Industrial Crops & Products xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Industrial Crops & Products

journal homepage: www.elsevier.com/locate/indcrop

Research Paper

Classification of biomass through their pyrolytic bio-oil composition using

FTIR and PCA analysis
Eliane Lazzaria, Tiago Schenaa, Marcelo Caetano Alexandre Marceloa, Carmem Tatiane Primaza,
⁎
Aline Nunes Silvaa, Marco Flôres Ferrãoa,b, Thiago Bjerkc, Elina Bastos Caramãoa,c,d,
a
UFRGS, Instituto de Química, ZIP 91501-970, Porto Alegre, RS, Brazil
b
INCT-BIO, Instituto Nacional de Ciência e Tecnologia em Bioanalítica, Instituto de Química, Universidade Estadual de Campinas, Campinas, SP, Brazil
c
UNIT, Programa de Pós-Graduação em Biotecnologia Industrial (PBI), Aracaju, SE, Brazil
d
INCT-E & A, Instituto Nacional de Ciência e Tecnologia em Energia e Meio Ambiente, Salvador, BA, Brazil1

A R T I C L E I N F O A B S T R A C T

Keywords: Fourier transform infrared (FTIR) spectroscopy, combined to principal components analysis (PCA), was applied
Bio-oil in the classification of biomasses through the composition of their bio-oil. Bio-oils were produced through
FTIR pyrolysis in a bed fixed reactor, using fifteen biomass sources available in Brazil and its characterization was
PCA made using gas chromatography coupled to mass spectrometric detector (GC/qMS). Around two hundred
GC/qMS
compounds were tentatively identified in the fifteen bio-oil samples. As expected, the chemical compositions in
each bio-oil were distinct. Through the chromatographic information and PCA of the FTIR spectra it was possible
observed the similarity and dissimilarity of biomasses according their bio-oil compositions. PCA revealed that
FTIR spectra of biomasses fell into three different groups representing distinct bio-oil chemical compositions.
The biomasses that belong to group 1 showed bio-oil compositions rich in carboxylic acids, the group 2 showed
bio-oil compositions consisting predominantly of phenols and group 3 showed bio-oils with a significant amount
of nitrogen compounds. Such clustering information allow exploring bio-oil quality prior to pyrolysis process.

1. Introduction
The use of biomass as a renewable source of energy can reduce the
dependency on fossil sources (Jahirul et al., 2012). In Brazil and in the
world, the agro-industrial wastes are an interesting biomass source due
to the large amount generated per year and to the environmental pro-
blems associated with them. In the year 2015, only for sugarcane,
cassava and coconut crops Brazil was responsible for more than 750,
22, and 5 million tons, respectively (IBGE, 2017; Rambo et al., 2015).
For every ton of sugarcane, rice and coconut produced, approximately
250 kg of waste is generated and for cassava 490 kg of waste is gen-
erated (Rambo et al., 2015; Pattiya and Suttibak, 2012). Brazil also is
known as one of the largest producers of rice, with an annual produc-
tion of approximately 12 million tons. Agro-industrial wastes from rice
processing are equivalent to 3 million tons annually (Rambo et al.,
2015; IBGE, 2017). Some of other agro-industrial wastes generated in
high amount in Brazil are coffee husk, peanut shell, mango waste,
pineapple leaves, cottonseed, peach pit, peanut shell and wood.
Pyrolysis is a promising way to convert these biomasses into high-
value products. The process of pyrolysis consists of the thermal de-
composition of biomass at high temperatures and in the complete ab-
sence of oxygen to obtain gas, solid (bio-char) and liquid (bio-oil)
products (Bridgwater, 2012). In recent years, bio-oil has received a lot
of attention due to its potential uses as biofuel (after upgrading process)
or as a starting material for producing chemicals (Czernik and
Bridgwater, 2004).
Bio-oil is a complex mixture of water and a hundred of organic
compounds that can be classified into the following categories: phenols,
ketones, acids, esters, aldehydes, alcohols, furans, anhydrous-sugars,
nitrogen containing compound, hydrocarbons, carboxylic acids (Jahirul
et al., 2012). The distribution of these compounds in the bio-oil de-
pends mainly on the variability of different proportions of lignin,

Abbreviations: AB, aquatic biomass; AMS, almond of mango seed; CF, coconut fibers; CP, cassava peel; CS, crambe seed; CSK, coffee silverskin; CTS, cotton seed; DCM, dichloromethane;
ES, eucalyptus sawdust; ETS, energetic tobacco seeds; FTIR, Fourier transform infrared spectroscopy; GC/qMS, gas chromatography coupled to mass spectrometric detector; LTPRI, linear
temperature programmed retention indexes; PC, peach cores; PCA, principal components analysis; PL, pineapple leaves; PLS, partial least squares; PS, peanut shell; RH, rice husk; SCB,
sugarcane bagasse; SCG, spent coffee grounds
⁎
Corresponding author at: UFRGS, Instituto de Química, ZIP 91501-970, Porto Alegre, RS, Brazil.
E-mail address: elina@ufrgs.br (E.B. Caramão).
1
Home page: http://www.inct.cienam.ufba.br.

http://dx.doi.org/10.1016/j.indcrop.2017.11.005
Received 13 June 2017; Received in revised form 26 October 2017; Accepted 3 November 2017
0926-6690/ © 2017 Published by Elsevier B.V.

Please cite this article as: Lazzari, E., Industrial Crops & Products (2017), http://dx.doi.org/10.1016/j.indcrop.2017.11.005
E. Lazzari et al. Industrial Crops & Products xxx (xxxx) xxx–xxx

Table 1
Details of the investigated biomasses.

Biomasses Samples ID Provided by local/industrya

Almond of Mango Seeds AMS Food industry, Dourados, MS, Brazil.

