Process Optimization for Aqueous Ethanosolv Pretreatment of Coffee Husk Biomass Using Response Surface Methodology

ABSTRACT

This study targeted to optimize process factors using response surface methodology and determining their effect on delignification and cellulose recovery. Optimization of pretreatment conditions with desirability value of 0.828 was done at temperature (A) 150°C, contact time (B) 2 h, and liquid-to-solid ratio (LSR) (C) 15 ml/g. The sample was soaked in an aqueous ethanol solution (48% ethanol). Chemical composition was determined to be 1.87 ± 0.20, 7.03 ± 0.99, 28.05 ± 0.27, 39.29 ± 3.39, and 23.77 ± 3.91 wt% for ash, extractives, lignin, hemicellulose, and cellulose, respectively. Experimental results were validated by comparing actual value with predicted value. The model was verified as the best for improving delignification and cellulose yield. Coefficient of determination (R²), analysis of variance (ANOVA), response plots, and optimization nodes were used to examine the experimental results. The quadratic equation was used to test the model, and ANOVA was used to evaluate the model’s statistical performance and correctness. Temperature, interaction term (BC), and quadratic terms (A², B², and C²) showed significance, while contact time, LSR, and interaction terms (AC and AB) indicated insignificance for cellulose recovery. Temperature, contact time, interaction term (AB), and quadratic terms (B² and C²) showed significance, but LSR, interaction terms (AC and BC), and quadratic term (A²) showed insignificance for delignification. Pretreatment efficacy for delignification (6.0%) and cellulose recovery (151.80%) were achieved at optimal conditions. The pretreatment method is able to produce cellulose-rich coffee husk residues.

摘要

本研究旨在使用响应面法优化工艺因素，并确定其对脱木素和纤维素回收的影响. 在温度（A）150°C、接触时间（B）2小时和液固比（C）15 ml/g下进行预处理条件的优化，期望值为0.828. 将样品浸泡在乙醇水溶液（48%乙醇）中. 灰分、提取物、木质素、半纤维素和纤维素的化学组成分别为1.87 ± 0.20、7.03 ± 0.99、28.05 ± 0.27、39.29 ± 3.39和23.77 ± 3.91 wt%. 通过将实际值与预测值进行比较，验证了实验结果. 经验证，该模型是提高脱木素和纤维素收率的最佳模型. 决定系数（R2）、方差分析（ANOVA）、响应图和优化节点用于检验实验结果. 二次方程用于检验模型，ANOVA用于评估模型的统计性能和正确性. 温度、相互作用项（BC）和二次项（A2、B2和C2）显示出显著性，而接触时间、液固比、相互作用项数（AC、AB）显示出对纤维素回收的不显著性. 温度、接触时间、相互作用项（AB）和二次项（B2和C2）对脱木素有显著影响，但液固比、相互作用项数（AC、BC）、二次项数（A2）对脱木质素无显著影响. 在最佳条件下，预处理的脱木素率（6.0%）和纤维素回收率（151.80%）达到了预期效果. 该预处理方法能够产生富含纤维素的咖啡外壳残留物.

KEYWORDS:

• Aqueous ethanosolv
• pretreatment
• response surface methodology
• delignification
• coffee husk
• central composite design

关键词:

• 含水乙醇溶剂
• 预处理
• 响应面方法
• 脱木素
• 咖啡外壳
• 中央组合设计

Introduction

Lignocellulosic biomasses are the most abundant biopolymers that are by-products of agricultural industries and offer a cheap source of sugar for fermentation processes. The production of biochemicals from lignocellulosic wastes requires pretreatment, hydrolysis, fermentation, and product separation steps. Pretreatment technologies are designed to break down the recalcitrant structure of lignocellulosic biomass and makes accessible for subsequent processes such as hydrolysis and fermentation in the extraction process for biochemical synthesis (Ethaib et al. 2015).

Cellulose and hemicellulose are a good sources of fermentable sugars that are necessary for fermentation processes; however, they are tightly adhered by lignin, so this inhibits the transformation of sugar from the lignocellulosic biomass to hydrolysable sugars. Lignin is a protective cover of cellulose against microbial attack except a few organisms like rot-fungi and some bacteria that can degrade it. The structure of lignin is a highly branched and three-dimensional phenolic macromolecule polymer that cannot be fermented (Behr and Seidensticker 2020; Inan, Turkay, and Akkiris 2014). To make carbohydrate more accessible for enzymatic or microbial action to convert into fermentable sugars, pretreatment is an important tool to remove lignin from the biomass. Pretreatment has been noticed as one of the most expensive processing steps in lignocellulosic biomass for fermentable sugar conversion. Pretreatment has great potential to improve the efficiency of fermentation to make the biomass more available to microbe’s consumption and lower the cost of production for biomass-based products (Mosier et al. 2005). The main bottleneck in the commercialization of biochemical products is the development of cost-effective and eco-friendly pretreatment technology for the efficient conversion of biomass into industrially valuable biochemicals (Alexander et al. 2020; Heggset, Syverud, and Øyaas 2016). Other problems are the use of expensive solvents that would result in a number of problems: formation of inhibitors, loss of sugar, unaffordability of the costs of production, corrosion of equipment, and potential risks (Zhao, Xiaorong, and Wang 2021).

Organosolv pretreatment is one of the effective methods for lignocellulosic biomass; a large number of organic solvents (called organosolv) are included, viz. ethanol, methanol, glycerol, acetone, acetic acid, and formic acid. They have been investigated to degrade lignocellulosic biomass more acquiescent to enzymes and microbes (Park et al. 2010; Zhao, Xiaorong, and Wang 2021). The advantage of the organosolv pretreatment method is the ease with which organic solvents can be recovered by distillation and recycled for pretreatment. Particularly low-molecular-weight aliphatic alcohols are the most common solvents used for organosolvic pretreatment. Normal primary alcohols have proved to be better delignification agents than secondary or tertiary alcohols although mixtures of n-butyl alcohol–water appeared to be the most efficient in removing lignin from woody biomass (Rabelo et al. 2023; Zhao, Xiaorong, and Wang 2021). However, methanol and ethanol seem to be the most favored alcohols for alcohol-based organosolv pretreatment because of their low cost and ease of recovery. Pretreatment of lignocellulosic biomass using ethanol is safer than methanol because of less toxicity and can be easily recovered through distillation (Brosse, Hazwan Hussin, and Abdul Rahim 2017; Kim et al. 2011). On the other hand, some polyhydric alcohols are also employed for pretreating biomass under atmospheric pressure with or without catalysts. In order to alleviate the corrosion caused by mineral acids in the acid-catalyzed ethanol pretreatment, acetic acid was used as a catalyst instead of mineral acids. The mineral acid-free ethanol organosolvic pretreatment was found to be very helpful for increasing the enzymatic digestibility of biomass (Brosse, Hazwan Hussin, and Abdul Rahim 2017; Kim et al. 2011; Zhao, Xiaorong, and Wang 2021). Ethanol organosolv pretreatment method enhance the recovery of cellulose (glucose) by improving the digestibility of biomass; it also produces high-purity lignin, which can be processed into other valuable products. Several organosolv pretreatment studies were carried out with Miscanthus, pitch pine, wheat straw, olive tree trimmings, and other lignocellulosic biomass; however, aqueous ethanosolv pretreatment has not been sufficiently studied on coffee husk biomass pretreatment for cellulose recovery in bioconversion process (Goh et al. 2011).

