Alkali transformation during single pellet combustion of soft wood and wheat straw

Fuel Processing Technology 143 (2016) 204–212

Contents lists available at ScienceDirect

Fuel Processing Technology journal homepage: www.elsevier.com/locate/fuproc

Alkali transformation during single pellet combustion of soft wood and wheat straw Jonathan Fagerström ⁎, Erik Steinvall, Dan Boström, Christoffer Boman Thermochemical Energy Conversion Laboratory, Department of Applied Physics and Electronics, Umeå University, SE-901 87 Umeå, Sweden

a r t i c l e

i n f o

Article history: Received 3 May 2015 Received in revised form 21 November 2015 Accepted 23 November 2015 Available online 10 December 2015 Keywords: Biomass Combustion Ash Alkali Release Single pellet

a b s t r a c t Controlling slag and deposit formation during thermochemical fuel conversion requires a fundamental understanding about ash transformation. In this work, a macro-TGA reactor was used to determine the release of ash forming elements during devolatilization and char combustion of single pellets. Soft wood and wheat straw were combusted at two temperatures (700 °C and 1000 °C) and the residual ashes were collected and analyzed for morphology, elemental and phase composition. The results showed that the single pellet combustion exhibit similar release character as in grate boilers. The temporal release was found to be both temperature and fuel dependent. For wood, the release of potassium occurred mostly during char combustion regardless of furnace temperature. Similar results were found for straw at 700 °C, but the temperature increase to 1000 °C implied that the release occurred already during devolatilization. The differences are presumably explained by different fuel phase compositions. The residual ash were composed of three different categories of phases; crystalline compounds, molten ash (glass) and char, and the work concludes that K was captured by crystalline K/Ca-carbonates as well as in amorphous glassy silicates for wood, and by almost fully molten ash of glassy silicates for straw. The fuel conversion processes occurring on a grate influence the fuel combustibility in terms of e.g. burnout, slag formation and release of fine particle and deposit forming matter, and the present work has given novel insights into the specific alkali behavior during biomass fuel conversion. © 2015 Elsevier B.V. All rights reserved.

1. Introduction Ash related operational problems, like slag and deposit formation in combustion appliances as well as fine particle emissions, are major concerns in biomass combustion that decreases the performance and may hinder an extension of the biomass resource base for heat and power production. The phenomenon of fouling is described in general terms by e.g. Bryers [1], and specifically for grate boilers by Yin et al. [2] and Bashir et al. [3]. The mechanisms of slag formation in grate boilers are not covered to the same extent in the literature but Näzelius et al. [4] depict the present understanding. The distribution and transformation of alkali metal compounds within the burner and boiler during biomass combustion determines to a large extent the type of operational problem and degree of severity. Deposit formation on heat transfer surfaces and formation of fine particle (PM1) emissions is mainly influenced by the release of alkali species. Suppressing the release, i.e. capturing alkali metals to the bottom ash as non-volatile compounds, is thus one way to control fouling and PM1 emissions. However, a drawback of capturing alkali metals to the bottom ash during fixed bed combustion is an increased risk of slagging

⁎ Corresponding author. E-mail address: [email protected] (J. Fagerström).

http://dx.doi.org/10.1016/j.fuproc.2015.11.016 0378-3820/© 2015 Elsevier B.V. All rights reserved.

due to the potential formation of sticky ash melts [4–5]. Controlling both slag and deposit formations during biomass combustion requires a fundamental understanding about ash transformation, especially when the fuel resource base is extended towards new ash rich assortments and biomass mixtures for co-combustion situations. To elucidate the underlying phenomenon and mechanisms related to these ash chemical effects, controlled studies in lab-scale reactors are a useful approach. For wood and straw fuels, some experimental work has been performed in specific reactor set-ups to elucidate the effect of temperature on the release of mainly K, S, and Cl by analyzing the elemental composition of the residual ash after full conversion of powder samples [6–8]. The experimental procedure applied in these studies included a step-wise thermal treatment of the fuel/ash with different atmospheres that resulted in long (200 min) conversion times. The capturing mechanisms for certain ash forming elements were discussed through the support of thermodynamic equilibrium calculations. In a recent study, the same experimental approach was used to study the release behavior of K, Cl, S and P during combustion of the energy crops poplar and brassica [9]. Furthermore, the composition of wood ash during thermal treatment has been studied in a similar manner and reactor also by other authors [10]. Another approach to study the alkali release behavior is by measuring alkali species in the evolving gas during thermal treatment, which has been performed for small biomass powder samples [11–12]. However, the composition of the

