Oxygen carrier aided combustion (OCAC) of two waste fuels Experimental and theoretical study of the interaction between ilmenite and zinc, copper and lead

Zinc, copper and lead are amongst the more abundant trace metals in waste fuels such as municipal solid waste and recovered waste wood. The ashes from waste fuels could contain high contents of these metals, which could be valuable but also toxic in certain environments. Oxygen carrier aided combustion, OCAC, is a novel tech- nology for combustion of biomass and waste. Utilizing oxygen carriers could affect the fate of these metals and have implications for stability and recycling. The aim of this work is to study the fate of zinc, copper and lead during oxygen carrier aided combustion of two waste fuels utilizing ilmenite as an oxygen carrier. In total, four samples have been obtained from two different industrial fluidized bed boilers using ilmenite as bed material. Due to low concentrations, bulk analysis methods are not suitable for speciation, i.e. SEM/EDX and XRD. Hence, this investigation utilizes high resolution x-ray photoelectron spectroscopy (XPS), coupled to detailed thermodynamic modelling, with the aim of understanding trace metal speciation, distribution and phase composition. Characterization of the four samples show that iron at the surface of ilmenite particles interact with both copper and zinc to form ferrites, CuFe 2 O 4 and ZnFe 2 O 4 . Lead, on the other hand, is more prone to end up in the fly ash as condensed PbCl 2 , but the mixed oxide PbTiO 3 could be identified at the oxygen carrier surface. Thermodynamic calculations were shown to be in line with the identified compounds.


Introduction
In order to limit the global average temperature increase to below 2 • C above pre-industrial levels as stated in the Paris agreement, greenhouse gas emissions need to be reduced to zero in less than twenty years [1]. This is a difficult task unless active removal of carbon dioxide (CO 2 ) from the atmosphere is enforced. The Nordic countries are world leading with respect to heat and power generation from sustainable biofuels. The location and available resources is well suited for development of Bio Energy with Carbon Capture and Storage (BECCS) [2]. To mitigate the greenhouse gas effect, one needs to consider energy sources beyond fossil fuels. Amongst these, biomass is one potentially sustainable option. It is therefore likely that the transition to renewable energy will result in an expansion of biomass resources [3,4]. On one hand the diverse nature enables biofuels to be utilized all over the world. On the other hand, this expansion leads to larger variation in both ash content and composition [5].
One technology which is capable of capturing carbon dioxide from combustion facilities is Chemical Looping Combustion (CLC) [6]. The technology is based on the use of metal oxides and their ability to be oxidized or reduced depending on the oxygen partial pressure of the surrounding gas. This way, combustion is carried out in two separate reactors without gas mixing between the two. The metal oxides are oxidized in the air reactor and transported to the fuel reactor where the fuel is oxidized using the oxygen supplied by the metal oxides. These metal oxides are referred to as oxygen carriers [7].
Another use for oxygen carriers is bed material for conventional combustion in circulating fluidized bed (CFB) boilers. This technology is referred to as Oxygen Carrier Aided Combustion (OCAC), see Fig. 1. OCAC is realized by replacing the commonly used bed material, such as silica sand, with oxygen carriers. This setup has several advantages. The active bed material will be reduced in fuel rich parts and oxidized in oxygen rich parts leading to a more even oxygen availability and temperature in the boiler. Additionally, it can increase both combustion efficiency and capacity [8]. It has also been shown that this replacement reduces risk for agglomeration and has positive effects on emissions levels [9]. Since CFB boilers are used in industrial applications, OCAC can be a bridging technology towards CLC, and provide valuable insights on the performance of oxygen carriers and fuel. The OCAC-concept has been successfully operated in industrial application where ilmenite has been used as a bed material [10,11].
The main advantage with OCAC is that the technology can take advantage of existing CFB boilers. In Sweden there are 35 waste combustion facilities receiving around 6 million tons of waste yearly with high fractions of bio-based material such as wood and paper [12]. One of the dominating techniques used for energy recovery is fluidized beds which implies a great potential for OCAC globally. OCAC is especially suitable for biomass combustion as it has an ability to handle a wide range of biomass fuels. In recent years OCAC has been demonstrated at industrial scale, still the research around the technology is relatively limited [8,11] Biomass often contains significant amounts of moisture, volatiles and ash [5]. When burning highly volatile fuels the lateral mixing may be insufficient when using conventional bed material such as silica sand, resulting in a requirement of excess air. Although silica sand distributes the heat in the boiler the oxygen carriers also have the possibility to even out the oxygen distribution. This results in decreased amount of excess air required and increases the boiler efficiency. There are some studies conducted on the ash layer build-up and sulfur interaction when using ilmenite in OCAC-applications [13][14][15], but in general there is limited research conducted around this important technology, especially at industrial scale. Investigating samples obtained from industrial applications which utilize complex waste fuels can provide unique and highly relevant information on the solid-state chemistry and fate of important elements.

Background
The flexibility, efficiency and stability of CFB boilers is advantageous for biomass combustion. The ability to handle a wide range of fuels with different shapes and sizes allows for the use of heterogenous waste fuels such as municipal solid waste (MSW). However, these fuels also contain impurities. After a fuel has been combusted the remaining ash fraction will contain inorganic compounds. Both the amount and composition of ash varies widely depending on the source of the fuel. Biomass, for example, can contain an ash fraction between 0.1 and 40% [16]. The most common major (>1%) and minor (0.1-1%) elements in biomass ashes are Si, Ca, K, P, Al, Mg, Fe, S, Na and Mn [17]. Trace elements (<0.1%) commonly includes heavy metals such as Cd, Cu, Hg, Ni, Pb and Zn [17]. The chemistry of these elements in presence of oxygen carriers is unknown [18]. A complex chemistry can be expected with chemical transformation of metals in the gas-phase as well as interaction between bulk gas phase and ash components [19,20].
Problems which could arise with complex fuel compositions with respect to CLC and OCAC could be associated with deposition and corrosion problems, emissions of heavy metals, agglomeration of bed material and possibly other operation problems. Compared to forest fuels, waste fuels may have a varying composition with higher amounts of heavy metals and chlorides. Waste wood for example may contain elevated amounts of zinc, lead or copper if surface treated or impregnated but also higher amounts of chlorine compared to forest fuels [21]. Municipal solid waste is a complex heterogenous fuel. A typical analysis of MSW consist of food waste, PVC, cardboard, newspapers, wood waste, yard waste, shoes, leather, batteries, shredder waste etc. [22] Even though the total amount of trace elements may be relatively low they may cause unwanted effects in the boiler. For example, an enrichment of lead or zinc in the fly ash may cause it to become sticky and in worst case corrosive [23]. It may also lead to unwanted heavy metal emissions [24].
Zinc and lead are two of the more abundant metals in both recovered waste wood (RWW) and municipal solid waste (MSW). Copper is also one of the metals found in higher concentrations in MSW [25]. The fate of these elements under CLC and OCAC conditions is not known. If they end up in the ashes, and these are landfilled it will result in a loss of both monetary value and natural resources. Further, there may be a potential benefit if valuable metals are recovered or recycled which for example, could produce ashes with lower concentrations of heavy metals. Recovery and recycling may even become necessary as strict legislation of recycling targets, landfilling directives and waste handling are implemented by government agencies, such as the European Union. Thus, the chemistry between zinc, copper and lead and oxygen carriers need to be further understood.

