*Article* **Wet Flue Gas Desulphurization (FGD) Wastewater Treatment Using Membrane Distillation**

**Noah Yakah 1,\*, Imtisal-e- Noor 2, Andrew Martin 2,\*, Anthony Simons <sup>1</sup> and Mahrokh Samavati <sup>2</sup>**


**Abstract:** The use of waste incineration with energy recovery is a matured waste-to-energy (WtE) technology. Waste incineration can reduce the volume and mass of municipal solid waste significantly. However, the generation of high volumes of polluting flue gases is one of the major drawbacks of this technology. Acidic gases are constituents in the flue gas stream which are deemed detrimental to the environment. The wet flue gas desulphurization (FGD) method is widely employed to clean acidic gases from flue gas streams, due to its high efficiency. A major setback of the wet FGD technology is the production of wastewater, which must be treated before reuse or release into the environment. Treating the wastewater from the wet FGD presents challenges owing to the high level of contamination of heavy metals and other constituents. Membrane distillation (MD) offers several advantages in this regard, owing to the capture of low-grade heat to drive the process. In this study the wet FGD method is adopted for use in a proposed waste incineration plant located in Ghana. Through a mass and energy flow analysis it was found that MD was well matched to treat the 20 m3/h of wastewater generated during operation. Thermal performance of the MD system was assessed together with two parametric studies. The thermal efficiency, gained output ratio, and specific energy consumption for the optimized MD system simulated was found to be 64.9%, 2.34 and 966 kWh/m3, respectively, with a total thermal energy demand of 978.6 kW.

**Keywords:** waste-to-energy; municipal solid waste; flue gas desulphurization; membrane distillation; thermal performance; thermal efficiency; gained output ratio; specific energy consumption

#### **1. Introduction**

Waste-to-energy (WtE) technology has been established to be an appropriate method of dealing with municipal solid waste (MSW) worldwide [1]. Most developed countries have embraced the use of it as an attractive means of treating non-recyclable and non-reusable waste because not only it minimizes the risks and environmental concerns associated with disposing copious quantities of MSW into landfill sites, but also it allows production/recovery of useful energy (e.g., electricity). The use of MSW as fuel to generate energy can reduce the over-dependence on fossil fuels as sources of energy.

Waste incineration is the most matured and widely used WtE technology [1–3]. Waste incineration can reduce the volume of MSW by 80 to 95% and the mass by 70 to 75% [4]. The major setback with the use of this technology is the significant amount of pollutants that are produced. There are some constituents (e.g., hydrofluoric (HF), hydrochloric (HCl), and sulphur dioxide (SO2)) in the flue gases emanating from waste incinerators which are proven to have a detrimental impact on the environment [5]. Therefore, strict limits on the amount of such constituents are set. Different technologies have been considered for removal of such emissions (e.g., wet, semi-dry, and dry scrubbing).

Absorption and adsorption are two distinct processes by which acidic gases are cleaned from flue gas streams emanating from waste incinerators. These processes are classified either as non-regenerative or regenerative. The non-regenerative technologies are further

**Citation:** Yakah, N.; Noor, I.-e.; Martin, A.; Simons, A.; Samavati, M. Wet Flue Gas Desulphurization (FGD) Wastewater Treatment Using Membrane Distillation. *Energies* **2022**, *15*, 9439. https://doi.org/10.3390/ en15249439

Academic Editor: Alessandra Criscuoli

Received: 13 October 2022 Accepted: 9 December 2022 Published: 13 December 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

divided into three; dry, semi-dry and wet methods [6,7]. Figure 1 depicts the various methods in non-regenerative flue gas desulphurization (FGD) technologies that are used in acid gas cleaning. Wet scrubbing is a FGD method that is widely employed in cleaning acidic gases in flue gas streams, with Japan, USA and Germany enjoying the most patronage [8]. Advantages of this technology includes higher rates of desulphurization, relative ease of operation, and smaller equipment. The limestone wet method of FGD is reported to be the most widely used technology in the cleaning of acidic gases in flue gas streams from waste incineration plants [9]. A study by Lecomte et al. [10] indicates that removal efficiencies of up to 99% can be achieved using this technology. The major disadvantage with the use of this method is the production of wastewater, which must be treated prior to reuse or release into the environment. There are several wastewater treatment methods that are employed in the treatment of various types of wastewater. Membrane separation methods employed in the treatment of wastewater include microfiltration, ultrafiltration, reverse osmosis (RO), etc. RO has enjoyed the most patronage among the various types of membrane separation techniques. However, drawbacks include higher electricity demand in providing high pressures at the intake, consequently affecting the membrane's long-term performance [11]. There is, therefore, the need to explore other membrane separation processes.

Membrane distillation (MD) is a promising novel technology used in separation processes where only water molecules can pass through a porous hydrophobic membrane material. The application of MD technology is widespread in the desalination of seawater and brackish water [12–15], and the treatment of wastewater that is polluted with radioactive substances [16,17]. MD is also reported to have potential in the treatment of oily wastewater from industries [18–20]. However, relatively few works have been carried out on the application of MD technology in the treatment of flue gas condensate. Chuanfeng et al. [21] investigated the use of MD in the framework of combined heat and power (CHP) plants in Sweden. Subsequently, a pilot unit was installed at the Vattenfall Idb .. a cken CHP plant (a biofuel-fired plant) during 2006 and 2007 [22]. Later, a follow-up investigation considered water recovery from flue gas condensate in MSW-fired cogeneration plants using MD [11], and employed both laboratory and pilot-scale air gap MD modules in combination with techno-economic analyses. The aforementioned studies demonstrated that MD offers equal or superior separation efficiency as compared to RO with higher specific costs. The availability of low-grade heat placed an upper limit on the capacity of the MD system to about 100 m3/h. However, their work was carried out on cogeneration plants, which supply both heat and electricity, and not for WtE facilities operating in condensing mode, where electricity is the only energy service provided. Therefore, there is the need for further investigation if the MD technology is to be integrated in a WtE plant in a tropical country such as Ghana, where the heat demand is low and thus cogeneration is unprofitable when selected. Moreover, to meet the MD energy demand, the possibility of using the recovered waste heat during the cooling of the flue gas stream prior to the particulate matter (PM) separation process is considered.

