Thermochimica Acta 618 (2015) 36–47

Contents lists available at ScienceDirect

Thermochimica Acta
journal homepage: www.elsevier.com/locate/tca

Synergism among biomass building blocks? Evolved gas and kinetics

analysis of starch and cellulose co-pyrolysis
Junjie Xue a,b , Selim Ceylan c , Jillian L. Goldfarb a,d,∗
a
Department of Mechanical Engineering, Boston University, 110 Cummington Mall, Boston, MA 02215, United States
b
College of Engineering, China Agricultural University, Beijing 100083, People’s Republic of China
c
Ondokuz Mayıs University, Faculty of Engineering, Chemical Engineering Department, 55139 Kurupelit, Samsun, Turkey
d
Division of Materials Science and Engineering, Boston University, 15 St. Mary’s Street, Brookline, MA 02446, United States

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

Article history: Debate surrounds biomass co-pyrolysis: can thermal decomposition be modeled as the sum of individ-
Received 19 June 2015 ual components, or do synergistic reactions promote or hinder devolatilization? Activation energies of
Received in revised form 31 August 2015 mixtures of starch and cellulose pyrolyzed at 10, 50 and 100 K/min were determined via the distributed
Accepted 2 September 2015
activation energy model. Reaction kinetics suggest that blending may promote devolatilization, seen
Available online 6 September 2015
through lower activation energies. Yet, evolved gas analysis shows no evidence of synergism as a result
of blending, at least at lower temperatures. As the percentage of cellulose increases, the temperature at
Keywords:
which the peak mass loss rate occurs and peak evolved gases emerge are linearly related. As such, there is
Pyrolysis
Biomass little evidence of chemical reaction synergism during the pyrolysis of these two biomass building blocks,
Cellulose but rather synergistic behavior is perhaps a result of the starch physically promoting the devolatilization
Starch of cellulose at lower temperatures when present in larger quantities.
Synergy © 2015 Elsevier B.V. All rights reserved.
Evolved gas
Activation energy
Distributed activation energy model

1. Introduction A number of studies demonstrate the ability to alter bio-oil

and pyrolysis gas composition and yield from pyrolysis, the ther-
The 2007 United States Independence and Security Act man- mal decomposition in the absence of oxygen, by changing reaction
dates that 16 billion gallons of cellulosic biofuel be blended into temperatures, pressures and heating rates [3,4] as well as explo-
traditional transportation fuels by 2022, a portion of which must iting reaction synergy among solid fuels to tailor the properties
be biodiesel produced from biomass [1]. The widespread use of of the pyrolysis products by co-pyrolyzing biomasses [5]. This is
biomass as a renewable feedstock for the production of liquid critically important as we move forward with industrial scale pro-
fuels and syngas depends on several factors. For the pyrolytic duction of bio-oils via thermal treatment. However, we require a
conversion of solids, these include the ability to design appro- deeper understanding as to the nature of these synergistic reac-
priate reaction systems, insure adequate supplies of feedstocks, tions. For example, does the physical compilation of solid fuels
standardize biomass-derived pyrolysis liquid products and, per- that devolatilize at lower temperatures than others cause changes
haps most importantly, develop a comprehensive understanding in overall reaction rates and kinetics? Does blending fuel streams
of the thermochemical pathways underlying the decomposition of change the pyrolysis gas compositions evolving from solid blend
solid biomass to liquid and gaseous fuels [2]. The ability to pre- pyrolysis? Can we predict the activation energies required to
dict the behavior of mixed biomass streams during co-pyrolysis pyrolyze blended fuel streams from knowledge of their pure com-
is imperative to insuring successful large-scale implementation of ponent characteristics, or does synergy impact activation energies?
thermochemical conversion of biomass to biofuels. These and other questions must be answered on a fundamen-
tal level to better design industrial pyrolysis units and determine
optimal feedstock blends. For that reason, here we investigate
the kinetics and evolving gas compositions of two fundamental
∗ Corresponding author at: Department of Mechanical Engineering, Boston Uni-
biomass building blocks with the same chemical formula but dif-
versity, 110 Cummington Mall, Boston, MA 02215, United States.
E-mail addresses: [email protected], [email protected] (J.L. Goldfarb). ferent molecular geometry: starch and cellulose.

http://dx.doi.org/10.1016/j.tca.2015.09.002
0040-6031/© 2015 Elsevier B.V. All rights reserved.
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 37

