viscosity collision

Stokes’ test New Parcle Size distribuon Decrease in available wet surface Fig. 1. Flow chart of Penn State ash agglomeration model.


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57













Wt. percent slag (%)

Change in thickness in µm








60 50




30 20 10


0 840

Temperarture in


FactSage™ calculations. The particles may be fully or partially covered by the slag–liquid based on the amount and surface tension. The probability of a wet collision between particles is calculated based on the fraction of surface covered by slag–liquid. Knowing the number of such wet collisions in a given time based on the collision frequency, changes in the number of particles are tracked. These steps, involved in the development of the model, have been illustrated in the algorithm in Fig. 1. The changes to an initial particle size distribution with time are tracked. The initial particle size distribution is taken in terms of the number of particles within bins based on particle size. All possible two-particle collisions in the system are subjected to the Stokes’ test to check for particle sticking. If the particles stick as per the Stokes’ test, the diameter of the agglomerate is calculated with a consideration of the mass balance and the particles are relocated into bins of appropriate size. Thus, the change in the average diameter of particles in a given distribution of ash particles with time can be tracked, accounting for both the chemistry of mineral matter and particle physics.


The slag–liquid formation tendency was studied under equilibrium conditions. Fig. 2 shows considerable slag–liquid formation around particles of all the four classes. A significantly high amount of slag–liquid is formed from the SG4 fraction which is rich in iron phases. Alumino-silicate rich SG1 and SG2 fractions also show higher slag–liquid formation between 900 and 990 °C. The SG4 fraction shows distinct behavior from the other fractions. Each of the four fractions is also different from the whole coal and shows higher amounts of slag–liquid up to 1000 °C. SG3 and SG4 fractions show more slag–liquid formation than that detected by whole coal analyses over the entire temperature range (Fig. 3). At temperatures of about 900 °C, the SG3 and SG4 fractions show a lower level of slag–liquid formation. The slag–liquid formation level in these fractions goes up after 990 °C. The temperature at which slag–liquid formation begins was determined by running several simulations at steps of 10 °C until the presence of any equilibrium slag–liquid phase was detected. Slag–liquid is formed in SG4 fractions at very low temperatures, although it is low in amount (Table 2). This could be significant from the standpoint of bed agglomeration at lower temperatures where it is not expected. Slag–liquid formation in the SG3 fraction also begins at relatively lower temperatures than the SG0 and SG1 fractions. Small amounts of slag–liquid are formed initially in this fraction with a sudden steep increase after about 990 °C. The SG3 and SG4 fractions have high ash contents although they make up a

Wt. percent slag (%)

Fig. 2. Estimation of reproducibility of TMA measurements.

3.1. FactSage™ thermodynamic analyses



Temperature (oC)


3. Results




(b) SG0 SG1


SG2 6

SG3 4 2 0 840




Temperature in oC Fig. 3. Slag–liquid formation tendencies under reducing atmosphere.

Table 2 Slag–liquid formation onset temperature for each fraction under reducing atmosphere. Ash

Density fraction

Slag–liquid formation onset temperature in °C


Whole <1.3 1.3–1.6 1.6–2.6 >2.6

910 910 890 890 840

smaller amount of the feed (Table 3) [18]. Hence, their contribution to the total slag formed in the bed is significant. The amount of slag in the bed is computed using attribute tracking and the slag is seen even at low temperatures in the attribute-tracked graph, unlike the whole averaged bulk coal (Fig. 4). The amount of slag formed at a given temperature using attribute tracking is higher compared to that formed from the whole coal. The slag amount was then used as input in the Penn State ash agglomeration model to predict the average particle size. As shown in Fig. 5, a larger average particle size was obtained through the attribute tracking analyses. Thus bulk coal analyses underestimated the agglomeration potential. This ash agglomeration model is still being developed further and is currently suitable only for comparative analyses. 3.2. Experimental 3.2.1. Thermo mechanical analyses The fourth gravity fraction begins to sinter at temperatures that are considerably lower than the other fractions and also the bulk coal as seen in Table 4 and Fig. 6. It also softens completely and begins to flow at lower temperatures than the other fractions as seen from the sudden drastic decrease in thickness at about 1000 °C in Fig. 6. This tendency of SG4 fractions to form slag-liquids at lower temperatures as well as the greater amount of

A.B. Khadilkar et al. / Fuel 152 (2015) 48–57

Table 3 Distribution of ash content in coal amongst gravity fractions [18].


