Production of biogas (methane and hydrogen) from

0 downloads 0 Views 2MB Size Report
hemicellulosic hydrolysate, rich in oligomers and C5 sugars, such as xylose and .... supplied and calibrated by the company Ozone & Life (Brazil), with a.
Bioresource Technology 263 (2018) 601–612

Contents lists available at ScienceDirect

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

Production of biogas (methane and hydrogen) from anaerobic digestion of hemicellulosic hydrolysate generated in the oxidative pretreatment of coffee husks

T



Lívia Caroline dos Santos, Oscar Fernando Herrera Adarme, Bruno Eduardo Lobo Baêta , Leandro Vinícius Alves Gurgel, Sérgio Francisco de Aquino Environmental and Chemical Technology Group, Department of Chemistry, Federal University of Ouro Preto, Campus Universitário Morro do Cruzeiro, Bauxita, s/n, 35400-000 Ouro Preto, Brazil

G R A P H I C A L A B S T R A C T

A R T I C LE I N FO

A B S T R A C T

Keywords: Coffee husks Pretreatment Ozonolysis Biogas Anaerobic digestion

Ozone pretreatment of coffee husks (CH) was evaluated to generate hydrolysates for biogas production and to preserve cellulose of the solid phase for 2G ethanol production. Pretreatment variables included liquid-to-solid ratio (LSR), pH and specific applied ozone load (SAOL). Considering single-stage anaerobic digestion (AD), the highest methane production (36 NmL CH4/g CH) was achieved with the hydrolysate generated in the experiment using LSR 10 mL/g, pH 11 and SAOL 18.5 mg O3/g CH, leading to 0.064 kJ/g CH energy recovery. Due to the presence of toxic compounds in the hydrolysate, the addition of powdered activated carbon (4 g/L) to the reactor enhanced biogas production, leading to 86 NmL CH4/g CH yield and 0.58 kJ/g CH energy recovery. When twostage AD was applied, methane production resulted in 49 NmL CH4/g CH, with additional 19 NmL H2/g CH production, resulting in a net 0.26 kJ/g CH energy recovery.

1. Introduction For developing countries such as Brazil, lignocellulosic biomass is a promising alternative source because it is renewable and abundant (Couto et al., 2004). According to the Brazilian National Supply Company (CONAB), Brazil is the world’s largest exporter of coffee and, consequently, generates a considerable amount of waste (coffee husks).



Corresponding author. E-mail address: [email protected] (B.E.L. Baêta).

https://doi.org/10.1016/j.biortech.2018.05.037 Received 6 March 2018; Received in revised form 5 May 2018; Accepted 9 May 2018 0960-8524/ © 2018 Elsevier Ltd. All rights reserved.

It is estimated that Brazilian coffee production in 2017 generated about 2.7 million tons of coffee husks as waste (CONAB, 2017) that represent a natural, cheap and abundant source of lignocellulosic biomass that can be used for sustainable production of bioenergy, biofuels and bioproducts of high added values. Therefore, from an environmental point of view, the production of second-generation (2G) bioethanol and biogas from coffee husks would be an interesting and strategic

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

2. Material and methods

alternative. Some studies for the use of coffee husks have been made, such as direct use as a fuel in farms, animal feed, solid state fermentation and biodiesel production; however, considering the high amount of waste generated, there is still a need to seek better alternatives and more profitable uses. Lignocellulosic biomass presents a high resistance to chemical and biological degradation, which is mainly explained by the presence of a complex interaction between the major compounds such as lignin, hemicelluloses and cellulose in its structure. As a result of the structural complexity presented by this type of biomass, its use as a raw material for the production of bioenergy such as biogas and bioethanol requires a pretreatment step. The main objective of this step is to disaggregate or disrupt interactions within the lignocellulosic complex (reduce particle size, increase surface area and pore volume, reduce lignin content and decrease cellulose crystallinity), facilitating access and enzymatic and microbiological digestibility (Mussatto & Dragone, 2016). Ozonation has been used as a pretreatment technique for different types of biomass to improve their biodegradability, for example, wheat straw, oats, barley, rice, sugarcane bagasse, grasses and sawdust from different tree species. Moreover, ozonation has been used as a treatment technique in cellulose and paper industry in the bleaching stage aiming at removing lignin residues and fragments from the fiber surface (Hung & Sumathi, 2004; Travaini et al., 2014). However, no studies were found that evaluated the use of oxidative pretreatment for adding value to coffee husks through the production of renewable fuels, which shows the relevance of the present study. In the oxidation of lignocellulosic biomass major compounds with ozone, the most susceptible material to degradation is lignin, followed by hemicelluloses, with little preference for cellulose (Taherzadeh & Karimi, 2008). Thus, ozone becomes effective, breaking the association between several components of lignocellulosic biomass and producing a substrate with better reactivity for enzymatic hydrolysis (Kumar et al., 2009; Sun & Cheng, 2002; Taherzadeh & Karimi, 2008). The mechanisms of organic matter oxidation by the application of ozone can be classified in direct (molecular reaction) or indirect (radical reaction). The direct mechanism called ozonolysis (Criegee mechanism) occurs when ozone molecule (O3) promotes an electrophilic addition to a double bond (π) between carbons, forming ozonides that decompose into carbonyl compounds and hydrogen peroxide in the presence of water. Due to its electrophilic nature, ozone also reacts with structures with high electron density such as aromatic compounds, promoting addition to the double carbon–carbon bonds and causing ring opening and formation of unsaturated byproducts with carbonyl and/or carboxylic functional groups (Mussatto & Dragone, 2016). The indirect mechanism prevails in alkaline media or in the presence of some agents (hydrogen peroxide, UV radiation) that lead to the formation of radicals such as hydroxyl (%OH), superoxide (O2%−) and hydroperoxide (HO2%) (Gottschalk et al., 2010). These radicals are very reactive and not selective, and oxidation is considered an advanced oxidative process (AOP) that has aroused considerable interest in several applications (Mussatto & Dragone, 2016; Nascimento et al., 1998). The pretreatment processes of lignocellulosic biomass can generate a solid fraction, rich in cellulose and a liquid fraction, which consists of hemicellulosic hydrolysate, rich in oligomers and C5 sugars, such as xylose and arabinose. Biogas production by anaerobic digestion of liquid fraction is a technological option to be considered in order to maximize energy recovery from lignocellulosic biomass (Baêta et al., 2016a,b; Barakat et al., 2012; Travaini et al., 2016) and the generated solid fraction can be used for the production of 2G ethanol after passing through a stage of enzymatic hydrolysis followed by alcoholic fermentation. Thus, the aim of this study was to investigate and optimize the ozonation process of coffee husks as a pretreatment technique for the production of hydrolysates (liquid fraction) that would be used in the production of biogas (CH4 and/or H2) via anaerobic digestion.