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

residual ashes in those experiments was not analyzed and considered in relation to the release behavior of certain elements, and total quantifications of the released fractions during different fuel conversion stages have so far not been presented in the literature. Thus, a detailed understanding of the time-resolved quantitative release behavior of critical ash forming elements in relation to the fuel conversion process, are to a large extent still lacking. Such information are vital for the development of generally applicable models of the ash transformation that can be applied on different biomass fuels to predict technical and emission related problems. The objective with the present study was therefore to determine the release of major ash forming elements after both the devolatilization phase and the char combustion using single pellets of soft wood and wheat straw. The phase compositions of the residual ashes were further determined after full conversion to study the capturing mechanism of alkali metals and to explore the overall ash transformation processes on a single fuel pellet level. 2. Material and methods 2.1. Fuels Two biomass fuel pellets were used, softwood without bark consisting of a spruce/pine mixture from northern Sweden, and wheat straw delivered from Denmark. The pellets were 8 mm in diameter and weighed 700 ± 50 mg. The ash content and the concentration of ash forming elements in the two fuels is presented in Table 1 together with reference data for the two fuels. As seen in Table 1, the composition of the two fuels is similar to the previously reported reference data. Thus, both of the included fuels, i.e. softwood and wheat straw, can be considered to be representatives. Besides the fuel bulk analysis (Table 1), additional fuel analysis was performed to determine the deviation of moisture and ash content within the pellet batch. 12 different pellets were placed in different ceramic containers and inserted into a muffle furnace at room temperature. The moisture contents were achieved (7.9 ± 0.2% and 13.5 ± 0.4% for soft wood and wheat straw respectively) by heating the pellets at 105 °C until no weight loss could be detected. The ash contents (0.33 ± 0.02% and 4.63 ± 0.09% for wood and straw respectively) were achieved by increasing the temperature to 550 °C in 1 h and keeping it at 550 °C for an additional1 h or until no unburned material was visible. The low standard deviation for the ash contents indicated that the ash matter was evenly distributed within the pellet batch and consequently that a good consistency could be achieved for the replicates during fuel conversion.

regulators and a type N thermocouple situated 10 mm above the grate and 30 mm below the fuel sample. The internal dimensions of the furnace were 200 × 130 × 130 mm and a cylindrical quenching tower was separated from the furnace zone by a slide hatch to enable the use of different atmospheres in the quenching tower and furnace. A pneumatic cylinder underneath the furnace was used to lower and raise the furnace into position of combustion or quenching. The sample basket, made of platinum to avoid interactions between the fuel ash and sample holder, was hung from an on-line analytical balance with resolution of 1 mg. A window on the front side facilitated visual inspection and video monitoring of the fuel conversion stages. The gas supply consisted of two sources, N2 and air, both controlled by calibrated volume flow meters. A schematic sketch is presented in Fig. 1. 2.3. Combustion experiments 2.3.1. Experimental procedure Three replicates were performed for each experiment. The experiments were performed with an oxygen concentration of 10 vol-%, a total gas supply of 15 l per minute (0 °C and 100 kPa), and at furnace temperatures of 700 °C and 1000 °C. The combustion temperature, i.e. the temperature in the pellet, was measured with a type N thermocouple (Ø 1.5 mm) inserted into a drilled hole in the pellet. These measurements were performed for both fuels and for both furnace temperatures prior to the release experiments since the set-up did not enable simultaneous temperature and gravimetric analysis. The ash release was determined for two fuel conversion stages by performing two separate experimental procedures. The experiments involving the first stage (devolatilization) was quenched before the char combustion started, i.e. when the fuel pellet stopped flaming and turned glowing red by char oxidation. The extinction of the flame coincided with the “break”, i.e. deflection point, of the weight curve (Fig. 3) and

2.2. Macro-TGA reactor The macro-TGA reactor was recently described by Biswas et al. [13] and in previous studies [14–17] a similar reactor has been applied. The furnace was resistively heated by two side panels controlled by PID Table 1 Total ash content (wt-% of ds) and the concentration of ash forming elements (mg/kg of ds) in the soft wood and wheat straw fuels. Reference mean and standard deviation values adapted from [37].

Ash content K Na Ca Mg Al Si P S Cl Zn a

Softwood

a

Wheat straw

a

0.40 564 22 751 140 34 420 44 72 59 12.7

n.a. 561 ± 157 39 ± 46 1118 ± 351 178 ± 39 135 ± 198 662 ± 1307 71 ± 28 n.a. n.a. n.a.

4.48 6970 201 3010 688 217 11100 385 913 2300 6.6

n.a. 8658 ± 6881 359 ± 1080 3910 ± 2287 1099 ± 1072 358 ± 403 18,960 ± 27,066 1283 ± 2600 n.a. n.a. n.a.

Reference (11)

Number in parenthesis denote number of samples.

205

Reference (7)

Fig. 1. Schematic sketch of the macro-TGA reactor.

