Characteristics of olivine as a bed material in an indirect biomass gasifier

Nevertheless, the results from the first run will be used here for comparison, to assess the repeatability of the activa-tion process of olivine in the Chalmers boiler.. To study the influ

Characteristics of olivine as a bed material in an indirect biomass gasifier

Department of Energy and Environment – Division of Energy Technology, Chalmers University of Technology, 412 96 Göteborg, Sweden

h i g h l i g h t s

After a period of activation, olivine shows good characteristics as a catalyst of tar destruction

When activated, olivine increases the yield of H2

Enrichment of the olivine by inorganics from the fuel has noteworthy effects on the activity of olivine and the overall process

Sulfur addition to the combustion side of the system shows a positive effect on the total tar reduction

Addition of silica-sand to the system results in a poorer performance of the system

a r t i c l e i n f o

Article history:

Received 26 January 2015

Received in revised form 14 May 2015

Accepted 18 May 2015

Available online 23 May 2015

Keywords:

Biomass

Gasification

Bed material

Olivine

Gas

Tar

a b s t r a c t

The use of untreated olivine as the bed material in a biomass gasifier is investigated in this work, in which activation of the material is the main focus The experiments were carried out in the Chalmers 2–4-MWth

indirect biomass gasification unit and comprised analyses of the gas composition and bed material, as well as changes in tar yield Starting from the raw material, the first signs of activation, in the form of

a reduction in the tar yield, were observed already during the second day of the operation The tar yield continued to decrease with time, and by the fourth day it was reduced by 30%, as compared to the yield

on the first day of the operation Analysis of the bed samples showed accumulation of inorganics within the bed material, with a share of potassium being present in leachable form Thermodynamic calculations support the indications from the experiment that potassium can be released under gasification conditions and may play an important role in the activation of olivine To examine the impacts of S and silica on the activity of olivine, two experiments were conducted The addition of S to the combustion side gave a pos-itive effect in terms of the tar levels in the raw gasification gas The addition of silica sand revealed, as expected given the affinity of potassium for silicone, negative influences on the tar yield and gas compo-sition that could not be attributed to mere dilution, as compared with the gas produced during operation with pure olivine

Ó 2015 Elsevier B.V All rights reserved

1 Introduction

Recently, biomass-based fuels have attracted great interest as

carbon dioxide-neutral energy sources, as opposed to fossil fuels

In this context, the use of biomass in countries where it is

abun-dant represents a route to reduce the dependency on oil from

for-eign sources utilizing domestic resources instead Among the

different conversion techniques, thermochemical conversion of

biomass via gasification is seen as highly promising[1–5]

The primary gas produced via the gasification step is commonly

referred to as the ‘raw gas’, and it consists of CH4, H2, CO, CO2, light

hydrocarbons, and heavy hydrocarbons The heavy hydrocarbons

are usually referred to as ‘tar’; the usual definition is that the tar fraction includes hydrocarbons with a molecular mass greater than that of benzene This fraction may represent 5–10% of the total energy content of the fed biomass[6] These tar species start to condense already at temperatures of around 360 °C, and they cre-ate operational problems, such as process shut-down and/or downstream catalyst deactivation[2,7–10]

The overall gasification reaction is endothermic and requires energy input As a solution to the partial burning of the fuel in the presence of pure oxygen or air, with the latter diluting the raw gas with nitrogen, indirect gasification is used This technology uses the circulating bed material as a carrier of heat between the two reactors The material is heated up in the combustion process and transferred to the gasifier supplying the heat for the endother-mic gasification reactions By use of catalytically active bed http://dx.doi.org/10.1016/j.cej.2015.05.061

⇑ Corresponding author.

E-mail address: jelenam@chalmers.se (J Marinkovic).

Contents lists available atScienceDirect

Chemical Engineering Journal

j o u r n a l h o m e p a g e : w w w e l s e v i e r c o m / l o c a t e / c e j

materials one can even actively influence the gas composition and

tar yield in the produced gas These materials can be produced

syn-thetically or may have a naturally occurring character, e.g., ores

[10] Given the severity of the conditions in the system, in terms

of high temperatures, the presence of ash, and material circulation,

the material undergoes significant physicochemical changes A

commonly applied bed material in this process is silica sand, which

is a natural material that is relatively cheap, readily available, and

not associated with disposal issues However, silica sand is not

expected to influence the chemical reactions within the system

so as to increase the quality of the produced gas, e.g., attain high

yields of hydrogen and minimize tar content Moreover, the

pres-ence of ash components that chemically bond to the silica sand

results in operational problems, such as agglomeration [11–13]