2.1. Chemicals Cyclohexane, ethanol (99.5%), hydrochloric acid (36–37 wt% in water), sodium hydroxide and sodium thiosulphate were purchased from Synth (Brazil). Sulfuric acids (95–98% and 99.999%) were purchased from Synth and Sigma-Aldrich (Brazil). Chromatography grade standards cellobiose, D-glucose, D-xylose, L-arabinose, acetic acid, formic acid, propionic acid, isobutyric acid, butyric acid, valeric acid, isovaleric acid, 5-hydroxymethyl-2-furfuraldehyde (HMF) and 2-furfuraldehyde (FF) were purchased from Sigma-Aldrich (Brazil). Oxygen (purity 99.99%) used during ozonolysis was purchased from White Martins/Praxair (Brazil). The powdered activated carbon (PAC) was purchased from Synth (Brazil). 2.2. Coffee husks Coffee husks (CH) were collected at Jangada farm, Criminoso, Lavras, Minas Gerais, Brazil. They were previously dried (under sunlight) to remove moisture until a less than 10% moisture content was achieved. After drying, CH were ground (42–60 mesh, 0.355–0.250 mm) in a knife mill (Marconi, model MA048) and stored at room temperature prior to use. 2.3. Characterization of biomass and fractions obtained in oxidative pretreatment The quantification of the main components of CH and of the solid and liquid fractions generated after pretreatment were performed according to the standard methodologies. Ash content (inorganic) was determined according to the standard methodology “Ash in wood, pulp, paper and paperboard” (TAPPI T211 om-02). In this method, three samples of CH (1.000 g, on dry-weight basis) were weighed into porcelain crucibles, transferred to a muffle furnace, and heated to 25 °C from 525 °C at a heating rate of 1.25 °C/min and kept at 525 °C for 2 h. Quantitative determination of extractives in CH was performed according to TAPPI T204 cm-07. In this method, three samples of CH (10.000 g, on dry-weight basis) were weighed into paper cartridges and continually extracted with a mixture of cyclohexane and ethanol (1:1, v/v) for 24 h using a Soxhlet apparatus. The insoluble lignin content in the solid fractions was determined according to the standard methodology “Determination of acid-insoluble lignin in biomass” (NREL LAP004). In this method, three samples of CH (0.700 g, on dry-weight basis) were weighed into glass cylindrical tubes and 10.7 mL of 72% (w/w) H2SO4 solution was added to the tubes. The mixture was magnetically stirred at 25 °C for 2 h. Then, 400 mL of distilled water is added to each tube. The tubes were placed into an autoclave for 1 h at 121 °C. The liquid and solid fractions obtained after hydrolysis were separated by vacuum filtration on Büchner glass crucibles. The solid fraction (insoluble lignin) was dried in an oven at 80 °C for 4 h while the liquid fraction (soluble carbohydrates and carbohydrate degradation products) were analyzed by high performance liquid chromatography (HPLC). Ash content in insoluble lignin was determined as previously described. Soluble lignin was analyzed on a UV–Vis spectrophotometer (NREL LAP-004). The concentration of sugars (cellobiose, glucose, xylose and arabinose), organic acids (acetic and formic) and sugar degradation products (FF and HMF) were determined by HPLC in a Shimadzu HPLC system equipped with Aminex HPX 87H column (300 × 7.8 mm Bio-Rad) maintained at 55 °C (Shimadzu column oven, model CTO-30A) using refractive index detector (Shimadzu, model RID-6A) for sugars and a UV–Vis detector (Shimadzu, model SPD-10AV) set at wavelengths of 210 nm and 274 nm for organic acids and sugar degradation products (FF and HMF), respectively. The mobile phase was composed of 5 mmol/L sulfuric acid at a flow rate of 0.6 mL/min. 602

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Then, a desirability tool was used to predict values of the independent variables that would result in a liquid fraction with favorable conditions for biogas production after pretreatment, and a suitable solid fraction to obtain higher yields in the enzymatic hydrolysis. To reach the best results in the enzymatic hydrolysis, it was desirable to maintain a high removal of hemicelluloses and a small removal of cellulose in the solid fraction. For the production of biogas, it was desirable to maintain high removal of cellulose and hemicelluloses from the solid fraction, since these compounds when dissolved in the liquid fraction are more easily degradable. In both cases, lignin removal levels (low, medium and high) varied, generating three desirable conditions that could be used for production of 2G ethanol and three desirable conditions used for biogas production from liquid fraction. The conditions generated by the desirability tool are presented in Table 1.

The mass of sugars, organic acids and sugar degradation products were used to calculate cellulose and hemicelluloses content in the solid fraction, whose conversion coefficients were described by Gurgel et al. (2014) and Lima et al. (2018). Chemical oxygen demand (COD) analysis of the liquid fraction was performed according to the closed reflux colorimetric method, as described in the Standard Methods for the Examination of Water and Wastewater (APHA, 2005). 2.4. Experimental design and desirability conditions Oxidative ozone pretreatment (OOP) of CH was performed according to conditions generated by Doehlert experimental design (DED). The selected operational variables (independent variables) were pH, liquid-to-solid ratio (LSR, mL/g) and specific applied ozone load (SAOL, mg O3/g CH). SAOL was a variable that included ozone concentration ([O3], mg/L), gas flow (Q, L/min), reaction time (t, min) and CH weight (wCH, g) for each experiment (Eq. (1)).

mg L 1 ⎞ × t (min) × ⎞ × Q O3 ⎛ SAOL = [O3 ] ⎛ wCH (g ) ⎝ L ⎠ ⎝ min ⎠

2.5. Oxidative ozone pretreatment (OOP) OOP was carried out in a complete mixing reactor, made of glass, with a volume of 500 mL, batch operated and continuously fed with ozone. Ozone entered the reactor through a porous stone for better distribution in the aqueous phase. Ozone leftover was directed to a flask containing an aqueous KI solution (2 g/L) so that it could be converted to O2 before its release to the environment. Ozone gas was obtained through electric discharge at an ozone generator, model O&L3.0RM, supplied and calibrated by the company Ozone & Life (Brazil), with a production capacity of 3 g O3/h, using oxygen with purity of 99.99%. The equipment allows regulating ozone concentration produced with frequency doses and an adjustable oxygen flow regulator in the cylinder (0–2 L/min), generating concentrations from 3 to 54 mg O3/L. For each experiment, a sample of 25.0 g (on dry-weight basis) CH (42–60 mesh) and distilled water in different volumes and pH values were used as described in Table 1. Aqueous NaOH and/or HCl solutions were used to

(1)

These values were chosen based on preliminary tests, taking into account reactor volume, mixing capacity, residual ozone quantification capacity and results obtained in studies that used AOP in lignocellulosic biomass, as in the studied performed by Adarme et al. (2017). The effect of the pretreatment in the biomass was evaluated considering as variable responses (dependent variables) the percentages of lignin, cellulose and hemicelluloses removal of the coffee husks, the chemical oxygen demand (COD, g/L) and C5 and C6 sugars (g) in the liquid fraction. Results were analyzed using the Statistica® software (StatSoft Inc, version 12) at 95% significance level. Matrix of experiments of DED is presented in Table 1.

Table 1 Experimental conditions, composition of the solid and liquid fractions and pretreatment effect on cellulose, hemicelluloses and lignin generated from Doehlert design and desirability conditions for ozone oxidative pretreatment of the coffee husks. Test

Doehlert design conditions Experimental condition

Solid fraction composition

Pretreatment effect

Liquid fraction

LSR (mL/ g)

pH

SAOL (mg O3/ g CH)

Time (min)

Cellulose (%)

Lignin (%)

Hemicelluloses (%)

Mass balance (%)

Cellulose removal (%)

Lignin removal (%)

Hemicelluloses removal (%)

Weight loss (%)

COD hydrolysate (g/L)

Total sugars (g)

20.0 17.5 17.5 10.0 12.5 12.5 17.5 17.5 12.5 15.0 12.5 15.0 15.0 15.0 15.0 15.0 15.0 15.0 15.0

7.0 7.0 11.0 7.0 7.0 3.0 7.0 3.0 7.0 3.0 11.0 11.0 7.0 7.0 7.0 7.0 7.0 7.0 7.0

43.9 81.0 56.3 43.9 6.8 31.5 6.8 31.5 81.0 68.6 56.3 19.1 43.9 43.9 43.9 43.9 43.9 43.9 43.9

32.5 60.0 41.7 32.5 5.0 23.3 5.0 23.3 60.0 50.8 41.7 14.2 32.5 32.5 32.5 32.5 32.5 32.5 32.5

49.1 48.3 46.4 43.0 42.5 47.6 46.5 46.8 48.5 46.3 45.8 46.2 46.5 44.5 46.7 49.2 48.2 46.8 49.5

28.8 28.7 29.4 29.5 30.5 31.3 31.3 32.2 28.8 32.3 28.9 30.3 30.4 29.3 30.9 30.3 30.8 30.5 29.8

22.1 22.7 22.9 22.1 22.3 21.8 22.4 21.6 22.8 21.2 23.2 21.4 20.9 20.8 21.2 21.8 22.0 21.5 21.9

100.0 99.6 98.8 94.6 95.3 100.7 100.1 100.5 100.2 99.8 97.9 97.9 97.8 94.7 98.8 101.3 100.9 98.8 101.1

2.7 14.4 13.2 17.9 18.6 2.5 6.6 1.7 8.4 14.4 20.9 31.5 11.1 9.7 9.9 14.1 9.6 5.2 13.3

31.6 39.0 34.0 32.4 29.8 23.0 24.7 18.9 34.8 28.3 40.1 46.0 30.3 28.8 28.3 36.7 30.7 25.8 37.4

31.6 37.2 33.0 34.1 33.3 30.4 29.9 29.2 32.8 38.8 37.5 50.4 37.8 34.1 36.1 40.7 35.6 31.9 40.2

35.7 42.5 39.3 38.0 37.8 33.5 34.8 31.8 38.8 40.0 44.0 51.9 38.0 34.2 33.4 43.4 39.1 33.7 43.1

13.4 20.6 18.5 29.4 24.4 21.9 15.4 16.4 25.4 20.4 26.5 23.3 18.6 19.9 18.2 20.6 21.4 17.2 18.2

