A Review of Caffeine Adsorption Studies onto Various Types of Adsorbents

A systematic literature review of publications from 2000 to 2020 was carried out to identify research trends on adsorbent materials for the removal of caffeine from aqueous solutions. Publications were retrieved from three databases (Scopus, Web of Science, and Google Scholar). Words “adsorption AND caffeine” were examined into titles, abstracts, and keywords. A brief bibliometric analysis was performed with emphasis on the type of publication and of most cited articles. Materials for the removal of caffeine were classified according to the type of material into three main groups: organic, inorganic, and composites, each of them subdivided into different subgroups consistent with their origin or production. Tables resume for each subgroup of adsorbents the key information: specific surface area, dose, pH, maximum adsorption capacity, and isotherm models for the removal of caffeine. The highest adsorption capacities were achieved by organic adsorbents, specifically those with granular activated carbon (1961.3 mg/g) and grape stalk activated carbon (916.7 mg/g). Phenyl-phosphate-based porous organic polymer (301 mg/g), natural sandy loam sediment (221.2 mg/g), composites of MCM-48 encapsulated graphene oxide (153.8 mg/g), and organically modified clay (143.7 mg/g) showed adsorption capacities lower than those of activated carbons. In some activated carbons, a relation between the specific surface area (SSA) and the maximum adsorption capacity (Qmax) was found.


Introduction
Emerging contaminants (ECs) include a wide range of chemical compounds, pharmaceuticals, personal care products, surfactants, industrial additives, plasticizers, and pesticides, among other, and their possible consequences for human health and the environment effects in many cases are still unknown [1][2][3][4]. e problem with ECs present in water sources is that plants for purification and wastewater treatment are not able to eliminate them completely, so their persistence in the environment is continuous. Caffeine is continuously detected, due to the advancement and development of instrumental analysis methods [5], and it is considered an indicator of anthropogenic contamination due to its common use by peoples [6]. Caffeine is an EC commonly found in drinking water [7][8][9], groundwater [4], wastewater [10][11][12], effluents from wastewater treatment plants [12], rivers, lakes, seas [13], and even in the Antarctic waters [14]. Caffeine is a persistent compound, routinely detected even in countries where coffee is not cultivated but consumed. For example, caffeine concentrations of 0.29 μg/L and 564 ng/L were determined in drinking water in California (USA) [9] and China [15], respectively. Also, caffeine concentration values in the 1644-3344 ng/L range have been measured in the Italian river Lambro after receiving the wastewater discharge from the city of Milan [16].
Caffeine is a stimulant of the central nervous system, a chemical compound from the group of methylxanthines (as illustrated in Figure 1), and the mostly consumed psychoactive substance in the world [19]. Caffeine is a naturally occurring alkaloid in approximately 60 plant species, including coffee, tea, and cocoa [20]. It also appears in some analgesic and bronchodilator drugs, and even in shampoos [6,17,18].
According to the information in Figure 1, the dipole moment of caffeine is high, increasing with the polarity of the medium that holds it [21]. It means that the positive charge of nitrogen in caffeine electrostatically interacts with any negatively polarized functional group [17,22]. e pKa and pH affect the chemical behavior of pollutants, and a protonated form of caffeine will be produced when pH < pKa [23,24]. e pKa and pH affect the chemical behavior of contaminants, so a protonated form of caffeine will be produced when pH < pKa.
Chemical treatments for emerging contaminants in water, as caffeine, include ozonation, photo-Fenton processes, photoelectrolysis, and electrochemical oxidation [15,17,25,26]. However, these technologies tend to consume high energy and do not achieve a complete mineralization [27]. In contrast, adsorption is efficient, inexpensive, versatile, and environmentally friendly. Adsorption is widely used to remove contaminants in water, especially those that are not biodegradable such as heavy metals and ECs, being caffeine one of them [28][29][30][31][32]. Recent proposals for the removal of caffeine include batch and fixed bed adsorption treatments [26,33]. Powdered activated carbon (PAC) and granular activated carbon (GAC) have been the most widely used adsorbents due to their high specific surface area and chemical surface properties.
Adsorption is a surface phenomenon, where contaminants dissolved in a liquid phase (adsorbate) interact with a porous solid surface (adsorbent). Generally, the adsorbent surface contains functional groups that allow physical or chemical interaction with the adsorbates present in the fluid [34]. In industrial processes, the contaminated fluid passes through a fixed bed where solid adsorbate particles retain the contaminant. When the bed becomes almost saturated, it regenerates inducing desorption of the adsorbate by heating or other methods, so the adsorbent becomes ready for another adsorption cycle [35]. Adsorption is influenced by several factors, including pH, ionic strength, temperature, amount of adsorbent, particle size, contact time, initial solute concentration, specific surface area, and stirring speed [36][37][38]. e selection of the operating ranges for these factors is essential in the study of the adsorption process [39].
Interest in the use of adsorbents for the removal of caffeine from aqueous solutions has led to the publication of several reviews on topics related to the very same purpose of this paper. Anastopoulos et al. [40] published a review focusing on emerging contaminants such as caffeine, nicotine, and amoxicillin.
at review covered the toxic effect of caffeine in humans and animals and described some of the main adsorbents used for its removal. Isothermal and kinetic models for the analysis of caffeine adsorption were also presented, as well as the maximum adsorption capacity and possible adsorption mechanisms. Rigueto et al. [17] prepared a review on methods for removing caffeine from aqueous solutions and real effluents, examining major findings and limits for each process. e research concluded that, despite encouraging application trends, current technologies for caffeine removal have significant limitations, including the complexity of adsorption mechanisms, quantification of contaminants in real effluents, and the low sustainability of the technique. Finally, Bachmann et al. [41] performed a systematic review of the removal of caffeine by adsorption, emphasizing on the evolution of adsorbents used and the kinetic, equilibrium, and thermodynamic studies. e pseudo-second-order (kinetic model) and the Langmuir isotherm models yield the best fit of the experimental data in most studies. On the other hand, our review presents a brief bibliometric analysis of publications of the last two decades, a classification of adsorbents in three groups, including a novel analysis of the process variables (pH and mass of adsorbent), characterization of the adsorbent (SSA), and the maximum adsorption capacity (Q max ).
Current trends in caffeine adsorption point the development of materials with adsorbent properties (affinity for the pollutant and high specific surface area), which are abundant and inexpensive. In this sense, the strengthening of scientific research related to the application of new adsorbent materials for the removal of caffeine, contributes to the implementation of clean and environmentally friendly industrial technologies and processes [32,40,42,43]. is document provides a systematic review of literature about caffeine adsorption, presenting the diversity of adsorbents used in research reported in the last two decades, as well as a brief bibliometric analysis. An analysis of the results of the specific surface area of adsorbents was performed, as well as the variables that affect the adsorption process, specifically pH and adsorbent mass (pH is commonly controlled and evaluated in studies of caffeine adsorption). e results of the maximum adsorption capacities and their adjustments to the model's adsorption isotherms are also presented. is review article will allow researchers to identify the types of adsorbents mostly used for caffeine adsorption in aqueous medium, in addition to adsorption conditions such as adsorbent dose and pH. An analysis of the maximum adsorption capacity (Q max ) and its relationship with the specific surface area (SSA) of each adsorbent is also shown to determine the performance of the different types of materials used for adsorption.