This study, therefore, mainly focuses on different models of a waste incineration plant simulated using Aspen Plus® software (version 11). In other words, three of the developed models used for the study are presented, with emphasis on the simulation of acid gas cleaning using wet FGD and the treatment of the produced wastewater via MD technology. The current study is a subsequent part of a research investigation with the broader aim of proposing optimal integration of WtE in Ghana.

**Figure 1.** Non-regenerative flue gas desulphurization technologies [23].

#### **2. System Description**

This section describes the overall flow chart used in the study, and the system design of a wet scrubbing system and how it is integrated to clean acidic gases in the flue gas stream.

#### *2.1. Block Flowchart of the Incineration Plant Used in the Study*

The integrated system block flow chart for the waste incineration plant used in this research is depicted in Figure 2. MSW is first fed into the incinerator/boiler, where sufficient air is added to aid in the complete oxidation of the MSW. After combustion of the MSW in the incinerator/boiler, the flue gas (which carries with it a high energy and particulate matter) and ash are produced. While ash is collected at the bottom, the flue gas stream exits into a heat recovery steam generator (HRSG). After heat exchange with high pressure water to produce high pressure steam, the flue gas is first cooled down further before entering a particulate matter (PM) separation device. At this stage, PM is separated from the flue gas stream using either a cyclone, filter bag, electrostatic precipitator PM separation device, or a combination of these. Then the flue gas stream goes into the wet FGD where water and aqueous calcium carbonate (CaCO3) are added to clean out acidic gases. The flue gas stream is further cleaned before its release to the environment through a stack. The produced wastewater after acid gas cleaning, on the other hand, is sent to the MD system for treatment prior to reuse or disposal into the environment. The wastewater that goes into the MD system is treated and produces a cleaned water (permeate), and the captured solids and the remains in the concentrate (retentate) can be returned to the MD system for further cleaning or disposed.

#### *2.2. Design of a Wet Scrubbing System*

A typical design of a limestone wet method of FGD technology has two basic stages. In the first stage, only water (acidic condition) is added. This stage is only effective in the removal of HF, HCl and sulphur trioxide. The removal of SO2 in the first stage is low due to the presence of HCl, which affects its absorption. Therefore, a second stage is incorporated where a liquid with higher pH (neutral or alkaline condition) is added to

remove the SO2. Studies [6,7,24–26] indicate that the second stage is capable of removing other acidic compounds that are present in the flue gas stream.

**Figure 2.** Block flowchart of incineration plant used for the study.

#### **3. Methodology**

The integrated system is divided into three subsystems that are simulated using Aspen Plus®. This section describes these three subsystem models simulated.

The first is the waste incineration plant, which is to determine the volume of emissions, including the parameters of the various constituents, that were generated after the combustion of the MSW. The second model is the wet scrubbing process, employed to determine the volume of water needed to clean acid gases (limited to only HCl and SO2 in this study) from the flue gas stream and subsequently the volume of flue gas condensate (wastewater) generated in the process. The third model is the MD technology which is used to treat the flue gas condensate produced during the acid gas cleaning.

The relations used in performing the thermal analysis of the MD system are also presented in this section.

#### *3.1. Models Used in the Study*

#### 3.1.1. Model of the Waste Incineration Plant

The waste incineration plant Aspen Plus® flowsheet [27] is shown in Figure 3. The simulated plant has a nominal incineration capacity of 240 metric tons of MSW per day, generating 30 MW of electrical power with an efficiency of approximately 31%. The waste incineration plant model has four stages: (1) drying of the MSW, (2) combustion of the MSW, (3) steam generation for only electricity generation, i.e., condensing mode and (4) PM separation.

In the waste incineration plant, the wet MSW (WET-MSW) is sent into a vessel (DRY-REAC), where hot air (HOTAIR) is mixed with WET-MSW. A calculator block is defined in Aspen Plus to control the drying process in another vessel (DRY-FLSH) and the by-products from this vessel are a dry MSW (DRY-MSW) and an exhaust vapour (EXHAUST), which is discharged into the atmosphere.

DRY-MSW is now ready to be combusted. As its composition can vary based on the source and regional factors (e.g., topography, seasons, food habits etc.), it has been defined as non-conventional in the model. Consequently, for successful simulation of combustion process, DRY-MSW first needs to be defined based on its content. Therefore, an extra vessel (DECOMP) is included in the flowsheet where DRY-MSW is broken down into its various elemental constituents (Q-DECOMPOST). Q-DECOMPOST is then sent into the combustion chamber (BURN), where sufficient air (ATM-AIR) is added to achieve a complete oxidation of the MSW. Energy is recovered from the flue gases (CPROD-H) from the combustion process in the heat exchanger (HRSG) for the generation of superheated steam (HPSTEAM), which turns a steam turbine (ST-TURB) for the generation of electrical power (WT-TURB). After the recovery of heat energy from CPROD-H, there is a drop in temperature in the flue gas (CPROD-C) before entering particulate matter (PM) separation devices. In the model depicted in Figure 3, all three types (cyclone, bag filter and the electrostatic precipitator (ESP)) of PM removal devices are incorporated. The flue gas stream after the PM separation (ESP-GAS) then goes into the wet scrubbing system for cleaning of acidic gases.

