**Characterization and Modelling Studies of Activated Carbon Produced from Rubber-Seed Shell Using KOH for CO2 Adsorption**

#### **Azry Borhan 1,\*, Suzana Yusup 1, Jun Wei Lim <sup>2</sup> and Pau Loke Show <sup>3</sup>**


Received: 22 October 2019; Accepted: 11 November 2019; Published: 14 November 2019

**Abstract:** Global warming due to the emission of carbon dioxide (CO2) has become a serious problem in recent times. Although diverse methods have been offered, adsorption using activated carbon (AC) from agriculture waste is regarded to be the most applicable one due to numerous advantages. In this paper, the preparation of AC from rubber-seed shell (RSS), an agriculture residue through chemical activation using potassium hydroxide (KOH), was investigated. The prepared AC was characterized by nitrogen adsorption–desorption isotherms measured in Micrometrices ASAP 2020 and FESEM. The optimal activation conditions were found at an impregnation ratio of 1:2 and carbonized at a temperature of 700 ◦C for 120 min. Sample A6 is found to yield the largest surface area of 1129.68 m2/g with a mesoporous pore diameter of 3.46 nm, respectively. Using the static volumetric technique evaluated at 25 ◦C and 1.25 bar, the maximum CO2 adsorption capacity is 43.509 cm3/g. The experimental data were analyzed using several isotherm and kinetic models. Owing to the closeness of regression coefficient (R2) to unity, the Freundlich isotherm and pseudo-second kinetic model provide the best fit to the experimental data suggesting that the RSS AC prepared is an attractive source for CO2 adsorption applications.

**Keywords:** rubber-seed shell; activated carbon; CO2 adsorption; isotherms; kinetics modeling

#### **1. Introduction**

With the fast escalation of the overall industrialization and population in many countries, the utilization of energy is exclusively expending. Presently over 85% of the international energy requirement is being financed by the burning of fossil fuels [1]. The reasons for this dependence on energy sources are a result of instinctive energy density, supply, and dependency of modern society on the procurement and exchange of these resources. Fossil fuels will still dominate in the predictable future, primarily in power production and industrial manufacturing. The utilization of these fossil fuels, especially in electricity generation, residentiary, transportation, and industrial area discharges massive amounts of carbon dioxide (CO2) into the atmosphere, and thus upsets the carbon balance of our planet, which has been stable over millions of years. Although anthropogenic emissions of CO2 can be considered comparatively limited related to the natural carbon changes such as photosynthetic fluxes, its escalation has distinct impacts on the global climate over a very short duration of time. The concentration of CO2 in the atmosphere recorded in December 2018 has increased from 280 to 408 ppm since the beginning of the industrial revolution [2]. The rise of CO2 concentration dominates

the balance of incoming and outgoing energy in the earth's atmosphere. Hence, CO2 has often been pointed out as the main anthropogenic greenhouse gas (GHG) as well as the leading offender in climate change. Due to this serious environmental problem, there is pressure all over the world to address this issue. International agreement known as the Kyoto protocol was reached and signed by 192 nations under the United Nations Framework Convention on Climate Change (UNFCC) to limit GHG emissions of industrialized nations by 5.2% [3]. Despite Malaysia contributing 0.2% in global greenhouse gases emissions [4], the increment of CO2 emission is not to be taken lightly, because the continuous increase of the CO2 rate to the surroundings can cause global warming as well as affect human health due to long term exposure to high concentrations of CO2 in the future.

Many CO2 capture technologies are proposed and being investigated these years, including chemical and physical adsorption, cryogenic separation, and membrane separation [5,6]. At the time being, chemical absorption (scrubbing), utilizing amine-based solvents, is the most considered technology used in industries. Although it is viewed as the most practical technology for CO2 capture in post-combustion processes and has been used for more than 60 years, its application brings some negative impact, such as equipment corrosion, it requires high absorber volume, and is harmful to human health. In addition, this method of CO2 separation is energy consuming and expensive as it requires large amounts of low-pressure steam for adsorbent regeneration [6,7]. Among the viable technologies for CO2 capture, adsorption using solid material is chosen due to low energy prerequisites, low essential and running cost, including controlled waste generation. Due to its diverse benefits, such as economical, accessible for regeneration, indifferent towards the moisture, high CO2 adsorption uptake at ambient situation, high specific surface area, high mechanical durability, sufficient pore size distribution, as well as low in energy needed, [8] activated carbon (AC) is one of the up-and-coming solid adsorbents that can be employed to capture CO2. Based on the advantages by AC, it has been extensively engaged in different utilizations including in gas and liquid phases. The capability of AC in CO2 occupation also depends on several criteria such as the nature of the activating method and the quality of starting materials, which in turn alters the surface chemistry and porosity and of the synthesized AC [8–10].

Many researchers have investigated the manufacture of low-cost adsorbents from inexhaustible and economical precursors, which are mainly industrial and agricultural derivatives ranging from palm kernel [11], banana peel [12], coconut shell [13], doum seed coat [14], walnut shell [15], etc., and their findings are quite conclusive. Meanwhile, rubber plantations in Malaysia have increased ever since the year 2010. The increase of rubber plantations have led to the increase in rubber production. Apart from producing latex, rubber seeds with hard shells were produced at the same time. It is estimated that about 800 to 1200 kg of rubber seed per ha per year is produced in a rubber plantation. The increase of rubber seed is causing significant environment and disposal problems [16]. To reduce the waste disposal issue, the rubber-seed shell (RSS) is proposed to be used as AC for CO2 capture, since there is limited research on its preparation, and to address the problems.

In this study, a chemical activation method by potassium hydroxide (KOH) was selected due to its numerous advantages over physical activation methods and favorable conditions compared to other chemical activating agents. The research study was set in the direction towards evaluating the potential of using RSS as AC material in the removal of CO2. Since chemical activation was adopted in this study, impregnation ratio (IR), temperature (Tact), and activation time (tact) would be the main factors affecting the extent of reaction. Therefore, these parameters were also investigated to evaluate the effects of operation conditions on pore advancement of AC prepared from RSS. Samples with the highest surface area are preferred to examine on the performance of CO2 adsorption at ambient pressure and temperature. In addition, the CO2 adsorption isotherm is evaluated through several models, such as Langmuir, Freundlich, and Temkin. Lastly, the kinetic property of the sample with the highest adsorption capacity of purified CO2 is evaluated using the pseudo-first order, pseudo-second order, and Elovich kinetic models.

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

#### *2.1. Materials and Pre-Treatment*

The raw RSS is collected from the Rubber Industry Smallholders Development Authority (RISDA) rubber plantation located at Chemor, Perak. It is then washed using distilled water to remove impurities and dried overnight in an oven at 110 ◦C to remove surplus water content. Once dried, the RSS is crushed and pulverized and then sieved into a particle size of 250 μm and stored in airtight plastic containers for further use. All chemicals were of industrial reagent grade and acquired from R & M Chemicals supplier located at Semenyih, Selangor.

#### *2.2. Activation and Carbonization*

The RSS prepared was impregnated by mixing it with a desired ratio (1:1, 1:2, and 1:3) of KOH based on the weight of the dry sample. The mixing and impregnation processes were allowed to sit for overnight to ensure complete reaction takes place between the chemical reagent and raw material. The impregnated material was then carbonized in a fixed bed activation unit with heating temperature ranging from 400, 500, 600, 700, 800 to 900 ◦C and activation time between 60 and 120 min. The one-factor-at-a-time (OFAT) method was adopted in this study so that reduction of sample is achievable; it is a method of designing experiments involving the testing of factors one at a time instead of all simultaneously. Throughout the process, N2 gas, which acts as the carrier gas and promotes the pore formation in RSS [17], was allowed to flow in the rotary kiln. After the heat treatment, the material was left to cool to room temperature and subsequently washed with distilled water to discard the excess KOH solution and ash. The AC sample produced was preserved in an oven for overnight at 90 ◦C and kept in a desiccator to prevent moisture.

#### *2.3. Characterization*

To investigate the best operating parameter for producing AC from RSS, several analytical equipments were employed. To investigate the morphology surface structure, a Zeiss EVO-50 Field Emission Scanning Electron Microscope (FESEM), model Supra 55 VP acquired from Zeiss Jena, Germany, is used to compare the structural images of RSS before and after activation by generating real space enhanced images of its surface. In addition to the standard electron microscope detectors, the instrument is also furnished with Energy Dispersive X-ray (EDX) Spectroscopy for elemental analysis investigation.

For specific surface and porosity analysis, Micrometrics ASAP 2020 is used to determine the surface area (SBET), pore size distribution (D), and the total pore volume (VT) through nitrogen adsorption–desorption isotherms analysis. N2 (99.9% purity) gas is applied as adsorbate and the condition was allowed to flow at 350 ◦C for 2.5 h [16]. The specific surface area of the AC samples is determined using the Brunauer–Emmett–Teller (BET) method using nitrogen adsorption isotherm data tabulated from a computer.

#### *2.4. CO2 Adsorption Capacity Analysis*

CO2 adsorption analysis were carried via purified CO2 (99.98% purity and supplied by Linde Malaysia Sdn. Bhd.) volumetric adsorption method using High Pressure Volumetric Analyzer (HPVA II) supplied by Particulate System. To ensure all impurities were removed from the samples, about 0.5 g of AC was inserted inside a 5 cm3 sample glass cylinder and degassed at 170 ◦C for 8 h under vacuum. A 60 μm filter gasket was then planted on top of the sample cylinder, to avoid the fine particles from entering the valve [17]. After completion of the degassing step, the samples were cooled to ambient temperature and prepared for adsorption studies. The CO2 adsorption process was started by introducing the gas adsorbate (CO2) into the system. This is accomplished by granting the valve between the loading and sample cylinder to open and allow the CO2 to interact with the AC material. To ensure equilibrium of the adsorption process, the holding time at each pressure interval was fixed at

45 min. By taking the differences between the amounts of dosed gas into and the amounts of gas staying in the system upon adsorption process, the volumetric CO2 sorption capacity during experimental run was calculated [18]. For isotherm studies, using the resulting points of volumes adsorbed at equilibrium pressures, several isotherm models such as Langmuir, Freundlich, and Temkin [19] were plotted and fitted to experimental data to find suitable representation of adsorption process of RSS based AC. The suitability of the above-mentioned models is assessed by R<sup>2</sup> values that are close to unity. Table 1 outlines the non-linear and linear equations of these three models.


**Table 1.** Model isotherm equations.

*Pe* is equilibrium pressure (bar); *qe* and *qm* are the amount capacity of CO2 adsorbed at equilibrium and at maximum, respectively (cm3/g); *kL* is the Langmuir constant (1/bar); *kF* (cm3/g·bar1/n) and *<sup>n</sup>* is the Freundlich constant; *<sup>B</sup>* <sup>=</sup> *RT*/*bT; bT* (J/mol); and *kT* (cm3/g·bar) is the Temkin constant.

#### *2.5. Kinetic Studies*

CO2 adsorption kinetics of KOH impregnated RSS AC are desirable to evaluate the accomplishment of sorbents and to understand the overall mass transfer in the CO2 adsorption process. In addition, it served as baseline to predict CO2 adsorption/desorption kinetics for the rational simulation and design of gas-treating systems. In this study, the usefulness of Lagergen's pseudo-first order model, pseudo-second order model, and Elovich model approaching the experimental values using purified CO2 adsorption at 25 ◦C is examined. The compliance of the predicted adsorption capacity was evaluated by the magnitude of coefficient regression R2 closeness towards unity.

