Bioresource Technology 101 (2010) 5601–5608

Contents lists available at ScienceDirect

Bioresource Technology
journal homepage: www.elsevier.com/locate/biortech

Thermal behaviour and kinetics of coal/biomass blends during co-combustion

M.V. Gil, D. Casal, C. Pevida, J.J. Pis, F. Rubiera *
Instituto Nacional del Carbón, CSIC, Apartado 73, 33080 Oviedo, Spain

a r t i c l e i n f o a b s t r a c t

Article history: The thermal characteristics and kinetics of coal, biomass (pine sawdust) and their blends were evaluated
Received 21 October 2009 under combustion conditions using a non-isothermal thermogravimetric method (TGA). Biomass was
Received in revised form 28 January 2010 blended with coal in the range of 5–80 wt.% to evaluate their co-combustion behaviour. No significant
Accepted 3 February 2010
interactions were detected between the coal and biomass, since no deviations from their expected behav-
Available online 1 March 2010
iour were observed in these experiments. Biomass combustion takes place in two steps: between 200 and
360 °C the volatiles are released and burned, and at 360–490 °C char combustion takes place. In contrast,
Keywords:
coal is characterized by only one combustion stage at 315–615 °C. The coal/biomass blends presented
Biomass
Coal
three combustion steps, corresponding to the sum of the biomass and coal individual stages. Several
Co-combustion solid-state mechanisms were tested by the Coats–Redfern method in order to find out the mechanisms
TG responsible for the oxidation of the samples. The kinetic parameters were determined assuming single
Kinetics separate reactions for each stage of thermal conversion. The combustion process of coal consists of
one reaction, whereas, in the case of the biomass and coal/biomass blends, this process consists of two
or three independent reactions, respectively. The results showed that the chemical first order reaction
is the most effective mechanism for the first step of biomass oxidation and for coal combustion. However,
diffusion mechanisms were found to be responsible for the second step of biomass combustion.
Ó 2010 Elsevier Ltd. All rights reserved.

1. Introduction order to provide a stable flame (Biagini et al., 2002), which could
be attained by using biomass. The ash deposition and fouling prob-
Recent awareness of CO2 emissions has resulted in a shift from lems on hot surfaces, which are commonly encountered in the
less environmental friendly fossil fuels to renewable and sustain- combustion of biomass can be reduced or altogether eliminated
able energy alternatives. Among these, biomass is considered to by burning coal/biomass blends (Haykiri-Acma and Yaman,
be one of the few viable replacement options (Munir et al., 2008). Furthermore, existing coal-fuelled power plants may con-
2009). Biomass can be grown in a sustainable way through a cycli- tinue to be used with very few modifications (Biagini et al.,
cal process of fixation and release of CO2, thereby mitigating global 2002), and as a final argument, the co-utilization of biomass or
warming problems (McKendry, 2002). Biomass fixes CO2 in the waste in existing coal-fired plants is likely to result in a number
form of lignocellulosics during photosynthesis, and the CO2 of environmental, technical and economical benefits (Kastanaki
emitted from the combustion of these materials makes no net con- and Vamvuka, 2006).
tribution to the accumulation of CO2 in the atmosphere or to the Thermogravimetric analysis (TGA) is one of the most common
greenhouse effect. techniques used to rapidly investigate and compare thermal events
Many technologies have been studied in recent years for their and kinetics during the combustion and pyrolysis of solid raw
possible use with biomass, such as combustion, pyrolysis, gasifica- materials, such as coal and woods (Pis et al., 1996; Rubiera et al.,
tion and liquefaction. Co-combustion is also one of the most prom- 1997; Haykırı-Açma, 2003; Skodras et al., 2007; Wang et al.,
ising options for application with renewable fuels. There are 2009). It is able to measure the mass loss of a sample as a function
several reasons to blend biomass with coal or with other types of of time and temperature. The temperatures at which combustion
fuel prior to burning. The co-combustion of coal/biomass blends or decomposition reactions in the sample start can also be followed
will help to reduce the consumption of fossil fuels. Sometimes by TGA. Moreover, quantitative methods can be applied to TGA
biofuel products are mixed with coal to achieve better control of curves in order to obtain kinetic parameters, and the kinetics of
the burning process (Wang et al., 2009). In co-combustion the thermal events can be determined by applying the Arrhenius
processes, a volatile matter content greater than 35% is sought in equation to the separate slopes of constant mass loss (Zhou et al.,
2006; Shen et al., 2009; Wang et al., 2009).
* Corresponding author. Tel.: +34 985 118 975; fax: +34 985 297 662. Cumming and McLaughlin (1982) indicated that the informa-
E-mail address: frubiera@incar.csic.es (F. Rubiera). tion obtained from TGA combustion profiles can be used for an