Gravity fraction

Wt. percent of fraction in whole coal

Wt. percent of ash in fraction


100 47.8 47.6 3.5 1.1

8.9 2.6 10.71 51.6 65.4

Change in Thickness in µm


0 -200

-600 -800 -1000 200







Temperature in oC

2 Attribute tracked


Weight percent slag (%)

Reducing_SG0 Reducing_SG1 Reducing_SG2 Reducing_SG3 Reducing SG4


Fig. 6. Thermo-mechanical analyses of gravity separated fractions.

Whole coal

30 25

500 850




15 10 5 0






Temperature (oC) Fig. 4. Differences in slag formation tendencies of coal using attribute tracking (Inset: Magnification over initial range).

Change in Thickness in µm



0 -500 -1000 -1500




-2500 700






Temperature in oC

Average diameter in µm


Fig. 7. Comparison of slag–liquid formation tendency of SG4 fraction under different gaseous atmospheres.

981 931 881 831

Attribute tracked at 870 deg C

781 731

Attribute tracked at 950 deg C

681 631

Whole coal at 950 deg C

581 531 0





Time in mins Fig. 5. Prediction of average particle diameter using Penn State ash agglomeration model.

Table 4 Initial sintering temperature of gravity separated fractions obtained using TMA. Sample

Initial sintering temperature in °C


790 710 770 770 680

slag–liquid formed under reducing conditions as seen in Fig. 7 validate the slag–liquid formation tendencies obtained from FactSage™ thermodynamic simulations. 3.2.2. High temperature X-ray diffraction Each of the fractions shows a gradual reduction of hematite (H) (Figs. 8–12). Hematite first gets reduced to magnetite and then to wustite and finally to iron. The formation of these new iron phases is clearly seen in the iron rich SG3 and SG4 fractions, as seen in

Figs. 11 and 12. The extent of reduction is increased in the SG4 fraction than the SG3 fraction. These iron-based transformations are not easily detected through analyses of the whole coal alone as is shown in Fig. 8. Noteworthy differences are also seen in the detection of muscovite in these fractions (Figs. 9–14). All the fractions show the presence of muscovite (Fig. 13). However, the muscovite peak is not seen at higher temperatures of 1000 and 1100 °C in the SG1, SG2 and SG4 fractions. SG3 fraction shows the presence of muscovite till 1100 °C (Fig. 14). Similar experiments under oxidizing conditions done in this study, but reported in a companion paper, had shown that muscovite decomposed at low temperatures for SG1 and SG2 fractions. However, the muscovite in the SG3 and SG4 fractions remained stable up to higher temperatures. In addition to these transformations, SG4 fraction also shows the decomposition of calcium carbonate to calcium oxide (Fig. 12). 4. Discussion 4.1. Slag–liquid formation at fluidized bed operating conditions Alumino-silicates along with other oxides lead to slag–liquid formation at different temperatures. In SG3 and SG4 fractions slag–liquid formation is expedited by the iron phases present. Alumino-silicates form slag–liquid at different temperatures in the presence of iron-phases depending on the oxidation state of iron as discussed in the next section. The eutectic composition of 23.25 wt.% CaO, 14.75 wt.% alumina and 62 wt.% silica forms at 1170 °C [19]. The proportions of CaO, Al2O3 and SiO2 present in the SG4 fraction (Table 1) are similar to the composition at which this eutectic forms. HT-XRD shows presence and decomposition of calcium carbonate in SG4 fraction


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57




1100 C



1000 C Muscovite Leucite Montmorillonite



900 oC

Anorthite CaCO 3 CaSO 4


800 C Ilmenite

Albite o

700 C

Fig. 8. HT-XRD of PS3SG0 fraction.