#### 2.5.1. Pseudo-First Order Kinetic Model

The linearized Lagergen's pseudo-first order model was the first adsorption rate equation depicted for sorption of a liquid/solid system and one of the most frequently used adsorption rate models. It is expressed by the equation below [19,20]:

$$
\log(q\_{\epsilon\eta} - q\_t) = \log q\_\epsilon - \frac{k1}{2.303}t,\tag{1}
$$

where *qeq* and *qt* are the amount of adsorption at equilibrium and at that particular time *t*, respectively. It has the unit of mg/g. *k1* represents the rate constant for pseudo-first order adsorption. A linear plot will be obtained from the graph when this model is applicable. In addition, the slope and interception point of the plot can be used to determine pseudo-first order parameters.

#### 2.5.2. Pseudo-Second Order Kinetic Model

The pseudo-second order model is expressed by the equation as follows [19,20]:

$$\frac{t}{q\_t} = \frac{1}{k2qc^2} + \frac{1}{q\_e}t,\tag{2}$$

where *qe* and *qt* are the amount of adsorption at equilibrium and at that particular time *t,* in mg/g, respectively. *k2* represents the overall rate constant for pseudo-second order adsorption with the unit of g/mg min. Similar to first order, a linear plot will be achieved from the graph of *t*/*qt* versus time, *t* when it is applicable.

#### 2.5.3. Elovich Kinetic Model

Elovich kinetic model is usually used in a gas–solid system and is expressed by [19,20]:

$$\frac{dq\_t}{dt} = \alpha \exp(-\beta q\_t) \tag{3}$$

where *qt* is the amount of CO2 adsorbed in mg/g at a particular time, *t.* α represents the initial adsorption rate in mg/g min, while β is the extent of surface coverage in g/mg and the process activation energy.

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

#### *3.1. Elemental Composition Analysis*

Aside from examining the surface morphology of the samples, FESEM is also furnished with Energy-Dispersive X-ray (EDX) spectroscopy for detecting the elemental composition. Table 2 below shows the corresponding elemental content before and after activation. Two main elements were detected on the samples before activation, which are carbon and oxygen. It was described in specific that the acceptable range of carbon presence should lie between 40 and 80% [17]. The result shows that RSS fulfills the criteria of producing AC. It was found that the percentage of carbon content has escalated after carbonization and activation due to the release of more volatile matter during the process. An additional element identified as potassium was detected in the sample after activation. Presence of potassium element is as a result the use of KOH as chemical activating agent during the activation process. Repetitive washing can greater reduce the potassium element, but complete elimination is hardly possible. Calcium content in RSS is expected and considered normal as the RSS is rich in protein, minerals, and amino acid. Eka et al. [21] has mentioned this in a report that rubber seed has a low content of calcium.


**Table 2.** Comparison of elemental composition.

#### *3.2. Characterization Study*

The preparation conditions and results of AC samples produced from KOH impregnated RSS are exhibited in Table 3. The result demonstrated that sample A6, which is arranged at an impregnation ratio (IR) of 1:2, Tact of 700 ◦C, and tact of 120 min, yields the highest values of specific surface area, SBET of 1129.60 m2/g, average pore diameter, D of 3.46 nm, and total pore volume, VT as high as 0.412 cm3/g. This is followed by samples A5 at 826.31 m2/g, 3.21 nm, and 0.376 cm3/g, and A8 at 731.06 m2/g, 3.21 nm, and 0.301 cm3/g, respectively. In comparison, samples A1 and A2, which are developed at an IR of 1:1, all exhibit reduced SBET and VT values in comparison to A3. The reason is due to the IR of KOH to RSS which plays a critical role in the pores formation. High IR supposedly helps to increase the amount of potassium metal that can be intercalated and thus develops more pore formation [22]. Nevertheless, the outcome of the analysis confirms that there is an utmost number of ions that can be accepted above which would reduce pore progression. The reason behind this is because additional or excess activating agents probably form an insulating layer (or skin) coating the AC particles, and therefore lowering the activation process and the influence with the surrounding atmosphere [23]. This phenomenon is particularly observed in sample A4 where most likely activation

is hindered, resulting in a lower surface area (SBET of 492.55 m2/g) and pore volume formation (VT of 0.182 cm3/g) as compared to A3.


**Table 3.** Surface area and porosity results.

As for activation temperature, the optimum temperatures have been reported to be between 500–800 ◦C by most of the earlier researchers [24]. Hence, the experiments were conducted by varying temperatures from 400 to 900 ◦C, and it turned out that the recommended temperature is reliable due to verification provided by samples A1 to A3 and A5 to A9. Sample A1, which is carbonized at 400 ◦C, yields the lowest SBET (203.81 m2/g) and VT (0.113 cm3/g) than any additional samples. It was reported that 400 ◦C is the starting carbonization temperature in the development rudimentary of pores of AC material. Increasing the activation temperature to 700 ◦C will intensify the expulsion of molecular weight of unstable compounds and further generating new pores, resulting in the hastening of porosity growth of the AC. Sample A5 and A6 have the highest SBET (826.31 m2/g and 1129.60 m2/g) and VT (0.376 cm3/g and 0.412 cm3/g) values, respectively. Nevertheless, when the activation temperature is elevated to 900 ◦C, the excessive heat energy supplied to the carbon will result in the collapsing and knocking of some porous wall [17]. The outcome is the decrease quantity of SBET and VT. This can be interpreted by comparing sample A9 with other samples' characterization result.

Extended activation time during the carbonization process may cause in over-activation, where surface erosion is accelerated more quickly than pore formation. Sample A7 shows rapid decreasing in SBET, VT, and D after an activation time of 180 min. Considering the well-developed porous structure at a temperature of 700 ◦C, any increment in activation time will causes the carbon structures to break between its cross-links, resulting in pore collapsing [25,26]. Although sample A6 has the highest SBET, VT, and D, only 31.67% of its volume existed as micropores. Sample A3 has the highest micropore volume, with 85.29% of its total pore, while sample A9 has the lowest micropore volume with 15.13%. According to The International Union of Pure and Applied Chemistry (IUPAC) classification pore size are categorized as macro if the size >50 nm, meso if the size is between 2–50 nm, and micropore when <2 nm [26]. All samples, except raw and A1, show pore diameter in the range between 2–4 nm, and thus distinctly categorize that the pores belong to the mesopores classification.

#### *3.3. Morphology*

Figure 1 shows the microscopic morphology structure of the raw RSS and some selected samples prepared at different operating parameters. By using FESEM, structural images with magnification up to 300 times are taken. The structural image of fresh RSS in Figure 1a indicates that the raw material before undergoing activation shows no noticeable pores. However, the image clearly shows the existence of fine pores on the surface, which is one of the important aspects for manufacturing AC. After activation, a greater distribution of pores emerged to become active sites for adsorption to take place more readily. The difference in pore structure before activation and after activation is clearly

illustrated through comparison made between Figure 1a with Figure 1b,c. The canal structure on the surface of sample A6 has been partially broken, which indicates carbonization process occurred and the surface was eroded by longer activation temperature and time than sample A1. Therefore, the SBET of sample A6 is much higher. As shown in Figure 1c, sample A6 has the most well-developed structure compared the other three AC. This justified that sample A6 has the highest SBET, VT, and D compared to other samples. On the other hand, collapse of porous wall due to excessive heat exposure is observed in Figure 1d with sample A9 being activated at highest temperature. Two promising samples, A5 and A6, are selected for further analysis.

**Figure 1.** FESEM images of selected samples, (**a**) fresh RSS, (**b**) sample A1, (**c**) sample A6, and (**d**) sample A9.

#### *3.4. Nitrogen Adsorption–Desorption Isotherms Study*

Figure 2 shows the N2 adsorption–desorption isotherms of the selected three AC. The quantity of N2 adsorbed is projected against the relative pressure p/po where p = pressure at given condition and po = saturated vapor pressure of N2. Based on IUPAC classification of adsorption isotherm [25], it can be seen that for raw AC the isotherm follows Type II classification, which signify the presence of microporous pores existed and within the micropores the surface resides almost exclusively. Once it was fully occupied by N2 adsorbate, very few or no exterior surface left for further adsorption. Samples A5 and A6, however, show a combination of Type I and Type III classification where the trend line was initially following Type I with Type III trend line appearing at the end of high relative pressure. This combination is associated with a combination of microporous and mesoporous structures with mesopores as the dominant species [26]. The result is consistent with the D of 3.21 and 3.46 nm, respectively, as shown in Table 3, where it complies with the IUPAC classification of mesoporous material.

**Figure 2.** N2 adsorption–desorption isotherms.

#### *3.5. CO2 Adsorption and Isotherm Modeling*

Figure 3a depicts the CO2 adsorption capacity for samples A5 and A6 at an ambient temperature and pressure. According to Estaves et al. [27], it is ideal to adsorb CO2 at lower a temperature (25 ◦C in this research) as CO2 adsorption process onto AC is exothermic due to the physical adsorption (physisorption) process. Weak van der Waals forces are basically involved in physisorption. At high temperature condition, these weak forces are easily broken and result in the decrease of the adsorption capacity. Instability of the CO2 adsorbate on the carbon surface will result in a desorption process, owing to higher surface adsorption energy and molecular diffusion at high temperature. According to Hauchhum et al. [28], the rise in bed temperature of AC will expedite the internal energy of the adsorbent, and therefore CO2 molecules are released from the surface. To summarize, exothermic process during adsorption is controlled by physisorption when there is reduction in the adsorption capacity with respect to the temperature [29]. Based from Figure 3a, higher CO2 adsorption capacity for sample A6 is recorded compared to A5. This finding is parallel to the nitrogen adsorption analysis carried out earlier. This verifies that higher surface area does lead to higher adsorption capacity. The highest CO2 adsorption capacity for sample A6 is 43.5094 cm3/g at a pressure of 1.2523 bar, while for the A5 sample it is 39.2496 cm3/g at a pressure of 1.2525 bar, respectively.

**Figure 3.** (**a**) CO2 adsorption capacity at 25 ◦C; (**b**) weight uptake of CO2 adsorption at 25 ◦C.

The effect of pressure on CO2 capacity is also observed where higher adsorption occurs at higher pressure. This phenomenon happens because high pressure tends to push CO2 molecules onto adsorption site within the pore. In Figure 3b, the highest CO2 sorption quantity that is exhibited by samples A6 and A5 is 8.5431 wt% at a pressure of 1.2523 bar and 7.7067 wt% at a pressure of 1.2525 bar, respectively. The disparity in CO2 weight uptake of these solid AC as a result of the surface area value is shown in Table 3. Sample A6, which has the largest SBET, is noticed to adsorb more CO2 compared to other samples, which implies that there are more available surface sites for CO2 adsorption processes to take place.

Table 4 shows the calculated isotherm constants and their corresponding R<sup>2</sup> values for CO2 adsorption using linear regression method. According to Perez et al. [30], the Langmuir constant *kL* and Freundlich constant *kF* are all related to the adsorption affinity. Its value decreases with increases in temperature, which signify physisorption behavior. This CO2 adsorption capacity reduction can be interpreted by Le Chatelier's principle, where for an exothermic process (physisorption), low temperature is preferred during adsorption. The exothermic behavior of the CO2 adsorption is aligned with the *qm* value that is likely to decline with the rise in the adsorption temperature. The Langmuir adsorption model explains that adsorption does not occur after monolayer adsorbate formation on the adsorbent surface. This model assumes constant adsorption energies onto the surface and no adsorbate movement on the surface planes [29–31].