206

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

Fig. 2. Illustration of combustion experiment procedure.

strengthened therefore the assumption that this was an appropriate time point for separation of the two conversion stages, i.e. devolatilization and char combustion. The quenching was rapid and stopped the fuel conversion in a couple of seconds. Thus, the experiments involving the second stage (char combustion) consisted of both devolatilization and char combustion and is hereby denoted full conversion. The full conversion experiments were completed after the end of char combustion, i.e. when the fuel particle stopped glowing. A similar methodology has previously been applied [18]. As discussed in the literature, the particle size may affect the thermal history of specific fuel particle [12,19], and for such large particles as the used fuel pellets, most probably there exist an overlap of the devolatilization and char combustion phases. This means that the surface of a fuel particle is believed to experience partial char combustion even though the fuel particle as a whole is in the devolatilization stage. The experimental procedure was initiated by weighing a fuel pellet with a precision of 0.1 mg. The pellet was thereafter placed in the sample basket and attached to an analytical balance. Subsequently, the fuel pellet was inserted to the furnace by lowering the furnace whereby ignition occurred within a few seconds. The fuel conversion stages were observed through the inspection window to visually identify the end of the char combustion. After completion of combustion, the furnace was lowered and the sample basket containing the residual ash was removed and weighed. The residual char/ash retained its shape and rigidity well after the combustion and enabled thus a transfer of the sample basket from the furnace to the analytical balance without losses of material. The weight of the sample basket was used in the release quantifications after subtracting the weight of the empty basket. The residual ash sample was thereafter moved to a glass vial for transport to

the laboratory for chemical analyses. A final weighing was performed to calculate the recovery of ash from sample basket to glass vial and thus ensure that most of the residual ash was sent for analysis. All experiments reached N 95% recovery except for wood 1000 °C that showed a recovery of 70–83%, i.e. the ash analysis were considered to represent the residual ash formed in the sample basket. The occasional unaccounted residue left in the sample basket was assumed to have the same ash composition as the sample sent to the laboratory for chemical analysis. The baskets were thereafter washed with de-ionized water and when needed scraped with steel tweezers. The same basket was used throughout all experiments. The results from total organic carbon (TOC) analysis showed that the mass content of organic carbon in the residual ashes from full conversion of wood was b0.01% in all cases except one replicate showing about 3%. The straw experiments showed about 6% TOC at 700 °C and about 1% for 1000 °C. Fig. 2 illustrates the main steps of the experimental procedure, and Fig. 3 presents a combined weight and temperature profile of single pellet combustion in the macro-TGA. To collect a reasonable amount of ash from the wood fuel experiments and thereby enable a more accurate subsequent analysis with XRD, combustion of 10 additional single pellets were performed in the way as described above. The devolatilization experiments were performed identically as the full conversion experiments but the fuel particle was quenched with N 2 in the tower at the end of the devolatilization. The char yields from devolatilization experiments were calculated as weight of char after quenching divided by weight of initial fuel pellet before combustion and the results are presented in Fig. 4. The moisture and ash contents were excluded from the initial fuel pellet weight to enable a comparison with other fuels.

Fig. 3. Weight loss (solid lines) and pellet temperature (dotted lines) measured with thermocouple (drilled into the center of a pellet) during combustion of soft wood (left) and wheat straw (right) single pellets in furnace temperatures of 700 °C (black lines) and 1000 °C (gray lines), respectively.

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

207

inherent phase composition or phases formed early in the fuel conversion. This information may be used as support for the interpretation of the information gained by the macro-TGA experiments to assess the ash transformation and release behavior. The instrument used was a PT7160 RF Plasma barrel etcher by Quorum Technologies. Pulverized fuel pellet samples (5 g wood and 2 g straw) were evenly distributed on a quartz plate and subjected to the plasma for about 24 h in an O2/He atmosphere. An ashing temperature of 120 °C was sustained by applying a microwave power of 150 W and a chamber pressure of 6 Pa. The ashing was interrupted a couple of times during the 24 h for mixing of the sample.

Fig. 4. Char (dry and ash free basis) yield from devolatilization experiments (light gray) and ash yield from full conversion experiments (dark gray) with single pellet combustion of soft wood and wheat straw at 700 °C and 1000 °C in the macro-TGA. Ash yield presented as ratio between collected ash residue weight from the combustion experiments and a calculated ash residue weight obtained from initial pellet weight and ash content from ashing experiments at 550 °C. The values are mean values based on three replicates with standard deviations.

2.3.2. Quantification of release The release values were obtained by determining the captured amounts of ash forming elements via weight measurements of the initial fuel pellet and the residual char/ash, and concentrations of the respective ash forming elements in initial fuel pellet and residual ash. The collection of ash residues was a crucial step in the experimental procedure with regards to both precision and accuracy of the results. An initial evaluation of the method was therefore performed using the ratio between the collected ash residue weight from the macro-TGA combustion experiments and a calculated ash residue weight obtained from initial fuel pellet weight and average ash content from ashing experiments (12 individual pellets) at 550 °C. The ash yields are presented in Fig. 4. The higher furnace temperature resulted in lower ash yields, as shown previously [20].