Fate of trace elements
The fate of trace elements (TEs) have been commonly studied using thermodynamical equilibrium approaches [26][27][28][29]. However, some experimental studies have also been performed. For example, the release of TEs during waste combustion of wood, shoes, shredder waste and PVC has been studied by Pedersen et al. [30] where lab-scale release experiments were performed on dedicated waste fractions to predict trends and identify release mechanisms attributed to the specific waste fractions.
Dong et al. [31] studied the partitioning of heavy metals in MSW pyrolysis, gasification and incineration using lab-scale tubular furnace, fluidized bed reactors and thermodynamic calculations. Silica sand was used as bed material for the experiments. The study showed that reducing atmosphere has a positive effect on the evaporation of Cd, Zn and Pb while Cu, Ni and Cr were retained in bottom ash. Pb however becomes less volatile under fluidized bed conditions. This could be caused by incorporation reactions of metals with SiO 2 which is enhanced with increasing oxygen concentration as in equation (1) [31].
Under fluidized bed conditions the mixing is more intense and contact between feedstock and bed material is enhanced. The interaction with bed material forming compounds such as Pb 3 Ca 2 Si 3 O 11 or PbZnSiO 4 could prevent Pb and Zn accumulating in the gas phase [31]. The metal-matrix reaction (1) is important as it affects the transfer of metals and is correlated to the presence of mineral substances, alkalis and the interactions among different existing metals. The study of Dong et al. summarized three main parameters affecting the partitioning of heavy metals under MSW thermal processes in the following points [31].
i. The metal speciation; which is related to the redox atmosphere, temperature and presence of alkali, Cl, S and mineral substances. ii. The system characteristics, such as furnace type. iii. The mechanical entrainment of particles caused by gas velocity.
The bed material plays an important role in heavy metal partitioning. Silica sand is an inert material with respect to oxygen carrying capability. If an oxygen carrier is to be used additional reactions with TEs may be possible. Further, the oxygen partial pressure in CLC/OCAC may vary in a wide range, which could affect the chemistry.
TEs in fluidized bed combustion systems utilizing coal have been studied by Alvin et al. [32] using thermodynamic calculations as early as in 1977 with focus on lead, mercury, beryllium and fluorine. In 1994 Frandsen et al. [29] extensively investigated the equilibrium distribution of 18 TEs under both coal combustion and gasification conditions. While early research has focused on coal combustion more recent ones have investigated more complex fuels such as biomass and waste fuels. Studies have shown that combustion of waste derived fuels in fluidized bed boilers may result in increased amounts of TEs in the fly ashes besides the common alkali and alkaline earth metal compounds [33]. One important parameter is the availability of chlorine which couples the volatilization behavior of TEs with the chemistry of alkali metals and calcium which govern the availability of chlorine [33]. Thus, the condensation behavior and partitioning of TEs are important in these systems.
Enestam et al. [27] evaluated the condensation behavior of lead and zinc in BFB combustion of recovered waste wood. The study found that zinc and lead mainly vaporize as Zn(g) and Pb(g) in the reducing part of the furnace. Thus, condensation of metallic Pb, Zn and PbS can be expected in the lower reducing parts of the furnace if the temperature on the surface drops below the condensation temperature, while condensation of PbCl 2 and ZnCl 2 may be expected on the heat transfer surfaces later in the convection section.
A thermodynamic study performed by Becidan et al. [26] studied all the separate waste fractions of MSW and their influence on the behavior of alkali metals and TEs during combustion. MSW fractions having significant Na concentrations are constituents of the biogenic fraction of waste. When chlorine is available most of the Na forms NaCl. When Zn is added to the mixture it will mostly form ZnO(s) (during oxidizing conditions) since most of the chlorine is bounded as NaCl suppressing the formation of ZnCl 2 . If most of the fuel consists of PVC (very small amount of Na) then ZnCl 2 will be the most thermodynamically stable compound. The same behavior is observed for lead which forms PbCl 2 . When the sulfur and chlorine fraction is minor, Pb forms PbO(g). One common feature between the different MSW waste fractions is that lead is overwhelmingly found to be in the gas phase during combustion [26].

Interaction between biomass ash and oxygen carriers
A limited number of studies are available regarding the interaction between biomass ash and iron-based oxygen carriers [14,[34][35][36][37]. Previous research on interactions between oxygen carriers and ash components has mainly focused on coal ashes and most of the studies are conducted in bench scale experiments [38][39][40][41]. For example, brown coal ash, rich in silica, iron and magnesium has been reported to react with both iron ore and ilmenite [42]. Gu, H et al. have investigated the interaction between iron ore and biomass ashes during CLC operation. Silica rich ashes were shown to produce molten potassium silicates which caused particle sintering and reactivity deterioration of the oxygen carrier. On the other hand, potassium rich ashes with relatively low amount of silica improved the reactivity during operation [35].
There are some studies available based on OCAC which have investigated the interaction between ilmenite and biomass ash [14,15]. The main interaction path has been reported to be with the major ash components calcium and potassium. Calcium forms a double ash layer on the ilmenite particle with increased iron content in between [15]. With increased time of exposure potassium diffuses into the ilmenite particle core forming titanates. Other observations relate to the segregation of iron to the surface and enrichment of titanium in the particle core [14,15,43]. Clearly, the chemistry of the main reactive ash components could be tightly connected to the fate of TEs, something which is investigated in this work.