Response surface methodology (RSM) is recognized as one of the best strategies for most researchers, for the improvement, development, and optimization of technological and chemical processes. It is a mixture of statistical and mathematical protocols, and this technique can also be used for analysis of specific problems where different process variables influence the response (Alexander et al. 2020; Karabaş and Boran 2019). The RSM was employed for this study to optimize the process variables such as temperature, contact time, and solid-to-liquid ratio for delignification and cellulose recovery of the coffee husk biomass.

To the best of our knowledge, optimization of process variables (temperature, time, and liquid-to-solid ratio [LSR]) on coffee husk biomass has not yet been studied in any of the previous studies. So far, studies have been focused on the optimization of process variables on different biomasses rather than coffee husk biomass. In this study, coffee husk biomass was treated by employing the aqueous ethanol organosolv method, and optimization of process factors could be done. Therefore, the main objective of this study was to optimize process variables to enhance cellulose yield for fermentable sugar production; furthermore, the effect of variables on chemical compositions of coffee husk was investigated by using RSM.

Materials and method

Sample collection

Coffee husk samples were collected from Zege (Bahr Dar, Ethiopia) local coffee market and transported to the research laboratory. The coffee husk was obtained from a dry processing. A disk miller (FRITSC, D-55743, Idar-Oberstein, Germany) was used to reduce the particle size of coffee husk sample. A sieve size of 2.36 mm and 1.70 mm (Retsch, AS200 Control, Germany) was used to have an average particle size of 2.03 mm. The sample was washed by tap water to remove any dirt and foreign materials and then dried in an oven at 70°C until reached constant weight, and it was packed by a high-density polyethylene (HDPE) bag and stored in a desiccator until further analysis (Kandasamy et al. 2016).

Pretreatment of coffee husk biomass

The pretreatment of coffee husk was done using stainless-steel reactor (Amar Equipments PVT. LTD Mumbia-400070, India). The coffee husk sample (50 g) was soaked in an aqueous ethanol solution (96% ethanol in distilled water at a ratio of 1:1, v/v) (Chu et al. 2021) and then added to the reactor. The pretreatment conditions were generated from design of experiment (DoE-V12) software (Table 1). Pretreatment was done at a reaction temperature (110–150°C), contact time (1–2 h) (Chu et al. 2021), and LSR of 10:1–15:1 ml/g with a continuous agitation (200 rpm) (Park et al. 2010; Zhao, Xiaorong, and Wang 2021). Finally, the reactor was cooled down immediately in an auto cooler, and the mixture was taken out from the reactor and further cooled down to room temperature. The solid fraction was separated from the liquid fraction with vacuum filtration, and the solid residue was washed using 800 ml distilled water. The washed solid residue was dried in an oven at 70°C until constant weight (Parchami, Agnihotri, and Taherzadeh 2022) and kept in a dry HDPE plastic bag for chemical composition analysis and further study.

Table 1. Factors and their coded levels used for optimization in CCD.

Variables	Unit	Coding	Coded level
			-α	−1	0	+1	+α
Temperature	°C	A	96.4	110	130	150	163.6
Contact time	h	B	0.7	1	1.5	2	2.3
Liquid/solid ratio	ml/g	C	8.3	10	12.5	15	16.70

Experimental design

Process condition optimization for delignification and cellulose recovery was done using the RSM of central composite design (CCD) at a rotatable α value of 1.68179 with six center points. The independent variables were examined at low and high levels of –α and +α respectively. Temperature (110–150°C), contact time (1–2 h), and LSR (10–15 ml/g) were the independent variables of the process. Quadratic equation was developed to fit the experimental data and to evaluate the effect of each independent variable on the responses; the optimum process condition was identified by the following general model terms:

(1) $\mathrm{y}=\beta \mathrm{o}\hspace{0.17em}+\hspace{0.17em}\mathrm{\Sigma }{\beta }_{\mathrm{i}}{\mathrm{x}}_{\mathrm{i}}+\hspace{0.17em}\mathrm{\Sigma }{\beta }_{\mathrm{j}}{\mathrm{x}}_{\mathrm{j}}+\mathrm{\Sigma }{\beta }_{\mathrm{i}\mathrm{i}}{{\mathrm{x}}_{\mathrm{i}\mathrm{i}}}^2+\mathrm{\Sigma }{\beta }_{\mathrm{j}\mathrm{j}}{x_{\mathrm{j}\mathrm{j}}}^2+\hspace{0.17em}\mathrm{\Sigma }{\beta }_{\mathrm{i}\mathrm{j}}{\mathrm{x}}_{\mathrm{i}}{\mathrm{x}}_{\mathrm{j}}+\epsilon$(1)

where y is the response function, x_i and x_j refer to the independent variables, β₀ is the model intercept term, β_i and β_j are linear effect coefficient terms, β_ij is different interaction coefficients between the input factors x_i and x_j, β_ii and β_jj are coefficients of quadratic terms, x_ii² and x_jj² are the quadratic terms, and ε is the error of the model. For this study, the independent variables are coded as A, B, and C, and thus, the equation could be described as follows:

(2) $\mathrm{y}\hspace{0.17em}=\hspace{0.17em}\beta \mathrm{o}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{a}}\mathrm{A}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{b}}\mathrm{B}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{c}}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{a}\mathrm{b}}\mathrm{A}\mathrm{B}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{a}\mathrm{c}}\mathrm{A}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{b}\mathrm{c}}\mathrm{B}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}{\beta }_{\mathrm{a}}{\mathrm{A}}^2+\hspace{0.17em}{\beta }_{\mathrm{b}}{\mathrm{B}}^2+\hspace{0.17em}{\beta }_{\mathrm{c}}{\mathrm{C}}^2+\epsilon$(2)

Each of the actual variables were investigated at five coded levels (−α, −1, 0, +1, and +α) (Table 1).

The coefficient of determination (R²), analysis of variance (ANOVA), response plots, and optimization were used to examine the experimental results. The experimental design matrix and the average of replicates are shown in Table 2.