Therefore, alternative bed materials are needed for optimization

of the product gas at a reasonable cost

Natural materials, such as dolomite, magnesite, limestone, and

olivine, all of which have tar cracking abilities[14], fulfill the above

requirements In particular, olivine, which is a naturally occurring

iron-magnesium mineral, has shown catalytic effects on tar

decomposition in several previous studies[15–19] For this reason,

it is used as the active bed material in the plant located at Guessing

(Austria), which is operated commercially[20] One drawback of

olivine is its ability to transfer oxygen between the combustion

side and the gasification side of the indirect gasification system

[16,21], which lowers the lower heating value (LHV) of the product

gas Furthermore, traces of heavy metals in olivine are

environ-mentally problematic, making disposal more difficult

Olivine has been investigated by several groups, with studies

being carried out on non-calcined, calcined, and doped material

[3,10,22–24], and it is known that olivine needs to undergo

activa-tion to produce the catalytic effects on tar species in the raw gas

The changes in activity of olivine is commonly attributed to Fe

migration in the particle, which has been shown to be a result of

calcination [3], and to the CaO layer that is formed during the

interaction with biomass ash [20,21,25,26] Besides calcium, the

fuel contains varying amounts of potassium (K), silicon, and sulfur

(S), which may also have effects on the chemistry within the

sys-tem On the one hand, high levels of silicon and potassium have

negative influences on agglomeration in the system[11,12]and S

disrupts catalysis downstream of the gasification process[27] On

the other hand, the presence of potassium increases carbon

con-version, as demonstrated by several groups and in different process

set-ups[28–31]

In the Chalmers 2–4-MWthdual fluidized bed gasification

sys-tem, in which silica sand is usually applied as the bed material,

the process has been thoroughly evaluated and reported upon

[6] Olivine has been tested on two occasions in consecutive years

in the Chalmers unit under comparable operational conditions The

materials used differed in terms of particle size, although they

orig-inated from the same mine The first test lasted only 3 days and

was terminated due to operational issues Therefore, it was

repeated in the following year using material with a slightly

smal-ler particle size Nevertheless, the results from the first run will be

used here for comparison, to assess the repeatability of the

activa-tion process of olivine in the Chalmers boiler

In the present work, non-calcined olivine was investigated over

the course of 9 days in the Chalmers unit The focus was on the

activation of the material and the consequences for the

perfor-mance of the gasifier in terms of gas composition, tar load, and

effi-ciency In this context, the roles of inorganics in the gasification

process were investigated through analyses of the physicochemical

changes of the interactions between bed material and ash

ele-ments Furthermore, thermodynamic modeling was performed

based on the analyses of bed material samples, to predict the behaviors of certain ash components

2 Experimental setup The Chalmers boiler and gasifier system (Fig 1) has been used

as the experimental unit in the present work The boiler is a 12-MWth circulating fluidized bed boiler that provides heat to the Chalmers campus The gasifier is a 2–4-MWthbubbling flu-idized bed coupled to the boiler via two loop seals Detailed description of the system can be found elsewhere[6] The letters A–D inFig 1represent the locations at which the bed material is sampled in the system In the present study, the focus was on the collection of samples from Loop Seal 1 (LS1; point B) and Loop Seal 2 (LS2; point C), corresponding to the sites at which the bed material enters and exits the gasifier, respectively The heat that is needed for the gasification reaction is provided

by the hot bed material that circulates from the boiler to the gasi-fier In order to match the normal output of the boiler, fuel is added

to both the boiler and the gasifier Wood chips are used as fuel in the boiler and wood pellets are used as fuel in the gasifier The results of the analyses of the wood chips and the pellets used in the experiment are given inTable 1 The gas that is produced in the gasifier is fed to the boiler and combusted therein

The olivine that was used in the present work was delivered by SIBELCO Nordic AB from a pit in Norway and it had a particle diam-eter that ranged from 180lm to 355lm The chemical composi-tion of the olivine is listed in Table 2 During this series of experiments, the Chalmers system contained approximately 4 metric tons of the material

Throughout the experiments, the gas composition was moni-tored online and the results are presented as average values over

a period of stable operation In addition, tar samples were taken

at each operational point and the bed material samples were col-lected from two points in the system The gas was sampled from

a slipstream that is led from the raw gas line From the same line tar samples were taken for the analysis In theFig 1sampling point

is indicated with red X sign The composition of the raw gas was measured online using a Varian CP4900 micro-Gas Chromatograph (microGC) Sampling of tar species in the raw gas produced in Chalmers gasifying unit was achieved by solid phase absorption (SPA) In order to capture all the fractions, including benzene, for sampling, amines with an active carbon layer were used For each operational point a set of 6 amines was taken Further analyses of the eluted samples were performed in a Gas Chromatograph equipped with Flame Ionization Detection (GC– FID) and average result is used for the presentation of the results Detailed procedures for the sampling and analyses can be found elsewhere[32] To structure and present the results of the tar anal-yses, different approaches are commonly used[33–35] In the pre-sent work, the results are prepre-sented according to the following groups: benzene; 1-ring components; naphthalene; 2-ring compo-nents; 3-ring and higher compocompo-nents; and phenolic aromatics In order to be able to relate the composition of the produced gas to the fuel conversion and to avoid wrong conclusions that can be caused by an dilution effect of the product gas, all results were cal-culated to yields and expressed as mol/kgdry ash free fuelfor perma-nent gases and g/kgdry ash free fuelfor tar species For calculation of the flows, a known amount of helium was used as a tracer gas The collected bed materials were analyzed to follow changes in the elemental composition (ALS Scandinavia AB, Luleå, Sweden) Solid samples were first dissolved in LiBO2according to standard-ized method The elemental analysis was then performed using Inductively Coupled Plasma Sector Field Mass Spectrometry

