Ho

JMa

b

c

d

a

ARRAA

KGHHPMF

1

fmawd

tzhahsaHtm

h0

Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277

Contents lists available at ScienceDirect

Journal of Analytical and Applied Pyrolysis

journa l h om epage: www.elsev ier .com/ locate / jaap

ydrothermal conversion of grape pomace: Detailed characterizationf obtained hydrochar and liquid phase

elena Petrovic a, Nebojsa Perisic b, Jelena Dragisic Maksimovic c, Vuk Maksimovic c,ilan Kragovic a, Mirjana Stojanovic a, Mila Lausevic d, Marija Mihajlovic a,∗

Institute for Technology of Nuclear and Other Mineral Raw Materials, 86 Franchet d’Esperey St., 11000 Belgrade, SerbiaVistin Pharma AS, Stuttlidalen 4, Fikkjebakke, 3766 Sannidal, P.O. Box 98, NO-3791 Kragerø, NorwayInstitute for Multidisciplinary Research, University of Belgrade, 1 Kneza Viseslava St., 11000 Belgrade, SerbiaFaculty of Technology and Metallurgy, University of Belgrade, 4 Karnegijeva St., 11000 Belgrade, Serbia

r t i c l e i n f o

rticle history:eceived 20 October 2015eceived in revised form 10 February 2016ccepted 11 February 2016vailable online 18 February 2016

eywords:

a b s t r a c t

In this study, carbonization products of grape pomace (hydrochar and process water) have been thor-oughly characterized in order to assess its fuel properties, physico-chemical composition and to optimizeits production. The obtained detailed insight into transformations of the biomass during hydrothermalconversion between 180–220 ◦C revealed that the hydrochar obtained at 220 ◦C exhibits a considerableenergetic potential, increased porosity and re-adsorption abbility. Hydrothermally induced structuralchanges in the obtained hydrochars were unveiled by thermal and morphology analysis, FTIR and NIR

rape pomaceydrothermal carbonizationydrocharrocess-waterultivariate analysis

TIR

spectroscopy. Temperature increment caused a decrease in antioxidative capacity, anthocyanin andorganic acid content in process water and simultaneous increase in total phenolic and individual organiccomponents content. The overall effect of the reaction temperature on products characteristics wasassessed by multivariate data analysis. Obtained results substantiated the suitability of hydrothermalconversion of grape pomace into highly valuable fuels and versatile products.

© 2016 Elsevier B.V. All rights reserved.

. Introduction

Hydrothermal carbonization (HTC) is a promising technologyor conversion of wet lignocellulosic biomass into highly functional

aterials [1]. HTC process is carried out in a suspension of biomassnd water at saturated pressure, whereby subcritical water reactsith fibrous components of lignocellulosic biomass, leading to itsegradation and reconstruction [2].

HTC treatment of biomass (associated with hydrolysis, dehydra-ion, decarboxylation, condensation, polymerization and aromati-ation) [3], generates two main products: (i) insoluble carbon-richydrochar (HC) and (ii) process water (PW), whereas only smallmounts of gases are formed [4]. Hydrochar is a homogeneous,ydrophobic, energy-dense solid, which contains micro- to nano-ized carbon spheres, oxygen-containing functional groups and has

porous structure [2]. Due to its physico-chemical characteristics,

C has great potential for numerous practical applications. So far

his material has been tested for use as solid fuel [1,4], soil supple-ent [5], as feedstock for pellets [6] etc. On the other side, the PW

∗ Corresponding author. Fax: +381 11 3691 722.E-mail address: m.mihajlovic@itnms.ac.rs (M. Mihajlovic).

ttp://dx.doi.org/10.1016/j.jaap.2016.02.010165-2370/© 2016 Elsevier B.V. All rights reserved.

from HTC process is comprised mostly of source-related organicacids and other intermediate products such as furfurals, phenolsand monomeric sugars [7]. Various nutrients from biomass mayalso be present. However, subsequent usage of PW is question-able because it may as well contain potentially genotoxic and/orcytotoxic substances [5].

Until now, HTC has been employed on a wide range of dif-ferent biomass feedstock [3,6,8] and it has been shown that thecharacteristics of HTC products are strongly dependent on reactionconditions and feedstock type.

One of raw materials that is highly suitable for HTC processingis grape pomace (GP), which is a wet biomass (>60% water) that iscomposed of seeds, skins, and stems remaining after grape process-ing. GP is produced in significant amounts by wine industry globallyas a 20–25% feedstock waste [9]. Regardless of its reputationas environment-friendly, current inadequate waste managementpractice within the wine industry induces a large number of envi-ronmental issues such as production and handling of solid wastestream, energy usage, generation of greenhouse gas emissions etc.

[10]. In order to reduce these negative environmental impacts,appropriate utilization of GP is indispensable. There are severalpotential applications of untreated GP, such as animal feed [11] andsoil fertilizer [12], while several studies have examined its antiox-

2 and A

iHmvcaieGaemaaw

isnaHcttFsfLstaRtwat

2

2

gBdws

2

NiR1aioswtvsw

68 J. Petrovic et al. / Journal of Analytical

dant activity [13] and its suitability for biofuel production [14].owever, they all require drying and storage of pomace, whichakes its utilization costly and complex, thus hindering the moti-

ation for its reuse. This is where the HTC has the biggest advantageompared to other carbonization processes such as torrefractionnd pyrolysis [9]: application of wet biomass, relatively low operat-ng temperatures, the absence of CO2 emission and high conversionfficiency. Besides the emphasis on the energetic potential of theP-HC [9,15], literature regarding comprehensive characterisationnd potential utilization of HTC-GP process streams is scarce. Forconomically beneficial GP-HTC process and efficient treatmentethods of PW, a detailed characterization of compounds obtained

long the HTC process streams is highly needed. In this way anppropriate and more sustainable reuse of this vastly availableaste material could be more effectively outlined.

The aim of this study was to provide a detailed insightnto the physico-chemical and fuel properties of HTC productstreams of GP in relation to different HTC temperatures. Theovelty of this study is reflected in the holistic multi-methodpproach, where the raw material (GP), solid products (GP-C) and liquid products (GP-PW) are characterised. Detailedharacterization of GP-HCs was performed by Atomic Absorp-ion Spectrometry (AAS), Thermal gravimetric analysis/differentialhermal analysis (TGA/DTA), Scanning electron microscopy (SEM),ourier-transform infrared spectroscopy (FTIR) and Near-infraredpectroscopy (NIR). Detailed characterization of GP-PWs was per-ormed by using High-performance liquid chromatography (HPLC,C–MS), FTIR and UV–Vis spectroscopy. Since the literature on thepectroscopic characterisation of the biomass materials is still inhe developing phase, multivariate data analysis methods suchs Principal Component Analysis (PCA) and Partial Least Squareegression (PLSR) were applied. This was done in order to bet-er understand the HTC process and correlate spectroscopic dataith structural and chemical changes in GP-HTC product streams,

s well as to provide additional interpretation of the obtained spec-roscopic data.