Table 2. The completed experimental design matrix of CCD runs and corresponding results.

No.	Experimental variables			Experimental responses
	A: Temperature	B: Contact time	C: Liquid/solid	Lignin (%)		Hemicellulose (%)		Cellulose (%)
	(°C)	(hr.)	(mL/g)	Actual	Predicted	Actual	Predicted	Actual	Predicted
1	150.0	1.0	10.0	29.76	29.94	41.48	39.96	23.31	22.19
2	150.0	2.0	10.0	26.56	27.25	46.11	45.33	21.59	17.95
3	163.6	1.5	12.5	27.88	27.62	38.50	40.70	22.05	24.38
4	130.0	1.5	16.7	29.02	28.47	38.47	42.41	20.83	20.89
5	110.0	1.0	10.0	26.37	27.11	46.78	44.75	21.76	19.94
6	150.0	2.0	15.0	26.76	27.14	41.23	39.50	27.44	26.41
7	96.4	1.5	12.5	26.52	25.21	47.16	50.27	12.96	14.66
8	110.0	2.0	15.0	26.17	27.10	48.33	46.09	18.83	17.10
9	150.0	1.0	15.0	29.62	29.72	45.38	42.66	20.94	20.50
10	130.0	1.5	8.3	29.48	28.46	43.44	44.82	14.00	17.98
11	130.0	1.5	12.5	25.85	25.94	53.54	53.24	13.64	13.41
12	110.0	2.0	10.0	25.84	26.86	52.69	51.65	14.36	11.95
13	130.0	2.3	12.5	28.25	26.99	46.44	48.07	11.98	15.85
14	110.0	1.0	15.0	26.81	27.23	50.70	47.72	14.14	14.93
15	130.0	0.7	12.5	29.69	29.37	41.24	44.92	17.42	17.59
Control	0	0	0	29.84	-	37.86	-	9.44	-

Characterization of coffee husk biomass

Chemical composition analysis of coffee husk

The major chemical components of untreated and pretreated coffee husk, such as ash, extractives, lignin (acid soluble and acid insoluble), hemicelluloses, and cellulose, were determined using the National Renewable Energy Laboratory’s (NREL) standard protocols (Sluiter et al. 2005, 2008, 2012; Lin et al., 2010).

Hemicellulose determination

Extractive free coffee husk biomass (1.0 g) was transferred into a 250 ml Erlenmeyer flask with 150 ml NaOH solution (0.5 M). The mixture was heated at a reaction temperature of 80 $\pm 3$ °C for 3.5 h; after heating, it was cooled to room temperature and then washed with distilled water under vacuum filtration until the solution’s pH reached neutral. The residue was dried at 105°C, cooled in a desiccator, and weighed repeatedly until it reached a constant weight (Ayeni et al. 2013; Lin et al. 2010). Then, the hemicellulose content was calculated by taking the weight difference before and after pretreatment.

(3) $Hemicellulose\left(\mathrm{\%}\right)=\frac{W_o-W_r}{W_o}\ast 100$(3)

where W_O = weight before pretreatment and W_r = weight of the residue after pretreatment.

Lignin determination

Extractive-free coffee husk was used to determine acid-insoluble and acid-soluble lignin content. Total lignin was determined by summing up the acid-insoluble and acid-soluble lignin according to NREL/TP-510-42618 protocols. The coffee husk sample of 0.3 g was weighed using an analytical digital balance (FA2104, China) and added to a 125 ml conical flask with 3 ml of H₂SO₄ (72%). Acid hydrolysis was done to occur by keeping the sample in digital water bath (XMTD- 204, China) at a temperature of 30°C for 1 h with stirring of every 10 -min interval. After 1 h acid hydrolysis, 84 ml of distilled water was added to each flask that brings the total volume to 87 ml and autoclaved for 1 h at 121°C/103.4 kPa.

After the second weak acid hydrolysis step, the product was cooled to room temperature. The solid and liquid part was separated by vacuum filtration. The filtrate (liquid) part was used for acid-soluble lignin (ASL) determination using UV/VIS spectrometer (PerkinElmer Lambda 850, USA). The absorbance of liquid fraction was measured at a recommended wave length of 320 nm to determine ASL. The solid residue was used to determine acid-insoluble lignin (AIL) by drying the residue at 105°C until it reached constant weight and then calculate the acid-insoluble residue (AIR). The residue was used for ash determination by incinerating at 600°C for 4 h in a muffle furnace (Nabertherm GmbH, Bahnhafstr 20, 28865, Germany). AIL was calculated from the difference of AIR and ash of the residue (Sluiter et al. 2005, 2008, 2012).

(4) $AIL\left(\mathrm{\%}\right)=\frac{W_{AIR}-W_{ash}}{ODW}\ast 100$(4)

where AIL = acid-insoluble lignin, W_AIR = weight of acid-insoluble residue,

W_ash = weight of ash of the acid-insoluble residue, ODW = oven dry weight

(5) $ASL\left(\mathrm{\%}\right)=\frac{U{V}_{absorbance}\ast V_{filtrate}\ast D}{e\ast OD{W}_{sample}\ast Pathlength}\ast 100$(5)

where ASL = acid-soluble lignin, e = absorptivity at recommended wavelength (λ = 320 nm) = 30 L/g cm. ODW = oven dry weight; Vfiltrate = volume of filtrate = 86.73 ml; D = dilution.

$D=\frac{V_{sample}+V_{dillutingsolvent}}{V_{sample}}$

Cellulose determination

The cellulose content was calculated by differences; assuming that extractives, hemicellulose, lignin, ash, and cellulose are the only components of the entire biomass (Lin et al. 2010)

(6) $Cellulose\left(\mathrm{\%}\right)=100-\mathrm{\Sigma }\left(ash\mathrm{\%}+extractives\mathrm{\%}+lignin\mathrm{\%}+hemicellulose\mathrm{\%}\right)$(6)

FT-IR Spectroscopy Analysis

Fourier transformed infrared (FT-IR) spectroscopy was used for analytical tool to qualitatively determine the functional group and structural changes in the lignocellulosic material upon pre-treatment. FT-IR spectra of untreated and pretreated coffee husk was obtained by direct transmittance using the KBr pellet technique (Shimadzu). The spectra range of 400–4,000 cm⁻¹ was used at a spectral resolution of 1 cm⁻¹ (Kandasamy et al. 2016). The spectra of FT-IR was drawn by using origin software (Origin 2022).