Materials and Methods
A systematic literature review of publications (research articles, review articles, book, book chapters, and conference papers) from 2000 (January 1) to 2020 (July 29) was carried out to identify the research trends on caffeine adsorption from   [44]. e references were obtained from three search and indexing databases: Scopus, Web of Science, and Google Scholar, which together cover approximately 95% of the publications worldwide. Scopus indexes the largest number of journals (20% more coverage than WoS), but WoS performs a more open search so the filters had to be more specific and personalized [45,46]. ese databases included research articles, review articles, books, book chapters, and conference papers. e search was performed in the title, abstract, and keywords of publications. For search equation "caffeine AND adsorption," 528 results were found. e 528 preliminary results were manually filtered to remove repeated articles in the databases, to exclude publications that were not within the scope of the present work, and most importantly, to identify the most relevant publications. Finally, 133 publications were used for the preparation of this review, mainly journal articles on environmental sciences, chemistry, chemical engineering, engineering, and material sciences.  Table 1 shows that most publications are research articles, concentrated in the areas of are chemistry, environmental science, and chemical engineering, which add up to 56.3% of the total number of publications. Most of the publications (52.8%) come from three countries: China, Spain, and the USA. Besides, 93.8% of all publications are published in English, while 4.9% in Chinese, and the remaining in other languages: Spanish, French, Japanese, German, Portuguese, and Italian. Twothirds of the publications on caffeine adsorption are concentrated on four journals: Science of the Total Environment (23.1%), Chemosphere (18.7%), Environmental Science (13.2%), and Chemical Engineering Journal (12.1%). Also, a ranking of institutions per number of publications on caffeine adsorption is shown in Table 2. e ten most cited publications are listed in Table 3, eight of them describe remotion with carbonaceous adsorbents: graphene nanoplatelets, activated carbons, and carbon xerogels. e remaining two publications describe the application of low-cost adsorbents: carbon fibers prepared from pineapple plant leaves, and natural sediments.