There is discord in the literature about the nature of blended Table 1

Mixtures of starch and cellulose used in pyrolysis experiments.
solid fuel pyrolysis; some find that the yields and activation
energies of co-pyrolyzed fuels are linearly proportional to the Sample Cellulose Starch Mixture 1 Mixture 2 Mixture 3
contributions of the individual components [6,7]. Others detect Cellulose mass fraction 1.0 0 0.75 0.50 0.25
non-additive compositions and activation energies across heat- Starch mass fraction 0 1.0 0.25 0.50 0.75
ing rates and temperature ranges [8–10]. And yet still others,
our group included, find both additive and synergistic behavior
for the same samples, such as the peak mass loss reaction rates as received. Approximately 1 g of each mixture was fabricated by
found via derivative thermogravimetric curves displaying additive weighing the desired amount of cellulose and starch on a Sartorius
behavior (linearly increasing reaction rate as biomass percentage semi-microbalance to ±0.1 mg and placed into a clean glass vial, as
in coal–biomass blends increases) whereas the activation energy given in Table 1. Contents were homogenized by placing the vials on
and evolved gas compounds of the same blends may display syner- a vortex mixer for several minutes. Samples were stored at ambient
gistic behavior [11,12]. However, there is scant information in the conditions until used.
literature concerning whether varying the composition of blends
can induce reaction synergism. A survey of singular source biomass 2.2. Thermogravimetric analysis-differential scanning
components (i.e. feed corn stover, nut shells, wood samples) indi- calorimetry (TGA–DSC)
cates that this would be the case; each of these biomasses is
comprised of different ratios of cellulose, hemicellulose, lignin, Each pure solid fuel and blend was pyrolyzed in a high purity
starch, proteins, etc., and each displays a different thermal decom- nitrogen atmosphere (reactive + protective gas flow at 70 mL/min)
position profile. Despite this accepted generalizability, the specific in a 70 ␮L alumina crucible using a Mettler Toledo TGA/DSC1 with
contribution of different biomass constituents to the overall kinetic gas flow control, data output to the Mettler STARe Default DB
behavior has yet to be explored. V10.00 software. The DSC was calibrated with indium standard
Cellulose and starch are polymers of glucose, and represent the (Mettler Toledo) at 10 K/min. The mass is measured every second
carbohydrate building blocks of biomass [13]. They have the same to the 10−9 g, along with temperature to ±0.1 K. All samples were
unit molecular formula [C6 H10 O5 ]n , though different molecular ori- heated to 383 K and held for 15 min to insure moisture removal.
entations. Starch is comprised of repeating glucose units oriented Then the sample was heated at 10, 50 or 100 K/min up to 1173 K
in the same direction (alpha linkages), whereas in cellulose, neigh- and held for 30 min to obtain a stable mass reading. Every sam-
boring units are rotated 180◦ around the axis of the polymer chain ple was run at 10 K/min three times to insure reproducibility; each
backbone (beta linkages). The hydroxyl group attached to carbon- sample at 50 K and 100 K/min was run once, with random samples
1 is below the plane of the ring in starch, and above for cellulose. run a second and third time to insure reproducibility.
Hydrogen bonding in the ␤-linked polymer is what lends cellu- There are multiple methods available to analyze the pyroly-
lose its structural advantage, yielding a strong, fibrous nature, as sis kinetics of solid carbonaceous fuels. Many are based off of the
compared to starch, which solubilizes fairly easily in water. There Arrhenius equation, expressed in the general form as:
are a number of studies in the literature that describe the ther- k = Ae−E/RT (1)
mal decomposition of pure starch [14–16] and pure cellulose [17];
we could not locate any that probe the potential synergistic rela- where A is the frequency (or pre-exponential) factor, E the acti-
tionship that occurs among blends of these compounds. Starch vation energy, T the absolute temperature, R the universal gas
and cellulose are a model system to probe the potential synergism constant, and k is the reaction rate constant. It is often assumed that
between biomass constituents; they have the same composition the thermal decomposition of carbonaceous fuels such as biomass
but vastly different structural characteristics and known decompo- occurs as an infinitely large set of first order reactions, allowing for
sition profiles. They are both known to thermally decompose via a the calculation of an overall, or apparent activation energy assum-
dehydration → depolymerization → devolatilization pathway, but ing an overall, or apparent, first order reaction. Nonisothermal TGA
these reactions occur at different temperatures and rates for each data are transformed by defining the extent of conversion, x(t), as
material [15,17]. As such, we query whether or not the decom- a function of initial mass, mi , final mass, mf , and mass at any time
position of starch can synergistically impact the onset of cellulose t, mt :
decomposition. That is, if we note synergism between these two mi − mt
x(t) = (2)
components, then the starch may promote decomposition of the mi − mf
hydrogen-bonded cellulose, which would indicate the possibility
A large portion of the biomass pyrolysis literature calculates the
of producing pyrolysis bio-oil and syngas at lower temperatures by
activation energy using the reaction rate constant method (RRCM)
blending biomass materials with lower energy barriers to decom-
(see, for example [12,23–25]). In this case, the rate of material
position in the overall raw material mixture. However, the results
reacted at any given time is expressed as a function of the rate
presented herein apply beyond bio-fuel production; as we seek new
constant:
sources of renewable materials such as thermoplastics and biopoly-
mers, knowledge of how blended biomass building blocks such as dx(t)
= k(1 − x(t)) (3)
starch and cellulose behave under various thermal environments dt
may assist in renewable materials design [18–22]. A plot of ln k versus 1/T, often referred to as an Arrhenius plot,
will yield a straight line with a slope of E/R, if the reaction pro-
ceeds via an overall first order. Many biomasses, when subjected
2. Materials and methods to analysis via the RRCM, show multiple devolatilization regimes –
abrupt changes in slope of the Arrhenius plot at temperatures spe-
2.1. Materials cific to a given biomass – with each region having its own activation
energy. Dozens of biomass pyrolysis studies in the literature show
Microcrystalline Cellulose (CAS: 9004-34-6) was purchased a reaction order of approximately 1; this assumption is commonly
from Fisher Scientific, supplied by Alfa Aesar, Lot #10179415. Sol- applied to account for the simultaneous reactions [26,27].
uble Starch (CAS: 9005-84-9) was supplied by Fisher Scientific, However, the RRCM fails to capture the entire range of decom-
Lot #136971 as Certified ACS Regent Grade. Samples were used position; for biomass pyrolysis we often see multiple mass loss
38 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47

regimes over different temperature ranges accounting for the erroneous, as A depends on the collision frequency of molecules
stage-wise decomposition of the primary biomass constituents. during a chemical reaction – that is, it is contingent on the number
That is, the Arrhenius plots have discrete changes in slope that of molecules present in a control volume – and the success of the
occur over different (but similar) temperature ranges for each collisions between those molecules to result in a reaction. As the
biomass. For example, using the RRCM our laboratory found three amount of molecules participating in pyrolytic reactions increase
primary decomposition regions for the pyrolysis of cabbage palm with temperature and yet decrease as volatiles are lost from the
(Sabal palmetto), likely corresponding to devolatilization of primar- solid, it would seem improbable to expect a constant frequency
ily hemicellulose, cellulose and lignin [28]. However, we know that factor across the range of conversion. In this analysis we determine
lignin can devolatilize over a broad temperature range [29], which if synergism exists between these two biomass building blocks,
the limited “mass loss regime” approach of the RRCM cannot eluci- by comparing activation energies obtained using the DAEM across
date. In addition, there can be substantial (between 2 and 20%) mass both models and mixture sample sets.
lost between one “regime” and another depending on the temper-
ature ramp rate and the temperature range selected for analysis, 2.3. Evolved compound evolution
which is not captured in the RRCM analysis. Selecting a tempera-
ture range that encapsulates the most mass loss will decrease the To improve the reliability of the evolved compound measure-
linearity of the mass loss regime segment due to curvature of the ments, we needed to increase the sample size beyond the maximum
line at tail ends, hence the inability to completely capture the entire of ∼15 mg possible in the TGA to increase the signal to noise ratio of
conversion. Another drawback of the RRCM is the dependence of the quadrupole mass spectrometer (MS). As such, the biomass was
the activation energy on the temperature ramp rate. At lower tem- pyrolyzed in a 1-in. diameter quartz tube furnace (Lindberg/Blue
perature mass loss regimes, we find higher activation energies at M® -Mini-Mite) under high purity nitrogen at a flow rate of approx-
slower temperature ramp rates; the opposite occurs at high tem- imately 105 ± 5 mL/min, monitored by a digital Omega flowmeter.
perature mass loss regimes [28]. Though the differences are often Approximately 0.45 g of sample was placed in a porcelain boat at
less than 10%, this underscores the importance of noting that acti- the center of the hot zone of the furnace, reaching the same relative
vation energies calculated in this manner are likely subject to heat height in the porcelain boat as in the TGA’s crucible. A tempera-
and/or mass transfer limitations, and are thus overall or apparent ture ramp rate of 10 K/min was used to avoid any diffusion effects
activation energies at a given heating rate and particle size. For arising from heat transfer and to mirror the TGA experiments. Sam-
these, and other reasons discussed by the ICTAC Kinetics Commit- ples were initially heated at 393 K for 30 min to insure moisture
tee, isoconversional methods that use at least three temperatures removal, and then at 10 K/min up to the final TGA temperature
or heating rates, such as the distributed activation energy model, of 1173 K. The exhaust was sampled via a fused silica capillary
are recommended [30]. tubing (40 ␮m inside diameter) connected to a Quadrupole Mass
A commonality among the many methods used to determine Spectrometer (Extorr XT Series RGA, XT300M) via a turbomolecu-
activation energies is the assumption that the countless reactions lar pump at 10−7 Torr. The mass spectrometer was in the electron
occurring during pyrolysis are all irreversible first order parallel impact (EI) ionization mode at the electron energy of 70 eV and
reactions with different activation energies that occur simulta- provided mass spectra up to 300 amu. Mass spectrometers are com-
neously, which can be represented by a distribution function, f(E). monly connected to a TGA to measure evolved gases from pyrolysis
The literature is replete with examples of the distributed activation [43]. It is recognized that the gas volume from biomass pyrolysis
energy model (DAEM) as applied to pure solid fuel pyrolysis, includ- is quite limited in the TGA, such that packed bed experiments may
ing that of various biomass, coal and other carbonaceous fuels (see, present a better picture of total compound evolution [44]. With the
for example [31–36]). The distribution function routinely assumes tube furnace we were able to use larger samples than in the TGA,
the form [37,38]: and were able to see compounds evolving at lower relative amounts
⎛ ⎞ than the small TGA sample sizes allow.
∞ t  
E
X(t) = 1 − exp ⎝−A exp − dt ⎠ f (E)dE (4) 3. Results and discussion
RT
0 0
To probe the pyrolysis behavior of cellulose, starch, and their
If the experiment is conducted nonisothermally at a constant tem- blends, two primary biomass components with the same chemical
perature ramp rate, ˇ = dT/dt, Eq. (4) can be rewritten as: formula but different structures, we employed thermogravimetric
⎛ ⎞
∞ t  
analysis in combination with evolved gas analysis. If these two com-
A E pounds decompose in an additive manner, we would expect to see
X(t) = 1 − exp ⎝− exp − dT ⎠ f (E)dE (5)
ˇ RT a linear relationship between the composition (mass and molar,
0 0 given that they have equal molecular weights) and activation
energy; peak thermal decomposition mass loss rates and temper-
It is commonly assumed that the activation energy is normally
atures and evolved gas compounds. Conversely, if the compounds
distributed with average activation energy, Ea and a standard devi-
react synergistically, promoting or inhibiting thermal decomposi-
ation, , whereby:
tion, we would see non-additive trends in the pyrolysis kinetics. If

