**Impact of Alkaline H2O<sup>2</sup> Pretreatment on Methane Generation Potential of Greenhouse Crop Waste under Anaerobic Conditions**

### **N. Altınay Perendeci 1,\*, Sezen Gökgöl <sup>1</sup> and Derin Orhon 2,3**


Received: 25 June 2018; Accepted: 16 July 2018; Published: 20 July 2018

**Abstract:** This paper intended to explore the effect of alkaline H2O<sup>2</sup> pretreatment on the biodegradability and the methane generation potential of greenhouse crop waste. A multi-variable experimental design was implemented. In this approach, initial solid content (3–7%), reaction time (6–24 h), H2O<sup>2</sup> concentration (1–3%), and reaction temperature (50–100 ◦C) were varied in different combinations to determine the impact of alkaline H2O<sup>2</sup> pretreatment. The results indicated that the alkaline H2O<sup>2</sup> pretreatment induced a significant increase in the range of 200–800% in chemical oxygen demand (COD) leakage into the soluble phase, and boosted the methane generation potential from 174 mLCH4/g of volatile solid (VS) to a much higher bracket of 250–350 mLCH4/gVS. Similarly, the lignocellulosic structure of the material was broken down and hydrolyzed by H2O<sup>2</sup> dosing, which increased the rate of volatile matter utilization from 31% to 50–70% depending on selected conditions. Alkaline H2O<sup>2</sup> pretreatment was optimized to determine optimal conditions for the enhancement of methane generation assuming a cost-driven approach. Optimal alkaline H2O<sup>2</sup> pretreatment conditions were found as a reaction temperature of 50 ◦C, 7% initial solid content, 1% H2O<sup>2</sup> concentration, and a reaction time of six h. Under these conditions, the biochemical methane potential (BMP) test yielded as 309 mLCH4/gVS. The enhancement of methane production was calculated as 77.6% compared to raw greenhouse crop wastes.

**Keywords:** alkaline H2O<sup>2</sup> pretreatment; breakdown of lignocellulosic structure; greenhouse crop waste; methane generation; process optimization

#### **1. Introduction**

In the last few decades, there was a drastic change in the conceptual understanding of waste management. Waste is no longer considered as matter to be disposed of at the expense of additional cost, but as a resource. Perhaps the most significant resource component is energy in view of the present and future energy shortages expected, due to demands of rapid population expansion and escalating industrial activities in the world. Therefore, energy recovery from waste is now a hot topic, both in terms of scientific efforts and practical applications.

Recently, renewable energy sources, such as solar energy, wind energy, and geothermal energy, are now being largely explored and exploited. Among these categories, biomass energy should be given specific emphasis mainly due to its accessibility; the energy recovery from biomass is also quite sustainable as the proper disposal of biomass requires costly technical processes. Agricultural waste is

an important component of the wide spectrum of waste sources considered within the scope of biomass energy [1]. This study focused on greenhouse agriculture, a significant agricultural practice in areas with a suitable climate, like the Antalya region on the southern coast of Turkey. The Mediterranean region is one of the most important areas in terms of protected cultivation because the mild winter makes production under simple structures possible [2]. Greenhouses provide a protected growing environment that can be controlled during the year. This allows intensive culture with annual yields many times higher than that of field production [3]. Turkey holds an important place in the world for the production of fresh fruit and vegetables, having close to 752,000 decares of greenhouse-covered land, placing it fifth in the world after China, South Korea, Spain, and Japan. About 278,000 decares of greenhouse land is located in the Antalya province, which corresponds to approximately 36.97% of greenhouse land in Turkey. Furthermore, 51% of Turkey's greenhouse vegetable production (3.2 million tons) is provided by Antalya. Greenhouse agriculture is very significant in the districts of Alanya, Aksu, Elmali, Gazipa¸sa, Kepez, Korkuteli, Kumluca, Manavgat, and Serik.

While total greenhouse production (tomato, pepper, cucumber, eggplant, and zucchini) was 2,256,325 tons, 1,087,247.75 tons of greenhouse crop waste was produced in the production year of 2005–2006 [4]. Unfortunately, greenhouse cultivation waste lignocellulosic residue is improperly disposed into the environment in Turkey. The conventional disposal methods for most of this waste, such as unconfined storage in forests and road edges, landfilling, and uncontrolled burning, cause significant environmental problems [5]. A limited quantity of greenhouse crop waste is also used for mulching. However, growers prefer not to apply mulching, due to the spread of some diseases and the transfer of non-biodegraded pesticides, herbicides, and others for the subsequent cultivation period. Landfilling is the most applied waste management practice, and results in the release of CH<sup>4</sup> which is around 20 times more potent as a greenhouse gas (GHG) than CO2. Landfilling was shown to be the greatest source of GHG emissions, contributing more than 75% of total emissions associated with waste management [6]. Uncontrolled burning and/or incineration of greenhouse crop waste emits CO<sup>2</sup> and N2O, a GHG gas 310 times more powerful in atmospheric warming than CO2. In addition, uncontrolled burning and/or incineration diverts waste from landfill, reducing the amount of methane generated. However, combustion also produces waste in the form of ash. Eventually, waste crops disposed from greenhouses were found to be a renewable and cost-free source of lignocellulosic biomass, whose management is necessary to prevent environmental pollution and to gain an alternative utilization as a fuel biogas. Greenhouse crop waste involves all parts left in the field after the harvest, including roots, stems, leaves, rotten/spoiled vegetables, etc. What makes this category of agricultural waste interesting is its complex lignocellulosic structure, whereby the residue contains cellulose (35–50%), hemicellulose (20–35%), lignin (10–25%), and minor fractions of proteins, oils, and ash [7,8] in such a way that the cellulose is embedded in a lignin–polysaccharide sheet [9]. This structure resists microbial destruction and hydrolysis, and requires pretreatment before an energy recovery process.

Many pretreatment technologies were suggested in the literature, such as physical pretreatment, which generally involves mechanical methods such as shredding and grinding [10,11]. Ultrasonic and microwave methods were also tested [12], but were not recommended due to phenolic by-products and the high energy costs involved [13]. Some physico-chemical methods, based on pretreatment with ammonia [14], hot water, and steam explosion [15,16] were reported, all claiming success; however, they also depend on conditions consuming high energy. Pretreatment conducted under acidic and alkaline conditions [17,18] was also found to be effective in breaking down the lignocellulosic structure.

The delignification process as a means of lignin removal is widely used to bleach high-lignin wood pulps in the pulp and paper industry [19,20]. The application of alkaline H2O<sup>2</sup> is one of the most effective chemical pretreatment approaches for energy recovery from wastes and residues with a lignocellulosic structure. During the alkaline H2O<sup>2</sup> pretreatment, while H2O<sup>2</sup> plays the role of an oxidant, the role of alkaline is to reduce or remove lignin, acetyl, and other uronic substitutions in the hemicellulosic portions of the biomass via swelling, salvation, and saponification, so that the accessibility and digestibility of holocellulose is enhanced [19]. Thus, theH2O<sup>2</sup> delignification of agricultural wastes is strongly pH-dependent, with an optimal pH of 11.5 for the dissociation reaction of H2O2. During the treatment, alkaline H2O<sup>2</sup> reacts rapidly with lignin to form low-molecular-weight, water-soluble oxidation products. The lignin-oxidizing species is a highly reactive hydroxyl radical (HO·), formed during the degradation of H2O<sup>2</sup> in a reaction with the hydroperoxy anion (HCOO−). HCOO<sup>−</sup> is the active species and is responsible for the bleaching action of H2O<sup>2</sup> under alkaline conditions. On the other hand, hydroperoxyl and hydroxyl radicals generated by the decomposition of H2O<sup>2</sup> are responsible for solubilizing hemicelluloses [21]. This process also has the advantage of not leaving H2O<sup>2</sup> residue, and it is considered as an environmentally friendly and low-cost application [22]. While a large number of studies were conducted using alkaline H2O<sup>2</sup> pretreatment on various types of agricultural waste, such as corn stover, wood waste, soft wood, cashew apple bagasse, energy crops, sugar cane bagasse, agricultural crop stalks, and cotton stalks [19,22–29], this method, although quite promising, remains untested for greenhouse crop wastes.

In this context, the main objective of the study was to carry out an experimental assessment of the effect of alkaline H2O<sup>2</sup> pretreatment on the biodegradability and the methane generation potential of greenhouse crop wastes. A central composite design (CCD) of response surface methodology (RSM) was applied to determine the optimal process conditions of alkaline H2O<sup>2</sup> pretreatment for maximum biogas production in the most cost-effective way. H2O<sup>2</sup> concentration, initial solid content, reaction temperature, and reaction time were selected as independent variables. The effects of these four independent variables on soluble chemical oxygen demand (COD), soluble reducing sugar, total lignin on an extractives free bases, and methane generation potential were investigated in detail. The alkaline H2O<sup>2</sup> pretreatment process was optimized to enhance methane production assuming a cost-driven approach. The effects of the alkaline H2O<sup>2</sup> pretreatment process on the molecular-bond characterization and surface properties of greenhouse crop waste were also examined via Fourier-transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM). To the best of our knowledge, this is the first study on biogas production from greenhouse crop waste with the integration of an alkaline H2O<sup>2</sup> pretreatment process.

#### **2. Results and Discussion**

#### *2.1. Chemical Composition*

The greenhouse crop waste used in the experiment contained around 13.6% dry matter, indicating an average moisture content of more than 86%. The organic fraction of the dry solids, i.e., volatile solids (VS), was measured as 68.7%, mostly composed of lignocellulosic material. The characteristics of the greenhouse crop waste, expressed in terms of major parameters, are presented in Table 1. The cellulose, hemicellulose, lignin, and soluble matter contents of the fresh greenhouse crop waste were measured as 19.49%, 3.89%, 0.03%, and 76.58%, respectively. The elemental composition of the fresh greenhouse crop waste was found to be 29.23% C, 4.89% H, and 2.96% N. The general composition profile reflected in the Table 1 is different from a previous assessment of the same waste [5], which had a different composition. While the composition of mixed greenhouse crop waste was 61.71% tomato, 22.44% cucumber, 7.92 % eggplant, 5.72 % pepper, and 2.21% zucchini in the previous work [5], the composition in this work was 72% tomato, 14.31% cucumber, 5.11% eggplant, 6.69% pepper, and 1.88% zucchini. Furthermore, the green house crop waste used in the previous study [5] was obtained from the Kumluca region, located in west Antalya. On the other hand, the green house crop waste in this study was acquired from the Gazipa¸sa region, located in east Antalya. Conclusively, even though the sampling period was the same, the location and composition of the collected greenhouse crop waste was different. Specifically, the cellulose and hemicellulose contents, together with the carbon content, were found to be lower. The reason is most likely due to sampling done from different cultivation areas, with a different sample composition.


**Table 1.** Average characteristics of the greenhouse crop waste.

\* Determined in extractives soluble in water.

The total COD equivalent of the organic matter in the crop waste was determined as 1.49 gCOD/gVS. This is a significant stoichiometric ratio, quite similar to the *f<sup>X</sup>* value of 1.4 gCOD/gVS, characteristic of biomass in activated sludge systems. This ratio corresponds to the traditional empirical formula of C5H7NO2, which is still in use for the basic stoichiometry of activated sludge [30]. While noting that the measured nitrogen content remains somewhat lower, it would be acceptable to adopt this simplified formula for the COD–VS relationship in greenhouse crop waste.

Table 1 also indicates the magnitude of COD leakage into the solution (soluble COD (sCOD), *ST*) as 61 mgCOD/gVS, and the soluble reducing sugar (sRedSugar) content in this leakage as 7.6 mgCOD/gVS. It should be noted that the soluble sugar component is basically the same as the readily biodegradable COD fraction (*SS*) identified in wastewater [31,32]. It is interesting to note that Sözen et al. [33] reported 5250 mg of COD leakage from 90 g of domestic sludge, quite similar to the 58 mg of *S<sup>T</sup>* per g of dry sludge in "eluate tests" performed for evaluating compliance with the limitation of dissolved organic carbon for the landfilling of municipal treatment sludge.

#### *2.2. Effect of Alkaline H2O<sup>2</sup> Pretreatment*

The directly observable effect of alkaline H2O<sup>2</sup> treatment was the substantial increase in the magnitude of sCOD, as illustrated in Figure 1a. All values in Figure 1a were compared with the sCOD value of 61 mgCOD/gVS in the original raw greenhouse crop waste, in order to visualize the effect of alkaline H2O<sup>2</sup> treatment. Basically, Figure 1a shows that (i) sCOD (*ST*) was increased above 200 mgCOD/gVS in all tests; (ii) the most noticeable increase was observed in experiments conducted at 100 ◦C; in a few experimental runs, *S<sup>T</sup>* exceeded 500 mgCOD/gSV, corresponding to more than an 800% increase compared with the initial COD leakage capacity of the greenhouse crop waste; (iii) the sCOD increase always remained higher when the reaction time was raised to 24 h while other parameters remained the same. This observation is particularly important, since it shows that the H2O<sup>2</sup> dosage was adjusted to increase the amount of sCOD, but not to oxidize and chemically remove the sCOD generated.

**Figure 1.** Increase in soluble chemical oxygen demand (sCOD; (**a**)) and increase in soluble reducing sugar (sRedSugar; (**b)**) due to alkaline H2O<sup>2</sup> treatment.

Figure 1b shows that alkaline H2O<sup>2</sup> treatment also increased the soluble sugar (sRedSugar) leakage. The highest sRedSugar concentration was found to be 32.47 mg of glucose/gVS from the greenhouse crop waste pretreated at a reaction temperature of 100 ◦C, an H2O<sup>2</sup> concentration of 3%, a reaction time of 24 h, and 3% initial solid content, which are the same pretreatment conditions where the maximum increase in sCOD was observed (Figure 1a). It should be remembered that the sRedSugar/sCOD ratio of the greenhouse crop waste before treatment was 12.4% (Table 1). The values displayed in Figure 1b indicate that, while sRedSugar values also increased with H2O<sup>2</sup> treatment, the sRedSugar/sCOD ratio decreased from 12.3% to in the range of 3.9–7.8%.