2.7 3.2 3.1 3.2 3.3 3.0 2.8 3.5 3.1 3.5 3.1 2.9 3.2 3.1 3.0 2.9 3.1 3.0 3.3

Desirability conditions D1 15.8 3.0 65.4 D2 14.2 3.0 73.2 D3 10.0 11.0 81.0 D4 20.0 11.0 81.0 D5 10.0 9.0 6.8 D6 10.0 11.0 18.5

48.4 54.2 60.0 60.0 5.0 13.7

48.9 49.6 53.3 50.5 50.4 51.4

32.5 32.1 28.0 29.6 30.5 27.3

21.0 21.9 22.2 22.7 21.7 20.6

102.4 103.7 103.4 102.8 102.5 99.4

2.2 3.7 6.3 3.2 7.7 2.5

17.0 24.8 41.0 32.0 33.0 37.9

34.3 33.7 39.2 32.0 38.0 38.9

35.1 37.0 43.0 37.8 40.5 38.5

23.2 24.5 43.8 22.3 38.4 39.5

3.3 3.6 2.0 3.2 1.5 2.3

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19

603

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Treatment Plant (ETE) Arrudas, Belo Horizonte, Minas Gerais, Brazil). For the BHP experiments, the inoculum was subjected to a heat pretreatment step at 90 °C for 10 min in a thermostated water bath (SOLAB®), which aimed to inactivate hydrogen-consuming microorganisms and to select hydrogen and acid producing microorganisms (Lazaro et al., 2014). After inoculation, the flasks were purged with N2(g) for 3 min and hermetically sealed with a rubber stopper and aluminum seals before being kept in a shaker incubator (Thoth®, model 6440) under constant stirring at 150 rpm and temperature of 35 °C. The biogas volume was evaluated frequently by means of the headspace pressure provided by a differential manometer (pressures up to 1 bar), whereas the biogas composition was measured on a gas chromatography system (Shimadzu GC, model 2014/TCD) equipped with CP-Molsieve 5 Å and CPPoraBOND Q columns, using nitrogen as the carrier gas in the conditions described by Jung and Theato (2013). The results of biogas production were expressed at standard temperature and pressure conditions (STP: 273 K, 1 atm). Initially, samples were taken every 8 h and after the exponential production phase, they were taken once every 24 h until the experiments were completed, when the daily production of CH4 was less than 1% of the accumulated CH4 production. At the end of the experiments, samples of the liquid phase were collected to measure the concentration of volatile fatty acids (VFAs) by HPLC as described in Section 2.3.

adjust the pH of the suspensions and to overcome the buffering effect of CH before ozone addition. The solid and liquid fractions were separated by vacuum filtration and stored at −20 °C for their characterization, as described in Section 2.3. 2.6. Experiments of methane and hydrogen production After its characterization, a part of the liquid fraction, resulting from CH pretreatment in the conditions generated by the statistical tool of desirability (Table 1), was submitted to anaerobic biodegradation experiments to evaluate the biochemical methane potential (BMP) and/or the biochemical hydrogen potential (BHP) of hydrolysates. Biogas production from CH hydrolysates was also evaluated by both single-stage (methanogenic) and two-stage (acidogenic followed by methanogenic) processes, as it was carried out in the study of Baêta et al. (2016b). Additionally, the effect of the addition of PAC was also tested to enhance the methane production in a single-stage AD process. For this, 15 mL of the CH hydrolysate obtained from condition D6 (see desirability conditions, Table 1) was used. The concentration of PAC in the batch reactor during the experiment was fixed at 4 g/L (Baêta et al., 2013). BMP and BHP experiments were carried out in 275 mL amber glass bottles, 150 mL of which corresponded to the net volume and 125 mL to the headspace. The experiments were performed in duplicate plus a blank one. The blank corresponded to the control test (without substrate) to verify the methanogenic activity of the inoculum. No other tests were performed with the inoculum. The flasks were filled with the hydrolysate (15 mL) from CH pretreatment, micro- and macro-nutrients solution prepared in sodium bicarbonate (buffer) according to the methodology described by Angelidaki et al. (2009), and inoculum, whose amounts depended on the COD of the hydrolysate, to maintain a food-to-microorganisms ratio of 0.7 g COD/g SSVinoculum for BMP experiments (Baêta et al., 2017) and 1.8 g COD/g SSVinoculum for BHP experiments (Lamaison et al., 2015). In order to evaluate the effect of the AD process in two-stages, the BMP experiment was carried out using as substrate the remaining liquid fraction of BHP experiment. For BMP experiment with PAC addition, the hydrolysate generated from the pretreatment condition that produced the higher methane amount during the BMP experiments (condition D6, Table 2) in a singlestage was used as substrate. PAC was added to the glass bottle to yield a concentration of 4 g/L. Further procedures were the same described before. For inoculum preparation, a mixture of bovine manure and anaerobic sludge was made, maintaining a ratio of 1:1 (w/w) (Lima et al., 2018). The bovine manure was collected at a livestock farm (Conselheiro Lafaiete, Minas Gerais, Brazil) and the anaerobic sludge came from a UASB reactor fed with sanitary sewage and operated on a demonstration scale in mesophilic conditions (∼25 °C) at the UFMGCOPASA Sanitation Research and Training Center (CePTS) (Sewage

2.7. Kinetic analysis Kinetic parameters of methane and hydrogen production experiments were estimated by fitting the modified Gompertz and exponential two-phase kinetic models, respectively, to the experimental data. The modified Gompertz model (Eq. (2)) is an empirically based model widely used to express biogas production in anaerobic reactors fed with hemicellulose hydrolysate as substrate (Baêta et al., 2016a,b). This model allows the determination of important parameters of the AD process such as the lag phase, maximum specific rate of biogas production (μm) and its maximum accumulated biogas production (P0) (Zwietering et al., 1990).

μ e P = P0 exp ⎧−exp ⎡ m (λ−t ) + 1⎤ ⎫ ⎥ ⎢ ⎨ P0 ⎦⎬ ⎣ ⎭ ⎩

(2)

where P is the accumulated biogas yield (NmL biogas/g CODapplied), P0 is the maximum biogas yield (NmL biogas/g CODapplied), t is the incubation time (h), λ is the lag phase (h), μm is the maximum rate of biogas production (NmL biogas/(h g CODapplied)) and e is the Euler’s constant (2.71828). The exponential two-phase model (Eq. (3)) is also used to monitor anaerobic processes and considers that biogas production can occur in two or more stages. The model describes the conversion of substrate into biogas for each phase, not considering the interaction between them (Lima et al., 2018; Pellera & Gidarakos, 2016).

Table 2 Production of biogas in single-stage anaerobic digestion with and without PAC addition and in two-stage anaerobic digestion from hydrolysate generated by desirability conditions. Hydrolysate

D1 D2 D3 D4 D5 D6

Single-stage anaerobic digestion

Two-stage anaerobic digestion

Single-stage anaerobic digestion with PAC (hydrolysate D6)

CH4 production (NmL CH4/ g CODapplied)

CH4 production (NmL CH4/g CH)

H2 production (NmL H2/ g CODapplied)

H2 production (NmL H2/g CH)

CH4 production (NmL CH4/ g CODacidogenic)

CH4 production (NmL CH4/g CH)

CH4 production (NmL CH4/ g CODapplied)

CH4 production (NmL CH4/g CH)

72.6 88.5 67.0 52.8 81.7 91.1

26.7 30.7 29.3 23.6 31.3 36.0

43.3 53.3 54.8 40.2 55.8 48.5

15.9 18.5 24.0 18.0 21.4 19.2

185.9 163.0 277.2 176.1 363.8 284.6

34.2 26.1 52.4 41.5 42.2 49.0

218.2

86.2

604

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Fig. 1. Response surfaces for (a) lignin removal, (b) hemicelluloses removal and (c) cellulose removal (c) as function of specific applied ozone load (SAOL, mg O3/g CH) and pH, maintaining LSR fixed at center point (15 mL/g).