1 (E − Ea )2 there are chemical reaction synergies between the two compounds,

f (E) = √ exp − (6) we might encounter variations in non-condensable gases in terms
 2 2 2
of peak evolution temperatures and concentrations.
Non-Gaussian distributions are sometimes used to represent f(E),
including Weibull [39], Gamma [40], and Maxwell–Boltzmann [41] 3.1. Peak devolatilization temperatures and rates (DTG curves)
distributions.
The frequency factor is often considered to be a constant for Throughout the literature we note a common trend among the
all reactions; Miura and Maki [42] allow for a compensation effect devolatilizaiton behavior of many solid fuel blends such as coal and
between A and E through their Integral method, though Cai et al. biomass: namely, the peak mass loss rates, as determined through
[38] question the accuracies of the estimations of A and f(E). How- derivative thermogravimetric (DTG) curves, are often an additive
ever, it stands to reason that a constant frequency factor is also combination of the amount of each fuel present [11,45], whereby
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 39

Table 2
Derivative thermogravimetric analysis (Peak DTG temperature and mass conversion rate) of cellulose, starch and blends at 10, 50 and 100 K/min.

10 K/min 50 K/min 100 K/min

DTG Peak 1 DTG Peak 2 DTG peak DTG peak

Temp Rate Temp Rate Temp Rate Temp Rate

K s−1 K s−1 K s−1 K s−1

Cellulose 613 5.35E−03 655 4.11E−02 672 1.04E−01

Starch 574 7.06E−03 595 9.83E−02 600 1.89E−01
Mixture 1 612 4.52E−03 655 3.83E−02 660 7.74E−02
Mixture 2 586 3.00E−03 611 3.84E−03 609 3.53E−02 634 7.90E−02
Mixture 3 577 5.19E−03 601 1.58E−03 602 9.11E−02 597 1.33E−01

the peak mass loss rate is linearly proportional to the amount of 5.35 × 10−3 s−1 at 340 ◦ C (613 K), agreeing with cellulose decom-
each component. However, it was unclear if such a trend were to position values in the literature [17,46]. As expected – given the
hold for two pure compounds such as starch and cellulose. Table 2 absence of strong hydrogen bonding – the starch decomposes over
presents the analysis of the DTG curves for the coal–biomass blends a lower temperature range, from 269 ◦ C to 345 ◦ C (542–618 K), and
investigated each at 10, 50 and 100 K/min. Fig. 1 presents the DTG there is one DTG peaks at 574 K corresponding to a conversion rate
curves of cellulose, starch and their blends pyrolyzed at these three of 7.06 × 10−4 s−1 . These are also in good accord with the literature
heating rates. [5]. At 10 K/min the peak conversion rate is within the same order
For pyrolysis at 10 K/min, cellulose decomposes between 290 of magnitude for the cellulose and blends, and is an order of mag-
and 387 ◦ C (563–660 K) with a peak mass conversion rate of nitude lower for starch. Though not a linearly additive relationship,

Fig. 1. Derivative thermogravimetric (DTG) curves for cellulose ( ), starch ( ), and blends Mixture 1: 75 wt% cellulose ( ); Mixture 2: 50 wt% cellulose (
); Mixture 3: 25 wt% cellulose ( ), pyrolyzed at 10, 50 and 100 K/min.
40 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47

Fig. 2. Peak DTG temperature (a) and conversion rate (b) for pyrolysis of cellulose, starch and blends as a function of mass fraction cellulose at 10 K/min (䊉); 50 K/min ( );
100 K/min ( ) (for peak 1 only).

the peak pyrolysis temperature does increase as the percent of cel-

lulose in the mixture increases, as seen in Fig. 2. For example, in
the 50:50 mixture, there are two DTG peaks observed. The first is
at 586 K – almost 12 K higher than pure starch, but 27 K lower than
pure cellulose; the second peak is at 611 K – the same temperature
as the cellulose peak. This suggests that the cellulose is slowing
the devolatlizaiton of the starch in the mixture at this composi-
tion while it itself is not impacted. Conversely, in Mixture 3 (75 wt%
starch), the first peak occurs a mere 3 K higher than the pure starch,
and second peak at 601 K – suggesting that the starch devolatliza-
tion is not impacted by the presence of cellulose, yet the cellulose
decomposes at a lower temperature, though same order of magni-
tude in rate, as compared to pure cellulose. Finally, when cellulose
is the dominant mixture component, we see only one DTG peak at
roughly the same temperature and rate as the pure cellulose, sug-
gesting that the decomposition kinetics of the cellulose control the
blends’ overall decomposition.
When we increase the heating rate to 50 K/min, cellulose
decomposes from 308 ◦ C to 439 ◦ C (581–712 K), and there is a single
DTG peak observed at 382 ◦ C (655 K) with a rate of 4.11 × 10−2 s−1 .
For the starch, the pyrolysis process mainly occurs in the tem-
Fig. 3. Combined DTG/DSC plot for cellulose, starch and mixtures.
perature range from 304 ◦ C to 367 ◦ C (577–640 K), and there is
only one peak at 322 ◦ C (595 K) with a rate of 9.83 × 10−2 s−1 . For
pyrolysis at 100 K/min, cellulose decomposes from 293 ◦ C to 448 ◦ C relationships between both peak temperature and peak conversion
(566–721 K), and there is one peak at 399 ◦ C (672 K) (Table 2). The rate and the blend weight fraction of cellulose are not “additive” in
corresponding peak conversion rate is 0.104 s−1 . For the starch, the nature. The peak reaction rate of pure starch is considerably higher
pyrolytic decomposition spans the temperature range of 286 ◦ C to at 50 and 100 K/min than that of pure cellulose. It appears that as
387 C (559–660 K), with one peak at 327 ◦ C (600 K) at a conversion the fraction of cellulose increases beyond 50 wt% in the mixture,
rate of 0.189 s−1 . As with the 10 K/min pyrolysis, the peak tempera- the peak reaction rate is decreased by the presence of the cellu-
ture for 50 and 100 K/min pyrolysis increases following an increase lose in the mixture beyond a simple additive trend. This behavior
of the percentage of cellulose in the mixture while the opposite is magnified when we look at the combined DTG–DSC curves, as
trend occurs for peak DTG conversion rate. However, as can be seen seen in Fig. 3 for the 10 K/min pyrolysis. In this case we see that the
in Fig. 2, the peak rate of decomposition is highest for the pure DTG peaks correspond to the endothermic troughs for each solid
starch at 50 and 100 K/min, in opposition to the 10 K/min results. and blend. Pure starch has a gradually decreasing endotherm from
With the exception of Mixture 3, the peak temperatures increase ∼600 to 950 K, with a sharp decline after 950 K. The 75 wt% starch
following the increase of heating rate for all of the samples (the mixture (Mixture 3) follows this heat flow curve. The pure cellulose
peak DTG temperature of Mixture 3, 25 wt% cellulose, is slightly curve, on the other hand, remains fairly flat between 625 and 800 K,
lower for the 100 K/min than the 50 K/min heating rate – a differ- then has a slight exothermic peaks and then falls gradually at 900 K.
ence of 7 ◦ C). Overall this is not an unexpected relationship – the Similar behavior was noted by Yang et al. for the pyrolysis of pure
slower the heating rate, the more time each sample has to react at cellulose [44]. Likewise, the 75 wt% and 50 wt% cellulose mixtures
each temperature. In a similar vein, the peak reaction rate is high- more closely follow the shape of the cellulose endotherm, though
est at 100 K/min and lowest at 10 K/min pyrolysis. Interestingly, the both dip below the cellulose to a small trough at ∼875 K and peak
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 41

