Bioresource Technology 159 (2014) 355–364

Contents lists available at ScienceDirect

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

Use of Empty Fruit Bunches from the Oil Palm for bioethanol production: A thorough comparison between dilute acid and dilute alkali pretreatment S. Chiesa ⇑, E. Gnansounou Bioenergy and Energy Planning Research Group (BPE), Ecole Polytechnique Fédérale de Lausanne (EPFL), CH-1015 Lausanne, Switzerland

h i g h l i g h t s  Oil palm Empty Fruit Bunches shown as good candidates for bioethanol production.  Two pretreatment techniques compared (acid/alkali): the first option is superior.  Optimized dilute acid treatment (161 °C, 10 min, 1.5% acid): 85% glucose yield.  Alkali treatment is seriously hampered by high lignin content of the feedstock.  Fate of the different components of biomass monitored during each treatment.

a r t i c l e

i n f o

Article history: Received 11 December 2013 Received in revised form 18 February 2014 Accepted 26 February 2014 Available online 7 March 2014 Keywords: Lignocellulosic bioethanol Oil Palm Empty Fruit Bunches (EFBs) Dilute sulfuric acid treatment Dilute sodium hydroxide pretreatment Surface response analysis

a b s t r a c t In the present work, two pretreatment techniques using either dilute acid (H2SO4) or dilute alkali (NaOH) have been compared for producing bioethanol from Empty Fruit Bunches (EFBs) from oil palm tree, a relevant feedstock for tropical countries. Treatments’ performances under different conditions have been assessed and statistically optimized with respect to the response upon standardized enzymatic saccharification. The dilute acid treatment performed at optimal conditions (161.5 °C, 9.44 min and 1.51% acid loading) gave 85.5% glucose yield, comparable to those of other commonly investigated feedstocks. Besides, the possibility of using fibers instead of finely ground biomass may be of economic interest. Oppositely, treatment with dilute alkali has shown lower performances under the conditions explored, most likely given the relatively significant lignin content, suggesting that the use of stronger alkali regime (with the associated drawbacks) is unavoidable to improve the performance of this treatment. Ó 2014 Elsevier Ltd. All rights reserved.

1. Introduction Over the last decades, novel forms of energy alternative to petroleum based products have been sought for, principally to cope with the growing energy demand and environmental concern caused by fossil fuels. Lignocellulosic ethanol, i.e., alcohol obtained from plant residues, has emerged as an attractive option. This is mainly due to its ease of use in the transportation sector, its renewable character and its relatively even distribution worldwide. In several tropical countries in Africa, Asia and South America, relevant amounts of biomass residues are generated from the palm oil industry. Among these, in particular, the leftovers obtained ⇑ Corresponding author. Present address: Station de recherche Agroscope Liebefeld-Posieux ALP, Rte de la Tioleyre 4, Case postale 64, CH-1725 Posieux, Switzerland. Tel.: +41 (0)26 407 71 11; fax: +41 (0)26 407 73 00. E-mail address: [email protected] (S. Chiesa). http://dx.doi.org/10.1016/j.biortech.2014.02.122 0960-8524/Ó 2014 Elsevier Ltd. All rights reserved.

when oil palm fruitlets are stripped from the bunches holding them together have been given growing attention over the last few years. Palm trees are principally cultivated for their fruits from which an edible oil can be extracted. Beyond local consumption in traditional dishes, palm oil is widely used in the food industry worldwide, given its moderate price combined with a semi-liquid status, which is highly desirable for industrial applications. Some of the residues that are generated during the processing of the fruits at the oil plant are already used to generate power for the plant itself. Others, however, are considered as a waste and remain at present underused. Oil Palm tree Empty Fruit Bunches belong to this second category. They are humid as they undergo a sterilization process in autoclave before stripping. Noteworthy, EFBs seasonality is not particularly pronounced, and fruits may be available, depending on the country, for several months of the year. Also, this feedstock is already available at the chemical plant where palm oil is produced, and it is natural to think of an extension of