Brunauer–Emmett–Teller (BET) surface area analysis

The surface area, pore volume, and the pore size distribution of raw and pretreated coffee husk samples were measured using BET analyzer (Nova 4000e, USA). The BET analyzer used nitrogen as an analysis gas. The degassing time was 8 h at a temperature of 300°C. The pressure tolerance for the analysis was 0.100/0.100 (ads/des) equilibrium time of 60/60 s and the equilibrium time out of 240/240 s (ads/des). The analysis contact time was 80.9 min (Awoyale and Lokhat 2021). The BET surface area, pore volume, and pore radius were obtained by the multipoint BET and non-local density functional theory (NLDFT) methods, respectively.

Statistical analysis

All the experiments were carried out in replicates, and the analysis was done using design expert-12 and origin software (V2022) for graphical illustration.

Results and discussions

Chemical composition analysis of coffee husk

The raw (untreated) coffee husk analysis results of ash, extractives, lignin, hemicellulose, and cellulose were 4.46 ± 0.11, 18.40 ± 0.65, 29.84 ± 0.92, 37.86 ± 0.06, and 9.44 ± 0.09 wt%, respectively, as shown in Table 3. The untreated coffee husk lignin content of 29.84 wt.% is in line with the previous report of 29.55 wt.% by Veiga et al. (2017). The other study for ash (4.6 wt.%) and extractives (17.67 wt.%) reported by Baêta et al. (2017) is very similar to the results of this study. Some previous studies also reported a slight difference in the chemical composition analysis of coffee husks. For instance, de Carvalho et al. reported about cellulose (26.5 wt.%), hemicellulose (25.5 wt.%), lignin (33.5 wt.%), ash (0.2 wt.%), and extractives (6.7 wt.%) (Oliveira et al. 2018). Gouvea et al. (2009) reported cellulose (43 wt.%), hemicellulose (7 wt.%), lignin (9 wt.%), and minerals (3–7 wt.%). Veiga et al. (2017) reported as extractives (47.38 wt.%), ash (7.75 wt.%), and lignin (29.55 wt.%). Bekalo and Reinhardt (2010) reported about cellulose (24.5 wt%), hemicellulose (29.17 wt.%), lignin (23.70 wt.%), and ash (6.20 wt.%). The reason for variation in chemical composition reported by different scholars could be due to different factors such as agronomical conditions, growth environment, crop variety, harvesting season, and coffee processing method (Morales-Martínez et al. 2021).

Table.3. Chemical composition analysis of untreated coffee husk.

Chemical composition	Ash (wt.%)	Extractives (wt. %)	Lignin (wt. %)	Hemicellulose (wt. %)	Cellulose (wt. %)
Raw coffee husk	4.46 $\pm$ 0.11	18.40 $\pm$ 0.65	29.84 $\pm$ 0.92	37.86 $\pm$ 0.06	9.44 $\pm$ 0.09
Treated coffee husk	1.87 $\pm$ 0.20	7.03 $\pm$ 0.99	28.05 $\pm$ 0.27	39.29 $\pm$ 3.39	23.77 $\pm$ 3.91
*Literature	4.6	17.67	29.55	25.5	26.5

*(Baêta et al. 2017) and (Veiga et al. 2017).

At the optimized process, the chemical composition of treated coffee husk was determined as ash (1.87 ± 0.20), extractive (7.03 ± 0.99), lignin (28.05 ± 0.27), hemicellulose (39.29 ± 3.39), and cellulose (23.77 ± 3.91 wt%) as shown in Table 3. The extractive content (7.03 wt%) reduced by 2.62 times to the content of untreated coffee husk (18.40 wt.%). The percentage of lignin (28.05 wt.%) was reduced by 1.06 times from the untreated coffee husk (29.84 wt.%). The cellulose content of the treated coffee husk is 2.52 times higher than that of the untreated coffee husk (9.44 wt. %); this reveals that the aqueous ethanol pretreatment improved the cellulose percentage.

Structural changes in the coffee husk biomass

FT-IR spectroscopy analysis

FT-IR spectroscopy is used to evaluate the functional group rearrangement and structural changes of the coffee husk before and after pretreatment. The FT-IR spectra of pretreated and untreated coffee husk was displayed within the wavenumber range of 4,000–400 cm⁻¹ as shown in Figure 1.

Figure 1. FT-IR spectra of coffee husk biomass: (a) raw coffee husk and (b) pretreated coffee husk.

A broad and very strong band is observed at an approximate wavenumber of 3,436 cm⁻¹ for both untreated and pretreated coffee husks. The band is predictable to the presence of O–H stretching vibrations of aliphatic hydroxyl groups, phenol, and carboxylic acid hydroxyl groups, which are present in cellulose, hemicellulose, and lignin (Awoyale and Lokhat 2021; Cruz et al. 2013; Stuart 2004). The weak band at 2,930 cm⁻¹ and the very weak band around 2,850 cm⁻¹ are observed, and this corresponds to the methyl (–CH₃) and methylene group (–CH₂–) for the asymmetric and symmetric C – H stretching, respectively; these groups are also found in aliphatic aldehydes, one of the functional group for cellulose and hemicellulose. The presence of –CH₃ and –CH₂– groups are an indicator for the existence of the structures of cellulose and hemicellulose in untreated and pretreated coffee husks (Cruz et al. 2013; Salim, Asik, and Sani Sarjadi 2021; Stuart 2004; Tolesa, Gupta, and Jer Lee 2018).

For both untreated and pretreated coffee husks, various bands were observed in the fingerprint region; a broad and very strong band around 1,084 and 1,028 cm⁻¹ affirms for C–O–C stretching. This strong and sharp band is attributed to the ether group. The peaks observed between 1,710 and 1,028 cm⁻¹ as well as 1,035 and 1,200 cm⁻¹ attributed for the existence of hemicellulose and cellulose in the untreated and pretreated coffee husk (Awoyale and Lokhat 2021; Taleb et al. 2020). However, there was a difference in peak strength between the pretreated (strong, wide, and tiny peaks) and untreated (less strong and narrow) coffee husks.

The strong band observed between 1,710 and 1,514 cm⁻¹, around 1,633 cm-1, is related to the aliphatic ketone C=O stretching; this is a ketone functional group of hemicellulose and cellulose in the coffee husk for both untreated and pretreated (Stuart 2004). Tolesa et al. reported that a band ranging around 1,330 cm⁻¹ belongs to the characteristic of syringyl and guaiacyl units of lignin monomers (Tolesa, Gupta, and Jer Lee 2018). Hence, the band 1,330 cm⁻¹ was observed in the untreated coffee husk but not in the pretreated coffee husk; consequently,the lignin aromatic monomers are removed during pretreatment. Similar results were reported by Mukherjee et al. (2022) on the torrefied coffee husk at temperature of 200–300°C.