Fig. 4. Representative isoconversional plots for pure cellulose and 50:50 mixture at mass fractional conversions X = (䊉) 0.1; ( ) 0.15; ( ) 0.20; ( ) 0.25; ( ) 0.30; (
) 0.35; ( ) 0.40; ( ) 0.45; ( ) 0.50; ( ) 0.55; ( ) 0.60; ( ) 0.65; ( ) 0.70; ( ) 0.75; ( ) 0.80; ( ) 0.85.

Fig. 5. Mean activation energy for each pure sample and blend calculated via DAEM
(error bars represent 1 standard deviation).
Fig. 6. Predicted (Eq. (4)) versus measured (DAEM) activation energy at each mass
around 1000 K before tapering off to meet the pure cellulose curve. fraction conversion level for starch–cellulose mixtures.
The DSC data therefore also point to a non-additive thermal degra-
dation (DSC data for 50 and 100 K/min pyrolysis available in online Activation energies of each pure compound and blend were
Supplemental Information). To further explore this potential reac- determined at mass fractional conversions from 0.10 to 0.85
tion synergy between these two glucose biomass building blocks, at intervals of 0.05, detailed in Table 3 (full analysis with
we turn our attention to global activation energies as calculated via slope, intercept, and regression coefficients provided in the
an isoconversional method. Supplemental Information, Table S1; results presented graphically
in Supplemental Fig. S2).
3.2. Activation energy of starch–cellulose blends For the pyrolysis of pure cellulose, we find activation energies
ranging from 104.2 ± 6.9 to 169.2 ± 12.8 kJ/mol. For pure starch
As heating rate has a significant impact on activation ener- the DAEM returned activation energies between 197.2 ± 69.7 and
gies of even these pure biomass components, we turn to an 270.9 ± 34.3 kJ/mol. This range of values – and indeed the average
isoconversional analysis of the data using the distributed activa- activation energies (Fig. 5) measured – is in agreement with the
tion energy model to remove this experimental variable from our vast pure cellulose pyrolysis literature. Using a first order Arrhenius
analysis of reaction synergism. Overall good linearity was found for model, literature values of the activation energy of pure cellu-
each pure compound and blend using the integral method of the lose range from 110.0 kJ/mol [47] to 238.0 kJ/mol [48]. Hashimoto
DAEM, illustrated in Fig. 4. et al. [49] report activation energy of 167.7 kJ/mol via a nonlinear
42 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47

Table 3
Activation energies and pre-exponential factors for each pure fuel and blend determined via the Distributed Activation Energy Model.

Conversion Ea (kJ/mol) A (s−1 ) Ea (kJ/mol) A (s−1 )

Cellulose Mixture 2
0.10 169.2 ± 12.8 4.10E+10 ± 1.26E+01 132.3 ± 0.6 4.00E+07 ± 2.16E+00
0.15 153.0 ± 8.7 1.14E+09 ± 5.56E+00 147.2 ± 4.4 6.86E+08 ± 2.42E+00
0.20 142.7 ± 1.7 1.19E+08 ± 1.39E+00 158.1 ± 7.0 5.30E+09 ± 4.04E+00
0.25 128.5 ± 17.1 5.60E+06 ± 2.72E+01 163.9 ± 7.6 1.51E+10 ± 4.52E+00
0.30 123.4 ± 17.4 1.82E+06 ± 2.84E+01 173.0 ± 9.5 7.99E+10 ± 6.59E+00
0.35 119.6 ± 20.8 7.78E+05 ± 5.29E+01 173.2 ± 21.7 6.00E+10 ± 7.12E+01
0.40 116.9 ± 15.1 4.19E+05 ± 1.77E+01 172.8 ± 15.4 6.61E+10 ± 2.08E+01
0.45 115.9 ± 12.6 3.19E+05 ± 1.10E+01 174.6 ± 29.3 6.77E+10 ± 3.09E+02
0.50 112.8 ± 13.7 1.61E+05 ± 1.33E+01 186.3 ± 23.8 5.55E+11 ± 1.03E+02
0.55 112.8 ± 14.0 1.50E+05 ± 1.39E+01 191.1 ± 22.1 1.11E+12 ± 7.18E+01
0.60 111.1 ± 11.3 1.01E+05 ± 8.34E+00 189.9 ± 11.6 7.11E+11 ± 9.25E+00
0.65 110.5 ± 11.5 8.61E+04 ± 8.55E+00 187.9 ± 21.7 4.17E+11 ± 6.48E+01
0.70 108.7 ± 8.8 5.62E+04 ± 5.14E+00 184.2 ± 24.8 1.74E+11 ± 1.14E+02
0.75 108.5 ± 8.8 5.17E+04 ± 5.18E+00 179.7 ± 25.0 6.29E+10 ± 1.15E+02
0.80 105.5 ± 9.4 2.69E+04 ± 5.66E+00 174.9 ± 10.3 2.04E+10 ± 7.01E+00
0.85 104.2 ± 6.9 1.94E+04 ± 3.56E+00 160.6 ± 16.5 1.11E+09 ± 2.20E+01