. Materials and methods

.1. Biomass

The red grape was grown on a test plot Radmilovac near Bel-rade, Serbia (property of the Faculty of Agriculture, University ofelgrade). The used GP (skin, stalk and seeds) was sampled ran-omly from landfill sites. The biomass was air-dried until constanteight and extensively grinded in order to obtain homogenous

amples. Sieved fraction of 0.5 mm was used in HTC experiments.

.2. HTC experiment

HTC process was carried out in 2000 mL autoclave (Deutsch &eumman, model 10253, Germany) equipped with external and

nternal thermometers, analogue manometer and a cooling system.eaction load was composed of 250 g of raw biomass mixed with250 mL of distilled water (5:1 mass ratio) [15], which ensures andequate HC yield and effective stirring of the mixture. A heat-ng rate of 2 ◦C min−1 was applied to reach reaction temperaturesf 180, 200 and 220 ◦C (with starting temperature of 20 ◦C). Theet temperatures were maintained constant for 60 min. During thehole experiment, the stirrer was operated at 60 rpm. Afterwards,

he reactor was cooled to room temperature, remaining gas wasented and solid and liquid products were collected. The HC waseparated from liquid by filtration, rinsed three times with distilledater and dried at 105 ◦C for 24 h. After weighing, samples were

pplied Pyrolysis 118 (2016) 267–277

placed in a sealed container. The PWs were collected in glass bottlesand stored at 4 ◦C.

2.3. Physico-chemical characterization of GP and HCs

Proximate analysis of moisture (wt%), volatile matter (VM), andash were determined in three replicates per sample by standardprocedure ASTM D1762-84 (2007). Fixed carbon (FC) was calcu-lated as difference from 100% [6]. Mass yield was defined as massratio of dried HCs and dried GP, multiplied by 100% [3].

Inorganic analysis (K, Mg, Ca, Na, Fe, Si, Pb, Cu, and Ni) wascarried out using AAS (Perkin Elmer, AAS Analyst 300). Sam-ples (in three replicates) were previously dissolved using thenitric-perchloric acid digestion method [16] and inorganics weredetermined directly from the solution. A part of undissolved Siwas subsequently determined by the hydrochloric acid dehydra-tion (gravimetric) technique [17]. Phosphorus and highly solublephosphorus were determined from acid digested samples solutionsusing UV–vis spectroscopy (Jena Analytic, Spekol 1300) [18].

Elemental analysis (C, H, N and S) of solid samples was per-formed in three replicates using Vario EL III; C, H, N, S/O ElementalAnalyzer equipped with a thermal conductivity detector (TCD).Operating ranges varied between the elements: 0.03–20 mg for C,0.03–3 mg for H, 0.03–2 mg for N and 0.03–6 mg for S. Oxygen con-tent was obtained by subtracting the sum of the obtained elementalvalues from 100%. Higher heating values (HHV), energy densifica-tion (ED) and energy yield (EY) of the samples were also calculated[9,19].

Thermal analysis (TGA-DTA) was performed on a Netzsch STA409 EP (Selb, Germany). Samples were heated from 20 ◦C to 1000 ◦Cin an air atmosphere at a heating rate of 10 ◦C min−1 and kept in adesiccator at relative humidity of 23%, prior to analysis.

Morphological studies were carried out on JSM-6610 JEOLscanning electron microscope. All samples were coated with gold,and placed on the adhesive carbon disc. After coating with gold,samples were placed under vacuum conditions and scanned.

Spectroscopic analysis: FTIR analysis of the GP and HCs wasperformed using a Thermo Scientific Nicolet iS50 FTIR spectrometerin transmission mode by producing KBr pastilles with 0.8 mg sam-ple and 80 mg KBr. The spectra were obtained in three replicates persample in the spectral range of 4000–400 cm−1. NIR analysis of theGP and HCs was performed using a Bruker Optics TANGO FT-NIRspectrometer, equipped with an integrating sphere. The sampleswere analysed in a reflection mode without any pre-treatment. Thespectra were obtained in three replicates per sample in the spectralrange of 11540–3950 cm−1.

2.4. Physico-chemical characterization of PWs

Inorganic analysis (K, Mg, Ca, Na, Fe, and Si) of each PWs wasperformed in triplicates by AAS (PerkinElmer, AAS Analyst 300).Phosphorus was determined by using UV–vis spectroscopy method(Jena Analytic, Spekol 1300) [18].

Spectroscopic analysis was performed by FTIR on the sameinstrument as for GP and HCs, by using the ATR technique, with dia-mond. Each PW sample was recorder in three replicates, with usingdistilled water as background, in order to emphasize the signals ofrelevant components.

Antioxidant assays: Total antioxidant capacity (TAC) of PWswas quantified by using UV–Vis spectrometer 2501 PC Shimadzu,Japan, at 730 nm following the ABTS method [20] and results wereexpressed as milligrams of ascorbic acid equivalent per millilitre

(mg AsA eq mL−1). Total anthocyanin content (TACY) was mea-sured by using UV–Vis spectrometer Multiscan Spectrum, Thermoelectron corporation, Finland, at 510 and 700 nm, following themodified pH differential absorbance method [21] and results were

and A

ecuaeG

risa(adTufi7t(ppptawfCmwntt(

2

2

atbNtapN

2

ebdlfi(

sduwb

UL

J. Petrovic et al. / Journal of Analytical

xpressed as �g cyanidin-3-glucoside equivalents per millilitre (�g-3-g eq mL−1). Total phenolic content (TPC) was determined bysing UV–Vis spectrometer 2501 PC Shimadzu, Japan, at 724 nmccording to the Folin-Ciocalteu’s procedure [22] and results werexpressed as milligrams of Gallic acid equivalent per millilitre (mgallic acid eq mL−1).

Individual organic components in PWs were determed byeversed phase HPLC and LC-MS analysis. Samples were injectedn Waters HPLC system consisted of 1525 binary pumps, thermo-tat and 717+ autosampler connected to the Waters 2996 dioderray and EMD 1000 Single quadropole detector with ESI probeWaters, Milford, USA). Separation of phenolics was performed on

Symmetry C-18 RP column 125 × 4 mm size with 5 �m particleiameter (Waters, USA) connected to appropriate guard column.wo mobile phases, A (0.1% formic acid) and B (acetonitrile) weresed at flow of 1 mL min−1 with the following gradient profile: therst 20 minutes from 10 to 25% B; next 40 min of linear rise up to5% B, followed by 10 minutes reverse to initial 15% B with addi-ional 5 minutes for column equilibration. Post column flow splitterASI, USA) with 5/1 split ratio was used to obtain optimal mobilehase inflow for ESI probe. For LC/MS analysis, signals for each com-ound were detected in negative ESI scan mode with followingarameters: capillary voltage 3.0 kV, cone voltage −35 V, extrac-or and RF lens voltages were 3.0 and 0.2 V respectively. Sourcend desolvation temperatures were 130 ◦C and 380 ◦C respectively,ith N2 gas flow of 400 L h−1. Separation of organic acids was per-

ormed on anion exchange column, Aminex HPX-87H (Bio-Rad Lab.,A) 300 × 7.8 mm. Isocratic elution was made with 5 mM H2SO4 asobile phase, with flow rate of 0.6 mL min−1 at 40 ◦C. DAD detectoras set to 210 nm. Each component was quantified by an exter-al standard method using pure compound standards, retentionimes and characteristic UV spectra. The data acquisition and spec-ral analysis were carried out by the Waters Empower 2 SoftwareWaters, USA).