Aim of study
The aim of this work is to study four industrial solid samples derived from OCAC of different waste fuels using the benchmark oxygen carrier ilmenite. Four ash samples will be characterized using experimental techniques, which will be compared to thermodynamic calculations.
More specifically, the objective is to study the interaction between the important trace elements copper, lead and zinc and the bed material, ilmenite. Further investigation of trace element behavior is of interest especially since no previous studies have focused on oxygen carrier interaction with trace elements. Due to low concentrations, bulk analysis methods are not suitable for speciation, i.e. SEM/EDX and XRD. Hence, this investigation utilizes high resolution x-ray photoelectron spectroscopy (XPS), coupled to detailed thermodynamic modelling, with the aim of improving understanding.

Method
There are several combined heat and power plants in Sweden that currently utilizes or have tested OCAC. This study will investigate the TE interaction with industrial bed samples extracted from plants located at Ö rtofta and Händelö, Sweden. Both plants utilize waste fuels and use ilmenite as bed material although in these specific cases the ilmenite was mixed with sand.
The plant at Ö rtofta consist of a 115 MW th CFB boiler which supplies electricity and heat to the local district heating grid by burning biomass. The boiler is owned by the company Kraftringen. During operation around 60 ton of bed and biomass waste material is present in the system. Usually silica sand with an average particle size of 0.25 mm is used as a bed material. In this case however, Norwegian rock ilmenite was used. The biomass fuel consisted of a mixture of recovered waste wood and wood chips. The boiler operates at temperatures around 850 • C. Further information can be found elsewhere [8]. Two samples are analyzed from this plant. One contains higher fractions of ilmenite and the other one was sampled when silica sand was fed back to the boiler.
The plant at Händelö consists of a 75 MW th CFB boiler which provides electricity and heat to the district heating network. The boiler is owned by the company E.ON. Since November 2014 the boiler has been operating regularly with ilmenite. The boiler normally operates with a mixture of MSW and industrial waste but allows sewage sludge and rubber to be used as well. The boiler temperature is kept above 850 • C. Further information can be found elsewhere [11,44]. From this plant one bottom ash and one fly ash sample are obtained for further analysis. During this time, the fuel consisted of a mixture of municipal solid waste and industrial waste.

Sample selection
Four industrial samples have been selected for investigation, presented in Table 1. All bottom ash samples have been sieved below <710 μm to remove larger pieces and magnetically separated two times to obtain an ilmenite rich fraction [43]. The magnetic separation was performed using a conveyor belt stretched across two horizontal cylinders. One cylinder constitutes of an electric motor while the other cylinder contains Rare Earth Roll (RER) magnets. The bed material was evenly fed to the belt using a distributor. The non-magnetic material follows a parabolic path in the direction of the magnetic drum where it is collected in a bin, while the magnetic material follows the magnetic drum making it possible to collect it in a separate bin. Using this method, it is possible to divide the bed material in two fractions.
Two samples were collected from the plant at Händelö where a combination of silica sand and ilmenite was used as bed material. The boiler at the Händelö plant was initially operated with silica sand whereas a transition to ilmenite-operation then took place. The bed material was gradually replaced with ilmenite after which the samples were extracted. Two samples were obtained 100 h after the start of the ilmenite feeding. One bottom ash and one fly ash sample, referred to as MSW-FA, were extracted. The bottom ash sample was magnetically separated, and the magnetic fraction was 26%. This sample is referred to as MSW-ilmenite.
The two Ö rtofta samples were extracted at different occasions. The boiler was operated using ilmenite as bed material for the first 380 h. Thereafter a transition to silica sand operation was initiated and sand was fed back to the boiler. Two bottom ash samples were extracted after 200 h and 580 h from start. Both samples were magnetically separated. The magnetic fraction was 70% for the sample extracted after 200 h of operation and 25% for the sample extracted after 580 h. The sample extracted after 200 h is henceforth referred to as RWW-Ilmenite (3). The sample obtained after 580 h consists of mainly silica sand and was extracted 200 h after the first silica sand input, this sample is referred to as RWW-Sand (4). This sample is used for comparison with MSW-Ilmenite since the bed was mixed with sand. Fresh rock ilmenite, REF, was used as reference for the XPS-analysis.

Characterization techniques
The total elemental composition of the samples was analyzed using ICP-SMFS according to the standards EN ISO 17294-2: 2016 and EPAmethod 200. 8: 1994. The soluble chlorine content was determined according to EN ISO 10304-1:2009. These characterizations were performed by the company ALS Scandinavia.
Crystalline bulk phases were determined using powder x-ray diffraction (XRD) with CuK α1 radiation in a Bruker D8 Advanced system. Bottom ash samples were lightly crushed in a mortar before being analyzed. Scans were made over a 2θ-range between 10⁰ and 90⁰ with a step size of 0.05⁰ and counting time 2s/step. The same settings (40 kV, 40 mA) were used for all samples.
The morphology was examined using scanning electron microscopy (SEM). The microscope is coupled to an energy dispersive x-ray spectroscopy (EDX) which allows study of elemental distribution. Samples were mounted on carbon tape. The system used for this investigation was Quanta 200FEG coupled with an Oxford EDX system. For elemental distribution both point-and elemental map analyses were used.
X-ray photoelectron spectroscopy (XPS) was used for surface material characterization, providing both elemental and chemical state information for solid samples. Another strength with this method is the low detection limits which allows detection of elements down to 0.1 at %. The PHI 5000 VersaProbe III Scanning XPS Microprobe (Base pressure of 1 × 10 − 9 bar) has been used with monochromated Al-source (25 W) as excitation source. Dual beam flood of low-energy electrons and low-energy argon ions were used to provide charge neutralization during measurements. Spectra was recorded with a 100 μm beam size and pass energy of 224 eV for surveys and 26 eV for high energy resolution measurements. Analysis points were chosen based on x-ray induced secondary electron imaging (SXI) allowing to specify points with good accuracy. Several analyses were performed in each fraction on different particles.
Charge referencing was made to the adventitious C 1s line at 284.8 eV in order to calibrate the binding energies of other elements. The peak areas of each element were normalized by the atomic sensitivity factors (ASF) in the quantitative analyses of the surface composition [45]. Selected region spectra were recorded covering Pb4f 7/2 , Cu2p 3/2 and Zn2p 3/2 for the quantification and chemical state identification [45,46]. The high-resolution spectra were fitted using the Multipak software with a Smart background due to the multilevel data. Peak constraints regarding the area ratios and spin-orbit separation were defined as reported in literature [45].