0960-8524/$ - see front matter Ó 2010 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biortech.2010.02.008
5602 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608

initial evaluation of the combustion behaviour at industrial scale. and a slow heating rate were used to avoid heat transfer limita-
TGA techniques operate at different conditions to those encoun- tions and to minimize mass transfer effects. Duplicate experiments
tered in a pulverised coal combustor. Other bench equipment such for each test were performed in order to test the reproducibility of
as drop tube furnaces and entrained flow reactors simulate more the results. The mass loss (TG) and derivative curves (DTG) of the
closely the combustion conditions of industrial pf combustors. samples were represented as a function of temperature.
However, since the pilot- and full-scale tests are costly to operate, In order to find out whether interactions between the compo-
the above techniques can help in the understanding of coal com- nents of the blends occurred, the theoretical DTG curves of the
bustion behaviour. Although extrapolation to other devices at lar- blends were calculated as the sum of the decomposition curves
ger scale cannot be performed directly, thermogravimetric analysis of each individual component. Thus:
is very useful from a fundamental viewpoint, and for comparison
ðdm=dtÞ ¼ x1 ðdm=dtÞcoal þ x2 ðdm=dtÞbiomass
between samples (Arenillas et al., 2004; Rubiera et al., 2002).
A knowledge of the thermal characteristics of biomass and of where (dm/dt)coal and (dm/dt)biomass are the mass loss rates of the
biomass combustion kinetics is essential for understanding and individual fuels and x1, x2 are the proportions of coal and biomass
modelling combustion in furnaces at industrial scale, both in the in the blend, respectively.
case of co-firing with coal or alone (Munir et al., 2009). Such
knowledge is also necessary for the design and operation of con- 2.2. Kinetic analysis
version systems (Cai et al., 2008). According to Shen et al. (2009),
a good understanding of the decomposition of biomass during The thermal devolatilisation curve is usually obtained as a sum
thermochemical conversion is important for developing an effi- of the contributions of the corresponding individual components
cient processing technology. Most studies related to the kinetics (Heikkinen et al., 2004). However, combustion is a more complex
of biomass decomposition are focused on pyrolysing these materi- process, since the presence of oxygen generates a variety of addi-
als under inert atmospheres. However, the most recent research tional phenomena (Skodras et al., 2007). These include the appear-
has been directed towards the study of biomass decomposition ance of gas-phase reactions between the volatiles released at low
in oxidative environments (Safi et al., 2004; Shen et al., 2009; temperatures and oxygen, and the combustion of the char gener-
Yorulmaz and Atimtay, 2009), a subject about which information ated in the early stages of the solids degradation (Bilbao et al.,
is still scarce. 1997).
The aim of this work is to compare the thermal properties and A two-stage reaction kinetics scheme consisting of two inde-
kinetic behaviour of coal, pine sawdust and their blends in an oxi- pendent reactions is proposed for the thermal decomposition of
dative atmosphere using a thermogravimetric analyzer in order to the biomass under an oxidative atmosphere (Liu et al., 2002; Shen
investigate the combustion characteristics of coal/biomass blends. et al., 2009), represented by the DTG curves. The kinetic scheme in-
cludes the two separate reactions:
2. Methods
A ðsolidÞ ! B ðcharÞ þ C1 ðgasÞ ðfirst stageÞ
2.1. Experimental setup B ðcharÞ ! C2 ðgasÞ þ D ðashÞ ðsecond stageÞ
The approach used in the present work to calculate the kinetic
The feedstock materials used in this work were a high-volatile parameters was based on the Arrhenius equation, which has been
bituminous coal and pine sawdust. The samples were collected used by other researchers in order to obtain kinetic parameters of
from the Aboño Power Plant located in Asturias (Spain), where thermal events under combustion conditions (Munir et al., 2009;
co-combustion trials are currently being conducted. Ultimate and Shen et al., 2009; Wang et al., 2009; Yorulmaz and Atimtay,
proximate analyses together with the heating values of the coal 2009). These two separate reactions are thought to be governed
and pine samples are presented in Table 1. Different mixtures of by the first-order Arrhenius law and so, the kinetics of the reaction
both materials were prepared. These included 5, 10, 20, 50 and is described as:
80 wt.% of pine sawdust (5P95C, 10P90C, 20P80C, 50P50C and
80P20C, respectively). The raw materials are named 100C (coal) dx=dt ¼ kf ðxÞ ð1Þ
and 100P (pine sawdust). First, the samples were ground and k ¼ A expðE=RTÞ ð2Þ
sieved in order to obtain a particle size fraction of 75–150 lm.
The coal/pine mixtures were mixed in appropriate proportions where f(x) represents the hypothetical model of the reaction mech-
and manually homogenised. anism, k is the reaction rate, A the pre-exponential factor (min1), E
The techniques employed in this study were thermogravimetric the activation energy (kJ mol1), R the gas constant (8.314 J K1
analysis (TG) and derivative thermogravimetry (DTG). Non-iso- mol1), T the absolute temperature (K), t the time (min) and x the
thermal TGA was performed using a Setaram TAG24 analyzer. loss in mass fraction or mass conversion ratio, which can be calcu-
The analyses were carried out under a 50 cm3 min1 air flow at a lated by the following relationship:
heating rate of 15 °C min1 from room temperature to 1000 °C. x ¼ ðm0  mt Þ=ðm0  mf Þ ð3Þ
Approximately 5 mg of sample was used for each experiment
and it was dispersed flatly on a crucible, which had a flat bottom where m0 is the initial mass of the sample, mt the mass of the sam-
8 mm in diameter and 3 mm in depth. A small amount of sample ple at time t and mf the final mass of the sample.

