Deep Eutectic Solvent Pretreatment for Cellulose Enrichment from Coffee Waste

Abstract

The treatment of coffee silverskin (CSS), a waste stream from the coffee industry, through deep eutectic solvents (DES) was studied. Choline chloride-based DES mixtures were checked by using a lactic acid (LA) hydrogen bond donor. The effect of time, temperature and biomass:DES ratio on the dissolution of lignin and hemicellulose was analyzed. Maximum hemicellulose reduction was from 28.15% to 3.37% at 60 °C and 1 h and maximum lignin reduction was from 33.1% to 20.03% at 90 °C and 5 h. Additionally, it was found that the optimum operating conditions obtained were 3 h, 120 °C and biomass:DES ratio (1:10), which led to 8.02%, 29.31% and 36.48% hemicellulose, lignin and cellulose content, respectively. Consequently, DES pretreatment can be considered a promising approach to obtain cellulose-enriched coffee silverskin solids under lower-temperature conditions than many conventional lignocellulosic biomass pretreatments. However, further downstream validation is required to assess their suitability for paper, textile, or food-packaging-related applications.

1. Introduction

Coffee is one of the most valuable primary products in world trade and, after water, is the second most popular beverage worldwide and the second-largest commodity in the stock exchange after oil [1]. According to the last report of the International Coffee Organization (ICO), in 2024, around 10.5 Mt of coffee was consumed worldwide, increasing by 1.4% from 2023, whereas the EU countries accounted for 30.59% of global coffee consumption [2]. As a result, coffee production increased by 5.2% from 2023 to 2024. This increase in production would lead to an increase in residue generation; in fact, during coffee manufacture 50% of the coffee is discarded since only the inner part of the bean is used [3]. The parts that are discarded as residues correspond to the outer layers of the coffee bean, namely: parchment, mucilage, pulp, skin and silverskin [4]. Regarding these residues, coffee silverskin (CSS) is attracting attention due to several factors. It is accessible and available, since from 120 kg of processed coffee about 2.5 kg of CSS is produced, representing about 5–6% of the coffee bean [5]—the biggest by-product from the coffee-roasting industry disposed as industrial waste [6]. However, traditional disposal of this residue leads to soil and groundwater contamination [7] and also can act as an inhibitor of plant growth due to some of its components, such as tannins, caffeine and chlorogenic acids [8]. Therefore, a green alternative for CSS management is needed. Traditionally, the treatment of lignocellulosic residues in general and in coffee, in particular, has been carried out by using extraction, hydrolysis, or a combination of both [9,10,11].

Hydrolysis breaks down hemicellulose and cellulose chains for the obtention of monosaccharides [12] so they can be used as substrates for biofuel or bioplastics production [13]. However, hydrolysis can generate inhibitors like acetic acid, formic acid, furfural or hidroxymethyl furfural which hinder microorganism activity; therefore, a purification step is needed [14,15].

Extraction, on the other hand, is able to obtain high value-added products without degradation of biomass’ macrocomponents. This process has been traditionally performed using organic compounds which can be potentially hazardous and with high energy demand [16]. Therefore, biomass processing has focused on using deep eutectic solvents (DES) as a greener alternative. These green solvents present some benefits like environmental degradability, low production costs, and simple synthesis routes [17]. Additionally, the use of DES allows for working at low temperatures and low energy consumption, and results in low- or moderate-severity factors in comparison with other biomass pretreatments. Some studies reveal that working at low temperatures avoids degradation of macrocomponents [18]. In previous research, the authors observed that DES causes cleavage of lignin ether and ester bonds while causing almost negligible dissolution of cellulose, making biomass suitable for further processing [19]. This study demonstrated that DES increased Fock’s reactivity of dissolving pulp, from 78.86% to 93.70%, without dissolution of cellulose. In addition, DES is much easier to reuse [20,21,22]. After testing different DES systems, the authors of this paper showed promising results using chlorine chloride (ChCl) as HBA and lactic acid (LA) as HBD when used as pretreatment of dissolving pulp. This behavior goes in agreement with the literature: on one hand ChCl-based DES provides chlorine anions that form new hydrogen bonds while destabilizing the lignin-carbohydrate complex, and also the hydrogen bonds between lignin and cellulose, resulting in a more porous structure [23]. With respect to the HBD, DES performance depends on the HBD acidity, polarity, and hydrogen-bonding capacity [17]. Among the DES systems assayed for biomass processing, there are acid-based, alkali-based, and neutral DES categories. In this case, since the goal is to remove hemicellulose and lignin, lactic acid was selected as the best choice. High-polar acid-based DES facilitates deeper penetration into the biomass matrix [24]. Acid-based DES possesses higher proton-donating ability for cleavage of the glycosidic bonds of the xylan chains [25].