P = P1 [1−exp(−k1 t )] + P2 [1−exp (−k2 t )]

(3)

generated by the combustion of CH4 and H2 produced in the AD, considering lower calorific values (LCV) of 34.5 MJ/Nm3 and 10.8 MJ/ Nm3, respectively. Considering a typical value for a commercial combined heat and power (CHP) system, 15% of the biogas energy is lost and, from the rest, 35% is converted into electric energy (EE), 30% in an exhaust gaseous stream at 400 °C (EG), which can be reused in the pretreatment, and 35% in a liquid stream at low temperature. The CHP system considered for this study consists of N-CHP units each of which includes a prime mover-generator set and heat recovery system, Lauxiliary boilers, a heat-storage tank, and the local grid connection (Pérez-Elvira & Fdz-Polanco, 2012). The electric energy consumed during the generation of ozone for each pretreatment condition was estimated using the SAOL values used in the experiments, taking into account that the production of 100 g of ozone requires 1.65 MJ energy (Travaini et al., 2016). The energy viability of the process was calculated by subtracting the electric energy generated by the biogas burning in the CHP system of the spent energy during the ozone pretreatment.

where Pn is the accumulated biogas production of phase n (NmL biogas/ g CODapplied) and kn is the kinetic constant for phase n (h−1). The nonlinear regression of the experimental data with the modified Gompertz and exponential two-phase models was performed using MATLAB® 2010a (Mathworks, Inc) software. The choice of the model that best described biogas production by single- and two-stage AD processes was made using the coefficient of determination (R2), root mean square error (RMSE), normalized root mean square error (NRMSE) and Akaike Information Criterion (AIC) as reported by Lima et al. (2018). 2.8. Energy balance An energy balance was applied for all AD experiments using the hydrolysates generated in the OOP of CH under conditions of desirability (Tables 1 and 2). It was considered that the output energy was 605

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

of hydroxyl radicals. In this case, oxidation reactions occur via radicals (indirect reaction), and they are not selective. The reaction of ozone with aromatic compounds in alkaline media can also result in the formation of radicals such as superoxide (O2%−) and hydroperoxide (HO2%), causing chain reactions to start generating more radicals that can attack the organic compounds (Assalin & Durán, 2007; Gottschalk et al., 2010; Ragnar et al., 1999). Thus, the radicals are powerful oxidants and can be effective in the degradation of resistant compounds such as lignin. According to Table 1, the largest removal of total lignin (46.03%) occurred in experiment 12 which applied alkaline conditions. In the study performed by Adarme et al. (2017), who pretreated sugarcane bagasse with ozone, the experimental condition that most removed total lignin (45.2%) had pH 11, LSR 13.33 mL/g and SAOL 12 mg O3/ g bagasse, conditions close to those observed in the present study. No study was found with oxidative pretreatment of CH, but when comparing to OOP of other biomasses, it is possible to notice that the time of 15 min was significantly lower. García-Cubero et al. (2009) pretreated wheat and rye straw with ozone for ethanol production, and the optimal ozonation time was 150 min for the removal of 45% insoluble lignin. Souza-Corrêa et al. (2013) and Travaini et al. (2013) used 60–240 min for ozonation of sugarcane bagasse and obtained delignification efficiencies between 50 and 67%. Kaur et al. (2012) used oxidation on cotton stems and obtained delignification of 42% with ozonation time of 150 min. In this aspect, it is possible to notice that molecular oxidation processes, which use gaseous ozone in a neutral aqueous medium, require a greater reaction time for delignification when compared to AOP, as used in the present study. The highest removal of hemicelluloses (50.43%) and cellulose (31.55%) also occurred in experiment 12. Some authors have reported that in the oxidation of lignocellulosic compounds, the most susceptible material to degradation is lignin, followed by hemicelluloses, but with little preference for cellulose (Li et al., 2015; Schultz-Jensen et al., 2011; Souza-Corrêa et al., 2013; Travaini et al., 2016). However, according to Mussatto and Dragone (2016), in aqueous medium and alkaline pH, the ozone dissolved in water generates hydroxyl radicals that react with carbohydrates, resulting in random cleavage of glycosidic bonds and formation of acids, alcohols and aldehydes (Ragnar et al., 1999; Travaini et al., 2014) being able to solubilize even part of the cellulose (Doménech et al., 2001). Therefore, when the objective is the production of 2G ethanol, the conditions with greater removal of cellulose are not so interesting, because part of the cellulose removed could be hydrolyzed in the stage of enzymatic hydrolysis and later converted into ethanol. However, when the aim is the biogas production from anaerobic digestion using hydrolysate from pretreated biomass as substrate, the dissolution of cellulose can contribute to enhance methane generation. It is important to report the observed relationship between removals of lignin and hemicelluloses. The solubilization of hemicelluloses may increase with increasing delignification because the hemicelluloses chains can be attacked by radicals generated in the aqueous phase, which may lead to the solubilization of both constituents of the ligninhemicelluloses complex. This behavior was also reported by as GarcíaCubero et al. (2009) and Li et al. (2015). In the study of Adarme et al. (2017), the experiment that most removed lignin was also the one that most solubilized hemicelluloses (48.28%), which is in agreement with the results found in the present study. In a study performed by Binder et al. (1980), wheat straw was ozonized aiming its delignification and the authors reported that this process improved cellulose degradability. These authors observed that when cellulose fibers were exposed to ozone, there was the formation of several shorter cellulose chains, improving enzymatic hydrolysis in the case of 2G ethanol production. The reaction mechanism for the ozone attack to carbohydrates is explained by Olkkonen et al. (2000) and Travaini et al. (2014) and has several fragmentation routes. Degradation occurs by means of ozone

The amount of electric energy (EE) and thermal energy (TE) that could be generated by biogas burning in the CHP system was calculated using Eqs. (4) and (5), respectively.

MJ ⎞ 3 3 EE / ⎛⎜ ⎟ = LCVbiogas (MJ / Nm ) × Nm biogas / kgCODhydrolysate ⎝ kgCH ⎠ × kgCODhydrolysate / kgCH × 0.85 × 0.35

(4)

MJ ⎞ 3 3 TE / ⎛⎜ ⎟ = LCVbiogas (MJ / Nm ) × Nm biogas / kgCODhydrolysate ⎝ kgCH ⎠ × kgCODhydrolysate / kgCH × 0.85 × 0.65

(5)

3. Results and discussion 3.1. Characterization of coffee husks and fractions obtained in the OOP Chemical characterization of CH in natura was performed on a dryweight basis, wt%) and presented the following average results: 32.5( ± 1.1)% cellulose, 20.8( ± 0.5)% hemicelluloses, 27.1( ± 0.8)% lignin, 22.0( ± 1.6)% extractives and 4.5( ± 1.7)% ash. These values are within the range of values reported in the literature (cellulose: 24.5–43%, hemicelluloses: 7–29.7%, lignin: 9–30.2%, ash: 3–10.7%) for this kind of lignocellulosic material (Mussatto et al., 2012; Pandey et al., 2000). Based on the results presented in Table 1, it was possible to observe that the ozonation conditions that led to the highest removal of lignin, cellulose and hemicelluloses were those used in the experiment 12: LSR 15 mL/g, pH 11 and SAOL 19.14 mg O3/g CH. Further removal of these compounds was expected in the alkaline pH experiments due to the fact that at this pH the process is considered an AOP (Glaze, 1986) and the release of hydroxyl radicals probably favored the attack on biomass components resulting in rapid and non-specific reactions, as typical in radical oxidation. It is worth mentioning that experiment 12 had a relatively short time (14.2 min), which contributes economically to the process, taking into account that the generation of ozone demands energy expenditure. Pareto’s chart of the standardized effects (see Supplementary Material) confirmed that only the pH had a significant positive effect on lignin removal (4.712). For the removal of hemicelluloses, the independent variable (pH) had a significant positive effect (2.887). However, the interaction between pH and SAOL had a significant negative effect of greater magnitude (−4.111). In addition, Pareto’s chart showed that pH and the interaction between LSR and SAOL had a significant positive effect on cellulose removal (6.598 and 3.079, respectively). On the other hand, LSR and the interaction between SAOL and pH, had a significant negative effect (−3.851 and −4.826, respectively), with the pH effect being the most pronounced among all parameters. Response surface graphs (Fig. 1a–c) generated by the quadratic model resulting from DED were created. The adjusted model presented coefficient of determination (R2) values of 0.7898, 0.7980 and 0.9285 for removal of lignin, hemicelluloses and cellulose, respectively. Analyzing the results, it is possible to observe that in order to achieve significant removals of lignin, hemicelluloses and cellulose, maintaining mean values of LSR, it was necessary to use pH values above 8 and low SAOL values, representing a radical reaction. On the other hand, when using pH values lower than 8, with high SAOL values, the constituents of the lignocellulosic biomass were also removed, although such removal values were not as high as those observed at alkaline pH, as well as the need of longer ozonation times. In this case, ozone reacted directly with CH, promoting a slow and selective reaction (Doménech et al., 2001). The presence of hydroxyl ions (HO−) in high concentration can initiate the decomposition of molecular ozone, leading to the formation 606