.5. Data analysis

.5.1. Pre-processing of IR spectraFTIR spectra were corrected for the absorption of water vapour

nd CO2 by the Thermo Omnic software that was used for spec-ral acquisition. Afterwards, the FTIR spectra were pre-processedy Extended Multiplicative Signal Correction (EMSC) method. FT-IR spectra were standardised by vector normalization (SNV), and

he second derivative was calculated by applying Savitzky–Golaylgorithm, with window size of 21 smoothing points [23,24]. Pre-rocessing was performed by using The Unscrambler v10.1 (Camo,orway).

.5.2. Multivariate analysis of IR dataPCA is a multivariate analytical method that is often used in

xploratory data analysis, for which the extended explanation cane found elsewhere [25]. In this study the PCA was used to (i) outlineifferences between the analysed samples (by producing scores and

oadings plots) and (ii) to outline the causes of these differences bynding correlations between HTC temperatures and wavenumbersby producing correlation loadings plots).

PLSR is a multivariate analytical method that uses PCA as a firsttep on one data set (X) and thereby obtained results uses for pre-icting the variation in another data set (Y). In this study PLSR wassed to: (i) find correlations between the bands from NIR regionith the bands from mid-IR region, and (ii) to find correlations

etween the different HTC temperatures and the vibrational bands.All PCA and PLSR calculations were performed by using The

nscrambler v10.1 (Camo, Norway) and routines written in MAT-AB (v R2010a, The Math Works, USA).

pplied Pyrolysis 118 (2016) 267–277 269

3. Results and discussion

3.1. Grape pomace and hydrochars

3.1.1. Physico-chemical and fuel characteristicsDetermined composition and certain characteristics of the GP

and HCs after HTC at 180 ◦C (HC-180), 200 ◦C (HC-200) and 220 ◦C(HC-220) are shown in Table 1.

Observed reduction of solid yield and the VM in the samples withan increasing temperature may be attributed to dehydration anddecarboxylation of lignocellulosic biomass during thermal treat-ment [19,4]. Hydrothermal reduction of VM improves the efficiencyof solids’ direct combustion [1], thus boosting energetic poten-tial of the obtained HCs. Higher yields of fixed carbon observedin HCs additionally confirm high carbonization efficacy. Similarobservations were found for HCs of palm-empty fruit bunches [3].Anaerobically digested sludge showed smaller increases in FC con-tent during HTC than the GP [19], which put GP-HCs among easilyaccessible and energy-dense carbon materials.

During HTC treatment, subcritical water behaves as a non-polarsolvent with high efficiency of breaking �-(1-4) glycosidic bondsin hemicellulose and cellulose, producing a porous hydrochar [2].Since hemicellulose contains most of inorganic elements in ligno-cellulosic biomass, its degradation initiated at 180 ◦C leading toleaching of inorganics from biomass into the liquid phase [2,6].Accordingly, our results showed that the HC-200 had the low-est content of inorganics. However, at temperatures above 200 ◦C,some of the removed inorganics were subsequently reabsorbed ona porous HC surface, which was confirmed through the increasedash content and concentrations of Ca, Mg, P, Fe and Si in the HC-220 sample (Table 1). Similar to our results Reza et al. [2] reportedan increase of Ca, P and Mg in the HC of corn stover, while Poer-schmann et al. [7] observed an increase of Ca and P in carbonizedElodea nuttallii plants.

In contrast to this, the content of sulfur in HCs decreased withtemperature (Table 1) and thereby preventing generation andemission of harmful sulfur oxides, SOx, due to their dissolutionof under hydrothermal conditions [19]. This is a highly beneficialaspect of HTC during potential large-scale biomass-conversion pro-cesses. Amounts of Pb, Cu, Cd and Ni remained unchanged upon HTCof GP.

Highest amount of P and highly soluble P were observed in HC-220 (Table 1). Similarly, Kambo and Dutta [6] and Reza et al., [2]observed an increase of P in HCs obtained from miscanthus andcorn stover carbonized above 200 ◦C.

The ability of some biochars to adsorb leached phosphate ionsfrom the solution have been demonstrated in previous studies[26,27]. There are two proposed mechanisms of this phenomenon:(i) the formation of bridge bonds between phosphate ions and elec-trostatically attracted or ligand-bonded di- and tri-valent cations(Ca2+, Mg2+, Al3+ and Fe3+) on biochar surface, and (ii) the phosphateadsorption due to increased pore volume of biochar [26,27]. Boththe increased porosity (confirmed by SEM analysis given below)and the increased Ca, Mg and Fe content in the HCs obtained at220 ◦C (Table 1) may hold up hypothesis that herein observed re-adsorption of P by HC-220 follows the suggested mechanisms.

However, the amount of P in HC-220 is indeed higher, but stillinsufficient for direct utilisation of HC as a fertilizer. Therefore,a modification of GP-HC surface with specific cations in order toincrease its phosphates sorption ability might be a useful procedurefor its beneficial application as a soil amendment.

Fuel properties of HCs are obtained on basis of atomic H/C and

O/C ratios of the GP, which were further plotted in a van Krev-elen diagram (Fig. S1). Expectedly, H/C and O/C ratios decreasedwith temperature increase, as a result of water and carbon dioxide

270 J. Petrovic et al. / Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277

Table 1Chemical characteristics and fuel properties of the GP and tested hydrochars.

Parameter GP HC-180 HC-200 HC-220

Yielda (wt%) – 86 78 66

Elementalanal-y-sis(%)

Carbon 49.48 ± 0.18 56.14 ± 0.03 58.63 ± 0.07 60.51 ± 0.05Hydrogen 6.86 ± 0.03 6.92 ± 0.05 7.07 ± 0.09 6.50 ± 0.06Oxygenb 34.34 ± 0.19 31.19 ± 0.07 28.78 ± 0.13 23.87 ± 0.09Nitrogen 2.84 ± 0.04 2.07 ± 0.02 1.97 ± 0.06 2.37 ± 0.02Sulfur 0.22 ± 0.01 0.19 ± 0.00 0.16 ± 0.00 –HHVc (MJ kg−1) 21.64 ± 0.13 24.43 ± 0.06 25.72 ± 0.11 26.13 ± 0.08Energy densificationd – 1.13 ± 0.01 1.19 ± 0.01 1.21 ± 0.01Energy yielde (%) – 97.08 ± 0.62 92.71 ± 0.67 79.69 ± 0.52

Proximateanal-y-sis(wt%)