Starch Mixture 3
0.10 214.2 ± 22.9 1.89E+15 ± 1.14E+02 123.2 ± 9.1 1.01E+07 ± 6.52E+00
0.15 197.2 ± 69.7 3.54E+13 ± 1.57E+06 121.1 ± 15.4 5.20E+06 ± 2.30E+01
0.20 211.1 ± 66.2 5.07E+14 ± 7.15E+05 135.6 ± 18.2 8.06E+07 ± 3.93E+01
0.25 221.5 ± 68.8 3.60E+15 ± 1.18E+06 136.0 ± 24.4 8.29E+07 ± 1.36E+02
0.30 230.6 ± 59.2 1.99E+16 ± 1.62E+05 135.9 ± 23.8 7.40E+07 ± 1.19E+02
0.35 236.9 ± 56.4 6.25E+16 ± 8.80E+04 134.6 ± 22.4 5.24E+07 ± 8.85E+01
0.40 243.1 ± 53.0 1.91E+17 ± 4.33E+04 134.3 ± 21.7 4.55E+07 ± 7.67E+01
0.45 247.2 ± 49.1 3.89E+17 ± 1.92E+04 147.8 ± 32.9 7.26E+08 ± 7.23E+02
0.50 250.3 ± 39.5 6.28E+17 ± 2.75E+03 148.6 ± 33.0 7.79E+08 ± 7.36E+02
0.55 254.1 ± 38.6 1.16E+18 ± 2.26E+03 149.4 ± 33.2 8.43E+08 ± 7.50E+02
0.60 251.4 ± 31.0 5.51E+17 ± 4.79E+02 150.5 ± 33.5 9.38E+08 ± 7.70E+02
0.65 252.0 ± 25.4 4.78E+17 ± 1.53E+02 159.3 ± 41.7 5.20E+09 ± 3.86E+03
0.70 270.2 ± 31.4 1.15E+19 ± 4.81E+02 165.2 ± 46.9 1.45E+10 ± 1.05E+04
0.75 270.2 ± 35.8 5.73E+19 ± 1.41E+03 174.8 ± 56.9 8.08E+10 ± 7.27E+04
0.80 270.9 ± 34.3 5.47E+19 ± 1.02E+03 174.1 ± 52.9 5.10E+10 ± 3.09E+04
0.85 252.9 ± 24.5 1.12E+18 ± 1.38E+02 179.5 ± 57.6 1.15E+11 ± 6.91E+04

Mixture 1
0.10 143.5 ± 18.6 2.65E+08 ± 4.07E+01
0.15 138.7 ± 18.0 7.39E+07 ± 3.51E+01
0.20 138.0 ± 11.2 5.11E+07 ± 8.90E+00
0.25 135.7 ± 8.5 2.73E+07 ± 5.20E+00
0.30 135.7 ± 15.5 2.27E+07 ± 1.98E+01
0.35 139.3 ± 13.7 4.13E+07 ± 1.41E+01
0.40 138.0 ± 15.4 2.78E+07 ± 1.92E+01
0.45 133.4 ± 13.1 9.95E+06 ± 1.22E+01
0.50 132.4 ± 9.9 7.51E+06 ± 6.60E+00
0.55 131.4 ± 6.8 5.59E+06 ± 3.64E+00
0.60 126.0 ± 5.2 1.77E+06 ± 2.67E+00
0.65 123.8 ± 2.6 1.07E+06 ± 1.63E+00
0.70 123.0 ± 0.4 8.36E+05 ± 1.07E+00
0.75 121.0 ± 2.9 5.34E+05 ± 1.72E+00
0.80 115.9 ± 3.9 1.81E+05 ± 2.08E+00
0.85 113.8 ± 6.2 1.14E+05 ± 3.16E+00

least squares analysis of nonisothermal data; “semi-global” mod- starch? We turn to an analysis of the primary gas components
els yield values ranging from 61.0 kJ/mol [50] to 242.4 kJ/mol evolved during pyrolysis to probe this behavior further.
[51]. Finally, isoconversional models show a variety of activation
energies, from 164.0 kJ/mol [52] to 175.6 kJ/mol via the DAEM 3.3. Gaseous products evolved from pyrolysis
[53].
Using the DAEM calculated activation energies, we predicted, The evolved gases: H2 , CH4 , C2 H2 , C2 H6 and CO2 (corresponding
via Eq. (4), an activation energy as a function of blend composi- to AMU signals of 2, 16, 26, 30 and 44, respectively) were mon-
tion at each mass fraction conversion, plotted against the measured itored in this study by pyrolyzing 400 ± 30 mg each of the pure
activation energy in Fig. 6. We find that deviations from predicted compounds and mixtures in a tube furnace at 10 K/min (the maxi-
values are all above the y = x line, such that the predicted values are, mum heating rate of the furnace) with high purity nitrogen flowing
with the exception of a few points for Mixture 2, all higher than the at 105 ± 5 mL/min through the furnace. The data are presented in
measured values. This suggests some sort of synergy between the Table 4 using natural logarithm form of the ratio of the evolved gas
two components. However, it is not clear whether the two com- divided by the sum of the gases (log of the partial pressure = log con-
pounds are interacting chemically or physically; is the “interaction” centration), normalized to the fraction of total sample converted
behavior due to chemical interactions of volatiles promoting reac- (between 0.818 for Mixture 2 and 0.864 for pure starch.)
tion among the solids, or is it a transport limitation related to the Fig. 7 presents relative concentrations of each evolved
higher activation energy barrier for cellulose to pyrolyze versus gas marker compound as a function of temperature.
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 43

Table 4
Peak temperatures and concentrations normalized to fraction devolatilized of gaseous species evolved for each pure compound and mixture.

H2 Peak 1 H2 Peak 2

Temp (K) Log Pi/P Temp (K) Log Pi/P

Cellulose 642.7 9.57E−09 1055.5 1.46E−07

Starch 586.0 1.33E−08 1035.7 2.53E−07
Mixture 1 619.3 1.49E−08 1033.8 2.42E−07
Mixture 2 605.6 1.78E−08 1041.6 2.78E−07
Mixture 3 597.8 1.12E−08 1039.7 2.32E−07

C2 H2 Peak 1 C2 H2 Peak 2

Temp (K) Log Pi/P Temp (K) Log Pi/P

Cellulose 640.8 1.13E−09 758.1 1.14E−09

Starch 584.1 1.44E−09 734.6 2.15E−09
Mixture 1 617.3 1.70E−09 738.4 2.12E−09
Mixture 2 607.5 1.91E−09 738.5 2.32E−09
Mixture 3 619.3 9.71E−10 746.4 1.89E−09

CH4 Peak 1 CH4 Peak 2

Temp (K) Log Pi/P Temp (K) Log Pi/P

Cellulose 634.9 1.32E−08 830.4 1.11E−08

Starch 582.1 3.40E−08 805.0 2.18E−08
Mixture 1 613.4 2.61E−08 805.3 2.03E−08
Mixture 2 603.6 3.07E−08 803.1 1.93E−08
Mixture 3 603.6 1.88E−08 807.0 1.86E−08

C2 H6 Peak 1 C2 H6 Peak 2

Temp (K) Log Pi/P Temp (K) Log Pi/P

Cellulose 609.5 9.43E−10 756.1 1.04E−09

Starch 580.2 9.57E−10 734.6 1.26E−09
Mixture 1 611.5 1.07E−09 738.5 1.32E−09
Mixture 2 603.6 1.19E−09 740.5 1.53E−09
Mixture 3 599.7 9.02E−10 746.4 1.23E−09

CO2 peak

Temp (K) Log Pi/P

Cellulose 634.9 9.57E−08

Starch 582.1 2.83E−07
Mixture 1 613.4 2.03E−07
Mixture 2 601.7 2.37E−07
Mixture 3 599.9 1.43E−07