(ICP-SFMS), with a confidence interval of 95% To observe changes

in the bed material and interactions between the material and ash

components, the sampled material was analyzed by Scanning

Electron Microscopy and Energy Dispersive X-ray spectroscopy

(SEM/EDX) The particles were mounted in epoxy and polished so that a cross-section of the particles could be investigated in an FEI Quanta 200 Field Emission Gun (ESEM) For further analysis and processing of the obtained data, the INCA software was used Furthermore, a leaching experiment was carried out, in which 1 g

of the material was leached with 10 g of deionized water for

24 h, and the leachate was analyzed by Inductive Coupled Plasma Optic Emission Spectroscopy (ICP-OES) As a reference, a fresh oli-vine sample was subjected to the same treatment The analysis was performed with three replicates for each sample

3 Experimental procedure

As different biomasses contain different amounts of inorganic elements, it is important to examine the influences of these species and try to understand their roles in the overall process

3

o o

13 12

2

1

9

5

10

4

22

Gasifier turned 90°

14

16

6 15

11

1 Furnace

2 Fuel Feeding (Furnace)

3 Wind Box

4 Cyclone

5 Convection path

6 Secondary Cyclone

7 Textile Filter

8 Fluegasfan

9 Particle Distributor

10 Particle Cooler

11 Gasifier

12 Particle Seal 1

13 ParticleSeal 2

14 Fuel Feeding (Gasifier)

15 Fuel Hopper (Gasifier)

16 Hopper

17 Fuel Hopper 1

18 Fuel Hopper 2

19 Fuel Hopper 3

20 Sludge Pump

21 Hopper

22 Ash Removal

A

B

C

B Loop seal 1

C loop seal 2

D secondary cyclone

X sampling point

X

Fig 1 Schematic of the Chalmers dual fluidized bed system.

Table 1

Analyses of the wood chips and pellets used during the two weeks of the experiment.

Chips_9 and Pellets_9 correspond to the analysis of the fuel used in experimental

Days I–IV while Chips_10 and Pellets_10 correspond to the analysis of the fuel used in

experimental Days VII–IX.

Dry basis Chips_9 Pellets_9 Chips_10 Pellets_10

S, wt% <0.02 <0.02 <0.02 <0.02

Cl, wt% <0.01 <0.01 <0.01 <0.01

Ti, wt% <0.001 <0.001 <0.001 <0.001

Table 2 Analysis of the raw olivine provided by supplier.

wt%

* Weight reduction after 30 min at 900 °C in air.

To study the influences of inorganics and their interactions with

olivine, three different experiments were carried out: (i) aging of

the olivine in the system over a period of 4 days; (ii) the addition

of S to the combustor; and (iii) the addition of silica sand to the

system The process was run each day under the same conditions

to observe the effects of material activation on the performance

of the gasifier The main operational parameters for this

experi-ment are listed inTable 3 It can be seen that the circulation of

the bed material differs between some experiments To maintain

the desired temperature in the gasifier, some operational

parame-ters which affect the solids circulation rate had to be adjusted

dur-ing the operation

The olivine circulated within the system for a total of 9 days

The gasifier was in operation during 7 of these days, while over

the weekend only the boiler was in operation The results from

Day III are considered to be unreliable due to uncertainties linked

to unstable operation, and are therefore not included in this

eval-uation During the night and over a weekend fuel feed to the

gasi-fier is stopped and the bed is fluidized with the flue gases from the

boiler To the boiler, fuel is fed continuously

Over the first 4 days, the process was continuously run without

any replacement of a certain percentage of the bed with fresh bed

material After this 4-day period, and over the weekend,

replace-ment of the bed material was done whereby approximately

880 kg of the bed inventory (roughly 22% of the total material)

was exchanged with fresh olivine The effects of the replacement

were determined on Day VII To observe the influence of S on the

system, 800 g of elemental S were added directly into the boiler

fuel feed at the middle of Day VII of the measurement The addition

of silica sand was carried out in two steps on the evenings of Day

VII and VIII and involved the replacement of approximately 500 kg

of bed material with fresh silica sand The addition was done in two steps to observe dilution effect (decrease of the olivine content

of the system) The changes were determined during the subse-quent days (Days VIII and IX)