Moisture 0.74 ± 0.02 2.73 ± 0.02 4.12 ± 0.03 3.23 ± 0.04Volatiles 75.49 ± 0.05 67.76 ± 0.04 66.60 ± 0.10 63.38 ± 0.08Ash 6.48 ± 0.04 3.67 ± 0.02 3.55 ± 0.03 6.75 ± 0.04Fixed carbonb 17.29 ± 0.07 25.84 ± 0.05 25.73 ± 0.10 26.65 ± 0.09

Inorganics(%)

K 1.02 ± 0.01 0.34 ± 0.01 0.12 ± 0.02 0.62 ± 0.01Mg 0.12 ± 0.04 0.06 ± 0.01 0.05 ± 0.01 0.11 ± 0.02Ca 0.64 ± 0.02 0.59 ± 0.02 0.51 ± 0.04 0.93 ± 0.04Na 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00Fe 0.12 ± 0.02 0.12 ± 0.01 0.10 ± 0.01 0.27 ± 0.01Si 0.94 ± 0.04 0.91 ± 0.04 0.88 ± 0.02 1.06 ± 0.08Pb 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00Cu 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00 0.01 ± 0.00Cd <0.01 <0.01 <0.01 <0.01Ni <0.01 <0.01 <0.01 <0.01P 0.67 ± 0.03 0.52 ± 0.02 0.41 ± 0.06 0.84 ± 0.02Phighsoluble 0.35 ± 0.01 0.11 ± 0.01 0.23 ± 0.01 0.47 ± 0.01

a Yield = %Weight of the biomass was measured before and after the HTC.b Content was calculated by difference.

roiC(se

Itr

3T

meeptfipedao

meneiso

c HHV = 0.3491C + 1.1783H + 0.1005S−0.1034O−0.015N−0.0221Ash [9].d ED = HHVhydrochar/HHVfeedstock [3,19].e EY = (Mhydrochar × HHVhydrochar)/(Mfeedstock × HHVfeedstock) [3].

emoval during HTC [19,6]. Furthermore, HC reached a maximumf HHV at 220 ◦C of 26.13 MJ kg−1. In fact, all three GP-HCs exhib-ted better energetic potential (Table 1) then lignite (16.3 MJ kg−1).oncurrently, HC-220 had higher calorific value than coconut fiber24.7 MJ kg−1) and sewage sludge (18.3 MJ kg−1) carbonized at theame temperature [1,19], making GP a more promising and cost-ffective biomass material for solid biofuel production.

The ED ranged from 1.13 for the HC-180 to 1.21 for the HC-220.n contrast to this, EY decreased from 97.08% to 79.69% within aemperature range of 180 ◦C to 220 ◦C. Similar observations haveeported Liu et al. [1] for coconut fiber and eucalyptus leaves.

.1.2. Thermal behaviour and morphology of GP and HCs:GA-DTA and SEM analysis

In DTA diagram of the GP (Fig. S2), an endothermic peak withinimum at 107 ◦C was observed, which is attributed to moisture

vaporation and volatilization of small organic molecules. Threexothermic peaks are observed above 150 ◦C due to oxidation ofresent organic matter. From the literature [28] it is well knownhat during thermal treatment of the lignocellulosic materials, therst volatilized components are hemicellulose and cellulose com-onents, followed by decomposition of lignin. Thus, for GP thexothermic peak with maximum at 342 ◦C may be attributed toegradation of hemicellulose, at 542 ◦C to oxidation of cellulosend residues of hemicellulose, and the peak at 626 ◦C to oxidationf the lignin and residues of cellulose.

From DTA diagrams of the HCs (Fig. S2) it can be seen that ther-al properties of the GP changed significantly: the intensity of

ndothermic peak decreased with temperature, suggesting elimi-ation of weakly bound moisture and small organic molecules. The

xothermic peak with maximum at 342 ◦C was not observed in HCs,ndicating decomposition of hemicellulose during HTC. The inten-ity and position of the exothermic peaks derived from degradationf cellulose and residual hemicellulose with maximum at 542 ◦C

in GP was reduced and shifted to lower temperatures in HC-180(472 ◦C) and in HC-200 (531 ◦C), while completely absent in HC-220. That confirms the complete degradation of hemicellulose until200 ◦C and degradation of cellulose at higher temperatures. Finally,increasing the HTC temperature from 180 to 220 ◦C caused a shift ofthe exothermic peak at 626 ◦C attributed to oxidation of lignin andresidues of the cellulose to 574, 600 and 564 ◦C for HC-180, HC-200and HC-220, respectively.

SEM images of raw GP revealed a relatively heterogeneoussurface structure typical for lignocellulosic biomass (Fig. S3). Car-bonization disrupts the smooth structure of the biomass, creatingcracks and pores on the HC surface. Morphological display of HC-220 showed the highest porosity compared to other samples, whichconfirmed that progressive degradation of fibres with temperaturecauses a pronounced increase in porosity of HCs compared to GP.

3.1.3. Spectroscopic analysis of GP and HCs: FTIR and FT-NIRThe FTIR spectra of analysed GP and HC samples, averaged over

spectral replicates per sample are presented in Fig. 1a in relevantspectral regions (3700–2800 cm−1 and 1850–500 cm−1). Besidesvisual inspection of the spectra, PCA was performed in two ways inorder to investigate the overall effect of HTC on the structure of GP:(i) a PCA model was calculated on all samples together (hereafterreferred to as PCA-GP + HC), and (ii) another PCA model was calcu-lated without the GP samples in order to emphasise the detecteddifferences between the HTC samples (hereafter referred to as PCA-HC). Significance of FTIR bands are determined by inspecting theloadings plot. From the PCA-GP + HC scores plot (Fig. 1b) it can beseen that the effect of temperature dominates the variation alongPC1 (79% of explained variance): from GP samples on the left, over

the HC-180 samples in the centre to the HC-200 and HC-220 sam-ples on the right. This means that the PC1 contains all the variationcaused by HTC process and therefore only the loadings for PC1 arepresented in Fig. 1c. Further, in the PCA-HC scores plot (Fig. 1d) it

J. Petrovic et al. / Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277 271

F ) scors

cHl

iwabTe

ig. 1. Results of FTIR analysis of grape pomace and its hydrochars: (a) spectra, (bcores plot for PCA-HC model, and (e) loadings for PC1 and PC2 for PCA-HC model.

an be seen that there is a clear difference between the differentTC temperatures in the first two PCs (92% explained variance). The

oadings for these PCs are presented in Fig. 1e.As it can be seen from the analysis of FTIR bands (summarized

n Table S1), the O-H stretching band between 3600–3000 cm−1,hich can be attributed to hydroxyl and carboxyl groups, is highly

ffected by different HTC temperatures. This is most likely causedy dehydration of raw grape pomace during the HTC process [19].he position of this broad band shifts to lower wavenumbers (lowernergy) as the HTC temperature increases, which is most proba-

es plot for PCA-HC + GP model, (c) loadings for PC1 of the PCA-HC + GP model, (d)

bly caused by disassembly of intermolecular H-bonds establishedby OH groups, which are disrupted when the temperature rises[29]. From the corresponding loadings plot (Fig. 1c), the shift ofthis band is found highly significant. The band of alkenes C-Hstretching vibration at around 3010 cm−1 is also highly affectedby temperature: it is almost absent in GP and HC-220, while hav-

ing a maximum in HC-200, most likely due to formation of aromaticstructure. This band is not found significant in PCA-GP + HC loadingsplot in (Fig. 1c), but is quite significant for distinguishing betweenthe effects of different HTC temperature instead (PCA-HC load-