Table 1
Ultimate and proximate analyses and HHV of the samples.

Sample Ultimate analysis (wt.%, db) Proximate analysis (wt.%, db) HHV (MJ/kg, db)
C (%) H (%) N (%) Oa (%) S (%) Ash (%) FCa (%) VM (%)
Coal (100C) 69.3 4.2 1.8 8.9 0.8 15.0 55.1 29.9 27.8
Pine (100P) 44.1 5.9 0.7 45.5 0.0 3.8 16.4 79.8 18.9
a
Calculated by difference; db: dry basis.
 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608 5603

Table 2 tics favour clean combustion conditions (Vamvuka et al., 2003b).

Expressions of g(x) for the kinetic model functions usually employed for the solid- The nitrogen content in the biomass is low, and therefore, emis-
state reactions.
sions of nitrogen oxides from this element will probably be mini-
Mechanism and model g(x) mal. In comparison with coal, biomass contains a higher
Reaction order proportion of oxygen and hydrogen and less carbon (Table 1),
O1 ln(1  x) which reduces its heating value because the amount of energy con-
O2 (1  x)1 tained in carbon–oxygen and carbon-hydrogen bonds is lower than
O3 (1  x)2
in carbon–carbon bonds (Munir et al., 2009). However, the higher
Phase boundary controlled reaction oxygen content in the biomass indicates that it will have a higher
R2 1  (1  x)1/2
R3 1  (1  x)1/3
thermal reactivity than coal (Haykiri-Acma and Yaman, 2008).
Biomass typically has a volatile matter/fixed carbon (VM/FC) ra-
Diffusion
D1 x2
tio of >4.0, while the VM/FC ratio for coal is almost always <1.0
D2 (1  x)ln(1  x) + x (Tillman, 2000). In this work, the VM/FC ratio for the biomass is
D3 [1  (1  x)1/3]2 approximately 5.0 and for the coal 0.5. Thus, for biomass fuels,
D4 1  2x/3  (1  x)2/3 the predominant form of combustion will take place via the gas-
phase oxidation of the volatile species (Wang et al., 2009).
For a constant heating rate b (K min1) during combustion, The experimental DTG curves for the coal and pine sawdust
b = dT/dt, Eq. (1) can be transformed into: samples and their blends, under air atmosphere, are shown in
Fig. 1. In this figure, the curves of the blends are situated between
dx=f ðxÞ ¼ ðk=bÞdT ð4Þ those of the individual fuels. Before ignition, some increase in mass
Integrating Eq. (4) gives: may occur due to the chemisorption of oxygen (Haykiri-Acma and
Z Z Yaman, 2008), as can be observed from the negative values of the
x T
DTG curves of coal sample (100C) and the samples with a high per-
gðxÞ ¼ dx=f ðxÞ ¼ A=b expðE=RTÞdT ð5Þ
0 T0 centage of coal (5P95C, 10P90C and 20P80C).
From the DTG curves (Fig. 1), it can be seen that an initial mass
where g(x) is the integral function of conversion. loss (stage A) occurred between the temperatures of 25 °C and
Eq. (5) is integrated by using the Coats–Redfern method (Coats 105 °C for all samples, due to moisture evaporation. After that,
and Redfern, 1964), yielding: two-step mass losses (stages B and C) took place in biomass sam-
ln½gðxÞ=T 2  ¼ ln½AR=bEð1  2RT=EÞ  E=RT ð6Þ ple, 100P, compared to only a one-step mass loss (stage D) for coal
sample, 100C. Stage B would be due to the release of volatiles and
Since it can be demonstrated that for most values of E and for the their combustion, and stage C would be due to the char oxidation.
temperature range of combustion, the expression ln[AR/bE(1  2RT/ However, all the coal/biomass blends displayed three step mass
E)] in Eq. (6) is essentially constant (Zhou et al., 2006), if ln[g(x)/T2] losses (stages B, C and D). In the case of the blends, the first two
is plotted versus 1/T, a straight line should be obtained. Moreover, if mass losses, stages B and C, were mainly due to the burning of bio-
the correct g(x) is used, the plot of ln[g(x)/T2] against 1/T should give mass, whereas the last one (stage D) was mostly due to the com-