Results and Discussion
To conclude the bibliometric analysis, the cloud shown in Figure 3 summarizes the appearance of keywords in the publications. e five most used keywords were adsorption (102 times), caffeine (94 times), water pollutants (54 times), drug (21 times), and activated carbon (19 times). erefore, this word cloud illustrates the interest in water and wastewater treatment and the use of adsorbents such as activated carbons for the removal of pharmaceutical products (caffeine, ibuprofen, diclofenac, and carbamazepine). e term pH also stands out, indicating that it is one of the most frequently studied variables in caffeine adsorption processes, being a keyword in 20 of the 133 publications. Finally, the appearance of kinetic and isotherm highlights the interest in understanding the caffeine adsorption mechanism.

Classification of Adsorbents.
All the adsorbents mentioned in the publications follow the common definition, i.e., solid materials with micro-and mesopores that can take a     [56,57]. Figure 4 shows the classification of adsorbents in three groups (with their respective subgroups): Organic (68.3% of publications): carbon-based, they can be of natural origin such as agricultural residues and biochar or of synthetic origin such as polymers. Some of them are of mixed origin, such as biopolymers obtained from chitosan compounds [58]. Inorganic (20.8%): mainly minerals, for example, silica, metal oxides, and materials such as mineral clays, sediments, and soils. By origin, they can be synthetic or natural [59]. Composite (10.9%): hybrids that combine two or more materials, of organic and/or inorganic type [60]. is group of adsorbents have been investigated with promising results for the removal of dyes [61,62], heavy metals [63,64], and emerging contaminants [65,66].
For each subgroup of adsorbents, the following parameters were analyzed: specific surface area (SSA), adsorbent dose, pH, maximum adsorption capacity, and isothermal and kinetic models for caffeine adsorption to which the data were adjusted.