2 and A

ifaatdba(wwawct(iswgHa

sdabctasaat

HsIftsaWsPnt

d4aitb4agbC(ieidat

72 J. Petrovic et al. / Journal of Analytical

ngs plot for PC2 in Fig. 1e), particularly in distinguishing HC-200rom the rest (Fig. 1d). Furthermore the −C-H stretching bandst 2921 cm−1 and 2850 cm−1 of the methylene groups, which aren indication of increasing unsaturation, are more intense in HCshan in GP. These bands are also weaker in HC-220 than in HC-200ue to following aromatization, similar to 3010 cm−1 band. Theseands are found highly significant for both PCA models (Fig. 1cnd e). The peak of C O from carboxylic groups of hemicellulose1730 cm−1) almost completely disappeared after carbonization,hile C O band at 1541 cm−1 and C H band at 1383 cm−1 wereeaker and shifted to lower wavenumbers. The band at 1711 cm−1

ttributed to C O bonds from cellulose and lignin [3] also increasedith temperature, reaching a maximum at 200 ◦C (Table S1). This

onfirms that decomposition of hemicellulose in HC starts at loweremperatures than the decay of cellulose and lignin, whose bands1515 cm−1, 1455 cm−1) become sharper. Increased lignin contentn HCs obtained at higher temperatures is also confirmed by thehift of alkyl aril ethers (-OCH3) band at 1247 cm−1 towards higheravenumbers 1280 cm−1. A decreased intensity of cellulose C O

roups at 1711 cm−1 and C-O-C band at 1161 cm−1 [3,29–31] inC-220 indicates that the degradation of cellulose began alreadyt 220 ◦C, which is in line with DTA results.

The C C vibration band (1650 cm−1) exhibited a higher inten-ity in GP and HC-180, while in HC-200 and HC-220 its intensityecreased. The shift of this band towards 1613 cm−1 (C C fromromatic bond) as well as increase of aromatic C H deformationand (782 cm−1) with temperature, indicate an increase of carbonontent and formation of aromatic structures in tested HCs. All ofhese bands are found highly significant for both PCA models (Fig. 1cnd e). Furthermore, the bands around 1054 cm−1 and 1034 cm−1

harpened with temperature. It is also observed that the band atbout 1054 cm−1 attributed to C-O stretching of primary alcoholsnd/or alkyl substituted ethers in HCs is also very important alonghe PC2 in Fig. 1.

The FT-NIR spectra (second derivative) of the analysed GP andC samples, averaged over spectral replicates per sample are pre-

ented in Fig. 2a, in relevant spectral regions (7500–4000 cm−1).n addition to visual inspection of the spectra, a PCA was per-ormed, in order to investigate the overall effect of HTC. Indeed,he observed patterns in the PCA model performed on GP and HCamples together presented in Fig. 2b and c (98.3% explained vari-tion) are highly similar to those observed in the FTIR PCA models.hen only the HTC samples were analysed by PCA, a very distinct

eparation of the samples is also observed in the combination ofC1 and PC2 (97.4% explained variance). However, this model didot provide any additional explanation of the effect of differentemperatures and is therefore omitted.

As it can be seen from the NIR spectra (Fig. 2a), the mostominant bands with the maximum at about 4340 cm−1 and265 cm−1, which are assigned to combinations C-H stretchingnd/or deformation group frequencies of polysaccharides [32]. Its also observed that the intensity of these bands increases withemperature (Table S2). Similarly, an increase in intensity of severalands was observed with temperature increment: bands between625 cm−1 and 4028 cm−1, which are associated with C-H stretchnd C O of carbohydrates (4545 cm−1), OH stretch and C O stretchroup of cellulose (4407 cm−1), C H stretch and C C stretch com-ination from C-C-H or CH2 (4134 cm−1, 4028 cm−1), as well asH (1st overtone) from lignin (5645), C O stretch (2nd overtone)5264) [32–35]. Increase in intensity of these bands indicates anncrease in carbon and lignin content which is in accordance withlemental and FTIR analysis. All of these bands are found significant

n the loadings plot (Fig. 2c). On the other hand, as a result of dehy-ration, the intensity of the absorption bands with the maximumt about 5214 cm−1 (OH stretching/HOH deformation combina-ion of cellulose), 5192 cm−1 (OH stretching/OH deformation of

pplied Pyrolysis 118 (2016) 267–277

water) and 5025 cm−1 (OH 2nd overtone), were reduced duringtemperature increase [34–36]. These bands are indeed found highlysignificant for explaining the effect of HTC temperature in theloadings plot, whereas the shift of the 5214 cm−1 band, whichdominates the loadings plot, was not as visible in the NIR spectra(Fig. 2c). Band at around 5264 cm−1 originating from C O in COOHgroups [32] is in like manner found highly significant. Furthermore,C-H vibration from hemicellulose (4880 cm−1) and C-O and O-Hbands from lignin (4415 cm−1) are also highly affected by temper-ature increment. Differences in C-H bands (4265 cm−1, 4134 cm−1,5820 cm−1, 5940 cm−1), C O from carbohydrates (4545 cm−1) andOH vibrations (around 7000 cm−1) are systematically affected bytemperature (Table S2). Furthermore, the water band occurring ataround 6868 cm−1 is found stronger in GP samples compared toHC samples, which confirms the dehydration of GP during the HTCprocess. As it can be seen in the loading for the PC1, both the shiftand the changes in intensity of this band were found significant(Fig. 2c).

3.1.4. Overall effect of temperature on HTC of GP: PLSR analysis ofFTIR and

In order to obtain better understanding of NIR spectra of GPs,a PLSR of NIR and FTIR spectra was performed. Based on that thecorrelations between the fundamental vibrations occurring in mid-IR region and overtones and combination bands occurring in theNIR region are determined and visualized in a correlation load-ings plot (presented in Fig. 3). In correlation loadings plot, twomain features are determined: (i) significance of a variable for theobserved variation, which is visualised by the proximity of a vari-able to the outer circle of the plot, because the ring between theinner circle (dashed line) and the outer circle designates 50–100% ofexplained variance, and (ii) interrelations between variables, whichare visualised by position of a variable relative to other variables(i.e. when two variables are plotted close to each other, they arestrongly positively correlated, and vice versa). By analysing thisplot, the relations between vibrational bands and different HTCtemperatures are determined, thereby explaining the effect of thesetemperatures on changes in molecular structure of GP. The PLSRmodel was calculated for all samples together, only this time usingFTIR and NIR spectra together, where FTIR data is used as predictor(X) and NIR as response (Y). First two principal components explain88% of variance in FTIR and 94% in NIR data. Variables that arenot found significant (plotted very close to the centre) are plottedwithout labels for better readability.