Coconut Fibers CF Embrapa, Aracaju, SE, Brazil
Pineapple Leaves PL Local market, Porto Alegre, RS, Brazil.
Sugarcane Bagasse SCB Food industry, Porto Alegre, RS, Brazil.
Cotton Seeds CTS Planalto Farm, Costa Rica, MS, Brazil.
Coffee Silverskin CSK Marata Industry, Itaporanga D’Ajuda, SE, Brazil.
Rice Husk RH Food industry, Pelotas, RS, Brazil.
Eucalyptus Sawdust ES CMPC Cellulose Industry, Guaíba, RS, Brazil.
Peach Cores PC Schrann Food Industry, Pelotas, RS, Brazil.
Spent Coffee Grounds SCG Local market, Porto Alegre, RS, Brazil.
Peanut Shell PS Local market, Porto Alegre, RS, Brazil.
Cassava Peel CP Food industry, Dourados, MS, Brazil.
Aquatic Biomass (duckweed) AB Pelotas, RS, Brazil
Crambe Seeds CS UERGS, Porto Alegre, RS, Brazil.
Residual Cake from Energetic Tobacco Seedsb ETS UNISC, Vale do Rio Pardo, RS, Brazil.

a
Different Brazilian regions which show high biomass production.
b
Residual cake was obtained by extraction of seeds of energetic tobacco.

polysaccharides (cellulose and hemicellulose), protein, triglycerides in Brazil and to use for the first time the FTIR data and PCA analysis to
etc., in the biomass (Jahirul et al., 2012; Sharma et al., 2015). classify the biomass through their bio-oil composition.
Bio-oil applications are strongly influenced by their chemical com-
position and by the type of upgrade used to improve the quality of this
2. Materials and methods
bio-oil (Bridgwater, 2012; Czernik and Bridgwater, 2004). As its com-
position depends largely on the composition of the original biomass, it
2.1. Biomass samples
is necessary to know it and to associate it with the characterization of
the bio-oil produced. Many authors demonstrated the production of bio-
Fifteen biomasses were investigated in this study. Among the bio-
oil by pyrolysis and characterization their chemical composition using
masses evaluated twelve are agro-industrial waste: almond of mango
chromatographic techniques (Muradov et al., 2010; Lazzari et al., 2016;
seed (AMS); coconut fibers (CF); pineapple leaves (PL); sugarcane ba-
Almeida et al., 2013; Bispo et al., 2016; Faccini et al., 2013; Moraes
gasse (SCB); cotton seed (CTS); coffee silverskin (CSK); rice husk (RH);
et al., 2012). However, the biomass-to-bio-oil conversion process can be
eucalyptus sawdust (ES); peach cores (PC); spent coffee grounds (SCG);
optimized using rapid methods for biomass characterization that allow
peanut shell (PS) and cassava peel (CP). These wastes represent a
pre-defining bio-oil quality prior to pyrolysis process.
considerable environmental problem in Brazil, due to the large amounts
In this context, FTIR is a non-destructive, simple, fast and cheap
produced per year.
instrumental technique that has been used successfully in the analysis
Other biomasses evaluated were: aquatic biomass (AB) (commonly
of the chemical composition of several biomass sources (Xu et al., 2013;
known as duckweed), crambe seed (CS) and residual cake from en-
Chadwick et al., 2014). It is likely that FTIR spectra might contain small
ergetic tobacco seeds (ETS). Aquatic biomass does not compete with
differences between two biomasses that could not be distinguished vi-
agriculture for land usage (Muradov et al., 2010), crambe is an inter-
sually, thus, multivariate data tool as PCA, are necessary for spectra
esting non-food source of seed oil (Onorevoli et al., 2014) and energetic
analysis (Liu et al., 2015, 2016). PCA is an effective variable reduction
tobacco seeds are a new biodiesel source available.
technique for spectroscopic data (Liu et al., 2015). In the PCA their
The great diversity of assessed biomass allows the representation of
scores on new principal components (PCs) evidence the correlations
different biomass sources available in Brazil, which show distinct che-
among spectra (or samples). Similar samples appear as clusters in the
mical composition, and exhibit potential for the production of bio-oil
score plot, while different samples appear segregate from each other
via pyrolysis.
(Liu et al., 2015, 2016). In a number of studies, spectroscopic data were
Some of these biomasses have already been studied in our research
associated with multivariate techniques for the rapid characterization
group through pyrolysis as AMS (Lazzari et al., 2016), CS (Onorevoli
of biomass samples for energy purposes. Everard et al. (2012) reported
et al., 2014), CF (Almeida et al., 2013; Bispo et al., 2016), SCG (Bispo
the use of online vis-NIR spectral sensing techniques in conjunction
et al., 2016), PC (Moraes et al., 2012), RH (Moraes et al., 2012) and ES
with partial least squares (PLS) and PCA for the rapid analysis of ash,
(Faccini et al., 2013; Schneider et al., 2014).
carbon content and moisture of Miscanthus and two varieties of willow.
Details of the biomasses investigated in this study are described in
In the results obtained, the PCA allow distinguished the different bio-
Table 1.
mass types and the PLS gave strong prediction for the carbon content
All the biomasses were ground in an industrial blender to produce a
and moisture of the samples. Chen et al. (2010) reported satisfactory
sub-sample at 28–60 mesh and dried in an oven at 110 °C during 24 h to
results for the prediction of lignin, cellulose and hemicellulose contents
remove moisture.
in wood samples using the FTIR spectral data in conjunction with PLS.
Through the FTIR and PCA, the wood samples were discriminated into
hardwoods and softwoods, in addition, wood samples with and without 2.2. Pyrolysis process
chemical treatments could also be differentiated. More recently, Liu
et al. (2016) described the application of the FTIR associated with PCA All the fifteen biomasses were subjected to intermediate pyrolysis in
to discriminate cotton plant biomass fractions. The results allowed a homemade vertical furnace containing a tubular fixed-bed quartz
evaluating the differences in the accumulation of carbohydrates among glass reactor. Schematic diagram of the reactor and furnace system was
the samples. illustrated elsewhere (Faccini et al., 2013). An electrical furnace heated
The aims of the present study were to obtain and characterized by the reactor and nitrogen was used as carrier gas. The pyrolysis vapors
GC/qMS bio-oils from the pyrolysis of fifteen biomass sources available were cooled through condenser at −10 °C, using a mixture of ethylene
glycol and water (1:1).