Although ChCl:LA DES has been widely investigated for several lignocellulosic feedstocks, its application for coffee silverskin remains comparatively less explored. Other authors have used differently formulated DES for CSS treatment [26,27,28]. Nevertheless, this is the first research testing a ChCl:LA DES mixture with CSS wastes. Additionally, CSS differs from conventional lignocellulosic residues due to its origin as a roasting by-product, its high availability in the coffee industry, and its particular composition, including relevant amounts of hemicellulose, lignin, extractives, and bioactive compounds. Therefore, evaluating whether ChCl:LA can selectively enrich cellulose in CSS is relevant for the valorization of this specific residue.

Therefore, in this article, the fractionation of a by-product side stream from coffee production using DES extraction will be studied. The effect of DES fraction will be confirmed by analyzing the composition of biomass: cellulose, lignin and hemicellulose, and, as independent variables, time, temperature, and DES:biomass ratio were chosen.

2. Materials and Methods

2.1. Materials

Coffee silverskin (CSS) was supplied by Cafetería Primos de Origen, a coffee shop located in Santander, Spain, which imports 100% Arabica specialty green coffee beans. The CSS was ground in a cutting mill, and the fraction with particle sizes below 0.4 mm was selected for the experiments. Choline chloride (ChCl, 99%) was purchased from Thermo Scientific (Waltham, MA, USA), while lactic acid (88%) and sulfuric acid (95%) were obtained from Fisher Scientific (Waltham, MA, USA). Acetylacetone was purchased from Sigma-Aldrich (St. Louis, MO, USA). Acetic acid, hydrochloric acid (37%), sodium chlorite (25% w/w), and 1,4-dioxane (99.5%) were supplied by Panreac (Barcelona, Spain).

2.2. DES Preparation and Experimental Work

DES choline chloride-based systems were tested by using lactic acid (LA) hydrogen bond donor. The mole ratio for each DES system was 1:9 for lactic acid (ChCl:LA) [19]. First, the DES reactants were weighed, mixed, heated at 80 °C, and magnetically stirred until a clear liquid was formed. Then, the dry CSS samples were milled. Once the DES treatment was finished, CS samples were filtered (sieves of 160 µm) and washed using tap water until neutral pH.

Initially, a three-level-three-variable Box–Behnken design (BBD) factorial experimental design was performed using the statistical software Statgraphics Centurion XVI.II. The independent variables considered were temperature, time and biomass:DES ratio and the response variables considered were cellulose, lignin and hemicellulose content. This resulted in 15 experiments with 3 center experiments used to determine the experimental error.

Table 1. Factors and levels of the BBD-RSM design.

Experiment	Temperature	Time	Biomass:DES Ratio
Number	°C	min	w/w
1	−1	−1	0
2	1	−1	0
3	−1	1	0
4	1	1	0
5	−1	0	−1
6	1	0	−1
7	−1	0	1
8	1	0	1
9	0	−1	−1
10	0	1	−1
11	0	−1	1
12	0	1	1
13	0	0	0
14	0	0	0
15	0	0	0
Factors	Levels
	−1	0	1
X1:Time (h)	1	3	5
X2:T (°C)	60	90	120
X3:Ratio (w/w)	1:10	1:20	1:30

shows the levels and the conditions of the experiments.

All the experimental conditions are shown in Table 1. Response variables analyzed at the end of the CSS treatment were Seifert cellulose [29], lignin [30], and holocellulose [31,32].

The upper temperature level was set at 120 °C to promote biomass fractionation while remaining below the decomposition temperature reported for ChCl:LA DES. Sazali et al. [33] reported decomposition temperatures of 179.73–192.14 °C for ChCl:LA DES, depending on the molar ratio. Therefore, 120 °C was selected as the highest temperature in the experimental design to avoid conditions close to severe DES thermal decomposition.