As it can be seen, the effect of the temperature is quite distinctand highly similar to patterns observed in FTIR and NIR indi-vidually: the PC1 explains the overall effect of temperature onthe structure of GP (difference between the GP on the left andHC samples on the right side), while the PC2 explains the dif-ference between the HC temperatures (180 ◦C top, 200 ◦C middleand 220 ◦C bottom). Raw GP is associated with high water con-tent (positively correlated to FTIR 3300–3600 bands and NIR bandsaround 7000 cm−1), native C O groups in cellulose, hemicelluloseand lignin (1735 cm−1, 1541 cm−1), alkens (3010 cm−1). NIR bandsat 4545, 4553, 5059, 4031 and 5218 cm−1 are also strongly posi-tively correlated, implying that these bands are probably associatedwith native structures in raw GP. Temperature of 180 ◦C mostlyaffects aromatic rings (strong positive correlation to the shift ofbands around 1520 cm−1 from 1525 to 1516 cm−1 and C C bondsaround 1650 cm−1 toward 1615 cm−1 band). NIR bands 5679 and4406 cm−1 are also strongly positively correlated, indicating that

these bands can as well be associated with aromatic rings. Tem-perature of 200 ◦C was not found as significant since it is plottedclose to the centre, but is associated with changes in CO groups inalcohol and carboxylic groups (positively correlated to 740, 1035,

J. Petrovic et al. / Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277 273

F , (b) scores plot for the PCA-HC + GP model, and (c) loadings for PC1 of the PCA-HC + GPm

1ia1aw4tc

3

3

aat(pwatw

Table 2Inorganic elements in tested PWs.

Parameter PW-180 PW-200 PW-220

Inorganics(mg L−1)

K 1930 ± 3 2140 ± 2 2525 ± 5Mg 164 ± 2 163 ± 1 135 ± 1Ca 160 ± 2 110 ± 2 55 ± 1Na 11.0 ± 0.5 12.0 ± 1.5 15 ± 1Fe 9.0 ± 0.6 17.0 ± 0.3 18.0 ± 0.7

ig. 2. Results of FT-NIR analysis of grape pomace and its hydrochars: (a) spectraodel.

054 and 1701 cm−1 bands). Temperature of 220 ◦C has a strongnfluence on methylene groups, changes in cellulose, hemicellulosend lignin structure (strongly positively correlated to FTIR bands at108, 1161,1246, 1281, 1383, 1561, 2852, 2922 cm−1, which arelso found highly significant on right hand side of the plot) andater (NIR bands at around 7000 cm−1). NIR bands at 4186, 5010,

341 and 4251 cm−1 are also positively correlated, indicating thathese bands can also be associated with changes in cellulose, hemi-ellulose and lignin structure.

.2. Characterization of process waters (PWs)

.2.1. Inorganic analysisConcentrations of detectable inorganics in the GP-PWs obtained

t different HTC temperatures 180 ◦C (PW-180), 200 ◦C (PW-200)nd 220 ◦C (PW-220) confirmed that the HTC led to the par-ial translocation of the tested elements from HCs into the PWsTable 2). Enhanced leaching of Na, K, Fe and Si from HCs with tem-erature resulted in their highest concentrations in the PW-220,

hile K was almost completely transferred. On the other hand, re-

dsorption of Ca, Mg, and P by HC-220 was once again confirmedhrough the decreasing trend of their concentrations in PW-220ith the temperature increase.

Si 105 ± 1 127 ± 2 131 ± 1P 356 ± 2 329 ± 2 182 ± 2

3.2.2. Spectroscopic analysisFTIR spectra of PWs (averaged over spectral replicates per sam-

ple) are presented in Fig. 4a. In addition to visual inspection of thespectra, PCA was performed for better insight into the effects of HTCtemperatures. As it can be seen in scores plot (Fig. 4b) different PWsamples are well separated along PC1 (72% explained variance),which also delineates the temperature increase (marked with greyarrow). This is why the loadings for only this component are rel-evant and therefore plotted on top of the FTIR spectra in order tobetter emphasize the significance of each of the FTIR band to thePCA model (Fig. 4a). PC2 (17% explained variance) is also analysed

and the variation that this component explains is mostly irrelevantphysical differences between samples (not shown).

From the analysis of FTIR spectra, summarized in Table S3, it isapparent that the C O vibration, most likely from −COOH group

274 J. Petrovic et al. / Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277

Fig. 3. PLSR results of FT-NIR and FT-IR spectra presented as correlation loadings plot for GP and HTC samples together.

ctra an

iiPpdafdtapett1d

Fig. 4. Results of process water characterization: (a) FTIR spe

n organic acids (1707 cm−1) is heavily affected by temperature: itsntensity decreases with temperature, and is almost absent in theW-220. As expected, this band is highly significant in the loadingslot. Aromatic ring stretching of the C C C bonds at 1580 cm−1,etected in PW-180 and PW-200 samples, decreases in intensitynd shifts towards lower energies in the PW-220. Most likely reasonor this is reduction of the unbound aromatic clusters/fragmentsue to enhanced aromatization of HCs at higher temperatures. Fur-hermore, dehydration of the hydroxyl groups of the monomerst 220 ◦C leads to the formation of C O band at 1550 cm−1 that isresent as a shoulder to the 1574 band. This band is not as appar-nt in the FTIR spectra, but is highly significant for the explaining

he effect of temperature on composition of PWs, as observed inhe loadings plot. The intensity of methyl C H bands at 1456 and420 cm−1 increases with temperature, which implies the abun-ance of methyl groups in PW-200 and PW-220. An opposite trend

d loadings plot, (b) scores plot of PCA model on FTIR spectra.

is observed for the band at 1354 cm−1, while the dimethyl C H bandat 1389 cm−1, due to disruption of the dimethyl linkage, completelydisappeared in PW-220. All of these bands are found highly signifi-cant for the PCA model (Fig. 4a). In addition to that, methoxy groups’band at 1285 cm−1 in PW-180 samples is present as a broad feature,with not very high intensity. In PW-200 and PW-220 samples thisband becomes sharper, slightly more intensive and shifted towards1300 cm−1, as a result of hydrolyzed lignin products formation. Theshift of this band is highly significant in the loadings plot. A promi-nent difference between the different PW samples is observed inthe bands at 1018, 1050, 1078 and 1114 cm−1. The bands at 1018and 1050 cm−1 are pronounced in intensity in PW-180, while their

intensity decreases with the temperature, probably due to cleavageof ether or alcoholic (C O) bonds. Conversely, the 1114 cm−1 bandtentatively assigned to secondary alcohol content, increases in PW-220. All of these bands are found highly significant in the loadings

J. Petrovic et al. / Journal of Analytical and Applied Pyrolysis 118 (2016) 267–277 275

orresp

pt

Fig. 5. HPLC-MS chromatograms of PWs (a) and c

lot (Fig. 4a). The C H aromatic band at 939 cm−1 decreases withemperature, which is a consequence of fiber components prod-

onding mass spectra of detected compounds (b).

ucts decomposition. However, this band is moderately significantfor the PCA model (Fig. 4a).

