**Petrographic Characteristics of Sandstones as a Basis to Evaluate Their Suitability in Construction and Energy Storage Applications. A Case Study from Klepa Nafpaktias (Central Western Greece)**

**Petros Petrounias 1,\*, Panagiota P. Giannakopoulou 1, Aikaterini Rogkala 1, Maria Kalpogiannaki 1, Petros Koutsovitis 1, Maria-Elli Damoulianou <sup>1</sup> and Nikolaos Koukouzas <sup>2</sup>**


Received: 4 February 2020; Accepted: 27 February 2020; Published: 2 March 2020

**Abstract:** This study investigates how the petrographic features of Klepa Nafpaktias sandstones affect their behavior in construction applications such as concrete, in environmental applications such as energy storage as well as whether they are suitable for the above uses. For achieving this goal, sandstones (ten samples) were collected in order to study their petrographic characteristics using petrographic microscope and GIS software, as well as their basic physical, mechanical and physicochemical properties were also examined. Concrete specimens (C25/30) were made according to international standards including the investigated aggregate rocks in various grain sizes. Various sandstones were tested and classified in three district groups according to their physicomechanical features as well as to their petrographic and microtopographic characteristics. Concrete strength's results determined the samples into three groups which are in accordance with their initial classification which was relative to their grain size (coarse to fine-grained). As the grain size decreases their physicomechanical and physicochemical properties get better resulting in higher concrete strength values (25 to 32 MPa). Furthermore, the proposed ratio *C*/*A* (crystals/mm2) seems to influence the aggregate properties which constitute critical factors for the final concrete strength, presenting the more fine-grained sandstones as the most suitable for concrete aggregates. Concerning the use of Klepa Nafpaktias sandstones as potential energy reservoirs, the studied sandstones presented as suitable for CO2 storage according to their physicomechanical characteristics.

**Keywords:** petrographic characteristics; sandstones; physicomechanical properties; concrete petrography; CO2 storage

#### **1. Introduction**

Applied petrography constitutes an essential tool for the assessment of natural rocks or recycling materials for different useful applications such as concrete and energy storage. Petrography, generally, using a combination of methods such as microscopic observations (polarizing and scanning electron microscope) and chemical analysis examines the nature of each given rock/material showing the main relationship of texture, structure, composition and alteration degree [1–9]. Through these relationships, petrography may explain the physicomechanical and physicochemical properties of materials/rocks as well as the relationships among them. It is well-known that the already above-mentioned properties

are the critical ones that define the particular use of each given material/rock either construction or environmental applications.

Concrete is one of the most important and useful composite material, which is made from a mixture of cement, aggregates, water and sometimes admixtures in required proportions [10–13]. Cement and water bonded with aggregates constitute the concrete. Aggregates constitute the main components of concrete and occupy between 70% and 80% of the concrete volume [10]. Nevertheless, concrete performance depends on the aggregate particles quality [11–14]. Crushed rocks derived from various geological sources are the natural aggregates [15]. Critical factors for the suitability of aggregates in construction use are their physicomechanical behavior. This behavior is affected by the mineralogical and textural features of aggregates which play critical role in its strength and therefore in concrete strength. The most common types of rocks used in concrete production are classified into igneous, sedimentary and metamorphic rocks. Aggregates can be expected to have an important influence on the concrete's properties [16]. Such rocks are mainly limestone, granite, sandstone, quartzite, dolomite, marble, dacite etc. Each of these rock types is more or less suitable for uses as concrete aggregates, based on their petrographic characteristics and therefore on physicomechanical properties which contribute to reinforcing the strength of the concrete.

Sandstone is a widespread aggregate presenting various construction applications such as concrete. The physicomechanical properties of the aforementioned rock lithotype are quite different and aggregates such as quartzite, subarcose and greywacke can produce various behaviors of hardened concrete. Thus, it can be seen that these sandstone aggregates can obviously be characterized to acquire relatively predictable aggregate and concrete performances [17]. Sandstone used as aggregate of variable sizes within concrete may result in influencing the corresponding strength. In addition, it is regarded as significant for these aggregates to be graded when applied for concrete production. Moreover, it is well-known that sandstone is vastly affected by moisture content which results in the decrease of the identified mechanical property behavior in brittle construction material. Additionally, these sedimentary rocks have the tendency to display smaller strength compared to conventional aggregate material such as limestones. Sandstone performance behavior under anhydrous conditions is regarded as being good, which is not in the case of hydrous conditions since it is regarded as poor behavior especially in sandstones that are not well cemented [18,19]. Quartz content in concrete prepared by sandstone aggregates determines the concrete application [20]. Yilmaz and Tugrul [21] reported that concretes produced by comparable qualities and quantities cements exhibit variability in strength values depending on: mineralogy assemblage of aggregate, type of cement, textural and physico-mechanic characteristics.

Many researchers have investigated the correlations between the percentages of specificmineralogical compositions of aggregates and the final compressive concrete strength. Petrounias et al. [22,23] when investigating igneous rocks from Greece concluded that the alteration products of serpentinite-bearing rocks and andesitic-intermediate rocks have a profound impact on their mechanical behavior, that apparently affect their ability to be characterized as suitable concrete aggregate material. On the other hand, Yılmaz and Tugrul [21] evaluated sandstone aggregates from Turkey based on the parameters that include: physical properties, mechanical properties and compositional variations. They concluded that concretes produced by comparable qualities and quantities of cements present variabilities in the obtained strength values regarding mineralogy assemblage, the textural and physico-mechanic characteristics of the tested sandstones used in the size of aggregate.

During the last decades, only a few attempts to combine the database and visualization facilities of Geographic Information System (GIS) software and petrographic features of rocks have been carried out. In these studies, polarizing microscope images have been used in order to identify and visualize rock textures on microscopic scale. Li et al. [24] using GIS software to segment and analyze boundaries in the perimeters of grains, proposed a procedure to be applied on examined samples. Barraud [25] has used GIS techniques to process vectorized textural features, whereas Fernandez et al. [26] used the same application to parametrize single grain crystals through map development methods. An innovative methodology combining GIS and petrographic characteristics of various rocks has been applied by Tarquini [27].

Economic growth and a rising global population means that the worldwide demand of energy will be rising with very fast pace. This increases concerns that the extensive use of fossil-fuels should be mitigated, allowing space for further development of renewable energy sources. The problem that arises with the use of the latter is that most of these sources are intermittent and therefore energy storage applications are necessary to make them available around-the-clock for uninterrupted power supply [28]. Suitable subsurface geological formations can serve as energy storage reservoirs depending on the storage purpose and the type of energy source. Energy storage systems include that of thermal energy, CO2, compressed air, hydrogen storage, natural gas and underground pumping of water.

Rocks consisting geological formations must fulfill certain criteria to be considered as a candidate reservoir for potential thermal-energy storage (TES), compressed air energy storage (CAES) and carbon capture and storage (CCS) applications. These criteria have been noted by researchers (e.g., [28–30]) stating that rocks should display high values of thermal conductivity, specific heat capacity, and density to enable high storage efficiency. Low porosity values correlate positively with high values of bulk density and uniaxial compressive strength, which are necessary to ensure not only the optimum energy storage criteria but to avoid fracture development and disintegration [28]. Research conducted by Allen et al. [31] and Tiskatine et al. [32] suggest that formations consisting of sandstones may serve as proper energy storage reservoirs, provided that they meet compositional (e.g., calcium-or silica-rich), textural and structural and also not having been significantly affected secondary alteration processes.

Aim of this study is to highlight the effect of petrographic characteristics of sandstones from Klepa Nafpaktias (central western Greece) as a decisive factor in the final strength of the produced concrete specimens by sandstones aggregates and also to examine their potential use as geological reservoir for carbon capture and storage (CCS) applications.

#### **2. Geological Setting**

The study area is Klepa Nafpaktias that geographically belongs to the regional unit of Aitolia and Akarnania and geologically to the Pindos Geotectonic Zone, which comprises an intricate thrust belt with allochthonous Mesozoic and Tertiary deep-water tectono-stratigraphic units [33], which are developing along the central Western Greece (Figure 1) and extend into Albania and former Yugoslavia to the north [34] and Crete, Rhodes [35] and Turkey [36,37], toward the south and southeast. The Pindos Zone sedimentary sequence was deposited in an extended oceanic rift related basin that likely was created during the Middle Triassic era, located in the Apulia extensive marginal platform [38] and more specifically in the Gavrovo-Tripolis sedimentary platform [33,39] that emerged periodically, and now lies westwards of Pindos, and the Pelagonian continental block in the east [40–43]. The progressive closure of the Pindos oceanic basin were initiated during the end of Maastrichtian, as recorded by the gradual alteration, from predominantly carbonates intercalating with radiolaria to siliciclastic/turbiditic lithofacies (Paleocene flysch deposition) derived from the north and east sectors [33,39,44]. The complete closure of the Pindos Ocean during the Eocene led to the detachment of the deep-sea sedimentary cover from the accretionary oceanic prism, which was later overthrusted in a westward direction onto the Pelagonian geotectonic platform, forming multiple layers of thrust sheets [38,42].