1100 oC


1000 oC Montmorillonite 900 oC



Anorthite CaSO 4 Mullite

CaSiO 3

CaCO 3 CaSO4 Wuste Ilmenite

800 oC Albite 700 oC

Fig. 9. HT-XRD of PS3SG1 fraction.

and thus further supports the possibility of eutectics and phase transformations involving these phases that lead to slag–liquid formation. CaO and CaSO4 can also form a eutectic mixture around 1390 °C [20]. The CaSO4 phase has been shown in Figs. 8–12. In the oxygen rich zones in the gasifiers, formation of CaSO4 is possible. The presence of CaS is also likely [21]. In addition to the iron phases detected using HT-XRD, FeS from pyrite decomposition may also be present in small quantities in bed ash. Thus several equilibrium liquid phases result due to combinations of these calcium and iron-based phases with the alumino-silicates. Some of these are listed in Table 5. These equilibrium compositions are the likely cause of slag–liquid formation at extremely low temperatures in the SG4 and SG3 fractions. Study of the slag–liquid phase composition obtained from FactSage™ shows that there is a significant increase in the FeO, CaO, SiO2 and Al2O3 content of the slag formed as shown in Fig. 15. Presence of

CaO, FeO and SiO2 in the proportions mentioned in Table 5 is the likely cause of low temperature slag–liquid formation in SG4. In addition, eutectics and other equilibrium phases involving more than 3 of these components also occur and further decrease slag– liquid formation temperatures beyond those shown in Table 5. The presence of this slag–liquid, seen only in SG3 and SG4 fractions, would not be detected if only bulk fuel analysis was undertaken. 4.2. Influence of hematite on alumino-silicate Both SG2 and SG3 fractions have large amount of alumino-silicates as seen from Table 1 and Fig. 13. However, the SG2 fraction forms a larger amount of slag–liquid between 900 and 1000 °C. SG3 has a high iron content in the form of hematite, in addition to the high amounts of alumino-silicates as seen from Table 1 and


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57



1100 C Illite


Wuste o

1000 C Muscovite o

900 C Magnete Cordierite

Anorthite CaSO 4


800 C

CaCO 3 Ilmenite

Albite o

700 C

Fig. 10. HT-XRD of PS3SG2 fraction.



FeSi 2


FeSi 2 1100 oC



Wuste 1000 oC



Anorthite CaSO 4

900 oC Albite

CaCO 3 Illite & Sodium

aluminium oxide

800 oC 700 oC

Fig. 11. HT-XRD of PS3SG3 fraction.

Fig. 11. The presence of hematite increases the stability of aluminosilicates such as muscovite. The HT-XRD results shown in Figs. 11 and 14 indicate that muscovite is present up to higher temperatures in SG3 than the other fractions, supporting this hypothesis. Due to this, the amount of slag formed from the SG3 fraction is lower than SG1 and SG2 fractions up to 990 °C. Muscovite bears partial negative charges [22]. Thus, the strongly positively charged ferric ions interact with these clays and thus increase the melting point. This is the same phenomenon responsible for initial suppression of slag formation in SG3 and SG4 under oxidizing conditions and the increased muscovite stability. Under reducing conditions, this effect of hematite presence prevails until significant transformation to wustite occurs. Thus, upon further heating, as the hematite undergoes reduction to magnetite, wustite and iron, the alumino-silicates begin to form more slag–liquid. This phenomenon as well as the other iron-based eutectics described previously causes a sudden rise in the slag–liquid at higher temperatures.

This interaction is less prominent in SG2 fraction due to the lower hematite to alumino-silicate ratio. This is further illustrated by the absence of the muscovite peak at higher temperatures in SG2 and SG1 fractions in HT-XRD. The hematite reduction in SG3 fraction also appears to be slower than the SG4 fraction. The interaction of the iron species with alumino-silicates in SG3 contributes to this delay. There is less hematite in SG3 as compared to SG4 and thus it is entirely involved in this interaction, slowing down its reduction process. Once the hematite has undergone reduction, this effect of muscovite stability is weakened. This also explains the reason for the stability of muscovite in both SG3 and SG4 fractions under oxidizing conditions (shown in a separate paper submitted for publication) but only in the SG3 fraction under reducing conditions. Although the SG4 fraction also contains hematite, it is in large quantities and begins to get reduced immediately. This reduced form (wustite) then leads to slag–liquid formation


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57


Wuste o

1100 C Iron

Wuste o

1000 C


900 C Quartz Gibbsite

CaSO 4


CaCO 3

800 C CaSO 4 o

700 C

Fig. 12. HT-XRD of PS3SG4 fraction.