Brunauer–Emmett–Teller (BET) surface area analysis

The surface area, pore volume, and pore size (radius) of the investigated coffee husk sample were presented in Table 4. The BET surface area and non-local density functional theory (NLDFT) pore volume of treated coffee husk are 568.03 m²/g and 0.1400 cm³/g, respectively. The pretreated coffee husk samples showed a slight change in surface area and pore volume but not in pore size compared with untreated sample. This shows that the aqueous ethanol organosolv pretreatment method affects mainly the chemical composition of the coffee husk samples. A closely related result was reported by Mukherjee et al. (2021) on BET surface area (539 m²/g), pore volume (0.32 cm³/g), and pore size (3.2 nm) by pyrolysis pretreating the spent coffee ground for biochar. Joul et al. reported BET surface area (424.1 m²/g), pore volume (0.94 cm³/g), and pore diameter (8.90 nm) as worked on the ethanol-based pretreatment of pine wood as well as BET surface area (328.1 m²/g), pore volume (1.15 cm³/g), and pore diameter (13.97 nm) of the ethanol-based pretreatment of barley straw (Jõul et al. 2022).

Table 4. Surface area and porosity of the coffee husk.

Sample types	Surface area (m^2/g)	Pore volume (cm^3/g)	Pore radius (nm)	Category
Untreated (raw)	567.300	0.1384	1.324	Microporous
Pretreated	568.027	0.1400	1.324	Microporous

Model fittingness analysis

The CCD was applied to study the effect of factors such as temperature, contact time, and LSR on the pretreatment of coffee husk biomass. The second-order polynomial model was employed. The linear terms, interaction terms, and quadratic terms of the models were fitted to the experimental data to obtain the regression equations of each response. The second-order polynomial model was used, but cubic model was found to be aliased. Two different tests namely sequential model sum of squares (see Annex Table A1) and model summary statistics (see Annex Table A2) were employed to decide about the adequacy of various models to represent lignin, hemicellulose, and cellulose content in the pretreated coffee husk biomass.

The test result of response lignin is shown in Annex Tables A1 and A2. For the suggested quadratic models, the p value was <.05, so it is highly significant. As per sequential model sum of squares test, it could be used for next analysis. As per model summary statistics, the quadratic model was selected as the best for predication of lignin composition.

Similarly, the hemicellulose and cellulose test analysis for the fittingness of the models are shown in Annex Tables A3 & A4, and Annex Tables A5 & A6, respectively. As per the sequential model sum of squares test, the models were highly significant (p < .0001) and the quadratic model was suggested for both hemicellulose and cellulose.

The model summary statistics (see Annex Tables A4 and A6 about the quadratic model was found to have R² (0.8114), adjusted R² (0.7407), and predicted R² (0.5964) for hemicellulose and R² (0.8007), adjusted R² (0.7260), and predicted R² (0.5727) for cellulose, respectively. The difference of adjusted R² and predicted R² less than 0.2 is in a reasonable agreement for the models. Therefore, quadratic models were selected as the best for analysis of the response hemicellulose and cellulose.

Statistical analysis of the experimental design

The experimental design matrix using the three factors of temperature, contact time, and LSR with the responses of lignin, hemicellulose, and cellulose were determined based on CCD. As shown in Table 2 (section 2.3), the lignin content of the pretreated samples is in the range between 25.84% and 29.76%. The highest delignification percentage of 13.40% was observed at the pretreatment condition of 110°C, 2h, 10 ml/g. While the lowest delignification of 0.27% was observed at the pretreatment condition of 150°C, 1 h, 10 ml/g. The hemicellulose recovery was recorded in the range 38.47% and 53.54%. The cellulose yield was observed in between 11.98% and 27.44%. The lowest (11.98%) and the highest (27.44%) cellulose yield was observed at the pretreatment condition of 130°C, 2.3 h, 12.5 ml/g and at the pretreatment condition of 150°C, 2 h, 15 ml/g, respectively.

Furthermore, to analyze the effects of variables on the responses, mathematical model is generated in the form of polynomial equation (Eqns. 7, 8 & 9); this represents the experimental relationship between the responses (lignin, hemicellulose, and cellulose yields) and the variables (temperature – A, contact time – B, and LSR – C). The generated polynomial equation in terms of coded factors and the relationship between the responses and the factors are explained in the following mathematical equations.

(7) $\mathrm{L}\mathrm{i}\mathrm{g}\mathrm{n}\mathrm{i}\mathrm{n}\hspace{0.17em}\left(\mathrm{\%}\right)\hspace{0.17em}=\hspace{0.17em}45.51\hspace{0.17em}+\hspace{0.17em}0.04\mathrm{A}\hspace{0.17em}-\hspace{0.17em}3.27\mathrm{B}\hspace{0.17em}-\hspace{0.17em}3.37\mathrm{C}\hspace{0.17em}-\hspace{0.17em}0.06\mathrm{A}\mathrm{B}\hspace{0.17em}+\hspace{0.17em}0.02\mathrm{B}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}3.17{\mathrm{B}}^2+\hspace{0.17em}0.14{\mathrm{C}}^2$(7)

(8) $\begin{array}{rl}& \mathrm{H}\mathrm{e}\mathrm{m}\mathrm{i}\mathrm{c}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{u}\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{e}\hspace{0.17em}\left(\mathrm{\%}\right)\hspace{0.17em}=\hspace{0.17em}53.24\hspace{0.17em}-\hspace{0.17em}2.85\mathrm{A}\hspace{0.17em}+\hspace{0.17em}0.93\mathrm{B}\hspace{0.17em}-\hspace{0.17em}0.\hspace{0.17em}72\mathrm{C}\hspace{0.17em}-\hspace{0.17em}0.38\mathrm{A}\mathrm{B}\hspace{0.17em}\\ & -\hspace{0.17em}0.07\mathrm{A}\mathrm{C}\hspace{0.17em}-\hspace{0.17em}2.13\mathrm{B}\mathrm{C}\hspace{0.17em}-\hspace{0.17em}2.74{\mathrm{A}}^2-\hspace{0.17em}2.38{\mathrm{B}}^2-\hspace{0.17em}3.40{\mathrm{C}}^2\end{array}$(8)

(9) $\begin{array}{rl}& \mathrm{C}\mathrm{e}\mathrm{l}\mathrm{l}\mathrm{u}\mathrm{l}\mathrm{o}\mathrm{s}\mathrm{e}\hspace{0.17em}=\hspace{0.17em}13.41\hspace{0.17em}+\hspace{0.17em}2.89\mathrm{A}\hspace{0.17em}-\hspace{0.17em}0.52\mathrm{B}\hspace{0.17em}+\hspace{0.17em}0.86\mathrm{C}\hspace{0.17em}+\hspace{0.17em}0.94\mathrm{A}\mathrm{B}\hspace{0.17em}\\ & +\hspace{0.17em}0.83\mathrm{A}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}2.54\mathrm{B}\mathrm{C}\hspace{0.17em}+\hspace{0.17em}2.16{\mathrm{A}}^2+\hspace{0.17em}1.17{\mathrm{B}}^2+\hspace{0.17em}2.13{\mathrm{C}}^2\end{array}$(9)