The sedimentary alternating strata of Pindos Zone consist of deep-water carbonate, siliciclastic and siliceous rocks of Late Triassic to Eocene age [38,39,42], mainly including the following units (Figure 1): (1) Ophiolite complex of Pindos (Jurassic age); (2) Limestones of Orliakas (Late Cretaceous age); (3) A Mélange formation, namely that of the Avdella (Late Triassic to Late Jurassic age), (4) the Sediments of deep water type, namely that of Dio Dendra (Late Jurassic to Late Cretaceous age); and (5) a thick flysch formation assigned to the Pindos zone (Late Cretaceous to Tertiary age) according to Jones et al. [45]. More extensively, the Pindos flysch consists of thin- to thick-bedded sandstones and mudstones in alteration with marly-oolitic limestones and cherts (reference). According to Konstantopoulos and

Zelilidis [46], the sandstones of the Pindos Flysch were possibly cited within an accretionary prism located close to an active marginal basin, supplied by predominantly basic/ultra-basic and less felsic material. Furthermore, Faupl et al. [47] conducted a heavy mineral examination, suggesting that the clastic material of the Pindos flysch has an eastwards origin, while a petrographic and geochemical study by Vakalas et al. [48] on sandstone samples from Epirus and Akarnania regions suggests a granitic source and a supply from the Pelagonian Zone correspondingly. Detrital modes of sandstone suites reveal the lithological composition of source terranes and the tectonostratigraphic level reached by erosion in space and time.

**Figure 1.** Geological map of the Klepa Nafpaktias [49] (Central Western Greece) region (modified after fieldwork mapping by using ArcMap 10.1).

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

#### *3.1. Materials*

Ten samples from different type of sandstones (coarse-grained and fine-grained) were collected from the studied area. These samples were tested for their petrographic characteristics, physicomechanical and physicochemical properties in order to be classified for their suitability as concrete aggregates. Normal Portland cement (CEM II 32.5N) was used in this study, which conformed to EN 197-1 [50] standard with the collected aggregates in order to produce concrete. For the mixing as well as for the curing of concrete, a potable water with pH = 7, characterized as free of clay, salt, organic matter and silt was chosen to be used. The same volume of aggregate per m<sup>3</sup> of the mixture was retained for keeping consistent composition in all the produced specimens. The proportions used for cement, aggregates and water ratio were those of 1/6/0.63. The same collected sandstones were also investigated for their potential use as reservoir for thermal-energy storage (TES) and compressed air energy storage (CAES) applications.

#### *3.2. Methods*

#### 3.2.1. Rock Material Tests

Polished thin sections of the collected sandstones were first tested via a petrographic microscope (EN 932-3 [51]) for identifying their mineralogical and textural characteristics. The used microscope was of the type of Leitz Ortholux II POL-BK Ltd., Midland, ON, Canada one. In a further stage, their petrographic characteristics were investigated as well as the quantification of their mineralogical composition was calculated using the ArcMap 10.1 software, in which six representative thin sections of the studied groups (two per sample) were investigated.

Secondary electron images (SEI) were used for the identification of the microtopography of the aggregates (BS 812 Part 1 [52], whereas six qualitative categories were outlined (porous, honeycomb, crystalline, rough, granular, smooth and glassy).

A hammer was used in order to crush the studied sandstones into smaller pieces. Then, a laboratory jaw crusher was used for extra fractioning of the samples. For the preparation of the cylindrical specimens with diameters varying from 50 to 54 mm and with ratio of length to diameter varying from 2.2 to 2.4 mm, laboratory core drill combined with saw machines were used.

The investigated physicomechanical and physicochemical properties of this study were the total porosity (nt) (ISRM 1981 standard [53]), the magnesium sulfate (MgSO4) (S) (EN 1367-2 [54]), the water absorption (wa) (EN 1097-6 [55]), the resistance in abrasion and attrition-Los Angeles (LA) (ASTM C-131 [56]), and the uniaxial compressive strength (UCS) (ASTM D 2938-95 [57]), properties crucial for further quality of the aggregates.

#### 3.2.2. Concrete Tests

Twenty normal concrete cube specimens having dimensions of 150 × 150 mm were produced by the ten investigated sandstone aggregates (ACI-211.1-91) (Table 1) [58]. For all the produced concrete specimens, the used parameters retained fixed. The aggregates were crushed and sieved via standard sieves and separated to achieve the sizes: 2.00–4.75, 4.45–9.5 and 9.5–19.1 mm. After 24 h, the samples were transferred from the mold and for 28 days they were cured in water at 20 ± 3 ◦C. A compressive testing machine, having an increasing rate of load of 140 kg/cm2 per minute, was used in order to calculate the concrete strength by the division of the value of the load at the moment of failure over the area of specimen. The concrete strength was calculated via BS EN 12390-3:2009 standard [59].

In this stage, the examination of the concrete textural features was carried out when using polished thin sections in a petrographic microscope (ASTM C856–17) [60]. A 3D depiction of the petrographic characteristics of the concrete as well as of the studied sandstone aggregates was carried out by the 3D Builder software using thin sections.


**Table 1.** Quantification of the modal composition of the representative investigated groups of sandstones.

#### **4. Results**

#### *4.1. Test Results of Aggregates*

4.1.1. Petrographic Features of Aggregates Using Petrographic Microscope

The studied sandstones derived from Klepa area have been divided according to the petrographic analysis into three district groups. These groups are based on the grain size of the collected sandstones and they were characterized as coarse to fine-grained ones.

Group I: Coarse-grained sandstones (KL.5, KL.9)

These sandstones comprise sub-angular to angular grains (Figure 2a,b). They are generally moderate to poor sorted. The mineralogical composition mainly includes quartz, K-feldspars, plagioclase, calcite, mica and in minor amounts muscovite, chlorite and biotite as well as lithic fragments (Table 1). These sandstones present mainly siliceous cement. Quartz is mostly displayed as undulose monocrystalline and less as polycrystalline grains. The monocrystalline quartz grains vary from sub-angular to angular, while the polycrystallines range from sub-angular to sub-round. The grain contacts presented as straight to suture. K-feldspars grains vary in size, from small to large with euhedral to subhedral shape, whereas plagioclase is observed in smaller grains. In general, the fragments are sub-rounded and sub-angular to angular and they are mainly comprised of clasts of quartz, feldspars as well as by rock-fragments of basalt and gabbro. Traces of carbonate fossils are also observed in several samples (e.g., KL.5).

Group II: Medium-grained sandstones (KL.1, KL.2, KL.3, KL.6)

The medium-grained sandstones can be classified as quartz sandstones. They are moderately sorted and their grains are sub-angular to sub-round. The main mineralogical composition includes quartz which forms monocrystalline and polycrystalline grains, K-feldspars, calcite and muscovite (Figure 2c). Polycrystalline quartz shows interlocking texture. Feldspars (mainly microcline) are presented in lesser amounts, including the weathered varieties (Table 1). Cement is mainly siliceous and locally calcareous.

Group III: Fine-grained sandstones (KL.4, KL.7, KL.8, KL.10)

In the fine-grained quartz sandstones, framework grains are mainly sub-angular to sub-round. They are characterized as well sorted quartz sandstones. The modal composition mostly comprises of quartz, K-feldspars, calcite, and mica (mainly muscovite) which is presented in bigger amounts in contrast to the other two groups. Cementing material is mainly siliceous (Table 1).

**Figure 2.** Photomicrograph of textural characteristics of sandstone aggregates (Nicols+) and 3D depiction of the studied sandstones respectively: (**a**) clastic texture presented in a coarse-grained quartz sandstone with quartz (Qz), K-feldspars (K-Fs), plagioclase (Plg), muscovite (Ms) and calcite (Cc); (**b**) 3D depiction of coarse-grained sandstone; (**c**) clastic texture presented in a coarse-grained sandstone containing large particles of carbonate fossils; (**d**) 3D depiction of coarse grained sandstone; (**e**) clastic texture presented in a medium-grained quartz sandstone with quartz (Qz), K-feldspars (K-Fs) and calcite (Cc); (**f**) 3D depiction of coarse-grained sandstone; (**g**) clastic texture presented in a fine-grained quartz sandstone with muscovite (Ms); (**h**) 3D depiction of fine-grained sandstone.