SG4 Muscovite SG3



Fig. 13. Prevalence of large quantities of muscovite in SG2 and SG3 fractions at 700 °C.

due to the eutectics and equilibrium liquid phases mentioned previously. 4.3. Implications to fluidized bed agglomeration Neither the high alumino-silicate content nor the high iron oxide content independently determines the slag–liquid formation tendency and the resultant initiation of agglomeration in fluidized bed conditions. The interactions involving alumino-silicates, which are also affected by the oxidation state of the iron, play an important role. Particles with high contents of reduced iron phases begin slag– liquid formation at low temperatures such as SG4. Significant amount of slag–liquid formation occurs around particles of specific compositions such as alumino-silicate rich particles like SG2. Particles rich in reduced iron phases yield several low melting eutectics under reducing atmospheres. On the other hand, iron present as hematite may slow down alumino-silicate slag formation as well as itself take longer to get reduced. Subsequent lower amounts of wustite are not adequate for slag–liquid formation at low tempera-

tures. This balance becomes critical toward an understanding and prediction of agglomeration under fluidized bed conditions. This is unlike entrained flow gasification wherein most of the iron would exist in the reduced form at high operating temperatures, and thus iron oxide content would directly correlate to higher slagging tendencies. The different stages of iron oxide reduction that exist at low fluidized bed operating conditions add complexity to the prediction of slag–liquid formation and the relevant transformations under different gasifying atmospheres. The differences in the slag formation tendencies of each of the gravity fractions imply that certain fractions initiate agglomeration as discussed below. It should be noted that this analyses does not consider the effect of carbon content in the fuel on agglomeration. Presence of carbon may create highly reducing atmospheres locally. This may lower slag formation temperatures and thereby increase agglomeration. However, the presence of carbon will also decrease the wettability of the particles [23]. This is believed to decrease the probability of wet collisions between particles and thereby the potential for agglomeration. Experiments by Hseih


A.B. Khadilkar et al. / Fuel 152 (2015) 48–57

SG4 Muscovite SG3 SG2 SG1

Fig. 14. Presence of muscovite in SG3 at 1100 °C.

Table 5 Equilibrium liquid phases possible around particles with calcium and iron phases [25–32]. Components forming liquid

Liquid formation temperature in °C

Reference number

CaSO4–CaS FeS–FeO Fe–FeS Fe2O3–SiO2–Al2O3 FeO–Al2O3–SiO2 (47–12.5–40.3) CaO–FeO–SiO2 (wt.% in melt = 17–46–37) CaO–FeO–SiO2 (wt.% in melt = 11.5–43–45) FeO–Fe2O3–CaO–SiO2 (wt.% in melt = 41.76–1.15–9.77–47.32) CaO–FeO FeO–Fe2O3–SiO2 CaO–SiO2–Al2O3

850 924 988 1073 1073 1093 1105 1105 1133 1147 1170

[26] [27] [28] [26] [29] [26,30] [30] [30] [26] [31] [32]

4.00 FeO

Weight in g

3.50 3.00






1.50 1.00 0.50 0.00 850



Temperature in oC Fig. 15. Slag–liquid composition of SG4 fraction.

and Roberts [24] indicated that agglomeration did not occur on heating coal samples until at least 80 wt.% of the carbon was removed. Thus, as a simplification, this study was restricted to the agglomeration of ash without carbon in the bed. SG2 particles are rich in alumino-silicates and form slag–liquid at higher temperatures than the SG4 fractions that are iron rich. The SG3 fraction has both a high alumino-silicate content as well as a high content of iron phases. In SG3 the slag–liquid formation is only limited by the co-presence of hematite and alumino-silicates. Until the hematite is reduced, the high slag formation tendency of this fraction is limited to some extent. The SG3 and SG4 fractions of coal are mineral-rich although they make up a smaller portion of the feed than SG1 and SG2. Since

the SG4 fraction forms slag–liquid at low temperatures, the percentage of bed ash that forms slag is higher than that predicted by bulk analysis alone. Thus, the average particle size at a given time predicted through attribute tracking of individual fractions was bigger than the average particle size that resulted from bulk coal analyses alone. Bulk analysis would underestimate the potential of the coal to agglomerate as the initiation of agglomeration occurs around SG4 particles at low operating temperatures. Once initiated, increase in bed particle size will promote defluidization and disturb the heat balance. The subsequent rise in bed particle temperatures, granular temperatures and collision frequencies would increase the possibilities of catastrophic agglomeration. This propagation of agglomerate growth is being further investigated. An ash agglomeration model is being developed to facilitate this study and it has been briefly described in this paper. The effects of both ash chemistry as well as particle hydrodynamics are being systematically integrated into this model. This will help to completely understand the mechanism of agglomeration including the initiation and propagation due to presence of particle classes with varying properties.