2
E. Lazzari et al.
The conditions of the pyrolysis process were based on previous
papers (Onorevoli et al., 2014; Moraes et al., 2012; Lazzari et al., 2016;
Almeida et al., 2013; Bispo et al., 2016; Faccini et al., 2013). Then, each
experimental run used 7 g of sample (28–60 mesh) with a heating rate
of 100 °C min−1, nitrogen flow of 100 mL min−1 and final temperature
of 700 °C.
The residence time at final temperature was 10 min (until no more
vapors are formed). The final temperature (700 °C) was established in
order to guarantee the total degradation of the biomass constituents.
The mass yield (w/w%) was measured by weighing the liquid pro-
duct (crude bio-oil) and solid product (bio-char + ashes) with reference
to the initial weight of biomass, while the yield of the non-condensable
products (gases + losses) was determined by difference and include the
losses on the system. The water was removed from crude bio-oils by
liquid–liquid extraction using 5 portions of 2 mL of dichloromethane
(DCM). The dry bio-oils were characterized by GC/qMS.
2.3. Bio-oil characterization by GC/qMS
Chromatographic analyses of the bio-oils were carried on a GC/qMS
system (Agilent Technologies, Palo Alto, CA, USA) equipped with an
AOC20 auto injector (split/splitless). The chromatographic separation
was performed on the DB-5 (60 m length, 250 μm i.d. and 0.25 μm p.t)
non-polar analytic column purchased from Agilent Technologies–J & W
Scientific (Palo Alto, CA, USA). The GC conditions were as follow: in-
jector, interface and ion source were kept at 280 °C; the oven tem-
perature program began at 50 °C for 4 min, heating at 4 °C min−1 until
280 °C, where it was kept for 10 min. Splitless mode was used in the
injection of 1 μL of samples solutions (10000 μg mL−1) dissolved in
DCM. Helium (ultra-pure, 99.999%, Linde Gases, Porto Alegre, RS,
Brazil) was used as carrier gas at a flow rate of 1.0 mL min−1. Mass
spectrometric detector operated in the full scan mode from m/z 45 to
m/z 400. The GCMS-solution software 2.6.1 (Shimadzu, Japan) pro-
cessed all data.
The compounds were tentatively identified using the following
parameters: spectra comparison with the library NIST and Wiley (only
peaks with mass spectra similarity from at least 70% were considered),
elution order and Van den Dool and Kratz LTPRI (Linear Temperature
Programmed Retention Indexes) indexes. These indexes are obtained at
temperature gradient conditions and not isothermally, which provides a
more robust index for identification purposes.
For the calculation of LTPRIs it was used a standard mixture of n-
alkanes (C6-C30) injected under the same conditions used for the sam-
ples. The LTPRIs were calculated for each compound in the samples
Fig. 1. Distribution of the product yields (crude bio-oil, biochar and gases) from the pyrolysis of biomass samples.
3
Industrial Crops & Products xxx (xxxx) xxx–xxx
according to Van den Dool and Kratz’s equation (Van den Dool and
Kratz, 1963) (Eq. (1)) which is automatically calculated by the software
of the GC/MS system.
RIx = 100n + 100(tx − tn)/(tn+1 − tn) (1)
Where: RIx = retention index of the compound X
n = number of carbons of the linear alkane that appears just before
X in the chromatogram
t(n+1) = retention time of de linear alkane that appears just after X
in the chromatogram, and
tn = retention time of de linear alkane that appears just before X in
the chromatogram
The values obtained were compared to indexes found in the NIST
web site (NIST Mass Spectral Library).
A semi-quantitative analysis using only the relative area of each
compound was applied for each chromatogram. This is not a complete
quantitative analysis but can indicate a comparative distribution of the
amount of each compound in each sample. For this semi-quantitative
analysis, it was used Eq. (2) below.
Ai
Area(%) = x100
At (2)
Where At is the area of the total chromatographic peaks in the sample
and Ai is the area of the tentative identified compound.
2.4. FTIR analysis of biomasses and chemometric approach
The infrared spectra of biomass samples were obtained in a
PerkinElmer 400 IR spectrometer (PerkinElmer, USA) using a universal
attenuated total reflectance (ATR) sampling accessory. Absorbance was
measured in the spectral range of 700 and 4000 cm−1. The grinded
biomass samples were directly analyzed. The spectra were acquired in
random order, in triplicate, with 24 scans and resolution of 4 cm−1.
The softwares Chemostat (Helfer et al., 2015), Matlab (Mathworks
Inc, USA) and PLS_toolbox 6.2.1 (Eigenvector Research Inc, USA) were
used for multivariate analysis. The standard normal variate (SNV) was
applied to the spectra for removing vertical shifts, before multivariate
analysis, and the spectra were mean-centered. The PCA was performed
in the pre-processed spectra (range of 700–4000 cm−1) to investigate
the similarities between the samples (Bro and Smilde, 2015).
E. Lazzari et al.
3. Results and discussion
3.1. Pyrolysis process: product yields
Fig. 1 shows the yields from the products of the pyrolysis of biomass
samples.
As observed in Fig. 1, the highest crude bio-oil yields obtained were
45% and 47% to ES and SCB, respectively. Other authors have also
achieved good bio-oil yield for ES (49 mass% Torri et al., 2016) and (50
mass% Faccini et al., 2013). The SCG, CTS and PC showed crude bio-oil
yields near 40%. For CF, RH and PS the crude bio-oil yields obtained
were 33%, 35% and 30%, respectively. Other biomasses showed crude
bio-oil yields between 25% and 29%. Only the AB shows crude bio-oil
yield below 20%.
The bio-oil yields presented in our study showed to be compatible to
the ones reported for AMS (28.1 mass% Lazzari et al., 2016), PC (32.4
mass% Moraes et al., 2012), RH (27.3 mass% Moraes et al., 2012) using
the same bed-fixed reactor.
The bio-char yields for different biomasses indicated that RH and AB
exhibited the maximum yield (approximately 35%) and the SCB ex-
hibited the minimum yield (16%), whereas the other biomasses ex-
hibited yields of solid product between 21% and 32%. Regarding to the
gas products, the maximum yield was found to CS (52%) and the
minimum yield was found to RH (30%).
In a general way, the mass yields varied in the follow ranges: solid
product (from 16 to 35%), crude bio-oil (from 14 to 47%) and gas
products (from 30 to 52%).
3.2. Bio-oils characterization by GC/qMS
As mentioned above the characterization of bio-oils were performed