#### 4.1.2. Petrographic Features of Aggregates Using GIS Method

In this paper, GIS method was used as a new approach for petrographic analysis of the investigated sandstones. For this reason, six thin sections, representative of the investigated sandstones (two sections for each group) were used in order to be analyzed via GIS method. More specifically, a part of the same size and in the same site of the thin section has been chosen to be digitized via ArcMap 10.1 software. Each digitized polygon corresponds to a different grain of the sandstone. The result of that process is the creation of a map that shows the modal composition of the tested rocks as well as their textural characteristics such as the grain size (Figure 3). In a next stage, the semi-quantification of the mineralogical composition of the studied sandstones was carried out, showing that Group III presents higher content of quartz than the other two groups, Group I displays intense higher content of feldspars in contrast to Group II and III (Table 1), while Group II is presented as more enriched in calcite (Table 1). Furthermore, Group III displays significant high content of muscovite. Concerning the containing cement, Group III presents higher content of silica cement in contrary to the other two groups (Table 1). After the petrographic analysis via the GIS proposed method, the ratio *C*/*A* was calculated (Table 1). *C*/*A* (crystals/mm2) is the ratio of the sum of the measured crystals to the measured area (mm2). As can be seen in Table 1, Group I, which contains the coarser grains, presents an average of *C*/*A* 11.60 in contrast to samples of Group III which presents values of *C*/*A* from 55.70 to 56.50 and this group consists of the smallest size grains.

**Figure 3.** Representative images from the studied groups derived from the ArcMap 10.1 software showing: (**a**) The part of the thin section of the coarse-grained sandstone which has been analyzed; (**b**) the output after the digitalization of the investigated part of each thin section where each mineral phase has been attributed by different colors; (**c**) the part of the thin section of the medium-grained sandstone which has been analyzed; (**d**) the output after the digitalization of the investigated part of each thin section where each mineral phase has been attributed by different colors; (**e**) the part of the thin section of the fine-grained sandstone which has been analyzed; (**f**) the output after the digitalization of the investigated part of each thin section where each mineral phase has been attributed by different colors.

#### 4.1.3. Microtopographic Characteristics of the Tested Sandstones

The surface textural characteristics of the examined sandstone particles were used to classify the quartz sandstones in groups consistent with the above mentioned Groups I to III, as it is shown in Figure 4. The abundance of the coarse size grains of quartz, feldspars and carbonate fossils in the poor sorted sandstone (Group I) is responsible for the smooth surfaces. Sandstones of Group, II constituting medium-grained rocks, are characterized by rough surface texture. The surface of Group III samples can be characterized as rougher in contrast to the other two groups because of the existence of higher content of evenly distributed mica and quartz expressing topographic low areas combined with feldspars which express lower topographic areas.

**Figure 4.** Secondary electron images (SEI) depicting the microtopography of representative sandstone samples observed to their textural and mineralogical characteristics: (**a**) coarse-grained; (**b**) medium-grained; (**c**) fine-grained sandstone.

#### 4.1.4. Physicomechanical Properties of Aggregates

Three distinct groups of sandstones were determined following the results of their physicomechanical properties (Table 2). The calculated mechanical as well as the calculated physical results of the examined rocks present quite wide range. These results have classified the investigated sandstones into groups which are in agreement with those derived from their petrographic study. Group I includes coarse-grained sandstones, which presented the worst values of mechanical properties among all the studied rock samples. Among the examined sandstone aggregates of Group I, sample KL.5, owing to its lower content in quartz, led to a lowest value of total porosity (nt) as well as in resistance in abrasion. Group II was composed of medium-grained sandstones presented variable physicomechanical performance relative to their mineralogical composition. Group III included fine-grained sandstones, which displayed high physicomechanical parameters among all the determined groups. The fine-grained sandstones, such as KL.4 and KL.8, presented better mechanical characteristics against the coarse-grained sandstones such as KL.5 and KL.9.


**Table 2.** The results of the physicomechanical properties of the tested sandstones.


**Table 2.** *Cont.*

#### *4.2. Concrete Test Results*

#### 4.2.1. Uniaxial Compressive Strength of Concrete

Uniaxial compressive strength test was carried out for the produced specimens, whereas their results can be observed in Table 3. These values varied from 24 to 32 MPa after 28 days of curing. The concrete strength results are in relevant accordance with the strength of their aggregates. Concrete specimens displayed the lowest strength values that have been made by sandstones of Group I as aggregate particles (Tables 2 and 3). More specifically, the coarse-grained sandstone that contains carbonate fossils of big size displays the worst strength value (24 MPa), value which is below the permitted limit for the concrete class C25/30. The concrete specimens, made by medium grained aggregates from Group II, displayed variety on strength values (26 to 31 MPa), whereas on the other hand those made with the finer grained aggregates from Group III showed the highest compressive strength values (30 to 32 MPa).

**Table 3.** Uniaxial compressive strength test results of the concrete specimens.


4.2.2. Petrographic Characteristics of the Investigated Concretes

Careful microscopic analysis of the thin sections of the tested concrete specimens which were studied by using polarizing microscope as well as through the 3D depiction of the same thin sections showed, in general, satisfied cohesion between the aggregate particles and the cement paste among all the concrete specimens produced by the coarse-grained, the medium-grained as well as the fine-grained sandstones (Figure 5). The existence of intense content of silica cement may enhance the bonding between the sandstone aggregates and the cement paste. Even neither in concrete specimens produced by aggregates of Group I nor in those produced by aggregates of Group II and III, loss of material can be observed nor extensive interaction zones.

**Figure 5.** Photomicrographs of representative tested concrete specimens produced by: (**a**) Coarsegrained sandstone aggregate; (**c**) medium-grained sandstone aggregate; (**e**) fine-grained sandstone aggregate and 3D depiction of representative tested concrete specimens produced by: (**b**) coarse-grained sandstone aggregate; (**d**) medium-grained sandstone aggregate; (**f**) fine- grained sandstone aggregate.

#### **5. Discussion**

#### *5.1. The Impact of Petrographic Characteristics on the Sandstone Aggregate Properties and on the Quality of Concrete*

Petrographic characteristics such as mineralogical composition, texture, particle size, alteration and weathering degree of rocks which are used as aggregate materials, constitutes the main factors influencing their properties that are critical for their suitability in various construction and industrial applications [61]. Numerous researchers have correlated physical and mechanical properties of rocks used as aggregates [62–64] giving clear interpretations of the relationships between them which are based on the type of the contained minerals and on their size. Petrounias et al. [65] have proved that the type of the secondary phyllosilicate minerals contained in mafic, ultramafic and intermediate and acidic volcanic rocks is the critical factor that predominately determines the physico mechanical properties of the studied rocks. Additionally, Tugrul and Zarif [66] have reported that the mean grain size is presented as a primary factor influencing the mechanical behavior of the granites which are used as concrete aggregates. More specifically, they have proved that as the mean grain size decreases, the strength of the rock increases respectively. The statistical method that is widely used to determine the relationships between different engineering parameters of rocks is the regression analysis [66–68]. In this paper, where sandstones from Klepa Nafpaktias were studied, strong relationships between the physical and mechanical parameters as well as among mechanical, physical, and physicochemical ones were observed using regression analysis. These correlations as they presented in Figure 6 are mainly dependent on the grain size and lesser on the mineralogical composition and on the amount of the cement. The diagrams of Figure 6 indicates that as the grain size of the investigated sandstone increases, the values of their physical properties increase while the values of their mechanical properties decrease respectively. For example from the diagram of Figure 6a we can observe that Group I, as it is classified after petrographic analysis through the petrographic microscope and verified after the new proposed petrographic analysis via GIS method and is characterized as the more coarse-grained group, presents ratio *C*/*A* 11.60 on average and higher values of porosity (nt) (Table 2) and lower resistance in abrasion and attrition (LA). Likewise in the diagram of Figure 6b Group I presented as more capable to absorb water (wa) and with lower values of uniaxial compressive strength (UCS). Diagrams of Figure 6a,b show the interaction between the physical and mechanical properties which are directly depended on the grain size of the similar mineralogical composition tested sandstones. The lower value of the mechanical strength of the coarse-grained sandstones may be a result of the low, and maybe because of the microtopography, internal attrition between the grains combined with the small percentage of cement, which leads to small angles of attrition relative to the density.

**Figure 6.** (**a**) LA vs. total porosity (nt) of rocks; (**b**) compressive strength (UCS) vs. water absorption (wa) of rocks; (**c**) LA vs. the compressive strength (UCS) of rocks; (**d**) water absorption (wa) vs. the soundness test (S) of rocks.