356

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

the industrial plant that could process this residue to produce ethanol. A rational use of EFB involves using its high polysaccharide content, mainly cellulose, to produce bioethanol by fermentation. As any other substrate for 2nd generation ethanol production, EFB contain sugar polymers that must be broken down to fermentable sugars by enzymes (saccharification) before being fermented via ordinary techniques (fermentation). Unfortunately cellulose is located in cell walls whose resistant structure must be disrupted to make cellulose attack possible by enzymes: therefore, a physicochemical pretreatment must be performed at the beginning of the ethanol production chain. Industrially, the pretreatment step is critical both for its cost and for the impact on all following operations performed. This operation must be studied in detail and optimized for each particular plant residue. Particular emphasis should be given to those treatments which could reasonably be scaled to a real industrial context. EFB have been recently investigated for bioethanol production using various techniques (Lau et al., 2010; Hassan et al., 2013; Park et al., 2013). Some difficulties arise in interpreting the results, possibly given the dissimilarity of the material used. Among the different studies, geographical origin and the composition of the samples vary, and sample preparation is not always thoroughly described. EFBs are peculiar in that they are usually left in the fields for variable times (usually a few days) after harvesting, and they undergo a sterilization step before separation of the fruits from the bunch takes place. These operations, which are demanded by the oil production, can presumably have an impact on the composition and susceptibility to successive treatments of the feedstock; this point, however, has not been given much attention in the literature. Also, a complete characterization of the biomass in terms of its chemical composition is not always given. A few studies (Rahman et al., 2007; Yunus et al., 2010) have performed a dilute acid hydrolysis of this substrate, but then have focused exclusively on the fermentation of the liquid stream obtained upon pretreatment, neglecting the cellulose-rich solid residue. In other instances, no statistical optimization of the processes has been carried out (Millati et al., 2011; Jung et al., 2013a,b), although good glucose yields have been obtained from the solid pretreated at temperatures around 190 °C. Alkali treatments have given far less impressive results. This could be partially interpreted considering the composition of EFB, which have relatively high lignin content, generally between 20% and 30%: lignin is known to reduce the efficiency of treatments using basic solutions. In the case of ammonia fiber explosion (AFEX) (Lau et al., 2010), a post-treatment size reduction was essential to later reach good glucose yields: such behavior is peculiar, as this operation is not demanded by other feedstocks, and is probably justified by the fibers’ toughness. The economic impact of this post-AFEX operation should be thus assessed. A number of studies have used NaOH solutions of relatively high concentrations, reporting variable outcomes. Hassan et al. (2013) could obtain glucose yields around 50% after a NaOH-enhanced (5%) steam pretreatment at 110 °C. Han et al. (2011) reported higher glucose conversion (85% of the original glucan) upon enzymatic saccharification following a relatively concentrated (almost 3 M NaOH) alkali treatment at 128 °C and 22 min. Similarly, Park et al. (2013) claim to have attained an economically attractive (>40 g/L) ethanol concentrations with simultaneous saccharification and fermentation of NaOH-pretreated EFBs at 121 °C: although in this case the alkali loading used was milder (1 M), the use of a very basic solutions raises concern about possible economic and environmental impacts. Discrete results (Choi et al., 2013) have been obtained with prolonged overnight soaking in a 3% NaOH solution before pretreatment: the main drawback of such an approach is given by the longer times.

The present study was conducted in the frame of a larger project involving the comparison of three different strategies for pretreatment (namely using pure water, dilute acid or dilute alkali): in the present work the results from the dilute acid and dilute alkali treatments are presented, whereas those from the pure water treatment will be the object of another publication. These pretreatments have been chosen as they are considered of interest for a potential future extension to the industrial case (Mosier et al., 2005), particularly as the use of exotic reagents is not needed. At present, indeed, most of the commercial plants do not employ any reagent at all, preferring hydrothermal approaches. In this optics, the dilute acid and alkaline treatments have been carried out in this work with the cheapest chemicals available (sulfuric acid and sodium hydroxide) and a particularly dilute regime has been chosen for the solutions to be employed. This point is crucial for limiting both the economic and the environmental impacts. We incidentally note that oil palm cultivation is already bitterly criticized on the ground of its impact on the environment. Dilute regimes must be clearly coupled with higher temperatures than in other studies (Han et al., 2011; Park et al., 2013). This would also allow to increase the hourly productivity in an hypothetical industrial context, although the generalization of laboratory results, especially concerning the optimal working conditions, is generally very complex as results critically depend on the set-up adopted. Rigorous compositional analyses have been employed to fully characterize the feedstock during each of the stages studied (pretreatment and saccharification). Mass balances have been used to measure the compositions of both the solid and liquid streams generated during biomass processing. Pretreatment processes have been standardized to ensure reproducible conditions and to limit external influences. In particular, special attention has been paid to the temperature profile of each treatment, a point which is often underestimated in the literature and that can nonetheless induce variations in the response to the treatment. Enzymatic saccharification of pretreated samples has been carried out to assess the pretreatment performance. A statistical design for the pretreatment runs was employed, and optimization was performed for each treatment in order to find the best operational conditions. Finally, the feedstock used in this study was collected in Africa, which is also the region of origin of the oil palm. This is different from mostly of the works in the literature that employ EFBs from South-Eastern Asia. Also, EFBs underwent a preliminary preparation similar to that encountered in the industrial context. In this study, in fact, we considered as raw feedstock EFBs that have been harvested, left a few days in the fields and then sterilized for 1 h at 134 °C. These parameters are very similar to those employed in the case of industrial processing: the residue thus mimics at best what could be reasonably available for a hypothetical industrial plant.

2. Methods 2.1. EFB samples The EFBs from oil palm used in this study were collected in the nearby of Sakété, Republic of Benin (Africa), in the month of June 2011, almost at the end of the harvesting season. They had been previously collected in the nearby palm fields and left a few days on the ground to start fermentation; then, the fruits were separated by hands and empty bunches set aside. EFBs were carried to Switzerland the day following their collection and sterilized in the laboratory at the EPFL using an autoclave operating at 134 °C for 1 h. Samples were air-dried to avoid spoilage, then defibered by hand to a typical length of 4–5 cm; the fibers from different bunches have then been mixed thoroughly and stored in closed zip plastic bags screened from the sunlight. A small portion has