The effect of alkaline H2O<sup>2</sup> treatment could only be quantified and evaluated in comparison with the methane generation of the raw greenhouse crop waste without pretreatment. The volume of methane produced from the raw greenhouse crop waste was 174 mLCH4/gVS. The experimental outcomes for the biochemical methane potential (BMP) test from the pretreatment experiments are presented in Figure 2. After pretreatment, the highest BMP value was 370.9 mLCH4/gVS, obtained at a reaction temperature of 50 ◦C, an H2O<sup>2</sup> concentration of 2%; a reaction time of 15 h, and 5% initial solid content, while the the lowest BMP value (256.6 mLCH4/gVS) was obtained from the greenhouse crop waste pretreated at a reaction temperature of 100 ◦C, an H2O<sup>2</sup> concentration of 3%, a reaction time of 24 h, and 3% initial solid content. It can be concluded that the dependent variables of sCOD and sRedSugar, which had the maximum values under these conditions. behaved differently than the variable of BMP.

It should be remembered that an initial sCOD amount of 61 mg/gVS was also measured in the greenhouse crop waste. Based on the ratio of 0.35 LCH4/gCOD, now universally recognized as the relationship between sludge COD utilized and methane generated [34], the utilization of the available sCOD would only correspond to 21 mLCH4/gVS. The generation of the remaining 153 mLCH4/gVS has to be related to the hydrolysis of the particulate organics, requiring 0.437 g of particulate COD/gVS. This particulate COD consumption may be converted to 0.31 gVS/gVS, using the previously selected ratio of 1.4 gCOD/gVS. In short, biochemical reactions for raw greenhouse crop waste depleted all available sCOD, and broke down/hydrolyzed 31% of the existing volatile solids, converting them into methane.

The increase in magnitude of methane generation was obviously a direct observation of the effect of alkaline H2O<sup>2</sup> treatment. The first important observation is the escalation in the volume of collected methane to a narrow bracket of 250–350 mLCH4/gVS as a result of alkaline H2O<sup>2</sup> treatment. The second is the relatively lower methane volumes of around 250 mLCH4/gVS associated with the experimental runs conducted at 100 ◦C, despite much higher sCOD levels achieved in the same experiments.

**Figure 2.** Methane generation due to the impact of alkaline H2O<sup>2</sup> pretreatment.

This effect may be further evaluated in terms of (i) the increase in the sCOD levels, and (ii) changes in the levels of particulate organic matter hydrolysis for this purpose. The related evaluations are plotted in Figure 3a,b, which show both the relative contributions of sCOD and the particulate matter hydrolysis. From a different perspective, in the experimental conditions describing a reaction temperature of 50 ◦C, an H2O<sup>2</sup> concentration of 3%, 7% initial solid content, and a reaction time of 24 h, only 78.8 mLCH4/gVS was related to the available sCOD, while 258.8 mLCH4/gVS was produced from the hydrolysis of 52.8% VS. Whereas at a reaction temperature of 100 ◦C, an H2O<sup>2</sup> concentration of 3%, 3% initial solid content, and a reaction time of 24 h, the increased amount of sCOD produced 199.7 mL of the 256.6 mLCH4/gVS generated, while particulate organic matter hydrolysis remained limited to 11.6%. On this basis, the role of the particulate COD breakdown and hydrolysis seemed

reversed at high temperatures. The limitation of methane generation under these conditions may be attributed to the formation of inhibitory by-products likely to be formed during H2O<sup>2</sup> oxidation.

**Figure 3.** Methane production from sCOD and volatile solid (VS) destruction (**a**) and VS breakdown (**b**) due to the impact of alkaline H2O<sup>2</sup> treatment.

The utilization rate of particulate organic matter under anaerobic conditions is an important parameter that reflects the biodegradability characteristics of the waste. The chemical structure of the greenhouse crop waste, dominated by lignocellulosic material, is too complex for biodegradation under natural conditions. In fact, the experiments indicated that only 31% of the waste could be utilized to generate methane without any pretreatment. Alkaline H2O<sup>2</sup> treatment breaks down this complex chemical structure and hydrolyzes it into simple/soluble compounds, detectable by the increase in the magnitude of sCOD. This process significantly affects and increases the biodegradation of the waste. The destruction of the volatile solids takes place in two steps: (i) initial conversion into sCOD, and (ii) partial utilization of volatile solids under anaerobic conditions. For example, at a reaction temperature of 50 ◦C, an H2O<sup>2</sup> concentration of 3%, 7% initial solid content, and a reaction time of 24 h, the incremental sCOD increase between the pretreated and raw samples (∆sCOD) was 164.4 mgsCOD/gVS, corresponding to a VS hydrolysis (∆VS) of 0.117 gVS/gVS. The generation of 338 mLCH4/gVS additionally consumed 0.528 gVS/gVS, with an overall VS destruction calculated as 64.6%. Furthermore, at a reaction temperature of 100 ◦C, an H2O<sup>2</sup> concentration of 3%, 3% initial solid content, and a reaction time of 24 h, ∆sCOD was measured as 509.8 mg/gVS, representing an initial VS hydrolysis of 0.364 gVS/gVS. An additional amount of volatile solids (∆VS) of 0.116 gVS/gVS was also converted into methane, resulting in a lower VS destruction of 48%. These values should be compared with the 40–50% volatile matter utilization in the anaerobic digestion of sewage sludge [35]. The VS utilization profile achieved with alkaline H2O<sup>2</sup> treatment is plotted in Figure 4a. The decrease in utilization rate at high sCOD levels also confirmed the presence and effect of inhibitory oxidation by-products. Furthermore, the experimental outcomes for the total lignin on an extractives free bases are presented in Figure 4b. As plotted in Figure 4b, the lowest total lignin on an extractives free bases was measured as 13.1% from the greenhouse crop waste pretreated at a reaction temperature of 100 ◦C, an H2O<sup>2</sup> concentration of 3%, a reaction time of 24 h, and 7% initial solid content. It should be remembered that the second lowest BMP value of 264.2 mLCH4/gVS was also observed under these conditions.

**Figure 4.** Destruction profile for volatile solids due to the impact of alkaline H2O<sup>2</sup> treatment (**a**). Experimental outcomes for the total lignin on an extractives free bases (**b**).

### *2.3. Alkaline H2O<sup>2</sup> Pretreatment Process Optimization*

The accuracy of the models was explained by the determination coefficient (*R* 2 ) and coefficient of adjusted determination (Adj-*R* 2 ). The *R* <sup>2</sup> values were found to be 0.9682, 0.7740, 0.8376, and 0.5728 for the sCOD, sRedSugar, total lignin on an extractives free bases, and BMP, respectively, whereas the Adj-*R* <sup>2</sup> values were calculated as 0.9562, 0.6966, 0.7762, and 0.4112. The *R* <sup>2</sup> and Adj-*R* <sup>2</sup> values for the models of sCOD, sRedSugar, and total lignin on an extractives free bases in Table 2 indicated that acceptable fits were obtained between the response and the independent variables. However, only moderate *R* <sup>2</sup> and Adj-*R* <sup>2</sup> values were calculated for the BMP model. Quadratic regression models were strongly considerable, as it was apparent from Fisher's *F*-test with very low probability outcomes (*p*-value *>* F = 0.0001 for sCOD, sRedSugar, total lignin on an extractives free bases, and BMP).

Since the objective of alkaline H2O<sup>2</sup> pretreatment was the enhancement of methane production with a reasonable process cost, process optimization of alkaline H2O<sup>2</sup> pretreatment was executed based on minimizing the cost of the process (cost-driven approach) using the models developed for sCOD, sRedSugar, total lignin on an extractives free bases, and BMP. In the cost-driven optimization approach, the dependent variables of sCOD and total lignin on an extractives free bases were set in range, whereas sRedSugar (+) and BMP (+) were maximized. On the other hand, the independent variables of reaction temperature (+++++), reaction time (+++++), and H2O<sup>2</sup> concentration (+++++) were minimized, while VS content (+++++) was maximized.

Optimal alkaline H2O<sup>2</sup> pretreatment conditions were determined with the highest desirability of 0.917 at a reaction temperature of 50 ◦C, 7% initial solid content, an H2O<sup>2</sup> concentration of 1%, and a reaction time of six h under these restraints. The optimal values for sCOD, sRedSugar, total lignin on an extractives free bases, and BMP were predicted to be 296.4 mgsCOD/gVS, 102.1 mg sRedSugar/gVS, 28.7%, and 318.6 mLCH4/gVS, respectively, using the models. An alkaline H2O<sup>2</sup> pretreatment experiment using a cost-driven approach conditions was performed for validation of the process optimization. The values of sCOD, sRedSugar, total lignin on an extractives free bases, and BMP were measured as 290.3 mgsCOD/gVS, 106.9 mg sRedSugar/gVS, 28.1%, and 309 mLCH4/gVS, respectively, supporting the predictive power of the developed models. The BMP enhancement was calculated as 77.6% compared to the raw greenhouse crop waste under the conditions optimized for the process cost.


**Table 2.** ANOVA results for sCOD, sRedSugar, total lignin on an extractives free bases, and biochemical methane potential (BMP) models.

sRedSugar = +844.41473 − 18.34946 × Reaction temp. + 16.89274 × Solid content − 136.48065 × H2O<sup>2</sup> concent. – 13.05242 × Reaction time − 0.17577 × Reaction temp. × Solid content + 1.01831 × Reaction temp. × H2O<sup>2</sup> concent. + 0.063115 × Reaction temp. × Reaction time − 2.93797 × Solid content × H2O<sup>2</sup> concent. + 0.27415 × Solid content × Reaction time + 0.12308 × Reaction temp.<sup>2</sup> − 1.32058 × Solid content<sup>2</sup> + 18.98017 × H2O<sup>2</sup> concent.<sup>2</sup> + 0,29090 × Reaction time<sup>2</sup>

Adeq Precision 11.705 C.V% 41.85

**Table 2.** *Cont.*


1/(Lignin) = +0.0736566 + 5.8380149 × 10−<sup>5</sup> × Reaction temp. − 0.0284772 × Solid content − 7.8491088 × 10−<sup>3</sup> × H2O<sup>2</sup> concent. − 4.5014496 × 10−<sup>4</sup> × Reaction time + 3.4923132 × 10−<sup>5</sup> × Reaction temp. × Solid content + 2.3900179 × 10−<sup>4</sup> × Reaction temp. × H2O<sup>2</sup> concent. − 2.2097257 × 10−<sup>7</sup> × Reaction temp. × Reaction time + 8.6433922 × 10−<sup>4</sup> × Solid content × H2O<sup>2</sup> concent. − 4.3403413 × 10−<sup>5</sup> × Solid content × Reaction time − 3.95154005 × 10−<sup>5</sup> × H2O<sup>2</sup> concent. × Reaction time − 6.01390484 × 10−<sup>6</sup> × Reaction temp.<sup>2</sup> + 2.664218431 × 10−<sup>3</sup> × Solid content<sup>2</sup> − 1.7142995 × 10−<sup>3</sup> × H2O<sup>2</sup> concent.<sup>2</sup> + 2.74301920 × 10−<sup>5</sup> × Reaction time<sup>2</sup>


1/(BMP) = +4.20476 × 10−<sup>3</sup> − 1.31145 × 10−<sup>5</sup> × Reaction temp. − 4.36888 × 10−<sup>5</sup> × Solid content − 9.28724 × 10−<sup>4</sup> × H2O<sup>2</sup> concent. − 4.17111 × 10−<sup>5</sup> × Reaction time + 4.08924 × 10−<sup>7</sup> × Reaction temp. × Solid content + 3.96470× 10−<sup>6</sup> × Reaction temp. × H2O<sup>2</sup> concent. − 8.50445 × 10−<sup>8</sup> × Reaction temp. × Reaction time − 3.64937 × 10−<sup>5</sup> × Solid content × H2O<sup>2</sup> concent. − 2.27112 × 10−<sup>6</sup> × Solid content × Reaction time + 4.32086 × 10−<sup>6</sup> × H2O<sup>2</sup> concent. × Reaction time + 1.02675 × 10−<sup>7</sup> × Reaction temp.<sup>2</sup> + 1.19229 × 10−<sup>5</sup> × Solid content<sup>2</sup> + 2.09008 × 10−<sup>4</sup> × H2O<sup>2</sup> concent.<sup>2</sup> + 1.83088 × 10−<sup>6</sup> × Reaction time<sup>2</sup>


sCOD = +1045.11218 − 24.76191 × Reaction temp. − 44.99164 × Solid content + 88.00049 × H2O<sup>2</sup> concent. − 3.98184 × Reaction time − 0.64327 × Reaction temp. × Solid content + 1.48441 × Reaction temp. × H2O<sup>2</sup> concent. + 0.022507 × Reaction temp. × Reaction time − 27.45672 × Solid content × H2O<sup>2</sup> concent. − 0.000607639 × Solid content × Reaction time − 0.62899 × H2O<sup>2</sup> concent. × Reaction time + 0.19592 × Reaction Temp.<sup>2</sup> + 11.76235 × Solid content<sup>2</sup> + 6.39440 × H2O<sup>2</sup> concent.<sup>2</sup> + 0.21604 × Reaction time<sup>2</sup> .


sRedSugar = +844.41473 − 18.34946 × Reaction temp. + 16.89274 × Solid content − 136.48065 × H2O<sup>2</sup> concent. − 13.05242 × Reaction time − 0.17577 × Reaction temp. × Solid content + 1.01831 × Reaction temp. × H2O<sup>2</sup> concent. + 0.063115 × Reaction temp. × Reaction time − 2.93797 × Solid content × H2O<sup>2</sup> concent. + 0.27415 × Solid content × Reaction time + 0.12308 × Reaction temp.<sup>2</sup> − 1.32058 × Solid content<sup>2</sup> + 18.98017 × H2O<sup>2</sup> concent.<sup>2</sup> + 0.29090 × Reaction time<sup>2</sup>