**Figure 3.** Waste Incineration Aspen Plus® model flowsheet [27].

3.1.2. Model of the Wet Scrubbing Process

An Aspen Plus® model (Figure 4) of a wet scrubbing process was developed in order to simulate the cleaning process of acid gases from produced flue gas in the previous model. The main output of this model is to determine the amount of flue gas condensate (wastewater) that would be generated. The base method used in Aspen Plus® is ELECNRTL.

**Figure 4.** Wet scrubbing process Aspen Plus® flowsheet [27].

As can be seen in Figure 4, the developed model has two stages. RadFrac (WTSCRUB1) from the Aspen block built-in library, which is the acidic scrubber, is selected where the

flue gas stream from the waste incinerator (GASFEED) and water (LIQFEED1) are the feed streams. The products then are wastewater (LIQPROD1), and the flue gas stream (GASPROD1). The second wet scrubber (WTSCRUB2), selected as the same as the first scrubber in Aspen Plus, is the alkaline (or neutral) scrubber with two feeds: the partially cleaned flue gas (GASPROD1) from WTSCRUB1 and a liquid feed (LIQFEED2). The latter is an alkaline solution, which in this study is considered to be calcium carbonate (CaCO3).

Table 1 lists the operating condition used in the modelling of the wet scrubbing process. It is assumed that there is no temperature and pressure drop in GASPROD1 from WTSCRUB1 to WTSCRUB2.


**Table 1.** Operating conditions in the wet scrubbing process.

#### 3.1.3. Model of the MD System

There are no available blocks in the Aspen Plus built-in library that can readily be used in the simulation of an MD unit. Hence, it was modelled using a customised USER Model2 in Aspen Plus. An excel file sheet built into Aspen Plus was modified with data obtained from the simulation of the wet scrubbing model. Table 2 is a list of the operating parameters used in the simulation of the MD system. This simulation work is based on the MD system presented by Imtisal-e-Noor et al. [28]. However, there are a few differences between that simulation work and this current work. The current model has a single air gap MD module relative to the dual-cascaded MD modules used in that work. There are also differences in parameters, such as the feed inlet temperature, coolant inlet temperature, density of the feed, and the composition of the flue gas condensate. The feed and coolant inlet pressures, however, remain the same.

**Table 2.** Operational parameters of MD system.


The Aspen Plus® MD model flowsheet is shown in Figure 5, and the base method used for this model is IDEAL. The flue gas condensate or wastewater (WWSCRUB) generated from the wet scrubbing process is collected into a tank (TNK) at a temperature of 56.7 ◦C. The wastewater stored in the tank (FD1) is then passed through a heat exchanger (HX) and heated up to a temperature of 85 ◦C using heat from the cooling of the flue gas stream (from 440 ◦C to 160 ◦C) before particulate matter separation. The heated flue gas condensate (FD2) then goes into the MD module (MD). The temperature of the flue gas stream drops because of the latent heat of vaporization which corresponds to the permeate flux passing through the membrane. The concentrate and permeate streams from the MD module are

referred to as RET and PERM, respectively. The treated water (PERM) is then collected into another tank for reuse.

**Figure 5.** MD system Aspen Plus® flowsheet [27].

During the simulation of the MD system, the coolant stream is considered to be water to be pumped from the surroundings (e.g., from a river) with an ambient temperature of between 25 to 27 ◦C (the ambient temperature of water in Ghana).

#### *3.2. Thermal Energy Analysis of MD Systems*

The permeate flux, or simply flux, of an MD (*Jp*) is considered the most relevant metric used in the assessment of any membrane technology. It is defined as the flow rate of permeate flowing through the membrane measured in kg/m2 s and can be expressed mathematically as [29] .

$$J\_p = \frac{m\_p}{A} \tag{1}$$

where . *mp* is the mass flow rate of the permeate measured in kg/s and *A* is the effective membrane area measured in m2.

The thermal efficiency (TE) or evaporative thermal efficiency of MD systems is defined as the ratio of the latent heat of vaporization to the total (latent and conduction) heat. The TE of MD systems is considered an effective tool in the measurement of desired thermal transport. It can be expressed mathematically as [29]

$$TE\left(\%\right) = \frac{\dot{m}\_p \,\Delta\!\!\!H\_{v,w}}{Q\_m} \times 100\tag{2}$$

where Δ*Hv*,*<sup>w</sup>* refers to the enthalpy of vaporization of the water in kJ/kg, and *Qm* is the total heat flux through the membrane in kW, which can be determined by using the relation in Equation (3). 

$$Q\_m = \dot{m}\_f \mathbb{C}\_p \left( T\_{f,in} - T\_{f,out} \right) \tag{3}$$

In this expression, . *mf* refers to the feed mass flow rate measured in kg/s, *Cp* refers to the feed water specific heat measured in kJ/kg ◦C, while *Tf* ,*in* and *Tf* ,*out* refer to the inlet and outlet feed water temperatures, respectively, measured in ◦C.

Another important parameter for evaluation of the thermal performance of an MD system is specific energy consumption (SEC). It is defined as the energy required to produce 1 m<sup>3</sup> of distillate water in MD systems and can be determined using the following relation [30]

$$SEC\left(\frac{\text{kWh}}{\text{m}^3}\right) = \left[\frac{Q\_m \rho}{f\_p \text{ } A}\right] / 3600\tag{4}$$

where *ρ* refers to the density of water measured in kg/m3.