by GC/qMS.
Around two hundreds of different compounds were identified in the
fifteen bio-oil samples, using the procedure discussed in Section 2.3.
From these, 166 compounds were confirmed by LTPRI system with a
good agreement between LTPRIs calculated and literature values (dif-
ferences lower than 15). The compounds identification is improved by
combining mass spectra analysis with retention indexes comparison
(LTPRI system) (Lazzari et al., 2016; Djokic et al., 2012). When values
of LTPRIs did not match with those from the database used in this
study, the compounds were identified in a general way without a clear
definition of their isomeric positions according to Onorevoli et al.
(2014). Each compound identified was classified into the following
categories: phenols, ketones, aldehydes, furans, nitrogen compounds,
alcohols, ethers, esters, carboxylic acids and hydrocarbons. The list of
tentatively identified compounds by GC/qMS is summarized in Table S1
(Supplementary Electronic Material).
As expected, high differences in the compositions of bio-oils were
found due to the distinct chemical composition of the biomasses used.
This fact reinforces the utility of the proposed study of previously
characterization of the biomass using a simpler and inexpensive as the
FTIR, before the pyrolysis, allowing a pre-definition of the quality of the
bio-oil that will be produced.
On the other hand, a great predominance of phenols: four alkylated
phenols (phenol, ortho and para cresol and ortho ethyl phenol) were
identified in all the fifteen bio-oil samples, among the major com-
pounds. These compounds are present in the bio-oils independently of
the feedstock composition. Other compounds as guaiacol, dimethyl
phenol, hydroxy-methyl-cyclopentenone, dimethyl- cyclopentenone
and hexadecanoic acid also appeared in most of the samples (fourteen
of fifteen bio-oil samples).
For comparative purposes, a list of five major identified compounds
in each of bio-oil samples is presented in Table 2.
3.2.1. Semi-quantitative analysis
Regarding the semi-quantitative analysis, Fig. 2 shows the chemical
4
Industrial Crops & Products xxx (xxxx) xxx–xxx
classes distribution according relative areas (relative concentration) in
the fifteen bio-oil samples.
As can be observed in Fig. 2, the classes of phenols, carboxylic acids
and ketones appeared in all bio-oil samples while alcohols and ethers
appear only in four (CS, PL, CP and AB) and three (PC, PS and ES) bio-
oil samples, respectively. Phenols and carboxylic acids were the major
classes for most analyzed bio-oil samples. The compounds belonging to
the class of aldehydes showed selectivity with respect to the different
samples, which appeared in only eight bio-oil samples (RH, PC, CF, PS,
ES, AMS, SCB and PL). Among these samples, the bio-oil derived from
SCB showed highest content of aldehydes (14.6%).
It was also possible note, in the semi-quantitative analysis, simila-
rities among the bio-oils composition. The Fig. 3 shows the distribution
of classes of phenols (including alkylphenols, guaiacols and catechols),
carboxylic acids, hydrocarbons and nitrogen containing compounds (N-
compounds) in bio-oils.
One can observe in Fig. 3 that the bio-oils from AMS, CTS, SCG and
CS are characterized by a high amount of carboxylic acids, which re-
present from 30 to 70% of the sample. Hydrocarbons also appear in
high amount in these samples. In contrast, the bio-oils derived from CF,
RH, PC, PS, ES, SCB and PL showed composition predominantly of
phenols (including alkyl phenols, guaiacols and catechols), representing
at least 40% of the total bio-oil sample. In the case of ETS, CP, CSK and
AB, the apparent bio-oil distinction, compared to other biomasses, is the
significant amount of N-compounds (from 6.4 to 25.4% of the sample).
The high differences in the compositions of bio-oils is due to the
distinct chemical composition of the biomasses used in pyrolysis pro-
cess. Lignin, cellulose and hemicellulose with other components such as
proteins and triglycerides appear in differing proportions in various
biomass types and these proportions influence in composition of bio-oil
obtained by pyrolysis (Sharma et al., 2015). Thus, the predominantly
phenolic composition of CF, RH, PC, PS, ES, SCB and PL bio-oils is due
to the content of lignin in these biomasses. Lignin, which is composed of
three phenylpropane units, by depolymerization leads to various phe-
nols (e.i. benzenediols, methoxyphenols, alkylphenols) (Collard and
Blin, 2014). In contrast, the large amount of carboxylic acids and hy-
drocarbons in the AMS, CTS, SCG and CS bio-oils is due to the higher
contents of triglycerides in these biomasses. Triglycerides degradation
gives carboxylic acids, which via decarboxylation produce hydrocarbon
(alkanes and alkenes) plus CO, CO2 and H2O (Silva et al., 2014). Fi-
nally, the significant amount of N-compounds in the ETS, CP, CSK and
AB bio-oil is possibly due to the presence of proteins in these biomasses.
The degradation of amino acids, which constitute the proteins, gives N-
compounds, such as single ring N-heterocyclic compounds (Silva et al.,
2014).
Some research work has been published regarding bio-oils compo-
sitions from biomasses studied in this paper. Onorevoli et al. (2014), Li
et al. (2014) and Lazzari et al. (2016) have also described the high
amount of carboxylic acids found in bio-oils derived from CS, SCG and
AMS, respectively. Phenols were also found to be the major class in bio-
oils from RH (Moraes et al., 2012), CF (Almeida et al., 2013; Bispo
et al., 2016), ES (Faccini et al., 2013), and PC (Moraes et al., 2012).
Muradov et al. (2010) and Pattiya (2011) described the characteriza-
tion of bio-oil from AB and CP, also identifying high amount of N-
compounds.
3.3. FTIR analysis of biomasses and chemometric approach
3.3.1. Biomasses spectra
The FTIR spectra of the biomass samples are shown in Fig. 4. De-
tailed bands positions and their assignments are listed in Table 3 (Liu
et al., 2016; Chen et al., 2010; Xu et al., 2013).
In Fig. 4 can be visualized that the most absorption bands were
observed in the “fingerprint” region. In general, the vibration at
1745 cm−1 is originate from the C]O stretching modes of free ester.
The broad band at 1620 cm−1 is mainly attributed to the OH bending
E. Lazzari et al. Industrial Crops & Products xxx (xxxx) xxx–xxx