2.4.4. SEM-EDS A scanning electron microscope (Phillips XL 30 ESEM) was used to study the morphology and size of the residual ash particles. The ash samples were placed directly on a carbon tape attached to aluminum sample holders without further preparation. The acceleration voltage was set to 20 kV and the working distance to 10 mm. The back scatter detector was applied with a magnification of 100 times for all the samples. The SEM was equipped with an energy dispersive X-ray (EDS) detector and it was used on some of the samples for information about the elemental composition. 2.4.5. P-XRD Powder X-ray diffraction analysis was performed on the fuels after low temperature ashing, standard ashing (550 °C) and on the residual ashes after the macro-TGA experiments. The samples were prepared by grinding in an agate mortar and subsequent application on a Si low-background sample holder. The instrument used was a Bruker d8 Advance X-ray diffractometer, Cu Kα-radiation with a line-focus tube (Ɵ–Ɵ mode) and Våntec-1 detector. The crystalline phases were initially identified using the PDF-2 databank [22] together with Bruker software. Refinement of the data was performed using the Rietveld technique (TOPAS, Bruker software) [23] and crystal structure data were obtained from ICSD [24] to enable a semi-quantitative determination of the phase composition. The results from fuel and ash residue analysis are presented in Table 2. 3. Results 3.1. Release of ash forming elements

2.4. Analytical methods 2.4.1. ICP-MS/OES Major and minor ash forming elements in the fuels were determined by wet chemical methods at BE2020+ in Graz, Austria. The procedure is well evaluated and was developed to suit biomass fuels and ashes [21]. The fuel sample was initially prepared according to CEN/TS 14780 and thereafter digested in a multi-step pressurized microwave with HNO3/ HCl/HF, and measured by inductively coupled plasma optical emission spectroscopy (ICP-OES) and/or mass spectroscopy (ICP-MS). Chlorine was determined by ion chromatography after bomb combustion and absorption in NaOH (0.05 M). The ash residues were ground in a quartz mortar prior to the above specified procedure. Water soluble chlorine was eluted with deionized water prior to ion chromatography. 2.4.2. TOC/TIC The analysis of total organic carbon (TOC) was performed to evaluate the amounts of unburned fuel in the residual ash samples. A carbon/ hydrogen analyzer (LECO RC-612) was used to detect CO2 formed from oxidation in a temperature range from 200 °C to 600 °C. Total inorganic carbon (TIC) was analyzed similarly but from 600 °C to 900 °C. 2.4.3. LTA The low temperature ashing (LTA) experiments were performed to obtain ash samples for P-XRD analysis and information about the fuel

The release quantifications of ash forming elements from devolatilization and full conversion experiments with single pellets in the macro-TGA are presented in Figs. 5 and 6 for soft wood and wheat straw respectively. The K release occurred both during devolatilization and char combustion. For wood, the devolatilization release was about 10% regardless of temperature. The K behavior for straw was slightly different since the release during devolatilization increased with temperature, from ~5% at 700 °C to ~15% at 1000 °C. The release character for Na was clearly different compared to that of K. About 90% of Na in straw was released already during devolatilization, regardless of temperature. The Na-release for wood is not presented in Fig. 5 since the analyses showed a negative release of 270%. The reason for this is probably explained by the low concentration of Na in the wood fuel which makes this kind of release quantification rather difficult from an analytical point of view. Thus, no further evaluation of Na is made here. 3.2. Morphology of ash The ash residues were analyzed for morphology by scanning electron microscopy and four representative micrographs are presented in Fig. 7 where differences in particle size and degree of melting can be assessed. Clear differences were observed both between fuel type and temperature for the respective fuels.

208

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

Table 2 Phase composition of soft wood and wheat straw ashes from low temperature plasma ashing (LTA), standard ashing method (STA), macro-TGA at 700 °C (HTA 700 °C), and macro-TGA at 1000 °C (HTA 1000 °C). The samples were analyzed with P-XRD and presented as weight percent of the crystalline fraction. Wood LTA 120 °C SiO2 (quartz) SiO2 (cristobalite) KAlSi3O8 (microcline) NaAlSi3O8 (albeite) CaO (lime) MgO (periclase) Ca(OH)2 (portlandite) CaCO3 (calcite) CaK2(CO3)2 (buetschliite) CaK2(CO3)2 (fairchildite) CaSO4 (anhydrite) K2SO4 (arcanite) KCl (sylvite) KClO3b KClO4b Ca5(PO4)3(OH) (apatite) Ca2SiO4 (larnite) Ca3Mg(SiO4)2 (merwinite) Ca7Mg(SiO4)x (bredigite) Ca2MgSi2O7 (akermanite) CaSiO3 (wollastonite) KAlSiO4 (kalsilite) Fe2O3 (hematite) a b

I

a

Straw STA 550 °C 4 3 2 2 5 47 2

I

a

3

HTA 700 °C 5

1 11 8 3 6 2 1 3

HTA 1000 °C 1

LTA 120 °C

STA 550 °C

22

22

9

6 5 2 13

6 6 7 4 7

HTA 700 °C I

a

HTA 1000 °C Ia Pa

5 3

1 20

8 19 4

Ia

20 26 13 11 5

21 13 13

29 16 16 5

9 8

14 1 3

I stands for “identified” (i.e. clear diffraction pattern) and P for “possible” (i.e. uncertain diffraction pattern). Quantifications were not possible for I and P. Probably an oxidation product of KCl during the plasma ashing method.