Analysis of responses (lignin, Hemicellulose, and cellulose)

The statistical significance of the generated model terms for lignin, hemicellulose, and cellulose were evaluated using analysis of variance (ANOVA) as shown in Annex Tables A7–A9). For lignin removal, the model is significant (p < .0001) (Table A7) and the main factors such as temperature (A), contact time (B), AB interaction terms, and quadratic terms (B² and C²) are significant (p < .05). However, the model terms (C, AC, BC, and A²) are not significant (p > .1000). The value of R² (0.7866), adjusted R² (0.7066), and predicted R² of 0.5431 clearly demonstrated that the model used to represent the relationship between the factors and the response lignin were well correlated and defined. The adjusted R² of 0.7066 and predicted R² of 0.5431 difference less than 0.2 is in a reasonable agreement. Adequate precision represents the measure of signal-to-noise ratio and the ratio greater than 4 is desirable; hence, in this case, the adequate precision was found to be 10.608, and this is an indication of the fitness of the developed model; therefore, the model was used to direct the design spaces. Besides, the lower values of coefficient of variance (3.00%) precisely exhibited a high degree of precision and good agreement of reliability on the experimental results (Alexander et al. 2020).

For the response hemicellulose, the analysis of variance (see Annex Table A8) described that the model is significant (p < .0001). The linear term temperature (A), the interaction term (BC), and the quadratic terms (A², B², and C²) are highly significant, whereas the model terms B, C, AB, and AC are not significant (p > .05). The R² (0.8114), adjusted R² (0.7407), and predicted R² (0.5964) verified that the model used to represent the relationship between the factors and the response was well correlated and defined. The predicted R² (0.5964) is in a reasonable agreement with the adjusted R² (0.7407) due to their difference less than 0.2. Similarly, adequate precision analysis shows a good fitness of the developed model; therefore, the model was used to navigate the points in the design spaces.

For the response cellulose, the statistical significance of the model terms was evaluated using ANOVA table as shown in Annex Table A9. The ANOVA table described that the model is significant (p < .0001). The model terms (A, BC, A², B², and C²) are significantly (p < .05) influencing the cellulose yield in the pretreatment process, while the model terms (B, C, AB, and AC) have no significant effects (p > .05). The R² (0.8007), adjusted R² (0.7260), and predicted R² (0.5727) demonstrated that the model used to represent the relationship between the factors and the response were well defined. Similarly, adequate precision analysis (11.2144) shows a good fitness of the developed model; therefore, the model was used to navigate the points in the design spaces (Alexander et al. 2020).

Effect of temperature, contact time, and LSR on chemical composition of the coffee husk biomass

Effect of temperature, contact time, and LSR on lignin removal

As shown in Figure 2(a-c), temperature, contact time, and LSR have an effect on the content of lignin removal. At high temperature (150°C) and long contact time (2 h), the delignification process is high, whereas high temperature (150°C) and short contact time (1 h) cause an increase in lignin percentage and vice versa in the solid residue of the pretreated coffee husk biomass (Figure 2a). This might be due to the re-polymerization (condensation) of lignin fragments at short contact time (Huang, Shiyu, and Gan 2019. In the re-polymerization process, there might be the formation of ether, ester, and glycosidic linkages between lignin and polysaccharides, which leads to the development of lignin–carbohydrate complexes linkages, and this supports to increase the concentration of lignin in the pretreated solid matrix (Trajano et al. 2013). Temperature, contact time, interaction term (AB), and quadratic terms (B² and C²) showed significance, but the LSR, the interaction terms (AC and BC), and quadratic terms (A²) showed insignificance for delignification process.

Figure 2. Contour plots for the effect of factors on percentage of lignin concentration. (a) Effect of temperature (°C) versus contact time (h). (b) Effect of temperature versus liquid-to-solid ratio (ml/g). (c) Effect of contact time (h) versus liquid-to-solid ratio (ml/g).

However, significant delignification process was observed in the temperature range of 110–140°C with an extended contact time from 1.4 to 2 h. This may possibly be the result of depolymerization at the β-O-4 linkage, which is chemically the easiest to cleave. The β-O-4 linkage plays a key role in the depolymerization of lignin due to high unstability as well as the bond cleavage at the specific lignin linkages, like β-β, β-5, and preferring demethoxylation reactions into monomeric units (Abejón, Pérez-Acebo, and Clavijo 2018; Gopalakrishnan, Kim, and Ceylan 2010; Kouris et al. 2021).

Effect of temperature, contact time, and LSR on hemicellulose content

The hemicellulose content increased gradually with the rise in contact time and decreased with an increase in temperature (Figure 3a). The highest hemicellulose content was observed between 10 and 14 ml/g of the LSR, but beyond 14 ml/g the hemicellulose content decreased gradually (Figure 3b). The highest concentration of hemicellulose was found above 1.2 h of contact time versus the LSR between 10 and 14 ml/g (Figure 3c).

Figure 3. Contour plots for the effect of factors on hemicellulose content. (a) Effect of temperature (°C) versus contact time (h). (b) Effect of temperature versus liquid-to-solid ratio (ml/g). (c) Effect of contact time (h) versus liquid-to-solid ratio (ml/g).

The analysis result depicted by the contour plot (Figure 3) shows that the concentration of hemicellulose increased below the temperature of 140°C; however, at the temperatures above 140°C, the concentration of hemicellulose decreased, and this may be due to the formation of furfural and hydroxymethylfurfural (HMF), which are the hydrolysis products of pentose (C5) and hexose (C6) sugars, respectively (Parchami, Agnihotri, and Taherzadeh 2022). By increasing contact time from 1.2 to 2 h at a temperature below 140°C, the highest concentration of hemicellulose was observed. A similar result was reported by Parchami et al., who worked on the aqueous ethanosolv pretreatment of brewer’s spent grain; the hemicellulose concentration decreased above 140°C (Parchami, Agnihotri, and Taherzadeh 2022).

Effect of temperature, contact time, and LSR on cellulose content

As shown in Figure 4 (a, b, and c), the yield of cellulose increased significantly as the temperature (A) rises to 150°C (Figure 4a); temperature is a significant factor for cellulose yield, but there is no significant differences in the yield of cellulose in the contact time (B) range of 1.2 to 2 h and the solid-to-liquid ratio (C) range of 10–15 mL/g (Figure 4a,b). Similarly, the ANOVA table (Table A9) shows that temperature has a high level of significance (p < .0001), whereas the contact time and LSR (p > .05) show low levels of significance. Therefore, temperature, the interaction term (BC), and quadratic terms (A², B², and C²) showed significance for cellulose recovery, but contact time, LSR, and the interaction terms (AC and AB) indicated insignificance for cellulose recovery.