**Total Lignin on an Extractives Free Bases Model**


1/(Lignin) = +0.0736566 + 5.8380149 × 10−<sup>4</sup> × Reaction temp. − 0.0284772 × Solid content − 7.8491088 × 10−<sup>3</sup> × H2O<sup>2</sup> concent. − 4.5014496 × 10−<sup>4</sup> × Reaction time + 3.4923132 × 10−<sup>5</sup> × Reaction temp. × Solid content + 2.3900179 × 10−<sup>4</sup> × Reaction temp. × H2O<sup>2</sup> concent. − 2.2097257 × 10−<sup>7</sup> × Reaction temp. × Reaction time + 8.6433922 × 10−<sup>4</sup> × Solid content × H2O<sup>2</sup> concent. − 4.3403413 × 10−<sup>5</sup> × Solid content × Reaction time − 3.95154005 × 10−<sup>5</sup> × H2O<sup>2</sup> concent. × Reaction time − 6.01390484 × 10−<sup>6</sup> × Reaction temp.<sup>2</sup> + 2.664218431 × 10−<sup>3</sup> × Solid content<sup>2</sup> − 1.7142995 × 10−<sup>3</sup> × H2O<sup>2</sup> concent.<sup>2</sup> + 2.74301920 × 10−<sup>5</sup> × Reaction time<sup>2</sup>


**Table 2.** *Cont.*


− − − −

Three-dimensional (3D) graphs were employed to emphasize the impacts of independent variables under optimal conditions. The effects of independent variables on BMP are demonstrated in Figure 5a–f. In Figure 5a, BMP decreased due to increasing H2O<sup>2</sup> concentration at a reaction temperature of 100 ◦C, whereas BMP increased due to decreasing reaction temperature (from 100 ◦C to 50 ◦C) within the range of 1–3% H2O<sup>2</sup> concentration. A maximum predicted BMP enhancement of 106.9% compared to the raw greenhouse crop waste was observed at a reaction temperature of 68 ◦C and an H2O<sup>2</sup> concentration of 2%. In Figure 5b, c, BMP decreased when the reaction temperature was increased to 100 ◦C at a reaction time of 24 h and 7% initial solid content. When the reaction time was maintained at 24 h, a decrease in BMP was observed when the temperature was increased to 100 ◦C. Similarly, when the initial solid content was kept constant at 7%, the decrease in BMP was temperature has a negative impact on BMP. Furthermore, as seen in Figure 5d–f, BMP was not affected by the interactive effects of H2O<sup>2</sup> concentration with initial solid content, reaction time with initial solid content, and reaction time with H2O<sup>2</sup> concentration. A maximum BMP was obtained at 4–6% initial solid content, H2O<sup>2</sup> concentrations of 1.5–2.5%, and reaction times of 10–18 h.

**Figure 5.** *Cont.*

**Figure 5.** Effects of independent variables on biochemical methane potential (BMP). (**a**) H2O<sup>2</sup> concentration and temperature; (**b**) reaction time and temperature; (**c**) solid content and temperature; (**d**) H2O<sup>2</sup> concentration and solid content; (**e**) reaction time and solid content; (**f**) reaction time and H2O<sup>2</sup> concentration.

#### *2.4. Chemical Structure and Morphological Changes of Biomass*

The FTIR spectra and SEM images of greenhouse crop waste pretreated with alkaline H2O<sup>2</sup> under different conditions (50 ◦C, 5% VS, 15 h, 2% H2O<sup>2</sup> for maximum CH<sup>4</sup> production; 100 ◦C, 3% VS, 24 h, 3% H2O<sup>2</sup> for maximum sCOD and sRedSugar production, along with minimum CH<sup>4</sup> production; and 50 ◦C, 7% VS, 6 h, 1% H2O<sup>2</sup> for cost optimization) compared to those of the raw greenhouse crop waste are presented in Table 3 and Figure 6.

− − As seen in Figure 6, the spectral profiles and relative intensities of the bands belonging to the raw greenhouse crop waste and that pretreated with alkaline H2O<sup>2</sup> were found to be very similar under conditions of 50 ◦C, 7% VS, 6 h, and 1%H2O<sup>2</sup> for cost optimization. On the other hand, the spectral profiles were different from the raw greenhouse crop waste for that pretreated with alkaline H2O<sup>2</sup> under conditions of 50 ◦C, 5% VS, 15 h, and 2% H2O<sup>2</sup> for maximum CH<sup>4</sup> production, and that pretreated with alkaline H2O<sup>2</sup> under conditions of 100 ◦C, 3% VS, 24 h, and 3% H2O<sup>2</sup> for maximum sCOD and sRedSugar production, along with minimum CH<sup>4</sup> production. New peaks were observed after alkaline H2O<sup>2</sup> pretreatment, indicating that the chemical composition of greenhouse crop waste changed. In particular, the prominent absorbances at 895–900, 1050, 1270, 1430–1460, 1510–1600, 2920–2925, 3420, and 3446 cm−<sup>1</sup> in the spectra were relatively different from the spectrum of raw greenhouse crop waste. As clearly seen in Table 3, the lignin-related absorbance values observed at 1270, 1430–1460, and 1510–1600 cm−<sup>1</sup> revealed that the alkaline H2O<sup>2</sup> pretreatment was effective on lignin disintegration. Sun et al. [19] also stated that the delignification of agricultural crop stalks could occur during the alkaline H2O<sup>2</sup> pretreatment process, while the macromolecular structure of cellulose did not show any noticeable change. Results from this study confirm the findings of Sun et al. [19].

As seen in Figure 6, the raw greenhouse crop waste exhibited a smooth, non-porous, compact, and rigid surface structure. There was no separation of fibers, or ruptures and scars. On the other hand, the pretreated greenhouse crop waste demonstrated a rough and porous structure. In particular, the fibrils of greenhouse crop waste pretreated with alkaline H2O<sup>2</sup> under conditions of 100 ◦C, 3% VS, 24 h, and 3% H2O<sup>2</sup> were completely deformed, and their structural integrity was disrupted. The SEM examination revealed that the morphological changes, along with the tissue damage, resulted from the alkaline H2O<sup>2</sup> pretreatment. Similar to our findings, Rezende et al. [36] also stated that alkaline and NaCl pretreatment dissolved the inter-fibrillar or bulk lignin, while disrupting the initial fiber structure, leading to the disaggregation of micro-fibrils from their neighboring fibers.

**Figure 6.** Fourier-transform infrared (FTIR) spectra and SEM images of raw and pretreated greenhouse crop waste.


**Table 3.** Comparison of Fourier-transform infrared (FTIR) spectra of waste pretreated with alkaline H2O<sup>2</sup> under different conditions with with that of raw greenhouse crop waste.

+++++ to +: Max to Min.

#### **3. Materials and Methods**

#### *3.1. Experimental Rationale*

Antalya is the largest area for greenhouse cultivation in Turkey, providing tomato, pepper, cucumber, eggplant, and zucchini. Greenhouse crop waste, consisting of roots, stalks, leaves, and fruits from cultivation, is generated in the region, creating environmental problems. The greenhouse

−

crop waste was supplied by the growers, and fresh waste was sliced into approximately 1cm pieces, and was stored in sealed plastic bags at −20 ◦C until used for composition analyses, alkaline H2O<sup>2</sup> pretreatment experiments, and methane generation potential tests.

The first phase of the experiments involved the characterization of the greenhouse crop waste in terms of the parameters that would be used as major indicators for the extent of energy recovery achieved by means of alkaline H2O<sup>2</sup> treatment.

Analyses of the total solids (TS) and volatile solids (VS) were performed based on standard methods 2540C [40]. Analyses of the total chemical oxygen demand (COD) were done according to standard methods 5220B [40]. The Kjeldahl nitrogen was determined using a Kjeldahl nitrogen analyzer (Büchi Digest Automat K-438, Büchi Auto Kjeldahl Unit K-370 and Radiometer TitraLab 840, Büchi, Flawil, Switzerland). The contents of lignin, cellulose, hemicellulose, and soluble matter were determined according to the Van Soest procedure [41] using a Gerhard FBS6 (Gerhard, Königswinter, Germany). Analyses of the total free lignin of extractives (acid-insoluble and acid-soluble) were performed according to the "Determination of Structural Carbohydrates and and Lignin in Biomass, NREL/TP-510-42618" [42]. The protein concentration was determined using the Lowry method [43]. The extractive matter and lipid contents of samples were determined using Soxhlet extraction [44]. The soluble chemical oxygen demand (sCOD) was determined using a Hach-Lange DR5000 spectrophotometer (Hach Lange GmbH, Duesseldorf, Germany) and a Lange LT200 (Grasscht, Germany) with COD kits. The concentrations of soluble reducing sugar (sRedSugar) were determined via the Dinitrosalicylic acid (DNS) method [45]. The elemental composition of the greenhouse crop waste was identified using a CHNS elemental analyzer (LECO, CHNS-932, St. Joseph, MI, USA). All composition analyses were executed in triplicate, and the quotable outcomes are demonstrated as means.

### *3.2. Alkaline H2O<sup>2</sup> Pretreatment Experiments*

The greenhouse crop waste was pretreated in a Parr reactor (Parr Instrument Company) with a 200 mL working volume. The independent variables with a potential impact on alkaline H2O<sup>2</sup> pretreatment were selected as reaction temperature (50–100 ◦C), H2O<sup>2</sup> concentration (1–3%), reaction time (6–24 h), and initial solid content of greenhouse crop waste (3–7%). The pretreatment experiments were done in duplicate under each condition. The calculated amount of fresh greenhouse crop waste and H2O<sup>2</sup> solution (*w*/*w*) was loaded into the pretreatment reactor, and initial pH values were set to 11.5 using 6M NaOH solution, with the reactors heated to the appropriate reaction temperature. When the predetermined temperature was attained, the experiment time was started. After reaching the determined reaction time, the reactor was put into ice and a water bath to cool down and stop the reaction. The pretreatment process was evaluated according to sCOD, sRedSugar, total free lignin of extractives, and BMP as objective functions related to pretreatment yield. The samples were centrifuged at 15,000 rpm for 10 min for the sCOD and sRedSugar analyses. The amount of sCOD was determined using a Hach-Lange DR5000 spectrophotometer and a Lange LT200 (Grasshut, Germany) with COD kits. The sRedSugar concentrations were determined via the DNS method [45]. Analyses of the total free lignin of extractives (acid-insoluble and acid-soluble) were performed according to the "Determination of Structural Carbohydrates and and Lignin in Biomass, NREL/TP-510-42618" [42] using the solid phase of the pretreated samples. The remaining pretreated samples containing solid and liquid fractions were stored at −20 ◦C for the subsequent methane generation potential experiment.

#### *3.3. Methane Generation Potential Experiment*

The efficiency of alkaline H2O<sup>2</sup> pretreatment was determined using a biochemical methane potential test (BMP) based on methane production. The samples, including macro and micro nutrients, were incubated in a closed glass reactor with a specific quantity of seed sludge (inoculum). Mesophilic conditions (35 ◦C) were preferred for the BMP tests. The BMP protocol according to Carrère et al. and Us & Perendeci [5,46] was implemented. For the BMP tests, 500 mL glass reactors with a working

volume of 400 mL were filled with sample, seed sludge, nutrients, and a tampon solution. All BMP reactors were loaded with seed sludge from the anaerobic reactor of an Antalya city wastewater treatment plant. Fifty-six glass reactors were used in the study, and two of them were fed with only seed sludge and nutrients to specify the methane potential of seed sludge on its own. The 52 glass reactors were used with different pretreated samples, and two reactors containing raw greenhouse crop waste were used as controls. After the optimization of conditions for alkaline H2O<sup>2</sup> pretreatment, the BMP test was also conducted under optimal conditions with two duplicates for validation of the model. The food-to-microorganism ratio (F/M) was fixed at 0.5 (gVS waste/gVS inoculum) for the glass reactors. The initial pH was set to neutral for all reactors. To keep anaerobic conditions in the reactors, a gas mixture of N2/CO<sup>2</sup> (70/30%) was flushed. The BMP test lasted for 62 days. The produced biogas was measured based on a gas-water displacement method. The biogas composition was ascertained using gas chromatography (GC; Varian 4900). A standard gas consisting of 60% (*v*/*v*) CH<sup>4</sup> and 40% CO<sup>2</sup> was used for the calibration of gas chromatography. The gas production of seed sludge was counted in the computation of biogas production of the samples. The methane production was estimated as mL of methane per g of VS (mLCH4/gVS) added to the reactor.

#### *3.4. Optimization of the Alkaline H2O<sup>2</sup> Pretreatment Process*

The pretreatment process was optimized using a CCD of RSM. Three levels of four independent variables were applied for the CCD, using the Design-Expert® software (Minneapolis, MN, USA). The ranges of each independent variable were established based on information in the literature and on our previous experimental experience. The levels of the independent variables were coded as −1 and +1. The four independent variables were changed within the following ranges: 50–100 ◦C (reaction temperature), 6–24 h (reaction time), 1–3% (H2O<sup>2</sup> concentration), and 3–7% (initial solid content). A total of 52 runs, including four runs at the design center and duplicates of each run, were determined using a CCD.

The performance of the alkaline H2O<sup>2</sup> pretreatment process was evaluated based on sCOD, sRedSugar, total free lignin of extractives, and the BMP test as dependent variables. The outcomes from the pretreatment experiments were modeled using the Design-Expert® software (Minneapolis, MN, USA). Analyses of the regression coefficients, variance (ANOVA), and the *p*- and *F*-values were preferred for the model assessment. The adequacy of the model fit was presented by the coefficient of determination (*R* 2 ) and the adjusted determination coefficient (Adj-*R* 2 ).

The alkaline H2O<sup>2</sup> pretreatment process was also optimized using the optimization module of the Design-Expert® software (Minneapolis, MN, USA). The optimization of the alkaline H2O<sup>2</sup> pretreatment process was executed using the models developed for sCOD, sRedSugar, total free lignin of extractives, and BMP. The goal settings were carried out using the plus (+) symbols in the Design-Expert® program (Minneapolis, MN, USA).