Thermodynamic study
Thermodynamic calculations were performed using the software FactSage 7.2 [47]. The program consists of calculation and manipulations modules that access different solution databases and substances for calculations of phase diagrams or conditions for multiphase, multicomponent equilibrium. In this study, the module Equilib was used. This module requires reactant inputs, temperature, total pressure, possible product phases and solution databases to be specified. The module employs the Gibbs energy minimization principle to calculate the amount of each product at the equilibrium state [47].
Global multicomponent equilibrium calculations were used to study minor and TE interaction with ilmenite during OCAC of waste wood and wood chips and municipal solid waste, representing the conditions at the plant located at Ö rtofta and Händelö respectively. The thermodynamic databases used for this purpose were FactSage databases FactPS, FToxid and FTsalt. An additional database was included, named HSCA, which contains species originating from HSC Chemistry 9 [48]. The calculations included 59 solid and 79 gaseous species from HSCA. More specifically, this database includes trace elements compounds of As, Co, Cr, Cu, Mn, Mo, Ni, Pb, V and Zn. Hence, it is believed that the quality of the calculations is improved with respect to these species.
The global calculations were performed at 850 • C and atmospheric pressure. The fuels are of heterogeneous nature and there was no analysis done with respect to composition of the fuel for the two boilers. Hence, the elemental composition of the RWW fuel mixture consisting of 50% recovered waste wood with the rest being equal amounts of wood chips and treetops/branches was estimated from the mean compositions of each fuel fraction as reported by Strömberg, B [21]. The MSW-composition was estimated using the average of six different MSW elemental analyses as reported by Avfall Sverige [49]. As it is clear that the MSW fuel composition can vary, thermodynamic calculations were ---performed using the six reported fuel compositions and compared with the calculated mean value. The difference in phase formation and trace element speciation was studied. No major difference with respect to trace element speciation were found, and hence it is believed that the average value gives a reasonable description of the fuel and its components. The fuel contents are presented in normalized weight percent in Table 2. Fig. 2 illustrates the methodology for the two different thermodynamic calculations performed in this study. The global equilibrium calculations are performed using the fuel composition, Table 2, air and an excess amount of the bed material ilmenite (FeTiO 3 ) as input. The amount of ilmenite was defined as five times the weight input of the fuel. This resulted in the bed consisting of roughly 95% ilmenite and 5% ash. The excess is defined to maintain an ilmenite rich bed. The air flow was varied in order to study the effect of different reduction potentials, since the oxygen partial pressure may vary locally in the boiler and inside the particles. The reduction potential is expressed as the logarithmic ratio pCO/pCO 2 . Higher reduction potentials represent conditions further into the particle, closer to the fuel particles and oxygen deficient parts of a boiler. Additionally, higher reduction potentials may provide an insight for possible compound formation in a CLC fuel reactor. The output from the calculations provides equilibrium compositions of the gas, solids and solid solutions.
These calculations result in useful information about the general behaviour of the trace elements. The final global calculation consisted of 103 solutions and 1959 species distributed as 730 solution phases, 524 gas compounds and 705 pure solids. Local thermodynamic calculations, illustrated in Fig. 2, were performed with the same databases and settings. The calculation is performed locally on one particle and the input is based on the surface composition obtained from the XPS-analysis. These calculations were then performed in a similar way as above but using 71 solution phases and 1299 species distributed as 529 solution species, 220 gas compounds and 550 pure solids.

Results
The elemental composition of the samples is presented in Table 3.
The table shows that the bed sample MSW-Ilmenite (1) contain slightly higher amounts of Si and Al and slightly lower amount of Fe and Ti compared to RWW-Ilmenite (3). However, the Si contents are far from the one in RWW-Sand (4) which has a different composition with respect to all major elements. The main elements in the fly ash MSW-FA (2) are Ca, Si, Al and Cl and S.
The Zn-content is similar in all fractions with slightly lower amounts in the fly ash MSW-FA (2). Pb content is similar in all bottom ash fractions with higher amount in the fly ash MSW-FA (2). Higher amounts of Cu are detected in both samples from Händelö, likely due to the fuel.

Global thermodynamic equilibrium calculations
The phase composition of ilmenite will depend on the reduction potential in the boiler as different regions may be air-rich or air-lean. The oxygen carrying capability of ilmenite depends on the transition between Fe 2+ and Fe 3+ . The most reduced form is FeTiO 3 (FeO + TiO 2 ) and the most oxidized form is    as moderately reducing conditions and could represent conditions relevant in oxygen deficient areas in OCAC. The highly reducing conditions occur above the reduction potential − 2.0 where ilmenite (FeTiO 3 ) starts to form. Higher reduction potentials represent conditions further into the particle or closer to the fuel particles. At even higher reduction potentials, around log 10 [pCO/pCO 2 ] = 1.6 and log 10 [pO 2 ] = − 20.4 atm graphite is expected to form. Table 4 illustrates the different defined regions along with their corresponding reduction potential, oxygen partial pressure and the CO 2 -yield. Fig. 3 illustrates the solid phase distribution across different reduction potentials for the calculation with MSW. The corresponding figure for RWW is presented in Fig. S1 in supplementary material. On the secondary y-axis the logarithmic oxygen partial pressure is presented over a range from approximately 0.06 atm to 3.2‧10 − 18 atm. Further decrease of the oxygen content would result in formation of metallic species which are likely not relevant for the materials studied in this paper. Fig. 3 clearly shows that the bed consists of a large fraction ilmenite, allowing trace element interactions with ilmenite particles to be studied.
In the highly reducing environment the phases ilmenite (