4 Model description

In a second step, the relevant process conditions were evaluated

by means of thermodynamic modeling In order to understand bet-ter and predict the reactions between the bed mabet-terial and species present in the system, a simplified calculation was done with the thermochemical software FactSage 6.2 The EQUILIB mode was used[36] with an intention to explain possible changes in the bed material and paths of the main inorganics, i.e., Ca and K The program uses the minimum Gibbs free energy of the system to cal-culate the equilibrium concentration of the species and the stabile phase distribution Databases considered for the possible solutions were: FTsalt; Ftoxide; and FACT53 The phases considered from the chosen databases were: (1) slag-liquid oxide solution, FToxide slagA; (2) olivine solution, OlivA; (3) solid alkali carbonate solu-tion, FTsalt_KCO; and (4) molten carbonate sulfate solusolu-tion, FTsalt_LCSO

In the Chalmers unit, the bed material is exposed to an oxidiz-ing environment in the boiler (where the temperature is close to

900 °C) and to a highly reductive atmosphere in the gasifier (where the bed temperature is around 820 °C) A simplified schematic of the pathway taken by the material during one cycle in the system

is presented inFig 2 In the boiler, the material can react with both air and ash components Studies conducted on the Chalmers gasi-fier, where the fuel feeding is from the top, have shown that the fuel and char particles are moving on the surface of the bubbling

Table 3

Operational conditions used during experiment.

Date Operational

point (day)

Bed inventory Temperature in

the bed, gasifier (°C)

Solids circulation rate (t/h)

Fuel flow (kg/h)

Fluidization level (kg steam /h)

Operational hours

of the gasifier (h)

Fig 2 Ilustration of the pathway of the bed material during a single cycle in the Chalmers system.

bed where devolatilisation of the fuel occurs also[37] Therefore,

two separate regions have been analyzed for potential reactions

of the bed material in the gasifier: (1) the bottom bed; and (2)

the freeboard In the bottom part, the material is located in an

envi-ronment that is rich in steam, while the material in the freeboard is

in contact with the raw gas

To generate predictions of the behaviors of the main inorganics

in the different zones in the system, three cases are considered:

(1) Reaction between the activated olivine and flue gas,

repre-senting the boiler environment (T = 880 °C);

(2) Reaction between the solid alkali solution from the first case

and the steam, as an indication of the bottom of the bed in

the gasifier (T = 800 °C); and

(3) Reaction between the solid alkali solution from the first case

and the raw gas, as an indication of the freeboard in the

gasi-fier (T = 790 °C)

Conditions used for the calculation are summarized inTable 5

in the Appendix As input data for the software, input variable

for the activated olivine is created from the results of the elemental

analysis (ICP-SFMS) of the sample for Day IV (Table 6 in the

is presented in Table 7 in the Appendix was measured by

microGC For the calculation, a humidity level of 30% was assumed

in the raw gas

5 Results

5.1 Tar yields

The effect of olivine in reducing the amount of tar in the raw

gas, as compared to the silica sand, is one of the main advantages

of using this bed material In general, this effect increases with

ongoing activation throughout the operation

InFig 3, the results obtained for 6 different days over a span of

2 weeks are presented in the form of yields (gtar/kgdaf fuel) for the

specific tar groups The first three bars in Fig 3 (I, II and IV)

represent Days I, II and IV, respectively, during which period no replacement was conducted so that aging/activation of the mate-rial could be followed A clear trend of decreasing yield is evident, showing that all the tar species are affected, and the highest obtained reduction in total tar yield is 30% (Day IV in the figure),

as compared with Day I of operation The most affected group is that of the heterocyclic aromatics, which shows a 53% decrease

in yield The fourth bar inFig 3shows how replacement with fresh olivine over the weekend affects the yield of tar on Day VII Replacement with roughly 880 kg of fresh olivine, which corre-sponds to the exchange of approximately 22% of the total bed inventory of the system, resulted in a 9% increase in the total amount of tar

The effects of adding S or silica-sand on the total amounts of tar and the levels of the individual tar groups are also shown inFig 3 The addition of 0.8 kg of elemental S in the afternoon of Day VII resulted in an additional 20% decrease in the total tar content of the raw gas (Day VII_S) This trend is persistent across all the tar groups

Exchanging 1/8th of the olivine inventory with sand on the eve-ning of Day VII (with measurement on Day VIII) increased the yields of all the tar groups In contrast, the second addition of sand into the system (measurement on Day IX) caused no significant change in the total yield of tar

Considering that silica sand is commonly used as the bed mate-rial in the Chalmers unit, the results obtained in the present study for the yields of tar in the produced gas can be compared to the ref-erence case described by Larsson et al.[38] InFig 4, the yields of the individual tar groups for the first 4 days of the olivine experi-ment are presented, together with the sand reference levels (last bar) The y-axis presents the yields (gtar/kgdaf) of the specific tar groups It should be noted that owing to the different measure-ment technique used for the sand case, as well as the lower num-ber of identified species in the Larsson study, only naphthalene and the heavier tar fractions are presented and compared It is clear from this comparison that olivine is active in the decomposition

of heavy tar species, causing overall decreases in heavy tar compo-nents of 50% and 63% on Day I and Day IV, respectively This trend persists across all the tar groups