#### *3.5. Fourier-Transform Infrared (FTIR) Spectroscopy and Scanning Electron Microscopy (SEM)*

Changes in the molecular-bond characterization of greenhouse crop waste were evaluated using an ATR-FTIR-Varian 1000 model FTIR spectrometer. The measurements were analyzed by averaging the signal of 16 scans across the range of 500 cm−<sup>1</sup> to 4000 cm−<sup>1</sup> with a spectral resolution of 4 cm−<sup>1</sup> . The evaluation of deformations on the surface of the greenhouse crop waste was also investigated, using a Zeiss Leo 1430 scanning electron microscope at a voltage of 15 kV.

#### **4. Conclusions**

In the light of the experimental results and evaluations reported in the preceding sections, a number of concluding remarks could be drawn for this study.

The alkaline H2O<sup>2</sup> pretreatment partially destroyed the complex lignocellulosic structure of the greenhouse crop waste. The organic matter was initially broken down and then hydrolyzed into simple, soluble compounds. On this basis, the alkaline H2O<sup>2</sup> pretreatment induced a significant increase in

the range of 200–800% in COD leakage into the soluble phase, and boosted the methane generation potential from 174 mLCH4/gVS to a much higher bracket of 250–350 mLCH4/gVS. Similarly, the volatile matter utilization increased from 31% in the waste material before treatment to 50–70% after treatment, depending on the selected experimental conditions.

The alkaline H2O<sup>2</sup> pretreatment was optimized to determine the optimal conditions for the enhancement of methane generation assuming a cost-driven approach. The optimal alkaline H2O<sup>2</sup> pretreatment conditions were found to be a reaction temperature of 50 ◦C, 7% initial solid content, an H2O<sup>2</sup> concentration of 1%, and a reaction time of six h. Under these conditions, the BMP test yielded a production of 309 mLCH4/gVS. The enhancement of methane production was calculated as 77.6% compared to raw greenhouse crop waste.

The results obtained provide an optimistic perspective for the possibility of energy recovery from complex waste such as greenhouse crop waste. It is recommended that future studies be directed toward testing new pretreatment processes, as well as toward novel energy recovery technologies such as pyrolysis, instead of traditional anaerobic digestion.

**Author Contributions:** Conceptualization, N.A.P.; Methodology, S.G. and N.A.P.; Software, N.A.P. and S.G.; Formal Analysis, S.G. and N.A.P.; Investigation, N.A.P., S.G. and D.O. Writing-Original Draft Preparation, N.A.P. and D.O.; Writing-Review & Editing, N.A.P. and D.O.

**Funding:** This research was funded by Scientific Research Projects Unit of Akdeniz University (Grant Number FLY 2015-623).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

### **Acid Assisted Organosolv Delignification of Beechwood and Pulp Conversion towards High Concentrated Cellulosic Ethanol via High Gravity Enzymatic Hydrolysis and Fermentation**

**Konstantinos G. Kalogiannis 1,\* ID , Leonidas Matsakas 2,\* ID , James Aspden <sup>2</sup> , Angelos A. Lappas <sup>1</sup> , Ulrika Rova <sup>2</sup> ID and Paul Christakopoulos <sup>2</sup>**


Received: 19 June 2018; Accepted: 4 July 2018; Published: 5 July 2018

**Abstract:** Background: Future biorefineries will focus on converting low value waste streams to chemical products that are derived from petroleum or refined sugars. Feedstock pretreatment in a simple, cost effective, agnostic manner is a major challenge. Methods: In this work, beechwood sawdust was delignified via an organosolv process, assisted by homogeneous inorganic acid catalysis. Mixtures of water and several organic solvents were evaluated for their performance. Specifically, ethanol (EtOH), acetone (AC), and methyl- isobutyl- ketone (MIBK) were tested with or without the use of homogeneous acid catalysis employing sulfuric, phosphoric, and oxalic acids under relatively mild temperature of 175 ◦C for one hour. Results: Delignification degrees (DD) higher than 90% were achieved, where both AC and EtOH proved to be suitable solvents for this process. Both oxalic and especially phosphoric acid proved to be good alternative catalysts for replacing sulfuric acid. High gravity simultaneous saccharification and fermentation with an enzyme loading of 8.4 mg/gsolids at 20 wt.% initial solids content reached an ethanol yield of 8.0 *w*/*v*%. Conclusions: Efficient delignification combining common volatile solvents and mild acid catalysis allowed for the production of ethanol at high concentration in an efficient manner.

**Keywords:** beechwood; organosolv delignification; ethanol fermentation; enzymatic hydrolysis; high gravity

#### **1. Introduction**

Lignocellulosic feedstocks have attracted a lot of interest for the production of biofuels and other high added-value bio-based chemicals and materials. Production of biofuels from lignocellulosic biomass waste streams, such as agricultural or forestry residues, comprises the following steps: pretreatment, enzymatic saccharification, and microbial conversion of sugars to biofuels. Pretreatment is the first step towards overcoming the complexity and recalcitrance of lignocellulosic biomass, aiming to make cellulose susceptible to enzymatic hydrolysis [1]. The pretreatment process, aiming at removing lignin, is considered to be the costliest and most challenging part of the lignocellulose conversion scheme. Lignin, which is a polyphenolic polymer surrounds the cellulose and hemicellulose, and it is essentially responsible for making biomass highly recalcitrant to pathogens, microorganisms, and enzymes [2]. Hence, a pretreatment step is required in order to disrupt the carbohydrate–lignin complex and to allow for the hydrolytic enzymes to gain access to the carbohydrates [3–5]. Hydrothermal pretreatment, without the use of chemicals, efficiently degrades hemicelluloses and increases the biomass porosity, which, in turn, enhances enzymatic hydrolysis of the pretreated solids [6]. However, the lignin that cannot be removed via hydrothermal pretreatment is partly rearranged on the surface of the lignocellulosic biomass exhibiting an inhibitory effect on downstream enzymatic hydrolysis [7].

Organosolv pretreatment has attracted an increased research interest, as it offers an effective method to remove lignin with the use of organic solvents. Organosolv employs aqueous-organic solvent mixtures, with high solvent concentration (30–70%) at temperatures of 100–220 ◦C, with or without the addition of catalysts [8]. One of the main benefits of organosolv pretreatment is the isolation of high-quality lignin and high-purity lignin-free cellulose [9,10]. The lignin recovered is sulfur free, while the organic solvents used (ethanol, acetone, formic, and acetic acid, etc.) can easily be recovered which is a significant advantage for small scale biorefinery plants [11]. The addition of an organic solvent allows for better mass transfer and the dissolution of lignin [12], reducing its recondensation on the external surface area of the pulp [13]. In addition, organosolv pulps have bleachability and viscosity retention when compared to cellulose soda and kraft pulps [14].

For these reasons, there is significant research interest in investigating the best pretreatment method for lignocellulosic materials. Sequential hot water pretreatment for hemicelluloses depolymerization and organosolv delignification for the removal of lignin and the production of high purity pulps have been published [10,15,16]. These studies investigated the effect of the different pretreatment techniques on the physical and chemical properties of the pulps, together with the saccharification effect of the residual solid. The existence of a two-stage sequential pretreatment method has a negative impact in the economic feasibility of the process when compared with the one-stage pretreatment methods.

Typically, both hydrothermal and organosolv pretreatments are catalytically assisted with mineral acids and bases, such as NaOH, H2SO4, etc. Despite their wide use, there are some limitations; they are not environmentally friendly, they generate large quantities of acid wastes and require high energy inputs, thus increasing overall process cost [17–19]. For the above reasons, an effort is being made to replace or exclude highly corrosive mineral acids such as H2SO4. Use of milder acids, such as H3PO<sup>4</sup> or even O<sup>2</sup> combining organosolv and oxidation processes, are considered as interesting alternatives [20].

To make the production of ethanol economically viable and at the same time reduce the environmental impact of the process, the use of high solid concentration (high gravity—HG) during saccharification and fermentation can serve as a solution. The use of high solids concentration during saccharification can result in high glucose concentration in the broth and in turn in high ethanol production. It has been already argued that an ethanol content of at least 4% *w*/*w* is required for an economically feasible ethanol distillation [21]. Moreover, HG processes are advantageous from a water economy point of view [22]. Despite the obvious advantages of HG processes, they also present several challenges during their implementation. The high solids content create a very viscous material, practically without any free water, which is hard to mix and pump, leading to insufficient mass and heat transfer [23]. Various alternatives have been proposed to overcome these issues and achieve efficient saccharification of lignocellulosic biomass under HG conditions, such as fed-batch hydrolysis [24]. Towards this direction, Luleå University of Technology (LTU) group has previously developed and implemented a free-fall mixing reactor that was successfully used for the saccharification of various lignocellulosic materials, such as sweet sorghum bagasse [25], food waste [26], corn stover [27], wheat straw [28], and beech wood [20] at high solids content prior to ethanol fermentation. Other groups have also developed high gravity processes, successfully fermenting steam pretreated spruce to ethanol [29] or beechwood to biobutanol and dicarboxylic acids in a Terrafors reactor [30].

In this work, different organic solvents were tested for the pretreatment of beechwood sawdust in an effort to efficiently delignify the biomass. The pretreatment conditions were optimized by studying the effect of the organic solvent, concentration, and type of acidic catalyst. The aim was to maximize lignin removal, while achieving high cellulose purity and recovery in the resulting pulps. The pulps were tested for their potential in enzymatic release of glucose. The materials demonstrating the highest saccharification yields were used in HG saccharification and fermentation at a solid content of 20 wt.%. Saccharification was done in a HG custom made reactor and it resulted in the production of an aqueous solution containing up to 8.0 wt.% ethanol in the subsequent fermentation. In addition, the removed lignin was easily recovered via solvent distillation and precipitation, and found to be potentially of high quality, being suitable for further conversion towards added value products.

#### **2. Results and Discussion**

#### *2.1. Effect of the Type of Organic Solvent*

Table 1 presents the experimental conditions of all runs conducted, while Table 2 presents the lignin, cellulose, and hemicellulose content of the pretreated pulp, together with the recoveries of each individual component into the pretreated pulps. It should be noted that, in some cases, the recoveries of the constituents are calculated at above 100%, due to the experimental errors of the analytical methods.


**Table 1.** Experimental conditions for the organosolv pretreatment.

\* Reaction temperature: 175 ◦C, reaction time: 60 min, liquid to solid (LSR) ratio: 10.

**Table 2.** Biomass constituents pulp content and % retrieved in the solid pulp.


\* Untreated Lignocel extractives are 9.1%.

Varying degrees of delignification were achieved depending on the solvent used and the presence or absence of acids that act as catalysts. Figure 1 presents graphically the pulp compositions and biomass constituents' recoveries when employing different organic solvents with and without homogeneous acidic catalysis.

**Figure 1.** Pulp compositions (**A**) and biomass constituents' recoveries (**B**) in solid pulps in batch autoclave runs at 175 ◦C, 1 h reaction time, LSR = 10, effect of organic solvents without and with use of 1 wt.% H2SO<sup>4</sup> , data labels have been rounded for clarity of presentation.

– Organosolv pretreatment was found to be very efficient in the pretreatment of beech wood biomass, as high cellulose low lignin content was achieved in all the treatment conditions. In some cases, the cellulose content exceeded 80 wt.% (runs 2, 7, 8, 9, and 11). Accordingly, the lignin content was very low, ranging from 4.2 to 10.7 wt.%. The pairs of runs 1–2, 3–4, and 6–7 employed different organic solvents, specifically, ethanol (EtOH), methyl- isobutyl- ketone (MIBK), and acetone (AC), without and with the use of 1 wt.% H2SO<sup>4</sup> as catalyst. EtOH and AC are water miscible solvents, typically used in organosolv processes. MIBK is water immiscible, forming a biphasic system with water, which on one hand, can impact the fractionation efficiency of the system, but on the other hand, can significantly simplify the separation process of the organic lignin rich fraction from the aqueous carbohydrates rich fraction. MIBK has been used in biphasic systems for production of chemicals from biomass [31], as a co-solvent during fractionation of organosolv lignin in single phase systems [32,33] and as an extracting agent for the isolation of lignin from liquors rich in lignin and hemicellulose [34].

Regardless of the solvent used, the use of H2SO<sup>4</sup> resulted in lower hemicellulose and lignin content and higher cellulose content in all of the pulps. Clearly, the hydrolyzing effect of the catalyst allowed for easier and more effective hemicellulose hydrolysis and removal. This, in turn, made the removal of lignin easier since it is closely connected to hemicellulose through a variety of bonds, such as ether and hydrogen bonds [35,36]. Among the three solvents used, both AC and EtOH proved to be effective in delignifying the biomass. AC was slightly more effective probably due to its higher solvent strength. On the other hand, MIBK was not as effective in delignifying the biomass. Compared to AC and EtOH, MIBK has a Hildebrand solubility parameter of 8.4, which is lower than typical lignin solvents, which is in the 10.5–12.5 range [32]. In addition, MIBK is not soluble in water, making this a two liquid phase reaction system. MIBK's insolubility in water is responsible for its poor performance, however it is this property that makes it very interesting for use as delignifying agent. Since the organic phase, which contains the dissolved lignin, and the aqueous phase, which contains the hemicellulose hydrolysate, can be very easily separated by spontaneous phase separation, this simplifies the separation process, and in turn, reduces the energy demands for lignin recovery. Hence, the 50% delignification degree (DD) achieved, although low, is satisfactory enough to justify further investigation in future work. Teng et al. [37] used the H2O/MIBK biphasic system successfully to delignify different biomasses such as corn cob and rice straw. They found that the use of acidic ionic liquids (IL) was significantly more efficient when compared to the use of mineral acids. Use of H2SO<sup>4</sup> achieved a DD of 61.5%, while the use of the IL [C4H8SO3Hmim]HSO<sup>4</sup> resulted in a DD of 76.3%. Pretreatment without the use of any catalyst resulted in poor delignification with a DD of 24% for corncob. They attributed the lower efficiency of the mineral acids to their miscibility in MIBK, which resulted in a reduction of their actual concentration in the aqueous solution, lowering their catalytic efficiency. In our work, use of mineral acids in the case of MIBK increased the DD from 37.2% to 49.5%, which is in accordance to the findings by Teng et al. [37]. Another interesting note is that in runs 3 and 4 where MIBK is used, there is a significant reduction of the hemicellulose that is retrieved in the solid pulp without H2SO<sup>4</sup> (19 wt.%) and with H2SO<sup>4</sup> (3.4 wt.%) when compared to EtOH (85.8 and 20.5 wt.%, respectively) and AC (70 and 13.6 wt.%, respectively). The immiscibility of the MIBK with water resulted in the stronger solvent power and hydrolysis effect of the water towards the biomass hemicellulose. In contrast, EtOH and AC that are water miscible act as antisolvents, in part reducing the hydrolysis achieved by H2O. This is also validated by the cellulose recovery in the pulps, which in the case of runs 3 and 4 with MIBK drops to 92.1 and 83.8 wt.% without and with H2SO<sup>4</sup> respectively. Cellulose, which is much more recalcitrant compared to hemicellulose [31], is not that affected, but part of it is solubilised in the aqueous fraction, especially when H2SO<sup>4</sup> is employed. Apparently, the use of one phase systems with EtOH and AC results in even lower solubilisation of cellulose, hence most of it is recovered in the solid pulp.