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

been further processed with a laboratory grinder operated intermittently before undergoing compositional analyses. 2.2. Compositional analyses The chemicals used in this work have been purchased from Sigma–AldrichÒ, Switzerland. Compositional analyses have been carried out on raw and pretreated biomass according to a series of protocols developed by the NREL whose central part is represented by the acid hydrolysis of biomass samples to quantify carbohydrate and lignin (Sluiter et al., 2008a). Slight modifications were adopted whenever needed, especially concerning the extraction step. In fact, two consecutive extractions, first with water for 18 h and then with alcohol for about 6 h have been performed as suggested for agricultural residues (Sluiter et al., 2008b). The amount of extractives removed from the samples was however measured removing the solvent with a rotary evaporator and measuring the loss in the solid phase after each extraction using gravimetric technique. This approach was chosen as our experience has shown that it was easier to perform and more accurate. A standard two-step acid hydrolysis followed by HPLC measurements (Sluiter et al., 2008a) has been employed for measuring the polysaccharide content and sugar degradation products. The apparatus used was a Hewlett-Packard Agilent series 1100 chromatographer. Sugars were measured with a Supelcogel™ Pb 30 cm  7.8 mm column kept at 80 °C, using Millipore water as the mobile phase (rate flow of 0.6 mL/min) and injecting samples of 50 lL. A H+/CO 3 deashing guard was used to avoid interferences in the chromatogram. Oppositely, HMF, furfural and of a number of organic acids were detected with a Biorad Aminex HPX-87H column at 65 °C, using 5 mM sulfuric acid as the mobile phase (rate flow of 0.6 mL/min) and samples of 25 lL. A refraction index detector (RID) at 55 °C was employed in both cases. Soluble lignin was measured via UV spectrophotometry in transmission, using a coefficient of 30 L g1 cm1 for the absorptivity at 320 nm (Hyman et al., 2008). Upon completion of pretreatment, solid residues were separated from the pretreatment liquids through vacuum filtration and extensively washed with deionized water. Compositional analyses were then repeated following the same protocols used for raw biomass, except that extractions have not been carried out as they had been proved unnecessary in preliminary trials: all the water extractives are hence assumed to be solubilized either during the pretreatment phase or in the washing step. Such procedure has given good mass closures. Aliquots of the pretreatment liquids have been screened for monosaccharides and degradation products using HPCL technique. For NaOH pretreatment, liquids have screened via UV–vis transmission for total phenolic content (absorbance at 260 nm, calibration against standards of gallic acid).

357

For pretreatment, samples of 4 g of biomass have been used, together with 66.7 mL of the appropriate solution to reach a 6% SL. For each of the treatments studied, 3 parameters have been varied: reaction temperature (140–210 °C for the acid, 121–195 °C for the alkali), reaction time (1–20 min for both treatments) and solution concentration (0.05–2% v/v H2SO4 for the acid, 0.04–2% w/v for the alkali). A Box-Behnken design (shown in Tables 1 and 2) with 5 repetitions of the center point has been adopted, this choice enabling the construction of a quadratic model and the related response surface. Note that the acid content is given in volume by volume percentage in analogy to some of the studies in the literature (Saha et al., 2005): thus, 2% v/v means that 2 mL of concentrated (98%) H2SO4 have been diluted with deionized water to reach 100 mL. However, the content given in terms of weight percentages (w/w) that can be found in other publications (Ballesteros et al., 2008; Sun and Cheng, 2005) is very different, due to the specific gravity of sulfuric acid. For the alkali treatment, the notation chosen means that, for instance, in the case of the most concentrated solution (2% w/v), 2 g of solid NaOH pellets have been dissolved in 100 mL deionized water. 2.4. Enzymatic assays and hydrolysis protocols A commercial cellulase cocktail from Trichoderma reesei ATCC 26921 (CelluclastÒ 1.5L) was purchased from Sigma–AldrichÒ Switzerland, as well as b-glucosidase lyophilized powder from almonds, and used for enzymatic saccharification. A stabilized, sterile-filtered penicillin–streptomycin solution was also used as antibiotic to inhibit microorganism growth. The Filter Paper Assay has been used for monitoring cellulase activity (Adney and Baker, 2008). b-Glucosidase activity was measured with a cellobiose assay (Zhang et al., 2009). Saccharification of raw and pretreated biomass was performed at 50 °C with an Infors Ecotron shaking incubator (agitation set at 150 rpm) over 72 h according to Selig et al. (2008). The cellulose content was of 100 mg for the fixed 10 mL final liquid volume. Cellulase and b-glucosidase were added at a loading of 40 FPU/g glucan and of 60 IU/g glucan, respectively, and the saccharification time started: HPLC was used to monitor the advancement of the process. 2.5. Data analysis Freely distributed R software was used in an RStudio environment to perform statistical data analysis and to produce graphical display of the results. 3. Results and discussion 3.1. Raw EFBs samples

2.3. Pretreatment reactor The pretreatment reactor was a stainless steel/Hastelloy temperature-controlled 300 mL Parr high pressure reactor. Biomass samples mixed with the aqueous solution of choice were placed in the reactor and the latter sealed. Special care was taken to ensure reproducible conditions: target temperatures (150–200 °C) were reached in a typical time interval of roughly 20–30 min, clearly depending on the final target temperature. The temperature was kept constant for the desired time. Upon completion of pretreatment, heating was stopped and the mantle removed to decrease the thermal inertia of the system; finally, the reactor was cooled down by circulating cold water through the cooling circuit to avoid lengthy cooling periods. Noteworthy, the interest of this procedure is naturally limited to a laboratory study, in order to obtain reproducible conditions of the temperature profile.