a straight line with a high correlation coefficient of linear regression bustion of coal. However, part of the mass loss is due to the coal
analysis, from which the values of E and A can be derived. The acti- because the second peak of the biomass (stage C) and the peak of
vation energy, E, can be calculated from the slope of the line, E/R; the coal (stage D) clearly overlap. Similarly, some of the mass loss
and by taking the temperature at which mt = (m0 + mf)/2 as the of stage D is due to the biomass.
intercept term of Eq. (6), the pre-exponential factor A can also be Table 3 shows the temperature ranges for the three different re-
calculated (Zhou et al., 2006). gions of mass loss after the initial loss of moisture during the com-
The function g(x), or f(x), depends on the mechanism controlling bustion of the samples, the mass loss produced in each of these
the reaction and the size and shape of the reacting particles. Table regions and the final residue after combustion. The initial temper-
2 shows the expressions of g(x) for the basic model functions usu- ature in stage B and the final temperature in stage D were taken as
ally employed for the kinetic study of solid-state reactions. By the temperature values at which the rate of mass loss was
means of these functions it was possible to estimate the reaction 0.005% s1 (Rubiera et al., 1999). The coal (100C) starts to devolat-
mechanisms governing the process of thermal oxidation of the ilise at a higher temperature compared to the biomass (100P),
samples from the TG curves. The form of g(x) that gives a straight 316 °C and 196 °C respectively. Stages C and D occur in a range
line with the highest correlation coefficient will be considered the of temperatures that are similar in all the samples, which suggests
function of the model that best represents the kinetics of mass loss that the blending process does not affect the combustion behav-
for each separate reaction. iour of the individual components. However, the initial tempera-
In most dynamic studies that use TGA, the first-order chemical ture of stage B decreased as the percentage of biomass increased,
reaction assumption (O1 model) is the most frequently used. In suggesting that combustion was brought forward due to the for-
addition to this reaction mechanism, other chemical reactions mation of volatiles from the biomass. Indeed, the biomass has
(O2 and O3 models), boundary controlled reactions (R2 and R3 much higher volatile matter content than the coal (Table 1), which
models) and diffusion mechanisms (D1, D2, D3 and D4 models) causes it to burn at lower temperatures.
are commonly applied to describe the combustion reactions of The mass loss from biomass (100P) is higher than in other sam-
biomass. ples in stage B, in the 200–350 °C temperature range (Table 3).
However, in stage D, above approximately 480 °C, the mass loss
3. Results and discussion from the biomass is over and the mass loss from coal (100C) is
now higher than from the blends. The mass losses corresponding
3.1. Thermal characteristics of the samples under air atmosphere to the blends lie between those of the individual fuels. The mass
loss in stage B increases with the biomass content due to the much
The biomass, pine sawdust (100P), has less fixed carbon and higher volatile matter content of the biomass (Table 1). Similarly,
more volatiles than the coal, 100C, and it also has a lower ash con- the mass loss in stage D increases as the coal content of the blend
tent and a negligible sulphur content (Table 1). These characteris- increases, due to the higher amount of char in the coal.
5604 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608