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

most produced biogas (91.07 NmL CH4/g CODapplied), which was expected by the desirability tool, whose desired conditions were higher removal of lignin, cellulose and hemicelluloses. This condition applied AOP (pH 11), lower amount of water (LSR 10 mL/g) and one of the shortest reaction times (13.7 min), with a SAOL of 18.50 mg O3/g CH. These results are consistent with the 19 experiments generated by the DED (Table 1), since experiment 12 was the one that most removed lignin, cellulose and hemicelluloses, where the conditions were LSR 15 mL/g, pH 11, SAOL 19.14 mg O3/g CH and reaction time of 14.2 min, demonstrating that the prediction made by the desirability tool was satisfactory. Comparing the experiments D6 and D4, in which the latter led to the lowest methane production, it is observed that total sugars in hydrolysate D4 was higher than in hydrolysate D6 (Table 1), indicating that the hydrolysate D4 contained more readily biodegradable material than the hydrolysate D6. Thus, the severity of the pretreatment does not guarantee greater biodegradability to the hydrolysate and the quantity of sugars does not guarantee a higher production of biogas. The lower production of methane in experiment D4 may be an indication that the hydrolysate contained toxic compounds which adversely affected AD. This hypothesis can be confirmed comparing the results of VFA concentration in the experiments D4 and D6. It is possible to observe that the CODVFA at the end of the BMP test for hydrolysate D4 (0.795 g/L) was greater when compared to that observed for hydrolysate D6 (0.580 g/L), indicating that the hydrolysate D4 contained some compounds that contributed to the thermodynamic imbalance of the anaerobic consortium. Depending on the reactivity of the functional group, oxidized compounds may react with excess ozone generating different inhibitory compounds during the OOP in aqueous medium (Travaini et al., 2016). Lignin degradation products represent a variety of aromatic and polyaromatic compounds which, depending on the reaction mechanism, can be converted into saturated or unsaturated carboxylic acids, thereby hampering the AD processes (Ben’ko et al., 2013). Compounds derived from furans such as furanone, 2,5-furanone, D-erythrotetrofuranose and D-erythro-pentofuranose, as well as lignin derivatives such as benzaldehyde and butanal adversely affect anaerobic digestion and were formed during sugarcane bagasse ozonation (Adarme et al., 2017). These authors investigated the biogas production from hydrolysates generated in the OOP of the sugarcane bagasse and showed anaerobic biomass was inhibited by the ozonation byproducts. In addition, Jönsson and Martín (2016), reviewed the formation of inhibitory byproducts and strategies for minimizing their effects for various types of pretreatments applied to lignocellulose. These authors reported the formation of furan derivatives such as furoic acid and phenolic acids such as 4-hydroxyphenoloic, vanillic and syringic acids as the main responsible for inhibition of microorganism’s activity. At the end of the single-staged BMP (Fig. 2a), analysis of VFAs showed that there was butyric and isobutyric acids accumulation. According to Aquino and Chernicharo (2005), the accumulation of VFAs is related to the absence of ideal conditions for biological growth (kinetic limitations) or to the presence of toxic compounds that affects slow growing microorganisms, such as acetogenic and methanogenic acetoclastic, resulting in low methane production. By comparing experiments D6 and D3, it can be inferred that the lower methane production in experiment D3 may be related to the toxicity of lignin fragments, which probably inhibited methanogenic and acetogenic microorganisms and resulted in higher isobutyric acid production (Fig. 2a), since experiment D3 was the one in which there was further lignin removal from the solid fraction. In this experiment a significant amount of lignin was transferred to the liquid fraction, which is consistent with the desirability conditions, and might have affected the methanogenic microorganisms. Considering experiments D5 and D6, it is possible to justify the lower methane production of experiment D5 due to the lower total sugar content (Table 1). Since there was no VFA accumulation in these

reaction with the carbohydrates at the anomeric carbon (C-1) or oxygen that are involved in the glycosidic bonds between anhydroglucopyranose units (AGU) of cellulose, leading to cleavage of these bonds and a series of subsequent reactions, thereby allowing the formation of different compounds containing hydroxyl, carbonyl and carboxyl groups. In alkaline medium, there are also non-selective reactions with radicals, which contribute to the degradation of carbohydrates. The final products may be monosaccharides, lactones, furans, acids and volatile compounds (Olkkonen et al., 2000). From the above, analyzing separately the dependent variables removal of cellulose, hemicelluloses and lignin, it was possible to note that the use of AOP contributed with higher values of removal. However, the best condition can only be chosen by multivariate evaluation by means of the use of desirability tools. 3.2. Desirability tool for OOP optimization for ethanol and/or biogas production For the three experiments (D1, D2 and D3) related to the possible production of 2G ethanol from the solid fraction, the desired conditions were high removal of hemicelluloses and a small removal of cellulose, since in the enzymatic hydrolysis, the cellulose is converted into glucose to be fermented to ethanol. For the other three experiments (D4, D5 and D6), aiming to obtain favorable conditions for the production of biogas from the residual liquid fraction, the desired conditions were high removal of cellulose and hemicelluloses, i.e., the transformation of these complex compounds into more soluble compounds such as glucose, xylose, arabinose, mannose, among others that could be used by anaerobic microorganisms. It was possible to observe in Table 1 that the largest lignin removals occurred in the experiments D3, D4, D5 and D6, both with pH above 8, which is in good agreement with the values predicted by the desirability tool and with the 19 experiments previously made (DED). Values for the removal of hemicelluloses were close to each other, since the desirability tool was adjusted to maintain high removal of such constituent. On the other hand, in the experiment in which there was a greater removal of lignin, there was also a greater removal of hemicelluloses (D3) which, according to Mussatto and Dragone (2016), is explained by non-selective radical reactions that solubilize hemicelluloses with lignin in aqueous alkaline media. Even in the experiments set to have high removal of cellulose, its removals were relatively low when compared to the higher cellulose removals of the previous 19 experiments (DOE, Table 1), as in experiment 12 (31.55% cellulose removal). However, it is worth mentioning that the model for cellulose removal has four-dependent variables with significant effects (pH, LSR, SAOL × pH and LSR × SAOL). Therefore, to meet all the desired conditions simultaneously, the system becomes complex. In addition, because of the structural characteristics, the hemicelluloses chains are more labile than the cellulose ones. While cellulose is crystalline, cohesive and resistant, hemicelluloses have a random and amorphous structure, being more prone to the reactions initiated by the hydroxyl radicals (Monlau et al., 2013). Therefore, hemicelluloses offer greater accessibility and less degree of polymerization than cellulose (Monlau et al., 2013), which might explain its higher removal. 3.3. Biogas production AD experiments were performed with the hydrolysates generated in the ozone pretreatment of CH under six conditions obtained by the desirability tool. Table 2 shows results of methane and hydrogen production related to the content of inoculated organic matter and dryweight of CH used in the pretreatment. 3.3.1. Production of methane by a single-stage anaerobic digestion According to Table 2, AD of the hydrolysate D6 was the one that 607

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Fig. 2. (a) Volatile fatty acids concentration in the end of the single-stage anaerobic digestion (BMP) experiments from desirability conditions, (b) Fitting of the Gompertz model to experimental data of single-stage anaerobic digestion for experiment D6, (c) Volatile fatty acids concentration in the end of the acidogenic (BHP) stage (two-stage anaerobic digestion), (d) Fitting of the two-phase exponential model to experimental data (BHP) of two-stage anaerobic digestion for experiment D5, (e) Volatile fatty acids concentration in the end of the methanogenic (BMP) stage (two-stage anaerobic digestion) and (f) Fitting of the modified Gompertz model to experimental data (BMP) of two-stage anaerobic digestion for experiment D5.