5. Conclusions The effect of heterogeneity in ash composition on agglomeration tendencies in fluidized beds was studied using gravity separated coal ash fractions. The FactSage™ and TMA results revealed the presence of ash particles that have distinct slag–liquid

A.B. Khadilkar et al. / Fuel 152 (2015) 48–57

formation tendencies. The SG4 particles begin to form slag–liquid at lower temperatures than other fractions. Once agglomeration begins by sticking of these particles, we believe that the subsequent rise in temperature due to heat imbalance will cause the other fractions to begin agglomerating. The method of attribute tracking used in this study also revealed that the amount of slag present in the bed is more than that predicted by whole coal analyses alone. Thus, the study showed that the agglomeration is initiated locally by a few wet particles sticking together. FactSage™ thermodynamic equilibrium modeling along with attribute tracking showed the presence of significant slag–liquid amounts at fluidized bed operating conditions. The XRF compositional analyses of the fractions, the differences in slag formation tendencies and slag phase composition calculated using FactSage™ and the phase transformations observed using HT-XRD along with the theoretical understanding of oxide phase thermodynamics and eutectic formation from the literature [19,25–32], showed that low temperature slag formation occurred due to the formation of iron and calcium-based eutectics. These tools are useful in the prediction of agglomeration and development of models to determine agglomeration kinetics. Under reducing conditions, iron rich particle classes such as SG3 and SG4 form slag–liquid at temperatures that are lower than typical operating temperatures of fluidized bed gasifiers. Higher alumino-silicate content in SG2 fraction and compounds with iron in a lower oxidation state such as in the SG4 fraction also cause higher slag–liquid amounts as seen from the HT-XRD, XRF and FactSage™ results. Differences in the ratio of ferric oxides to alumino-silicates were clearly distinguishable for the gravity separated coal ash fractions from the XRF analyses with a higher ratio for SG4 than SG3, which was in turn higher than SG2. The study of the slagging tendencies of these different fractions using FactSage™ thermodynamic simulations as well as experiments such as HT-XRD showed that the alumino-silicate to iron oxide ratio and the gaseous atmosphere prevalent are important factors for the prediction of agglomeration under fluidized bed conditions. Partial reduction and the presence of several iron phases with different oxidation states of iron are observed using HT-XRD in the SG3 and SG4 fractions under low temperature fluidized bed gasification conditions. Thus, it is not easily possible to relate the slagging rate directly to the iron content, unlike in high temperature entrained flow gasification. Based on the grind of the fuel and the composition distribution, the number of slag–liquid-covered sticky particles will differ for a given particle size based on the amount of slag–liquid. This leads to particle-level differences in ash agglomeration tendencies. Such differences remain undetected by bulk fuel analyses techniques as seen from the HT-XRD and attribute tracking results. Mathematical models that consider such particle level differences, such as the Penn State ash agglomeration model, will help to improve the accuracy in prediction as well as the fundamental understanding of the agglomeration phenomenon in fluidized beds. These in turn could improve the operational reliabilities of fluidized bed boiler and gasification systems. Acknowledgment Financial support for this work was partially provided by National Energy Technology Laboratory (USDOE) under the RES contract 0004000.