The chemical composition of the EFBs samples used in this work was measured in septuplicate. Values were calculated on a dry matter content basis including extractives: water and ethanol extractives constitute 15.7% and 1.2% of the total dry matter, respectively. Glucan (29.6%) and xylan (18.8%) together make up almost half (48.4%) of the content of EFBs, attesting the richness of this feedstock in these polysaccharides. Other minor hemicellulosic sugar polymers are present, namely galactan (0.6%), arabinan (1.2%) and mannan (0.2%), but they are however rather scarce. This is presumably due to the peculiar structure of the hemicellulose found in the EFBs under exam and to their geographical origin, but also to the post-harvest operations carried out, the autoclave sterilization in particular. The total lignin content was found to be 22.9%: acid insoluble (Klason) lignin accounted for 20.7% of the dry mass, whereas acid soluble lignin accounted for only

Furfural [g/L] Glc yield sacch. Glc yield sacch. Xyl yield sacch. HMF [g/L] Acetic acid Xylan Glucan Galactan Lignin Xylan % mass Klason lignin Glucan (% theor. xylan in [g/100 g raw (% theor. glucan (% glucan in [g/L] [g/100 g [g/100 g recov. recov. recov. (% pretr. solid) (% pretr. (% pretr. (% pretr. recov pretreated solid) raw biomass) pretr. solid) (% initial) (% initial) (% initial) raw biomass] raw biomass] biomass] solid) solid) solid) 72 h (48 h)

– 175 140 175 140 210 175 140 140 175 175 175 175 175 210 175 210 210

– 83.4 85.0 69.8 64.5 62.4 47.4 62.7 53.2 42.3 n.a. 44.5 43.0 45.0 30.9 38.0 33.5 34.6

Raw 1 10.5 20 1 10.5 1 20 10.5 10.5 10.5 10.5 10.5 10.5 1 20 20 10.5

– 0.05 0.05 0.05 1.025 0.05 2 1.025 2 1.025 1.025 1.025 1.025 1.025 1.025 2 1.025 2

– 2.46 2.35 0.70 0.20 0.67 1.49 1.52 1.61 2.23 2.25 2.25 2.25 2.27 2.30 2.79 3.56 3.59

20.7 25.2 33.0 31.4 38.3 36.4 49.4 41.4 43.1 48.5 57.4 53.2 59.1 51.5 61.0 78.6 82.3 89.2

29.6 36.4 34.1 36.0 41.6 51.8 43.9 37.9 53.3 36.5 36.3 37.0 38.4 39.5 5.0 9.1 3.0 5.0

18.8 23.2 23.4 15.5 10.1 4.9 1.6 3.0 1.9 1.2 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d.

0.6 n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. n.d. 1.2 1.4 n.d. n.d. n.d. n.d. n.d.

– 90 120 94 106 97 100 111 98 87 107 103 108 98 80 126 116 130

– 103 98 85 91 109 70 80 96 52 54 57 56 60 5 12 3 6

– 103 106 58 35 16 4 10 5 3 0 0 2 0 0 0 0 0

– 0.40 0.28 0.71 1.43 0.98 0.97 1.38 1.01 1.51 1.18 1.23 1.40 1.53 1.20 1.37 1.36 1.68

– n.d. n.d. 0.02 0.02 0.14 0.14 0.03 0.04 0.26 0.20 0.20 0.22 0.28 0.21 0.22 n.d. n.d.

(0.49) (0.34) (0.89) (1.79) (1.23) (1.21) (1.72) (1.26) (1.89) (1.47) (1.53) (1.74) (1.92) (1.5) (1.71) (1.7) (2.1)

(0.02) (0.02) (0.18) (0.18) (0.04) (0.05) (0.33) (0.25) (0.25) (0.28) (0.35) (0.26) (0.28)

– n.d. n.d. 0.10 (0.12) 0.1 (0.13) 1.38 (1.73) 3.20 (4.0) 0.56 (0.69) 1.47 (1.84) 3.47 (4.34) 3.77 (4.72) 3.50 (4.37) 3.88 (4.85) 3.90 (4.87) 2.94 (3.68) 2.37 (2.96) 0.92 (1.15) 0.55 (0.69)

9.5 (8.7) 17.0 (15.1) 7.3 (6.6) 19.6 (17.3) 31.2 (28.4) 53.1 (45.7) 77.3 (65.7) 57.8 (52.0) 71.8 (63.5) 83.0 (77.5) 85.3 (78.2) 83.2 (77.0) 85.5 (78.1) 82.6 (72.6) 39.0 (36.8) 32.4 (28.2) 10.0 (8.2) 16.0 (14.5)

– 17.4 7.1 16.6 28.3 58.0a 54.3 46.4 68.8 43.3 46.0 47.3 47.7 49.6 2.0 3.8 0.3 0.9

– 11 4 24 27 39 44 56 59 n.a.* n.a.* 30 20 n.a.* n.a.* n.a.* n.a.* n.a.*

n.d. = not determined n.a = not available * = no xylan present in the pretreated EFB sample used for saccharification. Table 2 Recapitulative results relative to the dilute alkali treatment: conditions, composition of the pretreated solids, corresponding mass recoveries, pretreatment liquid stream composition and yields upon enzymatic saccharification. In the first line, lignin, glucan and xylan contents in raw EFB are given to facilitate comparison.

a

Conditions

Temp (°C)

Time (min)

Alkali (% w/v)

Klason lignin (% pretr. solid)

Glucan (% pretr. solid)

Xylan (% pretr. solid)

% mass recov.