Organic Adsorbents.
e summary of characteristics of the subgroups of organic adsorbents and conditions of the adsorption process are shown in Table 4 (activated carbons), Table 5 (carbon-based), Table 6 (agricultural wastes directly as adsorbents or as precursors for activated carbons), Table 7 (biochar), and Table 8 (polymeric resins and biopolymers). Most of the organic adsorbents in this group are activated carbons (Table 4), either powdered (PAC) or granular (GAC). Activated carbons have a high specific surface area, between 578 and 2431 m 2 /g and good surface chemical properties, but they are expensive, and some of them cannot be regenerated [67]. eir adsorption capacity range is very wide, including low (161 ng/g-396 μg/g), medium (4.95-219.2 mg/g), and high (271-1961.26 mg/g) values. e pH described for caffeine adsorption varies between 3 and 9 for this subgroup of adsorbents, with the highest Q max values obtained at pH 7. e carbon-based adsorbents subgroup ( Table 5) covers materials that come from carbon but differ from activated carbons, such as xerogels, nanotubes, nanofibers, and graphene. Carbon-based adsorbents have a smaller specific surface area and removal capacity than activated carbons, although they are considered materials with potential application for the removal of caffeine [48,72,[82][83][84]. Carbon cloth has both the highest SSA and Q max values in this subgroup, 1560 m 2 /g and 369.0 mg/g, respectively [85].   e Scientific World Journal   e Scientific World Journal 5 Carbon xerogels and carbon cloth show adsorption capacities of 107.0 and 369.0 mg/g, respectively [52,85], while carbon nanotubes and commercial column C18 show adsorption capacities in the range of 8.14 to 41.6 mg/g and less specific surface area [72,82,83,86]. Following Table 5, the most used pH for removal of caffeine is close to neutrality, except for carbon xerogel modified with (CH 3 COO) 2 Cu, nanotubes, and carbon nanofibers, where the optimal pH at values of 2 and 3 was found [72,87].
Agricultural wastes (Table 6) are low-cost natural raw materials, abundantly available. e agricultural waste subgroup includes them used directly as adsorbents and the activated carbons produced from them, which have improved properties, mainly specific surface area (407.66-1099.86 m 2 /g) and adsorption capacity for caffeine (8.7-916.7 mg/g).
Activated carbons are produced from agricultural wastes by controlled pyrolysis combined with chemical treatment, taking advantage of its high carbon content and low percentage of inorganic matter [88][89][90]. Activated carbons from grape stalk and pine have the highest SSA and adsorption capacity [91,92], while those obtained from other agricultural wastes such as peach stones, dende coco, and babassu coco have a Q max between 186.9 and 270 mg/g, which is anyway higher than those of other carbon-based adsorbents such as nanotubes (41.6 mg/g) and nanofibers (28.3 mg/g) [72].
Regarding pH, acidic pHs in the range 2-5 tend to yield the best results, such as produced as follows: at pH 2, a maximum adsorption capacity of 13 mg/g for the activated carbon obtained from Elaeis guineensis (palm oil) [93]; at pH 3, a Q max of 212.3 mg/g for biomass-derived activated carbons (dende coco and babassu coco) [94]; at pH 4, a Q max of 916.7 mg/g for grape stalk activated carbon [92]; and at pH 5, a Q max of 500 mg/g for pine activated carbon with K 2 CO 3 [91]. For pH values between 6.3 and 7.7, a Q max of 30.3 to 270 mg/g has been reported, being the case of activated carbons obtained from residues of Acacia mangium wood, pineapple plant leaves, Eragrostis plana, and peach stones [51,54,95,96].
Biochar adsorbents (Table 7) are obtained from pyrolysis of animal or vegetable biomass at temperatures between 300 and 700°C with a low oxygen amount [102][103][104]. e interest in biochar adsorbents has recently surged due to their high specific surface area and adsorption capacity [105,106]. e adsorption capacities of the biochar adsorbent subgroup (Table 7) are between 6.54 and 40.2 mg/g, well below values of activated carbons but comparable with those of carbonbased adsorbents such as nanoplatelets [48], carbon nanofibers [72], and graphene [84]. e pH for caffeine adsorption in this subgroup of adsorbents is slightly acidic, with values between 3.5 and 5.9.
Polymeric resins, solid or liquid, are usually based on polystyrene, polyacrylamide, and polyvinyl alcohol. Recent improvements in polymerization processes have achieved enough resistance and chemical stability to make them suitable for the removal of contaminants from water [60]. Summary of characteristics of polymeric resins and  6 e Scientific World Journal biopolymers and conditions of the adsorption process are described in Table 8. e Q max in Table 8 is not very high (between 256.4 ng/g to 301 mg/g) compared with the Q max of activated carbons, but the high specific areas of the polymer resins (450-1000 m 2 /g) suggest that they could be very effective for the adsorption of other contaminants. Biopolymer adsorbents are polymers obtained from algae, plants, or agro-industrial products, hence the bio prefix. e most common one is chitosan, which is obtained from chitin, an abundant mucopolysaccharide and the support material of crustaceans and insect exoskeletons. It is biocompatible, biodegradable, and nontoxic and, most importantly, has good adsorption properties [122]. Chitosan and other biopolymers have been evaluated with good results for removal of contaminants such as metals and dyes, but their performance for caffeine adsorption is not promising [123][124][125]. e costs of obtaining biopolymers are very high and show no advantages over other adsorbents investigated for the removal of caffeine, except when used with other materials to form composites [123,126].

Inorganic Adsorbents.
is kind of adsorbents can be classified into two subgroups: aluminosilicate minerals (Table 9) and soil and sediment adsorbents (Table 10); and the most used for the removal of caffeine have been clay minerals (natural and modified), zeolites, soils, and sediments. It is also possible to change the naturally hydrophilic character of smectite clay minerals into organophilic, making them act as adsorbents for organic compounds. e intercalation of surfactant cations (quaternary ammonium salts) into layers of clay minerals not only changes the surface properties but also increases the interlamellar distance (basal spacing) of the layers, thus easing the adsorption of organic molecules [42,59,[127][128][129][130][131]. Bentonite (a montmorillonite) and saponite are the most frequently used clays as caffeine adsorbents.
Regarding aluminosilicates (Table 9), smectite minerals-such as montmorillonite and saponite and their modifications, organoclays, or thermally treated smectites-show a good caffeine adsorption capacity (99.0-143.7 mg/g), with the advantage of being abundant and low-cost materials [22,43,130,131]. Although aluminosilicates have a low SSA in comparison with activated carbons, sepiolite and bentonite have a good adsorption capacity for caffeine. For example, a sepiolite with a specific surface area of 221 m 2 /g has a maximum adsorption capacity of 48.7 mg/g [132], and bentonite with a low SSA (64.31 m 2 /g) has a Q max of 42.5 mg/g [127]. e pH for caffeine adsorption in the subgroup of aluminosilicate mineral adsorbents (Table 9) varies between 6.0 and 6.6, values close to neutrality.
Soil and sediments (Table 10) are sandy and silty minerals obtained from aquifers and rivers. ey are adsorbents of water contaminants, preferably those with a low content of organic matter and clay, which increase the specific surface area [50,53,136]. e maximum adsorption capacities of soils and sediments (in the range of 292 μg/kg to 221.2 mg/g) tend to be lower than the Q max of the other adsorbent types covered in this review. It has been also described that the equilibrium time among this group of adsorbents and adsorbates is very long, in some cases more than 24 h, which would explain why sediments do not retain ECs in rivers and water bodies [137,138]. e pH for caffeine adsorption in this subgroup of inorganic adsorbents (Table 10) varies between 6.0 and 7.98, values close to neutrality.