Table 2
Five major identified compounds in bio-oil samples.

Samples Compounds/area Five major compounds (area%) in each sample

%

AB compounds Hexadecanoic acid Phenol, 4-methyl- Phenol Indole Phenol, 3-ethyl-

area% 15.70% 11.43% 11.38% 3.71% 3.44%
AMS compounds 9-Octadecenoic acid, Phenol Hexadecanoic acid Phenol, 4-methyl- 9-Octadecenoic
isomer acid,isomer
area% 21.80% 7.6% 5.92% 4.78% 3.42%
CF compounds Phenol Dodecanoic acid Phenol, 4-methyl- Phenol, 2-methoxy- 1,2-Benzenediol
area% 23.13% 8.29% 6.07% 4.64% 4.39%
CP compounds Hexadecanoic acid Phenol Phenol, 4-methyl- Phenol, 4-ethyl- Indole
area% 8.17% 8.02% 7.21% 3.86% 3.31%
CS compounds Docosenoic acid 9-Octadecenoic acid, Docosanoic acid 9-Octadecenoic acid, isomer Hexadecanoic acid
isomer
area% 13.79% 9.88% 4.97% 4.54% 3.23%
CSK compounds Caffeine Phenol Hexadecanoic acid Phenol, 2-methoxy- Indole
area% 15.12% 7.5% 5.85% 4.04% 3.04%
CTS compounds Hexadecanoic acid 9-Octadecenoic acid, 9-Octadecenoic acid, isomer Octadecanoic acid Hexadecanamide
isomer
area% 45.29% 16.48% 4.36% 2.74% 2.36%
ES compounds Phenol, 4-methyl- Phenol, 2-methoxy-5- Phenol Phenol, 2-methoxy- 1,2-Benzenediol, 4-
methyl- methyl-
area% 9.11% 7.16% 6.32% 5.02% 4.99%
ETS compounds Phenol, 4-methyl- Phenol Phenol, 3-ethyl- Benzofuran, 2,3-dihydro- Indole
area% 17.44% 12.22% 10.25% 6.09% 5.76%
PC compounds Phenol, 2-methoxy-5- Phenol, 2-methoxy- 1,2-Benzenediol Methoxy-propenyl, Phenol- Phenol, 4-methyl-
methyl-
area% 7.6% 7.5% 7.42% 5.43% 4.54%
PL compounds Benzofuran, 2,3- Phenol Phenol, 4-ethyl- Ethanone, 1-(2-hydroxy-5- Phenol, 4-methyl-
dihydro- methylphenyl) (4.88%)
area% 9.67% 7.20% 7.09% 4.39%
PS compounds Phenol, 2-methoxy- Phenol, 2-methoxy-5- Ethanone, 1-(2-hydroxy-5- Methoxy-propenyl, Phenol- Phenol, 4-ethyl-2-
methyl- methylphenyl)- 8.24% methoxy-
area% 15.96% 8.5% 7.04% 6.38%
RH compounds Benzofuran, 2,3 Phenol Phenol, 4-methyl Phenol, 4-ethyl Phenol, 2-methoxy
dihydro
area% 8.74% 8.41% 7.23% 6.64% 6.61%
SCB compounds Benzofuran, 2,3- HMF Phenol, 4-ethyl- C2-Imidazolidinone Phenol, 4-methyl-
dihydro-
area% 14.00% 11.27% 6.26% 5.60% 5.09%
SCG compounds Hexadecanoic acid Octadecanoic acid 9-Octadecenoic acid, isomer Phenol 9-Octadecenoic
acid,isomer
area% 47.13% 7.93% 3.15% 3.71% 2.80%

mode of adsorbed water and protein amide I (C]O) (Liu et al., 2016).
The vibration at 1682 cm−1 is due to the C]O stretching and have
been associated with lignin bonds (Chen et al., 2010).
Bands at 1608, 1500 and 1313 cm−1 are close to the absorptions of
aromatic ring vibration of lignin subunits, and bands at 1327 cm−1
(CeO) also are associated to the absorption lignin subunits. Vibration at
1235 cm−1 results from the OH/NH deformation modes in proteins (Liu
et al., 2016). Bands at 1160 cm−1 is due to the CeOeC stretching
asymmetric results to cellulose and hemicellulose and the strong vi-
bration at 1032 cm−1 are attributed of the CeO, C]C and CeCeO

Fig. 2. Distribution of the chemical classes identified in bio-oils using GC/qMS technique.

5
E. Lazzari et al. Industrial Crops & Products xxx (xxxx) xxx–xxx

Fig. 3. Distribution of classes of phenols (including alkylphenols, guaiacols and catechols), carboxylic acids, hydrocarbons and N-compounds in bio-oil samples.