Figure 4. Contour plots for the effect of factors on cellulose content. (a) Effect of temperature (°C) versus contact time (h). (b) Effect of temperature versus liquid-to-solid ratio (ml/g). (c) Effect of contact time (h) versus liquid-to-solid ratio (ml/g).

Validation of optimal data

Process variable optimization was performed with the target goal to minimize lignin and hemicellulose contents and maximize cellulose content. The optimum process variables were obtained at temperature of 150°C, contact time of 2.0 h, and LSR of 15.0 ml/g at the desirability function value of 0.828. After experimental work, the actual percent of lignin (28.05%), hemicellulose (39.29%), and cellulose (23.77%) were obtained (Table 5). The data were validated by entering into confirmation node and verified within the 95% prediction interval (PI) for the fittingness of the optimal results. Hence, the data mean values were accurately fitted within the prediction intervals.

Table 5. Validation of the data obtained from optimal process conditions.

Response	Experimental data	Predicted mean	Predicted median	Std Dev	n	SE Pred	95% PI low	Data mean	95% PI high
Lignin	28.24	27.14	27.14	0.82	2	0.75	25.58	28.05	28.69
	27.86
Hemicellulose	36.89	39.50	39.50	2.59	2	2.37	34.61	39.29	44.39
	41.68
Cellulose	26.53	26.41	26.41	2.38	2	2.17	21.93	23.77	30.89
	21.00

Evaluation of the pretreatment method efficiency

The effectiveness of delignification and cellulose recovery of the pretreatment method were determined. The yields of hemicellulose and cellulose increased by 3.77% and 151%, respectively. The efficacy of the delignification process was determined by comparing the percentage of lignin present before and after pretreatment in the solid residue. The pretreatment method has a delignification effectiveness of 6.00%. The effectiveness of extractives removal is 61.79%. The amount of cellulose in the treated coffee husk was 3.02 times higher than in the untreated. Hence, the most significant improvement of the pretreatment was observed for the cellulose content (Figure 5).

Figure 5. Efficiency of the pretreatment methods.

At the optimal pretreatment conditions, 1.038 times of hemicellulose and 2.518 times of cellulose were improved by removing 1.064, 2.617, and 2.385 times of lignin, extractives, and ash, respectively. The lignin undertaking a repolymerization reaction during pretreatment might reduce the delignification process (Pielhop et al. 2015). During pretreatment, complete breakdown of lignin into its fragments is a challenging process because two parallel processes – the depolymerization and repolymerization of lignin fragments like the phenolic moiety and reactive groups like carbonyls – occur simultaneously (Kaur, Singh, and Kumar Arya 2022). At this condition, lignin may not be completely removed but remains in the solid residue together with polysaccharides in modified and weakened bond form (Pielhop et al. 2015). In the aqueous ethanol organosolv pretreatment, the lignin carbocation (which is known as carbonium ion) and free radical intermediates can also be generated from the lignin molecules during molecular interaction within the pretreatment matrix (Chu et al. 2021). The carbonium ion and the lignin fragment ions are highly responsible for the repolymerization (condensation) reactions of lignin in the matrix as shown in Figure 6 (Pielhop et al. 2015). According to several studies, in acidic conditions, the lignin carbocation intermediates may attack the nearby aromatic rings or condense with another aromatic ring, and this can lead to the repolymerization of insoluble lignin and/or the formation of pseudo-lignin that hinders the removal of lignin from the biomass (Chu et al. 2021). This phenomenon could explain the delignification efficiency of the pretreatment (6.00%). The pretreatment is more effective for cellulose yield enhancement (151%). Similarly, Zhang et al. (2020) reported 5.19% delignification while conducting the experiments on wheat straw with an ethanol concentration of 78%.

Figure 6. Schematic representation for reactions of lignin. (a) Cleavage of β-O-4 bond and formation of carbocation; (b) elimination of β-proton giving rise to an enol ether; (c) acid hydrolysis of enol ether; (d) repolymerization reaction. Adapted from Pielhop et al. (2015).

In this study, cellulose-rich solid residues were successfully obtained from coffee husks under optimized pretreatment conditions; however, the presence of lignin impedes enzymatic hydrolysis by physically blocking enzyme accesses as well as nonproductive binding with enzymes (Jia et al. 2020). It is possible to reduce the impact of lignin on nonproductive binding by using lignin-blocking agents (like proteins such as bovine serum albumin (BSA), CaCl₂, Tween 80, and activated carbon and surfactants) in enzymatic hydrolysis to protect the enzyme-binding sites on lignin (Qin et al. 2016; Qin, Clarke, and Kecheng 2014).

Conclusion

Coffee husk is regarded as a promising lignocellulosic feedstock source for a variety of biochemical production processes. The raw coffee husk contains high amounts of extractives and lignin compared with the minerals and cellulose content. The optimized pretreatment condition may have reduced the lignin content and significantly improved the cellulose content as well as hemicellulose. The coffee husk pretreated at the optimal process conditions (150°C, 2 h, 15 ml/g) showed an increase in pore volume and surface area, and this revealed that structural modification of the coffee husk was done. The CCD of RSM was used to model and validate the optimization of the process conditions (temperature, contact time, and LSR). Under optimized process condition, the cellulose content was improved from 9.44% to 23.77%, and the lignin content was reduced from 29.84% to 28.05%, respectively; the pretreatment efficiency for cellulose recovery and delignification was achieved to 151.8% and 6%, respectively. Therefore, the concentration of cellulose was significantly improved by the aqueous ethanosolv pretreatment method; however, the effectiveness of the delignification process might be influenced by repolymerization of lignin fragments. The pretreatment method also removed a large percentage of extractives (61.79%), indicating that the coffee husk contained a lot of extractives. In general, cellulose-rich solid residues from coffee husks were successfully obtained. The cellulose-rich solid residue of coffee husks can be used for bioconversion process to contribute into circular economy concept. The aqueous ethanol organosolv pretreatment method could be considered eco-friendly because it does not use any noxious chemicals.

Highlights

• Aqueous ethanosolv pretreatment conditions have been optimized.

• Highest cellulose yield (151%) was obtained at optimal conditions.

• Delignification of coffee husk was achieved at optimal process conditions.

• The effect of temperature, contact time, and LSR was investigated

• The effectiveness of the aqueous ethanosolv pretreatment method was investigated

Author’s contribution

All authors contributed equally to this manuscript.