608

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Table 3 Kinetic parameters estimated by modeling the experimental data (desirability conditions) of biogas production in single- and two-stage anaerobic digestion with the modified Gompertz and two-phase exponential models. Methane production in a single-stage anaerobic digestion Model

Parameter

Experiment D1

D2

D3

D4

D5

D6

78.75 0.1292 79.5 0.9929 2.1813 3.0028 84.9807

93.47 0.1692 55.2 0.9903 3.0962 3.4992 120.0080

70.66 0.1179 28.1 0.9955 1.5551 2.3205 51.1420

56.34 0.1023 21.3 0.9941 1.4317 2.6229 42.8794

87.42 0.1590 55.9 0.9928 2.4957 3.0537 98.4461

92.17 0.2156 53.4 0.9979 1.5272 1.6766 49.3329

Hydrogen production in a two-stage anaerobic digestion Two-phase exponential P1 (NmL H2/g CODapplied) P2 (NmL H2/g CODapplied) k1 (h−1) k2 (h−1) R2 RMSE NRMSE AIC

26.74 21.97 0.0822 0.0022 0.9764 1.4366 3.3143 39.4891

28.52 30.14 0.0899 0.0027 0.9797 1.6922 3.1754 53.1947

24.85 36.98 0.0713 0.0025 0.9904 1.2687 2.3153 29.1725

14.23 29.48 0.0954 0.0034 0.9902 0.9938 2.4713 8.7884

36.05 20.82 0.0831 0.0059 0.9757 1.9367 3.4679 64.5162

29.19 21.94 0.0796 0.0037 0.9801 1.5364 3.1661 45.1433

Methane production in a two-stage anaerobic digestion Modified Gompertz P0 (NmL CH4/g CODapplied) µm (NmL CH4/h g CODapplied) λ (h) R2 RMSE NRMSE AIC

177.81 0.4187 132.2 0.9981 2.8695 1.6177 82.8854

148.92 0.3353 114.5 0.9924 4.7685 3.0738 119.4519

279.38 0.5172 150.8 0.9965 5.3554 2.0806 127.8102

166.76 0.3638 160.2 0.9953 3.9545 2.4373 105.9755

349.56 0.7437 134.8 0.9959 7.9890 2.3507 156.6072

281.87 0.5591 140.4 0.9944 7.1883 2.6305 149.0029

Modified Gompertz

P0 (NmL CH4/g CODapplied) µm (NmL CH4/h g CODapplied) λ (h) R2 RMSE NRMSE AIC

production of this biogas in experiment D2, which was also close to that of experiment D5, it can be inferred that lower water content in the pretreatment also contributed positively to the hydrogen production. In these experiments (D2 and D5) water content was lower than in experiments D1 and D4, which obtained lower hydrogen production. The only difference between experiments D3 and D4 was the water content, which confirms the previous hypothesis. During the acidogenic stage of anaerobic digestion, short chain organic acids are the main products, since they are byproducts of biodegradable compounds (Baêta et al., 2016a). In the present study, at the end of the acidogenic stage, the main VFAs were acetic and propionic acids (Fig. 2c). It is observed that the anaerobic degradation of the hydrolysates obtained in all the experiments favored the acetic route, since the concentration of this acid was higher than the others. This is good from the standpoint of process stability since acetate is a direct substrate of methanogenesis. It is known that propionic acid is a common intermediate of sugar metabolism when there is accumulation of hydrogen in the bulk solution (Speece, 1996). It can also be formed from long chain organic acids and more complex organic compounds, and its accumulation normally reflects an imbalance between metabolic stages that govern anaerobic digestion (Mesquita et al., 2013). Somehow, this has occurred in all experiments, since the accumulation of acetic acid in the medium (due to the absence of acetoclastic methanogens) probably triggered metabolic changes in the acidogenic microorganisms to minimize the accumulation of hydrogen and to guarantee the internal recycling of the electrons carrier, such as NAD (Aquino and Chernicharo (2005)). The conversion of propionate into acetate is not thermodynamically favorable under standard conditions (ΔG° = +76.1 kJ/mol) and, therefore, this reaction is inhibited at high concentrations of hydrogen and acetate, causing longer chain organic acids accumulation in the medium (Aquino and Chernicharo, 2005). The two-phase exponential model was used to model the kinetics of hydrogen production and showed good fits to the experimental data, since the coefficients of determination (R2) were higher than 0.97.

experiments, the data indicates there was no toxicity in the conditions D5 and D6. In experiment D3 a significantly higher load of ozone was applied when compared to experiments D5 and D6, therefore the inhibition seems to be related to the severity of OOP, probably due to the higher lignin solubilization and byproducts formation. The modified Gompertz model best fitted the experimental data for the single-stage methane production, and the kinetic parameters estimated for this condition are presented in Table 3. Fig. 2b shows the accumulated methane production against time for the BMP test with the hydrolysate D6, which was the one with the best anaerobic biodegradability, along with the curve fit to the experimental data using the modified Gompertz model. The model was able to reproduce a welladjusted methane production, since the coefficient of determination (R2) was 0.9979. In addition, methane production observed in experiment D6 was 91.07 NmL CH4/g CODapplied, while the model predicted 92.17 NmL CH4/g CODapplied, which is very close to the experimental value. 3.3.2. Production of hydrogen and methane by a two-stage anaerobic digestion Spatial separation of the acidogenic and methanogenic stages is regarded as a promising strategy to increase biogas production from lignocellulosic biomass (Baêta et al., 2016a; Monlau et al., 2013). According to Ghosh et al. (1985), the separation of acidogenesis and methanogenesis enhances the hydrolysis step, leading to a better stabilization of organic material as well as a higher biogas production. According to data presented in Table 2, the experiment D5 was the one with higher hydrogen production (55.84 NmL H2/g CODapplied). This experiment employed AOP conditions (pH 9), less water (LSR 10 mL/g) and lower reaction time (5 min), with a SAOL of 6.75 mg O3/ g CH. In experiment D3, hydrogen production was close to that of experiment D5 and the difference between these experiments were the applied ozone loads, which was 81 mg O3/g CH (60 min) in D3. This shows that lower severity and shorter pretreatment time contributed to a higher hydrogen production. However, when analyzing the 609

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

with the hydrolysate of the condition D6, as it led to higher production of methane in the single-stage and had the second lowest pretreatment severity and, thereby, a lower energy cost. PAC was added to the glass bottle, according to the methodology described in Section 2.6 and the results are presented in Table 2. In a study with anaerobic submerged membrane bioreactor treating effluent containing azodye Baêta et al. (2013) used PAC (4.0 g/L) and obtained excellent performance in the removal of organic matter, color, turbidity and VFAs. As PAC was added directly into the glass bottle, the adsorbed compounds were slowly desorbed as the biodegradation occurred in the liquid phase due to the shift of chemical equilibrium, especially with more biodegradable compounds such as VFAs. Therefore, costs related to the PAC addition within the bioreactor are smaller because the biodegradation would promote a desorption and regeneration of adsorptive sites of PAC. As seen in Table 2, the use of PAC contributed to the reactor’s stability and decreased VFAs accumulation, since these intermediates were found at the end of the BMP test. In addition, when comparing the BMP tests using PAC with previous experiments (single-stage and two-stage BMP), it was possible to observe that PAC enhanced methane production by 2.4 times in a single-stage AD and by 1.8 times in the two-stage process. This strengthens the hypothesis that inhibitory compounds hampered methanogenesis and show the process could be mitigated by PAC adsorption. Future study may be done considering the use of adsorbents (activated carbon or even char – without activation) from coffee husks and other biomasses from coffee production.

Kinetic parameters estimated by the two-phase exponential model for the hydrogen production in two-stage AD are presented in Table 3. For experiment D5 (Fig. 2d), the two-phase exponential model assumes that hydrogen production occurred in two phases, with maximum yield of 55.84 NmL H2/g CODapplied. Using the hydrolysate from experiment D5, the first phase corresponded to 64.55% of the accumulated hydrogen production (36.05 NmL H2/g CODapplied), while the second phase corresponded to the remaining 35.45% (20.82 NmL H2/g CODapplied). The first phase lasted for approximately 30 h and the second phase corresponded to approximately 686 h. The higher initial hydrogen production may be related to the conversion of easily biodegradable substances such as sugars (glucose, xylose and arabinose) into acids (Abreu et al., 2012; Sá et al., 2014). On the other hand, the second phase may be related to the conversion of more complex substances as longer chain acids (e.g., butyric and propionic), oligomers, among others. After the acidogenic stage, the methanogenic step was performed using the BHP flask supernatant as substrate. The results of methane production were considered as a function of the organic matter content that came out from the acidogenic stage. Therefore, to compare the results with the single-stage BMP tests, data were normalized to NmL CH4/g CH (Table 2). According to data presented in Table 2, the two-stage anaerobic digestion improved methane production for almost all tests. Experiment D5 was the two-stage BMP test which produced the highest amount of methane (363.83 NmL CH4/g CODapplied), since the acetic acid concentration at the end of the acidogenic stage in experiments D3, D5 and D6 were the highest. This makes these experiments the most likely to obtain higher methane production in the two-stage BMP, since acetic acid is a direct precursor of methanogenic acetoclastic microorganisms (Aquino and Chernicharo, 2005). In Fig. 2c, it is possible to observe that experiments D5 and D6 accumulated higher concentrations of propionic and butyric acids, which indicates that these acids were converted into acetate and hydrogen by acetogenic microorganisms and later converted into methane by methanogenic microorganisms. Methane production values (Table 2) confirm these hypotheses. The VFAs (mainly isobutyric) that accumulated at the end of the two-stage BMP (Fig. 2e) were reduced by 3–7 times in relation to the single-stage BMP (Fig. 2a). According to Baêta et al. (2016a), the acidogenic stage probably acted detoxifying the medium, since the fast growing acidogenic microorganisms are more resilient and prone to assimilate part of the toxic compounds present in the hydrolysates. The conversion of toxic or recalcitrant compounds into more easily biodegradable ones in the acidogenic step contributes to a lower accumulation of longer chain VFAs and a greater and faster methane production in the methanogenic stage. The modified Gompertz model was also the kinetic model that best fitted the experimental data for the two-stage methane production. Kinetic parameters estimated by this model for the methane production in the two-stage AD are presented in Table 3. Fig. 2f shows the accumulated methane production for BMP test with hydrolysate D5, which was the one that exhibited the highest methane production in two-stage AD, along with the curve fit (R2 = 0.9959) to the experimental data using the modified Gompertz model.