stretching and have been associated with cellulose, hemicellulose and Table 3
lignin (Xu et al., 2013). Chen et al. (2010), used the absorption bands Main adsorptions on FTIR spectra of biomasses.
associated with lignin successfully for differentiation of wood sample
absorption bands origin
into hardwood and softwood. This classification was according to bio-
mass lignin content. 1745 cm−1 C]O stretching free ester (oil components)
It is also seen in Fig. 4 little bands in the region of 2000–4000 cm−1. 1620 cm−1 OH bending water and amide (C]O) in
Among these bands, the broad bands in 3000–3600 cm−1 region are proteins
1682 cm−1 C]O stretching lignin bonds
assignable to OH stretching vibrations related to cellulosic materials or 1608, 1500 cm−1 C]C aromatic lignin subunits,
water. The weak bands at 2850 and 2924 cm−1 are due to the CH2 1327 cm−1 (CeO) lignin subunits.
symmetrical and asymmetrical stretching, respectively (Liu et al., 1235 cm−1 OH/NH deformation proteins
2016). 1160 cm−1 CeO-C stretching cellulose and hemicellulose
1032 cm−1 (strong) CeO, CeCeO cellulose, hemicellulose and
The feature spectral demonstrates that the vast difference in the
stretching lignin.
biomass compositions generates characteristic absorption bands for 3000–3600 cm−1 OH stretching cellulosic materials or water.
each sample (Fig. 4). Clearly, CS and ETS have spectral feature quite 2850 and 2924 cm−1 CH2 stretching, aliphatic bonds
different from those of other biomass. In addition, it is observed that the
spectra of CS, CTS and CSG show much pronounced absorption bands at
1745, 2850 and 2924 cm−1 indicating the presence of fatty acids in the
samples (Liu et al., 2016).

Fig. 4. FTIR spectra from fifteen biomasses with the main absorption
bands highlighted.