Gained output ratio (GOR) is defined as the ratio of thermal energy that is required to produce distillate water in an MD system to the energy input to the system. GOR, a dimensionless parameter, can be expressed mathematically using the following equation [31]

$$GOR = \frac{\int\_{\mathcal{P}} A \bigtriangleup H\_{\upsilon, w}}{E\_{\text{in}}} \tag{5}$$

where *Ein* refers to the total power input to the system measured in kW.

#### **4. Results and Discussion**

In this section, results for both the wet scrubbing and MD processes (including a parametric analysis on the MD system) are presented and discussed.

#### *4.1. Results from the Wet Scrubbing Process*

Table 3 presents some key simulation results of the wet scrubbing process. The temperature of the flue gas condensate was 56.7 ◦C, at a pressure 1.01 bar. The temperature of the flue gas condensate was above its dew temperature. The temperature of the cleaned flue gas (after acid gas cleaning) which continues into the stack to be emitted into the atmosphere is 52.2 ◦C at a pressure of 1.01325 bar and this is also above its dew point. In all the process can achieve an overall cleaning efficiency of over 99% for SO2 and over 95% for HCl. There was also a little reduction in the other constituents of the flue gases exiting the wet scrubber at the end of the cleaning process. The total flow rate of liquid feed used in the wet scrubbing was 70,000 kg/h of water, 35,000 kg/h in the first scrubber and 35,000 kg/h in the second scrubber. It must be noted that in the second scrubber the liquid feed is a mixture of water and CaCO3. (The mole fraction ratio of water to CaCO3 is 0.9:0.1).

**Table 3.** Simulation results from the wet scrubbing process.


#### *4.2. MD Model Simulation Results*

This section presents the results obtained from the simulation of the optimized MD model used in the study. The thermal performance of the MD system is also presented and discussed. Table 4 list the results obtained after simulation of the MD model.

**Table 4.** Simulation results from the MD system.


#### 4.2.1. Thermal Efficiency

The thermal efficiency of the MD system was determined using Equation (2) when increasing the feed inlet temperature from 75 ◦C to 90 ◦C while maintaining the coolant inlet temperature at 26 ◦C. Values of the determined thermal efficiency against the corresponding feed inlet temperature are reported in Figure 6a. It can be observed in the figure that increasing the feed inlet temperature from 75 ◦C to 90 ◦C resulted in an increment of the TE from 50.7% to approximately 73%. This increment in TE can be attributed to the increase in permeation when the feed/concentrate inlet temperature increases. The thermal efficiency of the MD system was determined again using Equation (2), but this time increasing the coolant inlet temperature from 15 ◦C to 32 ◦C while maintaining the feed inlet temperature at 85 ◦C. Values of the determined thermal efficiency against the corresponding coolant inlet temperature are reported in Figure 6b. It can be observed that increasing the coolant inlet temperature from 15 ◦C to 32 ◦C decreases the TE from 75.4% to approximately 57%. This decrease in TE can be attributed to the low level of permeation due to the increase in the coolant inlet temperature. These results conform to the results obtained in research by Shahu et al. and Elmarghany et al. [32,33].

**Figure 6.** Effects on increasing feed/coolant inlet temperature versus TE: (**a**) Effects on increasing feed inlet temperature vs. resulting SEC. (**b**) Effects on increasing coolant inlet temperature vs. resulting TE.

#### 4.2.2. GOR

The GOR of the MD system was determined using Equation (3) by increasing the feed inlet temperature from 75 ◦C to 90 ◦C while maintaining the coolant temperature inlet temperature at 26 ◦C. The values of the determined GOR against the feed inlet temperature are reported in Figure 7a. It can be observed from that increasing the feed inlet temperature from 75 ◦C to 90 ◦C increases the GOR from 1.82 to 2.53. Increasing the feed inlet temperature increases permeation, which in turn increases the driving force of the permeate, as much less thermal energy is required to produce distillate water in the MD system. The GOR of the MD system was determined again using Equation (5), but this time increasing the coolant inlet temperature from 15 ◦C to 32 ◦C while maintaining the feed inlet temperature at 85 ◦C. Values of the GOR against the corresponding inlet temperature are reported in Figure 7b. It can be observed from the figure that increasing the coolant inlet temperature from 15 ◦C to 32 ◦C decreases the GOR from 2.72 to 2.06. This decrease can be attributed to the fact that increasing the coolant inlet temperature decreases the level of permeation and the driving force; therefore, higher amount of thermal energy is required to produce distillate water in the MD system. These results conform to the results obtained in research works carried out by Shahu et al. and Elmarghany et al. [32,33].

**Figure 7.** Effects on increasing feed/coolant inlet temperature versus GOR: (**a**) Effects on increasing feed inlet temperature vs. resulting GOR. (**b**) Effects on increasing coolant inlet temperature vs. resulting GOR.

#### 4.2.3. SEC

The SEC of the MD system was determined using Equation (3) by increasing the feed inlet temperature from 75 ◦C to 90 ◦C while maintaining the coolant temperature inlet temperature at 26 ◦C. The values of the determined SEC against the corresponding feed inlet temperature are reported in Figure 8a. It can be observed that increasing the feed inlet temperature from 75 ◦C to 90 ◦C decreases the SEC from 1237.4 kWh/m3 to approximately 860 kWh/m3. As mentioned earlier, increasing the feed inlet temperature increases permeation and thus increases the driving force of the permeate which in turn decreases the energy required to produce the distillate water, therefore decreasing the SEC. The SEC of the MD system was determined again using Equation (4), but this time increasing the coolant inlet temperature from 15 ◦C to 32 ◦C while maintaining the feed inlet temperature at 85 ◦C. The values of the determined SEC against the corresponding inlet temperature are reported in Figure 8b. It can be observed that increasing the coolant from 15 ◦C to 32 ◦C increases the SEC from 831.2 kWh/m3 to approximately 1074 kWh/m3. Increasing the coolant inlet temperature decreases the level of permeation which in turn decreases the driving force of the permeate, therefore more energy is needed to produce the distillate water. These results conform to the results obtained in research works carried out by Shahu et al. and Elmarghany et al. [32,33].