**Table 4.** Langmuir, Freundlich, and Temkin isotherm models via linearized technique.


Freundlich adsorption model is based on an empirical relationship, which explains adsorption isothermal variation with pressure. This model is often used to explain heterogeneous surface adsorption characteristics. The Temkin adsorption model accounts for the adsorbate–adsorbent interaction. This model assumes adsorption heat of all adsorbate molecules decreases linearly [29]. As for Freundlich constant, *n,* and sometimes known as heterogeneity factor, its value signifies the type of adsorption, where *n* > 1 is for physical adsorption while *n* < 1 corresponds to chemical adsorption. Both samples A5 and A6 have a R2 value closer to unity for the Freundlich model compared to Langmuir and Temkin. This shows that the adsorption process occurs in heterogeneous surfaces and is not restricted to monolayer adsorption as recommended by Langmuir. The summary for Table 4 suggests that the Freundlich model gives the best fitting correlation to the experimental data, owing its R<sup>2</sup> value approaching unity as it permits the CO2 adsorbate molecules to form a successive layer onto the surface of AC.

#### *3.6. Kinetic Analysis*

Figure 4 shows the straight line plot of log(qe − qt) versus *t* for RSS AC of sample A6 at 25 ◦C using Equation (4). The kinetic data are summarized in Table 5. The pseudo-first order kinetic model is established on the assumption that the rate of adsorption is proportional to the number of vacant sites available on the adsorbent surface and is used regularly in liquid–solid phase [29]. Due to its low R2 value of 0.8392, this kinetic model does not fit well with the CO2 adsorption experimental data. Sadaf et al. [32] verified that the pseudo-first order model was unsuitable in the adsorption process as it can be only applied during the beginning stage and not for the entire period.

**Figure 4.** Pseudo-first order kinetic model of sample A6 RSS AC.


**Table 5.** Summary of kinetic models.

For the pseudo-second order kinetics model, the rate of adsorption is assumed to be linearly related to the square of the number of vacant sites available on the adsorbent surface. This model has been used by many researchers for the modeling of experimental data of CO2 adsorption kinetics [33,34]. A plot of *t*/*qt* against time *t* will generate a straight line with 1/*h* and 1/*qe* as y-interception and slope, respectively, if the model is applicable in the adsorption process. By taking the R2 value shown in Figure 5, the pseudo-second order model fits the CO2 adsorption profile with the regression coefficient value of 0.9388 compared to the first-order model of 0.8392. In addition, the magnitude of *h* that represents the rate of adsorption has a value of 2.306 mg/g min and its value is expected to decline with respect to the operating temperature. According to Simon et al. [29], the CO2 molecules will gain an adequate amount of energies at elevated temperatures and be able to overcome the weak van der Waals bonding and finally will be moved back to the bulk gas phase. A similar trend was observed by Chao Ge et al. [25] when investigating the adsorption equilibrium capacity of CO2 at temperatures between −47 and 28 ◦C using both pseudo-model kinetics. Results predicted by the pseudo-second order model were much closer and coincided with the experimental values (R2 value greater than 0.996). Contrarily, larger deviation between the actual and calculated value resulted by using the pseudo-first order model and, thus, producing lower R<sup>2</sup> values.

**Figure 5.** Pseudo-second order kinetic model of sample A6 RSS AC.

Figure 6 displays the Elovich plot of CO2 adsorption using the same data. If it fits with the model, a straight line with gradient of 1/β and y-interception of 1/β ln(αβ) will be produced. The Elovich kinetic model assumes that the rate of adsorption exponentially decreases with the increase in the amount of CO2 adsorbed on the adsorbent surface without any interaction among the adsorbed species [35,36]. By applying boundary conditions and integrate in Equation (3), it results in a new equation as shown below:

$$q\_t = \frac{1}{\beta} \ln(a\beta) + \frac{1}{\beta}Int\tag{4}$$

**Figure 6.** Elovich model of sample A6 RSS AC.

As demonstrated in the figure, the Elovich kinetic model gave a poor regression coefficient of 0.8188 compared to the other two pseudo-models.

Based on the summary of kinetic models shown in Table 5, it was noticed that the pseudo-first order and Elovich model are not well suited to fit the CO2 adsorption kinetics data owing to their low R2 values. Significant deviations are observed when determining equilibrium adsorption capacity using the two kinetic models. The pseudo-second order kinetics model fits the adsorption kinetics data, with a R<sup>2</sup> value of 0.939. It can be concluded that the three kinetics models are found to be suitable for fitting the present adsorption kinetics data in the following subsequent order: Pseudo-second order > pseudo-first order > Elovich.

#### *3.7. Comparison Study with Other Biomass Activated Carbon Materials*

Figure 7 shows the CO2 adsorption comparison capacities on selected types of AC from agricultural waste ranging from coconut shell, banana peel, rice husk, palm shell, and coconut fiber [17,34]. From this relative study, it can be seen that the RSS AC of sample A6 has a noticeable higher adsorption capacity compared to banana peel and rice husk based AC. The highest CO2 adsorption capacity is coconut shell at 78.77 mg/g, followed by coconut fiber at 60.2 mg/g and palm shell at 58.52 mg/g. RSS is ranked fourth with an adsorption capacity of 54.31 mg/g of CO2. The usage of RSS waste biomass from rubber plantation for the development into AC is practical as it can overcomes the shortage of the non-renewable precursors, such as zeolites, metal-organic frameworks (MOF), mesoporous oxides, polymers, etc. To guarantee a long-term sustainability of this industry, Khalili et al. [37,38] acknowledged that the biomass-based AC may be synthesized from the renewable feedstock. In addition, the preparation of wastes from sustainable biomass precursors provides an environmentally friendly and sustainable passage for the advancement of CO2 sorbent materials.

**Figure 7.** CO2 adsorption capacity by different agricultural waste AC [18,30,33,34].

#### **4. Conclusions**

This study indicates the promising potential in producing a low-cost AC from RSS using KOH as activating agent. The resulting AC yield a high surface area, pore volume, and average diameter at an impregnation ratio of 1:2, activation temperature of 700 ◦C, and activation time of 120 min. According to the characterization analysis, the range of pore diameter (2–50 nm) is in the mesopores classification. The AC with a high surface area was also verified to have higher CO2 adsorption capacity using a static volumetric instrument. Moreover, the adsorption capacity test proved that adsorption using RSS based AC has a high prospective in reducing CO2 and is on par with some already existing conventional biomass AC. Based on isotherm model analysis, the Freundlich isotherm model fit best, and the kinetic analysis demonstrated that the CO2 adsorption onto the AC obeys the pseudo-second order model due to its closest proximity R2 value toward unity.

**Author Contributions:** Data curation, A.B.; formal analysis, A.B. and J.W.L.; funding acquisition, A.B.; methodology, A.B. and S.Y.; supervision, S.Y.; validation, A.B. and S.Y.; writing—original draft, A.B.; writing—review and editing, J.W.L. and P.L.S.

**Funding:** The authors extend their appreciation to the Ministry of Higher Education (MOHE), Government of Malaysia, under the Fundamental Research Grant Scheme, [FRGS No: FRGS/1/2018/TK10/UTP/02/9] for funding this research.

**Acknowledgments:** The authors gratefully thank the Universiti Teknologi PETRONAS and the Center for Biofuel and Biochemical Research (CBBR), UTP for providing financial assistance and support.

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

#### **References**


© 2019 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* **Thermophysical Properties and CO2 Absorption of Ammonium-Based Protic Ionic Liquids Containing Acetate and Butyrate Anions**

**Normawati M. Yunus 1,\*, Nur Hamizah Halim 2, Cecilia Devi Wilfred 1, Thanabalan Murugesan 3, Jun Wei Lim <sup>2</sup> and Pau Loke Show <sup>4</sup>**


Received: 17 October 2019; Accepted: 4 November 2019; Published: 5 November 2019

**Abstract:** Ionic liquids, which are classified as new solvents, have been identified to be potential solvents in the application of CO2 capture. In this work, six ammonium-based protic ionic liquids, containing ethanolammonium [EtOHA], tributylammonium [TBA], bis(2-ethylhexyl)ammonium [BEHA] cations, and acetate [AC] and butyrate [BA] anions, were synthesized and characterized. The thermophysical properties of the ammonium-based protic ionic liquids were measured. Density, ρ, and dynamic viscosity, η, were determined at temperatures between 293.15 K and 363.15 K. The density and viscosity values were correlated using empirical correlations and the thermal coefficient expansion, α*p*, and molecular volume, *Vm*, were estimated using density values. The thermal stability of the ammonium-based protic ionic liquids was investigated using thermogravimetric analyzer (TGA) at a heating rate of 10 ◦C·min-1. The CO2 absorption of the ammonium-based ionic liquids were measured up to 20 bar at 298.15 K. From the experimental results, [BEHA][BA] had the highest affinity towards CO2 with the mol fraction of CO2 absorbed approaching 0.5 at 20 bar. Generally, ionic liquids with butyrate anions have better CO2 absorption than that of acetate anions while [BEHA] ionic liquids have higher affinity towards CO2 followed by [TBA] and [EtOHA] ionic liquids.

**Keywords:** ammonium-based protic ionic liquids; density; thermal expansion coefficient; viscosity; thermal stability; CO2 absorption

#### **1. Introduction**

Natural gas consists mainly of methane as well as other higher alkanes in varied amounts. It is mainly used as a fuel and as a raw material in petrochemical industries [1]. While natural gas is principally a mixture of combustible hydrocarbons, many natural gases also contain impurities, such as carbon dioxide, CO2, hydrogen sulfide, H2S, and water. Refining processes are required to remove all of these unwanted impurities from natural gas. Besides water and higher-molecular-weight hydrocarbons, one of the most crucial parts of gas processing is the elimination of CO2 and this process is normally done by means of chemical absorption techniques using alkanolamine solutions. Despite the successful practice of using alkanolamines for CO2 removal, several disadvantages have been identified, such as solvent loss and degradation as well as corrosion issues [2]. In view of these

issues, there is a need to develop new alternative, yet effective solvents for the same purposes. Ionic liquids have emerged as new solvents that have potential to be used for CO2 removal due to their special features, namely non-detectable vapor pressure, wide liquid range, and remarkable thermal stability. Ionic liquids are low melting salts with typical melting points of below 100 ◦C. Furthermore, there are countless possible combinations of cations and anions that can yield ionic liquids and this flexibility is utilized to design ionic liquids based on the application. Ionic liquids are used in many applications, such as in chemical reactions and separation processes [3–6]. In the field of CO2 capture using ionic liquids, initial work on the CO2-ionic liquid system was done in 2001 [7] and, following this discovery, extensive works have been done to explore the absorption of CO2 in ionic liquids under various operating conditions [8–15]. Imidazolium-based ionic liquids were used in most of the study of CO2 absorption. They are highly unsymmetrical and therefore have low melting points. Recently, protic ionic liquids have been used in the study of CO2 capture [16,17]. Protic ionic liquids are formed by a proton transfer between an equimolar amount of a Brønsted acid and a Brønsted base. The modern type of protic ionic liquids have been described by Ohno and co-workers [18] and an extensive review on the properties and applications of protic ionic liquids has been provided by Greaves and Drummond [19] who also noted the ability of protic ionic liquids to support amphiphile self-assembly [20,21]. The interest in the ionicity of protic ionic liquids has come to light due to the unlikeliness of complete proton transfer between the acid and base contributing to the presence of a neutral acid and base mixture [19,22,23]. MacFarlane and Seddon [24] proposed a limit of a 1% neutral species presence in an ionic liquid to be called 'pure ionic liquid'. The term pseudo-protic ionic liquid was suggested by Doi et al. [25] after they discovered that a mixture of *N*-methylimidazole and acetic acid exhibited electrical conductivity behavior even though the mixture was mostly dominated by neutral species rather than ions when inspected using Raman spectroscopy.