Fate of trace elements for municipal solid waste
Results for the case at the Händelö plant using MSW as fuel are compiled in the figures below. Fate of trace elements were evaluated at different reduction potentials illustrated in Figs. 4-6 showing the phase distributions of zinc, copper and lead. Fig. 4 shows that zinc forms a spinel compound across the whole range. In the oxidizing environment ZnFe 2 O 4 is stable, but when increasing the reduction potential above − 1.9 ZnCr 2 O 4 becomes the stable phase. At the reduction potential − 1.9 a large fraction (53%) of willemite (Zn,Fe) 2 SiO 4 is formed. When going from moderately reducing to highly reducing conditions the main spinel phase (Fe 3 O 4 ) will decrease and the ilmenite phase (FeTiO 3 ) will increase, as seen in Fig. 3. This transformation will likely hinder zinc from entering the spinel phase. Therefore, Zn will start to interact with silica forming willemite. When increasing the reduction potential above − 1.9 (corresponding to a log 10 [pO 2 ] of − 13.5 atm) zinc will start to enter the gas phase, mainly as Zn (g). At the reduction potential − 1.1 (log 10 [pO 2 ] = − 15.1 atm) around 61% of the zinc is retained as solids (45% in the slag phase as ZnS and 14% as spinel, mainly ZnCr 2 O 4 ). Further reduction pushes the equilibrium of Zn towards the gas phase, Zn (g). Thus, for the more reducing conditions, as more zinc is released to the gas phase, it is more prone to end up in the fly ash.
The copper phase distribution is presented in Fig. 5. It shows that from oxidizing to moderately reducing conditions a gas phase is formed which mainly consists of (CuCl) 3 (g). Under oxidizing conditions copper preferably interacts with iron to form CuFe 2 O 4 (s) and with increasing reduction potential Cu 2 Fe 2 O 4 (s) and then Cu (s). When reaching highly reducing conditions, above the reduction potential − 1.9, a slag phase with Cu 2 S is formed accompanied by some Cu 3 As (s) under highly reducing conditions.
The lead phase distribution is presented in Fig. 6. Pb is distributed mainly in gaseous compounds which range from PbCl 2 (g) in the oxidizing and moderately reducing environment parts, to PbS (g) which is the main constituent in the highly reducing environment. At the reduction potential 0.1 PbS (g) is accompanied by Pb (g). A small fraction of lead (approximately 5%) is dissolved in a slag phase containing PbS under highly reducing conditions.

Fate of trace elements for recovered waste wood
Results for the case at the Ö rtofta plant utilizing the fuel mixture recovered waste wood and wood chips are compiled and presented in the supplementary material. For the case with RWW, presented in Fig. S2, most of the zinc will form spinel in combination with iron, ZnFe 2 O 4 , similar to what was seen for MSW in Fig. 4. When highly reducing conditions are employed, above the reduction potential − 1.9, formation of ZnCr 2 O 4 is preferred. At the reduction potential − 1.9, willemite, (Zn,Fe) 2 SiO 4 , forms together with Zn (g) and a small amount of ZnO dissolves into a slag phase. Further reduction pushes the equilibrium towards the gas phase which mainly consists of Zn(g). As opposed to the gas composition in Fig. 4, over 85% zinc enters gas phase as Zn (g) when reaching the reduction potential − 1.1. The gaseous compounds ZnCl 2 , KZnCl 3 and Zn(OH) 2 are formed at oxidizing and moderately reducing conditions but the amount of gas formed at these conditions is negligible.
The copper content is significantly lower in the recovered waste wood and wood chips compared to MSW, Table 2. For oxidizing and moderately reducing conditions the major constituent is Cu 2 O dissolved in a slag phase, presented in Fig. S3. The amount of gas never exceeds 0.6% showing that copper is retained in the bottom ash under all conditions. When entering highly reducing conditions, exceeding the reduction potential − 1.9, the solid compounds Cu 3 As (s) becomes the major phase (above 88%). RWW contains relatively high arsenic content, approximately one third of the copper concentration, which favors the formation of Cu 3 As (s). If the As-content is lowered, the amount of gas is expected to increase at high reducing potentials, forming a larger amount copper chloride accompanied with smaller amounts of copper in the slag phase. As likely originates from wood preservatives such as chromated copper arsenate (CCA). EU has introduced restrictions against CCA since 2004, which has lowered the amount of CCA-treated wood since [51]. However, arsenic may still enter the waste streams since substantial quantities of CCA have been used daily and for a long Table 4 The relation between the logarithmic reduction potential, logarithmic oxygen partial pressure and the yield of CO 2.

Logarithmic reduction potential
Corresponding oxygen partial pressure CO 2 yield log 10  time but could also be due to imported waste wood. There are currently no EU directives stating that CCA-treated wood should be removed from their current applications. Therefore, it is expected to remain in use for some years to come. Thus, CCA treated wood can be expected to be found in waste fuels and affect the speciation of Cu species. The MSW composition presented in Table 2 contains higher amounts of Cl and S compared to RWW, which has important implications on the results. This is reflected in the thermodynamic results where the gas phases of zinc, copper and lead in the case with MSW contain larger fraction of chlorides in comparison to RWW. It is known from literature that Cl may enhance the volatility of several heavy metals while S may decrease volatility due to formation of condensed sulphate phases [30,52]. When comparing the lead phase distribution between the fuels significantly less chlorides are formed for RWW, and significantly larger fraction is kept in the slag phase in oxidizing and at moderately reducing conditions, see Fig. S4. This is likely due to the higher fraction of K compared to Cl in RWW. This allows more K to be associated with Si, favoring slag phase formation. Pb is then selectively dissolved in this phase, and the small Cl content will hinder formation of lead chlorides in the gas phase. The slag phase, which is formed with RWW consists of oxides (ZnO, PbO and Cu 2 O) while the slag phase formed for MSW dominates in moderately and highly reducing conditions and contains sulfide compounds (ZnS, PbS and Cu 2 S).
An additional database, HSCA, was used in this study which includes more compounds not present in the standard FactSage databases. The database allowed additional gaseous and solid compounds to form, improving the calculations. The effects are noticed in the MSW calculations as several predicted gaseous compounds originate from HSCA. For example the gas phase distributions of Cu and Pb in Figs. 5 and 6, where CuCl 2 (g), Cu 2 Cl 2 (g) and Pb 2 O 3 (g) form. For the RWW calculation the effect is noticed in the zinc and copper distribution where Zn (OH) 2 (g) and CuH (g) form. Thus, the database is important in order to describe these systems accurately.