3.3. Elemental composition of ashes

3.4. Phase composition of ashes

The mole fraction of ash forming elements was calculated from the wet chemical results for both fuels and temperatures, and the results are presented in Fig. 9. The temperature had no major impact on the elemental ash composition.

Table 2 presents the phase composition in four different ashes; low temperature plasma ashing at 120 °C (LTA), standard temperature ashing method at 550 °C (STA), and macro-TGA ashing/combustion at 700 °C (HTA 700 °C) and 1000 °C (HTA 1000 °C), respectively.

Fig. 5. Release of ash forming elements from single pellet combustion of softwood in the macro-TGA at 700 °C (left) and 1000 °C (right), presented as mean values from three replicates with standard deviation for both devolatilization and full conversion. Na values were about −270%.

Fig. 6. Release of ash forming elements from single pellet combustion of wheat straw in the macro-TGA at 700 °C (left) and 1000 °C (right), presented as mean values from three replicates with standard deviation for both devolatilization and full conversion.

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

209

Fig. 7. Scanning electron micrographs of residual ash from single pellet combustion of soft wood (upper) and wheat straw (lower) at furnace temperatures of 700 °C (left) and 1000 °C (right) in the macro-TGA.

3.4.1. Wood The LTA from wood was mainly amorphous with weak identified peaks of SiO2 and K2SO4. The crystallinity of the STA sample was higher and ten additional phases were quantified. Three phases containing K were found, although in low relative fractions; KAlSi3O8, CaK2(CO3)2, and K2SO4. Furthermore, the combined analysis with XRD and SEMEDS applied for the HTA ashes from the macro-TGA, enables a further evaluation of the characteristics of these ashes. From the SEM micrographs shown in Fig. 7a and b, it is clear that the properties morphologically differ between the two temperatures applied. The ash from the experiments at 1000 °C consisted of larger particles than the ash from the 700 °C experiments, i.e. up to approximately 200 μm (at 1000 °C) instead of much smaller micrometer sized structures up to maximum around 20 μm (at 700 °C). Based on the XRD analysis (Table 2) the presence of several crystalline phases could be determined for both studied temperatures although with some, but rather small, variations in their distributions between the two cases. The major part of the crystalline fraction were composed of Ca/Mg-silicates, Ca- and Mg-oxides as well as Ca/Kcarbonates. Some un-reacted quartz (SiO2) were also seen in the ash from 700 °C, however almost disappeared in the ash at 1000 °C, which was also confirmed by SEM-EDS analyses.

3.4.2. Straw For the straw fuel, the STA ash showed the highest crystallinity, although the LTA from straw contained more crystalline phases than the LTA from wood. Six phases were quantified in the LTA ash for straw and close to 80 wt-% of the total crystalline fraction belonged to K-containing compounds. Half of the crystalline fraction was explained by KClO3 and KClO4. These phases have previously only been identified in plasma ashing [25] and should be considered as oxidation products of KCl. The large amounts of ‘KCl’ identified in LTA were not observed in

the STA. The macro-TGA ashes from wheat straw were clearly highly amorphous since the only detected crystalline phases were different forms of silica (SiO2), i.e. quartz (both at 700 °C and 1000 °C) and cristobalite (at 1000 °C), and K2SO4 (at 700 °C). 4. Discussion In the following, the method of single pellet combustion is initially discussed with respect to accuracy of results and consistency with data from full scale boilers. The discussion is thereafter focused specifically on the alkali behavior, with the support of release data for the full list of ash forming elements and information from SEM micrographs and P-XRD analyses. 4.1. Considerations of single pellet combustion 4.1.1. Method accuracy The determined release was considered to reflect the ash volatilization while other losses, e.g. entrainment of larger fuel particles, were kept to a minimum. This was based on the ash yield determinations (Fig. 4), visual observations of burning pellet, and weight curves (Fig. 3). Complete combustion was assumed during the full conversion experiments based on the low TOC-values. 4.1.2. Full scale comparison The study confirmed that single pellet combustion exhibits a similar distinctive release character between volatile and refractory elements as grate boilers. The group of volatile elements K, Na, S, Cl, and Zn were released to a larger extent than the more refractory group Ca, Mg, Al, Si, and P. The relative release quantifications were overall in line with what have been shown in grate boilers [26–27]. The comparison is, however, somewhat rough since the method used for release quantifications, i.e. fine ash aerosol particle measurements, is associated