#### *2.2. Effect of Catalyst Type and Concentration*

Runs 8–11 along with runs 1 and 2 aimed at understanding the effect that homogeneous catalysis can have on the removal of lignin and the depolymerization and hydrolysis of hemicellulose in the bid to produce a high cellulose pulp. For this purpose, three different types of acids were investigated and their effect on the composition of the pulps is graphically presented in Figure 2. H2SO<sup>4</sup> was tested as a base case scenario, since it is the most used acid for biomass pretreatment [38]. H3PO<sup>4</sup> was tested as an inorganic acid alternative. Its main advantages are the fact that it is much less corrosive, easier to recycle, and can yield more amorphous cellulose pulp [39]. Oxalic acid was tested as an organic acid alternative. Dicarboxylic acids exhibit some advantageous characteristics, such as controlled stepwise acidity, biodegradability, and diminished corrosivity. In addition, they can be produced from bio-based and renewable resources, making them particularly attractive catalysts for biomass conversion [40].

**Figure 2.** Pulp compositions (**A**) and biomass constituents recoveries (**B**) in solid pulps in batch autoclave runs at 175 ◦C, 1 h reaction time, LSR = 10, effect of homogeneous catalysis, data labels have been rounded for clarity of presentation.

Comparing run 1 and 2, the use of H2SO<sup>4</sup> at 1 wt.% on dry biomass basis as catalyst has a pronounced effect, increasing the removal of hemicellulose and lignin from 14 and 46 to 80 and 89%, respectively. As expected, it enhanced hemicellulose hydrolysis, which also facilitated the removal of lignin, since these two components are connected via ether bonds, removing one can significantly boost the removal efficiency of the other. Both phosphoric and oxalic acids were also tested as catalysts. Run 8 and 10 employed 1 wt.% of each acid on a biomass basis, while run 9 and 11 used 5.6 and

2.6 wt.% of phosphoric and oxalic, respectively. This was done in order to reach the same pH as in the case of 1 wt.% H2SO4, so as to test the three different catalysts at the same severity. Phosphoric acid proved to be quite efficient in enhancing hemicellulose hydrolysis and lignin removal, at the 5.6 wt.% addition it was marginally better when compared to H2SO<sup>4</sup> for delignification. The addition of oxalic acid also increased the efficiency of delignification when compared to the treatment without acid catalysis. However increasing its concentration had no further effect. Stein et al. [41] achieved delignification using oxalic acid as catalyst in a water/2-methyltetrahydrofuran (2-MTHF) biphasic system. Oxalic acid has been previously used to depolymerize the hemicellulosic part of biomass [41], leaving the cellulosic crystalline part intact even at temperatures as high as 180 ◦C [42]. The above is in accordance with our work. Cellulose recovery in the solid pulp was 100% when oxalic acid was added, however hemicellulose recovery in solid form dropped from ~86% of initial hemicellulose when no oxalic acid was used to ~42% with oxalic acid catalysis. Lignin was also successfully removed, its recovery in the solid pulp dropped from 53.5% to ~24% (run 1, 10, 11, in Table 1).

#### *2.3. Pulp and Lignin Quality*

Apart from the composition of the resulting pulps, their crystallinity index (CrI) was determined as an attempt to evaluate the effect of the pretreatment on the pretreated solids and their potentials for enzymatic saccharification. Table 3 presents the CrI of all the produced pulps.


**Table 3.** Crystallinity index (CrI) of pretreated pulps.

\* Standard deviation for CrI was ± 1.3%.

As expected, there is an overall trend that resulted in the increase of the CrI as the cellulose content in the pulp increased due to the inherent crystallinity of the cellulosic part of the biomass. Run No. 1, for example, had cellulose content of 60% corresponding to a CrI of 68.8%, while runs 7 and 9 with increased cellulose contents of 89 and 85% had CrI at around 78%. In addition, it is noted that it is the presence of hemicellulose rather than lignin in the pulp that lowers the CrI. Pulps with high hemicellulose content had lower CrI due to the hemicellulose amorphous regions. Figure 3 presents SEM images of the initial biomass and pulps retrieved from run 7 and 9, which employed H2SO<sup>4</sup> and H3PO4, respectively.

It appears that the removal of lignin and hemicellulose results in the partial change in the fiber morphology. Untreated beechwood (Figure 3A) has a relatively smooth surface, while AC-1%H2SO<sup>4</sup> and EtOH-5.6%H3PO<sup>4</sup> pulps have rougher surface. Especially in the case of EtOH-5.6%H3PO4, the pulp appeared to be partially defibrilated and individual cellulose fibers were exposed (Figure 3C). The surface area of the pulps was slightly increased when compared to the untreated beechwood. More specifically, untreated beechwood had surface area of 0.27 m2/g, while for pulps that are produced from run 7 and 9, this increased to 1.18 m2/g and 1.08 m2/g, respectively. This is a small increase in surface area but has been found to positively affect the enzymes' efficiency. Arantes et al. concluded that the topology/porosity of the pulp can limit protein penetration into the microfibril pores of

the pulp, and hence affect the enzyme efficiency [43]. This is in agreement with the findings of Thygesen et al. who showed that the enzymes first penetrated into the porous regions of the pulp, and subsequently hydrolysed the cellulosic parts towards mono and oligomeric sugars [44].

μ **Figure 3.** Scanning electron microscopy (SEM) images, bar scale of 100 µm (**A**) untreated beechwood, (**B**) AC-1%H2SO<sup>4</sup> , and (**C**) EtOH-5.6%H3PO<sup>4</sup> .

Lignins were retrieved from all runs and some selected samples were analysed via NREL to evaluate their purity. The lignins from run 7 and 9, which were found to be the most suitable for biomass delignification, were found to have very high lignin content at >94.5 wt.% and 92 wt.% purity, respectively. Lignin from run 7 had 0 wt.% cellulose content and only 0.8 wt.% hemicellulose content. For comparison, lignin from run 6 had lignin content of around 89 wt.% and hemicellulose content around 4.2 wt.%. The lack of an acid catalyst in the case of run 6 led to the sedimentation of some hemicellulose oligo- and poly- saccharides. The use of the severe H2SO<sup>4</sup> in the case of run 7 hydrolyzed hemicellulose to such an extent that none was retrieved in the solid fraction of lignin. Run 9, on the other hand, had 2 wt.% and 1.8 wt.% cellulose and hemicellulose content, respectively. The milder acidity of H3PO<sup>4</sup> was enough to solubilize a small part of cellulose and leave some hemicellulose intact, so as to receive it in the solid lignin. Overall, all of the lignins retrieved were very pure and well fractionated. Finally, the lignins from run 7 and 9 were also analysed via FTIR (Figure 4).

− − − From the spectra, it appears that the delignification treatment did not degrade the recovered lignin. The FTIR graphs have peaks at characteristic wavelengths below 1500 cm−<sup>1</sup> , corresponding to guaiacyl, syringyl, and some methyl- and methylene- side chains that are typically found at 1385, 1420, and 1463 cm−<sup>1</sup> [45]. Wavelengths at 1216, 1271, and 1328 cm−<sup>1</sup> , corresponding to stretching of C–C and C–O bonds in guaiacyl oligomers and condensed syringyl and guaiacyl rings typical of hardwood lignin are also detected [46], suggesting that the structures of the lignins remain intact. This is a very important finding, since this pure lignin product, which is easily recovered from the

solvent mixture, could be upgraded to high value chemicals towards the establishment of a holistic biorefinery approach.

**Figure 4.** Fourier Transform Infrared Spectroscopy (FTIR) spectra of lignins retrieved from beechwood delignification from run 7 (AC-1%H2SO<sup>4</sup> ) and 9 (EtOH-1%H3PO<sup>4</sup> ).

#### *2.4. Enzymatic Saccharification of Pretreated Pulps*

To evaluate the potential of the pretreated pulps as raw materials for ethanol production, their susceptibility to enzymatic saccharification was assessed under low solids content. Table 4 presents the cellulose conversion after 24 and 48 h of enzymatic saccharification. The numbers in parentheses in the 24 h column indicate how much of the total glucose production occurred in the first 24 h, which is an important parameter and is indicative of the conversion speed.


**Table 4.** Enzymatic hydrolysis to glucose at 24 and 48 h.

\* numbers in parentheses depict the percentage of the amount of cellulose hydrolyzed to glucose in 24 h to the amount hydrolyzed in 48 h.

**Figure 5.** Cellulose conversion to glucose via enzymatic hydrolysis at 24 and 48 h vs pulp cellulose content.

An overall trend is noted where the higher the cellulose content of the pulp, the higher the cellulose conversion percentage was achieved (Figure 5). This is attributed to the lower lignin content of the high cellulose content pulps. Lignin has been known to have significant impact on the enzymes used for cellulose hydrolysis, inhibiting the depolymerisation of cellulose and the production of monomeric sugars [2]. In addition, some interesting observations can be deduced from the combination of Table 4 and Figure 5. More specifically, run 2 and 7–11, where homogeneous acidic catalysis was employed, produced pulps that were enzymatically hydrolysed to glucose easier (higher conversion after 48 h), but also more rapidly (higher % of conversion in first 24 h). Run 2 and 7 have the highest conversion rates; ~95 and 100% of overall cellulose to glucose conversion occurs in the first 24 h, respectively. This can be attributed not only to the high DD achieved, but also to a partial depolymerization of the cellulose to lower molar mass cellulose that can be enzymatically hydrolysed more rapidly. Run 9, where 5.6 wt.% H3PO<sup>4</sup> was used, had the highest conversion of cellulose at 24 and 48 h, higher than that of run 7 at roughly the same lignin content. Work in the literature suggests that treating biomass with concentrated H3PO<sup>4</sup> results in the swelling of the fibres and the reduction of the cellulose crystallinity [47,48]. In our work, the CrI increased as a consequence of the increased cellulose content of highly delignified pulp. Pulps produced with the aid of H2SO<sup>4</sup> or H3PO<sup>4</sup> catalysis had no significant differences in the CrI at similar cellulose and lignin contents (runs 2, 7, 9). Sathitsuksanoh et al. treated biomasses with concentrated H3PO<sup>4</sup> and found that the CrI values varied greatly, depending on several parameters, such as measurement techniques, calculation approaches, and sample drying conditions. They concluded that the effects of CrI data obtained from dried samples on enzymatic hydrolysis should be interpreted with caution. On the other hand, they suggested that increase of the fibres surface area through lignin and hemicellulose removal and disruption of the hydrogen bonds found in crystalline cellulose could significantly increase the hydrolysis rates and efficiencies [49]. Hence, a possible explanation for the hydrolysability of pulps produced with H3PO<sup>4</sup> assisted catalysis is the disruption in part of hydrogen bonding, which is not necessarily depicted as a reduction in the CrI. Enzymatic hydrolysis proved to be dependent mostly on cellulose and lignin content and was irrelevant of the CrI. Lignin, which has been found to be a major inhibitor in cellulose saccharification should therefore be removed to achieve high glucose production [50,51].

#### *2.5. High Solids Hydrolysis and Fermentation*

Based on the results from the saccharification at low solid content, two different delignified pulps, specifically from run 7 and 9, which employed AC with 1 wt.% H2SO<sup>4</sup> and EtOH with 5.6 wt.% H3PO4, respectively, were selected for evaluation under high solids hydrolysis and fermentation towards ethanol. The pulps from run 7 and 9 were found to have the highest DD, lowest lignin content, and over 90 wt.% cellulose recovery in the solid pulp. They were thus deemed suitable for high solids simultaneous saccharification and fermentation (SSF). As noted in the Methods section, the liquefaction/saccharification duration was 8 h at an enzyme loading of 8.4 mg/gsolids. After 8 h of pre-liquefaction/saccharification, the concentration of glucose was 63.8 g/L and 74.7 g/L, corresponding to 32.1% and 39.5% cellulose saccharification for the H2SO<sup>4</sup> and H3PO<sup>4</sup> assisted runs, respectively. Efficient glucose production in the first 8 h meant that ethanol concentrations higher than 40 g/L could be reached; a required minimum for downstream low-cost distillation [17]. Figure 6 presents the evolution of ethanol concentration for a six-day period for both delignified pulps.

**Figure 6.** Ethanol concentration in high solids simultaneous saccharification and fermentation (SSF), an 8 h hydrolysis step preceded the SSF.