**Figure 8.** Effects on increasing Feed/Coolant Inlet temperature versus SEC: (**a**) Effects on increasing feed inlet temperature vs. resulting SEC. (**b**) Effects on increasing coolant inlet temperature vs. resulting SEC.

#### **5. Conclusions**

The results obtained from the simulation of the MD process indicates it is possible to achieve almost 100% separation of the various constituents of the flue gas condensate. Additionally, a thermal performance assessment of the MD process indicates that increasing the feed/concentrate inlet while minimizing that of the coolant inlet temperature is the optimum means of operating the MD. It is therefore imperative to state that the coolant inlet temperature of 25 ◦C is reasonable and the feed inlet temperature should be at 90 ◦C.

Results obtained earlier from the simulation of the wet FGD model indicate almost 100% cleaning of acid gases (HCl and SO2) from the flue gas stream, where a total volumetric flow rate of 20 m3/h was used in the process. The total flue gas condensate produced at the end of the process was approximately 19.44 m3/h.

In conclusion, it is worth noting that the energy recovered during the cooling of the flue gas stream prior to PM separation is adequate for the operation of the integrated MD system, and the wet FGD technology is an effective method of cleaning acid gases from flue gas streams (achieving separation efficiencies of over 95% for the HCl and over 99% for the SO2). In addition, the MD technology is an effective method that can be used in the separation of these acid gases in the flue gas condensate produced in the wet FGD technology.

It is recommended that a techno-economics analysis on both the wet FGD and MD technologies are performed on a typical waste incineration plant with energy recovery, with Ghana as the location for the operation of the plant.

**Author Contributions:** Conceptualization, N.Y. and A.M.; methodology, N.Y. and A.M.; software, N.Y., M.S, and I.-e.N.; validation, N.Y.; formal analysis, N.Y.; writing—original draft preparation, N.Y.; writing—review and editing, N.Y., M.S., I.-e.N., A.M., and A.S.; supervision, A.M., and A.S.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding. The APC was funded by a discount voucher received by A.M.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** Authors wish to thank University of Mines and Technology, (UMaT), Tarkwa, Ghana and Royal Institute of Technology (KTH), Stockholm, Sweden for their support.

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

#### **References**


### *Article* **Osmotic Membrane Distillation Crystallization of NaHCO3**

**Mar Garcia Alvarez 1,2,\*, Vida Sang Sefidi 1,2,\*, Marine Beguin 1, Alexandre Collet 1, Raul Bahamonde Soria 1,3 and Patricia Luis 1,2,\***


**Abstract:** A new crystallization process for sodium bicarbonate (NaHCO3) was studied, proposing the use of osmotic membrane distillation crystallization. Crystallization takes place due to the saturation of the feed solution after water evaporation on the feed side, permeating through the membrane pores to the osmotic side. The process operational parameters, i.e., feed and osmotic velocities, feed concentration, and temperature were studied to determine the optimal operating conditions. Regarding the feed and osmotic velocities, values of 0.038 and 0.0101 m/s, respectively, showed the highest transmembrane flux, i.e., 4.4 <sup>×</sup> <sup>10</sup>−<sup>8</sup> <sup>m</sup>3/m2·s. Moreover, study of the temperature variation illustrated that higher temperatures have a positive effect on the size and purity of the obtained crystals. The purity of the crystals obtained varied from 96.4 to 100% In addition, the flux changed from 2 <sup>×</sup> <sup>10</sup>−<sup>8</sup> to 7 <sup>×</sup> <sup>10</sup>−<sup>8</sup> m3/m2·s with an increase in temperature from 15 to 40 ◦C. However, due to heat exchange between the feed and the osmotic solutions, the energy loss in osmotic membrane distillation crystallization is higher at higher temperatures.

**Keywords:** NaHCO3; osmotic membrane distillation crystallization; membrane contactor

#### **1. Introduction**

Climate change is redirecting global objectives to regulate greenhouse gas emissions. Industry accounts for 21% of these emissions [1] Thus, in order to minimize waste production and its effect on the environment, the design of more efficient processes is required. Many industrial sectors are already focused on lower energy consumption, such as the pharmaceutical industry, food industry, fine chemicals industry, and construction. However, some of their processes are still far from sustainable [2,3]. That is the case for crystallization, a separation technique for producing or purifying solid products from a supersaturated solution. Crystals have high stability, are easy to store, and have a long life. For these reasons, there is an immense requirement for their production from industry [4].

On a larger scale, several principles are used to form crystals, such as cooling of the feed solution, evaporation of the solvent, and anti-solvent techniques. The conventional equipment for performing crystallization is a batch stirred tank, which has several drawbacks. Firstly, the conventional crystallizer cannot provide crystalline solid products of sufficient morphological quality (size, shape, and crystal size distribution), structure (polymorphism), and purity [5]. Secondly, there are some reproducibility issues such as imperfect mixing, where the solution is not homogeneous, and the supersaturation control is limited. Moreover, the points at which crystallization can be performed vary from one batch to another. Furthermore, a great deal of energy is needed either to heat/cool

**Citation:** Garcia Alvarez, M.; Sang Sefidi, V.; Beguin, M.; Collet, A.; Bahamonde Soria, R.; Luis, P. Osmotic Membrane Distillation Crystallization of NaHCO3. *Energies* **2022**, *15*, 2682. https://doi.org/ 10.3390/en15072682

Academic Editor: Alessandra Criscuoli

Received: 28 February 2022 Accepted: 28 March 2022 Published: 6 April 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 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 (https:// creativecommons.org/licenses/by/ 4.0/).

the solution in a conventional evaporator or to power vacuum systems, which are not efficient [6].