The results obtained from the experiments were analyzed with the software Statgraphics Centurion and a statistical model with response curves was obtained. The model is described by a second-order polynomial:

$$Y=b_0+\sum b_i·X_i+\sum b_{ij}·X_i^2+\sum b_{ij}·X_i·X_j$$

where Y is the dependent codified as: Y₁ cellulose (%), Y₂ lignin (%) and Y₃ hemicellulose (%) and X_i and X_j are the independent variables codified as shown in

Table 2. Experimental conditions and final composition of treated CSS.

			ChCl:LA
Temperature (°C)	Time (h)	Ratio Biomass:DES	Cellulose (%)	Lignin (%)	Hemicellulose (%)
0	0	0	17.59	33.01	28.15
60	1	1:20	27.80	23.70	29.26
120	1	1:20	31.95	27.42	32.36
60	5	1:20	37.73	26.10	20.97
120	5	1:20	48.35	25.82	21.14
60	3	1:10	32.17	28.50	3.37
120	3	1:10	35.81	29.07	7.22
60	3	1:30	28.74	27.05	7.06
120	3	1:30	35.49	22.76	9.01
90	1	1:10	28.72	24.23	11.87
90	5	1:10	34.07	26.42	9.25
90	1	1:30	32.41	28.08	10.63
90	5	1:30	25.63	20.03	7.04
90	3	1:20	34.82	30.49	25.18
90	3	1:20	34.83	30.61	24.87
90	3	1:20	35.05	32.57	23.45

. Finally, the analysis of the variance (ANOVA) was performed using the same software described before to assess the relationships of the variables.

2.3. Chemical Characterization

2.3.1. Seifert Cellulose

The cellulose content was determined using the Seifert procedure [34], which is directly applicable to samples with high lignin content, as in the coffee silverskin studied in this paper. Two grams of free-extractive CSS were mixed with 12 mL of acetylacetone, 4 mL of 1,4-dioxane and 3 mL of commercial HCl (37% w/w). Then, the mixture was heated and boiled for 35 min. Once the flask was cooled, the contents were filtered using a porous cresol, collecting the entire flask’s contents with methanol, then washed with distilled water.

2.3.2. Holocellulose

Holocellulose, which represents the total carbohydrate content, the sum of cellulose and hemicelluloses, was measured by means of the Wise chlorite method [31,32] but using the modified version of Cruz et al. [30], which uses lower quantities of biomass. A total of 0.25 g of the sample was mixed with 160 mL of deionized water. The mixture was heated to 80 °C in a water bath, and 10 drops of acetic were added.

Acid and 5 mL of sodium chlorite were added. Each hour, the same amount of each reactant was added to a total time of 3 h. The process was repeated until the product became white.

2.3.3. Lignin

Lignin was measured according to the method shown by Cruz et al. [30], which is based on a two-step hydrolysis. Initially, 0.35 g of the sample was weighed and mixed with 10 mL of 72% sulfuric acid, then homogenized by vortexing. The mixture was heated at 30 °C for an hour with vortex-assisted homogenization every 10 min. Then, the solution was transferred to a 250 mL ISO flask, and 144 mL of deionized water was added to obtain a 4% sulfuric acid concentration. The solution was sealed and heated at 120 °C for an hour. Finally, the mixture was left to cool, then vacuum-filtered through 1.2 µm glass-fiber filters. The solid was dried at 105 °C for 24 h and weighed.

3. Results and Discussion

The chemical characterization of CSS resulted in 17.6 ± 2.64% of cellulose, 33.018 ± 2.11% of lignin, and 45.7 ± 4.09% of holocellulose, which is the sum of cellulose and hemicellulose. Thus, the hemicellulose content of the CSS samples is approximately 30.4%. Table 2 presents the experimental conditions and results for the DES treatment of CSS.