Both pulps reached the 40 g/L ethanol concentration threshold in less than 24 h of SSF. The AC–H2SO<sup>4</sup> delignified pulp produced slightly more EtOH the first 24 h reaching a concentration of ~46 g/L. Afterwards, the ethanol production gradually leveled off to a final concentration of 76.3 g/L after six days of SSF, which is equal to approximately 75% of the maximum theoretical ethanol yield that could be attained for the cellulose content of the pulp. The EtOH-H3PO<sup>4</sup> (pulp No. 9), on the other hand, had a slightly lower production rate in the first 24 h, however it retained its high production rate for up to 48 h, reaching an ethanol concentration of 68.7 g/L after the first 48 h of SSF. After six days of SSF, the ethanol concentration reached a maximum of 80 g/L, which is equal to approximately 83% of the maximum theoretical ethanol yield. Pulp No. 7 demonstrated a slightly higher productivity during the first 24 h of fermentation. This difference in the initial ethanol productivity, can be attributed to a minor inhibition of the fermentation process by the higher initial glucose concentration; behavior that has also been observed elsewhere [27]. Pulp No. 9, which used H3PO4, had a steadier fermentation rate for up to 48 h. Even though its cellulose content was slightly lower when compared to pulp No. 7, it achieved higher final ethanol concentration. As explained above, the H3PO<sup>4</sup> may have disrupted in part the hydrogen bonds allowing for more efficient cellulose hydrolysis and consequently fermentation towards ethanol. The HG results are in good agreement with the initial enzymatic hydrolysis evaluation runs, where pulp No. 7 was found to quickly reach its maximum conversion in the first 24 h, while pulp No. 9 gave higher overall conversion in the 48 h period. Table 5 summarizes some of the work that has been done in HG SSF of different types of lignocellulosic feeds for the production of ethanol. The ethanol concentration of 80 g/L, as reported in our work, is one of the highest achieved in the literature.


#### **Table 5.** Work found in the literature on high gravity (HG) SSF for ethanol production.

#### **3. Materials and Methods**

#### *3.1. Raw Materials*

Commercially available beechwood sawdust with particle size 150–500 µm (Lignocel® HBS 150–500) and moisture content 8 wt.% was used as biomass feedstock. It was handled, as described by Kalogiannis et al. [45].

#### *3.2. Strains and Enzymes*

The *Saccharomyces cerevisiae* strain Ethanol Red® was used as fermenting microorganism during the current work. This specific strain was developed by Fermentis (Marcq-en-Barœl, France) for industrial fuel ethanol production, and therefore it exhibits high ethanol tolerance, making it suitable for use in HG fermentation processes. The commercial enzyme solution Cellic® CTec2 from Novozymes A/S (Bagsværd, Denmark) was used for the saccharification trials under low solids content and at HG conditions. The protein content of the enzyme solution was 100 mg/mL, as determined by using the Bradford assay [64]. All the other chemicals and reagents were of analytical grade.

#### *3.3. Organosolv Pretreatment*

Organosolv pretreatment of Lignocel was performed in metallic cylinders of 2.5 L size, which were placed in an air-heated multidigester apparatus [27] at 175 ◦C for 60 min. During the pretreatment, 110 g of biomass were mixed with 1.1 L of solvent-aqueous mixture. The following solvents were tested: ethanol, acetone, and methyl-isobutyl-ketone at a content of 60% *v*/*v* (with the acetone to be also tested to an acetone content of 25% *v*/*v*) with or without the addition of sulfuric acid (1 wt.% on dry biomass) as acidic catalyst. Replacement of sulfuric acid with phosphoric acid and oxalic acid was also tested with ethanol as the solvent. In that case, the concentration of the acid catalysts was either similar to sulfuric acid (1 wt.% on dry biomass) or was fixed to achieve the same pH as the sulfuric acid during the pretreatment (phosphoric acid, 5.6 wt.% on dry biomass; oxalic acid, 2.6 wt.% on dry biomass). At the end of the pretreatment, the pretreated solids were separated from the pretreatment liquor by vacuum filtration, washed with the same solvent used during the pretreatment, air dried, and stored until further use. The weight of the pretreated solids was measured to determine biomass solubilization and the composition of the solids was determined, as described in the *Analytical Methods* section.

The pretreatment liquor was collected and the solvent was evaporated (when ethanol and acetone were used) under vacuum in order to reduce lignin solubility. Lignin was then separated from the liquid by centrifugation (14,000 rpm, 29,416× g, at 4 ◦C for 15 min), and finally air-dried [53]. When MIBK was used as solvent, a different lignin isolation process was followed. MIBK is water immiscible at room temperature, resulting in phase separation with the lignin being recovered in

the solvent phase. The solvent was then evaporated under vacuum, leading to the recovery of the solid lignin.

All of the experimental conditions are presented in Table 1. The resulting pulps were dried and weighed, while the original biomass and the resulting pulps were analysed by the NREL method to determine (see analytical methods paragraph) cellulose, hemicellulose, and lignin content. The delignification degree (DD) can be calculated as 100%-lignin recovery (%).

#### *3.4. Enzymatic Saccharification Trials*

The pretreated solids were assessed for their enzymatic saccharification yields under a solid content of 2 wt.% in 50 mM citrate buffer (pH 5). The enzyme load was 8.4 mg/gsolids of the commercial enzyme solution Cellic® CTec 2. Sodium azide at a concentration of 0.02 wt.% was added in the solution to prevent microbial contaminations. Incubation took place in 2 mL Epperdorf tubes containing 1.5 mL of the solution in ThermoMixer C (Eppendorf, Hamburg, Germany) at 50 ◦C and 1200 rpm for 48 h. Samples were withdrawn at 0 h, 24 h, and 48 h, and analyzed for glucose concentration. All of the enzymatic hydrolysis trials were performed in duplicates. The enzymatic saccharification yield was expressed as the percentage of cellulose converted to glucose and was calculated according to the following equation:

$$Sacchariification\,yield = \left(\frac{\mathbb{C}\_{glucose} \times V\_{liquid} \times 0.90}{\mathfrak{m}\_{solids} \times \mathfrak{x}\_{cellulose}}\right) \times 100\%$$

where *Cglucose* is the concentration of glucose, *Vliquid* is the volume of the liquid during the trials, 0.90 is the correction factor for the conversion of glucose to cellulose, *msolids* is the mass of the dry solids, and *xcellulose* is the cellulose content of solids.

#### *3.5. High Gravity Saccharification and Fermentation*

The two most promising materials were further subjected to high gravity saccharification and fermentation trials. Saccharification took place at a freefall mixing saccharification reactor, as previously described [25]. More specifically, the dry material content used was 20 wt.% in 50 mM citrate buffer with an enzyme load of 8.4 mg/gsolids. Saccharification took place at 50 ◦C for 8 h. At the end of the saccharification the slurry was collected and supplemented with nutrients for the yeast growth at a final concentration of 1 g/L yeast extract, 0.5 g/L (NH4)2HPO4, and 0.025 g/L MgSO4·7H2O. The fermentation was initiated by inoculation with *S. cerevisiae* suspension (that was grown overnight at YPD media at 35 ◦C and 180 rpm) to achieve an initial dry cell weight concentration of 1 g/L. Incubation was carried out at 35 ◦C and 120 rpm, and the samples were withdrawn daily, diluted, filtered through a 0.2 µm syringe filter, and analyzed for ethanol. The fermentations were performed in duplicates.

#### *3.6. Analytical Methods*

The cellulose, hemicellulose, lignin and ash contents of lignocellulosic biomass were determined, according to the procedures provided by National Renewable Energy Laboratory (NREL; Golden, CO, USA) [65]. The sugars were analysed at a high pressure liquid chromatography (HPLC) apparatus, coupled with a refractive index detector equipped an Aminex HPX-87P column (Bio-Rad, Hercules, CA, USA). Analysis performed at 85 ◦C, with ultrapure water as mobile phase at a flow rate of 0.6 mL/min. Ethanol produced during the SSF was analysed by the same HPLC apparatus equipped with an Aminex HPX-87H (Bio-Rad, Hercules, CA, USA) chromatography column. The column was kept at 40 ◦C and the mobile phase was 5 mM sulphuric acid in degassed HPLC grade water at a flow rate of 0.6 mL/min.

Fourier Transform Infrared Spectroscopy (FTIR) (Nicolet 5700, Thermo Electron Corporation, Waltham, MA, USA) analysis was employed for further characterization of the lignin samples' structure. Details may be found elsewhere [46]. X-ray Diffraction analysis was done on a Siemens D500, copper ray with a Nickel filter (λ = 15,406 Å, voltage 40 KV, intensity 30 mA) (Bruker, Wien, Austria). The angle 2*θ* was between 5◦ and 50◦ with a step 0.04 and step time 2 s. Surface area of the pulps was measured on a Micromeritics Tristar 3000 (Micromeritics, Norcross, GA, USA) via the BET method after outgassing the biomass samples at 25 ◦C for 72 h. Scanning electron microscopy (SEM) images were obtained on a Jeol JSM-6300 microscope (Jeol, Peabody, MA, USA).

For the determination of the surface area (BET method), pore volume, and pore size distribution (BJH method) of the catalyst samples, N<sup>2</sup> adsorption/desorption measurements were carried out at −196 ◦C, using an Autosorb-1MP Automatic Volumetric Sorption Analyzer (Quantachrome, Boynton Beach, FL, USA).

#### **4. Conclusions**

In the present work, the efficiency of organosolv pretreatment on lignin and hemicellulose removal and its effect on the downstream biochemical conversion of the solid pulp to ethanol were evaluated. A hardwood feedstock, more specifically beechwood, was treated with mixtures of water and different organic solvents, namely AC, EtOH, and MIBK. In addition, the effect of homogeneous catalysis was investigated. Sulfuric, phosphoric, and oxalic acids were tested at different concentrations and their effect on the DD and the hydrolysability and fermentability of the resulting pulps was evaluated. Both AC and EtOH, which are water miscible, were found to be very efficient in removing lignin and hemicellulose from the initial feedstock. Both were able to remove almost 50% of the lignin found in the feedstock. MIBK, on the other hand, behaved poorly due to its non-miscibility in water. Use of sulfuric acid as catalyst significantly improved the DD; more than 90% of initial lignin was removed and pulps with high cellulose content (>85%) were produced. Phosphoric and oxalic acid were used as alternative catalysts and were both found to enhance lignin removal. In the case of phosphoric acid, partial defibrillation and exposure of the cellulose fibrils was also noted. Moreover, the lignin retrieved from the solvent system was found to be intact and of high purity and quality making it a valuable potential feedstock for production of bio-based chemicals and materials. High gravity SSF at 20 wt.% solids yielded highly concentrated ethanol solutions (8 wt.%), which is one of the highest reported in the literature for beechwood feedstock and stresses the potential of combining organosolv pretreatment with high solids fermentation on the basis of a biorefinery approach.

**Author Contributions:** Conceptualization, K.G.K. and L.M.; Methodology, K.K. and L.M.; Formal Analysis, X.X.; Investigation, K.G.K., L.M. and J.A.; Resources, P.C. and U.R.; Data Curation, K.G.K., L.M., P.C. and A.A.L.; Writing-Original Draft Preparation, K.G.K. and L.M.; Writing-Review & Editing, P.C., U.R. and A.A.L.; Supervision, P.C., U.R. and A.A.L.; Project Administration, K.G.K., P.C. and U.R.; Funding Acquisition, K.G.K., L.M., P.C., U.R. and A.A.L.

**Funding:** The authors wish also to acknowledge the support from COST Action FP1306 via the Short Term Scientific Mission of K. Kalogiannis in LTU.

**Acknowledgments:** The authors would like to thank Novozymes A/S, Denmark, for providing the Cellic® CTec2 enzyme solution and Lesaffre Advanced Fermentations, France, for providing the Ethanol Red® that were used during this work. L.M., U.R. and P.C. would like to thank Bio4Energy, a strategic research environment appointed by the Swedish government, for supporting this work.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the materials produced in the current work are available from the authors at a reasonable request.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Article* **Lime Pretreatment of Miscanthus: Impact on BMP and Batch Dry Co-Digestion with Cattle Manure**

### **Hélène Laurence Thomas, Jordan Seira, Renaud Escudié and Hélène Carrère \***

LBE, University of Montpellier, INRA, 102, Avenue des Etangs, 11100 Narbonne, France; helene.thomas@inra.fr (H.L.T.); jordanseira@hotmail.fr (J.S.); renaud.escudie@inra.fr (R.E.)

**\*** Correspondence: helene.carrere@inra.fr; Tel.: +33-468-425-168

Received: 18 May 2018; Accepted: 28 June 2018; Published: 2 July 2018

**Abstract:** In Europe, the agricultural biogas sector is currently undergoing fast developments, and cattle manure constitutes an important feedstock. Batch dry digester processes with leachate recirculation prove to be particularly interesting for small-scale plants. However, their startup being relatively slow, the process could be facilitated by co-digestion with energy crops. In this study, *Miscanthus x giganteus* was chosen for its high biomass yields and low input requirements. The carbohydrate accessibility of this lignocellulosic biomass is limited but may be improved with alkali pretreatment. The efficiency of lime (CaO) pretreatment with low water addition on the biochemical methane potential (BMP) of miscanthus was investigated through two experimental designs (CaO concentrations ranged between 2.5 and 17.5% and pretreatment lasted 1, 3, or 5 days). The pretreated miscanthus was then co-digested with cattle manure in dry leach bed reactors. CaO pretreatments led to a 14–37% improvement of miscanthus BMP, and a 67–227% increase in the first-order kinetics constant; a high contact time was shown to favor methane production. According to these results and to industrial requirements, miscanthus was pretreated with 5 and 10% CaO for 5 days, then co-digested with manure in dry leach bed reactors. Nevertheless, the promising results of the BMP tests were not validated. This could be related to the high water absorption capacity of miscanthus.