6
E. Lazzari et al.
Fig. 5. PCA of the FTIR spectra (4000–700 cm−1 region) of fifteen biomass samples: (a) score plot of PC2 (18.09%) vs PC3 (14.21%) and (b) loadings profile of PC2 and PC3 with the
bands that mostly contributed to the separation of biomasses highlighted. (1) 3300 cm−1 assignable to OH stretching, (2) 2850 and 2924 cm−1 associated to the CH2 stretching, (3)
1745 cm−1 related to the C]O stretching and (4) 1620 cm−1 related to the protein amide.
Although some differences were observed, FTIR spectra contain
small differences among the samples that could not be distinguished
visually, thus, chemometric approach are necessary to FTIR spectra
analysis.
3.3.2. PCA analysis
From FTIR spectra and using the adequate software, it was possible
to do a multivariate analysis, performing a PCA intending to investigate
some similarities and differences between the samples. In this study the
PCA was used to discriminate the biomass according their bio-oil
composition.
Fig. 5(a) shows the scores plot in the plane of the second and third
principal components (PCs) for the biomass samples, which together
account for 32.3% of the total variability and Fig. 5(b) shows the
loadings plot using PC3 and PC2 versus the variable number (sub-
spectrum of the data set). In the scores plot can be seen the re-pro-
ductivity of the data, in which the triplicates of analysis are closed in
the plot. The loading plot give information on what absorption bands
(wavenumber) mostly contributed to the separation of biomass samples
according the PC2 and PC3 values.
As observed in Fig. 5(a), the plot of PC3 versus PC2 provides a better
visualization of separation among the samples allowing the qualitative
recognition of different groups of samples. This classification was ac-
cording to bio-oils compositions obtained by GC/qMS.
Clearly, the PC3 (Fig. 5(a)), with 14.21% of the total variance,
identify three sample groups: G1 (represent by red points), G2 (re-
present by green point) and G3 (represent by blue point). The first
group (G1), formed by AMS, CTS, CS and SCG biomasses, has a high
positive score in the PC3. Among biomasses belong to G1, AMS showed
the lowest positive PC3 score. The second group (G2) have the scores in
the PC3 around zero and it is formed by CF, SCB, PS, PL RH, ES and PC
biomasses. It is also interesting note that CF has positive PC3 score in
this group. At last, the third group (G3) has negative score in the PC3
and it is formed by AB, CSK, CP e ETS biomasses.
The loadings of PC3 (Fig. 5(b) (red line)) showed a high positive
influence of the bands at 2850 and 2924 cm−1 associated to the CH2
symmetrical and asymmetrical stretching and of the bands at
1745 cm−1 related to the C]O stretching modes, which may corre-
spond to the fatty acids absorption bands. The PC3 negative side have a
high influence of the band at 1620 cm−1 which may be related to
amides (C]O), indicating the presence of proteins in samples with PC3
negative scores.
The PC2 (Fig. 5(a)), with 18.09% of the total explained variance,
depicted a tendency of separation between these groups: G1 is around
7
Industrial Crops & Products xxx (xxxx) xxx–xxx
zero, G2 in the positive side and G3 in the negative side. However, some
samples have different scores in this PC related to its own group: ES and
PC in the G2 and AB in the G3.
The loadings of PC2 (Fig. 5(b) (black line)) shows a high positive
influence of the broad bands in 3300 cm−1 assignable to OH stretching
vibrations modes, the bands at 1620 cm−1 attributed to the amide (C]
O) and the bands at 1032 cm−1 results from the CeO, C]C and
CeCeO stretching. The PC2 negative side have a high influence of the
bands at 1160 cm−1 related to CeOeC stretching asymmetric and
spectral region below than 1000 cm−1.
3.4. Correlation of GC/qMS, FTIR and PCA data
Using the chromatographic information (Figs. 2 and 3, Table S1)
and FTIR data together with PCA (Table 3 and Fig. 5a,b) it was possible
to propose some correlations. The Fig. 6 relates the separation of the
biomass samples in the PCA with their bio-oil composition obtained by
GC/qMS analysis.
The biomasses that belong to G1 in PCA showed bio-oils char-
acterized by a large amount of carboxylic acids and hydrocarbons in
GC/qMS analysis. The carboxylic acids in bio-oil are derived from
glycerides degradation that, via decarboxylation, produces hydro-
carbons plus CO, CO2 and H2O (Silva et al., 2014). The amount of
carboxylic acids found in bio-oils is in accord with the absorption bands
(1745, 2850 and 2924 cm−1) which may be related to fatty acids and
are associated with PC3 positive score of G1 samples. It is also inter-
esting note that in G1, AMS showed bio-oil composition with lowest
amount of acids carboxylic and highest amount of phenols and ketones,
these results are in accordance with less positive PC3 score of the AMS
in G1.
In contrast, the biomasses that belong to G2 in PCA analysis showed
bio-oils composition consisting predominantly of phenols which re-
presenting at least 40% of the bio-oils sample, as can be observed in
Figs. 2 and 3. It is also seen in these samples a significant amount of
ketones. The phenolic and ketones compounds are mainly derived from
lignin and hemicellulose degradation, respectively (Kim, 2015; Collard
and Blin, 2014). Among the biomasses belong to G2, CF showed bio-oil
composition with highest amount of carboxylic acids (20%), this result
is in agreement with the PC3 positive score of the CF sample in G2.
In the case of biomasses that belong to G3 their bio-oils showed high
diversity of chemical classes and a significant amount of N-compounds,
which is due to amino acids degradation (Silva et al., 2014). The sig-
nificant amount of N-compounds in biomasses on G3 group is according
to the absorption bands at 1620 cm−1 which may be related to proteins
E. Lazzari et al.
Fig. 6. Correlation between biomass PCA analysis and their bio-oil chemical composition obtained by GC/qMS.
and are associated with PC3 negative score of the G3 samples.
With regard to the PC2, is observed negative scores manly to the ES
and PC as well as the positive scores to the AB. The band at 3300 cm−1
assignable to the OH stretching vibrations related to cellulosic materials
or water absorbed are associated with PC2 positive score. It suggests
that among the biomass samples, the AB, due to be an aquatic biomass,
shows major content of water that may have influenced in its positive
PC2 scores. This PC2 separation can be related to yield results (Fig. 1).
Depending on the composition of the bio-oil it can be used for dif-
ferent applications. Thus, the bio-oils rich in phenols derived from
biomasses that belong to G2, can be used as a substitute for fossil
phenols, for example in resins which are widely used in industry
(Czernik and Bridgwater, 2004; Kim, 2015). While bio-oils with a sig-
nificant amount of carboxylic acids and hydrocarbons derived from
biomasses that belong to G1 are more attractive for use as a biofuel
after upgrading process. On the other hand, bio-oils with a great
amount of nitrogen compounds, derived of biomasses that belong to G3,
can be used as alternative sources to important compounds, for example
pyridines are important medicament precursors (Geraldo et al., 2010).
Previous studies (Liu et al., 2016; Chen et al., 2010; Everard et al.,
2012) also demonstrated the successful application of PCA on the FTIR
spectra to distinguish biomass samples, but this is the first study where
FTIR spectra data associated with PCA has been used to classify biomass
through their bio-oil composition to our knowledge.
4. Conclusions
FTIR in combination with PCA was employed for the first time to
classify biomass through their bio-oil composition. Bio-oils were pro-
duced by pyrolysis for the fifteen biomass source available in Brazil and
its chemical characterization was observed using GC/qMS. The bio-oils