The results presented in Table 2 show that DES treatment reduces lignin and hemicellulose content in CSS across all experimental conditions. As a result, since DES do not dissolve crystalline cellulose, the content of DES increases as the other two macrocomponents are removed. This effect has been reported in the literature on biomass delignification [35,36]. From Table 1, it can be concluded that hemicelluloses are removed to a greater extent, especially at reaction times longer than 3 h, with a maximum hemicellulose reduction of 75% at a 5 h reaction time, at which the minimum lignin content was reached (20.03%). Similar hemicellulose removal was reported in the literature, reaching removals of 74.19% from corn cons with ChCl/LA systems.

High selectivity is related to the hydrolytic action of protons on the glycosidic linkages [37]. Other authors also noticed the DES selectivity towards hemicellulose [38,39]. The amorphous nature of hemicellulose makes it more accessible to the reactants, thus resulting in higher elimination rates [40]. By contrast, lignin is recalcitrant, making it less reactive, so cleavage of the β-O-4 bond occurs at a lower rate [41]. Additionally, the results show an increase in cellulose content due to two main factors: the elimination of other macrocomponents present in lignocellulosic biomass and the negligible dissolution of cellulose in DES, as reported in the literature [42,43]. This behavior indicates a preferential removal of hemicellulose over lignin during ChCl:LA pretreatment. This selectivity is further discussed below considering DES acidity, hemicellulose accessibility, lignin recalcitrance, solvent–biomass interactions, and cellulose retention in the treated solid.

To evaluate the impact of operating conditions on the composition of silverskin, an ANOVA was conducted using Statgraphics Centurion. Table 2 displays the results of the statistical analysis and the ANOVA. The ANOVA analysis (

Table 3. Coefficients for the model and ANOVA table.

DES Treatment Factors	Y1:Cellulose			Y2:Lignin			Y3:Hemicellulose
	bij	F-Ratio	p Value	bij	F-Ratio	p Value	bij	F-Ratio	p Value
Intersection	27.94	-	-	−18.97	-	-	−51.78	-	-
X1:Time	3.13	3.92	0.10	9.46	0.88	0.39	−3.62	11.96	0.02
X2:Temperature	−0.47	4.00	0.10	0.46	0.00	0.96	0.27	1.49	0.28
X3:Ratio	1.73	0.46	0.53	1.71	3.67	0.11	7.02	0.07	0.79
X1X1	−0.16	0.08	0.79	−0.95	14.81	0.01	0.56	2.66	0.16
X1X2	0.03	0.53	0.50	−0.02	1.11	0.34	−0.01	0.31	0.60
X1X3	−0.15	1.86	0.23	−0.13	7.25	0.04	−0.01	0.03	0.86
X2X2	0.00	0.90	0.39	0.00	2.79	0.16	0.00	0.34	0.58
X2X3	0.00	0.12	0.74	0.00	1.63	0.26	0.00	0.13	0.73
X3X3	−0.04	3.06	0.14	−0.03	7.58	0.04	−0.20	154.92	0.00
Statistical data
Number of experiments			15			15			15
R2			75.23%			88.06			98.68
R2-adjusted			30.64%			66.57			96.3
Sum-Square errors			98.88			18.09			21.49
p-value			0.29			0.0674			0.0004
Standard error			4.45			1.9			2.07
Total d.f			14			14			14
Model d.f			9			9			9
Error d.f			5			5			5

) indicates that the experiments have no statistically significant effect on cellulose reduction, as the model p-value is greater than 0.05. Therefore, it can be concluded that LA:ChCl DES does not significantly affect cellulose content, as predicted by the literature [44]. Regarding hemicellulose, the ANOVA table shows a significant effect of the operating conditions, as the model p-value is less than 0.05. Furthermore, the variables studied explain 96.3% of the variation in hemicellulose content, as indicated by the adjusted R² value. The statistical analysis also shows that DES can extract hemicelluloses with higher selectivity than lignin. Therefore, it is possible to predict hemicellulose removal with very good accuracy.

However, the data obtained from the ANOVA table is not sufficient to observe how the variables affect each component separately. Pareto plots, shown in Figure 1, help to understand how each operating condition affects each macrocomponent.