Fig 3 Levels of tar components in the raw gas, presented as yields of the specific

tar groups in (g tar /kg daf ) versus time VII_S indicates the addition of sulfur to the

Fig 4 Levels of the tar components in the raw gas, presented as the yields of the specific tar groups in (g tar /kg daf ) versus time The sand reference data are taken from the paper of Larsson et al [38]

5.2 Effects on the gas composition

Fig 5presents the yields of the main gaseous components (in

mol/kgdaf fuel) of the raw gas for each of the experimental days

As additional information, the values from the olivine tests from

the previous season are included in the graph Results from this

experiment are marked as 2013 in the graph The reproducibility

of the trend of activation of the material, especially with respect

to the increase in H2production is clear The yields of CO were

somewhat higher, while the CO2yields were lower Even though

the material had different particle sizes, the amounts of the carbon

and hydrogen in the raw gas were comparable and followed

simi-lar trends Therefore, we conclude that the activation behavior of

olivine observed in the Chalmers system is reproducible

Considering the activity of the material with respect to tar decomposition, an increase in the levels of gaseous products in the system is expected when the olivine is activated Indeed, as shown inFig 5, the yields of H2and CO2are increasing, whereas that of CO is decreasing Decrease in CO and increase in the CO2

yields can be contributed as well to some extent to the oxygen transport of the bed material In a separate experiment the oxygen transport capacity of an olivine sampled from the Chalmers system was determined to around 0.015% mass The test was done in the small scale batch reactor with 10 g of olivine Setup of the system and procedure of the experiment is described in detailed in work of Leion et al.[39]

The levels of CH4and C2Hxshow minor decreases inFig 5 The yield of H2S in the dry raw gas is increasing with time of operation, which indicates an accumulation of sulfur in the bed material

InFig 5, the gas composition obtained after replacement of the material is shown (Day VII) In this case, H2 clearly exhibits an increasing trend, while CO, CO2 and CH4 show more moderate changes The yields of CH4and CO slightly increase and that of

CO2 decreases The yield of H2S in the raw gas dramatically decreased after replacement of material which shows that the main share of the sulfur in the raw gas originates from sulfur that

is stored in/on the bed material

The addition of S to the boiler (point VII_S), resulted in increases

in the H2and CO2yields, while the yields of all other gas compo-nents decreased The yield of H2S in the raw gas increased around 43% after S addition

The consequences of adding sand to the system are shown in

Fig 5for time-points VIII and IX The first addition of sand (on the evening of Day VII) resulted in a significant decrease in the yields of H2and CO2 The sand influenced to a lesser extent the

CO content, while the CH4content remained essentially constant The yields of the light hydrocarbons increased Moreover, the gas composition was similar to that obtained with non-activated olivine

With the second addition of sand (on the evening of Day VIII), the change in the gas composition was not so significant, although the same trend was observed as for the first addition

Fig 5 Compositions of the dry raw gases (in mol/kg daf fuel ) obtained at different

time-points during the experiments in the Chalmers gasification unit.

Fig 6 (a) Bed material composition expressed as oxides (in wt%) Percentage of SiO 2 and MgO are divided by 10 (b) Changes in the material compositions between Loop Seal

1 and Loop Seal 2 for Day IV, expressed as differences in the amounts of oxide (in %) The differences in iron and titanium levels are compared with the fresh olivine that

5.3 Bed material analysis

Fig 6a shows the results of the analysis of the bed material, as

provided by ALS Scandinavia AB The analysis encompasses the

main elements in the samples collected from the seal entering

the gasifier (LS1), which includes non-activated olivine (Day I),

activated olivine (Day IV), and bed material post-replacement

(Day VII) The levels of all the ash components increase over time

Already on the second day of operation, there are large

accumula-tions of Ca and K in the bed material

As the only source of Ca and K is the fuel, which is fed to both

the boiler and gasifier, it can be calculated that in total

approxi-mately 19 kg of potassium and 32 kg of calcium were added during

the first day of operation (calculations based on the fuel analysis

given inTable 1)

A comparison of the compositions of the materials entering and

exiting the gasifier, e.g., at LS1 and LS2, respectively, for the fourth

day (Day IV) of the operation is shown inFig 6b As it can be seen,

there is a small but consistent decrease in the ratios of oxides

(except for P2O5) entering and leaving the gasifier Due to the

non-uniformity of the elemental composition of an ore, differences

in the elements that originate from olivine material cannot be

cussed However, for the elements that originate from ash this

dis-cussion is acceptable The difference in K2O levels between the two

seals was significant (21%) The level of Na2O also showed a

signif-icant change, although the total amount of Na in the material was

much lower compared to the levels of the other elements, so it is

not expected that it does have as significant an impact on the

pro-cess as K does However, considering the activity of this metal, its

role should not be ignored When discussing the differences in ash

components presented inFig 6b, one has to be aware that one part

of the ash will be attrited and in form of fines will leave the gasifier

together with the raw gas

5.4 SEM/EDX

Fig 7shows the results of the SEM/EDX analysis of a collected

olivine sample Only the results for activated olivine (Day VII) from

the seal where material enters the gasifier (LS1) are presented This sample was chosen to demonstrate the elemental distribution in the cross-sections of the particles Analysis of the solid sample from the seal where material exits the gasifier (LS2) was analyzed

as well but showed no difference in the elemental distribution and

is not presented

As anticipated based on a previous investigation conducted by Kirnbauer et al.[20], the analysis revealed the formation of a Ca layer around the particles However, Fe did not seem to be migrat-ing to the outer layer, as was observed in a previous investigation