In contrast to the comparatively finer grained sandstones, molecular internal forces are developed during the loading presenting better cohesion and bonding among the grains. Furthermore, the porosity (nt) as well as the water absorption (wa) seems to significantly be increased in the coarse sandstone rocks against the fine ones, which indicates that larger grains exhibiting weaker cohesion in contrast to the smaller are capable to adsorb more percentage of water around each grain mainly in the form of a

surface layer. The diagram in Figure 6c illustrates the interaction of the mechanical properties LA and UCS directly dependent on the grain size. Several researchers have also reported similar relationships between these properties [23,65,69,70] when studying various types of rocks. In this diagram, it seems obvious that the mechanical characteristics of sandstones vary in a similar way under different type of mechanical loadings. For example, rocks of Group I (coarse-grained) present lower resistance in abrasion and attrition and lower compressive strength in contrast to those of Group III (fine-grained) which presented more resistance in abrasion and attrition and with higher strength values. In the diagram of Figure 6d, the relationship between the Soundness test (S) and the water absorption (wa) is presented, the trend of which is similar to other reported relationships between the Soundness test and physical properties by several researchers [61]. The interpretation given above regarding the ability of coarse-grained sandstones to adsorb water in their structure in contrast to the fine-grained sandstones has a strong effect on their resistance to temperature changes. All of the above theories regarding the effect of grain size on the physicomechanical properties of rocks are quantified and presented below in Figure 7. More specifically, the quantification of the number of minerals per mm2 (*C*/*A*) (calculated via GIS) seems to be strongly correlated with the physicomechanical properties of the sandstones. In Figure 7a, it can be seen that as the *C*/*A* increases, the strength of the rocks increases (Figure 7a) and their resistance in abrasion and attrition increases respectively (Figure 7b), whereas the number of minerals per mm<sup>2</sup> increases as their porosity decreases (Figure 7c). It is noticeable that the above mentioned relationships display high coefficient of correlation (R2 = 0.72 and R<sup>2</sup> = 0.71) (Table 4) a fact that shows that the grain size constitutes the principal but not the unique factor that influences these properties. This happens because the mineralogical composition of rocks also determines their physicomechanical properties.

**Figure 7.** (**a**) compressive strength (UCS) of rocks vs. the ratio *C*/*A*; (**b**) LA of rocks vs. the ratio *C*/*A*; (**c**) total porosity (nt) of rocks vs. the ratio *C*/*A*.

The results from the investigated sandstone concretes shows that they present satisfactory values of compressive strength (24.00 to 32.00 MPa) relative to other concrete specimens that have been made by andesites and serpentinites as aggregate particles [22]. These satisfactory strength results may attributed to the generally high microtopography of coarse-grained, medium-grained, as well as of fine-grained sandstones relative to the microtopography of other used rocks [22] (Figure 2b,d,f,h and Figure 4). The microtopography of the aggregates constitutes a crucial factor for the mechanical quality of the aggregate rocks and consequently for the quality of the produced concrete as it influences the cohesion and the bonding between the cement paste and the aggregate particles [1,22,23]. The only studied concrete specimen that displays lower strength (25 MPa) than the standard states is the specimen in which the used aggregate was the enriched in carbonate fossils coarse grained sandstone (Figure 2c). This fact may be the result of the lower resistance of the fossils which tend to be broken during the mechanical loading, combined with the low microtopography which they promote (Figure 3d). However, although all the investigated concrete specimens revealed satisfactory strength results, small differences in their values appeared to be dependent on the grain size of the sandstones. The diagrams in Figure 8 indicate that the aggregate properties, which are determined by the size and the number of the grains, as it is shown in Figure 7, determine the final strength of the produced concrete specimens.

**Figure 8.** (**a**) Total porosity (nt) of rocks vs. concrete strength (UCScon); (**b**) compressive strength of rocks vs. concrete strength (UCScon); (**c**) water absorption (wa) of rocks vs. concrete strength (UCScon); (**d**) LA of rocks vs. concrete strength (UCScon).


**Table 4.** Correlation equations of diagrams of Figures 6–8.

During the petrographic analysis of the tested concrete specimens, no significant failures and loss of material were observed in those produced by coarse-grained sandstones and nor extensive reaction zones, which typically occur in igneous high porosity aggregates. One possible interpretation that can be attributed is that the lower mechanical strength of concrete aggregates may depend on the higher porosity of the coarse-grained sandstones in contrast to the fine-grained ones (Table 2) which result in the greater adsorption of water which is useful during the 28 days of curing for the achievement of the optimum cohesion between the cement paste and the aggregate particles. However, such extensive areas of incomplete hydration of the cement around the grains were not observed during petrographic examination of the concrete using polarizing microscope. This may have happened because of the even distribution of the mineralogical composites of rocks as can be seen in the 3D depiction via GIS. This resulted in the even adsorption of water and consequently these zones cannot be easily perceived.

#### *5.2. A Potential Scenario for Storage of CO2 in Sandstones from Klepa Nafpaktias*

The studied area presents an appropriate geological basin environment for applying CO2 capture and storage (CCS) applications. It is well-known that the permeability of flysch formations is regarded as being generally low because of the presence of marl and clay intercalations within this type of formation. This practically impermeable sedimentary formation lies stratigraphically above the sandstones, thus providing an excellent seal caprock to keep the buoyant CO2 within the reservoir rock. This case presents many similarities with that described for the Mesohellenic Trough (NW Greece), which examined the potential of CO2 storage within porous sandstones that are overlaid be a less permeable cap rock formation [71–73]. In the latter case, a depth of over 800 m was regarded as suitable for trapping CO2 under supercritical conditions [73–75]. The sandstone samples provided from our study are highly comparable in terms of composition with sandstones from the Pentalofos formation of the Mesohellenic Trough [72]. Petrographic and mineral modal examinations reveal that the sandstones (Groups I–III) from Klepa Nafpaktias display the following modal compositions: Quartz = 24–29%; K-feldspar = 7–29%; Calcite = 1–8%; Muscovite = 1–4%; Plagioclase~0.5%; Siliceous and Calcite-bearing Cement = 42–57% (Figure 2, Table 1).

These results show that the sandstones examined include relatively higher quartz contents and less calcite compared to those located in the Mesohellenic Trough [73]. Furthermore, effective porosity of the Klepa Nafpaktias sandstones, as it was calculated through the total porosity, which is about 6% for the Group I presents higher values of effective porosity in contrast to the other two sandstone groups and tend to be lower than the Pentalofos sandstones of the Mesohellenic Trough ~ 9%.

Despite the relatively smaller storage potential presented in the region of Klepa Nafpaktias, the rather higher silica contents offers the ability of avoiding undesirable fracture development and disintegration phenomena. This is because CO2 is expected to react with calcite hosted within the sandstones; however, this would result in the formation of unstable bicarbonates, which would hinder their ability for permanent CO2 storage. Recent studies on CO2 geological storage within sandstone formations reveal the importance of feldspar and plagioclase minerals for permanent CO2 trapping [75–78].

The mineralogical composition of the studied sandstones of Group I as well as their general petrographic characteristics enhances their capacity for CO2 storage as the sufficient amounts of K-feldspars can react with injected supercritical CO2 with the following reactions (1)–(4):

$$2\text{KAlSi}\_3\text{O}\_8\text{ (K-feldspar)} + \text{CO}\_2 + 2\text{H}\_2\text{O} \Rightarrow \text{Al}\_2\text{(Si}\_2\text{O}\_5\text{)(OH)}\_4\text{ (kaolinite)} + 4\text{SiO}\_2 + \text{K}\_2\text{CO}\_3 \tag{1}$$

$$\text{\textbullet\text{KAlSi}\_3\text{O}\_8 \text{ (K-feldspar)} + \text{H}\_2\text{O} + \text{CO}\_2 \Rightarrow \text{KAl}\_3\text{Si}\_3\text{O}\_{10}\text{(OH)}\_2\text{(illite)} + \text{\textbullet\text{SiO}\_2} + \text{K}\_2\text{CO}\_3\tag{2}$$

$$2\text{NaAlSi}\_3\text{O}\_8\text{ (altitude)} + 2\text{CO}\_2 + 3\text{H}\_2\text{O} \Rightarrow \text{Al}\_2(\text{Si}\_2\text{O}\_5)(\text{OH})\_4\text{ (kaolinite)} + 4\text{SiO}\_2 + 2\text{Na}^+ + 2\text{HCO}^- \tag{3}$$

$$\text{NaAlSi}\_3\text{O}\_8\text{ (altitude)} + \text{CO}\_2 + \text{H}\_2\text{O} \Rightarrow \text{NaAlCO}\_3\text{(OH)}\_2\text{ (dawsonite)} + \text{3SiO}\_2\tag{4}$$

Thus, the dissolution of alkali feldspars will lead to the precipitation of clay minerals and silica (in the form of quartz). Plagioclase, although present in smaller amounts, is also expected to produce kaolinite, as well as calcite through the successive reactions (5) and (6):

$$\text{CaAl}\_2\text{Si}\_2\text{O}\_8\text{ (anorthite)} + 2\text{CO}\_2 + 3\text{H}\_2\text{O} \Rightarrow \text{Al}\_2(\text{Si}\_2\text{O}\_5)(\text{OH})\_4(\text{kaolinite}) + \text{Ca}^{2+} + 2\text{HCO}^- \tag{5}$$

$$\text{CaAl}\_2\text{Si}\_2\text{O}\_8\text{(anorthite)} + \text{H}\_2\text{CO}\_3 + \text{H}\_2\text{O} \Rightarrow \text{Al}\_2\text{Si}\_4\text{O}\_5\text{(OH)}\_4\text{(kaolinite)} + \text{CaCO}\_3\tag{6}$$