2 and A

3

pcadcbATTldtTpcado

3

qa

o(in(dPba(tMdps(timld

pFoAausfpctb(

4

p

[

[

[

[

[

76 J. Petrovic et al. / Journal of Analytical

.2.3. Antioxidant assaysThe antioxidant analysis of PWs has been performed, where

henolic compounds are the target analytes in TAC assay since theirontent contributes to overall antioxidant activity. The results ofntioxidant assays are presented in Fig. S4, where for the sake ofirect comparison the values are scaled and expressed as mg oforresponded equivalents per mL of gallic acid (GA) for TPC, ascor-ic acid (AsA) for TAC and cyanidin-3-glucoside (C-3-G) for TACY.s it can be seen, PW-180 and PW-200 showed higher TAC andACY than the PW-220. Temperature enhances the extraction ofPC from the GP, but also degrades the anthocyanins, which theneads to reduction of TAC. Previous studies have reported that theegradation of anthocyanins during extraction in water occurs atemperatures above 100 ◦C [13]. In this study, TACY was related toAC, implying that the antioxidant activity of GP is largely due to theresence of anthocyanins. Under certain conditions, anthocyaninsan degrade significantly to their constituent phenolic acids, which,mong other things, can affect the TPC [37]. This can explain theecrease in TACY with a simultaneous increase in TPC observed inur results.

.2.4. Individual organic components and organic acid contentHPLC-MS chromatograms (Fig. 5a) clearly display an increase in

uantity of organic compounds with the HTC temperature increase,lthough without significant qualitative differences between PWs.

From MS spectra given as inserts in Fig. 5b, tentative presencef guaiacol (A), catechol (B), vanillin (C), 2-furoic acid (E), 2,5-furandi) carboxylic acid (F), 5-formyl furoic acid (I) was inferredn accordance with literature [7]. Furthermore, the most domi-ant peak (G) possibly originates from 5-(hydroxymethyl) furfural5-HMF), while less pronounced peak (D) probably represents itserivate. As these peaks indicate the existence of furfurals in theWs, it can be concluded that 5-HMF and its derivatives maye obtained through degradation of various precursors, such aszelaic (187 mass-to-charge ratio (m/z)) and pentadecanoic acids241 m/z) (G and D, respectively). Furthermore, repetitive fragmen-ation pattern (–62 m/z, possibly loss of CO2 + H2O) is observed in

S spectra of various peaks indicating different levels of dehy-ration and decarboxylation of parent compounds. Presence of-coumaric acid (163 m/z) derivatives is also observed, since MSpectra of compound labelled as H displayed fragmentation pattern225–163 m/z). Compound H has almost two times longer retentionime than p-coumaric acid standard (data not shown), suggest-ng that it may represent p-coumaric acid dimer derivatives. The

ass spectrums of hydrophobic compounds (331 m/z and 293 m/z,abelled as (J) and (K) respectively), presumably arises as a result ofifferent degradation stage of lignin structures during HTC.

In addition to better extraction, a formation of various furanroducts is stimulated in PWs collected at higher temperatures.uroic acid can be produced by furfural oxidation [7], while thexidation of 5-HMF may produce 2,5-(furandi) carboxylic acid.dditionally, Tsubaki et al. [38] reported that vanillin and guaiacolre products from hydrothermally induced hydrolysis of guaiacylnits in lignin. These results are consistent with the FTIR analy-is of PWs. Besides, PWs are rich in organic acids (malic, succinic,umaric and acetic acid), which may originate from the conversionroducts of sugars and 5-HMF [7]. It is notable that the organic acidoncentration in tested PWs decreases with increasing tempera-ure (Table S4). As aforementioned, FTIR analysis showed that C Oonds in ketone and in carboxylic acid decreased with temperatureFig. 4a), which is thus confirmed with chromatographic analysis.

. Conclusion

A thorough physico-chemical characterization of raw grapeomace and the products obtained by hydrothermal carboniza-

[

pplied Pyrolysis 118 (2016) 267–277

tion has been performed. Hereby obtained results outlined highlyvaluable fuel properties of grape pomace hydrochars. Multivari-ate analysis of the results revealed that reactivity of hemicelluloseand cellulose caused by the temperature increase governs the mostsignificant structural changes of grape pomace during HTC process.The formation of energy-dense, coal-like hydrochars with prevail-ing aromaticity and with heightened ability for re-adsorption ofinorganics and phosphorus was achieved at 220 ◦C. These findingssupport utilization of GP-HC as solid biofuel, absorbent and poten-tial soil amendment. Although the temperature increase improvedtotal phenolic content in process waters, it was shown that theirtotal antioxidant capacity mostly arises from presence of antho-cyanins. Concurrently, several potentially high-valuable chemicalswere found in process waters, as various individual organic compo-nents. These results can be further used for determining the mostadequate reuse of the GP biomass, as well as for optimisation of theHTC process parameters in larger scale applications.

Acknowledgments

The authors are grateful to the Serbian Ministry of Education,Science and Technological Development for the financial support ofthis investigation included in the projects TR 31003 and OI 173040.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.jaap.2016.02.010.

References

[1] Z. Liu, A. Quek, S.K. Hoekman, R. Balasubramanian, Production of solid biocharfuel from waste biomass by hydrothermal carbonization, Fuel 103 (2013)943–949.

[2] M.T. Reza, J.G. Lynam, M.H. Uddin, C.J. Coronella, Hydrothermalcarbonization: fate of inorganics, Biomass Bioenergy 49 (2013) 86–94.

[3] G.K. Parshetti, S.K. Hoekman, R. Balasubramanian, Chemical, structural andcombustion characteristics of carbonaceous products obtained byhydrothermal carbonization of palm empty fruit bunches, Bioresour. Technol.135 (2013) 683–689.

[4] A. Funke, F. Ziegler, Hydrothermal carbonization of biomass: a summary anddiscussion of chemical mechanisms for process engineering, Biofuels Bioprod.Bioref. 4 (2010) 160–177.

[5] D. Busch, A. Stark, C.I. Kammann, B. Glaser, Genotoxic and phytotoxic riskassessment of fresh and treated hydrochar from hydrothermal carbonizationcompared to biochar from pyrolysis, Ecotox. Environ. Saf. 97 (2013) 59–66.

[6] H.S. Kambo, A. Dutta, Strength storage, and combustion characteristics ofdensified lignocellulosic biomass produced via torrefaction and hydrothermalcarbonization, Appl. Energy 135 (2014) 182–191.

[7] J. Poerschmann, B. Weiner, H. Wedwitschka, A. Zehnsdorf, R. Koehler, F.D.Kopinke, Characterization of biochars and dissolved organic matter phasesobtained upon hydrothermal carbonization of Elodea nuttallii, Bioresour.Technol. 189 (2015) 145–153.