Phase composition and surface morphology
The main crystalline phases as determined by XRD for both MSW-Ilmenite (1) and RWW-Ilmenite (3) are aligned with the results obtained from the thermodynamic calculations (Fig. 3). The phases detected in the two ilmenite bottom ash samples MSW-Ilmenite (1) and RWW-Ilmenite (1) were iron titanium oxide With respect to the observed bulk phases and their oxidation state, it is important to consider that the particles have likely been exposed to conditions between oxidizing and moderately reducing environments. Fig. 3 shows that the spinel and rutile phases are stable here. The particle surface on the other hand is in contact with the gas phase where the average oxygen partial pressure ranges between 0.21 and 0.04, i.e. oxidizing environment. However, the particles may be exposed to more reducing environments locally.
Phases with the elements copper and lead were difficult to observe with XRD due to the low content. However, some impurities could be observed, such as Mg 0.4 Zn 0.6 Fe 2 O 4 for both MSW-Ilmenite (1) and RWW-Ilmenite (3). Traces of zinc silicates could also be observed, which could be a possible interaction path when using silica sand as bed material. Thus, it is likely that locations with higher silica sand concentrations in the bed could result in zinc incorporating in silicates. This was also predicted in the thermodynamic calculations where willemite was formed at moderately reducing conditions. Surface features of the three industrial samples are illustrated in Fig. 7, as determined by scanning electron microscopy. Chemical mapping was performed on a large area to observe trends in composition. The elemental mapping, see Fig. S5 in supplementary material, indicated the presence of lead in submicron particles in MSW-FA (2) while larger particles, pieces of the bed material, could contain a major amount of potassium, aluminum and silicon or iron and calcium. Zinc and copper were also detected, unevenly distributed across the sample. XRD confirmed the formation of calcium sulfate, melilite and elemental aluminum in MSW-FA (2).
Mappings were performed on a larger area to observe trends in  composition. Thereafter several point analyses were performed and two of these, from each bottom ash fraction, are presented in Table 5.
Mapping of the particles from MSW-Ilmenite (1) Table 5 shows the main ash forming elements together with the elements Cu and Zn, with positions indicated in Fig. 7. The results in Table 5 are well aligned with the findings from the global equilibrium calculations where zinc is found in a spinel formation as seen in Fig. 4. In position 1, showed in Fig. 7, the Zn:Fe ratio is ~1:2 suggesting a spinel structure. The lower amount of copper makes detection with SEM-EDX difficult. Point analyses show that the average Zn-content is around 1 at% on the surfaces of particles in both bottom ash fractions. The content of Pb was below the detection limit and therefore it was not possible to draw a conclusion using this approach.

Chemical speciation of particle surfaces
X-ray Photoelectron Spectroscopy (XPS) was performed to evaluate the outermost layer of the particles in detail. This layer is in contact with the gas phase and can be one interaction channel for ash components in both the boiler and downstream of the main combustion chamber. Fig. 8 shows the XPS spectra for the four samples as well as the reference material. The XPS-analyses performed use a beam size of 100 μm which provide information about the surface of the bottom ash particles but a mean composition of several submicron particles for the fly ash sample. Several survey analyses were performed in each fraction, but one is shown here for comparison. The survey spectrum shows the photoelectron peaks which are characteristic for each element at the surface of each particle. Comparison with the fresh sample REF (5) shows accumulation of elements in both bed samples, i.e. MSW-Ilmenite (1) and RWW-Ilmenite (3). The Fe, Ti and Mg peaks have slightly lower intensities while Na, Ca, K and S increase accompanied by minor elements such as P, Zn, Cu and Pb. RWW-Ilmenite (3) showed slightly higher content of Fe and Ti at the surface compared to MSW-Ilmenite (1) which instead contained higher amounts of Ca, Al and Si. The survey for MSW-FA (2) shows that the submicron particles are enriched in Ca, K, Cl accompanied by minor amounts of Zn, Al, S, Cu and F.
The XPS survey quantification, presented in Table 6, allows detection of elements on the surfaces of the bed material which is not possible using other characterization techniques such as SEM-EDX. Several point analyses performed showed that Zn and Cu were detected to be slightly above 0.5 at%. Pb was also detected in all samples except MSW-Ilmenite (1) where the concentration was <0.1%.
In order to obtain more detailed information regarding the chemical states of zinc, copper and lead, high-resolution analysis was performed. The high-resolution analysis is performed with lower pass energy and narrow region, which is unique for each element. The spectral regions Cu 2p, Zn 2p and Pb 4f were selected for this purpose as no overlapping is detected for these species, and as they are also the focus of this specific work.

Chemical state investigation -copper
The spectral region for Cu 2p is located at binding energies between 930 eV and 960 eV. The XPS-quantification procedure for Cu-species incorporate the use of shake-up lines and binding energies of the main peak. It is known that Cu 2+ may be identified using the strong shake-up satellite adjacent to the main Cu 2p 3/2 peak [53]. This peak is almost not present in the case of metal Cu but may be present as a weak peak for Cu + compounds. Additionally, the corresponding Cu LMM line can be useful for further chemical state analysis. However, in this study the Auger peak was difficult to obtain because of the overlapping Ti 2s peak

Table 5
Elemental composition (at %) of selected points of the SEM-EDX analysis in Fig. 7.  [55]. However, the satellite for CuFe 2 O 4 is reported to occur at 942.8 eV with lower intensities compared to CuO [55]. Fig. 9 shows the high-resolution scans for each sample and specie together with the fitted components. Copper was observed in the samples MSW-Ilmenite (1) and MSW-FA (2). The two RWW samples did not indicate any copper content at the surface, most likely due to the low content in the fuel mixture. The intensity of the observed Cu 2p peaks were high and the peaks were narrow. The peaks were fitted using the spin orbit splitting together with the positions 933.2 eV, 952.8 eV and 942.8 eV, corresponding to the Cu 2p 3/2 , Cu 2p 1/2 and satellite peak as reported in literature for CuFe 2 O 4 [55]. The same full width half maximum was used for all chemical states (FWHM = 2 eV). An additional peak located at 935.2 eV was used for the fly ash, corresponding to CuCl 2 [56]. Results show that approximately 80% of Cu is bonded as an oxide with the rest being copper chlorides in the fly ash. The fitted components presented in Fig. 9 are well aligned with the experimental data from the literature, confirming the formation of CuFe 2 O 4 and CuCl 2 .