Composites.
is subgroup comprises adsorbents formed by the combination of materials of two or more types, such as organic and inorganic, to improve their morphological characteristics, specific surface area, and adsorption capacity [60,109]. ey fall in two categories: organic-organic composites (Table 11) and organic-inorganic composites (Table 12). Composite adsorbents have been investigated with promising results for the removal of dyes [61,62], heavy metals [64], and emerging contaminants [65,66].
Organic-organic composites used for the adsorption of caffeine include chitosan-graphene, chitosan-reduced graphene oxide, and chitosan-waste coffee-grounds. e synthesis of composites notably improves the specific surface area of adsorbents and increases the caffeine adsorption capacity. For example, when chitosan was used as an individual adsorbent, a specific surface area of 3.6 m 2 /g and a maximum adsorption capacity of 0.0062 mg/g were obtained [121], while the combination with graphene yields 214 m 2 /g and 14.8 mg/g, respectively [140]. For the chitosan-reduced graphene oxide composite, the SSA is not reported, but the Q max of caffeine was 63.6 mg/g [126]. e resin resulting from the copolymerization of N-vinyl-2-pyrrolidinone (PVC), ethylene glycol dimethacrylate (DEGMA), and triallyl isocyanurate (TAIC) has also been tested as a caffeineadsorbent material, although with low removal capacity (2 mg/g) [141]. e pH range for caffeine adsorption on chitosan-graphene composites is 6-7, values close to neutrality.
In general, in the group of organic adsorbents, activated carbons tend to have the highest specific surface area and Q max . Most of them are microporous in structure, although mesopores and macropores are also present, with a size distribution that mainly depends on three factors: origin of the raw material, type of activation, and duration of the activation process. e capacity of an activated carbon to retain a given substance depends not only depends on its specific surface area but also on the proportion of internal pores that the carbon has and the size of these pores (a suitable pore size should be between 1 and 5 times the diameter of the molecule to be retained) [150]. Adsorbents obtained from agricultural residues, biochar, and polymeric resins are modified by different processes developing porous structures that notably improve the maximum adsorption capacities, as it happens with the AC obtained from e Scientific World Journal lignocellulosic materials [55]. Regarding composite-type adsorbents, Delhiraja et al. [140] have used functionalized graphene oxide composites (GO) for the adsorption of caffeine and other pharmaceutical and personal care products. Density functional theory calculations indicated that the adsorption mechanism is typically accompanied by size-related diffusion and a minor contribution of synergetic combination of hydrophobic/hydrophilic, hydrogen bonding, electrostatic, and π-π interactions.