[25] In the samples extracted from the Chalmers unit, Fe was dis-tributed throughout the particles Potassium was observed in the melted layer between the particles and within the cracks in the particles In the spot analysis (Fig 7), it is clear that the melted layer consists mostly of Si, K, and Mg (point 2 in top-left panel), whereas the outer part of the particles consist mostly of Mg, Si, and (as a minor component) K (point 3 in bottom-left panel) Assuming that the surface of the particle is reactive and not the core, spots in the outer part of the particle were analyzed

InFig 8, the results of the mapping analysis of samples taken after sand addition to the system are shown (point Day VIII), together with a table with the spot analysis results By comparing particles in the images, it is possible to identify a quartz-sand par-ticle as in its core detected were only Si and O while Fe and Mg is absent In such a particle most of the K is located in the melted layer around the silica-sand core Moreover, it seems that K has dif-fused into the silica particle According to the spot analysis, the melted layer consists mainly of potassium-silicate Calcium still remains in the outer layer of all particles

5.5 Leaching analysis

To extract information about the presence of soluble potassium compounds in the bed material, a leaching analysis was performed

on the material collected on Day IV

Potassium, K, was found in the leachate in an amount that cor-responds to 52 mg K per 1 kg olivine However, one needs to be aware of the non-uniformity of the ash content of the particles

wt %

Point 3

wt %

Mg 11.4 27.9

e F g

a

Therefore, the results of this analysis should not be discussed

quantitatively, but should instead be used as supplementary

infor-mation to indicate how strong K is bound and whether K can be

released from the bed material When results of the leaching

anal-ysis are compared with the difference of potassium-oxide shown in

Fig 6b it can be concluded that difference measured by ICP-SFMS is

much bigger This more significant loss of potassium from the bed

in the gasifier can be explained by the partial attrition of the ash

layer from the surface of the particles

5.6 Thermodynamic calculations and predictions

To complement the experimental results, equilibrium

calcula-tions were made for different cases of the bed material interacting

with flue gas, steam, and raw gas The result of the equilibrium

cal-culation for the first case, e.g., activated olivine in flue gas, is

pre-sented in Table 4as the starting point, with indications of the

compounds most likely to be present and the respective gas, solid,

slag, and OlivA solutions

The calculation shows that a minor proportion of the K exists in

sulfate form, while the majority of the K is in the form of slag

Considering the low amount of sulfur present in the system,

com-pared to the amount of potassium, less than 1% of all potassium in

the bed material is predicted to be in form of the sulfate

(corre-sponds to the 86% of the sulfur) Given that 4t of the bed material

are circulating through the system, even this small share of the

potassium-sulfate in the material predicted can be expected to

affect the process To allow conclusions about possible further

phase transformations of K, a separate input variable for the slag,

based on the composition obtained in the previous step was made

By simulating a gasifier case for the slag, e.g., reaction with steam

and reaction with raw gas, the software predicted that K would not

be released in any part of the system; instead, it would remain in

the form of slag The same was done for the OlivA solution which remained unaffected by the raw gas and the steam As for the

K2SO4(s), the calculations showed possible decomposition in the gasifier In the environment that prevails in the gasifier, the stabil-ity of the aforementioned sulfate can be affected by reactions with the raw gas to form KOH(g) and K2CO3(s) In the part of the bed where the environment is rich in steam (assumed 100% steam),

no significant decomposition of the potassium-sulfate emerged from the calculations Less than 1% was predicted to go to the gas phase as KOH(g)

Based on the results of the elemental bed material analysis and the observed change in Na content (Fig 6b), the distribution of this element was examined The software predicted that all the Na would be in the slag in the form of Na2O and NaAlO2 In the case

of a reaction between the slag and raw gas, it was predicted that

Na would remain in the same form The same result was obtained for the reaction with the steam

With the intention to elucidate how additional S might affect the bed material, the same reaction system, comprising the olivine and flue gas input variables, with the addition of S was run with the FactSage software The calculation showed the formation of the same phases (seeTable 4), although with a higher level of solid potassium sulfate than in the first case

5.7 Operational issues Throughout the operation of the gasification system, there was

a problem with the high levels of CO emissions from the boiler, which tended to increase with the time of operation The boiler was operating at all times above stoichiometry, with 5% O2in the flue gas, which corresponds to a much more excessive air-to-fuel ratio than is usually used during normal operation with silica-sand, which is around 3.5% O in the flue gases At this point,

Fig 8 Results of the SEM/EDX analysis of used olivine particles mixed with sand The table contains the results of the elemental analysis of the chosen spot.