We provide preliminary calculations that estimate the CO2 that could be stored in the frames of a potential pilot project in the studied region. For this aim, we implement this function:

$$\text{CO}\_2\text{Storage Capacity} = \Sigma(\text{V} \times \phi \times \text{R}\rho \times \varepsilon)$$

With V symbolizing the sandstone reservoir volume (under the flysch cap rock); φ denoting the effective porosity; ρ specifying the sCO2 specific gravity properties; whereas ε stands for the sCO2 storage ratio capability. We undertake a likely scenario of a test pilot project, in which we assume a volume of 5000 m (the length) <sup>×</sup> 3000 m (the width) <sup>×</sup> 500 m (the depth) = 75 <sup>×</sup> 108. Based upon the estimations of Jin et al. [75] and with reference to the statistical values of USGS modeling, we can consider the CO2 storage ratio for sandstones to be 1%. The application of this discount factor is necessary in order to obtain a realistic estimation of the sandstone reservoir storage potential. Taking the aforementioned value into consideration, as well as additional parameters that include the average sandstone effective porosity values from our studied site (6%) and the specific gravity considered for CO2 under supercritical conditions (a value of 400 kg/m<sup>3</sup> in pressure and temperature conditions of 10 MPa and 50 ◦C respectively [79]), it is assessed that the demarcated area could potentially store an amount of 18 <sup>×</sup> 105 tons of CO2.

We also consider Equation (12) of Jin et al. [75] to calculate the quantity of CO2 trapped by feldspars (K-feldspar and plagioclase minerals, where these amounts K-feldspar = 23–34%; Plagioclase = ~1% resulted from the reduction of the initial amounts without the cement):

$$m\text{CO2 Feld} \text{spar} = \left[ p\_{\text{Feldspar}} \times V \times (1 - \varphi) \times X\_{\text{Feldspar}} \times M\_{\text{CC2}} \times R \right] / M\_{\text{Feldspar}}$$

With V symbolizing the sandstone reservoir volume, φ denoting the effective porosity, *p*Feldspar expresses the feldspar density (2.55–2.67 <sup>×</sup> <sup>10</sup>3, 2.55–2.60 <sup>×</sup> <sup>10</sup><sup>3</sup> and 2.75–2.76 <sup>×</sup> 103 kg/m3 for K-feldspar, albite and anorthite respectively), *M*Feldspar is molecular weight (279.07, 262.96 and 278.94 for K-feldspar, albite and anorthite respectively) *R* is the ratio of feldspar mineral to CO2 0.5, 1 and 1 for K-feldspar, albite and anorthite respectively), *X*Feldspar the proportions of feldspar minerals, *M*CO2 is the total CO2 storage capacity of mineral trapping. By applying this equation upon alkali feldspars and plagioclase the results calculated for the CO2 that can be permanently trapped within the sandstone formation is ~6 <sup>×</sup> 105 tons, which is less by almost approximately 1/3 of the storage potential calculated with the previous method. This is due to the fact that the latter equation does not consider calcite crystallization as a stable mineral phase. Nevertheless, considering both cases, it is evident that the sandstones of the Klepa Nafpaktias region are capable of storing sufficient amounts of CO2. This is even more evident taking into consideration that region's sandstones and flysch formations encompasses an even wider area and thus could allow for the deployment of larger-scale CO2 storage projects, provided that the proposed pilot test is deployed successfully.

#### **6. Conclusions**

In this paper, sandstones of various petrographic characteristics derived from Klepa Nafpaktias were examined in order to evaluate their suitability in construction (concrete) and environmental applications (CO2 storage). For the first time, the petrographic study of rocks such as of those sandstones were carried out by using classic petrographic methods (observation through polarizing microscope) combined with modern tools of quantification of modal composition (GIS proposed method) and 3D depictions of their petrographic features (3D Builder software). The above mentioned study leads to the following concluding remarks:


**Author Contributions:** P.P. was involved in the field data collection, participated in most of the laboratory testing's, result interpretation and paper writing; P.P.G. was involved in the field data collection, participated in most of the laboratory testing's, and contributed to paper writing; A.R. was involved in the field data collection, interpretation, and assisted in the paper writing; M.K. was involved in the field data collection, modified the geological map and performed the GIS analysis; P.K. participated by conducting data interpretation and assisted in part of the paper writing; M.-E.D. was involved in the field data collection and in some of the of the laboratory testing's and N.K. participated by conducting data interpretation and assisted in the paper writing. All authors have read and agreed to the published version of the manuscript.

**Funding:** Our research study has not received any form funding from external sources.

**Acknowledgments:** We would like to thank A.K Seferlis for his help concerning the SEM performed work (Lab of Electron Microscopy & Microanalyses, University of Patras).

**Conflicts of Interest:** No conflicts of interest are declared.

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Article* **Acid-Catalyzed Wet Torrefaction for Enhancing the Heating Value of Barley Straw**

#### **Antonios Nazos 1, Panagiotis Grammelis 2,\*, Elias Sakellis <sup>3</sup> and Dimitrios Sidiras <sup>1</sup>**


Received: 25 February 2020; Accepted: 24 March 2020; Published: 3 April 2020

**Abstract:** In the present study, the possibility of improving the higher heating value (HHV) of lignocellulosic biomass, especially barley straw, was examined. The research deals with the treatment of barley straw by acid-catalyzed wet torrefaction (ACWT), also called acid hydrolysis, in a batch reactor (autoclave) Parr 4553 3.75 L. In this case, two different simulation approaches were applied: (i) combined severity factor (CSF) and (ii) response surface methodology (RSM) based on Box–Behnken design of experiments (DoE). Sulfuric acid (SA) concentration, temperature and time were the ACWT parameters examined herein. An oxygen bomb calorimeter was used for the HHV measurement. The findings indicated that the composition changes of the straw due to ACWT had a significant effect on the HHV of the pretreated material. In this study, treatment conditions were 10–35 mM SA, 160–200 ◦C and an isothermal reaction time 0–40 min (preheating period not included in these values). In conclusion, there was a significant increase in the HHV up to 24.3 MJ/kg for the ACWT barley straw, compared to 17.5 MJ/kg for the untreated straw, at optimal conditions of 200 ◦C for 25 min (isothermal period) and 35 mM SA. This resulted in a 1.39 enhancement factor (EF) and 68% energy yield (EY).

**Keywords:** acid-catalyzed wet torrefaction; acid hydrolysis; barley straw; combined severity factor; enhancement factor; energy yield; higher heating value; response surface methodology

#### **1. Introduction**

The increase in worldwide energy demand has a significant effect on the fossil fuel contribution to environmental pollution and climate change, provoking a global interest in the use of environmentally friendly, renewable fuels [1,2]. Renewable biomass is one of the emerging energy resources with high potentials that can balance CO2 emissions [3]. Lignocellulosic biomass is currently a major energy source for fossil fuel partial substitution [4–6]. Biomass contributes about 10% of the global annual energy production [7,8]. Thus, biomass attracts considerable research interest in order to meet the increasing future sustainable energy demand [9]. Biomass is widely available in nature and accounts for around 100 billion tons per year [6]. Heat, electricity, fuel, chemicals and other high added value products can be produced from biomass [7,9].

Moreover, biomass disadvantages include heterogeneity in structure, low density, high humidity, low higher heating values (HHVs) and flexibility, all of which limit the use of biomass as a fuel [7,10]. These disadvantages make the production of heat and energy from biomass more complicated. They result in the increased cost of processing, transporting and storing of biomass fuels. Consequently, we need

to process the fuel and upgrade it through processes like the torrefaction process. Combined with understanding and studying the mechanisms of the biomass reaction through the torrefaction process, it will be easier to develop the technologies for thermal utilization of biomass [7,11–15].

There are mainly two torrefaction categories in the recent literature: dry torrefaction (DT) and wet torrefaction (WT). Both these techniques produce biomass fuels with improved chemical and physical qualities compared to the untreated biomass characteristics [7]. No references were found with regard to acid-catalyzed wet torrefaction (ACWT) but it could be considered as a subcategory of WT which uses, e.g., SA as catalyst (not autocatalyzed by organic acids produced during WT). On the other hand, acid hydrolysis was examined for production of fermentable to bioethanol sugars from lignocellulosic biomass [16,17]. However, DT has been extensively studied for decades and significant references appear in the international literature [13–15] while WT appeared later, and is associated with concepts such as autohydrolysis, hydrothermal carbonization, hydrothermal pretreatment, etc. In recent years, research on WT has increased compared to investigation on DT [7,18,19].