210

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

with potential drawbacks, most important the potential losses of released matter inside the boiler, on heat exchangers and/or in the flue gas channels. The literature commonly agrees that the release of K increases with temperature, although this has also been shown to be affected by fuel (ash) composition. In the present study, no significant influence of temperature on total K release was seen. This can possibly be explained by a combined effect of the relatively high temperature during char combustion and the availability of silicon in the fuels, as extrinsic minerals (e.g. quartz) for wood and as amorphous silica for straw. The char combustion temperature (~ 900–1100 °C) is above the melting point for both KCl and K2CO3 already at the lower furnace temperature. The relatively high Si/K ratio in both fuels, although more clear in the straw, enables formation of K-silicates and hence capture of potassium. 4.2. Phase composition in residual ash from wood About 70% of K (Fig. 5) was captured in the ash residue in the full conversion experiments for both 700 °C and 1000 °C. The only crystalline K-containing phases identified by XRD were the two forms of CaK2(CO3)2 (at both temperatures) and minor amounts of K2SO4 (at 700 °C). Thus, the results showed that potassium was captured, at least partly, in the ash as Ca/K-carbonates at both temperatures, corresponding to 7–8 wt-% of the crystalline matter. The presence of carbonates in the ash was however not supported by the TIC measurements which showed very low concentrations (b0.01%) of inorganic carbon. Since the TIC analysis only covers temperatures up to 900 °C, while the carbonates where found by XRD also in the ash from the macroTGA experiments at 1000 °C, the TIC method was obviously not applicable in this case. The fraction of amorphous material in these ashes is rather difficult to assess, although the higher fraction of larger blocks formed at 1000 °C may indicate the presence of amorphous glassy material. This is also strengthened by the SEM-EDS spot analyses in Fig. 8 (right) where the ash material denoted “slag” which showed increased Si and K contents and less Ca as compared to the “ash” matter. Whether amorphous glassy matter was present also at 700 °C, although within smaller morphological structures, is more difficult to determine. However, based on the fact that the elemental distribution (SEM-EDS) and phase composition (XRD) were similar at both temperatures, together with the SEMEDS spot analyses shown in Fig. 8 (left), it can be assumed that a melt rich in Si and K has been formed in the ash also during the experiments at 700 °C. Overall, it therefore can be concluded that the remaining noncarbonate fraction of K in the residual ash most probably was captured in amorphous K/(Ca)-silicates, both at 700 and 1000 °C. The formation of K/Ca-silicates is in line with the general description of ash transformation and slag formation in woody biomass [28]. However, the present results illustrate that this also happens on a single

pellet level during a conversion time of some minutes. The wood in the present study had somewhat elevated absolute concentrations of Si and a previous study [29] showed that quartz (SiO2) participated in melt and subsequent slag formation. The presence of quartz in the macro-TGA ash at 700 °C but not at 1000 °C, together with the formation of larger Si and K rich particles in the ash, indicates that the decreased levels of quartz seen by XRD at the higher temperatures, primarily is due to this phenomenon, i.e. dissolution into a K-rich silicate melt. These temperatures used here in the macro-TGA experiments are also relevant temperatures for real-life conditions in burners and grate boilers. 4.3. Phase composition in residual ash from straw About 80% of K (Fig. 6) was captured in the residual ash together with mainly Si and Ca (Fig. 9) as in the wood experiments. K2SO4 was identified at 700 °C (Table 2), but this was the only crystalline K-containing compound. Considering the low amounts of S in the ash residue, only trace concentrations of K2SO4 are expected to form and the majority of K should be found in other phases. Consequently, based on the low TOC-contents and the SEM micrographs (7c and 7d) together with EDS analysis, K is captured in a glass phase. In addition, the SEM micrographs also reveal non-molten flake-shaped particles with K and Si ratios that should result in melt formation under equilibrium conditions. 4.4. Potassium behavior during devolatilization and char combustion The devolatilization difference between the two fuel types is likely explained by differences in the K-speciation of the fuels, as shown by the XRD data on the LTA fuel ash samples for wood and straw respectively, given in Table 2. Based on these results, it can be assumed that K in wood is probably related to amorphous matter or alternatively too small crystallites that were not seen by the XRD, whereas K in straw is more likely to be composed of K-salts. Devolatilization of wood is believed to liberate K from carboxylates [6] independent of the two temperatures, as recently also shown [30] by in-situ gas measurements. However, during devolatilization of straw in the present study, the increase in release between 700 °C and 1000 °C might be explained by increased surface temperatures of the pellets and therefore an increase in the evaporation rate of KCl, which also has been shown previously [8]. The literature rather commonly agrees that KCl is the major K release species during combustion of both woody [31] and herbaceous [32–33] biomasses. However, based on the results [34] from pyrolysis of pulverized pine and sugarcane bagasse in a wire mesh reactor it was discussed that atomic K was a more dominant primary release species than KCl. It

Fig. 8. Close-up SEM micrograph of wood 700 °C (left) and 1000 °C (right).