Lignin recov (% initial)

Glucan recov. (% initial)

Xylan recov. (% initial)

pH

Glc yield sacch. (% theor. glucan pretr. solid) 72 h (48 h)

Glc yield sacch. (% glucan in raw biomass)

Xyl yield sacch. (% theor. xylan in pretreated solid)

Raw 121 °C; 121 °C; 121 °C; 121 °C; 158 °C; 158 °C; 158 °C; 158 °C; 158 °C; 158 °C; 158 °C; 158 °C; 158 °C; 195 °C; 195 °C; 195 °C; 195 °C;

– 121 121 121 121 158 158 158 158 158 158 158 158 158 195 195 195 195

– 1 20 10.5 10.5 1 20 10.5 10.5 1 20 10.5 10.5 10.5 1 10.5 20 10.5

– 1.02 1.02 0.04 2 2 0.04 1.02 1.02 0.04 2 1.02 1.02 1.02 1.02 0.04 1.02 2

20.7 21.1 17.3 27.3 18.7 18.8 34.9 15.1 16.5 26.4 12.3 20.1 17 19 10.6 30.6 9.9 9.9

29.6 39.6 42.2 34.6 41.7 45.3 30.9 42.7 42.1 34.6 48.6 43.5 43.2 43 49.8 43.6 52.4 61.3

18.8 24.8 25.1 23.1 25.8 26.7 21.6 25.6 25.1 24.2 26.9 26.2 26.0 26.2 26.0 16.2 26 25.1

– 73.8 73.1 81.4 64.1 70.2 91.4 65.5 69.5 80.2 51.2 76.4 70.0 69.1 50.7 70.2 47.9 40.8

– 68 55 97 52 57 139 43 50 92 27 67 52 57 23 94 21 18

– 99 104 95 90 107 95 94 99 94 84 112 102 100 85 103 85 84

– 97 98 100 88 100 105 89 93 103 73 106 97 96 70 60 66 54

– 12.6 12.53 7.5 12.95 13 6.5 12.23 12.24 7.17 12.82 12.36 12.7 12.6 11.11 5.01 10.49 12.47

9.1 (8.5) 21.9 (20.3) 38.4 (32.6) 7.4 (6.7) 30.3 (27.3) 32.7 (28.4) 14.7 (13.5) 31.3 (27.5) 31.3 (26.4) 9.4 (8.2) 42.6 (36.8) 31.3 (27.3) 31.4 (28.3) 31.0 (28.2) 42.0 (39.8) 18.0 (15.8) 25.0 (22.0) 35.0 (32.8)

– 21.7 39.9a 7.0 27.3 35.0a 14.0 29.4 31.0 8.8 35.8 35.1a 32.0a 31.0 35.7 18.5a 21.3 29.4

– 24 24 5 36 38 6 38 34 5 47 38 37 35 55 30 55 78

1.02%; 1 m 1.02%; 20 m 0.04%; 10.5 m 2%;10.5 m 2%; 1 m 0.04%; 20 m 1.02%; 10.5 m 1.02%; 10.5 m 0.04%; 1 m 2%; 20 m 1.02%; 10.5 m 1.02%; 10.5 m 1.02%; 10.5 m 1.02%; 1 m 0.04%; 10.5 m 1.02%;20 m 2%; 10.5 m

= values calculated using a mass recovery in excess of 100%.

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

Temp Time Acid CSF (°C) (min) (% v/v)

358

Table 1 Recapitulative results relative to the dilute acid treatment: conditions, composition of the pretreated solids, corresponding mass recoveries, pretreatment liquid stream composition and yields upon enzymatic saccharification. In the first line, lignin, glucan and xylan contents in raw EFB are given to facilitate comparison.

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

359

Fig. 1. Composition of the solid residues upon dilute acid pretreatment in terms of lignin, glucan and xylan (panel a) and mass recoveries of these compounds with respect to the content in raw biomass (panel b).

2.2%. Finally, the ashes measured in the raw samples constituted 5.4% of the biomass content: however, as a portion of the ashes (1.7%) could be solubilized by water it has been assigned to water extractives, and the true ashes content of the biomass was evaluated to be 3.7%. The last class of component detected was acetyl groups, accounting for 1.5% of the biomass content. A mass balance closure of 95.4% was hence reached, in line with the figures usually found for biomass feedstocks. 3.2. Acid pretreatment A single parameter named Combined Severity Factor (CSF) is routinely employed to resume the three different experimental conditions (time, temperature and acid concentration). It is defined as follows (Overend and Chornet, 1987):

CSF ¼ log R0  pH ¼ log½t  e

T r T 0

x

  pH

ð1Þ

where T0 and x are constants (100 °C and 14.75, respectively), Tr and t are the treatment temperature and time and pH refers to the final pH of the acidic solution. The CSFs relative to the runs performed in this study are summarized in Table 1, along with several parameters of interest. The calculated total mass recoveries upon pretreatment and washing steps never exceeded 85%, and were just above 30% for the harshest conditions explored in this work. A decreasing pattern of mass recoveries was found for increasing values of the CSF because of more extensive solubilization. The observed trend is approximately linear: for a linear regression a R2 larger than 0.91 was in fact found.