Fig. 1. Experimental and calculated DTG curves for biomass (100P), coal (100C) and blends (5P95C to 80P20C) in an air flow of 50 cm3 min1, at a heating rate of 15 °C min1.

Table 3
Temperature interval for different regions after loss of moisture, mass loss in these regions, and residues for all samples.

Sample Temperature interval (°C) Weight loss (%)

Stage B Stage C Stage D Stage B Stage C Stage D Residue (%)
100C – – 316–615 – – 82.8 15.7
5P95C 290–343 343–477 477–616 2.1 27.8 52.3 15.5
10P90C 273–353 353–481 481–617 4.6 28.6 48.8 15.3
20P80C 240–360 360–481 481–616 14.4 26.8 42.7 12.7
50P50C 215–365 365–487 487–606 34.4 26.9 23.4 10.3
80P20C 198–363 363–494 494–595 49.7 27.2 9.4 7.2
100P 196–364 364–487 – 63.9 25.1 – 3.2

C, coal; P, pine.

In the case of the biomass, the mass loss in stage B is due to oxi- samples. In stage D, which is due to the presence of coal, the max-
dative degradation – i.e. volatiles are released and then burned – imum rate of mass loss increases with the rise in the percentage of
whereas the mass loss in stage C is due to the combustion of char. coal in the blend, indicating that the higher the amount of coal in
Haykırı-Açma (2003) described this first stage as the burning re- the blend, the faster the mass loss in the 350–600 °C temperature
gion in which volatiles are released and burned. Zheng and Kozińs- range. This is due to the fact that coal has a higher carbon content
ki (2000) reported that biomass combustion consisted of two main (Table 1).
steps, the first characterized by the devolatilisation process and The temperature value at the maximum rate of mass loss is usu-
burning of the released light organic volatiles, and the second ally considered inversely proportional to the reactivity and com-
resulting from the oxidation of char. Liu et al. (2002) added that bustibility of the sample (Haykırı-Açma, 2003). In relation to the
the first DTG peak in air is largely due to the pyrolysis of hemicel- B and D stages, no significant variation is observed between the
lulose and cellulose, but also partly due to that of lignin, while the different samples (Table 4). However, in the C region, a decrease
second DTG peak is largely caused by oxidation. in the peak temperature can be observed with the increase in the
Table 4 shows the peak temperatures and the maximum rates
of mass loss for the three stages after the initial loss of moisture
during the combustion of all the samples. The maximum rate of Table 4
mass loss is considered directly proportional to the reactivity of Peak temperature and maximum rate of mass loss for all samples.
the sample (Zheng and Koziński, 2000). In stage B, the maximum
Sample Peak temperature (°C) Maximum rate of mass loss
mass loss rate increases with the percentage of biomass present (% min1)
in the blend, which suggests that the higher the amount of biomass
Stage B Stage C Stage D Stage B Stage C Stage D
in the blend, the faster the rate of mass loss between 200–350 °C
100C – – 507 – – 13.6
or, in other words, the higher the reactivity of the samples in this
5P95C 326 467 508 1.0 8.5 13.2
range. This indicates that a greater number of volatiles are formed 10P90C 323 467 507 1.7 8.5 12.4
and that they ignite at these temperatures due to the presence of 20P80C 323 464 510 4.7 6.9 10.5
biomass in the blends. In stage C, this is not the case because of 50P50C 323 464 510 10.6 5.6 5.8
the overlapping of the C and D stages. However, for biomass 80P20C 323 454 510 15.2 5.1 2.6
100P 323 447 – 20.1 5.5 –
(100P), the maximum rate of mass loss in stage C is lower than that
of stage B, a finding also reported by Ghaly et al. (1993) for straw C, coal; P, pine.
 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608 5605