Nevertheless, the simple synthesis pathway of protic ionic liquids, i.e., a one-step neutralization reaction, and their proven ability to absorb CO2 motivated us to explore this type of ionic liquids in the field of CO2 capture. Nonetheless, physical properties, such as density, viscosity, and thermal stability of ionic liquids, are very important prior to using these new solvents in any applications. Precise understanding on the thermophysical properties is important as it is required to evaluate the suitability of ionic liquids to be used at an industrial scale [26]. For instance, viscosity is an important property for the design of industrial processes involving heat and mass transfer and dissolution of compounds in fluids [27]. Therefore, the aim of our work was to synthesize several new ammonium-based protic ionic liquids using a 1-step neutralization reaction, measure their thermophysical properties, and, lastly, test their ability to capture CO2. In this work, six ammonium-based protic ionic liquids, containing acetate and butyrate anions, were synthesized using solvent-free, 1-step neutralization reaction. The density, dynamic viscosity, and thermal stability of these ionic liquids were determined. The density values enable the estimation of thermal expansion coefficient and the molecular volume of the ionic liquids. To assess the capability of these ionic liquids towards CO2, absorption of CO2 was done using a solubility cell and the screening was done in the CO2 pressure range up to 20 bar and at 298.15 K. Results showed that the ammonium-based protic ionic liquids synthesized in this study have the potential to absorb CO2.

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

#### *2.1. Chemicals*

Three amines and two organic acids from Merck Sdn. Bhd. were used in the production of the ammonium-based protic ionic liquids. All chemicals were of analytical grade. The amines and acids CAS numbers, abbreviations, and grades are as follows: ethanolamine (141–43–5, 99%), bis(2–ethylhexyl)amine (106–20–7, 99%), tributylamine (102–82–9, 99%), acetic acid (64–19–7, 99.8%), and butyric acid (107–92–6, 99%).

#### *2.2. Synthesis*

For the synthesis of each of the ammonium-based protic ionic liquids, an equimolar amount of the acid was added dropwise to the amine at ambient conditions and the mixture was consistently stirred for 24 h to facilitate mixing. The resulting solution was dried under vacuum at 65 ◦C for 6 h to remove remaining reactants. The final product was kept in a seal container until further use. The combinations of two acids and three amines produce six ammonium-based protic ionic liquids. Table 1 shows the structures and the abbreviation used for the ionic liquids. All ionic liquids exist as liquids except [BEHA][AC], which exists as a solid compound at room temperature.

**Table 1.** Structures of cations and anions, names and abbreviations.

#### *2.3. NMR and Water Content*

The structure of the ammonium-based protic ionic liquids was analyzed and confirmed via nuclear magnetic resonance (NMR) spectroscopy. About 5 mg sample of ionic liquid was dissolved in 6 mL deuterated solvent and the sample's purity was determined using 500 MHz Bruker NMR Oxford Instrument. Coulometric Karl Fischer autotitrator DL39 from Mettler was used to determine the water content of the ionic liquids.

#### *2.4. Thermophysical Characterization*

The viscosity and density of the ammonium-based protic ionic liquids were determined simultaneously using Anton Parr Stabinger Viscometer SVM3000 in the temperature range of 293.15 K to 363.15 K. The temperature measurement's accuracy was within 0.02 K while the reproducibility of the viscosity and density measurements were 0.35% and <sup>±</sup>5.10−<sup>4</sup> g·cm−3, respectively [28]. The decomposition temperatures of the ionic liquids were examined by means of thermogravimetric analyzer, TGA Perkin Elmer STA 6000. About 10 mg of sample was loaded into a platinum pan and the sample was heated at a heating rate of 10 ◦C·min-1 under nitrogen flow.

#### *2.5. CO2 Absorption Measurement*

The ability of the ammonium-based protic ionic liquids to absorb CO2 was investigated based on a pressure drop technique using a solubility cell as described in our previous publication [29]. The solubility cell consists of an equilibrium cell and a gas vessel immersed in a thermostatic bath. In a pressure drop method, the gas with a known pressure at constant volume is allowed to be in contact with the ionic liquid in the equilibrium cell and the pressure drop is monitored as the gas absorbs into the ionic liquid until equilibrium is attained. In a typical experiment, the equilibrium cell was loaded with a pre-weighed amount of the ionic liquid and the equilibrium cell was evacuated to remove any gases. In the gas vessel, CO2 was allowed to stabilize before being quickly charged into the equilibrium cell. The CO2-ionic liquid system was assumed to achieve equilibrium when the pressure attained a constant value. The system was maintained in that conditions for an additional two hours to ensure equilibration. Equation (1) was used to calculate the amount of CO2 absorbed in the ionic liquid, *n*<sup>2</sup> [30]:

$$m\_2 = \frac{P\_{\rm ini} V\_{\rm total}}{Z\_2(P\_{\rm ini}, T\_{\rm ini})RT\_{\rm ini}} - \frac{P\_{\rm cq} \Big| V\_{\rm total} - V\_{\rm liq} \Big|}{Z\_2(P\_{\rm cq}, T\_{\rm eq})RT\_{\rm eq}},\tag{1}$$

where *Pini* and *Tini* are the initial pressure and temperature of the system, *Peq* and *Teq* are the pressure and temperature of the system at equilibrium, *Vtotal* is the volume of the equilibrium cell, *Vliq* is the volume of ionic liquid, *R* is the gas constant, and *Z*<sup>2</sup> represents the compressibility factor of the gas. *Z*<sup>2</sup> can be calculated using Soave–Redlich–Kwong equation of state [31]. The mole fraction of CO2 absorbed in the ionic liquid (*x*2) was calculated using Equation (2):

$$\mathbf{x}\_2 = \frac{n\_2^{liq}}{\left(n\_2^{liq} + n\_1^{liq}\right)'} \tag{2}$$

where *n*<sup>2</sup> *liq* represents the mole of dissolved CO2 and *n*<sup>1</sup> *liq* is the mole of the ionic liquid.

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

In this work, six ammonium-based protic ionic liquids—ethanolammonium acetate [EtOHA][AC], ethanolammonium butyrate [EtOHA][BA], tributylammonium acetate [TBA][AC], tributylammonium butyrate [TBA][BA], bis(2-ethylhexyl)ammonium acetate [BEHA][AC], and bis(2-ethylhexyl)ammonium butyrate [BEHA][BA]—were synthesized and characterized. All the ammonium-based protic ionic liquids exist as liquids at room temperature except [BEHA][AC], which is a solid. The NMR and water content of each ionic liquids are presented as follows:

**[EtOHA][AC]**: 1H NMR (500 MHz, D2O): δ 3.613 [t, 2H, H2(-OH)], 2.922 [t, 2H, CH2(-NH2)], 1.726 [s, 3H, CH3]. Water content: 2.93%

**[EtOHA][BA]**: 1H NMR (500 MHz, D2O): δ 3.669 [t, 2H, CH2(-OH)], 2.988 [t, 2H, CH2(-NH2)], 2.010 [t, 2H, CH2(-COOH)], 1.375–1.448 [m, 2H, CH2], 0.747 [t, 3H, CH3]. Water content: 2.06%

**[BEHA][AC]**: 1H NMR (500 MHz, D2O): δ 2.861 [d, 4H, CH2(-NH)], 1.805 [s, 3H, CH3], 1.613–1.677 [m, 2H, CH], 1.181–1.309 [m, 16H, CH2], 0.786 [t, 12H, CH3]. Solid

**[BEHA][BA]**: 1H NMR (500 MHz, D2O): δ 2.856 [d, 4H, CH2(-NH)], 2.062 [t, 2H, CH2(-COOH)], 1.582–1.660 [m, 2H, CH], 1.412–1.486 [m, 2H, CH2(AC)], 1.236–1.339 [m, 16H, CH2], 1.214-1.229 [t,t 15H, CH3]. Water content: 0.15%

**[TBA][AC]**: 1H NMR (500 MHz, D2O): δ 3.002 [t, 4H, CH2(-NH)], 1.788 [s, 3H, CH3], 1.508–10571 [m, 6H, CH2], 1.207–1.282 [m, 6H, CH2], 0.805 [t, 9H, CH3]. Water content: 0.47%

**[TBA][BA]**: 1H NMR (500 MHz, D2O): δ 3.003 [t, 6H, CH2(-NH)], 2.031 [t, 2H, CH2(-COOH)], 1.510–1.573 [m, 6H, CH2], 1.394–1.468 [m, 2H, CH2(AC)], 1.211-1.285 [m, 6H, CH2], 0.809 [t, 9H, CH3], 0.768 [t, 3H, CH3(AC)]. Water content: 0.23%

#### *3.1. Thermophysical Properties*

The experimental density and dynamic viscosity values for all liquid samples of synthesized ammonium-based protic ionic liquids are presented in Tables 2 and 3. The experimental densities of [EtOHA][AC], [EtOHA][BA], [BEHA][BA], [TBA][AC], and [TBA][BA] as a function of temperature are shown in Figures 1–3. [BEHA][AC] is not included as it exists as solid. As can be seen from Figure 1, the density of all five ammonium-based protic ionic liquids decreased gradually and linearly with increasing temperature over the range of temperature studied. An increase in temperature caused higher mobility of the ions which, in turn, weakens the intermolecular forces between the constituent ions and correspondingly increases the unit volume for these ions [32]. The density of these ammonium-based protic ionic liquids was slightly affected by the length of the alkyl chain of the anion in which the density of ionic liquids with the [AC] anion was higher than of ionic liquids with the [BA] anion for a fixed cation, as shown in Figure 2a,b. This observation is consistent with the literature in which it has been shown that the density value drops as the alkyl chain gets longer [28,32–37]. Our experimental density value of [EtOHA][AC] is in good agreement with Kurnia et al. [35] and Hosseini et al. [38] with the value differences of less than 0.2% and 0.8%, respectively. Generally, effective arrangement of ions in a liquid can increase the density of the liquid due to a greater number of ions available in a unit volume [39]. Based on our experimental results, as shown in Figure 3, ionic liquids with the [EtOHA] cation have higher density values compared to the rest, suggesting a better packing of the ions due to the small size of the cation. [TBA][BA] had the lowest density values at all temperatures due to the combined effects of branching of the cation, which increases the asymmetricity and steric hindrance of the ionic liquid [32], along with the large size of the alkyl chain of the [BA] anion.

**Table 2.** Density (ρ) values of ionic liquids from 293.15 K to 363.15 K.


**Table 3.** Dynamic viscosity (η) values of ionic liquids from 293.15 K to 363.15 K.


**Figure 1.** Plot of experimental density (ρ) values of ammonium-based protic ionic liquids.

**Figure 2.** (**a**) Plot of experimental density (ρ) values of [EtOHA][AC] and [EtOHA][BA] and (**b**) plot of experimental density (ρ) values of [TBA][AC] and [TBA][BA] as a function of temperature.

**Figure 3.** Plot of experimental density (ρ) values of [EtOHA][BA], [BEHA][BA], and [TBA][BA] as a function of temperature.