To consider more in detail the fate of the different components in the sample, we can inspect Fig. 1a, which presents the composition of the pretreated solid in terms of glucan, xylan and lignin. Note that the percentages are given here on a pretreated DM basis. Acid treatments notoriously solubilize the hemicellulosic component to an extent related to their harshness, while the more resistant cellulosic and lignin portions are generally less affected. Coherently with what expected, the samples undergoing pretreatments corresponding to CSF up to 2 have been enriched in glucan proportionally to the treatment’s intensity: glucan was in fact found in the range 33–53% (compared with 29.6% in the raw biomass). This trend was however reversed for the harshest conditions explored, and in particular for CSF >2 the glucan content abruptly decreased. These losses can be reasonably explained in terms of both solubilization and degradation of the cellulosic fraction. Noteworthy, the decrease beyond the point CSF = 2 is extremely fast, although direct comparison with other works is not straightforward, given the variability of the treatments sharing the same severity factor but performed with different reactors. Oppositely, large losses of xylan have been measured for increasing treatment severity: xylan was actually efficiently removed from the solid phase for almost all runs except the lightest. Other hemicellulosic sugars have not been detected, or they were present only to trace levels. Noteworthy, the EFBs employed in this work are not particularly rich in hemicellulosic sugars other than xylose and these tiny amounts were hence almost totally lost during pretreatment, apart for the two lightest runs. Also, only a part of the ashes present in the raw samples were recovered after pretreatment, the other portion being solubilized in the acidic

360

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

solution. Ashes thus accounted for 2–4% dry mass in the pretreated samples, with the higher levels found for the harsher treatments. In Fig. 1b the recoveries for glucan, xylan and lignin are shown with respect to the initial content in the raw samples. Glucan recoveries spanned over a wide range from quantitative to less than 15% the original content for the four harshest pretreatments. Upon inspection of this graph, the region around CSF = 2 seems again of special interest under the present experimental setup, in that for harsher treatment relevant losses are encountered. Xylan was almost totally recovered in the solid for the two lightest pretreatments performed, taking into account the experimental uncertainty, but then the recovery dropped down very rapidly to values of a less than 10%, being practically zero for the harshest treatment in use. False recoveries above 100% were found in a few cases for sugar polymers and are mostly due to experimental uncertainty: they are anyway below 110%, which represents an acceptable level. Oppositely, lignin was as expected only slightly affected by the treatment: lignin recoveries were very high for all runs and in some cases exceeded 100%, especially for the harshest treatments. From the data measured we can infer that lignin is generally preserved in the solid phase for most treatments. False lignin recoveries are most likely motivated by condensation reactions of other components (extractives, for instance) onto lignin, as it has also been observed during dilute acid treatments of different feedstocks (Liao et al., 2007; Qi et al., 2010; Yan et al., 2009). Concerning ashes, it has been calculated that roughly one fourth of the original ashes were preserved in the solids, with the remaining part solubilized in the acidic solution. The liquids obtained from the pretreatment step were in most of the cases very acidic, with pH values lower than 1. The concentrations of sugar and selected sugar degradation products (inhibitors) measured via HPLC are given in Table 1 both in g/L and in grams/100 g of raw biomass undergoing pretreatment, for ease of comparison with other studies operating at different solid loadings. Acetic acid was detected up to 1.53 g/L and its dependency on the severity factor was not particularly evident. This compound is however probably of minor relevance with respect to possible inhibition on microorganisms. In fact, the maximum measured concentration of 1.53 g/L is far below the values reported to affect fermentation: concentrations up to 10 g/L have been found not to affect the ethanol production in a SSF process (Oliva et al., 2003), similarly to what reported by Palmqvist and Hahn-Hägerdal (Palmqvist and Hahn-Hägerdal, 2000a,b). Furfural from hemicellulose degradation, which is responsible for inducing a lag phase in the fermentation process, during which it is gradually converted to furfuryl alcohol, was found in this study at varying levels from lower than the detection range to almost 4 g/L. For low CSF treatments furfural concentration increased with harshness, whereas for CSF slightly higher than 2 it declined. We infer that the amount produced in the EFBs pretreatment liquids is to be carefully monitored if fermentation of this portion is of interest, as for some runs marked inhibition can be reasonably expected. In fact, in a study on the yeast Pichia stipitis (Roberto et al., 1991), the authors concluded that while at low concentrations below 0.5 g/L furfural may have positive effects on the growth of this yeast, above 2 g/L cell growth is totally inhibited. In a different study employing the same yeast (Nigam, 2001), levels above 1.5 g/L similarly interfered with the functioning of the cells and severely lowered the ethanol concentrations attainable. HMF, originating mainly from the cellulosic fraction, was detected at levels below 0.3 g/L and a less clear pattern was associated with this compound. For the two harshest runs, however, HMF levels dropped down again. Globally, HMF is found at levels which should not be of special alarm, considering that in their