In addition, the stirred tank mostly operates as a batch reactor, meaning that the process is not continuous and has to be stopped to recover the products. It would be more convenient and energetically more efficient to use a continuous process [5,7–9]. As conventional crystallizers have many inconveniences, research has been conducted to find alternatives allowing better control and performance during the crystallization process, and membrane distillation crystallization is one of these alternatives [4].

Osmotic membrane distillation crystallization (OMDC) is an innovative technique in which two liquids are brought into contact through a non-selective hydrophobic microporous membrane [10]. Because the concentration is not the same on both sides, this induces a water activity difference and leads to the evaporation of water from the feed to the osmotic side. Thus, the driving force is the vapor pressure gradient created by the water activity difference between the two sides of the membrane. Figure 1 depicts the mass transfer profile for the OMDC system [7,8].

**Figure 1.** Concentration profiles in osmotic membrane distillation crystallization.

OMDC has advantages over conventional distillation and crystallization processes. This technique has a very high specific contact area, promoting higher mass transfer with more compact equipment than in conventional crystallization or distillation. The main advantage of OMDC is lower energy consumption [11,12]. As the driving force is created through the partial pressure gradient, no additional pressure is required, which allows equipment costs to be reduced and process safety to be increased in comparison with pressure-driven processes. Residual heat or renewable energy can also be used, if available, which could reduce the overall cost and environmental impact [12–14]. Another benefit of OMDC is the use of polymer materials in the equipment, which decreases or even avoids erosion problems [4–7].

OMDC is presented as an alternative option for crystallizing sodium bicarbonate (NaHCO3). NaHCO3 is a salt obtained from a reaction between soda ash and carbon dioxide (CO2) [15,16]. NaHCO3 is used in various industries such as food, pharmaceuticals, agriculture, etc. However, the purity and the morphology of the obtained crystals play an important role in where the NaHCO3 salts can be used. To crystallize NaHCO3 in a conventional crystallizer, CO2 must be introduced to the tank atmosphere as NaHCO3, which can easily be converted to CO2 by heat or stirring [15]. Shifeng Jiang also studied the crystallization of NaHCO3 using a cooling crystallizer to generate more NaHCO3 crystals [17]. However, when using OMDC technology to crystallize NaHCO3, there is no need for the constant addition of CO2. Moreover, as the solution is not heated, less NaHCO3 is converted to CO2, which is the main advantage of OMDC for crystallizing NaHCO3. To the best of our knowledge, no studies have been performed on the crystallization of NaHCO3 using membrane distillation crystallization. However, OMDC has been used for other materials. Israel Ruiz Salmon et al. studied OMDC for the crystallization of sodium carbonate. It was observed that in OMDC, the main resistance was the membrane itself, and the process suffered from concentration polarization and possible wetting [18].

In this study, the main objective was to optimize the OMDC system for the crystallization of NaHCO3. Several operational parameters such as the feed and osmotic velocities, the effect of feed concentration, and the feed temperature were studied. Moreover, the purity, shape, and size of the crystals were analyzed using X-ray diffraction (XRD) and scanning electron microscopy (SEM).

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

#### *2.1. Chemicals*

The feed solution for each experiment was produced by dissolving NaHCO3 salt (sodium bicarbonate, ≥99.7%, AnalaR NORMAPUR, Leuven, Belgium) in ultrapure water, and the osmotic solution was obtained by dissolving sodium chloride (NaCl) (sodium chloride, ≥99.9%, AnalaR NORMAPUR, Leuven, Belgium) up to the maximum solubility in ultrapure water.

#### *2.2. Equipment*

Figure 2 shows the scheme for the distillation/crystallization setup. The membrane contactor used to carry out experiments was a 3MTM Liqui-CelTM MM-1 × 5.5 Series Membrane Contactor. The characteristics of the membrane are given in Table 1. The feed and osmotic solution were in contact with a countercurrent flow. The feed solution flowed on the lumen side and the osmotic solution was on the shell side. The weight of the feed reservoir was measured constantly using a balance (LP 4202I, VWR, Milano, Italy), and is used in Equation (1) for calculating the transmembrane flux and in Equation (2) for the mass transfer coefficient calculation. The feed solution was always kept in a closed-cap container. The feed and osmotic solutions were kept at room temperature for most experiments, except for the temperature study, in which a cooler (Corio CD-900F, Julabo, Seelbach, Germany) and a water bath (VWB2 12L, VWR, Poole, UK) were used to change the temperature in the range of 15 to 40 ◦C. The temperature was measured using thermocouple thermometers (2000, TME, Birmingham, UK).

**Figure 2.** Schematic diagram of the membrane distillation crystallization setup: A feed solution; B gear pump; C membrane contactor; D peristaltic pump; E osmotic solution; F balance; G water bath/cooler; T1–T4 thermometers.


**Table 1.** Characteristics of the membrane contactor and hollow fibers.

Scanning electron microscopy (SEM) (GEMINI, Zeiss, Ultra 55) was used to observe the NaHCO3 crystals produced at different feed temperatures. The SEM images studied were taken at 500× magnification with a signal A = E2.

X-ray diffraction (XRD) (Bruker, AXS D8 ADVANCE) was used to determine whether the feed temperature altered the crystal purity. First, a metal sputter deposition system (CEA030, Balzers, Liechtenstein) was used to coat the surface with a thin gold layer to produce a conductive surface. Subsequently, the analyses were performed with a LYNXEYE detector, with a 2Theta from 20◦ to 100◦.