Final Discussion
is section presents an analysis of the maximum adsorption capacity of caffeine on each adsorbent subgroup and its relationship with the specific surface area of the material and the pH at which adsorption was performed. e adsorption of caffeine-like molecules onto activated carbon, its interaction with adsorbents, and the regeneration and reuse of adsorbents for caffeine removal are also analyzed.     Figure 5 shows the Q max of caffeine for the adsorbents of each subgroup. Granular activated carbon (MG 1050 from ChiemiVall-Spain, SSA of 1100 m 2 /g and an average particle size of 1 mm) is the adsorbent with the highest adsorption capacity (1961.3 mg/g) followed by grape stalk activated carbon (916.7 mg/g) and activated carbon cloth (369 mg/g). e above three adsorbents are activated carbon-based organic adsorbents, and biochar is one of the organic adsorbents with the lowest Q max . Adsorbents such as natural sediment (221 mg/g) and aluminosilicate minerals (143.7 mg/g) are an option for removal of caffeine, due to their low cost and abundance. Figure 6 shows the Q max -SSA relation for the adsorbent subgroups. e highest Q max values, above 70 mg/g, correspond to activated carbons (carbon-based and from agricultural wastes) followed by polymeric resins and organicinorganic composites, with SSA in the 140-1900 m 2 /g range. A direct relationship between SSA and Q max was found only for some adsorbents of the activated carbon subgroup.
In the publications covered in this review, the adsorption of caffeine has been performed over a wide pH range (2-10), although it is customary to operate at pH values close to neutral, and even some studies do not evaluate the influence of pH [47,84,151]. Figure 7 illustrates how the highest Q max (activated carbons, included those produced from agricultural wastes) correspond to pH values between 5 and 7. It is also observed that the adsorption with carbon-based materials decreases with pH, and their higher Q max are in the 2-3 pH range. It is important to consider that pH affects the degree of caffeine ionization and the surface charge of the adsorbent [99]. On some carbonaceous adsorbents, pH has little effect on caffeine adsorption; however, some studies indicate that caffeine adsorption capacity decreases with higher values of pH due to electrostatic repulsion [47,48,92,133].
Regarding the adsorbent dose, its effect was not studied in most of the articles reviewed, but it was set in a very wide range, from 0.05 mg/L to 40000 mg/L, while the value used for the highest Q max (1961.3 mg/L, with activated carbon) was 1000 mg/L [68].
For most of the organic adsorbent subgroups, the adsorption isotherms were fitted with the Langmuir model followed by the Freundlich, Sips, Toth, and Liu models. e Dubinin-Radushkevich and Redlich-Peterson models were also used for fitting the adsorption isotherms of caffeine in the subgroups of aluminosilicate minerals and organic-inorganic composites.

Adsorption of Caffeine-Like Molecules.
Adsorption of similarly sized molecules can be an indicative of a suitable material for caffeine removal (similar to caffeine such as pharmaceuticals, stimulants, and personal use products), as molecular size and pore distribution influence the adsorption process, particularly on activated carbons [49]. For example, caffeine and diclofenac both have spherical equivalent diameters of 6.9Å and 7.7Å, respectively, and their adsorption in presence of powder carbonaceous   e Scientific World Journal 9 materials can be attributed to their volumes being considerably smaller than the pore size of such materials [72]. Sotelo et al. [72] studied the removal of caffeine and diclofenac as emerging contaminants with three powder carbonaceous materials: activated carbon (AC, S BET � 997.0 m 2 /g, V micro � 0.260 cm 3 /g), multiwalled carbon nanotubes (MWNT, S BET �162.2 m 2 /g, V micro � 0.016 cm 3 /g), and carbon nanofibers (CNF, S BET �199.1 m 2 /g, V micro � 0.012 cm 3 /g). Carbonaceous materials were effective for the removal of emerging contaminants since the size of these compounds is considerably smaller than the pore size of AC, MWNT, or CNF. Q max values for compounds in ultrapure water on AC, MWNT, and CNF were 271.0, 41.6, and 28.3 mg/g for caffeine and 329.0, 41.4, and 29.9 mg/g for diclofenac, respectively.
In a subsequent study, Sotelo et al. [49] investigated the adsorption of caffeine and diclofenac by granular activated carbon F-400 (S BET � 997 m 2 /g and V micro � 0.26 cm 3 /g). e molecular size of caffeine is 0.98 × 0.87 nm compared to that of diclofenac, 0.97 × 0.96 nm. In competitive adsorption, adsorption capacities for caffeine and diclofenac were 190.9 and 233.9 mg/g, respectively. Both adsorption capacities decreased compared to the single adsorption system by 32.1% for caffeine and 29.1% for diclofenac. Caffeine and diclofenac molecules accessed similarly sized pores and directly competed for the same adsorption sites. Higher values of octanol-water partition coefficient of a diclofenac molecule might be responsible for the stronger affinity of the adsorbent surface.
Mailler et al. [152] explored the removal of 15 micropollutants (including pharmaceutical, stimulant, and personal use compounds) from wastewater treatment plant discharges using 4 types of activated carbons with different micro-and mesoporosity ratios. e highest average removal of the pollutants investigated was 52%, which was achieved with PB 170 ® (S BET � 957 m 2 /g, V micro+meso � 0.5066 cm 3 /g), an activated carbon with the highest specific surface area and volume of micro-mesopores. e lowest average removal (26%) was obtained with PC 1000 ® (S BET � 458 m 2 /g, V micro+meso � 0.2435 cm 3 /g), which is the activated carbon with the lowest specific surface area and volume of micro-mesopores. For the other two activated carbons (WP 235 ® and W 35 ® ) with similar pore volume (0.4841 and 0.4876 cm 3 /g), the average removal was 45%. In conclusion, the removal of micropollutants is associated with chemical characteristics of the compounds and textural properties of the adsorbent material, being favored by adsorbents with high specific surface area and micro and mesoporous volume.
Gil et al. [74] studied the removal of six emerging contaminants from aqueous solutions using a commercial granular activated carbon as an adsorbent (S BET � 578 m 2 /g, V totalpore � 0.564 cm 3 /g, V micro � 0.206 cm 3 /g). e Q max calculated for salicylic acid, caffeine, diclofenac, and ibuprofen from the isotherm fitting to the Langmuir model were 33, 88, 64, and 34 mg/g, respectively. In conclusion, organic molecules used in this study had a similar chemical structure. erefore, the behavior during the adsorption by the activated carbon will also be very similar. Finally, Zhang et al. [153] evaluated the performance of powdered activated carbon (PAC, S BET � 852.94 m 2 /g) for removing 28 types of antibiotics from water. Results of the PAC adsorption experiments showed that the percentage of removal of 28 selected antibiotics ranged from 96.5 to 99.9% and 86.8-99.6% in deionized water and surface water, respectively.