review on inhibitors, Mussatto and Roberto (2004) report HMF concentrations of the order of 1 g/L as problematic with respect to different microorganisms, thus well above the values found in our study. With respect to solubilized sugars, low intensity treatments could not apparently solubilize any polysaccharide, while other treatments solubilized a relevant amount of xylose, with a strong variability among the runs. For harsher treatments, marked xylose losses were observed. Glucose was also present accounting for up to 40% of the initial content in one single case, being less in the other runs. The final (72 h) saccharification yields are shown in Table 1 for the different conditions pertaining to the experimental plan. Considering that saccharification of raw (thus not pretreated) EFB samples performed alongside has given also values around 10% theoretical, a vast range of values has been obtained. Two runs gave extremely limited glucose yields (7.3% and 10.0% of the theoretical glucose maximum): this has happened for extreme experimental conditions, the corresponding pretreatments being either way too harsh or too mild. Other samples oppositely showed very high digestibilities, with yields up to 85.5%. This was the case of the samples pretreated at conditions close to the center points of our experimental plan, where conditions are optimally balanced for the feedstock in use. Variable amounts of free xylose have been detected in the saccharification liquids for some runs as commercial cellulase preparation also displays some xylanase activity, sufficient to liberate a portion of the xylan initially present in the pretreated biomass. Note however that as xylan was not initially present in all experiments, for some of the runs the xylose yield could not be calculated. One should however consider that the enzyme loadings used in this work are quite high if considered from a commercial viewpoint. In the industrial context, operating at lower loadings will be necessary for the economy of the process. Surfactants may also constitute an option to cut on the amounts of enzymes used. A rigorous statistical analysis of the glucose yields was performed: a quadratic model has been fitted to the experimental data and an ANOVA analysis has been carried out to assess the validity of the model. As the latter was found significant (with a p-value of 0.0025), the contributions of single parameters have been examined (throughout this work A = temperature, B = time, C = acid): of the initial possible terms (A, B, C, A2, B2, C2, AB, BC and AC) the terms in AB and BC were not significant to the 95% confidence level and have not been included in the model. The term B and C were not significant at the first order but kept to preserve hierarchy of the model. The final equation describing the model in terms of coded factors is hence:

glcyieldð%theorÞ ¼ þ83:92  6:25  A  5:59  B þ 12:56  C  25:40  AC  24:47  A2  24:95  B2  22:40  C 2

ð2Þ

From the quadratic model given in Eq. 2 we can indeed see that at the first order the factor C (acid percentage) has the stronger influence on saccharification yields; however, larger coefficients are attributed to the second order terms, indicating a strong nonlinearity in the relation between the liberated glucose and each of the factors considered. Also, only one interaction is statistically significant (AC), namely that between temperature and acid. This is evident also when considering the runs carried out at the highest temperature (210 °C), for which only that conducted at the lowest acid content has given a satisfactory saccharification of 53.1% theoretical, with the remaining yielding less than 40%. Coupling high temperature and elevated acid content thus create conditions which are detrimental to following saccharification of EFBs. One

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

Fig. 2. Response surface for glucose yield upon saccharification for acid pretreated EFB, displayed as a function of temperature and acid concentration when time is fixed to 10.50 min.

plot relative to the response surface, for a reaction time of 10.5 min, is given in Fig. 2. Maximization of the saccharification yield with the only condition that the three parameters under study be in the experimental range studied, provides one solution at 161.5 °C, 9.44 min time and 1.51% acid loading. This is not far from the middle of the range explored (175 °C, 1.5 min, 1.025% acid) in the present experimental design. Fermentation of the sugars has not been performed in this work and pretreatment has been evaluated only based on the outcome of the saccharification step. This naturally indicates that caution should be used when trying to generalize these results to a different context. It is also interesting to examine the glucose yield expressed as percentage of the initial cellulose content of raw (rather than of the pretreated) biomass, as shown in Table 1. These yields are generally sensibly lower than those expressed in terms of pretreated biomass because of the relevant sugar losses occurred during the pretreatment phase. For instance, glucose yields of around 80% with respect to cellulose in a pretreated sample can be decreased to less than 50% in terms of the original cellulose content of the raw biomass. Naturally, in a perspective industrial situation, the fate of the portion of glucan solubilized in the pretreatment should be cautiously examined on a case-by-case basis to determine if this glucose could be somehow recovered. This process may be complex, especially when dealing with very acidic streams. 3.3. Alkali treatment Differently from the acidic case, the CSF parameter cannot be employed to describe an alkali treatment. Hence, runs are presented in order of increasing temperatures, and then of increasing basicity among runs having the same temperature. Clearly, also this representation takes into account to some extent the harshness of the treatment. The pH of the solution has been measured at the end of each run: whereas the pH of the most concentrated solution has been only slightly affected by the pretreatment, the middle-level solution passed from a starting pH of 13 to almost 10, depending on the case. Finally, for the lightest solution employed the final pH has been lowered to close to neutrality, with a couple of runs yielding a final pH in the acidic region. The measured final pH drop is motivated by the buffering property of the biomass: most probably, acetic acid and uronic acid derived from EFB hemicellulose are responsible for the lowering effect. It is