Fig. 2. Plots of ln[g(x)/T2] against 1/T that gave the highest correlation coefficient values for all samples.
5606 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608

Table 5
Thermal kinetic results of all samples.

Sample Stage B Stage C Stage D

2 2
E (kJ mol 1
) A (min 1
) R E (kJ mol 1
) A (min 1
) R E (kJ mol1) A (min1) R2
Model O1
100C – – – 97.9 9.9E+05 0.9986
5P95C 227.1 6.5E+19 0.9973 115.8 1.2E+07 0.9934
10P90C 149.5 7.4E+12 0.9987 120.2 2.4E+07 0.9928
20P80C 111.4 3.6E+09 0.9978 120.0 2.2E+07 0.9921
50P50C 103.2 7.4E+08 0.9977 134.6 2.0E+08 0.9924
80P20C 102.0 6.1E+08 0.9981 157.0 6.1E+09 0.9936
100P 102.3 6.6E+08 0.9983 – – –
Model D3
100C – – –
5P95C 228.1 8.7E+14 0.9980
10P90C 238.6 4.4E+15 0.9981
20P80C 238.4 5.2E+15 0.9988
50P50C 224.1 6.3E+14 0.9986
80P20C 217.4 3.2E+14 0.9983
100P 236.1 1.5E+16 0.9986
Model D4
100C – – –
5P95C 223.1 5.6E+14 0.9980
10P90C 229.2 1.2E+15 0.9988
20P80C 229.5 1.4E+15 0.9992
50P50C 208.1 4.3E+13 0.9982
80P20C 195.5 9.4E+12 0.9993
100P 219.0 6.0E+14 0.9983

C, coal; P, pine.

percentage of pine sawdust in the blend, reflecting the higher reac- responsible for the oxidation of the samples under study. The ki-
tivity of the biomass. netic parameters were determined assuming single separate reac-
tions for a particular stage of thermal conversion. According to the
3.2. Interactions between the components of the blends DTG plot (Fig. 1), a single reaction could be used to describe the
coal (100C) combustion process, whereas for biomass (100P) and
No significant deviations were observed between the experi- the coal/biomass blends (5P95C to 80P20C), two or three indepen-
mental and theoretical DTG curves of the coal/sawdust blends, as dent reactions, respectively, are necessary. Thus, Eq. (6) was ap-
can be seen in Fig. 1. Therefore, according to the DTG results, no plied separately to each of the stages, and the conversion, x, was
interactions between the components of the blends occurred, recalculated for each reaction. The form of g(x) which gives a
reflecting the additive behaviour of the coal/biomass blends and straight line with the highest correlation coefficient will be consid-
the absence of synergetic effects during the combustion process. ered to be the function of the model that best represents the ki-
Several authors, however, have observed interactions between netic mass loss for each separate reaction. Fig. 2 shows the plots
the components of coal/biomass blends (Zhou et al., 2006; Skodras of ln[g(x)/T2] against 1/T that gave the highest correlation coeffi-
et al., 2007), while others have reported the additive behaviour of cient values for all samples. From the slope of each line, the values
coal and biomass blends (Biagini et al., 2002; Kastanaki et al., 2002; of E and A were obtained.
Vamvuka et al., 2003a,b; Sadhukhan et al., 2008). It should be The mechanisms that yielded the highest thermal kinetics cor-
pointed out, however, that none of these studies were carried out relation coefficient are shown in Table 5, where R2 represents their
under combustion conditions but during pyrolysis experiments. correlation coefficients. The high coefficient values indicate that
Fitzpatrick et al. (2009) studied the co-combustion of coal and pine the corresponding reaction model satisfactorily fitted the experi-
wood in a fixed bed combustor and they observed synergy in mental data. Table 5 also shows the values of activation energy,
organics emissions from the coal/pine blends, with lower emis- E, and pre-exponential factor, A, for the three stages (B, C and D)
sions than would be expected on an additive basis. However, it for all the samples, each of which was obtained employing the
has to be considered that in fixed bed combustion, the particles most suitable model.
are larger and they would be close together and synergistic effects The solid-state reaction follows the first-order kinetics model
would be therefore observed. (O1) when the rate-determining step is the chemical reaction. In
In this work, the absence of any interaction indicates that the phase boundary controlled reactions, the reaction is controlled by
combustion reactions of biomass or coal are not significantly af- the movement of an interface at constant velocity and the reaction
fected by the presence of coal or biomass respectively. Each com- occurs almost instantaneously, with the result that the surface of
ponent of the mixture behaves independently and does not each particle is covered with a layer of the product. R2 is a function
interact with the other materials in the experiments carried out used for a circular disc reacting from the edge inward, whereas R3
with blends of pine sawdust and coal under the conditions estab- is used for a sphere which reacts from the surface inward. This
lished in this study. mechanism is sometimes assumed to be the governing conversion
model in the combustion of some carbonaceous materials (López-
3.3. Kinetic parameters Fonseca et al., 2006). In a diffusion-controlled reaction, D1 is the
function for a one-dimensional diffusion process governed by a
Several solid-state mechanisms (Table 2) were tested by the parabolic law, with a constant diffusion coefficient. For diffusion
Coats–Redfern method in order to determine the mechanisms in cylinders or spheres, it is necessary that all three dimensions
 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608 5607