WT is the treatment of biomass with hydrothermal media at 180–260 ◦C with pressures at 0.9–4.6 MPa, respectively [4,7,20,21]. WT seems to be a promising technology to produce low cost high quality solid biomass fuels from agricultural wastes. WT is synonymous with "hydrothermal carbonization" or "hydrothermal conversion" [22] or "hydrothermal treatment" [23]. Sometimes WT and hydrothermal carbonization terminologies are used alternatively. Although WT is applied to produce advanced solid fuels, hydrothermal carbonization is primarily used for coal production. WT gives fuels with an increased HHV, as well as low carbon emissions, which can also be considered as moderate biomass pyrolysis at 200–300 ◦C [24]. WT produces solid fuels with enhanced properties at comparatively milder conditions with regard to time and temperature [25]. Furthermore, WT was considered to be a low cost and thus cost-effective prefabrication method with promising equipment corrosion-limiting properties and simple operation [4,7]. After All, WT or hydrothermal pretreatment has attracted an interest, being an effective way to convert high moisture biomass [26]. Recently, Gan et al. [27] developed microwave-assisted ACWT (using various acids as catalysts) of microalgae for simultaneous production of char and sugars.

According to He et al. [4], issues associated with operating cost, pollutant emissions, re-design of industrially scale reactor, and system integration with downstream applications must be solved in order to make WT environmentally and commercially sustainable. The most critical issues deal with (i) requirements of reactor materials withstanding high temperature, moderate pressure and severe corrosion; (ii) clogging issues caused by inorganic precipitates; (iii) post-treatment of wastewater and downstream application of WT material; and (iv) heat recovery to minimize energy consumption.

Furthermore, in ACWT the recovery and recycling of the used acid (e.g., sulfuric acid) is crucial. According to Gan et al. [27] the corrosive nature of sulfuric acid requires a high capital cost for the pretreatment reactor, as well as high safety and handling measures due to using acid. Moreover, the generated byproducts inhibit bioethanol production from the sugars in the ACWT liquid phase and requires costly downstream waste treatment.

In our recently published work [28], the effect of DT in a muffle furnace on barley straw HHV was simulated using the combined severity factor (CSF) methodology (incorporating the effect of reaction time and temperature). In this study, in an effort to optimize the HHV of barley straw, the feasibility of ACWT was examined. Thus, the subject of this research is the modification of barley straw by acid hydrolysis using a batch reactor (autoclave). The pretreatment conditions were time, temperature and acid concentration. Moreover, two approaches were used to simulate the HHV enhancement during the performed experiments. These approaches were (i) the response surface methodology (RSM) based on a Box–Behnken design of experiments (DoE) and (ii) CSF. The HHV of the ACWT barley straw was compared to that of the untreated material. Afterwards, the ACWT conditions were optimized for maximizing the HHV of modified barley straw.

#### **2. Materials and Methods**

#### *2.1. Material Development*

The experiments performed using barley straw, originated in the Kapareli village of Thebes, Greece (38◦14 8" N 23◦12 59" E). The original straw was manually cut to particles of 2 and 3 cm. The specific fraction was considered to be suitable because in this way homogeneity could be achieved when the ACWT procedure was over. The untreated straw moisture was 6.0% *w*/*w* measured according to the procedure UNE-EN ISO 18134-1: 2015. Moreover, the barley straw used for the experiments in this work was the same raw material used in an earlier study of our team on DT [28].

#### *2.2. ACWT Process*

To carry out the ACWT treatment, a Parr 4553 batch reactor 3.75 L was used, being capable of reaching working temperatures up to 250 ◦C, while the planned experiments require a constant prevailing temperature of 160 or 180 or 200 ◦C. For each temperature level, there was a set of experiments which combine different acid concentration and reaction time. More specifically, the different SA concentrations tested herein were 0.01, 0.0225 and 0.035 M over 0, 20 or 40 min of reaction time each (isothermal reaction periods, not including the preheating and cooling period). When the sample has remained in the reactor vessel for the planned time, the cooling process kicks in, cooling the vessel to approximately 25 ◦C. It should be noted that throughout each experiment, the liquid/solid solution was stirred at 150 rpm, while the vessel preheating and the duration of the cooling process is neglected. The ACWT process was followed by a separation and drying process. In particular, the samples were extracted from the liquid solution using a No. 3 Whatman filter, washed and finally dried at 105 ◦C using an oven. The solid residue yield (SRY) was estimated by comparing the substrate weight, before and after the ACWT process, i.e., as the fraction substrate used for torrefaction.

#### *2.3. Bomb Calorimeter*

A Parr 1341 Plain Jacket Bomb Calorimeter was used for the HHV measurements. The original barley straw and modified barley straw samples produced by ACWT were tested for HHV determination according to the method ISO 1928:2009 "Solid mineral fuels—Determination of gross calorific value by the bomb calorimetric method and calculation of net calorific value". More details are given in a recent work of our team [28].

#### *2.4. Proximate and Ultimate Analysis*

The moisture content, the volatile matter (VM), the fixed carbon (FC), the ash, the carbon, the hydrogen, the nitrogen, the sulfur and the oxygen (by difference) of untreated and ACWT pretreated samples was determined by proximate and ultimate analysis.

#### *2.5. Scanning Electron Microscopy*

Scanning electron microscopy (SEM) was used to examine the surface topology changes of the untreated barley straw compared to the ACWT treated barley straw. For this purpose, a FEI INSPECT SEM equipped with an EDAX super ultra-thin window analyzer for energy-dispersive X-ray spectroscopy (EDS) was used.

#### *2.6. Combined Severity Factor*

In order to incorporate the effect of reaction conditions (SA concentration, time and temperature) into a single ordinate we applied the CSF. The CSF concept was based on the "*P* factor" or "reaction ordinate" presented by Brasch and Free [29] in 1965 and Overend and Chornet [30] in 1987:

$$P\text{ factor} = \left[\exp((T - 100)/14.75\right] \cdot t\tag{1}$$

where *T* represents the temperature and *t* is the time of the procedure.

In 1990, Chum et al. [31], and in 1992, Abatzoglou et al. [32] expanded this abovementioned formula, including the effects induced by how acidic or basic the solution is, according to the following equation:

$$R\_0' = 10^{-\text{pH}} \cdot t \cdot \exp[\left(T - 100\right)/14.75] \tag{2}$$

The CSF concept enables the comparison and\or unifications of different pretreatments. CSF has been widely used. For instance, it was used on acid hydrolysis (used as a first step before enzymatic saccharification) for softwood and corn stover samples, in the study Lloyd and Wyman [33], as well as for samples of wheat straw by Kabel et al. [34]. However, the latter study is the only one characterized by low CSF, to the best of our knowledge. In particular, CSF in the study of Kabel et al. ranges from −1.7 to 1.5, while in contrast Lloyd and Wyman's CSF ranges from 0.4 to 2.7.

CSF also can be used for non-isothermal acid treatment, based on previous studies [35–37], with the major difference from Equation (2) laying in expressing with more detail the change of temperature over time (including preheating and cooling period over the commonly used isothermal period):

$$R\_0^\* = e^{-\text{pH}} \int\_{-0}^{t} \exp(\frac{(T - 100)}{14.45}) dt\tag{3}$$

The pH was measured in the liquid phase derived after ACWT (i.e., acid hydrolysis pretreatment). In the present work, the CSF is *R*∗ <sup>0</sup> as expressed by Equation (3).

#### *2.7. Statistical Method*

As a second method (additional to CSF), the RSM is used to incorporate the effect of the ACWT variables (temperature, residence time and SA concentration) into the modified barley straw HHV. The RSM was based on the Box–Behnken DoE (BBD). This particular design is considered more efficient than other designs, such as central composite design (CCD) and three-level full factorial design [38]. The CCD is a five-level fractional factorial design, which comprises of a two-level factorial design, central designs and two axial designs. On the other hand, the BBD is a spherical, rotatable second-order design. It is based on a three-level incomplete factorial design, which consists of the center point and middle points like the edge of a cube. Although BBD can be derived from a cube, it can be represented spherically, making the vertices of the cube not covered by the design. It can be considered as three interlocking factorial designs along with center points. The BBD is said to be a more economical and viable tool than the CCD, because its design matrix is usually generated with a fewer number of experimental runs. BBD requires an experiment number according to *N* = 2*k*(*k* − 1) + *cp*, where *k* is the number of factors and *cp* is the number of the central points, while CCD requires *N* = *k*<sup>2</sup> + 2*k* + *cp*. More specifically, CCD requires experiment number *N* = 18 while BBD requires only *N* = 15 for *k* = 3 and *cp* = 3.

The BBD method is advantageous over other common experimental or optimization methods since it requires a relatively small number of experiments while it also enables the exploration of the interactive effects over the considered observables, hence enabling to depict the effects of each one [38].

The experiments are planned with Quantum XL (SigmaZone) software and are presented in Table 1. It should be noted that each of the 15 experiments was performed twice, with Table 1 containing the averaged outcomes.