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212

211

Fig. 9. Elemental composition of ash forming elements in residual ash from single pellet full conversion experiments of soft wood (left) and wheat straw (right) at 700 °C (bright gray) and 1000 °C (dark gray) in the macro-TGA presented as mole fraction of the displayed elements, determined from wet chemical analyses.

has also been suggested that atomic K is released from K–O–C bonds in the char [35]. 4.5. Potassium and sodium differences The behavior of Na during biomass combustion is often in the literature put on a par with K. However, the release was clearly different in the case of single pellet combustion for wheat straw. As mentioned previously, Na was excluded from wood. Previous studies [6,36] with pulverized samples have also reported a higher release for Na than for K. From a speciation point of view, large differences between K and Na in the fuels are unlikely since leaching experiments [37] with wheat straw have reported that both K and Na are mainly found in the water soluble fraction. Further, since KCl was the major K-species in the LTA sample (Table 2), and the leaching study [37] pointed at similar alkali speciation for K and Na, it is fair to believe that Na is also present as NaCl even though the low Na-concentration made it undetectable to the XRD-technique. Thus, a difference in speciation is not likely to explain the difference in release behavior observed in the experiments of the present study. However, from a thermodynamic point of view, differences between K and Na have been elucidated [38] and two possible explanations are outlined here. Firstly, the reduction potential of the two metals differ in the temperature range of 700–1000 °C, as seen e.g. in an Ellingham diagram [28]. The reduction potential for K is higher than for Na, implying that K to a larger extent would be found in atomic form whereas Na would be found oxidized. This difference might affect subsequent capture within the fuel particle. Secondly, the Gibbs free energy of reaction in the formation of M2SiO3 and M2Si2O5 (M represents K or Na) were lower for K than for Na at both 700 °C and 1000 °C. This indicates that K is more prone to form silicates than Na, and the large amounts of K would “capture” Si and reduce available silicate sites for Na, which could result in an extensive Na release. The two chemical phenomena outlined here behind the different K and Na release behavior are still not conclusive and more experimental work is obviously needed to clarify the mechanisms. 5. Conclusion In this study, a macro-TGA reactor was used with a newly developed and novel approach to determine the release of ash forming elements after the devolatilization phase and the char combustion, respectively, using single pellets of soft wood and wheat straw. The total K release was around 30% and 20% for wood and straw respectively and the furnace temperature did not influence the total release significantly. However, the temporal release behavior was shown to be both temperature and fuel dependent. For wood, the release of K occurred mainly during char combustion. A similar behavior was observed for straw at 700 °C, but at 1000 °C the majority of K was released during the

devolatilization. The differences are presumably explained by different compositions, i.e. form of occurrence of the ash matter in the fuels. The phase composition of the residual ash was composed of three different categories; crystalline compounds, molten ash (glass) and char. The study showed that similar ash transformation reactions as previously outlined for grate boilers, e.g. K/Ca-silicate and K/Ca-carbonate formation, also occur at a single pellet level during short conversion times. Overall, it can be concluded that K was captured by crystalline K/Ca-carbonates as well as in amorphous glassy silicates for wood, and by almost fully molten ash of glassy silicates for straw. The release of Na in the straw experiments was considerably higher than the release of K. Aspects related both to ash speciation and chemical thermodynamics are outlined as potential, although not conclusive, explanations. Furthermore, the results showed that the release character during single pellet combustion, of both refractory and volatile elements, was similar as shown in grate boilers. The total release of the volatile elements K, Na, S, Cl and Zn was also consistent with data from grate boilers. The fuel conversion processes occurring on a grate influence the fuel combustibility in terms of e.g. burnout, slag formation and release of fine particle and deposit forming matter, and the present work has given novel insights into the specific alkali behavior during biomass fuel conversion. Acknowledgments The financial support from ERA-NET Bioenergy, via the Swedish Energy Agency (32352-1), is gratefully acknowledged. Teagasc Agriculture and Food Development Authority are acknowledged for their kind support with pelletizing the fuels. Gustav Gårdbro and Mikael Granholm are acknowledged for performing parts of the single pellet combustion experiments. References [1] R.W. Bryers, Prog. Energy Combust. Sci. 22 (1) (1996) 29–120. [2] C. Yin, L.A. Rosendahl, S.K. Kaer, Prog. Energy Combust. Sci. 34 (2008) 725–754. [3] M.S. Bashir, P.A. Jensen, F. Frandsen, S. Wedel, K. Dam-Johansen, J. Wadenbäck, S.T. Pedersen, Fuel Process. Technol. 97 (2012) 93–106. [4] I.-L. Näzelius, J. Fagerström, C. Boman, D. Boström, M. Öhman, Energy Fuels 29 (2) (2015) 894–908. [5] J. Fagerström, I.-L. Näzelius, C. Gilbe, D. Boström, M. Öhman, C. Boman, Energy Fuels 28 (2014) 3403–3411. [6] S.C. van Lith, V. Alonso-Ramirez, P.A. Jensen, F.J. Frandsen, P. Glarborg, Energy Fuels 20 (2006) 964–978. [7] S.C. van Lith, P.A. Jensen, F.J. Frandsen, P. Glarborg, Energy Fuels 22 (2008) 1598–1609. [8] J.N. Knudsen, P.A. Jensen, K. Dam-Johansen, Energy Fuels 18 (2004) 1385–1399. [9] M. Díaz-Ramírez, F.J. Frandsen, P. Glarborg, F. Sebastián, J. Royo, Fuel 134 (2014) 209–219. [10] M.K. Misra, K.W. Ragland, A.J. Baker, Biomass Bioenergy 4 (2) (1993) 103–116. [11] D.C. Dayton, R.J. French, T.A. Milne, Energy Fuels 9 (1995) 855–865. [12] K.O. Davidsson, B.J. Stojkova, J.B.C. Pettersson, Energy Fuels 16 (2002) 1033–1039.