References [1] Basu Prabir. Combustion and gasification in fluidized beds. CRC Press; 2006. [2] Cargill P, DeJong G, Howsley T, Larson B, Leighton L, Woodward M. Pinon pine IGCC project, final technical report, DE-FC21-92MC29309; 2001. [3] Southern Company Services Inc., Power systems development facility topical report-gasification test campaign TC21, DE-FC21-90MC25140; 2008. [4] Southern Company Services Inc., Power systems development facility topical report- gasification test campaign TC24, DE-FC21-90MC2514; 2008. [5] Pete C, Gail D, Time H, Brian L, Leo L, Mike W. Pinon pine IGCC project report to DOE; 2001. [6] Huffman GP, Huggins FE, Dunmyre GR. Investigation of the high-temperature behavior of coal ash in reducing and oxidizing atmospheres. Fuel 1981;60:585–97. [7] Mclennan AR, Bryant GW, Stanmore BR, Wall TF. Ash formation mechanisms during pf combustion in reducing conditions. Energy Fuels 2000;14:150–9. [8] Chen J, Greenberg S, Poeppel RB. The viscosity of coal slags as a function of composition, temperature and oxygen partial pressure, report by Argonne National Laboratory, U.S. DOE W-31-109-Eng-38; 1984. [9] Gorbaty ML, Larsen JW, Wender I. Coal science, vol. 2. London: Academic Press; 1983. [10] van der Drift A, Olsen A. Conversion of biomass, prediction and solution methods for ash agglomeration and related problems, ECN fuels conversion and environment report; 1999. [11] Khadilkar A, Rozelle PL, Pisupati SV. Models of agglomerate growth in fluidized bed reactors: critical review, status and applications. Powder Technol 2014;264:216–28. [12] Shannon GN, Matsuura H, Rozelle P, Fruehan RJ, Pisupati S, Sridhar S. Effect of size and density on the thermodynamic predictions of coal particle phase formation during coal gasification. Fuel Process Technol 2009;90:1114–21. [13] Fact Sage Thermodynamic Software. . [14] Ariyapadi Siva, Shires Philip, Bhargava Manish, Ebbern David. KBR’s transport gasifier (TRIG™) – an advanced gasification technology for SNG production from Low rank coals. In: 25th Annual Pittsburgh coal conference; 2008. [15] de Graaf Jan. Overview of shell global solutions’ worldwide gasification developments. San Francisco: Gasification Technologies; 2002. [16] Jones Michael L. Gasification of lignite coal in North America; past experience and future opportunities. Lignite Energy Council 2012. [17] Gupta SK, Wall TF, Creelman RA, Gupta RP. Ash fusion temperatures and the transformations of coal ash particles to slag. Fuel Process Technol 1998;56(1– 2):33–43. [18] Soundarrajan N, Krishnamurthy N, Pisupati SV. Characterization of size and density separated fractions of a bituminous coal as a feedstock for entrained slagging gasification. Int J Clean Coal Energy 2013;2:58–67. [19] Finkelnburg, Dave. How glazes melt in search of the elusive eutectic. Ceramic arts daily; 2009. [accessed March 2014]. [20] Popescu M, Simion A, Mate V. Study of thermal behaviour up to 1550° of materials containing calcium sulfate. J Therm Anal 1985;30:297–303. [21] Wang Y, Sridhar S, Valdez M. Formation of CaS on Al2O3–CaO inclusions during solidification of steels. Metall Mater Trans B 2002;33(4):625–32. [22] Schlegel Michel L, Nagy Kathryn L, Fenter Paul, Cheng Likwan, Sturchio Neil C, Jacobsen Steven D. Cation sorption on the muscovite (0 0 1) surface in chloride solutions using high-resolution X-ray reflectivity. Geochim Cosmochim Acta 2006;70:3549–65. [23] Gibson LaTosha. Measurement and modeling the co-efficient of restitution of char under simulated entrained flow gasifier conditions. Dissertation in energy and mineral engineering. The Pennsylvania State University; 2013. [24] Hsieh CR, Roberts PT. A laboratory study of agglomeration in coal gasification. Chevron Research Company, American Chemical Society Fuels division, preprints, Fall Chicago Conference 1985;30:(3):468-479. [25] Huffman GP, Huggins FE, Shah N, Shah A. Behaviour of basic elements during coal combustion. Prog Energy Combust Sci 1990;16:243–51. [26] Spliethoff H. Power generation from solid fuels. Springer; 2010. [27] Marinov V et al. Ash agglomeration during fluidized bed gasification of high sulphur content lignites. Fuel Process Technol 1992;31(3):181–91. [28] Ryzhenko B, Kennedy GC. The effect of pressure on the eutectic in the system Fe–FeS. Am J Sci 1973;273:803–10. [29] Schairer JF, Kenzo Yagp. The system FeO–Al2O3–SiO2. Am J Sci 1952:471–512. [30] Bowen NL, Schairer JF, Posnjak E. The system, CaO–FeO–SiO2. Am J Sci 1933;5(25):192–283. [31] Turkdogan ET. Fundamentals of steelmaking. The Institute of Materials; 1996. [32] Mao Huahai, Hillert Mats, Selleby Malin, Sundman B. Thermodynamic assessment of the CaO–Al2O3–SiO2 system. J Am Ceram Soc 2006;89:298–308.