The dynamic viscosity of the ammonium-based protic ionic liquids, presented in Figure 4, dropped significantly as the temperature increased and the viscosity of ionic liquids with a [BA] anion was higher than that of ionic liquids with an [AC] anion for each type of cation studied in this work, as shown in Figure 5. The longer the alkyl chain in the ionic liquid structure, the higher the viscosity of the ionic liquids due to the increase in van der Waals attraction between the aliphatic alkyl chains [35]. On the other hand, [EtOHA] ionic liquids have remarkable high viscosity values compared to [BEHA] and [TBA] ionic liquids due to the presence of the hydroxyl group which enables a hydrogen bonding interaction between the structures of the ions.

**Figure 4.** (**a**) Plot of experimental viscosity (η) values of ammonium based ionic liquids and (**b**) plot of experimental viscosity (η) values of [BEHA][BA], [TBA][AC], and [TBA][BA].

**Figure 5.** (**a**) plot of experimental viscosity (η) values of [EtOHA][AC] and [EtOHA][BA] ionic liquids and (**b**) plot of experimental viscosity (η) values of [TBA][AC] and [TBA][BA].

The values of density, ρ and dynamic viscosity, η were fitted using Equations (3) and (4) [28]:

$$
\rho = A\_0 + A\_1 T\_\prime \tag{3}
$$

$$
\log \eta = A\_2 + A\_3 / T\_\prime \tag{4}
$$

where ρ is the density, η is the dynamic viscosity of the ionic liquids, T is temperature in K, and *A*0, *A*1, *A*2, and *A*<sup>3</sup> are correlation coefficients determined using the method of least squares. The calculated correlation coefficients together with standard deviations, SD are presented in Tables 4 and 5. The standard deviations, SD, were calculated using Equation (5) in which *Zexpt* and *Zcalc* are experimental and calculated values, respectively, while *nDAT* is the number of experimental points:

$$SD = \sqrt{\frac{\sum\_{i}^{n\_{DAT}} \left(Z\_{\text{exp}\,t} - Z\_{\text{calc}}\right)^2}{n\_{DAT}}}.\tag{5}$$

**Table 4.** Fitting parameters of Equation (3) for density (ρ) correlation and standard deviation (SD) from Equation (5).


**Table 5.** Fitting parameters of Equation (4) for dynamic viscosity (η) correlation and standard deviation (SD) from Equation (5).


The thermal expansion coefficient, α*p*, for the ammonium-based protic ionic liquids can be calculated using Equation (6) [28] while the molecular volume, *Vm* can be estimated from Equation (7) in which *M* is the molar mass of the ionic liquid and *NA* represents the Avogadro's number [32,36,37]:

$$\alpha\_p = -1/\rho \cdot (\partial \rho/\partial T)\_p = - (A\_1)/(A\_0 + A\_1T) \text{ and} \tag{6}$$

$$N\_m = M \slash N\_A \cdot \rho.\tag{7}$$

The calculated thermal expansion coefficients and molecular volume values of the ammonium-based protic ionic liquids are presented in Tables 6 and 7. The calculated values lie in the range of (5.3 to 9.8)·10-4 K-1 for all five ionic liquids. The thermal expansion coefficients were found to be quite consistent over the temperature range studied and therefore are considered to be temperature independent. The pattern of the results is consistent with other types of ionic liquids [27,28,36,40,41]. The molecular volume, *Vm*, of the [BA] ionic liquid was greater than that of [AC] for a fixed cation and this may be attributed to the presence of additional CH2 groups [36,37]. In this work, the *Vm* decreased in the sequence of [BEHA] > [TBA] > [EtOHA] for a fixed anion, i.e., [BA].

**Table 6.** Thermal expansion coefficients (α*p*) of the ionic liquids calculated using Equation (6).


**Table 7.** Molecular volume (*Vm*) of the ionic liquids calculated using Equation (7).


The thermal decomposition (*Td*) of the ammonium-based protic ionic liquids were measured at a heating rate of 10 ◦C·min-1. The *Td* was approximately determined by the intersection of the baseline weight from the beginning of the measurement and the tangent of the weight against the temperature curve as the decomposition process occurs. The *Td* of the ionic liquids are presented in Table 8 and the thermal decomposition curves for [EtOHA][BA] and [BEHA][BA] are given in Figure 6. The thermal stability of the ammonium-based protic ionic liquids in this study varied with the ion combination. The *Td* of [BA] ionic liquid was higher than that of [AC] ionic liquid for every type of cation studied while [EtOHA] ionic liquids displayed the highest *Td* followed by [BEHA] and [TBA]. However, ammonium-based protic ionic liquids in this work and from the literature [36] tend to possess lower thermal stability compared to other ionic liquids, such as imidazolium and pyridinium

ionic liquids [28,41,42]. However, generalization must not be made as the thermal stability depends largely on the combination of the cation and anion of the ionic liquids.

**Table 8.** Thermal decomposition (*Td*) temperature of the ionic liquids.

**Figure 6.** Plot of thermal decomposition of [EtOHA][BA] and [BEHA][BA].

#### *3.2. CO2 Absorption*

The experimental results of CO2 absorption in the ammonium-based protic ionic liquids are shown in Figure 7. Generally, the CO2 absorption in these ammonium-based protic ionic liquids increased with pressure following Henry's law; the solubility of a gas in a liquid is proportional to the partial pressure of the gas above the surface of the liquid. The mol fraction of CO2 absorbed in the ammonium-based protic ionic liquids was in the range of about 0.02 to 0.48 and up to 20 bar at 298.15 K. The effects of cation structure on the CO2 absorption in the ionic liquids are shown in Figure 8. For a fixed anion, the solubility of CO2 increased in the sequence of [EtOHA] < [TBA] < [BEHA] where the mol fraction of CO2 absorbed in [BEHA][BA] was 0.486 in comparison to 0.328 and 0.307 in [TBA][BA] and [EtOHA][BA], respectively. Meanwhile, Figure 9 indicates a slight increase in the CO2 solubility when the anion of a common cation was changed from [AC] to [BA]. Based on our experimental results, there is a relationship between absorption of CO2 with the density and the molecular volume of the ionic liquids. As the density decreases and molecular volume increases, the fractional free volume increases and, thus, the solubility of CO2 increases [43,44]. By using a common anion, the CO2 absorption in [BEHA][AC] was compared with 1-ethyl-3-methylimidazolium acetate, [EMIM][AC] [45] and 1-butyl-3-methylimidazolium acetate, [BMIM][AC] [11]. At about 20 bar and 298.15 K, there was only a marginal difference between the CO2 solubility in [BEHA][AC] compared to that of [EMIM][AC] and [BMIM][AC]. This result shows a positive indication that our newly synthesized [BEHA][AC] and [BEHA][BA] ionic liquids have comparable performance towards CO2 capture when compared to

more established type of ionic liquids. However, more experimental investigation and data are needed to further evaluate the potential ability of our ammonium-based protic ionic liquids in the application of CO2 capture.

**Figure 7.** Plot of CO2 absorption in ammonium-based protic ionic liquids at 298.15 K.

**Figure 8.** (**a**) Plot of CO2 absorption in ammonium-based protic ionic liquids with [AC] anion and (**b**) plot of CO2 absorption in ammonium-based protic ionic liquids with [BA] anion at 298.15 K.

**Figure 9.** (**a**) Plot of CO2 absorption in ammonium-based protic ionic liquids with the [BEHA] cation; (**b**) plot of CO2 absorption in ammonium-based protic ionic liquids with the [TBA] cation; (**c**) plot of CO2 absorption in ammonium-based protic ionic liquids with the [EtOHA] cation and; and (**d**) plot of CO2 absorption in [BEHA][AC], [EMIM][AC] [45] and [BMIM][AC] [11] at 298.15 K.

#### **4. Conclusions**

Six ammonium-based protic ionic liquids were successfully synthesized via solvent-free 1-step neutralization reaction. The density, viscosity, and decomposition temperature were measured. The thermal expansion coefficient and the molecular volume were calculated using the density values. The density and viscosity values were inversely proportional with temperature in the range of temperature studied at atmospheric pressure. The density decreased when the alkyl chain of the anion increased, while the viscosity increased with the alkyl chain of the anion. The decomposition temperature of the ammonium-based protic ionic liquids was affected by the combination of cation and anion and [EtOHA] ionic liquids had the highest thermal stability when compared to the other ionic liquids. The absorption of CO2 in the six ammonium-based protic ionic liquids was measured at 298.15 K and up to a pressure of 20 bar. The CO2 absorption values in the ammonium-based protic ionic liquids increased with pressure and both the cation and anion affected the solubility of CO2 in the ionic liquids. The amount of CO2 absorbed was affected by the length of the alkyl chain of the anion while [BEHA] ionic liquids displayed higher CO2 absorption capacity compared to [TBA] and [EtOHA] ionic liquids. Results indicate the potential of the ammonium-based protic ionic liquids to be used as solvents for CO2 capture.

**Author Contributions:** Conceptualization, N.M.Y. and N.H.H.; methodology, N.H.H. and N.M.Y.; validation, N.H.H.; formal analysis, T.M.; resources, C.D.W. and T.M.; data curation, N.M.Y. and T.M.; writing—original draft preparation, N.M.Y. and N.H.H.; writing—review and editing, J.W.L. and P.L.S.; supervision, N.M.Y.; project administration, N.M.Y. and C.D.W.; funding acquisition, N.M.Y.

**Funding:** This research was funded by the Ministry of Higher Education (MOHE), Government of Malaysia, under the Fundamental Research Grant Scheme, [FRGS No: FRGS/1/2016/STG01/UTP/03/1] entitled *The impact of cations and anions structures of ionic liquids on the CO2 absorption from natural gas* and also YUTP grant (cost centre 015LC0-054).

**Acknowledgments:** Financial assistance and support from Universiti Teknologi PETRONAS and Center of Research in Ionic Liquids (CORIL), UTP are greatly acknowledged.

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

#### **References**


© 2019 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* **Physical and Thermal Studies of Carbon-Enriched Silicon Oxycarbide Synthesized from Floating Plants**

#### **Guan-Ting Pan 1, Siewhui Chong 2, Yi Jing Chan 2, Timm Joyce Tiong 2, Jun Wei Lim 3, Chao-Ming Huang 4, Pradeep Shukla 5,\* and Thomas Chung-Kuang Yang 1,\***


Received: 6 September 2019; Accepted: 28 October 2019; Published: 2 November 2019

**Abstract:** In the present study, amorphous mesoporous silicon oxycarbide materials (SiOC) were successfully synthesized via a low-cost facile method by using potassium hydroxide activation, high temperature carbonization, and acid treatment. The precursors were obtained from floating plants (floating moss, water cabbage, and water caltrops). X-ray diffraction (XRD) results confirmed the amorphous Si–O–C structure and Raman spectra revealed the graphitized carbon phase. Floating moss sample resulted in a rather rough surface with irregular patches and water caltrops sample resulted in a highly porous network structure. The rough surface of the floating moss sample with greater particle size is caused by the high carbon/oxygen ratio (1: 0.29) and low amount of hydroxyl group compared to the other two samples. The pore volumes of these floating moss, water cabbage, and water caltrops samples were 0.4, 0.49, and 0.63 cm<sup>3</sup> g−1, respectively, resulting in thermal conductivities of 6.55, 2.46, and 1.14 Wm−<sup>1</sup> K−1, respectively. Floating plants, or more specifically, floating moss, are thus a potential material for SiOC production.