212 [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26]

J. Fagerström et al. / Fuel Processing Technology 143 (2016) 204–212 A.K. Biswas, M. Rudolfsson, M. Broström, K. Umeki, Appl. Energy 119 (2014) 79–84. C. Erlich, M. Öhman, E. Björnbom, T.H. Fransson, Fuel 84 (2005) 569–575. M. Öhman, C. Boman, H. Hedman, R. Eklund, Energy Fuels 20 (2006) 1298–1304. C. Rhén, M. Öhman, R. Gref, I. Wästerlund, Biomass Bioenergy 31 (2007) 66–72. D. Bergström, S. Israelsson, M. Öhman, S.A. Dahlqvist, R. Gref, C. Boman, I. Wästerlund, Fuel Process. Technol. 89 (2008) 1324–1329. J.M. Jones, L.I. Darvell, T.G. Bridgeman, M. Pourkashanian, A. Williams, Proc. Combust. Inst. 31 (2007) 1955–1963. J.M. Johansen, M. Aho, K. Paakkinen, R. Taipale, H. Egsgaard, J.G. Jakobsen, F. Frandsen, P. Glarborg, Proc. Combust. Inst. 34 (2013) 2363–2372. M.J.F. Llorente, J.E.C. García, Fuel 85 (2006) 1273–1279. G. Baernthaler, M. Zischka, C. Haraldsson, I. Obernberger, Biomass Bioenergy 30 (2006) 983–997. International Center for Diffraction Data (ICDD), The powder diffraction file PDF-2, ICDD, Newtown Square PA, 2004. TOPAS 2.1, Bruker AXS Gmbh, Karlsruhe Germany, 2003. Fachinformationzentrum (FIZ) Karlsruhe, Inorganic Crystal Structure Database (ICSD), FIZ Karlsruhe, Karlsruhe, Germany, 1978. S.V. Vassilev, D. Baxter, C.G. Vassileva, Fuel 112 (2013) 391–449. O. Sippula, Fine particle formation and emissions in biomass combustion (Ph.D. Thesis) University of Eastern Finland, 2010 (http://www.atm.helsinki.fi/faar/ reportseries/rs-108.pdf, accessed Jan 15, 2015).

[27] T. Brunner, Aerosols and coarse fly ashes in fixed-bed biomass combustion — formation characterisation and emissions(Ph.D. Thesis) Technical University of Eindhoven, 2006. [28] D. Boström, N. Skoglund, A. Grimm, C. Boman, M. Öhman, M. Broström, R. Backman, Energy Fuels 26 (2012) 85–93. [29] E. Lindström, M. Öhman, R. Backman, D. Boström, Energy Fuels 22 (2008) 2216–2220. [30] T. Sorvajärvi, N. DeMartini, J. Rossi, J. Toivonen, Appl. Spectrosc. 68 (2) (2014) 179–184. [31] D.C. Dayton, B.M. Jenkins, S.Q. Turn, R.R. Bakker, R.B. Williams, D. Belle-Oudry, L.M. Hill, Energy Fuels 13 (1999) 860–870. [32] J.M. Johansen, J.G. Jakobsen, F.J. Frandsen, P. Glarborg, Energy Fuels 25 (2011) 4961–4971. [33] D.C. Dayton, R.J. French, T.A. Milne, Energy Fuel 9 (1995) 855–865. [34] T. Okuno, N. Sonoyama, J.I. Hayashi, C.Z. Li, C. Sathe, T. Chiba, Energy Fuels 19 (2005) 2164–2171. [35] K. Hashimoto, K. Miura, J.J. Xu, A. Watanabe, H. Masukami, Fuel 65 (1986) 489–494. [36] M.J. Wornat, R.H. Hurt, N.Y.C. Yang, T. Headley, J. Combust. Flame 100 (1995) 131–143. [37] M. Zevenhoven, P. Yrjas, B.J. Skrifvars, M. Hupa, Energy Fuels 26 (2012) 6366–6386. [38] M. Becidan, E. Houshfar, R.A. Khalil, O. Skreiberg, T. Lovas, L. Sorum, Energy Fuels 25 (2011) 3223–3234.