Figure 1a shows that the most influential variables for hemicellulose removal are the quadratic effects of the ratio and time. As predicted by the ANOVA table and shown in Figure 1b, none of the operating conditions have a statistically significant effect on cellulose content. Regarding lignin, Figure 1c shows that the quadratic effects of time and ratio are the most significant variables, along with the interaction effect of the ratio and time. Using these graphs, some coefficients can be eliminated from the model equation, simplifying calculations to predict the final content. However, to avoid any accuracy issues, it was decided that all obtained coefficients be retained.

Once it was possible to obtain a mathematical equation describing the evolution of the macrocomponents throughout the process, it was decided that surface response graphs be presented to indicate the optimal operating conditions. Figure 2 shows the response surface graphs, representing the most affecting variables. Figure 2 helps to assess the range of values to optimize the process. Regarding hemicellulose content, it can be reduced to around 20% at low temperatures (60–70 °C) and over long treatment times (4–5 h).

To minimize hemicellulose content, these should be the treatment conditions. Additionally, the response surface indicates which variable most strongly affects hemicellulose content. This can be seen by comparing the slopes of the surface. Over time, the slope increases, becoming steeper than that for temperature, confirming the results obtained from the Pareto plots. Cellulose-wise, when plotted against temperature and time, the highest cellulose release is observed at 5 h and 120 °C, reaching 44%. However, the minimum cellulose content occurs at milder temperatures, around 80–100 °C. This may be contrary to the conditions observed for hemicellulose content, but since temperature does not affect it, the temperature can be set at 120 °C.

The lignin trend with the operating variables is shown in Figure 2c. The optimal region for lignin elimination is around a 1:30 biomass:DES ratio and at the longest time (5 h). Notably, under mild conditions (3 h and a biomass:DES ratio of around 1:18), lignin content reaches a maximum of approximately 32%. This observation provides insight into the kinetics of lignin elimination and the selectivity of DES towards hemicellulose removal. At short times (below 3 h), lignin reaches a maximum due to the selective removal of hemicellulose.

The results obtained in this study are consistent with several studies that used the same DES (ChCl:LA) at different ratios and conditions, showing hemicellulose removal to be higher than lignin removal [45]. Other researchers used a ChCl-based DES with lactic acid to increase the saccharification of Napier Grass, achieving 92.5% hemicellulose removal and 82.92% lignin removal, demonstrating the selectivity of DES towards hemicellulose elimination under conditions similar to those used in this study [46]. Another study used a ternary DES formed by ethylene glycol:lactic acid:ZnCl₂ at 120 °C as a pre-treatment to increase cellulose conversion, resulting in a 94.3% xylan (hemicellulose) reduction and a 60.8% lignin reduction [47]. Zhang et al. [48] reported higher hemicellulose removal at similar temperature conditions (110 °C), around 50% and 39%, for hemicellulose and lignin removal, respectively.

From the literature it was possible to compare the results in this study to those that use different residues from food industry. In this study, lignin removal in CSS (39.49%) is similar to that in corn stover (48.4%) [49] and in brewers’ spent grains (46.0%) [50]. Hemicellulose removal in CSS (88.02%) was exceeded by 100% [50], 100% [46], and 94.3% [47]. Koffi et al. [51] used DES to increase furfural production in rice husk, and the same trend was observed; 86% reduction in hemicellulose and 64% elimination of lignin were achieved. This behavior suggests that ChCl:LA treatment of CSS is particularly selective towards hemicellulose removal.

The differences observed among the feedstocks should not be interpreted only in terms of removal percentages, since pretreatment performance depends on the initial biomass composition, hemicellulose accessibility, lignin content and structure, and the chemistry and severity of the DES system used. Based on the compositional trends obtained in this work and on previous studies, the interaction mechanism between ChCl:LA and CSS can be interpreted as the combination of three main effects: acid-catalyzed hydrolysis of hemicellulose, disruption of biomass hydrogen-bonding networks, and different accessibility of the lignocellulosic fractions.

First, the acidic character of lactic acid favors the cleavage of hemicellulosic glycosidic linkages, while the amorphous, branched, and more accessible structure of hemicellulose makes this fraction more susceptible to removal [52,53]. This explains the higher hemicellulose removal observed in CSS compared with lignin. Second, the ChCl-based DES may disrupt hydrogen-bonding networks and lignin–carbohydrate interactions, increasing biomass accessibility during pretreatment. Third, lignin removal is more limited because lignin is a heterogeneous and recalcitrant aromatic polymer. Although labile ether linkages such as β-O-4 may be partially cleaved, lignin solubilization can be restricted by solvent–lignin interactions, DES viscosity, mass-transfer limitations, and possible acid-induced condensation under severe conditions [54,55]. Consequently, cellulose is comparatively retained in the solid fraction due to its more ordered and crystalline structure, resulting in cellulose enrichment as hemicellulose and part of the lignin are removed.