Fig. S1 (Supplemental Information) shows the concentration detector and (2) as gases devolatilize from the solid, they must dif-
of each evolved gas as a function of temperature with overlaid fuse out of the pores and from the surface of the solid sample,
heat flow data for each compound and mixture. For cellulose, which gives them ample time to experience secondary cracking.
there is a primary peak of evolved CO2 at 638 K, approximately Fig. S1 in the online supplemental information shows that such
25 K higher than the primary DTG peak at 612.5 K. As cellulose a trend continues through the peak endothermic behavior. That is,
decomposes, there are two peaks for H2 evolution. The first is the peak endotherms (requiring the highest amount of heat to com-
at 642.7 K and second at 1055.5 K. Likewise, there are two peaks plete the reaction at this point) loosely correspond to the first peaks
for CH4 , C2 H2 and C2 H6 evolution. Cellulose has the highest peak in each of the evolved gas components.
evolution temperatures for all of the gases monitored, and starch We note no apparent relationship between the peak concen-
has the lowest peak temperatures for each gas. trations and cellulosic fraction, with the general exception that
Of note is the correlation between the first peak temperature for the pure cellulose has the lowest concentrations of any sample,
each evolved gas species and the fraction of cellulose present in the though within the same order of magnitude for each solid. This is
blend (Table 5). These two variables have correlation coefficients, illustrated graphically in Fig. 9. Yang et al. [44], using FTIR to mon-
R2 , for simple linear regression models (“additive schemes”) for H2 , itor evolved gases, found that the total gaseous pyrolysis products
CH4 , and CO2 greater than 0.90, with a somewhat looser relation- released from cellulose were lower than those devolatilizing from
ship for C2 H2 and C2 H6 (R2 > 0.70). However, as seen in Fig. 8, this hemicellulose. Hosoya et al. [54] found slightly lower tar yield and
is clearly not the case for the second peak evolution temperatures, slightly higher char yield by mixing in a 2:1 cellulose:hemicellulose
where correlation coefficients are all less than 0.45. mixture than would be predicted in an additive scheme, which the
There is a loose correlation between peak DTG temperature authors suggest is because cellulose is covered by a molten hemi-
and peak evolution temperature for any gaseous species, and in cellulose phase, inhibiting volatile formation from the cellulose.
the expected direction. As temperature is increasing at a rate of Given the high volume expansion of starch, and the observation
10 K/min, we would expect the peak mass loss of the sample to that the total mass conversion of all the samples was approxi-
occur several degrees (and therefore seconds) before the evolved mately equal, we do not believe the lower (though same order
gases are detected in the MS because (1) there will be a lag as of magnitude) concentrations of H2 , CH4 , C2 H2 , C2 H6 and CO2 , of
evolving compounds travel the capillary compound and reach the cellulose and mixtures higher in cellulose is due to an analogous
44 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47

Fig. 7. Log of evolved gas concentrations from pyrolysis of cellulose, starch, and blends at 10 K/min for cellulose ( ), starch ( ), and blends Mixture 1: 75 wt% cellulose (
); Mixture 2: 50 wt% cellulose ( ); Mixture 3: 25 wt% cellulose ( ). (a) H2 , (b) CH4 , (c) C2 H2 , (d) C2 H6 and (e) CO2 .

Table 5
Linear regression equations for relationship between peak evolved gas temperature and mass fraction of cellulose.

Linear regression equation R R2

H2 Peak 1 Peak T (K) = 53.96 ± 6.52 × (xcellulose ) + 583.30 ± 3.99 0.97877 0.95799
H2 Peak 2 Peak T (K) = 13.48 ± 9.75 × (xcellulose ) + 1034.50 ± 5.97 0.62377 0.38909
CH4 Peak 1 Peak T (K) = 46.16 ± 8.31 × (xcellulose ) + 584.44 ± 5.09 0.95468 0.91141
CH4 Peak 2 Peak T (K) = 19.64 ± 12.19 × (xcellulose ) + 800.34 ± 7.47 0.68107 0.46386
C2 H2 Peak 1 Peak T (K) = 44.56 ± 15.53 × (xcellulose ) + 591.52 ± 9.51 0.85609 0.73289
C2 H2 Peak 2 Peak T (K) = 15.6 ± 10.31 × (xcellulose ) + 735.4 ± 6.31 0.65796 0.43291
C2 H6 Peak 1 Peak T (K) = 28.16 ± 8.26 × (xcellulose ) + 586.82 ± 5.06 0.89154 0.79484
C2 H6 Peak 2 Peak T (K) = 14.04 ± 9.14 × (xcellulose ) + 736.2 ± 5.60 0.66334 0.44002
CO2 Peak Peak T (K) = 47.64 ± 7.21 × (xcellulose ) + 582.58 ± 4.42 0.96728 0.93563
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 45

Fig. 8. Temperatures of peak evolution of evolved gases as a function of mass fraction of cellulose for each pure compound and mixture pyrolyzed at 10 K/min: H2 ( ), CH4
( ), C2 H2 ( ), C2 H6 ( ), CO2 ( ).

liquid-phase shielding effect. In addition, from the fairly linear terms of reaction rates and kinetics that the cellulose, when present
behavior for the first peak, it would appear that initially the blended as the majority component (75 wt%) inhibits the devolatilization
streams do not display evidence of reaction synergism between of starch, and perhaps at higher starch fractions the rates of
the starch and cellulose during co-pyrolysis in terms of the non- reaction are higher than a simple additive scheme would predict.
condensable gases formed. Rather, it is likely that the cellulosic Activation energies are not additive in nature, suggesting some
gases underwent repolymerization of anhydrosugars, increasing sort of synergism between the two components.
tar yield and decreasing non-condensable gas yield, which is sup- However, given the lack of chemical reaction synergism noted
ported through the non-linear relationship for the second peak for the first devolatilization peak – there is no evidence that the
devolatilization as observed in Table 5. evolved (noncondensable) gas profiles are drastically altered as
a result of the blends – we suggest that the synergies noted in
3.4. Kinetics and evolved gases: two sides to the synergism story terms of reaction kinetics are initially due to heat transport and
diffusional issues between the particles. For the 25 wt% cellulose
Starch and cellulose represent the building blocks of many and 50 wt% cellulose mixtures, the peak derivative thermogravi-
biomasses, and provide a foundational understanding of metric temperatures and rates indicate that the starch is perhaps
how different biomass constituents may influence pyrolytic promoting devolatilization of the cellulose; the peak temperatures
decomposition. There is no obvious chemical pathway by which are lower than predicted and peak rates higher than predicted
starch and cellulose, two glucose isomers and building blocks of by an additive model. For the 75 wt% cellulose mixture, we see
many biomasses, interact during pyrolysis. There is evidence in temperature considerable higher and rate considerably lower than