361

evident that in the case of the lightest alkali loading very mild conditions have been reached at the end of the treatment, and that for a couple of runs acid conditions were obtained in the very last stage. From what exposed, on could anticipate that this factor probably severely limited the efficiency of such treatments. The solid mass recovery is displayed in Table 2: it varies from the 40% found at the highest temperature and alkali concentration (195 °C and 2% NaOH) to the 91% found for one of the lightest treatments. Although a general decreasing pattern is visible for increasing temperature and alkali concentrations, this behavior is not exceptionally clear. The fate of the single different components of the biomass is shown in Fig. 3. The lignin content in the pretreated solid residues has generally decreased with increasing harshness. Three of the treatments performed at 195 °C, in particular, yielded a solid residue with less than 10.6% lignin (compared to the initial 22.9%). In the lightest treatments, however, lignin has been only marginally attacked and has been found at even higher percentages than in the raw biomass because of the solubilization of other components, notably the (water) extractives. Hemicellulosic sugars, xylan in particular, were found at levels in a narrow band around 25% for all runs except one. This represents a tiny enrichment with respect to the content of the raw biomass, 18.8%. On the other hand, glucan percentage increased, depending on the run, from 30.9% to 61.3%: this naturally constitutes a dramatic enrichment of the solid in this component. Globally, the residue from the alkali pretreatment of EFB has been mainly constituted of polysaccharides, consistently with the mode of action of a dilute alkali treatments which preferentially target and solubilize lignin. In fact, linkages between lignin and the polysaccharide matrix are generally broken upon this pretreatment, in particular through saponification of ester links (Lai, 2001), whereas cellulose and hemicellulose are relatively untouched. Very high recoveries were found for most of the treatments for sugars, both for glucan and xylan as visible upon inspection of Fig. 3b. Glucan was almost totally left in the solid phase: all the pretreatments yielded in fact recoveries higher than 80%, and most of them were close to 100%. Xylan recoveries were similar, except for the harshest runs that were able to solubilize around 40% of the xylan contained in the biomass. Lignin was generally effectively solubilized: taking apart the anomalous value of 139%, which is with all probabilities an outlier, one can see that for the lowest temperature runs at 121 °C a decreasing pattern of the solid recovery is associated with increasing alkali concentrations. For the middle temperature of 158 °C the trend is less clear as discrete variability is present among the runs; finally, the solution with the lightest concentration had very little effect on lignin, and more than 90% of the original content was recovered after pretreatment. The content of ashes has also been also measured and accounted for 2–3% of the dry matter content for all the runs. Part of the ashes has thus been solubilized during pretreatment. Small amounts of minor hemicellulosic sugars (arabinose and galactose) have been found for all runs, excluded those performed the highest temperature. Concerning the liquid fractions upon pretreatment, no measurable amount of sugar monomers has been detected via HPLC. As a consequence, measurement of potential sugar degradation products has not been performed. Given the lack of monosaccharides in the liquid phase, a possible fermentation of the latter appears of minor interest. In addition, concomitant degradation of lignin may generate phenolic compounds that also possess inhibitory effects on microorganisms. Phenolic compounds of this type have been estimated spectrophotometrically, and their concentration appears to be only to some extent related to the harshness of the treatment. The spectroscopic measurement performed in this work constitutes however only an approximate indicator of the potential inhibitory power during saccharification and fermentation, as phenolic compounds constitute in reality a vast category with

362

S. Chiesa, E. Gnansounou / Bioresource Technology 159 (2014) 355–364

Fig. 3. Composition of the solid residues upon dilute alkali pretreatment in terms of lignin, glucan and xylan (panel a) and mass recoveries of these compounds with respect to the content in raw biomass (panel b).

dissimilar properties. The highest amount of phenolics anyway detected for the samples treated at the highest temperature and middle/high alkali loadings. For the samples treated with 0.04% alkali loading and associated with the dramatic pH drop, no phenolic compounds could be detected, coherently with the minimal lignin solubilization attained. Conversions to glucose obtained upon 72 h enzymatic saccharification (see Table 2) spanned the interval 7.4–42.6% theoretical glucan available in the pretreated solid. The lower limit is comparable to the conversion found when performing enzymatic saccharification of raw EFB samples. In this sense, the recalcitrance of the native feedstock has almost been left unchanged by pretreatment, the latter being hence totally ineffective. At the other extreme, yields around 40% obtained under different pretreatment conditions suggest a roughly 4-fold enhancement of the saccharification due to pretreatment. Nevertheless, compared to the yield obtained for pure cellulose (around 85%), this value was very limited and indicates that disruption has only been partial. The recent works on EFB cited in the introduction have obtained higher glucose conversions using more concentrated alkaline solution for pretreatment: for instance, the concentration of 2.89 mol L1 used by Han et al. (2011) corresponds to more than 11% NaOH on a weight basis. This imposed lower operating temperatures, in the interval 110–121 °C. Although the totally different geographical origin of the EFB samples (Asia instead of Africa)

could be an additional reason for the discrepancy with our work, it appears more likely that for EFB biomass the use of an alkaline solution with higher concentration than 2% w/v is unavoidable to reach high glucose yields. This is unfortunately problematic at the industrial level given the very well-known drawbacks associated with concentrated solutions, both environmentally and economically. The statistical analysis of the glucose yields has indicated that a second degree model is actually significant, with a p value