**Keywords:** silicon oxycarbide; thermal conductivity; floating plants; SiOC; silica

#### **1. Introduction**

Silicon oxycarbide (SiOC) is an important material that finds application in semiconductor devices [1,2], electrode materials of lithium–ion batteries [3,4], micro electromechanical systems devices [5], high temperature sensors [6], conductive protective coatings [7], and super-capacitor [8] due to its superior mechanical properties. SiOC exhibits high strength [9], high chemical durability [10], excellent oxidation resistance under high temperature, good antioxidant crystallization property [11], and exceptional thermal expansion properties [12]. However, amongst these, thermal properties of silicon oxycarbide were rarely studied. Qiu et al. [13] reported a three-dimensional reticular macro-porous SiOC ceramic structure prepared by sol–gel process. The sample's porosity and specific surface area were reduced due to particle agglomeration, resulting in lower thermal conductivity that ranged from 0.041 to 0.062 Wm−<sup>1</sup> K<sup>−</sup>1. Recently, random porous structure has attracted an increasing

interest due to its radiation tolerance and very high crystallization temperature, which in turn resulted in good thermal conductivity [14,15]. Gurlo et al. [16] reported that the thermal conductivity of SiOC material containing zirconium and hafnium was 1.3 Wm−<sup>1</sup> K−<sup>1</sup> which was similar to that of silica. Mazo et al. [17] observed that silicon oxycarbide glasses that were synthesized by using spark plasma sintering had higher thermal conductivity values (≈1.38 Wm−<sup>1</sup> <sup>K</sup><sup>−</sup>1). Eom, et al. [18] prepared barium-added silicon oxycarbide (SiOC–Ba) via pyrolysis method and showed that the addition of Ba resulted in an increase in thermal conductivity values from 1.8 to 5.6 Wm−<sup>1</sup> K<sup>−</sup>1.

The traditional approach to producing amorphous SiOC microstructure has been via physical and chemical processing methods, such as powder metallurgy process [19], magnetron sputtering [20], sol–gel method [21,22], and chemical vapor deposition [23]. The raw materials used in synthesizing these SiOC materials were mainly chemical feedstocks such as tetraethoxysilane (TEOS)/hydroxyl-terminated polydimethylsiloxane (PDMS) organic–inorganic hybrid materials [17], polyxiloxane [18], methyltrimethoxysilane and dimethyldimethoxysilane [22], and dimethyl dimethoxy silane [21]. There are also limited studies on the thermal conductivity of SiOC materials. It could be expected that the material's structure affects its heat transfer properties as the type of particles and different pore sizes in the SiOC structure are the main reason to the different thermal conductivity abilities [12]. Low-cost sacrificial template has been widely used to generate replicated pore structures within such ceramic materials. Utilizing organic bio-mass, such as wood, or chemically modified biomass as a template for synthesizing SiOC with varying degree of porosity has been reported [24,25]. The use of selected biomass can also provide an added advantage of providing a source of silica for synthesis of SiOC.

These observations motivated our research for synthesizing SiOC using plant based biomass and to evaluate its physical properties and thermal conductivity. The main advantages of this method, as compared to other technologies, are its benign environmental nature and production of high-purity products via a low-cost process. In this study, silicon oxycarbide materials were synthesized by using a facile and inexpensive route from floating plants (floating moss, water cabbage, and water caltrops). The characterizations of these silicon oxycarbide materials were studied with regards to their structural, morphology, functional group, textural properties, and thermal conductivity.

#### **2. Experimental**

#### *2.1. Preparation of Materials*

All chemicals of analytical grade were purchased from Merck Chemicals. Three floating plants were used to synthesize carbon-enriched silicon oxycarbide (SiOC), namely, floating moss (Salvinia natans), water cabbage (Pistia stratiotes), and water caltrops (Trapa natans). The plants were collected from a pond at Kun Shan University, Tainan, Taiwan. The plants were washed and bone-dried in an oven at 100 ◦C before being grinded using a blender. For each sample synthesis, 0.071 mole potassium hydroxide (KOH) was dissolved in 40 mL of deionized water into which 1 gram of the dried and grounded biomass was added. The mixture was stirred for 2 h at 85 ◦C for activation before being annealed for 2 h in a tube furnace purged with nitrogen gas at 800 ◦C for carbonization. After that, the samples were added into 1.8 M hydrochloric acid (HCl) under magnetic stirring for 1 h. The resultant solid powder from the mixture solutions were washed by vacuum filtration until the pH value reached 7, and finally dried in vacuum at 110 ◦C overnight. The KOH activation enlarged the pores in these samples, thereby destroying their internal structure. These carbon-rich materials were then converted into pure carbon via high-temperature carbonization. Finally, acidic treatment (HCl) was used to alter the surface functional groups, surface morphology, and textural properties of the samples.

#### *2.2. Characterization of Carbon-Enriched Silicon Oxycarbide*

The crystalline structures of the carbon-enriched silicon oxycarbide samples were studied using an X-ray diffractometer (PANalytical X'Pert PRO, Almelo, the Netherlands) with copper K–alpha radiation (λ = 0.15418 nm) scanning from 10◦ to 70◦. The structures of all samples were analyzed using Raman Spectroscopy (DONGWOO DM500i, Gyeonggi-do, Korea). The morphology and compositions of the SiOC samples were analyzed via a field emission scanning electron microscopy (FESEM, JEOL JSM-6700F, JEOL, Peabody, MA, USA) with energy-dispersive X-ray spectroscopy (EDX). The morphology imaging was carried out at 5 k magnification with an accelerating voltage of 12 kV and a working distance of 12.1 mm. TEM observations of precipitates were performed on selected samples, using a transmission electron microscopy (TEM, JEM2100F, Akishima, Japan) at 200 kV. Fourier transform infrared (FTIR) spectroscopy (Perkin Elmer Spectrum GX, Shelton, CT, USA) was carried out to study the sample's structure using a diffuse reflectance infrared Fourier transform accessory (DRIFT) equipped with a heating cartridge. The pore textural characterizations of the samples were measured by a volumetric sorption analyzer (Micromeritics ASAP 2020, Micromeritics Instrument Corporation, Norcross, GA, USA). An adsorption/desorption isotherm of nitrogen gas was recorded at −196 ◦C and the pore size distribution was investigated by using the Barrett–Joyner–Halenda (BJH) model for the specific surface area and pore volume of the samples. The thermal conductivity absolute value was directly determined by a thermal conductivity probe (Mathis Instruments Ltd., Fredericton, Canada).

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

#### *3.1. Structural Analysis*

XRD patterns of the SiOC samples are shown in Figure 1. The patterns correspond to an amorphous Si–O–C material feature with disordered carbon and with broad (002) and (100) at 2θ angles of 22.5◦ and 44◦, respectively [26,27]. The peak at 44◦ is involved with small graphene sheets, and the peak at 22.5◦ means small graphene sheets stacked with random rotations or translations [27]. Raman spectra are also provided in Figure 2 for comparison. The Raman spectra of the prepared samples show the presence of two peaks at 1350 and 1582 cm<sup>−</sup>1, corresponding to D- and G-bands. The relative intensity/area ratio of D- and G-bands can be used to analyze the carbon material's structure. The G-band is a reflection of the presence of sp2 carbon–carbon bond hybridization within a graphitic ring structure, whereas the D band reflect a disordered carbon structure. It can be seen that the G band of these samples is broader than the D band, indicating the high graphitized carbon structure in these samples [28,29].

**Figure 1.** X-ray diffraction patterns of the silicon oxycarbide (SiOC) samples.

**Figure 2.** Raman spectra of the silicon oxycarbide (SiOC) samples.

#### *3.2. Morphology and Composition Analysis*

Figure 3 shows the field emission scanning electron microscopy (FESEM), TEM images, and EDX spectra of the prepared samples. Figure 3a shows that the floating mass sample contains a rough surface with irregular patches. Figure 3b, on the other hand, shows that the water cabbage sample contains a randomly distributed and overlapped structure. Contradictorily, Figure 3c shows an interconnected three-dimensional layered material for the water caltrops sample. Apart from the overall view by FESEM, all TEM images showed that these samples contained ordered hexagonal pore arrays with the majority of pores less than 1 nm.

The quantitative analysis was carried out with using energy dispersion analysis to determine the atomic ratios of C, O, and Si in these samples. As shown in Figure 3, the samples contain no contaminations other than C, O, and Si elements. Table 1 shows that the atomic ratios of C, O, and Si of the samples are 1:0.29:0.06, 1:1.13:0.48, and 1:2.24:1.07, respectively for the floating moss, water cabbage, and water caltrops samples. It is apparent that the element oxygen plays an important role to form structural layering. The surface roughness of the samples reduces with a decrease in the ratio of carbon/oxygen. The respective layers of water caltrops sample display a parallel arrangement due to the low ratio of carbon/oxygen. Shen et al. [30] reported that the surface structure becomes rougher when there is a deprivation of epoxide and hydroxyl groups, as well as a high C/O ratio.

**Figure 3.** *Cont.*

**Figure 3.** Field emission scanning electron microscopy (FESEM), transmission electron microscopy (TEM) and energy-dispersive X-ray spectroscopy (EDX) images (from left to right) of (**a**) Floating moss sample; (**b**) Water cabbage sample; and (**c**) Water caltrops sample.

**Table 1.** The C, O, and Si contents of the samples obtained by energy-dispersive X-ray spectroscopy (EDX) analysis.


#### *3.3. Fourier Transform Infrared Spectroscopy (FTIR) Spectra*

Figure 4 shows the FTIR spectra of the SiOC samples. Before the IR spectra analysis, the adsorbed water was removed by heating to 250 ◦C to advance the understanding of the functional group properties. The stretching vibration of siloxane bond (Si–O–Si) and the Si–O stretching vibration of the silanol group appear weak at 800 cm−<sup>1</sup> and 972 cm<sup>−</sup>1, individually [31,32]. The asymmetric vibration of the Si–O–Si obviously appears at 1100 cm−<sup>1</sup> [33] for water cabbage and water caltrops samples. However, in the floating moss sample, the Si–O–Si asymmetric vibration sharply reduces, due to the low amount of O and Si present in the structure. In addition, the floating moss sample has the lowest amount of hydroxyl groups, indicated by the low peak at 3450 cm<sup>−</sup>1. The water caltrops sample has the highest amount of hydroxyl groups. The amount of hydroxyl groups corresponds to the particle size from FESEM morphology in which the floating moss sample with the lowest amount of hydroxyl groups has a larger particle size compared to the water cabbage and water caltrops samples. It is speculated that the particle size of the floating moss sample is larger as the hydroxyl groups of the sample may generate some surface charge, thus resulting in agglomeration [34].

**Figure 4.** Fourier transform infrared (FTIR) spectra of the silicon oxycarbide (SiOC) samples.

#### *3.4. Porosity and Surface Area Study*

Figure 5a shows the nitrogen adsorption–desorption isotherms of the prepared samples. There is a very slow increase in N2 adsorption up to 0.80 of the relative pressure (P/P0) before the sharp rise in the adsorbed volume when capillary condensation occurred. A H2-type hysteresis loop consisting of a triangular shape and a steep desorption branch of the isotherms can be found, indicating the existence of highly meso-pores with narrow mouths and wider bodies (ink-bottle pores) [35]. Figure 5b shows the corresponding pore size distributions of the prepared samples. There is a narrow pore size distribution centered at around 2.75 nm with a small meso-pore part in the floating moss sample and water cabbage sample. In addition, the water caltrops sample has the most porous structures with pore diameters ranged from 4 to 32 nm, indicating that it belongs to hierarchical porous structure. Table 2 shows the specific surface area, pore volume, and average pore diameter calculated using the Barrett–Joyner–Halenda (BJH) equation.

**Figure 5.** *Cont.*

**Figure 5.** (**a**) Nitrogen adsorption/desorption isotherms and (**b**) pore size distributions of the silicon oxycarbide (SiOC) samples.