Fig. 9. Peak evolved gas concentration, normalized to total fraction evolved compounds as a function of mass fraction of cellulose for each pure compound and mixture
pyrolyzed at 10 K/min: H2 Peak 1 ( ), H2 Peak 2 ( ), CH4 Peak 1( ), CH4 Peak 2 ( ), Peak CO2 ( ), C2 H2 Peak 1 ( ), C2 H2 Peak 2 ( ), C2 H6 Peak 1 ( ), C2 H6 Peak 2
( ).
46 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47

would be predicted by an additive scheme, especially at the 50 [2] A. Oasmaa, C. Peacocke, S. Gust, D. Meier, R. McLellan, Norms and standards
and 100 K/min heating rates. This is further underscored by the for pyrolysis liquids. End-user requirements and specifications, Energy Fuels
19 (2005) 2155–2163.
differential scanning calorimetry results, where we see sharper [3] J.F. González, J.M. Encinar, J.L. Canito, E. Sabio, M. Chacón, Pyrolysis of cherry
and deeper endotherms for cellulose and the 75 wt% cellulose mix- stones: energy uses of the different fractions and kinetic study, J. Anal. Appl.
ture than the 50:50, 25 wt% cellulose and pure starch samples. The Pyrolysis 67 (2003) 165–190.
[4] F. de Miguel Mercader, M.J. Groeneveld, S.R.A. Kersent, R.H. Venderbosch, J.A.
primary peak evolved gas temperatures and mass fractions of cel- Hogendoorn, Pyrolysis oil upgrading by high pressure thermal treatment,
lulose in a mixture are strongly correlated; as the percentage of Fuel 89 (2010) 2829–2837.
cellulose increases, the temperatures at which the peak mass loss [5] T. Cornelissen, M. Jans, M. Stals, T. Kuppens, T. Thewys, G.K. Janssens, H.
Pastijn, J. Yperman, G. Reggers, S. Schreurs, R. Carleer, Flash co-pyrolysis of
and peak evolved gases emerge are linearly related. As such, there
biomass: the influence of biopolymers, J. Anal. Appl. Pyrolysis 85 (2009)
is not substantial evidence pointing to a high degree of chemi- 87–97.
cal reaction synergism during the pyrolysis of these two biomass [6] C. Meesri, B. Moghtaderi, Lack of synergistic effects in the pyrolytic
characteristics of woody biomass/coal blends under low and high heating rate
building blocks at lower temperatures. However, at higher tem-
regimes, Biomass Bioenergy 23 (2002) 55–66.
peratures, the relationship between cellulose fraction and peak [7] A.K. Sadhukhan, P. Gupta, T. Goyal, R.K. Saha, Modelling of pyrolysis of
evolved gases is not linear in nature. The most logical explana- coal–biomass blends using thermogravimetric analysis, Bioresour. Technol.
tion for the behavior is the cellulose undergoing char-forming side 99 (2008) 8022–8026.
[8] A.O. Aboyade, M. Carrier, E.L. Meyer, J.H. Knoetz, J.F. Görgens, Model fitting
reactions [55,56], which may inhibit starch devolatilization and kinetic analysis and characterization of the devolatilization of coal blends
increase peak pyrolysis temperatures. As Várhegyi explains, unfor- with corn and sugarcane residues, Thermochim. Acta 530 (2012)
tunately the physical properties (i.e. transport) of a solid sample 95–106.
[9] Q. Liu, Z. Zhong, S. Wang, Z. Luo, Interactions of biomass components during
change with temperature, and of course in our case, mixture com- pyrolysis: a TG-FTIR study, J. Anal. Appl. Pyrolysis 90 (2011)
position likely compounds this phenomenon [57]. We presume 213–218.
– like all similar TGA experiments – that the samples are under [10] T. Sonobe, N. Worasuwannarak, S. Pipatmanomai, Synergies in co-pyrolysis of
Thai lignite and corncob, Fuel Process. Technol. 89 (2008) 1371–1378.
kinetic control during the experiments, i.e. devolatilization and [11] A.M. Celaya, A.T. Lade, J.L. Goldfarb, Co-combustion of brewer’s spent grains
product formation is governed by chemical reactions and not trans- and Illinois no 6 coal: impact of blend ratio on pyrolysis and oxidation
port properties. However, the synergism in terms of mass loss data behavior, Fuel Process. Technol. 129 (2015) 39–51.
[12] P. Yangali, A.M. Celaya, J.L. Goldfarb, Co-pyrolysis reaction rates and
and lack thereof for pyrolysis gas data at lower temperatures, and
activation energies of West Virginia coal and cherry pit blends, J. Anal. Appl.
synergism for both data sets at higher temperatures, suggests that Pyrolysis 108 (2014) 203–211.
there might be both kinetic and transport forces at play between [13] L. Petrus, M.A. Noordermeer, Biomass to biofuels, a chemical perspective,
Green Chem. 8 (2006) 861–867.
these two biomass building blocks, especially at lower tempera-
[14] P. Aggarwal, D. Dollimore, A comparative study of the degradation of different
tures. This interplay is understood through the significantly higher starches using thermal analysis, Talanta 43 (1996) 1527–1530.
heat capacity and peak devolatilization temperatures of cellulose [15] A.A.A. Soliman, N.A. El-Shinnawy, F. Mobarak, Thermal behavior of starch and
versus starch, whereby a mixture with a significant amount of cel- oxidized starch, Thermochim. Acta 296 (1997) 149–153.
[16] X. Liu, L. Yu, F. Xie, M. Li, L. Chen, X. Li, Kinetics and mechanism of thermal
lulose may suppress the ability of starch to devolatilize at lower decomposition of cornstarches with different amylose/amylopectin ratios,
temperatures. Starch 62 (2010) 139–146.
[17] D.F. Arseneau, Competitive reactions in the thermal decomposition of
cellulose, Can. J. Chem. 49 (1971) 632–638.
4. Conclusions [18] H. Yang, R. Yan, H. Chen, C. Zheng, D.H. Lee, D.T. Liang, In-depth investigation
of biomass pyrolysis based on three major components: hemicellulose,
cellulose, and lignin, Energy Fuels 20 (2006) 388–393.
There is much evidence in the literature that blending biomass [19] B.G. Girija, R.R.N. Sailaja, S. Biswas, M.V. Deepthi, Mechanical and thermal
and other solid fuel streams can alter reaction rates and tempera- properties of Eva blended with biodegradable ethyl cellulose, J. Appl. Polym.
tures – as observed here – that can potentially increase conversion Sci. 116 (2010) 1044–1056.
[20] W. Jiang, X. Qiao, K. Sun, Mechanical and thermal properties of thermoplastic
efficiency. For example, mixtures of 50% or more of starch appear to acetylated starch/poly(ethylene-co-vinyl alcohol) blends, Carbohydr. Polym.
promote the pyrolysis of cellulose at lower temperatures. In addi- 65 (2006) 139–143.
tion, there is a consensus developing that we can advantageously [21] J.F. Mano, D. Koniarova, R.L. Reis, Thermal properties of thermoplastic
starch/synthetic polymer blends with potential biomedical applicability, J.
alter bio-oil composition by blending biomasses to improve stabil- Mater. Sci.: Mater. Med. 14 (2003) 127–135.
ity and fuel quality. However, what the data here suggest is that [22] R. Moriana, F. Vilaplana, S. Karlsson, A. Ribes-Greus, Improved
this reaction synergy may not be due to a chemical interaction thermo-mechanical properties by the addition of natural fibres in
starch-based sustainable biocomposites, Composites A 42 (2011) 30–40.
occurring during devolatilization, at least not at lower tempera- [23] Y.Z. Wan, H. Luo, F. He, H. Liang, Y. Huang, X.L. Li, Mechanical, moisture
tures. Rather, the ability to tune bio-oil product composition is absorption, and biodegradation behavior of bacterial cellulose
likely due to secondary reactions occurring to form or change tar fibre-reinforced start biocomposites, Compos. Sci. Technol. 69 (2009)
1212–1217.
phase constituents.
[24] M.J.J. Antal, G. Varhegyi, E. Jakab, Cellulose pyrolysis kinetics: the current
state of knowledge, Ind. Eng. Chem. Res. 37 (1998) 1267–1277.
[25] M. Van de Velden, J. Baeyens, A. Brems, R. Dewil, Fundamentals, kinetics and
Acknowledgement endothermicity of the biomass pyrolysis reaction, Renew. Energy 35 (2010)
232–242.
The authors acknowledge the generous support of the scholar- [26] C. Di Blasi, Modeling chemical and physical processes of wood and biomass
pyrolysis, Prog. Energy Combust. Sci. 24 (2008) 47–90.
ship from China Scholarship Council (CSC) under the Grant CSC NO.
[27] P. Parasuraman, R. Singh, T. Bolton, S. Omori, R. Francis, Estimation of
201406350169. hardwood lignin concentrations by UV spectroscopy and chlorine
demethylation, Bioresources 2 (2007) 459–471.
[28] L. Buessing, J.L. Goldfarb, Energy along interstate-95: pyrolysis kinetics of
Appendix A. Supplementary data Floridian cabbage palm (Sabal palmetto), J. Anal. Appl. Pyrolysis 96 (2012)
78–85.
[29] M. Brebu, C. Vasile, Thermal degradation of lignin – a review, Cellul. Chem.
Supplementary data associated with this article can be found, in
Technol. 44 (2010) 353–363.
the online version, at http://dx.doi.org/10.1016/j.tca.2015.09.002. [30] S. Vyazovkin, K. Chrissafis, M.L. Di Lorenzo, N. Koga, M. Pijolat, B. Rodiut, N.
Sbirrazzuoli, J.J. Suñol, ICTAC Kinetics Committee recommendations for
collecting experimental thermal analysis data for kinetic computations,
References Thermchim. Acta 590 (2014) 1–23.
[31] A. Aboulkas, K. El harfi, A. El bouadili, Kinetic and mechanism of Tarfaya
[1] N. Kauffman, D. Hayes, R. Brown, A life cycle assessment of advanced biofuel (Morocco) oil shale and LDPE mixture pyrolysis, J. Mater. Process. Technol.
production from a hectare of corn, Fuel 90 (2011) 3306–3314. 206 (2008) 16–24.
 J. Xue et al. / Thermochimica Acta 618 (2015) 36–47 47