**Table 1.** Box–Behnken experimental design.

Performing a polynomial regression with the method of least squares as described from Box and Wilson [39], enables the prediction of optimum conditions through RSM. More accurately, in order to draw a three-dimensional graph (the RSM) that will facilitate the identification of the optimum data, the most important steps include the determination of influencing variables and their range, the selection of a random point on sample as representative of input, encoding of the variables, implementation of the analysis of variance (ANOVA) technique and the lack-of-fit test.

The experiments outcome (*A*: temperature, *B*: time and *C*: SA concentration) fitted to the following equation:

$$\text{HHV} = a + a\_1A + a\_2B + a\_3C + a\_{\text{D}}A^2 + a\_{22}B^2 + a\_{33}C^2 + a\_{12}AB + a\_{13}AC + a\_{23}BC \tag{4}$$

where *a*, *axx* are the model's constant and coefficients, respectively.

#### **3. Results and Discussion**

#### *3.1. Simulation Based on CSF*

The pH before and after ACWT of barley straw depending mostly on the concentration of SA is given in Table 1. Furthermore, the HHV and the SRY of barley straw after ACWT are presented in Table 2. According to the results, ACWT had a mild effect on barley straw SRY. Moreover, in Table 2, the HHV as affected by the ACWT treatment conditions is presented. These findings indicate that there was a serious increase in HHV for high temperature and the maximum SA concentration while untreated barley straw HHV was measured and its average rate was 17.6 MJ/Kg. This was the average result of two HHV measurements.

The CSF of the present study incorporates the effect of temperature, time and SA concentration. In Figure 1a the reaction temperature profile vs. the reaction time is presented. The isothermal period was 40 min at 160, 180 and 200 ◦C. In Figure 1b the reaction pressure profile vs. time for the same experiments is presented. Table 2 shows the CSF values estimated for the specific ACWT conditions. The CSF in logarithmic form ranged from −0.70 to 2.61, incorporating the effect of temperature, time and SA concentration. Combined increase in temperature, time and SA concentration leads to higher values of the CSF. Relevant research on straw pretreatment [40] shows that severe hydrolytic conditions leads to the degradation of cellulose and hemicelluloses, breaks the lignocellulosic matrix, decreases the ash content and changes the elemental composition of the modified material. In another relevant work, Angles et al. [41] used the CSF (also called gravity factor) for the softwood vapor explosion at 176 to 231 ◦C for 2.5 to 5.5 min hydrolysis. In their work the CSF varied from log*R*<sup>0</sup> \* = 2.6

to log*R*<sup>0</sup> \* = 4.6. Moreover, in the Chornet and Overend [30] study CSF was 2.6–4.2. Decreased sugar yield and degree of polymerization for severe hydrolytic conditions were reported. Toussaint et al. [42] and Heitz et al. [43] proved that a higher reaction temperature and time (i.e., increased CSF) results in enhanced cellulose recovery, accessibility and digestibility in combination with higher lignin removal. This was followed by increased hemicellulose degradation.


**Table 2.** Combined severity factor (CSF) in logarithmic form (log*R*<sup>0</sup> \* ), higher heating value (HHV) and solid residue yield (SRY) of barley straw treated by acid-catalyzed wet torrefaction (ACWT).

**Figure 1.** The ACWT pretreatment's temperature (**a**) and pressure (**b**) profiles vs. time (isothermal period 40 min at 160, 180 and 200 ◦C, respectively).

Demirbas [44] reported that the HHV of a lignocellulosic fuel depends on lignin, cellulose and hemicellulose percentage. He found that holocellulose HHV was 18.60 MJ/kg, while lignin HHV was 23.26–25.58 MJ/kg. Consequently, the HHV of lignocellulosic fuels is improved when the lignin content is increased.

Changes in the elemental composition of lignocellulosics due to ACWT affects the HHV of the solid residue. According to Semhaoui et. al. [45] hemicelluloses solubilization depends on the severity of treatment and becomes significant for positive CSF values. The results of CSF show that the concentration of acid that led to negative levels CSF value was insufficient for hemicelluloses solubilization.

In the present work, the correlation between log*R*<sup>0</sup> \* and SRY % *w*/*w* (dry basis) was found by applying a nonlinear regression, fitting the equation

$$\text{SRY} = 3.857 \text{exp}(-1.480 \text{x}) + 47.92 \tag{5}$$

where *x* = log*R*<sup>0</sup> \* , and the standard error of the estimate (SEE) was equal to 2.885. In Figure 2, the SRY percentage decreases significantly when the CSF increases. This decrease becomes negligible for relatively high log*R*<sup>0</sup> \* values. The SRY shows higher values for log*R*<sup>0</sup> \* approximately 0.36 while the most sever conditions, 2.30 < log*R*<sup>0</sup> \* < 2.60, gave lower SRY values. The SRY decrease can be attributed to the fast hydrolysis of hemicelluloses of barley straw while this decrease is limited by resisting the hydrolysis lignin fraction and the slow hydrolysis of the crystalline cellulose fraction. Moreover, some of the produced soluble sugar degradation products (furfural and 5-hydroxymethylo-furfural) can be further degraded to insoluble byproducts (humic substances, etc.).

**Figure 2.** SRY of the ACWT-treated material as affected by the CSF.

In Table 2 the HHV and SRY values are presented, while the relationship between HHV and ACWT severity using a second-order polynomial function is shown in Figure 3. The equation with the best fighting among many others was as follows:

$$\text{HHV} = 0.5028 \text{x}^2 + 0.0087 \text{x} + 19.613 \tag{6}$$

where *x* = log*R*<sup>0</sup> \* . The correlation coefficient was *R* = 0.7891 and the SEE was 5.097. The increase in the HHV can be explained by the increase of the carbon content in the ACWT material, as well as by the ash percentage decreasing. In addition, it can be explained by the lignin percentage increase and the hemicelluloses percentage decreasing with regard to the pretreated barley straw.

Table 3 shows total ash content of ACWT barley straw compared to the untreated one. A decrease in ash content was observed from 8.4 for the untreated straw to 5.5% wt. for the treated one. At this point, the CSF of the ACWT was log*R*<sup>0</sup> \* = 2.87. Öhman et al. [46] also observed the decrease in ash content of hydrolysis residue. In addition, Jenkins et al. [47] also recognized that water-leaking biomass has decreased ash concentration. Acid hydrolysis disrupts the structure of lignocellulose. As a result, minerals presented in the biomass were released into the soluble processing fluid. Therefore, the decrease in ash content can be attributed to the disordered cell structure and the water treatment. A reduction in the total ash content can be accomplished by applying acid hydrolysis to dissolve the minerals during the process.

**Figure 3.** HHV vs. CSF in logarithmic form.

**Table 3.** Proximate and ultimate analysis of the original and ACWT-treated (at optimized conditions based on HHV) barley straw.


In the present work, sulfur content increased after pretreatment, from 0.11% for untreated barley straw to 0.14 for the pretreated one (see Table 3). The difference is not significant, so acid hydrolysis can be applied for high sulfur containing lignocellulosics like barley straw [47].

The enhancement factor (EF) is defined by the following equation:

$$\text{EF} = \text{HHV}\_{\text{t}} / \text{HHV}\_{\text{o}} \tag{7}$$

where HHVt represents the HHV for ACWT barley straw and HHVo the HHV for original straw. The energy yield (EY) can be calculated as follows:

$$\text{EF} = \text{EF} \cdot \text{SRY} \tag{8}$$

In Figure 4, the EF and the EY of ACWT barley straw vs. the CSF in logarithmic form are presented. The curves representing the calculated values of EF and EY (i.e., the theoretical values) occurred from the application of Equations (5) and (6), respectively, using the same parameters as in Figure 3. Moreover, in Figure 4a we can notice that the EF increases when the CSF increases. On the other hand, in Figure 4b, EY demonstrates improved values either for low CSF values or for higher CSF values, but not for moderate conditions. It must be observed that taking into account both Figures 3 and 4,

we get high HHV, EF and EY only for high CSF values, i.e., severe conditions. The increase in EF is a result of the increase in HHV, as can be seen from Equation (7). The initial decreasing of the EY can be explained by the significant decreasing of SRY and the insignificant increasing of EF for relatively low CSF values. On the other hand, the consequent increasing of EY can be explained by the significant increase of EF compared to the negligible SRY decreasing for higher CSF values.

**Figure 4.** ACWT barley straw (**a**) enhancement factor (EF) and (**b**) energy yield (EY) vs. CSF in logarithmic form.