showed high differences compositions and only four compounds which
belonging to the phenols class were identified in all the samples. The
use of PCA has enabled us to discriminate biomasses into three main
groups, which showed distinct bio-oil compositions, by their infrared
spectral data. These data revealed that carboxylic acids were found to
be the major class in bio-oils derived from biomasses belongs to group
1, phenols were the major class in group 2 and a significant amount of
nitrogen compounds were found in group 3. The knowledge derived
from this paper would benefit the use of spectroscopy along with che-
mometrics as an option to the more expensive chromatographic tech-
niques for exploratory purposes in bio-oil production.
8
Industrial Crops & Products xxx (xxxx) xxx–xxx
Acknowledgments
Authors thank to Petrobras, CNPq, FINEP and CAPES for financial
support.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.indcrop.2017.11.005.
References
Almeida, T., Bispo, M.D., Cardoso, A.R.T., Migliorini, M.V., Schena, T., Campos, M.C.V.,
Machado, M.E., López, J.A., Krause, L.C., Caramão, E.B., 2013. Preliminary studies of
bio-oil from fast pyrolysis of coconut fibers. J. Agric. Food Chem. 61, 6812–6821.
Bispo, M.D., Barros, J.A.S., Tomasini, D., Primaz, C.T., Caramão, E.B., Dariva, C., Krause,
L.C., 2016. Pyrolysis of agroindustrial residues of coffee, sugarcane straw and co-
conut-fibers in a semi-pilot plant for production of bio-oils: gas chromatographic
characterization. J. Earth Sci. Eng. 6, 235–244.
Bridgwater, A.V., 2012. Review of fast pyrolysis of biomass and product upgrading. Biom.
Bioenergy 38, 68–94.
Bro, R., Smilde, A.K., 2015. Principal component analysis. Anal. Methods 6, 2812–2831.
Chadwick, D.T., McDonnell, K.P., Brennan, L.P., Fagan, C.C., Everard, C.D., 2014.
Evaluation of infrared techniques for the assessment of biomass and biofuel quality
parameters and conversion technology processes: a review. Renew. Sust. Energy Rev.
30, 672–681.
Chen, H., Ferrari, C., Angiuli, M., Yao, J., Raspi, C., Bramanti, E., 2010. Qualitative and
quantitative analysis of wood samples by Fourier transform infrared spectroscopy and
multivariate analysis. Carbohydr. Polym. 82, 772–778.
Collard, F.-X., Blin, J., 2014. A review on pyrolysis of biomass constituents: mechanisms
and composition of the products obtained from the conversion of cellulose, hemi-
celluloses and lignin. Renew. Sust. Energ Rev. 38, 594–608.
Czernik, S., Bridgwater, A.V., 2004. Overview of applications of biomass fast pyrolysis oil.
Energy Fuel 18, 590–598.
Djokic, M.R., Dijkmans, T., Yildiz, G., Prins, W., Geem, K.M.V., 2012. Quantitative ana-
lysis of crude and stabilized bio-oils by comprehensive two-dimensional gas-chro-
matography. J. Chromatogr. A 1257, 131–140.
Everard, C.D., McDonnell, K.P., Fagan, C.C., 2012. Prediction of biomass gross calorific
values using visible and near infrared spectroscopy. Biomass Bioenergy 45, 203–211.
Faccini, S.C., Dalla Vecchia, I., Ribeiro, D., Zini, C.A., Caramão, E.B., 2013.
Comprehensive two-dimensional GC with TOF-MS detection: study of pyrolytic bio-
oil of kraft mill residues. J. Braz. Chem. Soc. 24, 1085–1098.
Geraldo, R.B., Bello, M.L., Dias, L.R.S., Vera-Divaio, M.A., Nagashima, T., Abreu, P.A.,
Santos, M.B., Albuquerque, M.G., Cabral, I.M., Freitas, A.C.C., Kalil, M.V., Rodrigues,
C.R., Castro, H.C.A., 2010. Antiplatelet activity and structure-activity relationship
study of pyrazolopyridine derivatives as potential series for treating thrombotic
diseases. J. Ather. Thromb. 17, 2–11.
Helfer, G.A., Bock, F., Marder, L., Furtado, J.C., da Costa, A.B., Ferrão, M.F., 2015.
Chemostat, um software gratuito para análise exploratória de dados multivariados.
Quim. Nova 38, 575–579.
IBGE, available in http://www.sidra.ibge.gov.br/bda/prevsaf/. (Accessed January 2017).
Jahirul, M.I., Rasul, M.G., Chowdhury, A.A., Ashwath, N., 2012. Biofuels production
through biomass pyrolysis—a technological review. Energies 5, 4953–5000.
E. Lazzari et al.
Kim, J., 2015. Production, separation and applications of phenolic-rich bio-oil—a review.
Bioresour. Technol. 178, 90–98.
Lazzari, E., Schena, T., Primaz, C.T., Maciel, G.P.S., Machado, M.E., Cardoso, C.A.L.,
2016. Production and chromatographic characterization of bio-oil from the pyrolysis
of mango seed waste. Ind. Crop Prod. 83, 529–536.
Li, X., Strezov, V., Kan, T.J., 2014. Energy recovery potential analysis of spent coffee
grounds pyrolysis products. J. Anal. Appl. Pyrol. 110, 79–87.
Liu, Y., He, Z., Uchimiya, M., 2015. Comparison of biochar formation from various
agricultural by-products using FTIR spectroscopy. Mod. Appl. Sci. 9, 246–253.
Liu, Y., He, Z., Shankle, M., Tewolde, H., 2016. Compositional features of cotton plant
biomass fractions characterized by attenuated total reflection Fourier transform in-
frared spectroscopy. Ind. Crop. Prod. 79, 283–286.
Moraes, M.S.A., Migliorini, M.V., Damasceno, F.C., Georges, F., Almeida, S., Zini, C.A.,
Caramão, E.B.J., 2012. Qualitative analysis of bio oils of agricultural residues ob-
tained through pyrolysis using comprehensive two dimensional gas chromatography
with time-of-flight mass spectrometric detector. J. Anal. Appl. Pyrol. 98, 51–64.
Muradov, N., Fidalgo, B., Gujar, A.C., Raissi, T., 2010. Pyrolysis of fast-growing aquatic
biomass—Lemna minor (duckweed): characterization of pyrolysis products.
Bioresour. Technol. 101, 8424–8428.
NIST Mass Spectral Library, available in http://webbook.nist.gov/chemistry/. (Accessed
April 2017).
Onorevoli, B., Machado, M.E., Dariva, C., Franceschi, E., Krause, L.C., Jacques, R.A.,
Caramão, E.B., 2014. A one-dimensional and comprehensive two-dimensional gas
chromatography study of the oil and the bio-oil of the residual cakes from the seeds of
Crambe abyssinica. Ind. Crop. Prod. 52, 8–16.
Pattiya, A., Suttibak, S.J., 2012. Production of bio-oil via fast pyrolysis of agricultural
residues from cassava plantations in a fluidized-bed reactor with a hot vapor filtration
Industrial Crops & Products xxx (xxxx) xxx–xxx
unit. J. Anal. Appl. Pyrol. 95, 227–235.
Pattiya, A., 2011. Bio-oil production via fast pyrolysis of biomass residues from cassava
plants in a fluidised-bed reactor. Bioresour. Technol. 102, 1959–1967.
Rambo, M.K.D., Schimidt, F.L.S., Ferreira, M.M.C., 2015. Analysis of the lignocellulosic
componentes of biomass residues for biorefinery opportunities. Talanta 144,
696–703.
Schneider, J.K., Cunha, M.E., Santos, A.L., Maciel, G.P.S., Brasil, M.C., Pinho, A.R.,
Mendes, F.L., Jacques, R.A., Caramão, E.B., 2014. Comprehensive two dimensional
gas chromatography with fast-quadrupole mass spectrometry detector analysis of
polar compounds extracted from the bio-oil from the pyrolysis of sawdust. J.
Chromatogr. A 1356, 236–240.
Sharma, A., Pareek, V., Zhang, D., 2015. Biomass pyrolysis-A review of modelling, process
parameters and catalytic studies. Renew. Sust. Energy Rev. 50, 1081–1096.
Silva, R.V.S., Casilli, A., Sampaio, A.L., Ávila, B.M.F., Veloso, M.C.C., Azevedo, D.A.J.,
2014. The analytical characterization of castor seed cake pyrolysis bio-oils by using
comprehensive GC coupled to time of flight mass spectrometry. J. Anal. Appl. Pyrol.
106, 152–159.
Torri, I.D.V., Paasikallio, V., Faccini, S.C., Huff, R., Caramão, E.B., Sacon, V., Oasmaa, A.,
Zini, C.A., 2016. Bio-oil production of softwood and hardwood forest industry re-
sidues through fast and intermediate pyrolysis and its chromatographic character-
ization. Bioresour. Technol. 200, 680–690.
Van den Dool, H., Kratz, P.D., 1963. Generalization of the retention index system in-
cluding linear temperature programmed gas-liquid partition chromatography. J.
Chromatogr. A 11, 463–471.
Xu, F., Yu, J., Tesso, T., Dowell, F., Wang, D., 2013. Qualitative and quantitative analysis
of lignocellulosic biomass using infrared techniques: a mini-review. Appl. Energy
104, 801–809.