#### *2.3. Overall Mass Transfer Coefficient and Transmembrane Flux Calculation*

Two parameters allow characterization of the operating conditions of the membrane system, namely, the transmembrane flux (*J*, m3/m2·s) and the overall mass transfer coefficient (*Kov*, m3/m2·Pa.s). *<sup>J</sup>* was calculated by measuring the weight of the feed tank over time and recorded in intervals of 20 min. The flux shown in the figures is an average of the fluxes during the experiment, calculated by Equation (1) [12,19,20].

$$J = -\frac{1}{A\rho\_{wf}} \frac{dw\_f}{dt} = \frac{1}{\rho\_{wf}} \frac{w\_{f(t\_{i+1}) - w\_f(t\_i)}}{t\_{i+1} - t\_i} \tag{1}$$

For calculating *Kov*, the following relation is used [12,19,20]:

$$J = K\_{\upsilon\upsilon} \Lambda p \;=\; K\_{\upsilon\upsilon} \left(p\_f^\* \; a\_f - p\_o^\* \; a\_o\right) \tag{2}$$

In this equation, *p\** and *a* are the vapor pressure and the activity coefficient of the feed (*f*) and osmotic (*o*) sides, respectively, which were computed following the procedure described by Hamer et al. [21] and by Sandler [22] when the values of the osmotic coefficients were not found in the literature [23]. The vapor pressure (mmHg) was calculated using Antoine's equation, with the temperature T given in ◦C:

$$\phi = \frac{-\ln(a\_w)}{vM\_mM} \tag{3}$$

where *aw* is the sum of the ions of the electrolyte (−), *Mm* is the molar mass of water (kg/mol), *M* is the molality (mol/kg), and *aw* is the water activity.

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

#### *3.1. Influence of the Fluid Dynamics*

Improving the mass transfer is the key to having a lower required contact area and reducing capital costs. There are three resistances to mass transfer in the OMDC system: the feed boundary layer, the membrane, and the osmotic boundary layer. An increment in the velocity has a positive effect on reducing the lumen- and shell-side boundary resistances and increasing Kov. Figures 3 and 4 show the flux and Kov versus the change in the osmotic/feed velocities, respectively, while the velocity at the other side was set at a constant value. In addition to the boundary resistances, membrane crystallization is significantly affected by the phenomenon of concentration polarization (e.g., when the concentration of the salt is higher on the surface of the membrane), and therefore higher velocities in the membrane are more suitable, since they result in higher turbulence and thus better mixing of the solution in the membrane contactor. However, it can be observed in Figure 3a that overall, the feed velocity presents a maximum flux at 0.04 m/s when the osmotic solution operates at 0.01 m/s. At higher velocities of the feed solution, there is a decrease in flux. This decrease might be because of partial wetting of the membrane pores. As previously reported [17], wetting of the pores results in a lower flux and a higher resistance to mass transfer. In Figure 3b, with increasing feed velocity, Kov decreases slightly, reinforcing the idea of potential membrane wetting. The error bars for the feed flow rate of 0.01 m/s are around 13%. Regarding the osmotic velocity, Kov increases slightly when a higher velocity is used, which is an indication of more turbulence on the osmotic side and lower resistance to mass transfer. It can be observed in Figure 4a that in general, there is a rise in flux with an increase in the osmotic velocity. The maximum flux was observed at an osmotic velocity of 0.01 m/s. In Figure 4b, Kov increases when the feed flow rate is higher, to overcome the resistance in the osmotic boundary layer. However, there is a drop after 0.015 m/s due to possible membrane wetting. It can be concluded that the effect of the osmotic flow rate is higher than that of the feed flow rate, and it is more favorable to have a higher osmotic flow rate than a lower feed flow rate to avoid membrane wetting.

**Figure 4.** (**a**) Flux for changes in osmotic flow rate. (**b**) Kov for changes in osmotic flow rate. The concentration of NaHCO3 was at a maximum, and the feed and osmotic solutions were at room temperature.

Another set of experiments was performed to check whether there was total membrane wetting. This would be the case if NaCl was found in the feed solution. Ultrapure water was placed in the feed container, and the conductivity of the feed was measured over time with a conductivity meter. It was concluded that there was a high mass transfer of NaCl salts to the feed container at high velocity. For example, when the feed and osmotic velocities were 0.078 and 0.02 m/s, respectively, the conductivity of the ultrapure water changed from 13 μS/cm to 32 mS/cm within 2 h. This also confirms the hypothesis of partial membrane wetting at higher flow rates. Thus, velocities of 0.038 m/s (200 mL/min) for the feed side and 0.01 m/s (~200 mL/min) for the osmotic side were chosen as the optimal conditions, leading to a high Kov without significant membrane wetting. These velocities were set as constant values for the rest of the experiments described in the following sections.

#### *3.2. Influence of the Feed Concentration*

Figure 5 shows the flux and Kov versus the change in NaHCO3 concentration. The average flux decreases with an increase in concentration. This is due to a decrease in the driving force. In osmotic membrane crystallization, the driving force for water evaporation is the vapor pressure difference between the two sides of the membrane, which is influenced by the water activity. To promote flux, the driving force must be increased. This can be achieved either by increasing the osmotic concentration or by decreasing the feed concentration. An increase in the osmotic concentration implies a lower water activity, while a decrease in the feed concentration induces a lower water activity. Globally, this results in a higher driving force [7]. Therefore, we expect to see a drop in flux with an increase in feed concentration, as can be observed in Figure 5a. By calculating Kov using Equation (2), the effect of the driving force will be removed, and a constant Kov is expected with a change in concentration. However, it can be observed in Figure 5b that Kov still decreases with an increase in the concentration. The factor that causes Kov to decrease could be concentration polarization.