**Keywords:** anaerobic digestion; biogas; lignocellulosic biomass; alkali pretreatment

#### **1. Introduction**

Within the context of having to mitigate global warming and reduce greenhouse effect gas emissions, anaerobic digestion (AD), which allows the production of renewable energy from various organic wastes, is undergoing rapid developments. In particular, the French government has set the target of 1500 biogas plants by 2020, including 1000 plants based on agricultural feedstocks [1]. In many agricultural anaerobic digestion plants, manure represents the main part of the feed. Furthermore, cattle manure is available in high quantities all over the country. Its production has been estimated at about 69 MT per year in 2010 [2]. Cattle manure is rich in straw and is characterized by a total solids (TS) content of about 20–30%. It is thus suited for dry AD [3], also called solid-state anaerobic digestion. A process occupying an important part in the development of the agricultural AD sector is the leach bed reactor (LBR) operated in batch mode [4]. In this high-solids process, the solid substrate is loaded into the reactor while a liquid phase, usually stored in a separate container, is regularly sprinkled over the solid bulk and percolates through it. However, batch mono-digestion of cattle manure usually takes time to start up and produces low amounts of biogas [5].

In this view, co-digestion would be a good option for improving biogas production and productivity. For example, Botji et al. (2017) [6] demonstrated how the co-digestion of poultry manure with maize silage improved methane production by 24% relative to mono-digestion. This is

presumably due to the improved C/N ratio. Nevertheless, the use of food or feed product-dedicated crops (e.g., cereals) or energy crops as AD plant feedstocks is limited to less than 15% of the total feed ration by the French national legislation [7]. Some exceptions are, however, still possible for catch crops and biomass cultivated on marginal lands that are not in conflict with food and feed production. Among these, miscanthus presents many advantages, including high biomass yields, low input requirements (i.e., water, fertilizers), prolonged soil cover, reduced soil disturbance, and increased soil carbon content [8,9]. This crop can also grow on polluted soils [10,11]. Few studies have used *Miscanthus x giganteus* as a co-substrate for manure AD. Moiceanu et al. (2016) [12] used miscanthus as a reference co-substrate to investigate the influence of different types of manure on biogas production.

Nonetheless, for most lignocellulosic biomasses, carbohydrate accessibility is limited and AD performance can be improved by pretreatment [13]. For example, a 170 ◦C hydrothermal pretreatment of miscanthus led to a 21% increase in biogas production [14]. In another study, Nges et al. (2016) [15] applied grinding, steam explosion, and acid and alkali pretreatments to *Miscanthus lutarioriparius*. The best result—i.e., 57% increase in methane production—was obtained with a mild alkaline pretreatment. Indeed, a high lignin content and lignin/polysaccharides links have been identified as main bottlenecks for lignocellulosic biomass AD [16]. Among the different kinds of pretreatment techniques (mechanical, biological, chemical) [17], alkali pretreatments have been recognized as the most efficient for degrading lignin [18,19]. Alkali pretreatments generally employ soda. Because digestates from agricultural AD plants are systematically used as organic fertilizers and returned to agricultural soils, sodium spreading into soils should be avoided. The aim of this study is to therefore investigate miscanthus alkali pretreatment with lime.

The first objective was to assess and optimize miscanthus pretreatment conditions compatible with a further application in dry AD (i.e., with low water content). In fact, high solid content pretreatments reduce waste generation, do not require a separation step before further processing, and reduce the environmental impact of the entire process [20]. In order to keep pretreatment costs as low as possible, the conditions were set to ambient temperature. The impact of lime concentration and pretreatment duration on the biochemical methane potential (BMP) of miscanthus was investigated using a response surface methodology. The second objective of this study is to evaluate the impact of a selected lime pretreatment of miscanthus on its batch co-digestion using cattle manure in an LBR. Startup performances, as well as methane production, are reported.

#### **2. Results**

#### *2.1. Impact of Lime Pretreatment on BMP*

The different pretreatment conditions were carried out at a TS content of 13%, CaO concentrations between 2.5 and 17.5%, and pretreatment durations of 1, 3, or 5 days. Two experimental designs were created consecutively: Design 1 (CaO concentrations between 7.5 and 17.5%, and durations of 1, 3, or 5 days) and Design 2 (CaO concentrations between 2.5 and 12.5%, durations of 1, 3, or 5 days). BMP tests were performed in duplicate using these pretreated substrates. Table 1 reports the BMP and first-order kinetics constant k values. The duplicates revealed a very good repeatability. In comparison with the BMP obtained for the non-pretreated biomass (158 ± 2 NmLCH4·gVS −1 ), the effect of the pretreatment was significant (*p*-value = 9.8 × 10−<sup>4</sup> ) and positive. An improvement in BMP was observed, ranging from +14% (for 15% CaO; 1 day) to +37% (for 5% CaO; 5 days) for the best-performing condition.

The adjustment of Equation (2) for estimating k was excellent over all experimental conditions (R<sup>2</sup> > 0.97; data not shown). With a focus on the first kinetics constant k, a strong and positive effect was also noticed as an improvement, ranging from +63% (for 10% CaO; 1 day) to +221% (for 17.5% CaO; 3 days). This was calculated by comparing k with that obtained for raw biomass (0.024 ± 0.002 NmLCH4·gVS −1 ·d −1 ). Even though a clear correlation did not emerge, the evolution of BMP and k presented similar trends, as the highest BMPs were mostly characterized by the highest

k values. These results suggest that lime pretreatment of the miscanthus does, indeed, induce an increase for both BMP and k. Despite these positive observations, it was not possible to assess which parameter was most relevant. The effect of each parameter therefore needs to be unraveled using an experimental design.


**Table 1.** Pretreatment conditions, biochemical methane potential (BMP), and first-order kinetics constant values and their improvement, compared to raw, for Design 1 and Design 2.

\* by comparison with raw (i.e., non-pretreated) sample.

#### *2.2. Mathematical Models to Describe Impact of Concentration and Pretreatment Duration on BMP Values*

2.2.1. Experimental Design 1 (CaO Concentration from 7.5 to 17.5% and Duration from 1 to 5 Days)

The effect of variables A (CaO concentration) and B (pretreatment duration) on BMP and first-order kinetics constant k was investigated by statistical analysis based on response surfaces. According to Oliveira et al. (2015) [21], response surface methodology (RSM) is a collection of both mathematical and statistical techniques that involves (i) designing and carrying out experiments with a reduced investment; (ii) building models; (iii) evaluating the relative significance of the studied variables; and (iv) assessing the optimal conditions for a favorable response. Using data displayed in Table 1 and multiple regression analysis, a polynomial equation was determined to predict BMP and k depending on the variables, as well as their interactions (Equation (1)). The different coefficients with their standard deviation, the Fisher value (F-value), and the coefficient of determination R<sup>2</sup> of the models are provided in Table 2 for each design.



−

The F-value for each case was far less than the Fisher parameter calculated at the 95% confidence level (161.45), thus indicating that the models were significant and fitted the data nicely. Joglekar and May [22] suggested that, for a good fit of a model, the R 2 should be higher than 0.8. The high R 2 (≥0.95) obtained for each case was a strong hint of suitability, as it indicated that 95% of the data were explained by the regressions, even reaching 99.9% for the prediction of k in Design 1. Consequently, all the models were validated. he high R² (≥

− −

−

Standardized Pareto charts displaying the effects of the different terms of the models are provided for BMP in Figure 1a and for the first-order kinetics constant k in Figure 1b. The duration of the pretreatment (variable B) was the only parameter that significantly affected the BMP (Figure 1a). Moreover, the effect being positive, long contact times between CaO and miscanthus should favor methane production. Interestingly, the pretreatment duration also produced a very significant and positive effect on the first-order kinetics constant k, although this effect was less significant than the effect of CaO concentration (Figure 1b). In addition, both quadratic terms proved to be significant for the first-order kinetics constant k, thus implying that the influence of the variables was not necessarily linear. Finally, the interaction term A × B was also significant, with a positive effect on the first-order kinetics constant. This result could not have been anticipated by using a univariate approach. Although both BMP and k were related, it is noteworthy that both selected variables affected these responses at different levels (positive or negative effect) and with various extents of contribution.

**Figure 1.** Pareto diagram showing the effect of different coefficient terms on BMP (**a**) and kinetics constant k (**b**) for Design 1. Red bars indicate a negative impact, and green bars show a positive impact. Bars exceeding the vertical line point to the significance of the coefficient terms (*p* < 0.05, corresponding to 4.3 according to Student t-test in our conditions).

Response surfaces were plotted in 3-D for each parameter (i.e., BMP and k) as a function of CaO concentration and pretreatment duration (Figure 2). The response surface plot for BMP (Figure 2a) led to the following observations: (i) the longer the duration of the pretreatment, the better the BMP, which is the same conclusion as that stated in a previous section; (ii) the more the CaO concentration increases within the experimental domain, the more the BMP decreases. Even if it was not significant, the negative effect of this variable could be anticipated from Figure 1. Consequently, in order to favor methane production, a combination of "low" CaO concentration (lower part of the experimental domain) with "high" contact time for pretreatment (upper part of the experimental domain) could be a viable option.

methane production, a combination of "low" CaO concentra

domain) with "high" contact time for pretreatment (upper part of the experimental domain) could

Focusing on the response surface plot for k (Figure 2b), the coupling of a "high" CaO

th a "long" pretreatment duration was linked to a high methane production rate.

**Figure 2.** Response surface plots showing the impact of lime concentration and pretreatment duration on BMP (**a**) and kinetics values (**b**) for Design 1.

Focusing on the response surface plot for k (Figure 2b), the coupling of a "high" CaO concentration with a "long" pretreatment duration was linked to a high methane production rate. Nevertheless, the effect was more pronounced in the upper part of the domain (CaO concentration > 16% and pretreatment duration > 3 days) where few experimental data were generated (Table 1). The prediction thus rather relies on extrapolation rather than on interpolation. Therefore, the selection of the experimental domain to be exploited was made where the trend described by the model is well established. In this case, it is the middle-upper part of the domain that was chosen. Application of a "medium" CaO concentration with a "long" contact time pretreatment was hence considered most relevant to achieve a high first-order kinetics constant k.

As the responses for BMP and k reflect different impacts, it is difficult to find a consensus regarding the values of the variables to select (lime concentration and pretreatment duration) in order to optimize both parameters simultaneously. Moreover, no optima could be determined within the investigated domain. Only one extreme stationary point was identified in the lower domain for the kinetics constant k, which is irrelevant when both BMP and k need to be maximized. Owing to its energetic relevance, BMP is the parameter to prioritize in this study. The first-order kinetics constant k will therefore not be discussed anymore in the further section.

#### 2.2.2. Experimental Design 2 (CaO Concentration from 2.5 to 12.5% and Duration from 1 to 5 Days)

According to the trend displayed in Figure 2a, the application of a lower CaO concentration could contribute to enhancement of the BMP and even lead to an optimal result. The experimental domain was thus extended to CaO concentrations ranging from 2.5 to 12.5% for a second design of experiments (DOE). Finally, and even though the duration of pretreatment had a positive effect on BMP, its experimental domain was not extended for the following reasons: (i) longer contact times would not be realistic for applications in full-scale plants; (ii) the extension of a Doehlert design with the reuse of certain points is only possible for one factor; (iii) the possible formation of refractory compounds, which could further impede methane production [23]; and (iv) the possible pre-degradation of accessible substrates by microorganisms already present in the bulk matrix, which could decrease the bioavailable fraction for methane production [24].

The results obtained for the second DOE are summarized in Table 2, and as a Pareto chart and response surface plot in Figure 3a,b. As depicted in Figure 3a, variable B (pretreatment duration) was the only one that significantly and positively impacted methane production, as was previously observed. The response surface (Figure 3b) confirmed this observation, as BMP increased with the application of longer contact times. Regarding variable A (CaO concentration), the effect was minor

within the investigated domain (from 2.5 to 12.5%), even though the highest concentrations appear to lead to an increase in BMP. Unfortunately, it was not possible to determine an optimum in the second DOE domain. However, high responses for BMP can be observed in a region of interest. According to these results, the application of a long contact time seems necessary for enhancing methane production. A duration of 5 days for the "pretreatment duration" variable was thus selected. For the CaO concentrations, response surfaces revealed that the additional increase in BMP was negligible when a concentration above 10% was applied, while a detrimental effect was even possible for concentrations below 5% (Figure 3). methane production. A duration of 5 days for the " retreatment duration" variable was thus selected.

**Figure 3.** Pareto diagram showing the effect of different coefficient terms on BMP (**a**) and response surface plot showing the impact of lime concentration and pretreatment duration on BMP in Design 2 (**b**).

#### *2.3. Co-Digestion*

According to RSM conclusions and due to economic incentives, a concentration of 5% CaO was first selected. In order to confirm the trends displayed by RSM (e.g., increase in k with increasing CaO concentration), it was also worthwhile to consider a higher CaO concentration (10%). As a consequence, the following combinations were retained for application in an LBR at lab scale: 5% CaO during 5 days of pretreatment and 10% CaO during 5 days of pretreatment (performed in duplicate).

−1 −1 −1 The concentration in volatile fatty acids (VFAs) in the leachate and the evolution of pH were measured each day at the beginning and on a regular basis thereafter (Figure 4). For duplicates, the VFA concentrations were similar (Figure 4a). VFAs were rapidly produced, and their maximum value was reached after 1 day for miscanthus pretreated with 5 and 10% of lime for 5 days (4.5 ± 0.5 g·L <sup>−</sup><sup>1</sup> and 4 g·L −1 , respectively). For raw miscanthus, the maximum was reached after 2 days (4.2 ± 0.1 g·L −1 ). For all reactors, pH variations were similar (Figure 4b). The evolution of VFA and pH can be separated into two steps. During the first step, the accumulation of VFA during the first 3 days induces, with a brief delay, a slight decrease in pH down to 6.8. Two hypotheses could explain this observation: either (1) the positive impact of higher alkalinity due to CaO or (2) the buffer capacity of manure. The second hypothesis, manure buffer capacity, seems to be more plausible, given that in the case of co-digestion with the raw substrate, the pH is similar to that of the pretreated substrate. The pH did not present a sharp drop and the VFA concentration was not very high either. Moreover, since the VFA/alkalinity ratio remains below 1 (Table 3), the risk of acidification is avoided [25] and the drop in pH is swiftly reduced. During the second step, the pH stabilized close to 7.3 due to the decrease in VFA concentrations. After 15 days, there were no more VFAs accumulated; they were simultaneously produced and consumed at the same rate.