To sum up,

Table 4. Comparison of this study with other food industry residues.

Biomass Feedstock	DES System	Temperature	Lignin Removal (%)	Hemicellulose/ Xylan Removal (%)	Main Observation	Ref.
Coffee silverskin	ChCl:LA	60–120 °C	39.49	88.02	High selectivity towards hemicellulose removal and cellulose enrichment	This work
Potato peels	DES-based pretreatment	n.r.	29.7	100.0	Complete hemicellulose removal with moderate lignin removal	[45]
Moso bamboo	EG:LA:ZnCl2	120 °C	60.8	94.3	High xylan removal and relevant lignin removal	[42]
Napier grass	ChCl:LA-based DES	n.r.	82.92	92.5	High removal of both hemicellulose and lignin	[41]
Corn stover	DES-based pretreatment	n.r.	48.4	59.2	Moderate removal of both lignin and hemicellulose	[44]
Brewers’ spent grains	DES-based pretreatment	n.r.	46.0	47.9	Similar lignin and hemicellulose removal	[45]
Apple residues	DES-based pretreatment	n.r.	57.7	46.7	Higher lignin removal than hemicellulose removal	[45]
Moso bamboo	DES-based pretreatment	110 °C	39–50	39–50	Comparable removal of lignin and hemicellulose under mild temperature conditions	[43]
Rice husk	DES + GVL	n.r.	64.0	86.0	High hemicellulose removal and substantial lignin removal	[46]

compares the different DES treatments of various lignocellulosic feedstocks. Finally, using the mathematical model, the optimum operating conditions were 3 h, 120 °C, and a 1:10 biomass-to-DES mass ratio, which yielded 36.5% cellulose, 8.02% hemicellulose, and 29.3% lignin. It should be noted that this optimum corresponds to a compositional optimum, since it was determined from the final cellulose, hemicellulose, and lignin contents of the treated solid. In this regard, the optimum biomass:DES ratio of 1:10 is favorable because it represents the lowest solvent loading evaluated in this study. However, DES recovery and reuse, detailed mass balances, cellulose recovery, energy demand, and techno-economic assessment were not included in the present work and should be addressed in future studies before scale-up.

4. Conclusions

This study evaluates the application of DES as a pretreatment for residual biomass from the coffee industry. Specifically, the DES composed of ChCl:LA in a 1:9 molar ratio was employed under various operational conditions, including different temperatures, biomass-to-DES ratios, and durations, to purify the biomass and diminish lignin and hemicellulose contents. It was observed that the DES does not influence the cellulose content of CSS. Indeed, this pretreatment increased cellulose content from 17.59% to 48.55% when conducted at 120 °C for 5 h with a biomass-to-DES ratio of 1:20. Statistically, the operational parameters were not significant for cellulose content. At 60 °C, utilizing a biomass-to-DES ratio of 1:10 for 3 h, the hemicellulose content was reduced by 3.37%. Conversely, the lowest lignin content achieved was 20.03%, at 90 °C, a biomass-to-DES ratio of 1:30, and a duration of 5 h.

DES are a cleaner option than conventional pretreatments and achieve very high delignification and hemicellulose removal rates, which enables the treatment of waste that previously went directly to landfills and allows these materials to be valued as different products through small biorefineries in Europe. In addition, the viability of using DES with coffee residues has been demonstrated, and its efficiency is comparable to that found in other biomass sources.

Although the compositional results indicate cellulose enrichment and preferential hemicellulose removal, complementary structural characterization such as FTIR, XRD, SEM, TGA, or lignin molecular analysis might elucidate the biomass deconstruction mechanism. These characterization tests will be carried out in future studies in which DES-pretreated biomass feedstocks will be used as adsorbents under the optimum DES pretreatment conditions found in this study.