Table 4

Equilibrium calculations for the reactions between activated olivine and air in the boiler.

Si K 2 O/Na 2 O/SiO 2 /Al 2 O 3 CaMgSi 2 O 6 /CaSiTiO 5 MgMnSiO 4 /CaMgSiO 4 /CaFeSiO 4 /Mg 2 SiO 4 /MgFeSiO 4

it is important to highlight that, unlike the gasifier used in Gussing,

the Chalmers system does not have an after-burner

InFig 9a–e, the CO profiles for all the days of operation of the

gasifier are shown Already after Day I a high level of CO is

appar-ent, while on Day V the signal reaches levels above the maximum

detection level of the CO analyzer An interesting observation is

that for each day in which fuel was introduced into the gasifier,

the level of CO emission from the boiler decreased

Hindiyrati et al.[40]performed a study, comprising

experimen-tal work and chemical modeling, in which the influence of the K

salt on CO oxidation was in focus The study indicated that alkali

could act as a strong inhibitor of CO oxidation at atmospheric

pres-sure and in the temperature range of 773–1373 K

In contrast to K, S was found to enhance CO oxidation,

depend-ing on the stoichiometry and the amount of SO2 in the system

Previous research [41,42]has shown that even small

concentra-tions of SO2 influence CO oxidation at the temperature levels

expected in the boiler Although different mechanisms have been

proposed for this, they have been mostly published in technical

reports For example, in one study[42]it was shown that S

addi-tion at 20 mg/MJfuelto the boiler enhanced CO oxidation

In the raw gas of the Chalmers gasifier, the concentration of H2S

is rather high [up to 0.25%vol(50 mgS/MJboiler fuel)] considering that

wood pellets with S content of <0.01%wtare used as the fuel As the

set-up of the system is such that the raw gas line is coupled to the

boiler where the raw gas is burnt, H2S that originates from the raw

gas will be oxidized to SO2in the boiler This addition of SO2to the

boiler may be the reason for the decreasing levels of CO in the flue

gas observed each time when fuel was introduced into the gasifier

(Fig 9a–d) As shown inFig 9e, the level of CO emission decreased

to zero after S addition to the boiler, albeit only temporarily This implies that S affects the inhibiting component for CO oxidation, most likely through the formation of compounds that can be decomposed in the system If decomposition occurs in the gasifica-tion part of the system, the products of decomposigasifica-tion will end up again on the boiler side, given the recirculation of the raw gas In line with the thermodynamic calculation used in the present work (seeTable 4), potassium sulfate is formed in the boiler and decom-posed in the gasifier These conclusions are also in line with the results of previous studies in relation to corrosion problems caused

by alkali[43–45] These studies have shown that the addition of S compounds to the boiler causes a decrease in the level of gaseous alkali in the gas by forming potassium sulfate, as suggested by the model of gas to particle transformation

Silica-sand addition (Fig 9e) resulted in a similar decrease in

CO emission, which implies that the same specie was affected as

in the case of S addition That the CO decrease caused by silica-sand addition was found to be permanent suggests that the inhibiting component for CO oxidation is non-reversibly bonded to the sand in the boiler Referring to the findings of the present work (seeFig 8), sand addition results in the forma-tion of potassium-silicate, a compound in which the K is chem-ically bonded into a silicate structure from which it is not readily released

Taking into consideration the findings reported in the literature

on the influences of inorganics on CO oxidation and the results of the present study, the hypothesis that K plays an important role

in the overall process can be confirmed

Fig 9 CO concentrations in the flue gas.