3.4. Energy balance The values of energy input and output are shown in Fig. 3a, which shows that in both single-stage and two-stage, the anaerobic digestion process was only energetically viable for experiments D5 and D6, since they were performed in milder OOP conditions, and the value of the applied ozone (SAOL) was the lowest for these experiments (Table 1). The aim of using PAC during the anaerobic digestion process with the hydrolysate generated in experiment D6 was to try to maximize methane production. As experiment D6 presented the highest methane production in the single-stage anaerobic digestion and lower severity (the second lowest), this implies that the additional costs related to the use of PAC would still be low when compared to those with higher severity. Fig. 3b shows the energy estimates for single- and two-stage anaerobic digestion experiments, considering the hydrolysate generated in experiment D6. In Fig. 3b, the use of PAC improved energy recovery by 9.7 times in comparison to the single-stage AD and by 2.2 times in comparison to the two-stage AD. Although the use of PAC increased energy recovery, it is necessary to perform an economic analysis to take into account the costs associated with the use of this adsorbent and the viability of its use. Considering the energy recovery obtained for the condition D6 with PAC (0.58 kJ/kg) and all coffee husks generated during the 2017 crop season (2.7 million tons), the potential of generation of energy if all waste is converted to methane is 0.435 GWh. 4. Conclusions

3.3.3. Use of powdered activated carbon to improve biogas production As seen before the toxicity of intermediated compounds generated by the OOP of coffee husks impaired anaerobic digestion of CH hydrolysates. The impairment may be due to the increase in redox potential and an oxidative stress put upon the anaerobic microorganisms, or due to the formation of inhibiting compounds during ozonation. One way of improving anaerobic digestion is removing the toxicants which are likely to be lignin degradation byproducts. Adsorption on activated carbon is a technological alternative to remove organic components of low molar mass, such as phenols and their derivatives, from aqueous media. Therefore, an experiment with PAC addition was performed

The OOP of CH was effective for solubilization of lignin and hemicelluloses and for cellulose preservation in the solid fraction. The highest biomass solubilization did not necessarily represent the best conditions for CH4 production in single-stage anaerobic digestion (AD) due to the presence of toxic compounds. The addition of powdered activated carbon or the adoption of two-stage AD process proved to be successful strategies to reduce hydrolysate toxicity and improve biogas production. For two conditions tested (D5 and D6), the energy recovered with biogas burning in CHP system offset the OOP energy expenditure. 610

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

Fig. 3. (a) Energy balance for anaerobic digestion of hydrolysates generated from oxidative ozone pretreatment (OOP) of coffee husks (CH) from desirability conditions and (b) energy estimate for experiments of anaerobic digestion of hydrolysate generated from OOP of CH from experiment D6.

Acknowledgments

(7), 1108–1113. Binder, A., Pelloni, L., Fiechter, A., 1980. Delignification of straw with ozone to enhance biodegradability. Eur. J. Appl. Microbiol. Biotechnol. 11 (1), 1–5. CONAB, 2017. Companhia Nacional de Abastecimento (CONAB). Observatório Agrícola. Acompanhamento da safra brasileira de café, quarto levantamento da safra 2017/ 2018. 4. Couto, L.C., Couto, L., Watzlawick, L.F., Câmara, D., 2004. Vias de valorização energética da biomassa. Biomass. Energia 1, 71–92. Doménech, X., Litter, M., Jardim, W., 2001. Procesos avanzados de oxidación para la eliminación de contaminantes. In: Blesa, M. (Ed.), Eliminación de contaminantes por Fotocatálisis Heterogénea. CYTED, La Plata, Argentina, pp. 3–26. García-Cubero, M.T., González-Benito, G., Indacoechea, I., Coca, M., Bolado, S., 2009. Effect of ozonolysis pretreatment on enzymatic digestibility of wheat and rye straw. Bioresour. Technol. 100 (4), 1608–1613. Ghosh, S., Ombregt, J.P., Pipyn, P., 1985. Methane production from industrial wastes by two-phase anaerobic digestion. Water Res. 19 (9), 1083–1088. Glaze, W.H., 1986. Reaction products of ozone: a review. Environ. Health Perspect. 69, 151–157. Gottschalk, C., Libra, J.A., Saupe, A., 2010. Ozonation of Water and Waste Water: A Practical Guide to Understanding Ozone and Its Applications, second ed. Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, Germany. Gurgel, L.V.A., Pimenta, M.T.B., Curvelo, A.A.d.S., 2014. Enhancing liquid hot water (LHW) pretreatment of sugarcane bagasse by high pressure carbon dioxide (HP-CO2). Ind. Crop. Prod. 57, 141–149. Hung, Y.-T., Sumathi, S., 2004. Treatment of pulp and paper mill wastes. In: Wang, L.K., Hung, Y.-T., Lo, H.H., Yapijakis, C. (Eds.), Handbook of Industrial and Hazardous Wastes Treatment, second ed. CRC Press, Boca Raton, FL, pp. 495–542. Jönsson, L.J., Martín, C., 2016. Pretreatment of lignocellulose: formation of inhibitory byproducts and strategies for minimizing their effects. Bioresour. Technol. 199, 103–112. Jung, B., Theato, P., 2013. Chemical strategies for the synthesis of protein–polymer conjugates. In: Schlaad, H. (Ed.), Bio-synthetic Polymer Conjugates. Springer Berlin Heidelberg, Berlin, Heidelberg, pp. 37–70. Kaur, U., Oberoi, H.S., Bhargav, V.K., Sharma-Shivappa, R., Dhaliwal, S.S., 2012. Ethanol production from alkali- and ozone-treated cotton stalks using thermotolerant Pichia kudriavzevii HOP-1. Ind. Crop. Prod. 37 (1), 219–226. Kumar, P., Barrett, D.M., Delwiche, M.J., Stroeve, P., 2009. Methods for pretreatment of lignocellulosic biomass for efficient hydrolysis and biofuel production. Ind. Eng. Chem. Res. 48 (8), 3713–3729. Lamaison, F.C., Reginatto, V., Antônio, R., 2015. Efeito de diferentes pré-tratamentos de cultura mista para o enriquecimento em bactérias produtoras de hidrogênio. Blucher Chem. Eng. Proc. 1 (2), 717–724. Lazaro, C.Z., Perna, V., Etchebehere, C., Varesche, M.B.A., 2014. Sugarcane vinasse as substrate for fermentative hydrogen production: the effects of temperature and substrate concentration. Int. J. Hydrogen Energy 39 (12), 6407–6418. Li, C., Wang, L., Chen, Z., Li, Y., Wang, R., Luo, X., Cai, G., Li, Y., Yu, Q., Lu, J., 2015. Ozonolysis pretreatment of maize stover: The interactive effect of sample particle size and moisture on ozonolysis process. Bioresour. Technol. 183, 240–247. Lima, D.R.S., Adarme, O.F.H., Baêta, B.E.L., Gurgel, L.V.A., de Aquino, S.F., 2018. Influence of different thermal pretreatments and inoculum selection on the biomethanation of sugarcane bagasse by solid-state anaerobic digestion: a kinetic analysis. Ind. Crop. Prod. 111, 684–693. Mesquita, P.D.L., Afonso, R.J.D.C.F., Aquino, S.F.D., Leite, G.D.S., 2013. Validação de método de cromatografia líquida para a determinação de sete ácidos graxos voláteis intermediários da digestão anaeróbia. Engenharia Sanitaria e Ambiental 18,