Chemical state investigation -zinc
The spectral region for Zn 2p is located at binding energies between 1019 eV and 1050 eV. The high-resolution spectra for Zn 2p are presented in Fig. 9. The chemical state analysis may be performed using the binding energies and corresponding Zn LMM peak. However, in this study the intensity of the Zn LMM Auger peak was too low for all samples in order to confirm the chemical state. Thus, chemical states are compared with experimental data reported in literature using the same experimental settings [57]. The binding energy for the main peak Zn 2p 3/2 of ZnFe 2 O 4 has been reported to be 1021.6 eV [58]. It is known that the photoelectron peak of Zn 2p 3/2 in tetrahedral and octahedral sites of ferrites can be observed with an energy difference around 1.9 eV. Zinc atoms occupying tetrahedral sites may be ascribed to the binding energy 1021.3 eV while peaks with the binding energy 1023.2 eV correspond to incorporation of Zn 2+ ions at octahedral sites [59]. The interaction between zinc ions and alumina has been studied by Strohmeier et al. [60] The study showed that low zinc loadings allows formation of a spinel phase where zinc ions occupy tetrahedral sites. One study, using similar settings as this paper, reported ZnO/ZnAl 2 O 4 composites to have the binding energies 1022.4 eV and 1045.7 eV for Zn 2p 3/2 and Zn 2p 1/2 respectively [61].
The MSW-Ilmenite (1) spectra was fitted using two parameters due to the broad peak. The peak positions 1021.3 eV and 1023.2 eV were used in the fitting process corresponding to Zn 2+ ions occupying either tetrahedral or octahedral sites. It is observed that 83% belongs to Zn 2+ ions occupying tetrahedral sites. The peak for MSW-FA (2) was 1022.5 eV, close to the reported value of ZnAl 2 O 4 . The peak position for Zn 2p 3/ 2 in RWW-Sand (4) is found to be 1021.3 eV. The XPS-quantification results for sample RWW-Sand (4),

Chemical state investigation -lead
The spectral region for Pb 4f is located at binding energies between 136 eV and 146 eV. The high-resolution spectra for Pb 4f are presented in Fig. 9. Previous studies have been made investigating various lead compounds using XPS [62]. The peak position for MSW-FA (2) is 139.2 eV which is in the range of PbCl 2 as reported in literature [62]. The position for RWW-Ilmenite (3) is found to be 137.9 eV corresponding to PbTiO 3 while RWW-Sand (4) has a position at 138.8 eV related to PbSiO 3 [62]. The sample MSW-Ilmenite (1) indicates that lead could be present, Fig. 9, but due to the low concentration the peaks were not fitted. However, the peaks are visually like those of RWW-Ilmenite (3). This could indicate that lead is buried beneath the top surface. A summary of the phases detected in each sample is presented in Table 7.

Comparison with local thermodynamic equilibrium calculations
Local thermodynamic equilibrium calculations are performed based on the quantification results from the survey analysis presented in Table 6. The final calculations consisted of 71 solution phases and 1299 species distributed as 529 solution species, 220 gas compounds and 550 pure solids. The temperature was set to 850 • C for the bottom ashes and 250 • C for the fly ash calculation. Results from the local thermodynamic calculations are presented in Table 8. The surrounding gas phase was set to consist of 5% O 2, which is a mean value for the concentration of oxygen in the gas outlet from the boilers.
The local thermodynamic calculations predict spinel-compounds in each fraction with some amounts of olivine in RWW-Sand (4). The chemical states are in accordance with the local thermodynamic calculations which predicted ferrites and lead oxides in the bottom ash and zinc aluminate and condensed lead chlorides in the fly ash. However, some discrepancies are also observed. The formation of PbO in RWW-Sand and (CuO)(Fe 2 O 3 ) in MSW-FA are two compounds not completely aligned with the experimental findings. Formation of (CuO) (Fe 2 O 3 ) is stable across temperature regions between 100 • C and 900 • C and could end up in the fly ash together with entrained bed particles. Several analyses showed that the concentration of copper was slightly higher in larger particles (>50 μm). Submicron particles are inevitably characterized along with these larger particles in the fly ash, which could affect the thermodynamic outcome as the mixing is not optimal.
Studying the PbO-SiO 2 phase diagram shows that lower amounts of Pb favors slag phase formation above 750 • C while lead silicates are favored below 750 • C explaining why thermodynamic calculations predict a slag phase. The PbO-TiO 2 phase diagram on the other hand shows that (PbO)(TiO 2 ) is stable across the whole region with one small region forming a slag phase at high contents of Pb (>99.5%). Other explanations could be due to not reaching equilibrium or kinetic limitations since the calculations assume perfect mixing and infinite reaction time. Nonetheless, these thermodynamic calculations have shown to be useful and consistent with the experimental findings. Thus, including silica sand during biomass combustion could provide some additional possibilities to retain lead in the bottom ash.