be taken into account. D2 is the function for a two-dimensional dif- culated for the blends with a coal content of more than 50%,
fusion-controlled process in a cylinder. D3 is Jander’s equation for increasing as the coal percentage increased. This shows that the
diffusion-controlled solid-state reaction kinetics in a sphere, where addition of pine sawdust to coal facilitates the volatilization and
diffusion in all three directions is all-important. D4 is Ginstling– gaseous phase combustion processes at low temperatures, since
Brounshtein’s equation for a diffusion-controlled reaction starting lower activation energies are sufficient for the processes to devel-
from the outside of a spherical particle (Alshehri et al., 2000). In op, which is in agreement with the decrease in the initial temper-
a diffusion-controlled reaction, numerous chemical reactions or ature of stage B as the percentage of biomass increases. Reactions
micro-structural changes in solids take place through solid-state with a high activation energy require a high temperature or a long-
diffusion, i.e., the movement and transport of gas molecules in er reaction time (Lázaro et al., 1998). The blends with 50 and
the solid phase (Yorulmaz and Atimtay, 2009). 80 wt.% of biomass (50P50C and 80P20C) presented similar E and
For stage B, models O2, O3 and D1 showed correlation coeffi- A values to that of the pure biomass (100P). Thus, in these cases
cients between 0.9186 and 0.9890 for all the samples studied (data the presence of coal does not significantly affect combustion dur-
not shown). On the other hand, models R2, R3, D2, D3 and D4 had ing this first stage. When the coal percentage is 80% (20P80C),
higher correlation coefficients, between 0.9879 and 0.9968 the influence of coal during stage B is small, since the values of E
(data not shown). However, model O1 showed the highest correla- and A are only slightly higher. But when the percentage of coal is
tion coefficients for all the samples with values exceeding 0.9973 higher than 80% (5P95C and 10P90C), significantly greater values
(Table 5). of E and A are observed in this first stage.
For stage C, models O2 and O3 showed correlation coefficients For stage D, where only coal is present, the values of E and A
between 0.8738 and 0.9437 for all the samples studied. Model (model O1) corresponding to the coal sample (100C) were lower
D1 presented even higher correlation coefficients, between than all the other values (Table 5). However, these values increased
0.9820 and 0.9942, while models O1, R2, R3 and D2 exhibited cor- with the increase in biomass content. When the percentage of bio-
relation coefficients between 0.9940 and 0.9978. None of these mass is 5–20% (5P95C, 10P90C and 20P80C), the values of E and A
models or their results has been included in Table 5. However, are only slightly higher than those of coal. However, when the per-
models D3 and D4 had the highest correlation coefficients, with centage of biomass is higher, greater values of E and A are
values higher than 0.9980 (Table 5). observed.
Finally, for stage D, models O2 and O3 showed correlation coef- For stage C, the calculated values of E and A (models D3 and D4)
ficients between 0.9178 and 0.9511 for all the samples studied were found to be consistently higher than those of the other stages
(data not shown). Models R2, R3, D1, D2, D3 and D4 displayed cor- (Table 5), but fairly similar to each other, both in the case of the
relation coefficients between 0.9427 and 0.9723 for all the samples biomass and the blends. This indicates that the char combustion
of coal/biomass blends (data not shown). For the coal sample stage requires a higher activation energy than the devolatilisation
(100C), all the models, except O2 and O3, presented correlation stage.
coefficients higher than 0.99 (data not shown), with model O1
showing the highest value, equal to 0.9986 for this sample (Table
4. Conclusions
5). Model O1 also presented the highest correlation coefficients
for the coal/biomass blends, >0.9921 (Table 5).
Pine sawdust was subjected to two combustion steps: between
Thus, the results confirm that the chemical first order reaction
200 and 360 °C, when the volatiles were released and burned, and
(O1 model) is the most effective solid-state mechanism for the first
between 360 and 490 °C, when char combustion occurred. How-
step of biomass oxidation (stage B) and for coal combustion (stage
ever, coal was subjected to only one combustion step: between
D). However, diffusion mechanisms (D3 and D4) were found to be
315 and 615 °C. The coal/biomass blends experienced all three
responsible for the second step of biomass combustion (stage C).
combustion steps. When the biomass percentage in the blend
The use of thermal analysis in order to find the mechanism
was 50 wt.% or more, devolatilisation was the predominant pro-
responsible for the oxidation process may result in more than
cess. No synergistic effect during coal and biomass co-combustion
one equation fitting the experimental results. Therefore, a combi-
was observed in these experiments. The combustion process of
nation of TG analysis with dynamic and isothermal studies could
coal consists of one reaction, whereas, for biomass and the coal/
be employed to determine the exact mechanisms and thermal con-
biomass blends, the process consists of two or three independent
stants of the oxidation processes (Yorulmaz and Atimtay, 2009).
reactions, respectively. The results of the kinetic analysis showed
As regards the kinetic parameters, the pre-exponential factor, A,
that the O1 mechanism (first-order chemical reaction) was as-
is more closely related with material structure, whereas the reactiv-
sumed to be the main mechanism responsible for the first stage
ity of samples is determined by the activation energy, E (Yorulmaz
of biomass oxidation and for coal combustion. On the other hand,
and Atimtay, 2009). The activation energy, E, in the first stage (stage
D3 and D4 diffusion mechanisms were observed to be the control-
B) of the biomass sample (100P) showed a value equal to
ling mechanisms in the second stage of biomass combustion.
102 kJ mol1 (Table 5), which is similar to the values recorded by
Shen et al. (2009) in the first stage of the oxidation of their biomass
samples (104–125 kJ mol1). These authors found higher values for Acknowledgements
E in the second stage of the oxidation process under low heating
rates (150–220 kJ mol1). In the present study, the activation en- Work carried out with financial support from the Spanish MIC-
ergy in stage C for the biomass (100P) was 219–236 kJ mol1. The INN (Project PS-120000-2006-3, ECOCOMBOS), and co-financed by
E values observed by Liu et al. (2002) reached 52–99 kJ mol1 for the European Regional Development Fund, ERDF.
the first stage of oxidation and 87–202 kJ mol1 for the second stage
with leaf and wood samples. For beech and Douglas fir, Branca and
References
Di Blasi (2004) found global E values ranging between 106 and
226 kJ mol1. Wang et al. (2009) found values of 88–115 kJ mol1 Alshehri, S.M., Monshi, M.A.S., Abd El-Salam, N.M., Mahfouz, R.M., 2000. Kinetics of
for the first stage of oxidation and 153–210 kJ mol1 for the second the thermal decomposition of c-irradiated cobaltous acetate. Thermochim. Acta
stage with wheat straw and coal and wheat straw mixtures. 363, 61–70.
Arenillas, A., Rubiera, F., Arias, B., Pis, J.J., Faúndez, J.M., Gordon, A.L., García, X.A.,
Table 5 shows that for stage B, the calculated values of E and A 2004. A TG/DTA study on the effect of coal blending on ignition behaviour. J.
(model O1) for biomass sample (100P) were lower than those cal- Therm. Anal. Calorim. 76, 603–614.
5608 M.V. Gil et al. / Bioresource Technology 101 (2010) 5601–5608