Interaction of Caffeine with Adsorbents.
e porous structure of the adsorbent material, energy heterogeneity, and surface chemical properties (presence of functional groups) are the main factors influencing the adsorption equilibrium [154]. Caffeine has a high dipole moment, and its positive charge of the nitrogen atom electrostatically interacts with any negatively polarized functional group [17]. e pKa of caffeine is 10.4, and the protonated form of this molecule in aqueous solution exists when pH < pKa [41].
In carbonaceous adsorbents, such as activated carbon or graphene, the carbon surface has polar groups with hydrophilic behavior such as −NH, −OH, −O, and −COO.
erefore, the adsorption of caffeine can be attributed to dipole-dipole interactions, where π-electrons and 2-nitrophenol aromatic rings of caffeine interact with the π-aromatic electrons present on the adsorbent surface [17,41,84]. Beltrame et al. [54] explained caffeine adsorption as the result of π-π interactions and the formation of hydrogen bonds between caffeine heterocyclic rings and the aromatic rings of activated carbon (from pineapple plant) in a pH range of 2-7. e removal of caffeine when the pH is lower than the zero charge point of the adsorbent (PZC) has been explained by hydrogen bonding between the adsorbent and adsorbate, as the surface of the material has a predominantly positive 10 e Scientific World Journal charge density and caffeine would not be electrostatically attracted to it [72]. e adsorption of caffeine with noncarbon adsorbents has been focused on low-cost materials such as sediments, polymer resins, and aluminosilicates, whose application is favored by the chemical properties of caffeine (high pKa and dipole moment). In these materials, the interaction of caffeine with the adsorbate is attributed to H-bonds, dipole-ion interactions, electrostatic interactions, and Van der Waals interactions [17,40,41,155]. e adsorption of caffeine on inorganic surfaces of the natural sandy sediment occurs because caffeine is positively ionized at the experimental pH (7.94) and the sediment surface is negatively charged. Furthermore, the distribution coefficient (Kd) of caffeine on the inorganic surface (Kd � 17.58) is greater than that on organic matter (Kd � 0.28), confirming that the interactions of caffeine with inorganic surfaces control its adsorption on the sediment [41,53].
In aluminosilicates, adsorption is facilitated in the interlayer space, even more so when these types of adsorbents are modified by expanding the interlayer space with quaternary ammonium sales [32,43]. However, caffeine was also removed in thermally modified bentonite at temperatures above 400°C, i.e., when it has undergone a collapse, so that, in such case, the interaction of caffeine does not occur in the interlayer space but on the surface of the bentonite [130].