Discussion
This study presents an investigation of the interaction between ilmenite and two different waste fuels at two commercial CFB units employing oxygen carrier aided combustion, or OCAC. The focus has been on the fate of trace elements, where both detailed characterization techniques and thermodynamic equilibrium approaches have been used for increased understanding. The trace elements Pb, Zn and Cu have been in focus, which are species that can be found at significant levels in various waste fuels. The use of oxygen carriers in normal fluidized bed combustion could have advantages with respect to combustion efficiency and gaseous emissions. Further, there could be potential benefits with respect to the fate of reactive ash components, as these could react with the oxygen carrier to form stable compounds. In normal combustion, these species are known to react with chlorine to highly volatile species, which could evaporate and be precursors for corrosion phenomena downstream of the main boiler. Hence it is important to study the fate of these elements using oxygen carriers. This study focuses on ilmenite, which is a benchmark oxygen carrier, and as it contains Fe, the results should be highly applicable for other types of Fe-based oxygen carriers [63].
The results here are applicable on OCAC-systems as discussed throughout this paper. However, the results can be highly relevant for other systems using oxygen carriers, for example chemical-looping combustion (CLC) and chemical looping gasification (CLG) [64]. The global thermodynamic calculations were performed at various reducing potentials, which are relevant for these three technologies. Further, it can be expected that the particles in a CFB boiler could be exposed to variable oxidizing and reducing conditions even in the same reactor [65]. Hence, it can be presumed that many of the reaction pathways of the oxygen carriers and ash components could be the same. Of course, the residence times at different reducing potentials and actual contact with ash components could be different. In a CLC and CLG system it is not expected that any fuel conversion would take place in the air reactor, and hence any transformations in this reactor would likely stem from transfer of particles from the fuel reactor, both oxygen carriers and ash.
The degree to which the oxygen carrier reacts with the trace elements is highly dependent upon the degree of reduction in the reactor. In Table 7 Summary of phases identified using XPS.  Table 7.
These mixed oxides could be stable in the oxygen carrier particles and they could also have oxygen carrying capabilities. For example, copper ferrites [66][67][68] and zinc ferrites [69] have previously been studied as oxygen carriers. This opens for the possibility to increase oxygen carrying capacity of the bed as the residence time increases, something which would be highly beneficial.
From the global equilibrium calculations, it is clear that there are some differences in the fate of the trace elements depending upon the gas phase. For example, Zn is more prone to enter the gas phase, as the degree of reduction potential increases, something which could be highly relevant if Zn is to be recovered, as seen in Fig. 4. Similarly lead displays more volatile behaviour with increasing reduction potential. However, if the amount of Cl in the fuel is high (as in MSW) lead could evaporate even at lower reduction potentials forming PbCl 2 (g), observed in Fig. 6. As the reduction potential increases Pb (g) and PbS (g) become major constituents. The volatility of Cu is also affected by the amount of Cl. Higher amounts of Cl, as in Fig. 5 for MSW, will favor (CuCl) 3 (g) formation while increasing reduction potential favors slag phase formation (Cu 2 S) or Cu 3 As (s). For CLC and CLG, where the air and fuel reactors are separate, corrosion and fouling in the air reactor could be reduced as the elements will be concentrated in gas outlet from the fuel reactor. The latter has a significantly lower gas flow compared to that of the air reactor, which is a big advantage.

Trace element reaction paths
There are potential benefits using an oxygen carrier which could interact with some trace elements. Under fluidized bed conditions the mixing is more intense and contact between feedstock and bed material is enhanced. Interaction with ilmenite could prevent evaporation of Zn, Cu and Pb. The global thermodynamic calculations indicate that the presence of ilmenite increases the amount of copper and zinc retained in the bottom ash as ferrites. The elemental composition in Table 3 shows that both bed materials, ilmenite and sand, contain some amount of lead. The ilmenite rich bottom ash contained PbTiO 3 and the silica rich ash contained PbSiO 3 at the particle surfaces. As discussed by Dong et al. the metal matrix reaction (1) may play a significant role for the Pbretention [31]. This is due to incorporation reactions of metals with SiO 2 , which is enhanced with increasing oxygen concentration as in reaction 1. With ilmenite rich beds one important interaction channel could be with TiO 2 and the thermodynamically favorable compound PbTiO 3 . Thermodynamic calculations also show that reaction between lead and TiO 2 is more favorable compared to SiO 2 . Considering the migration of iron and ash layer formation it is possible that lead interacts with TiO 2 at the surface of the particles, but with increasing residence time becomes a part of the titanium rich core. Both bottom ash ilmenite samples which are investigated in this work contain approximately the same amount of lead. However, the high-resolution scans showed that it was difficult to observe lead on the surface of the particles in MSW-Ilmenite (1) which could suggest that it is buried beneath the top surface. Further studies are required to investigate the influence of residence time and the distribution of these elements inside ilmenite particles.
Ilmenite particles change when undergoing continuous redox reactions. Iron will migrate towards regions with higher oxygen potential, the particle surface, forming an iron layer [70]. The iron layer could be both of homogenous and heterogenous nature. Nevertheless, this increases the availability of iron at the surface and is one important interaction channel with the ash components [70]. As shown, the iron at the surface of ilmenite interact with both zinc and copper forming the ferrites ZnFe 2 O 4 and CuFe 2 O 4 . Thus, it is proposed that zinc and copper are incorporated in the oxygen carrier during the redox reactions. It is of interest to further understand the effects that these elements have on the bed material and its properties.
From this study it is clear that ilmenite can interact with all investigated elements and hence be obtained as mixed oxides. The implication for combustion processes using oxygen carriers could be; • Formation of a more benign fly ash, as some heavy metals combine with the bed material, something which could be valuable if the ashes are landfilled or recycled to soils or forests. • Corrosion or fouling on downstream heat transfer surfaces could be less. • Enhanced fuel conversion as mixed oxides could result in new phases which could be oxygen carriers themselves. • Opens up for the possibility of combined process for oxygen regeneration and metal extraction, i.e. Cu and Zn.

Conclusions
This work has investigated trace element interaction with ilmenite using a thermodynamic modelling approach and detailed experimental work using x-ray photoelectron spectroscopy (XPS). Samples from two commercial plants utilizing two different waste fuels were investigated. XPS is found to be a good complement for sample characterization of trace elements, especially since the electron microscope and x-ray diffraction cannot detect phases and elements at lower concentrations. To summarize, the following interactions between ilmenite and zinc, copper and lead have been identified: • Lead interacts with the titanium in ilmenite forming PbTiO 3 • Copper and zinc interact with iron forming ferrites CuFe 2 O 4 and ZnFe 2 O 4 Thermodynamic calculations were used for further understanding Zn, Cu and Pb behavior mainly under OCAC conditions, but also at conditions applicable for CLC and CLG. To improve accuracy of the thermodynamic calculations, the standard FactSage databases were complemented with relevant data from additional sources. Global calculations were useful for predicting the overall behaviour of TEs under different reducing environments while the local calculations provided detailed information about the surface composition. Both calculations are important to capture the overall behaviour and surface interactions.