Biagini, E., Lippi, F., Petarca, L., Tognotti, L., 2002. Devolatilization rate of biomasses Munir, S., Daood, S.S., Nimmo, W., Cunliffe, A.M., Gibbs, B.M., 2009. Thermal analysis
and coal–biomass blends: an experimental investigation. Fuel 81, 1041–1050. and devolatilization kinetics of cotton stalk, sugar cane bagasse and shea meal
Bilbao, R., Mastral, J.F., Aldea, M.E., Ceamanos, J., 1997. Kinetic study for the thermal under nitrogen and air atmospheres. Bioresour. Technol. 100, 1413–1418.
decomposition of cellulose and pine sawdust in an air atmosphere. J. Anal. Appl. Pis, J.J., de la Puente, G., Fuente, E., Morán, A., Rubiera, F., 1996. A study of the self-
Pyrol. 39, 53–64. heating of fresh and oxidized coals by differential thermal analysis.
Branca, C., Di Blasi, C., 2004. Global intrinsic kinetics of wood oxidation. Fuel 83, 81– Thermochim. Acta 279, 93–101.
87. Rubiera, F., Morán, A., Martínez, O., Fuente, E., Pis, J.J., 1997. Influence of biological
Cai, J., Wang, Y., Zhou, L., Huang, Q., 2008. Thermogravimetric analysis and kinetics desulphurisation on coal combustion performance. Fuel Process. Technol. 52,
of coal/plastic blends during co-pyrolysis in nitrogen atmosphere. Fuel Process. 165–173.
Technol. 89, 21–27. Rubiera, F., Arenillas, A., Fuente, E., Miles, N., Pis, J.J., 1999. Effect of the grinding
Coats, A.W., Redfern, J.P., 1964. Kinetic parameters from thermogravimetric data. behaviour of coal blends on coal utilisation for combustion. Powder Technol.
Nature 201, 68–69. 105, 351–356.
Cumming, J.W., McLaughlin, J., 1982. The thermogravimetric behaviour of coal. Rubiera, F., Arenillas, A., Arias, B., Pis, J.J., 2002. Modification of combustion
Thermochim. Acta 57, 253–272. behaviour and NO emissions by coal blending. Fuel Process. Technol. 77–78,
Fitzpatrick, E.M., Bartle, K.D., Kubacki, M.L., Jones, J.M., Pourkashanian, M., Ross, A.B., 111–117.
Williams, A., Kubica, K., 2009. The mechanism of the formation of soot and other Sadhukhan, A.K., Gupta, P., Goyal, T., Saha, R.K., 2008. Modelling of pyrolysis of coal–
pollutants during the co-firing of coal and pine wood in a fixed bed combustor. biomass blends using thermogravimetric analysis. Bioresour. Technol. 99,
Fuel 88, 2409–2417. 8022–8026.
Ghaly, A.E., Ergüdenler, A., Al Taweel, A.M., 1993. Determination of the kinetic Safi, M.J., Mishra, I.M., Prasad, B., 2004. Global degradation kinetics of pine needles
parameters of oat straw using thermogravimetric analysis. Biomass Bioenerg. 5, in air. Thermochim. Acta 412, 155–162.
457–465. Shen, D.K., Gu, S., Luo, K.H., Bridgwater, A.V., Fang, M.X., 2009. Kinetic study on
Haykırı-Açma, H., 2003. Combustion characteristics of different biomass materials. thermal decomposition of woods in oxidative environment. Fuel 88, 1024–
Energ. Convers. Manage. 44, 155–162. 1030.
Haykiri-Acma, H., Yaman, S., 2008. Effect of co-combustion on the burnout of Skodras, G., Grammelis, P., Basinas, P., 2007. Pyrolysis and combustion behaviour of
lignite/biomass blends: a Turkish case study. Waste Manage. 28, 2077–2084. coal–MBM blends. Bioresour. Technol. 98, 1–8.
Heikkinen, J.M., Hordijk, J.C., De Jong, W., Splithoff, H., 2004. Thermogravimetry as a Tillman, D.A., 2000. Biomass coffering: the technology, the experience, the
tool to classify waste components to be used for energy generation. J. Anal. combustion consequences. Biomass Bioenerg. 19, 365–384.
Appl. Pyrol. 71, 883–900. Vamvuka, D., Kakaras, E., Kastanaki, E., Grammelis, P., 2003a. Pyrolysis
Kastanaki, E., Vamvuka, D., 2006. A comparative reactivity and kinetic study on the characteristics and kinetics of biomass residuals mixtures with lignite. Fuel
combustion of coal–biomass char blends. Fuel 85, 1186–1193. 82, 1949–1960.
Kastanaki, E., Vamvuka, D., Grammelis, P., Kakaras, E., 2002. Thermogravimetric Vamvuka, D., Pasadakis, N., Kastanaki, E., Grammelis, P., Kakaras, E., 2003b. Kinetic
studies of the behavior of lignite–biomass blends during devolatilization. Fuel modelling of coal/agricultural by-product blends. Energ. Fuel 17, 549–558.
Process. Technol. 77–78, 159–166. Wang, C., Wang, F., Yang, Q., Liang, R., 2009. Thermogravimetric studies of the
Lázaro, M.J., Moliner, R., Suelves, I., 1998. Non-isothermal versus isothermal behavior of wheat straw with added coal during combustion. Biomass Bioenerg.
technique to evaluate kinetic parameters of coal pyrolysis. J. Anal. Appl. Pyrol. 33, 50–56.
47, 111–125. Yorulmaz, S.Y., Atimtay, A.T., 2009. Investigation of combustion kinetics of treated
Liu, N.A., Fan, W., Dobashi, R., Huang, L., 2002. Kinetic modelling of thermal and untreated waste wood samples with thermogravimetric analysis. Fuel
decomposition of natural cellulosic materials in air atmosphere. J. Anal. Appl. Process. Technol. 90, 939–946.
Pyrol. 63, 303–325. Zheng, J.A., Koziński, J.A., 2000. Thermal events occurring during the combustion of
López-Fonseca, R., Landa, I., Elizundia, U., Gutiérrez-Ortiz, M.A., González-Velasco, biomass residue. Fuel 79, 181–192.
J.R., 2006. Thermokinetic modelling of the combustion of carbonaceous Zhou, L., Wang, Y., Huang, Q., Cai, J., 2006. Thermogravimetric characteristics and
particulate matter. Combust. Flame 144, 398–406. kinetic of plastic and biomass blends co-pyrolysis. Fuel Process. Technol. 87,
McKendry, P., 2002. Energy production from biomass (part 1): overview of biomass. 963–969.
Bioresour. Technol. 83, 37–46.