Ethical approval

This research work does not contain any studies performed on animals/humans subjects by any of the authors.

Consent

This research work does not involve information about any individual, e.g., case studies, survey, interview, etc., and no consent was needed.

Acknowledgements

The authors would like to acknowledge the Bahir Dar University Institute of Technology for providing the necessary laboratory facilities and resources for this study.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

No funding received for this study, but the Bahr Dar institute of Technology supports the required laboratory facilities and chemicals.

Appendix

Table A1. Sequential model sum of squares for the response lignin.

Source	Sum of squares	df	Mean square	F-value	p-value	Remark
Mean vs. total	25,578.48	1	25,578.48
Linear vs. mean	27.70	3	9.23	5.71	0.0032
2FI vs. linear	6.12	3	2.04	1.30	0.2947
Quadratic vs. 2FI	26.10	3	8.70	12.85	<.0001	Suggested
Cubic vs. quadratic	3.83	4	0.9565	1.54	0.2290	Aliased
Residual	12.43	20	0.6214
Total	25,654.64	34	754.55

Table A2. Model summary statistics for the response lignin.

Source	Std. Dev.	R^2	Adjusted R^2	Predicted R^2	PRESS	Remark
Linear	1.27	0.3637	0.3000	0.2041	60.62
2FI	1.25	0.4440	0.3204	0.2877	54.25
Quadratic	0.8229	0.7866	0.7066	0.5431	34.80	Suggested
Cubic	0.7883	0.8368	0.7308	0.4406	42.61	Aliased

Table A3. Sequential model sum of squares for the response hemicellulose.

Source	Sum of squares	df	Mean square	F-value	p-value	Remark
Mean vs. total	73,157.96	1	73,157.96
Linear vs. mean	259.07	3	86.36	4.34	0.0118
2FI vs. linear	75.17	3	25.06	1.30	0.2961
Quadratic vs. 2FI	360.62	3	120.21	17.86	<.0001	Suggested
Cubic vs. quadratic	19.90	4	4.98	0.7027	0.5993	Aliased
Residual	141.60	20	7.08
Total	74,014.32	34	2,176.89

Table A4. Model summary statistics for the response hemicellulose.

Source	Std. Dev.	R^2	Adjusted R^2	Predicted R^2	PRESS	Remark
Linear	4.46	0.3025	0.2328	0.1666	713.68
2FI	4.40	0.3903	0.2548	0.2762	619.86
Quadratic	2.59	0.8114	0.7407	0.5964	345.62	Suggested
Cubic	2.66	0.8346	0.7272	0.4330	485.54	Aliased

Table A9. ANOVA table for quadratic model of the response cellulose.

Source	Sum of squares	df	Mean square	F-value	p-value	Remark
Model	545.35	9	60.59	10.72	<.0001	Significant
A-Temp	228.23	1	228.23	40.37	<.0001
B-Contact time	7.34	1	7.34	1.30	.2658
C-L/S	20.45	1	20.45	3.62	.0693
AB	14.03	1	14.03	2.48	.1284
AC	10.99	1	10.99	1.94	.1761
BC	103.12	1	103.12	18.24	.0003
A^2	105.32	1	105.32	18.63	.0002
B^2	30.84	1	30.84	5.46	.0282
C^2	102.25	1	102.25	18.08	.0003
Residual	135.70	24	5.65
Lack of fit	135.70	5	27.14
Pure error	0.0000	19	0.0000
Cor total	681.05	33
R^2 = 0.8007		Adjusted R^2 = 0.7260			C.V. % = 13.36
Predicted R^2 = 0.5727		Adeq precision = 11.2144

Table A5. Sequential model sum of squares for the response cellulose.

Source	Sum of Squares	df	Mean Square	F-value	p-value	Remark
Mean vs Total	10767.58	1	10767.58
Linear vs Mean	256.02	3	85.34	6.02	.0024
2FI vs Linear	128.14	3	42.71	3.88	.0198
Quadratic vs 2FI	161.19	3	53.73	9.50	.0003	Suggested
Cubic vs Quadratic	53.98	4	13.50	3.30	.0312	Aliased
Residual	81.72	20	4.09
Total	11448.63	34	336.72

Table A6. Model summary statistics for the response cellulose.

Source	Std. Dev.	R^2	Adjusted R^2	Predicted R^2	PRESS	Remark
Linear	3.76	0.3759	0.3135	0.2035	542.47
2FI	3.32	0.5641	0.4672	0.3008	476.22
Quadratic	2.38	0.8007	0.7260	0.5727	291.04	Suggested
Cubic	2.02	0.8800	0.8020	0.5886	280.20	Aliased

Table A7. ANOVA for quadratic model of the response lignin of coffee husk biomass.

Source	Sum of squares	df	Mean square	F-value	p-value
Model	59.92	9	6.66	9.83	<.0001	Significant
A-Temperature	14.06	1	14.06	20.76	.0001
B-Contact time	13.64	1	13.64	20.14	.0002
C-Liquid/solid	0.0005	1	0.0005	0.0007	.9793
AB	5.98	1	5.98	8.83	.0066
AC	0.1260	1	0.1260	0.1861	.6700
BC	0.0132	1	0.0132	0.0195	.8900
A^2	0.6313	1	0.6313	0.9322	.3439
B^2	14.18	1	14.18	20.94	.0001
C^2	17.94	1	17.94	26.50	<.00001
Residual	16.25	24	0.6772
Lack of fit	16.25	5	3.25
Pure error	0.0000	19	0.0000
Cor total	76.17	33
R^2 = 0.7866		Adjusted R^2 = 0.7066			C.V (%) = 3.00
Predicted R^2 = 0.5431		Adeq precision = 10.6083

Table A8. ANOVA table for quadratic model of the response hemicellulose.

Source	Sum of squares	df	Mean square	F-value	p-value	Remark
Model	694.86	9	77.21	11.47	<.0001	Significant
A-Temp	221.20	1	221.20	32.87	<.0001
B-Contact time	23.86	1	23.86	3.55	.0719
C-L/S	14.00	1	14.00	2.08	.1621
AB	2.34	1	2.34	0.3479	.5608
AC	0.0729	1	0.0729	0.0108	0.9180
BC	72.76	1	72.76	10.81	.0031
A^2	169.30	1	169.30	25.16	<.0001
B^2	128.05	1	128.05	19.03	.0002
C^2	261.13	1	261.13	38.80	<.0001
Residual	161.50	24	6.73
Lack of fit	161.50	5	32.30
Pure error	0.0000	19	0.0000
Cor total	856.36	33
R^2 = .8114		Adjusted R^2 = .7407			C.V. % = 5.59
Predicted R^2 = .5964		Adeq precision = 9.7638