**Table 2.** Textural characterization for the silicon oxycarbide (SiOC) samples.


The pore volumes (Vpore) of the floating moss, water cabbage, and water caltrops samples are respectively 0.4, 0.49, and 0.63 cm3 g<sup>−</sup>1. The Brunauer–Emmett–Teller (BET) specific surface areas (SBET) of the floating moss sample is greater than the water cabbage and water caltrops samples. The BJH adsorption-average pore diameters (Dp) are 7.00, 6.04, and 4.75 nm, respectively, for the floating moss, water cabbage, and water caltrops samples. The texture properties of the prepared samples agree with the surface morphology results in which the floating moss sample having the largest grain size has the lowest pore volume and largest pore diameter.

#### *3.5. Thermal Conductivity*

The thermal conductivities of these samples were studied using Yang et al.'s method [36]. The thermal conductivities of these samples are shown in Table 2 and compared with other literature in Table 3. As we can see, the thermal conductivity of the floating moss sample is the highest among these samples due to lowest pore volume and larger grain size [12]. The water caltrops sample has the lowest thermal conductivity among these samples, indicating that the surface hydroxyl groups on silica reduced the thermal conductivity of the sample due to the increased amount of voids [37]. Eom et al. [38] also reported that the equivalent porosity is the reason for low thermal conductivity. Thermal conductivity of these SiOC samples reduces in the order of floating moss (6.55 Wm−<sup>1</sup> K−1), water cabbage (2.46 Wm−<sup>1</sup> K−1), and water caltrops (1.14 Wm−<sup>1</sup> K−1). However, when compared with other literature (Table 3), the thermal conductivity of our floating-plant derived SiOC samples is considerably higher and comparable with those made from chemical feedstock. It is also a key finding that the surface hydroxyl groups on silica decreases the thermal conductivity. Nevertheless, among the

three floating plants: floating moss, water cabbage, and water caltrops, the floating mass plant is the candidate with the most potential to replace the use of chemical precursors in making SiOC.


**Table 3.** Comparison of thermal conductivity with other literature.

#### **4. Conclusions**

In this study, silicon oxycarbide materials were synthesized via a simple and low-cost method from floating plants. Three floating plants: floating moss, water cabbage, and water caltrops were used as the precursors of the SiOC powders. XRD shows the amorphous Si–O–C material structure as the major phase. The morphologies and textural properties of the SiOC samples were porous micro-structures. The maximum thermal conductivity of the prepared samples in this study was found to be 6.55 Wm−<sup>1</sup> K−<sup>1</sup> from floating moss. The surface hydroxyl groups of the SiOC samples may have possibly reacted with the silica atom, resulting in interconnecting bonds and thus larger grain size which enhances the thermal conductivity. Overall, the results demonstrated the feasibility of using floating plants as the precursor in making SiOC powder with high thermal conductivity via a facile and low-cost procedure, making it a suitable process for industrial applications.

**Author Contributions:** Conceptualization and writing, G.-T.P.; Review, S.C.; Validation, Y.J.C.; Sample analysis, T.J.T.; Funding acquisition, J.W.L.; Resources, C.-M.H.; Editing and finalization, P.S.; Supervision, T.C.-K.Y.

**Funding:** The financial support Universiti Teknologi PETRONAS via YUTP-FRG with the cost center 0153AA-E48 is gratefully acknowledged. Funding from Ministry of Education Malaysia through HICoE awarded to the Centre for Biofuel and Biochemical Research, Universiti Teknologi PETRONAS is as well duly acknowledged.

**Acknowledgments:** The authors would like to thank Kuan Ching Lee, Ho-Shin Shiu, Ping-Chun Lin, and Yi-Hsuan Lai for their assistance in conducting the experiments.

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

#### **References**


© 2019 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* **Conversion Technologies: Evaluation of Economic Performance and Environmental Impact Analysis for Municipal Solid Waste in Malaysia**

#### **Rabiatul Adawiyah Ali 1, Nik Nor Liyana Nik Ibrahim 1,\* and Hon Loong Lam <sup>2</sup>**


Received: 22 August 2019; Accepted: 24 September 2019; Published: 16 October 2019

**Abstract:** The generation of municipal solid waste (MSW) is increasing globally every year, including in Malaysia. Approaching the year 2020, Malaysia still has MSW disposal issues since most waste goes to landfills rather than being utilized as energy. Process network synthesis (PNS) is a tool to optimize the conversion technologies of MSW. This study optimizes MSW conversion technologies using a PNS tool, the "process graph" (P-graph). The four highest compositions (i.e., food waste, agriculture waste, paper, and plastics) of MSW generated in Malaysia were optimized using a P-graph. Two types of conversion technologies were considered, biological conversion (anaerobic digestion) and thermal conversion (pyrolysis and incinerator), since limited data were available for use as optimization input. All these conversion technologies were compared with the standard method used: landfilling. One hundred feasible structure were generated using a P-graph. Two feasible structures were selected from nine, based on the maximum economic performance and minimal environmental impact. Feasible structure 9 was appointed as the design with the maximum economic performance (MYR 6.65 billion per annum) and feasible structure 7 as the design with the minimal environmental impact (89,600 m3/year of greenhouse gas emission).

**Keywords:** optimization; P-graph; municipal solid waste conversion technology

#### **1. Introduction**

Municipal solid waste (MSW) is material arising from human activities. It is generated commonly from different areas such as residential, commercial, and institutional zones, as well as public parks [1]. The generation of MSW is drastically increasing globally every year, by a factor of 2.6 [2]. In 2016, the world's MSW generated was around 2.01 billion tons, and this figure is expected to increase to 3.40 billion tons by 2050 [3].

In Asia, MSW generation is expected to reach 1.8 million tons every day in 2025, as more than 1 million tons of MSW is currently being generated every day [4]. Based on a survey conducted by the Malaysian government, MSW generation in Malaysia has increased from 23,000 tons/day in 2008 to 33,000 tons/day in 2012 [5]. The increases of MSW generation in Malaysia are caused by three significant factors: (i) the rapid increase in population; (ii) accelerated urbanization; and (iii) increased industrialization processes [6]. The total population of Malaysia in 2017, as mentioned by the World Bank [7], was 31.62 million. As the population increases, the per capita generation rate also increases. For Malaysia, the MSW per capita generation rate range was of 0.6–0.8 kg per capita per day between 2001 and 2005 [8]. This number is expected to increase to double digits by the year 2020 [9]. Based on the World Bank's report, the waste generation per capita in Malaysia increased by up to 1.00–1.49 kg

per capita per day by September 2018 [3]. At present, 54% of the world's population lives in urban areas, and this percentage will increase to 66% or more by 2050 [10].

Increasing MSW generation has become the most prominent environmental issue as MSW may contain dangerous substances that are harmful to our ecosystem and increase the potential risk to our health. MSW must be appropriately disposed of and managed efficiently. Many significant environmental issues may arise from this kind of waste, such as the generation of greenhouse gases (GHGs) released from MSW. Besides, the increasing number of landfills can increase the numbers of rodents and insects that may cause diseases to humans. In recent years, more landfill sites are needed to dispose of all the MSW generated [11]. The main issue we face this traditional disposal method is shortage of landfill sites inland [12]. An essential component of a healthy society and a sustainable environment is an efficient waste management system [13].

The main purpose to manage MSW efficiently include reducing (i) the amount of MSW generated, (ii) the impact on the environment with a lower cost of disposal of MSW, and (iii) the impact on human health [14]. In MSW management in developing countries, five typical problems can be identified: (i) inadequate service coverage, (ii) operational inefficiency of services, (iii) limited utilization of recycling activities, (iv) poor management of non-industrial hazardous waste, and (v) shortage of landfill disposal sites [15]. The present waste management method in Malaysia depends on landfill [11]. Only 5.5% of MSW is recycled and 1% is composted, while the remaining waste goes to landfill [16]. Currently, there are 174 landfills around Malaysia [14]. The recycling rate increased from 5.5% in 2009 to 10.5% in 2012 [17]. Malaysia's recycling rate was 17.5% in 2016, which is still far from the target of 22% by 2020 [18].

One method to solve problems related to landfills is to introduce sustainable and efficient waste management [6]. Integrated waste recycling and various conversion technologies could be effective waste management strategies [19]. There are several steps in sustainable and efficient waste management [20]. Possible waste generation and their conversion technologies are illustrated in Figure 1.

**Figure 1.** Possible waste generation relationship between their conversion technologies, reproduced with permission from [20]. Copyright Elsevier, 2017.

Based on Figure 1, there are many pathways for conversion technologies to manage MSW after it is collected from residences and processed before being sent to landfills. The three main steps to handle MSW are (i) the collection and transportation of MSW; (ii) the treatment and processing of MSW; and (iii) its final disposal [21]. Each stage has its own investment cost, operating cost, and energy recovery.

However, there are pros and cons to each conversion technology. It is useful to utilize analytical tools to synthesize a promising waste management strategy [22]. There are limited studies on the performance of conversion technology in the context of Malaysia. The optimization of MSW conversion technology will help decide the most favorable and useful method and pathway in managing MSW. Through optimization, we can introduce combined conversion technology to manage Malaysia's MSW.

A process network synthesis (PNS) problem is defined as specifying the raw materials, operating units, and desired products in chemical engineering problem, for example the conversion technologies problem. The PNS problem was developed as a mathematical model in which variables correspond to decisions, such as input and output flow rates, with a limitation corresponding to the mathematical description of the optimization criterion such as the material balance objective function [23]. The common problems in a PNS are (i) the reaction pathway; (ii) process design; (iii) the heat exchangers network; (iv) the water integration system; and (v) the separation unit [24]. The process graph, best known as the P-graph, is one method to solve the PNS problem [25]. The P-graph is a graphical optimization which is available in the software P-Graph Studio [26]. The P-graph is a bi-graph, meaning that its vertices are in disjunctive sets and there are no edges between vertices in the same set [27]. The vertices of the P-graph are denoted as the operating unit (O) and the material (M). This P-graph represents the material flow between the material and the operating unit. The P-graph methodology was originally developed for PNS problems in chemical engineering applications. The P-graph methodology is based on five axioms [28], as follows:


To summarize, the P-graph methodology is composed of the following algorithms: (i) maximal structure generation (MSG); (ii) solution structure generation (SSG); and (iii) accelerated branch and bound (ABB) [29]. The MSG algorithm identifies a network structure, which is the union of all possible solution structures of the problem. It can be generated in polynomial time using the information specified in the five axioms. The SSG algorithm generates all combinatorically feasible solution structures or networks. Each solution is a subset of the maximal structure and represents a potential network configuration for the PNS problem. The ABB algorithm identifies the optimal structure based on the solution structures, in conjunction with additional problem-specific information.

The P-graph framework enables rigorous model building and the efficient generation of optimal solutions [29]. The PNS problem primarily utilizes unique information. The P-graph is known as a user-friendly decision-making tool for PNS. This helps in better design and better operations that lead to (i) lower capital and operating cost (CAPEX and OPEX); (ii) higher profitability through increased output and better quality of the product; (iii) reduced technology risk; and (iv) better health, safety, and environmental requirements. These factors may thus help in optimizing MSW conversion technology.

The main objective of optimization is maximizing the efficiency of production by minimizing the cost of production. Therefore, it is essential to optimize MSW conversion technologies using a process graph to evaluate the selected pathway. Table 1 shows different types of optimization models for solid waste management based on previous studies. Data are tabulated based on the optimization method used, the objective of the study, the focus of the study, and the optimization of economic performance (EP) and environmental impact (EI).