[8] V. Benavente, E. Calabuig, A. Fullana, Upgrading of moist agro-industrialwastes by hydrothermal carbonization, J. Anal. Appl. Pyrol. 113 (2015) 89–98.

[9] M. Pala, I.C. Kantarli, H.B. Buyukisik, J. Yanik, Hydrothermal carbonization andtorrefaction of grape pomace: a comparative evaluation, Bioresour. Technol.161 (2014) 255–262.

10] K. Christ, R. Burritt, Critical environmental concerns in wine production: anintegrative review, J. Clean Prod. 53 (2013) 232–242.

11] T. Baumgartel, H. Kluth, K. Epperlein, M. Rodehutscord, A note on digestibilityand energy value for sheep of different grape pomace, Small Ruminant Res. 67(2007) 302–306.

12] J. Ferrer, G. Paez, Z. Marmol, E. Ramones, C. Chandler, M. Marin, A. Ferrer,Agronomic use of biotechnologically processed grape wastes, Bioresour.Technol. 76 (2001) 39–44.

13] J. Yu, M. Ahmedna, Functional components of grape pomace: theircomposition, biological properties and potential applications, Int. J. Food Sci.Tech. 48 (2013) 221–237.

14] D.J.N. Subagyono, M. Marshall, W.R. Jackson, A.L. Chaffee, Pressurized thermal

and hydrothermal decomposition of algae, wood chip residue, and grapemarc: a comparative study, Biomass Bioenergy 76 (2015) 141–157.

15] D. Basso, F. Patuzzi, D. Castello, M. Baratieri, E. Rada, E. Weiss-Hortala, L. Fiori,Agro-industrial waste to solid biofuel through hydrothermal carbonization,Waste Manage. 47 (2016) 114–121.

and A

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[recovery: implications for the analysis of clinical and experimental samples, J.

J. Petrovic et al. / Journal of Analytical

16] Z.Y. Hseu, Evaluating heavy metal contents in nine composts using fourdigestion methods, Bioresour. Technol. 95 (2004) 53–59.

17] J.I. Shultz, R.K. Bell, T.C. Rains, O. Menis, Methods of Analysis of NBS ClayStandards, Nat. Bur. Stand., Spec. Publ. 260–37, Washington, D.C.,US, 1972, pp.3–4.

18] N.S. Awwad, Y.A. El-Nadi, M.M. Hamed, Successive processes for purificationand extraction of phosphoric acid produced by wet process, Chem. Eng.Process. 74 (2013) 69–74.

19] D. Kim, K. Lee, K.Y. Park, Hydrothermal carbonization of anaerobicallydigested sludge for solid fuel production and energy recovery, Fuel 130(2014) 120–125.

20] M.B. Arnao, A. Cano, M. Acosta, Methods to measure the antioxidant activityin plant material. A comparative discussion, Free Radic. Res. 31 (1999) 89–96.

21] G.W. Cheng, P.J. Breen, Activity of phenylalanine ammonia-lyase (PAL) andconcentrations of anthocyanins and phenolics in developing strawberry fruit,J. Am. Soc. Hortic. Sci. 116 (1991) 865–869.

22] V.L. Singleton, J.A. Rossi Jr., Colorimetry of total phenolics withphosphomolybdic-phosphotungstic acid reagents, Am. J. Enol. Viticult. 16(1965) 144–158.

23] A. Kohler, C. Kirschner, A. Oust, H. Martens, Extended multiplicative signalcorrection as a tool for separation and characterization of physical andchemical information in Fourier transform infrared microscopy images ofcryo-sections of beef loin, Appl. Spectrosc. 59 (2005) 707–716.

24] H. Martens, J.P. Nielsen, S.B. Engelsen, Light scattering and light absorbanceseparated by extended multiplicative signal correction. Application tonear-infrared transmission analysis of powder mixtures, Anal. Chem. 75(2003) 394–404.

25] K. Pearson, On lines and planes of closest fit to systems of points in space,Philos. Mag. 2 (1901) 559–572.

26] A. Mukherjee, A.R. Zimmerman, W. Harris, Surface chemistry variations

among a series of laboratory-produced biochars, Geoderma 163 (2011)247–255.

27] S.E. Hale, V. Alling, V. Martinsen, J. Mulder, G.D. Breedveld, G. Cornelissen, Thesorption and desorption of phosphate-P, ammonium-N and nitrate-N in cacaoshell and corn cob biochars, Chemosphere 91 (2013) 1612–1619.

[

pplied Pyrolysis 118 (2016) 267–277 277

28] A. Garcia-Maraver, D. Salvachua, M.J. Martinez, L.F. Diaz, M. Zamorano,Analysis of the relation between the cellulose, hemicellulose and lignincontent and the thermal behavior of residual biomass from olive trees, WasteManage. 33 (2013) 2245–2249.

29] J. Coates, Interpretation of infrared spectra, a practical approach, in: R.A.Meyers (Ed.), Encyclopedia of Analytical Chemistry, John Wiley & Sons Ltd.,Chichester, UK, 2000, pp. 10815–10837.

30] M.T. Reza, B. Wirth, U. Luder, M. Werner, Behavior of selected hydrolyzed anddehydrated products during hydrothermal carbonization of biomass,Bioresour. Technol. 169 (2014) 352–361.

31] L. Kong, P. Miao, J. Qin, Characteristics and pyrolysis dynamic behaviors ofhydrothermally treated microcrystalline cellulose, J. Anal. Appl. Pyrol. 100(2013) 67–74.

32] W. He, H. Hu, Prediction of hot-water-soluble extractive, pentosan andcellulose content of various wood species using FT-NIR spectroscopy,Bioresour. Technol. 140 (2013) 299–305.

33] H. Liu, B. Xiang, L. Qu, Structure analysis of ascorbic acid using near-infraredspectroscopy and generalized two-dimensional correlation spectroscopy, J.Mol. Struct. 794 (2006) 12–17.

34] J. Xue, Z. Yang, L. Han, Y. Liu, Y. Liu, C. Zhou, On-line measurement ofproximates and lignocellulose components of corn stover using NIRS, Appl.Energy 137 (2015) 18–25.

35] C.C. Guimaraes, M.L.F. Simeone, R.A.C. Parrella, M.M. Sena, Use of NIRS topredict composition and bioethanol yield from cell wall structuralcomponents of sweet sorghum biomass, Microchem. J. 117 (2014) 194–201.

36] N. Morsy, D.W. Sun, Robust linear and non-linear models of NIR spectroscopyfor detection and quantification of adulterants in fresh and frozen-thawedminced beef, Meat Sci. 93 (2013) 292–302.

37] G. Woodward, P. Kroon, A. Cassidy, C. Kay, Anthocyanin stability and

Agr. Food Chem. 57 (2009) 5271–5278.38] S. Tsubaki, M. Sakamoto, J. Azuma, Microwave-assisted extraction of phenolic

compounds from tea residues under autohydrolytic conditions, Food Chem.123 (2010) 1255–1258.