Regeneration and Reuse of Adsorbents.
Although the adsorption capacity is an essential criterion in the selection of an adsorbent, its regeneration after use should be considered in a lifecycle analysis of the material, but only three publications out of the 133 selected included an analyzed reuse [68,86,97]. Similarly, in only one of the 45 manuscripts selected by Bachmann et al. [41] in their review on the removal of caffeine from aqueous media by adsorption, adsorbent regeneration tests were performed.
Batista et al. [97] used rapeseed activated carbons prepared by chemical activation with K 2 CO 3 for caffeine adsorption. Temperatures of 400, 500, and 600°C under N 2 flow were selected to carry out the regeneration assays. e thermal treatment at 400°C did not allow an effective regeneration of the activated carbon, and only 30% of the original adsorption capacity for caffeine was retained. For activated carbons regenerated at 500 or 600°C, an almost complete recovery of the caffeine adsorption capacity was observed (>95%), even after the second regeneration cycle. e N 2 adsorption isotherms of the samples obtained after two exhaustion-regeneration cycles showed that at high regeneration temperatures (500 and 600°C), the volume of micropores available is higher. It is important to consider that activated carbons are regenerated ex situ by heating or steaming, which is a high energy-consuming process and can limit the reuse of the adsorbent [41,94].
Multiwalled carbon nanotubes were used for the adsorption of diclofenac and caffeine. e adsorbent used was separated from the solution by filtration and treated with 0.1 M HCl for 2 h to desorb the organic retained compounds. e solid was then separated from the acid solution by filtration, washed with deionized water, and dried in an oven at 60°C for 48 h before being reused in the adsorption experiments. e reusability of this adsorbent was studied, and it was found to maintain its adsorption capacity after one cycle of reuse. However, data only referred to diclofenac [86].
Discarded granular activated carbon from a drinking water treatment plant was used for the adsorption of different pharmaceutical compounds, specifically caffeine, ibuprofen, and diazepam. e results were very promising since the reused carbon adsorbed caffeine (initial concentration of 1.23 mg/L, bed height of 10 cm, and weight of 13.51 g) and achieved a removal of about 40%. e reused activated carbon achieved a better performance for pharmaceutical compounds elimination as powder after grinding (at a concentration of 0.5 g/L) than as granular carbon in column. Caffeine was removed in percentage higher than 90%. us, future applications of used granular activated carbon, particularly in the framework of a circular economy, may be possible [68].
Abdel-Aziz et al. [156] synthesized bimetallic zero-valent iron/copper nanoparticles (FB-nZVFe/Cu), having used this material for caffeine adsorption. FB-nZVFe/Cu was used during five successive times for the adsorption of caffeine (5 mg/L). Between each adsorption cycle, FB-nZVFe/Cu was immediately collected from solution by centrifugation, washed with ethanol, and dried at 45 C, before being used for the next adsorption recycle. e caffeine removal was 82, 78, 83, 83, 70, and 69% in the 1st, 2nd, 3rd, 4th, and 5th cycle of use, respectively. Although a decrease in the removal efficiency was found with each reuse cycle, the FB-nZVFe/Cu material offered a high potential to be repeatedly used for caffeine removal without a considerable decrease in its removal efficiency.
is article was not included for the analysis of adsorbents since its publication occurred after the date established in the search strategy.

Conclusions
e systematic literature review of publications on caffeine adsorption indicates that it is a recent and developing research topic because caffeine is becoming an emerging contaminant found in different types of water (surface, drinking, and wastewater). Caffeine is listed, along with nicotine, paraxanthine, and cotinine, as an anthropogenic marker of contamination, hence the importance of researching on methods for its removal. Of the total number of publications analyzed, 91.7% are research articles, 93.8% are written in English, and 67.1% are concentrated in four journals: Science of the Total Environment, Chemosphere, Environmental Science, and Chemical Engineering Journal.
Adsorption is one of the most frequently used methods for the removal of emerging contaminants, including caffeine. Powdered activated carbon (PAC) and granular activated carbon (PAC) are the mainly studied adsorbents for removal of caffeine. eir wide use is associated with their high specific surface area (up to 2431 m 2 /g) and high adsorption capacities (up to 1961.3 mg/g). e high cost of activated carbon and its difficult regeneration have led to the search for low-cost, widely available, and ecofriendly adsorbents. Materials obtained e Scientific World Journal 11 from agricultural biomass waste, polymeric resins, clay minerals, soil and sediments, and organic-inorganic compounds emerge as an alternative adsorbent for the removal of caffeine. However, high doses (100 to 40000 mg/L) are required to achieve high removal of caffeine. In 49% of the investigations analyzed in this review, the adsorption isotherms for caffeine fit into the Langmuir model, which assumes that adsorption takes place in a monolayer and that there is homogeneity on the surface of the adsorbent. e adsorption capacity of caffeine depends on the properties of the adsorbent and other factors such as pH, adsorbate concentration, agitation speed, and contact time. Removal of caffeine increased with increasing adsorbent dose (or adsorbent amount) since the number of adsorption sites is greater. However, the capacity, in some cases, decreased with an increasing dose. is is because of the aggregation of particles, which leads to a decrease in active sites for adsorption. Determining the adsorbent dose is a useful factor in predicting the cost of the process per unit of solution to be treated.

Data Availability
All data used to support the findings of this study are included within the article.

Conflicts of Interest
e authors declare that there are no conflicts of interest regarding the publication of this paper.