**Table 1.** Optimization models for solid waste management.

Although there are various models for optimizing MSW conversion technologies, we still cannot manage MSW efficiently in Malaysia, as there are no integrated conversion technologies for solid waste treatment. Therefore, this study aimed to simulate the feasibility of MSW conversion technologies and analyzed the EP and EI of MSW conversion technologies. The study framework was based on the following factors:


The proposed processing network was designed using the P-graph model. There are a few types of MSW conversion technologies, including landfill, anaerobic digestion, incineration, and pyrolysis. The selected optimized conversion technology for MSW was then further assessed with respect to the impacts on feedstock and products on GHG emissions, demand, and prices. Two different scenarios were considered in this case study, which was designed for maximum EP and design for minimal EI.

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

Figure 2 shows the intracellular synthesis procedure for the process graph (P-graph). The procedure starts with the identification of materials and streams to yield the optimal MSW conversion technology network.

The intracellular synthesis procedure started with the identification of materials, streams, and operating units. After that, data input was required to generate the maximal superstructure and solution structure. The procedure ended with an optimal MSW conversion technology network.

**Figure 2.** Intracluster synthesis procedure for the process graph (P-graph), reproduced with permission from [36]. Copyright Elsevier, 2010.

#### *2.1. Identification of Materials and Streams*

This step produced the details for the inputs and outputs of the system. In this study, there were four types of process feedstock. There are six types of outputs or products, along with their intermediate products, as illustrated in Figure 3.

**Figure 3.** Superstructure for process flow managing MSW in this case study.

Figure 4 shows the circle nodes for the materials and the directed arrows represent as the streams. The number attached on the arrow signifies the consumption or production rate that represents the relationship between a material and an operating unit. Table 2 shows the list of raw materials, intermediate products, and products of the conversion technologies.

**Figure 4.** Graphical representation of materials in P-graph.



#### *2.2. Identification of Operating Units*

For this case study, 11 operating units were included in the flowsheet-generation problem shown to be solved algorithmically with P-graphs in Table 3. For this case study, anaerobic digestion, incineration, and pyrolysis were identified as the MSW conversion technologies in the model as an operating unit. The relationships between raw materials and operating units involved are shown in Figure 3.



#### *2.3. Input Data for Waste and Related Conversion Technologies*

For this study, four types of MSW were chosen, based on the largest MSW composition generated in Malaysia: food waste, agricultural waste, plastics, and paper. The composition of MSW is shown in Table 4. Organic waste is the main component of MSW in Malaysia, representing up to 50% of the total waste.


**Table 4.** Composition generation of MSW in Malaysia [5].

The study was conducted using data based on a literature review of MSW generation and MSW conversion technology. Data used were based on different studies and resources. The three types of conversion technologies considered in this case were pyrolysis, incineration, and anaerobic digestion, as limited data were available to be used for the optimization process. All these technologies were compared with the common method of MSW disposal in Malaysia, which are landfills. The allocation of waste to conversion technologies is illustrated in Table 5.

Both types of organic waste (i.e., food and agricultural) undergo an anaerobic digestion process. For inorganic waste, plastics and paper undergo two thermal treatments pyrolysis and incineration. After treatment, all waste is disposed of in landfills.


**Table 5.** Waste allocation to conversion technology.

#### *2.4. Optimization of Superstructure*

The results of the solution structure generated from the previous step are then utilized in the selection of an optimal network using the solution structure generation and linear programming (SSG + LP) algorithm, allowing for the design of optimal process networks based on the solution structures in conjunction with additional problem-specific information, such as flow rates and costs. Consequently, the solution that provides the selected pathways with the best and near-optimum solutions is obtained, as shown in Figure 5. The selected optimized conversion technology for MSW was then further used to access the impacts on feedstock and products on GHG emissions, demand, and prices. Two different scenarios were considered: scenario 1: a design for maximum EP; and scenario 2: a design for minimal EI.

**Figure 5.** A P-graph representation of the municipal solid waste process network.

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

One hundred feasible structures were generated using the P-graph with the SSG + LP algorithm. The SSG generates all combinatorically, feasible solution structures or networks. Each solution is a subset of the maximal structure and represents a potential network configuration for the PNS problem. The LP is the process of finding the best solution under specific conditions.

Of the 100 feasible structures generated, nine were selected to identify and analyze their EP and EI. These nine feasible structures convert all types of waste into the final products. Four types of MSW (food waste, agriculture waste, plastics, and paper) were converted using three different types of conversion technologies (i.e., anaerobic digestion, incineration, and pyrolysis) to generate six main products, i.e., solid fertilizer, liquid fertilizer, heat, electricity, GHGs, and biochar.

#### *3.1. Comparison of Di*ff*erent Feasible Structures*

Figure 6 shows the profit generated for each feasible pathway. The profit generated was calculated by the total gain of the product minus the total cost of raw materials. The highest profit generated was feasible in structures 7 and 9. Both these feasible structures gave the same profit value, which was MYR 6.65 billion per annum. However, as shown in Table 6, feasible structure 7 did not generate two products: electricity and heat. The products generated by feasible structure 7 were solid fertilizer, liquid fertilizer, GHGs, and biochar; while feasible structure 9 generated all six products: solid and liquid fertilizers, heat, electricity, GHGs, and biochar.

**Figure 6.** Economic analysis for each feasible structure.



The lowest profit was generated by feasible structure 1. The value gained from the product was much lower than the cost of raw materials. The cost of the raw materials was MYR 746 million, while the value gained from the product was MYR 301 million. There was around a 59.7% loss from this feasible structure because no fertilizer was generated from the digestate (a by-product of anaerobic digestion), as anaerobic digestion has a massive conversion of around 75% from raw materials to digestate before undergoing treatment to convert it into two types of fertilizer: solid and liquid.

For each feasible structure, at least five types of operating units were used. For a feasible structure that generates electricity and heat, at least one operating unit involved either a boiler or a gas turbine. For the generation of liquid and solid fertilizers, the digestate must undergo pre-treatment before it can be sold as products. Table 7 shows the operating units that affected the volume of GHGs generated. The GHG emissions were generated from landfills, gas turbines, and boilers. From landfills, 0.1605 m<sup>3</sup> GHGs per ton of MSW were released into the surroundings. The different types of technology involved in converting MSW gives a different ratio of GHG emissions produced.


**Table 7.** Operating units that affected the volume of GHG emissions.

The lowest three structures that generated GHGs were feasible structures 7, 8, and 9 with values of 89,600, 140,000, and 16,500 m3/year as shown in Figure 7. The GHGs generation was affected by only one piece of operating equipment for feasible structures 7 and 9. Since the conversion of GHG emissions from boilers was only 0.1 m<sup>3</sup> per combustible gas, the generation of GHGs for feasible structure 9 was lower than for feasible structure 7, which was affected by the generation of GHGs from landfills as mentioned earlier. Although feasible structure 8 had two pieces of equipment that affected the generation of GHGs, it was one of the top three feasible structures with a low generation of GHGs. Comparing feasible structures 3 and 4, both had a lower generation of GHGs because both used a gas turbine that affected their GHG generation. The gas turbine had a higher conversion of GHG emission, which was 0.505 m<sup>3</sup> per biogas. The highest GHG volumes were generated from feasible structures 1, 2, 5, and 6. All these feasible structures used three to four operating units, which affected the volume of GHGs generated.

**Figure 7.** Generation of greenhouse gas for each feasible structure.

#### *3.2. Scenario 1: Maximum EP*

Feasible structure 9, as shown in Figure 8, was selected as the structure with the maximum EP. Based on Figure 6, the maximum EP of the selected pathways was estimated to be MYR 6.65 billion per annum or considering the total population of Malaysia in 2017, as mentioned by the World Bank [7] (31.62 million), it was estimated to be MYR 210 per person. For this feasible structure, both organic wastes (i.e., food and agriculture waste) underwent anaerobic digestion in operating units Digester\_1 and Digester\_2, which produced biogas and digestate. The digestate was separated into two types of fertilizer, liquid and solid, after pre-treatment. However, biogas did not undergo further treatment to convert it into electricity and heat. Plastics were burned in operating unit Pyrolysis\_2 to be converted into biochar. Paper was burned in Incinerator\_1 to produce both ash and combustible gas. The combustible gas was used in the boiler to convert it into electricity and heat. From this feasible structure, GHGs were produced from the boiler. The highest EP yield products, which were electricity, heat, solid fertilizer, liquid fertilizer, GHG emissions, and biochar, had flow rates of 82,700 kWh/year, 215,000 J/year, 1,030,000 tons/year, 4,110,000 tons/year, 33,100 m3/year, and 795,000 tons/year, respectively.

**Figure 8.** Feasible structure with the maximum economic performance.

The capital cost for a year was estimated to be MYR 123,587,000,000. This capital cost value is for the first year only. The profit margin was calculated as the net profits divided by the revenue. The profit margin for this feasible structure was 89.9%. The payback period of this pathway was 16.7 years. The payback period was calculated to identify the time required to earn back the investment money on the project.

#### *3.3. Scenario 2: Minimal EI*

Feasible structure 7, as shown in Figure 9, was selected as the structure with the minimal EI. Based on Figure 6, the minimal EI of the selected pathways was estimated to be MYR 6.65 billion per annum or considering the total population of Malaysia in 2017 as mentioned by the World Bank (31.62 million), it was estimated to be MYR 210 per person. This selected feasible structure was the same as the feasible structure of the maximum EP. However, feasible structure 7 produced less GHGs as GHGs are emitted only from landfills. For this feasible structure, both types of organic waste, food and agriculture waste also underwent anaerobic digestion in operating units Digester\_1 and Digester\_2, which produced biogas and digestate. The digestate was again separated into two types of fertilizer,

liquid and solid, after pre-treatment. However, biogas did not undergo further treatment to convert it into electricity and heat. Plastics were also burned in operating unit Pyrolysis\_2 to be converted into biochar. Paper was burned in Incinerator\_1 to produce both ash and combustible gas Combustible gas was not converted into electricity and heat, and ash was removed to landfills. From this feasible structure, GHG emissions were produced from landfills only. The highest EP yield products were solid fertilizer, liquid fertilizer, GHG emissions, and biochar at flow rates of 1,030,000 tons/year, 4,110,000 tons/year, 89,600 m3/year, and 795,000 tons/year, respectively.

**Figure 9.** Feasible structure with the least environmental impact.

#### **4. Conclusions**

In this study, for the feasibility of MSW conversion technologies by PNS, "process graphs" were simulated. One hundred feasible structures were generated and nine of these were selected randomly for further analysis. Next, the EP and EI of the MSW conversion technologies were analyzed. Feasible structure 9 was chosen as the design with the maximum EP, with a total annual profit gain of MYR 6.65 billion, requiring up to 16 years as the payback period, with a constant flow of products. Feasible structure 7 was chosen as the design with the minimal EI, generating 89,600 m3/year of GHGs for the whole of Malaysia.

Further studies should be conducted using a real case study. This method of study would be more convenient for optimizing a small case study, focusing primarily on a district rather than on the whole country. This study was a feasible-structure-based preliminary study to treat MSW in the whole country. For example, the payback period of conversion technologies could take up to 16 years because of the high cost of developing conversion technologies to manage the whole country's MSW.

**Author Contributions:** Conceptualization, R.A.A. and N.N.L.N.I.; methodology and software, R.A.A. and H.L.L.; writing-original draft preparation, R.A.A.; writing-review and editing, R.A.A. and N.N.L.N.I; supervision, N.N.L.N.I.

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

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

#### **References**


© 2019 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/).