**Figure 4.** Volatile fatty acid (VFA) concentration (**a**) and pH variation (**b**) in leachate during the first 30 days.

− **Table 3.** VFA/alkalinity ratio at 3 days, methane production at 6, 10, 15, 24, 29, and 59 days, expected methane production calculated from BMP values, and first-order kinetics constants.


<sup>1</sup> HAceq means acetic acid equivalent, <sup>2</sup> from BMP values.

**/ − − −** Methane production from co-digestion of raw miscanthus and miscanthus pretreated with lime at 5 and 10% for 5 days with cattle manure was 158 ± 4, 150, and 167 ± 2 mL·gVS <sup>−</sup><sup>1</sup> after 59 days, respectively (Table 3). The 59-day period was selected because it corresponds to the duration of a batch in industrial plants. For this time span, there is no significant difference in the methane production between the control and the two different conditions of pretreatment (*p*-value = 0.7, 0.82) nor between both pretreatments (*p*-value = 0.82). However, after a shorter time, differences can be observed (Table 3). After 15 days of anaerobic co-digestion in an LBR, the methane production was higher for the miscanthus pretreated at 10% than for the raw miscanthus (+18%). It is related to a higher kinetics constant (+23%).

#### **3. Discussion**

#### *3.1. BMP and Pretreatment*

The BMP value of unpretreated miscanthus was 153 ± 7 NmLCH4·gVS −1 . This value lies within the lowest range of published miscanthus BMP values (170 mLCH4·gVS −1 [14] to 227 mLCH4·gVS −1 [26]). This low methane potential is most certainly linked to the high lignin content of the Floridulus clone [27]. Alkaline pretreatment may therefore be relevant for improving methane production from this clone.

Lime pretreatments have been far less studied than NaOH pretreatments, although sodium has detrimental effects on agricultural soils when digestates are used as biofertilizers. In addition, miscanthus has been scarcely employed in AD studies, which are extensively dedicated to agricultural residues such as rice straw, sugarcane bagasse, corn stover, wheat straw, or other energy crops, such as switchgrass [28]. To the best knowledge of the authors, there has not yet been any study on the impact of lime pretreatment on *Miscanthus x giganteus* BMP. Moreover, what makes this study innovative is that low inputs were set for the pretreatment conditions (low water input with high solid concentration and no heat energy input with ambient temperature conditions). In particular, for humidifying the entire biomass, a minimum amount of water was used. This corresponds to only 13% TS due to the high absorption capacity of miscanthus. Peces et al. (2015) [29] clearly demonstrated that total solids content is a significant parameter for the performance of sonication pretreatment, although it has been rarely considered in pretreatment optimization procedures. However, the doses of lime are consistent with those applied to other types of herbaceous biomass. For example, Jiang et al. (2017) [28] pretreated giant reed biomass with 1, 3, 5, 7, 12, and 20% (gCa(OH)2·ginitialTS −1 ) at 25 ◦C for 24 h. They observed an increase in methane yields between 7 and 34%. Another study obtained a 23% improvement of BMP with sunflower straw pretreated at 30 ◦C with 4% (gCa(OH)2·100gTS −1 ) for 1 day [18].

Although good performances have been achieved with lime, compared to potassium and sodium hydroxide at the same molarity, it has proven to be significantly less efficient for delignification [30]. Results indicate that with an equivalent molar basis of OH−, potassium and sodium hydroxide have a performance that is superior to calcium hydroxide [30].

The efficiency of pretreatments also depends upon the substrate. Indeed, a low lignin content is the main factor in promoting enhanced enzymatic saccharification [19] or enhancing anaerobic digestion [16]. Miscanthus is also widely used as animal bedding due to its high absorption capacity [31]. Thus, for an equivalent biomass TS content, less free water would be available with miscanthus than with other types of biomass. These, associated with a low lime solubility (1.65 g·L −1 at 20 ◦C, corresponding to 5.5% in this study), could reduce the amount of lime in contact with the substrate. In addition, Boix et al. (2016) [32] demonstrated that the absorption capacity of miscanthus increases with alkaline treatment. This can be explained by the removal of hydrophobic compounds, due to more exposed OH groups from cellulose or hemicellulose on the stem surface.

#### *3.2. Pretreatment and Co-Digestion*

Pretreatments are a promising solution in BMP test series. However, the performances in the LBR could not be confirmed if the methane production was estimated at 59 days. The BMP of manure used for the experiment was 202 ± 30 NmLCH4·gVS −1 , which was higher than the BMP of raw miscanthus and within a similar range to pretreated miscanthus BMP. The maximum expected methane production with a ratio of 40%VS miscanthus and 60%VS manure is presented in Table 3. While 87% of the expected methane production was reached after 59 days co-digestion with raw miscanthus, 75 and 80% of expected methane production were obtained with miscanthus pretreated at 5 and 10%, respectively. Riggio et al. (2016) [4] carried out the process using an LBR fed with spent cow bedding. They obtained 168 NmLCH4·gVS <sup>−</sup><sup>1</sup> after 60 days, which represents 86% of the BMP value. Thus, the overall performance of the LBR evaluated in this study is satisfactory, although the small increase in methane production at 59 days remains surprising. Dry anaerobic digestion

inoculum might require some adaptation to the pretreated biomass. Another explanation could be the high water absorption capacity of miscanthus. As BMP tests were carried out in diluted medium (5 gTSmiscanthus·L −1 ), the high amount of available water can favor contact between lime and biomass. This could enhance the action of lime if it continues during the AD process.

The improvement in methane production after a 10% pretreatment was quite low (6%). This was related to a higher VFA production at the beginning of the AD run (Figure 4a), thus revealing that, unlike a 5% pretreatment, the 10% pretreatment can lead to the release of easily biodegradable matter. Nevertheless, this increase in methane production is not sufficient to justify the application of a full-scale pretreatment.

#### **4. Materials and Methods**

#### *4.1. Miscanthus*

*Miscanthus. x giganteus* Floridulus was grown in the North of France (49◦53 N, 3◦00 E) [27] at the INRA experimental unit of Estrées-Mons and harvested in winter 2015 during its eighth year. The soil is a deep loam soil (Orthic Luvisol, Roma, Italia, FAO classification). The climate is oceanic. The stems were dried at 64 ◦C for 4 days in a ventilated oven and ground with a crusher (Viking, model GE 220, STIHL, Stuttgart, Germany) to a coarse size (around 6 cm). The TS and volatile solids (VS) content were 94% and 98%TS, respectively.

#### *4.2. Experimental Design*

To assess the effect of CaO pretreatment on BMP and the first-order kinetics constant k, a two-factor Doehlert-type uniform network was used to define the experimental matrix. The principle and strengths of such a design is described by Goupy et al. (2014) [33] and by Witek-Krowiak et al. (2014) [34]. Briefly, it consists of a two-variable (z = 2) Doehlert design and requires N = z<sup>2</sup> + z + C experiments, with z as the number of variables and C as the number of center points. Here, N was equal to 2<sup>2</sup> + 2 + 1 = 7. The center point was repeated once. As the experiments were performed in duplicate for each condition, the total number was 16. The two variables of interest (called factors) were defined as the CaO concentration applied for pretreatment (variable A) and the duration of pretreatment (variable B). The ranges to be studied for both factors were selected based on literature, sound reasoning, and preliminary experiments carried out at the laboratory. Thus, for variable (A), the range was between 7.5 and 17.5% (% mean gCaO per 100 gTS). For variable (B), the range was between 1 and 5 days. As no optimum was found within the investigated domain, the range for variable (A) was extended to a second set of experiments and defined between 2.5 and 12.5%. The experimental domains, expressed as coded (±0, 0.5, 0.866, and 1) and real values for each factor, are listed in Table 1 for both designs.

A full second-order polynomial equation was used to model the values obtained for BMP and the first kinetics constant k as a function of the applied lime concentration (A) and the duration of the pretreatment (B). The system can be described by the following equation (Equation (1)):

$$\mathbf{Y} = \mathbf{a}\_0 + \mathbf{a}\_1 \mathbf{A} + \mathbf{a}\_2 \mathbf{B} + \mathbf{a}\_{12} \mathbf{A} \mathbf{B} + \mathbf{a}\_{11} \mathbf{A}^2 + \mathbf{a}\_{22} \mathbf{B}^2 \tag{1}$$

where Y is the BMP or the first-order kinetics constant k, a<sup>0</sup> is the constant term, a<sup>1</sup> and a<sup>2</sup> are the linear coefficients associated with each variable, a<sup>12</sup> is the coefficient associated with the interaction between both variables, and a<sup>11</sup> and a<sup>22</sup> are the quadratic coefficients. A detailed calculation of the coefficients is already available in the literature [35].

The model was validated using a Fischer test. The significance of each coefficient in the model was tested using a Student's *t*-test [35]. The results were then compared using standardized effects in a Pareto chart.

#### *4.3. Alkaline Pretreatments*

The pretreatments were carried out at ambient temperature, without mixing, and in duplicate in 500 mL flasks using lime (CaO, Akdolit® Q90; purity ≥ 92%, Paris, France) and 2 gTS of miscanthus in conditions reported in Table 1. Another originality of this study is the high TS loading (130 g·L −1 ) selected to test conditions with low water input.

#### *4.4. Measure of Methane Potential*

All pretreated samples (solid and liquid fractions altogether) were digested in a 500 mL flask with a working volume of 400 mL. Bicarbonate buffer (NaHCO3, 50 g·L −1 ), macroelement and oligoelement solutions, anaerobic sludge at 5 gVS L −1 , and the substrate at 5 gTS·L <sup>−</sup><sup>1</sup> were added [36]. Degasification with nitrogen was carried out to obtain anaerobic conditions. Duplicate bottles were incubated at 35 ◦C for 60 days.

#### *4.5. Methane Production Kinetics*

All methane potentials are expressed in NmLCH4·ginitialVS −1 . Thus, the eventual losses of organic matter during pretreatments are included in the results. To quantify the impact of pretreatment on the kinetics of methane production, the first-order kinetic constants were calculated by using the least-squares fit of methane production data versus time (*t*) to the following equation:

$$V = Vma\mathbf{x}\left(\mathbf{1} - e^{-kt}\right) \tag{2}$$

where *V* is the volume of methane (NmLCH4·gVS −1 ), *Vmax* the maximum producible methane volume (NmLCH4·gVS −1 ), *k* the first-order kinetics constant (d−<sup>1</sup> ), and *t* is the digestion time (d). *Vmax* and *k* were determined using the Microsoft Excel Solver function.

#### *4.6. Leach Bed Reactors*

In order to represent farm batch dry anaerobic digestors used on farms, LBR systems were employed for these experiments. They were previously used and described by Riggio et al. [4]. Experiments were carried out in a 6 L LBR fed with 300 gTS of substrate and inoculum and 1.1 L leachate. The substrate was composed of 85% (in wet weight basis) manure and 15% miscanthus (corresponding to 40%VS). Cow manure from wheat straw bedding was collected from a dairy farm in the South of France and stored at −20 ◦C. Before feeding the reactors, two different lime pretreatments were applied to miscanthus at room temperature, with low water addition (to reach 13%TS) and no mixing: 5 and 10% for 5 days. The inoculum used came from a previous experiment. It was composed of a mix of digested manure, miscanthus, and raw sorghum and kept at 35 ◦C. The leachate originated from a previous experiment and was also maintained in mesophilic conditions. It did not contain VFAs and was diluted with water and buffered with NaCO<sup>3</sup> at 1.3 g·Laddedwater −1 . The substrate/inoculum ratio was 6 (gVS·gVS −1 ) and the TS content of the solid fraction in the reactor (including miscanthus, manure, and inoculum) was 19%. Taking the leachate volume into account, the overall TS content was 12%. The pretreated substrate at 10% CaO for 5 days and the raw substrate (control) were digested in the LBR in duplicate, whereas only one reactor was run for the other pretreatment. Degasification with nitrogen was carried out to obtain anaerobic conditions.

#### *4.7. Analysis*

The TS and VS contents were measured according to standard methods [37]. The leachate used was characterized in terms of alkalinity, VFA concentration, and pH. According to the APHA method, alkalinity was performed by 0.1 N hydrochloric acid titration [37]. VFAs were analyzed in a Perkin Elmer Clarus 580 gas chromatographer with helium as the gas vector [38]. The pH was measured with WTW pH-electrode SenTix 41 connected to a WTW inoLab pH 7110 operational manual transmitter. Biogas flow rate from reactors was recorded every 5 min with the use of a lab-made software connected to a Ritter Milligascounter MGC-1 V3. Biogas volume in BMP tests was monitored using a manometric device (LEO 2, KELLER) and biogas composition was determined by gas chromatography as described in Sambusiti et al. (2012) [39].

#### *4.8. Statistical Analysis*

For results obtained from the DOE, Wilcoxon tests were performed using R software (version 3.2, R Development Core Team, R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing, Austria, Vienna, 2004, ISBN 3-900051-07-0.). Effects were considered to be significant for *p*-values < 0.05.

**Author Contributions:** Conceptualization, H.C.; Formal analysis, H.L.T. and J.S.; Investigation, H.L.T.; Methodology, R.E. and H.C.; Writing—original draft, H.L.T.; Writing—review & editing, J.S., R.E. and H.C.

**Funding:** The authors acknowledge the French National Research Agency for funding the "Biomass For the Future" project (ANR, grant ANR-11-BTBR-0006-BFF).

**Acknowledgments:** Acknowledgements are addressed to Brancourt and Arnoult (INRA Estrée-Mons and Péronne) who provided miscanthus samples. The authors thank Théo Closet for his involvement in the experiments during his internship.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Sample Availability:** Samples of the compounds are not available from the authors.

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Review*