[32] J. Cai, R. Liu, New distributed activation energy model: numerical solution and [45] E. Biagini, F. Lippi, L. Petarca, L. Tognotti, Devolatilization rate of biomasses
application to pyrolysis kinetics of some types of biomass, Bioresour. Technol. and coal–biomass blends: an experimental investigation, Fuel 81 (2002)
99 (2008) 2795–2799. 1041–1050.
[33] Z. Li, C. Liu, Z. Chen, J. Qian, W. Zhao, W. Zhu, Analysis of coals and biomass [46] S. Baumlin, F. Broust, M. Ferrer, N. Meunier, E. Marty, J. Lédé, Continuous self
pyrolysis using the distributed activation energy model, Bioresour. Technol. stirred tank reactor: measurement of the cracking kinetics of biomass
100 (2009) 948–952. pyrolysis vapours, Chem. Eng. Sci. 60 (2005) 41–55.
[34] T. Mani, P. Murugan, N. Mahinpey, Determination of distributed activation [47] J. Lédé, Cellulose pyrolysis kinetics: an historical review on the existence and
energy model kinetic parameters using annealing optimization method for role of intermediate active cellulose, J. Anal. Appl. Pyrolysis 94 (2012)
nonisothermal pyrolysis of lignin, Ind. Eng. Chem. Res. 48 (2009) 1464–1467. 17–32.
[35] C.P. Please, M.J. McGuinness, D.L.S. McElwain, Approximations to the [48] G. Várhegyi, M.H. Antal Jr., E. Jakab, P. Szabó, Kinetic modeling of biomass
distributed activation energy model for the pyrolysis of coal, Combust. Flame pyrolysis, J. Anal. Appl. Pyrolysis 42 (1997) 73–87.
133 (2003) 107–117. [49] K. Hashimoto, I. Hasegawa, J. Hayashi, K. Mae, Correlations of kinetic
[36] T. Sonobe, N. Worasuwannarak, Kinetic analyses of biomass pyrolysis using parameters in biomass pyrolysis with solid residue yield and lignin content,
the distributed activation energy model, Fuel 87 (2008) 414–421. Fuel 90 (2011) 104–112.
[37] A. Dawood, K. Miura, Pyrolysis kinetics of ␥-irradiated polypropylene, Polym. [50] T.B. Reed, C.D. Cowdery, Heat flux requirements for fast pyrolysis and a new
Degrad. Stabil. 73 (2001) 347–354. method for generating biomass vapor, in: 193rd National Meeting of the
[38] J. Cai, T. Li, R. Liu, A critical study of the Miura–Maki integral method for the American Chemical Society, April 5–10, Denver, CO, USA, American Chemical
estimation of the kinetic parameters of the distributed activation energy Society Division of Petroleum Chemistry, 1987.
model, Bioresour. Technol. 102 (2011) 3894–3899. [51] R.S. Miller, J. Bellan, A generalized biomass pyrolysis model based on
[39] C.C. Lakshmanan, N. White, A new distributed activation energy model using superimposed cellulose, hemicellulose and lignin kinetics, Combust. Sci.
Weibull distribution for the representation of complex kinetics, Energy Fuels Technol. 126 (1997) 97–137.
8 (1994) 1158–1167. [52] L. Cabrales, N. Abidi, On the thermal degradation of cellulose in cotton fibers,
[40] A. Burnham, R. Braun, Global kinetic analysis of complex materials, Energy J. Thermal Anal. Calorim. 102 (2010) 485–491.
Fuels 13 (1999) 1–22. [53] L. Gašparovič, J. Labovskŷ, J. Markoš, L. Jelemenský, Calculation of kinetic
[41] P.J. Skrdla, Crystallizations, solid-state phase transformations and dissolution parameters of the thermal decomposition of wood via the distributed
behavior explained dispersive kinetic models based on a Maxwell–Boltzmann activation energy model (DAME), Chem. Biochem. Eng. 26 (2012) 45–53.
distribution of activation energies: theory, applications, and practical [54] T. Hosoya, H. Kawamoto, S. Saka, Cellulose–hemicellulose and
limitations, J. Phys. Chem. A 113 (2009) 9329–9336. cellulose–lignin interactions in wood pyrolysis at gasification temperature, J.
[42] K. Miura, T. Maki, A simple method for estimating f(E) and k0(E) in the Anal. Appl. Pyrolysis 80 (2007) 118–125.
distributed activation energy model, Energy Fuels 12 (1998) 864–869. [55] A. Broido, M. Weinste, Kinetics of solid-phase cellulose pyrolysis, in: H.G.
[43] J. Chattopadhyay, C. Kim, R. Kim, D. Pak, Thermogravimetric study on Wiedemann (Ed.), Proceedings of the 3rd International Conference on
pyrolysis of biomass with Cu/Al2 O3 catalysts, J. Ind. Eng. Chem. 15 (2008) Thermal Analysis, Birkhauser Verlag, Basel, 1971, pp. 285–296.
72–76. [56] A.K. Burnham, R.K. Weese, Kinetics of thermal degradation of explosive
[44] Y.F. Huanga, W.H. Kuanb, P.T. Chiueha, S.L. Lo, Pyrolysis of biomass by thermal binders Viton A, Estane, and Kel-F, Thermochim. Acta 426 (2005) 85–92.
analysis–mass spectrometry (TA–MS), Bioresour. Technol. 102 (2011) [57] G. Várhegyi, Aims and methods in non-isothermal reaction kinetics, J. Anal.
3527–3534. Appl. Pyrolysis 79 (2007) 278–288.