The authors would like to thank the Federal University of Ouro Preto (UFOP, Brazil), the Brazilian National Research Council (CNPq, Brazil) and the Minas Gerais State Research Funding Agency (FAPEMIG, Brazil) for their financial support. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at https://doi.org/10.1016/j.biortech.2018.05.037. References Abreu, A.A., Karakashev, D., Angelidaki, I., Sousa, D.Z., Alves, M.M., 2012. Biohydrogen production from arabinose and glucose using extreme thermophilic anaerobic mixed cultures. Biotechnol. Biofuels 5 (1), 6. Adarme, O.F.H., Baêta, B.E.L., Lima, D.R.S., Gurgel, L.V.A., de Aquino, S.F., 2017. Methane and hydrogen production from anaerobic digestion of soluble fraction obtained by sugarcane bagasse ozonation. Ind. Crop. Prod. 109, 288–299. Angelidaki, I., Alves, M., Bolzonella, D., Borzacconi, L., Campos, J., Guwy, A., Kalyuzhnyi, S., Jenicek, P., Van Lier, J., 2009. Defining the biomethane potential (BMP) of solid organic wastes and energy crops: a proposed protocol for batch assays. Water Sci. Technol. 59 (5), 927–934. Apha, A., 2005. WWE, Standard Methods for the Examination of Water and Wastewater, twenty first ed. American Public Health Association/American Water Works Association/Water Environment Federation, Washington DC, USA. Aquino, S.F.D., Chernicharo, C.A.L., 2005. Acúmulo de ácidos graxos voláteis (AGVs) em reatores anaeróbios sob estresse: causas e estratégias de controle. Engenharia Sanitaria e Ambiental 10, 152–161. Assalin, M.R., Durán, N., 2007. Novas tendências para aplicação de ozônio no tratamento de resíduos: ozonização catalítica. Rev. Anal. 26, 76–86. Baêta, B.E.L., Cordeiro, P.H.D.M., Passos, F., Gurgel, L.V.A., de Aquino, S.F., Fdz-Polanco, F., 2017. Steam explosion pretreatment improved the biomethanization of coffee husks. Bioresour. Technol. 245, 66–72. Baêta, B.E.L., Lima, D.R.S., Adarme, O.F.H., Gurgel, L.V.A., Aquino, S.F.D., 2016a. Optimization of sugarcane bagasse autohydrolysis for methane production from hemicellulose hydrolyzates in a biorefinery concept. Bioresour. Technol. 200, 137–146. Baêta, B.E.L., Lima, D.R.S., Filho, J.G.B., Adarme, O.F.H., Gurgel, L.V.A., Aquino, S.F.D., 2016b. Evaluation of hydrogen and methane production from sugarcane bagasse hemicellulose hydrolysates by two-stage anaerobic digestion process. Bioresour. Technol. 218, 436–446. Baêta, B.E.L., Luna, H.J., Sanson, A.L., Silva, S.Q., Aquino, S.F., 2013. Degradation of a model azo dye in submerged anaerobic membrane bioreactor (SAMBR) operated with powdered activated carbon (PAC). J. Environ. Manage. 128, 462–470. Barakat, A., Monlau, F., Steyer, J.-P., Carrere, H., 2012. Effect of lignin-derived and furan compounds found in lignocellulosic hydrolysates on biomethane production. Bioresour. Technol. 104, 90–99. Ben’ko, E.M., Manisova, O.R., Lunin, V.V., 2013. Effect of ozonation on the reactivity of lignocellulose substrates in enzymatic hydrolyses to sugars. Russ. J. Phys. Chem. A 87

611

Bioresource Technology 263 (2018) 601–612

L.C.d. Santos et al.

lignin and carbohydrate model compounds. Holzforschung 53 (3), 292–298. Sá, L.R.V.D., Cammarota, M.C., Ferreira-Leitão, V.S., 2014. Produção de hidrogênio via fermentação anaeróbia - aspectos gerais e possibilidade de utilização de resíduos agroindustriais brasileiros. Quim. Nova 37, 857–867. Schultz-Jensen, N., Kádár, Z., Thomsen, A.B., Bindslev, H., Leipold, F., 2011. Plasmaassisted pretreatment of wheat straw for ethanol production. Appl. Biochem. Biotechnol. 165 (3), 1010–1023. Souza-Corrêa, J.A., Ridenti, M.A., Oliveira, C., Araújo, S.R., Amorim, J., 2013. Decomposition of lignin from sugar cane bagasse during ozonation process monitored by optical and mass spectrometries. J. Phys. Chem. B 117 (11), 3110–3119. Speece, R.E., 1996. Anaerobic Biotechnology for Industrial Wastewaters. Archae Press, United States. Sun, Y., Cheng, J., 2002. Hydrolysis of lignocellulosic materials for ethanol production: a review. Bioresour. Technol. 83 (1), 1–11. Taherzadeh, M.J., Karimi, K., 2008. Pretreatment of lignocellulosic wastes to improve ethanol and biogas production: a review. Int. J. Mol. Sci. 9 (9), 1621–1651. Travaini, R., Marangon-Jardim, C., Colodette, J.L., Morales-Otero, M., Bolado-Rodríguez, S., 2014. Ozonolysis. In: Pandey, A., Negi, S., Binod, P., Larroche, C. (Eds.), Pretreatment of Biomass: Processes and Technologies. Elsevier, Amsterdam, Netherlands, pp. 105–135. Travaini, R., Martín-Juárez, J., Lorenzo-Hernando, A., Bolado-Rodríguez, S., 2016. Ozonolysis: an advantageous pretreatment for lignocellulosic biomass revisited. Bioresour. Technol. 199, 2–12. Travaini, R., Otero, M.D.M., Coca, M., Da-Silva, R., Bolado, S., 2013. Sugarcane bagasse ozonolysis pretreatment: effect on enzymatic digestibility and inhibitory compound formation. Bioresour. Technol. 133, 332–339. Zwietering, M.H., Jongenburger, I., Rombouts, F.M., van 't Riet, K, 1990. Modeling of the bacterial growth curve. Appl. Environ. Microb. 56 (6), 1875–1881.

295–302. Monlau, F., Barakat, A., Trably, E., Dumas, C., Steyer, J.-P., Carrère, H., 2013. Lignocellulosic materials into biohydrogen and biomethane: impact of structural features and pretreatment. Crit. Rev. Env. Sci. Technol. 43 (3), 260–322. Mussatto, S.I., Dragone, G.M. 2016. Biomass pretreatment, biorefineries, and potential products for a bioeconomy development. first ed. In: Marinakis, K., Fedor, John (Eds. ), Biomass fractionation technologies for a lignocellulosic feedstock based biorefinery. Amsterdam, Netherlands. pp. 1–22. Mussatto, S.I., Machado, E.M.S., Carneiro, L.M., Teixeira, J.A., 2012. Sugars metabolism and ethanol production by different yeast strains from coffee industry wastes hydrolysates. Appl. Energy 92, 763–768. Nascimento, E.A., Morais, S.A.L., Aquino, F.J.T., Piló - Veloso, D, 1998. Ozonólise das ligninas organossolve e kraft eucalipto. Parte II: cinética nos meios ácido e básico. Quim. Nova 21, 578–583. Olkkonen, C., Tylli, H., Forsskåhl, I., Fuhrmann, A., Hausalo, T., Tamminen, T., Hortling, B., Janson, J., 2000. Degradation of model compounds for cellulose and lignocellulosic pulp during ozonation in aqueous solution. Holzforschung 54 (4), 397–406. Pandey, A., Soccol, C.R., Nigam, P., Brand, D., Mohan, R., Roussos, S., 2000. Biotechnological potential of coffee pulp and coffee husk for bioprocesses. Biochem. Eng. J. 6 (2), 153–162. Pellera, F.-M., Gidarakos, E., 2016. Effect of substrate to inoculum ratio and inoculum type on the biochemical methane potential of solid agroindustrial waste. J. Environ. Chem. Eng. 4 (3), 3217–3229. Pérez-Elvira, S.I., Fdz-Polanco, F., 2012. Continuous thermal hydrolysis and anaerobic digestion of sludge. Energy integration study. Water Sci. Technol. 65 (10), 1839–1846. Ragnar, M., Eriksson, T., Reitberger, T., 1999. Radical formation in ozone reactions with

612