#### *3.2. Ultimate and Proximate Analysis Results*

Ultimate and proximate analysis results of the barley straw compositions are presented in Table 3. The purpose of the analysis was mainly to determine whether carbon appeared during the ACWT process. In the untreated barley straw, carbon and oxygen were 45.53% and 47.86%, respectively. After ACWT pretreatment (200 ◦C, 0.0035 N SA concentration, 25 min), the carbon content increased to 52.51% (optimal conditions) and the oxygen decreased to 40.71%. Zanzi et al. [48], Angles et al. [40] and Iroba et al. [49] obtained similar results from the ultimate analysis of the WT-treated lignocellulosic biomass feedstock. The C content of their steam-exploded samples increases while lignin condenses and carbonizes. The lignin condensation was with loss of H2O and reduction of O content. H and O losses were also reported due to the formation of H2O, CO and CO2. The observed increased N content was dropped by the increasing temperature and time of WT. The change in C, H and O content was significant. Reactions occurring during hydrolytic treatment generally produce volatiles and gases with a low energy density until they evaporate, increasing the energy density of the residue material by making it rich. The results obtained from the present ultimate and proximate analysis are consistent with the above literature results, demonstrating that carbonization occurs at higher temperatures and times in combination with degradation of holocellulose and destruction of the lignocellulosic matrix.

The increase in the HHV of the pretreated straw relative to the increase in its values of hydrogen, carbon and sulfur, as well as the reduction of oxygen, are in agreement with previous studies [44]. In the present work, the HHV of the ACWT-treated barley straw samples were calculated using the C, H and N results presented in Table 3. The following Friedl et al. [50] equation was used for these calculations:

$$\text{HHV} = 3.55 \text{C}^2 - 232 \text{C} - 2230 \text{H} + 51.2 \text{C} \cdot \text{H} + 131 \text{N} + 20600 \tag{9}$$

The HHV for the untreated and ACWT-treated barley straw, estimated by Equation (9), was 18.1 and 21.0 MJ/kg, respectively. The experimentally measured HHV was 17.5 of untreated barley straw and 24.3 MJ/kg for ACWT-treated material. Consequently, the estimated HHV was sufficiently close to the experimental measurements. Moreover, the estimated EF was 1.16, also not so close to the experimental one that was 1.39.

#### *3.3. Simulation Based on RSM*

In this work we performed 15 experimental trials as the Box–Behnken scheme demands. The results were fitted by an RSM model using a second-order polynomial equation. The model variables, where *A* temperature, *B* time and *C* SA concentration. The suggested polynomial was fitted to the ACWT experimental data. The equation parameters were estimated by multiple linear regressions analysis. The confidence level was 99%. The optimal conditions for ACWT of barley straw were found. The polynomial equation used herein was as follows:

$$\begin{aligned} \text{HHV} &= 20.397 + 1.18A + 0.585B + 0.4138 \text{C} + 0.2775AB - 0.38AC - 0.0575BC + \\ &\quad 1.0967A^2 - 1.1008B^2 + 0.7367 \text{C}^2 \end{aligned} \tag{10}$$

By solving this model using a partial differential equation (PDE), three 3D graphs were developed.

In Table 4 the effect of the three independent *A*, *B* and *C* variables on the response variables *P*, *T* and *F* is presented by ANOVA of the RSM. The coefficient of determination (*R*2) equaled 0.9035, showing sufficient fitting of Equation (9) to the experimental data. Joglekar and May [51] demands *R*<sup>2</sup> > 0.80 for the reliable model based on its parameters. The *R*<sup>2</sup> of the model presented herein has shown that it adequately represents the true relationship between the predicted and observed values of the variables under consideration. When the *R*<sup>2</sup> value is 0.9035 it indicates that 90.3% of the volatility is explained by the model and only 9.70% is due to luck. Model regression analysis showed that the effect of the temperature variable was significant, while the time and SA concentration variable had no effect on ACWT.


**Table 4.** RSM coefficients and P, T and F-values for the polynomial model.

\* not significant.

ANOVA of the performance index, including the probability value *P*-value, the value on the F distribution F-statistic or F-value, as well as the value of the T-distribution, T-statistic or T-value, shows that the high value of F indicates that the proposed model is valid and low *P* values indicate that the model is significant. Given the F value for the temperature, which was higher compared to the other two, it implies that the increase in the reaction temperature has very strong effect on the experiment. The appropriate T value (4.3368) gives the signal to noise ratio and when the value is greater than 4 is generally desirable [52]. When *P* becomes less than 0.05, it also indicates the significance of the model terms. On the other hand, values above 0.10 means that these terms are not significant. In addition, the RSM provided a sum of squares (SS) of 27.90, degrees of freedom (DF) for the regression of 9 and 5 for the residual; a mean square (MS) of 3.100 for the regression and 0.5960 for the residual, and a significance of F (SigF) of 0.0421. ANOVA analysis produced values within the range of experimental values, confirming that the RSM model can simulate the HHV of ACWT-treated barley straw.

In Figure 5, three 3D-graphs illustrate the RSM results as a function of two of the independent variables, while the third variable remains constant. The conditions 180 ◦C, 20 min and 22.5 mM SA concentration were chosen as constant values as their combination gives a moderate HHV value. The relationships between the time and temperature, temperature and SA concentration and the time and SA concentration are presented in Figure 5a–c, respectively. There was no significant effect of the combination of variables individually, namely the reaction time and the concentration of sulfate on the response value.

#### *3.4. Optimization*

The optimal conditions for ACWT of barley straw were determined by using Quantum XL commercial software. The findings were comparable to those of CSF methodology. The highest value of the HHV experimentally achieved at the expected by the Quantum XL software set point and it is achieved when relatively severe pretreatment conditions (200 ◦C, 20 min and 35 mol/m<sup>3</sup> SA concentration) were used (Table 5). Under these optimal conditions, log*R*<sup>0</sup> \* = 2.87 and SRY = 38% were estimated. The experimental result agreed with the optimal values predicted by the RSM model giving HHV = 25.5 MJ/kg. Both simulation approaches (CSF methodology and RSM) provided useful information on CSF and exact experimental parameters to optimize the HHV of ACWT-treated material.


**Table 5.** Optimal conditions (set point) for maximized HHV predicted by the RSM polynomial model.

#### *3.5. SEM Results*

The surface of the solid residue of untreated barley straw (Figure 6a,c,e) and the ACWT-treated one Figure 6b,d,f) was observed by a SEM, in the order to observe the physical changes in the structure of barley straw. The porous structure of the surface of solid residue is created during the ACWT process due to the production of volatile substances. As the process continues, the pores and cracks are appearing on the surface of the solid residue. The porous structure and the cracks of solid residue can be clearly shown on the SEM images. When the temperature increases at 200 ◦C, the hydrolysis process enters the second phase and most of organic materials are released gradually, resulting from a great loss of mass and the formation of the major hydrolysis products.

As the temperature rises, the volatile matter gradually evaporates, creating pores and transforming the surface of solid residue to be concave. The surface of straw is found to have more irregular porous structures in the higher hydrolysis temperature, and changes in surface morphology is observed (Figure 6b,d,f). This evidence indicates that the structure of the torrefied biofuel is downgraded or lit due to high temperature.

This results in the formation of an irregular structure of solid residue, destroying its uniformity. The solid residue, created from high temperature, is more fragile compared to the untreated straw and it cannot withstand the pressure due to its fragile structure.

**Figure 5.** HHV vs. temperature–time (**a**), temperature–SA concentration (**b**) and SA concentration–time (**c**).

**Figure 6.** SEM results of original barley straw at (**a**) ×750, (**c**) ×7500 and (**e**) ×20,000 magnification, and ACWT-treated material at (**b**) ×750, (**d**) ×7500 and (**f**) ×20,000 magnification.

#### **4. Conclusions**

In this study, it was shown that ACWT has the potential to convert biomass such as barley straw into biofuel with enhanced thermal, chemical and physical fuel properties compared to raw barley straw. The results were obtained by ACWT treatment of barley straw, in a batch reactor, showing significant improvements, such as an increased HHV and lower ash content. In addition, the ACWT is appropriate for cleaner fuels production from lignocellulosic residues. Within the biorefinery concept, during the fermentable sugars production via acid hydrolysis, ACWT barley straw with enhanced HHV can be obtained as a solid byproduct. Two simulation methodologies were successfully applied for ACWT process: (i) the combined severity factor approach and (ii) the response surface methodology. The first methodology gives the CSF *R*<sup>0</sup> \* or *R*<sup>0</sup> \* (which can be achieved by various combinations of time, temperature and acid concentration) for optimized HHV, EF and EY, while the second methodology determines exactly the optimal experimental conditions (time, temperature and acid concentration). The pretreatment conditions were optimized for modified barley straw production with a maximum HHV equal to 24.3 MJ/kg, EF = 1.39 and EY = 68% wt. dry basis. The optimal ACWT temperature, time and SA concentration were 200 ◦C, 25 min (with regard to the isothermal period only) and a 35 mM SA concentration of the solution. The above presented simulation method, which is the combination of two methodological approaches, can be possibly used in modelling/simulation and optimization of various processes of biomass thermal, chemical and thermochemical treatment.

**Author Contributions:** Investigation, A.N.; methodology, P.G.; software, E.S.; supervision, D.S. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** This work has been partly supported by the University of Piraeus Research Center.

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

#### **References**


© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