6 Discussion

The decrease in tar content observed after activation of the bed

material was expected However, considering that the experiment

started with raw, non-calcinated, olivine, it was surprising that

activation of the material occurred already after 1 day of operation

of the system Examination of the tar yield in the raw gas (see

Fig 3) reveals a trend of decreasing content for all the tar groups

The groups that involve P3-ring aromatic components are not

affected to the same extent as the other groups This result is in line

with previous findings, whereby it was concluded that olivine has

low catalytic activities for heavy tar components [5] Comparing

with sand as a reference, the olivine activity for tar decomposition

is significant Yields of tar components that are heavier than

naph-thalene are significantly lower in the olivine case than in the sand

case

As a significant decrease in tar content was observed over time,

the increase in H2can be attributed to steam reforming of

hydro-carbon species (seeFig 3) Moreover, the increases in H2and CO2

levels, together with the decrease in CO level reveal the influence

of the Water Gas Shift (WGS) reaction Consequently, the increased

activity improves the H2/CO ratio from 1.1 to 1.7 Considering that

Fe migration to the surface was not observed (seeFig 7), activation

of the olivine cannot be attributed solely to this element, as

men-tioned in the literature[3,25] One possible explanation for the lack

of Fe migration is that the olivine was not pre-calcinated, as it was

in the study of Lancee et al.[25] As the same behavior in terms of

tar decomposition was observed for the olivine used in the present

study, other explanations for the activity of olivine, such as ash

components, are more likely

In a combustion plant, ash is normally distributed between the

bottom ash and fly ash, and not all of the inorganics that enter the

system will be retained in the bed Furthermore, some fractions of

these inorganics are expected to be retained on the inside walls of

the furnace As the bed material analysis shows, in the present

study retention of the inorganics in the bottom bed material is

sig-nificant This is as expected, not only because of the expected

inter-action between the material and the biomass ash, but also due to

the fact that no replacement of the bed was conducted in the first

4 days of the experiment The subsequent decrease in the levels of

ash components (on Day VII) can be attributed to replacement of

the material and the removal of ash-coated olivine (seeFig 6a)

The overall activity for tar decomposition decreased after

replacement of the bed with fresh olivine, while a steady increase

of activity was observed in the period spanning Days I–IV

Considering the previous studies, in which the activity of the olivine

was connected to the formation of an iron layer and calcium layer

on the surface of the particles[20,21,25], the observed result after

bed replacement is anticipated due to the dilution of active bed

material in the system Considering the ability of olivine to transfer

oxygen, a side-effect of the dilution with silica-sand is the increase

in CO content and the decrease in CO2content as the oxygen

trans-port decreases protrans-portionally with increasing dilution However,

the addition of silica-sand did provoke a much stronger response

than that expected from a simple dilution The yields increased

for all the tar groups, and the gas composition was similar to that

for non-activated olivine, measured on the first day of the

experi-ment From the SEM/EDX images, it is clear that the sand particles

react with K (seeFig 8), and it is expected that the majority of the K

will be bound into the potassium-silicate structure, thereby

pre-venting the release of K Given the well-known catalytic effect of

K, formation of the potassium-silicates affects the carbon

conver-sion rate, as less K is free to react with the fuel and raw gas As

observed following the second addition of sand to the system, the

changes in gas composition and tar content are not linear This

implies that the reason for the decrease in the overall activity of the bed material is not due to a decrease in the olivine share in the system, but rather that the majority of the available K reacts with the sand already during the first addition

The result from the FactSage calculations show that all the com-ponents indicated in the literature as being potentially active with respect to tar reduction, with the exception of K, are in the solid phase in all parts of the system A part of the K in the system is pre-dicted to be in the slag form, as confirmed by the SEM/EDX analysis

of the sampled material (seeFig 7) In this form, the K should not

be easily released into the gas phase Another part of the K is in the solid phase and in the form of sulfate that can be transformed in active potassium specie in the environment that prevails in the upper part of the gasifier Indeed, as the leaching analysis shows,

K is present in the material in leachable form, meaning in the form

of salt According to the thermodynamic calculations, KOH and

K2CO3will be formed The catalytic effects of these species for car-bon gasification are well-known[28–31] Moreover, the activities

of K salts toward tar species are known from research on

K2CO3-catalyzed steam gasification [46], which showed a decreased total tar level In that work, the most affected species were phenolic and polycyclic Even though the positive effect is well-known, it remained unclear as to whether K is preventing the formation of tar species or is catalyzing their decomposition Since the addition of S to the boiler affected the gasifier perfor-mance, it seems reasonable to conclude that the added S reacts with the material in the boiler and is transported by circulation

to the gasifier According to the thermodynamic calculations, in the boiler environment, the S reacts with the K to form K2SO4 Subsequent decomposition of the transported K2SO4leads to more KOH and K2CO3 becoming available for the reactions within the gasifier This increase in activity is clearly evident in the gas and tar analyses both before and after S addition (IV and VII_S in

Figs 3 and 5) The increased level of H2S in the raw gas after S addi-tion is in line with the hypothesis Conclusively, the main effect of the sulfur in the system is to bond potassium and make it mobile in the system

7 Conclusions Considering the possible variations in fuel composition, it is of great importance to understand the influences of the inorganic components that enter the process together with the biomass In the present study, olivine was evaluated as a bed material for indi-rect gasification of biomass with the objective of studying the interactions with inorganics under realistic and steady-state condi-tions The experiment was carried out in the Chalmers gasification unit using wood as the fuel and raw olivine as the bed material for

a total of 9 days of operation The main conclusions from the study are as follows:

 After a period of activation, olivine shows good characteristics

as a catalyst of tar destruction, giving a 30% reduction in total tar yield compared to the first day of the operation

 When activated, olivine increases the yield of H2 The highest

H2/CO ratio obtained was 1.7

 Enrichment of the bed particles by inorganics from the fuel has noteworthy effects on the activity of olivine and the overall process

 Addition of silica-sand to the system results in a poorer perfor-mance of the system and the production of a gas of lower qual-ity; the effect is greater than would be expected solely from dilution of an active solid As non-active potassium-silicates are formed at the same time, K appears to be a crucial active component