**Figure 5.** Effect of the concentration of NaHCO3 in the feed solution on: (**a**) flux and (**b**) Kov. Feed velocity was 0.038 m/s, osmotic velocity was 0.01 m/s, and osmotic and feed solutions were at room temperature.

#### *3.3. Influence of the Feed Temperature*

The temperature of the feed solution was varied in a range between 15 ◦C and 40 ◦C, while the osmotic temperature was kept at 20 ◦C, equivalent to room temperature (see the evolution of temperatures shown in Appendix A). This study was limited to 40◦ by the membrane contactor characteristics. To investigate the effect of feed temperature on the flux and the overall mass transfer coefficient, experiments were conducted with a feed concentration of 0.8 mol/L (67.2 g/L) and an osmotic concentration of 6.16 mol/L (360 g/L). The feed flow rate and the osmotic flow rate were at their optimal values for this system. The results are presented in Figure 6. When the feed temperature increases, the flux increases due to the vapor pressure difference created by the temperature difference and the concentration difference across the membrane at the same time. After some time of

operation, the feed solution and the osmotic solution reach the same temperature, since the two streams are recirculated in the experimental setup. The membrane contactor acts as an excellent heat exchanger between the feed and osmotic solutions, which unfortunately is not the desired effect, since most of the energy is lost in heating the osmotic solution rather than evaporating the water in the feed. Thus, it is more efficient if a membrane with a lower heat conductivity is used. Unlike the flux, the mass transfer coefficient decreases when the temperature increases, which could be explained by the presence of the temperature polarization effect. These phenomena were also observed by Salmon et al. [18] and by Boubakry et al. [24].

**Figure 6.** (**a**) Flux for changes in temperature, for 0,8 mol/L of NaHCO3. (**b**) Kov for changes in temperature, for 0,8 mol/L of NaHCO3.

The results of previous studies on membrane crystallization are presented in Table 2, for comparison with the results obtained in this study. It can be observed that the flux obtained in this work was in good agreement with the values obtained with the same type of hollow fibers but was inferior to the flux obtained in bigger membrane contactors. The fact that most of the studies did not report the mass transfer coefficient is a critical limitation, in terms of making a fair comparison. Sparenberg et al. [20] used the same type of membrane contactor for direct contact and vacuum membrane crystallization. The vacuum membrane crystallization had a higher flux in comparison to OMCD and DCMD, as the heat losses during the process were lower. The same applied to Kov: the values obtained in previous studies using the same type of hollow fibers were near the values obtained in this study, ranging from 4.8 × <sup>10</sup> − 11 m3/m2·Pa·s to 6.53 × <sup>10</sup> − 11 m3/m2·Pa·s.

**Table 2.** Comparison of performance of previous membrane crystallization processes with the process in this study.


#### *3.4. Crystalline Phases*

Commercial crystals were observed via SEM for comparison with the crystals produced at 15, 20, 30, 35, and 40 ◦C. The images produced via SEM are shown in Figure 7a–f. Commercial sodium bicarbonate is a powder consisting of flat sheet crystals with no preferential shape, while the crystals produced by membrane distillation crystallization

were in the form of squares, sticks, and other shapes. The shape of the crystals obtained in this study was similar to that of crystals obtained in the literature using other novel crystallization processes. Therefore, the effect of OMDC on the shape of the crystals was not significant [17,29,30]. The temperature influenced the morphology and the size of the crystals, giving bigger crystals at 35 ◦C. This is due to the fact that temperature has a great effect on the nucleation and growth rate of crystals [16]. The experiment at 35 ◦C gave larger crystals. This experiment was repeated four times, and it was observed that on two occasions the crystal size was similar to the size in the experiment at 40 ◦C, while on the other two occasions bigger crystals were obtained. For this reason, the average size should be taken carefully. This is due to the effect that temperature, or the residence time of the crystals in the tank, has on the nucleation and growth rate of crystals. The crystal size obtained in this study agreed with the crystal size obtained by Adnan Abdel-Rahaman et al. [31]. To find an optimal temperature, higher temperatures may have to be tested, but this was not possible in this study due to the thermal limitations of the material of the module.

**Figure 7.** SEM images of NaHCO3, comparing commercial crystals (**a**) with crystals obtained using osmotic membrane distillation crystallization at different temperatures: (**b**) 15 ◦C; (**c**) 20 ◦C; (**d**) 30 ◦C; (**e**) 35 ◦C; (**f**) 40 ◦C.

XRD analysis was performed on various bicarbonate crystals. The first was the original sodium bicarbonate powder from the industrial supplier. The second to the sixth samples analyzed were NaHCO3 crystals obtained after membrane crystallization distillation at 15, 20, 30, 35, and 40 ◦C for the feed solution. A comparison of the different XRD spectra is shown in Figure 8. As previously reported [32,33], all the spectra compared showed peaks at 29.7, 35.4, and 40.8 (2Theta), attributed to the NaHCO3 crystal phase.

**Figure 8.** NaHCO3 crystals—X-ray diffraction analysis.

The purity of the crystals obtained was in a range between 96.4 and 100%. While the crystals obtained at 30◦ had the lowest purity, showing a composition of 96.4% pure NaHCO3 and 3.6% hydrated Na2CO3, the highest purity was observed at 20, 35, and 40 ◦C, with crystals of 100% NaHCO3. The quantitative analysis of purity is included in Appendix B. As observed by Wang et al. [34], two factors influence the decomposition process of NaHCO3 to Na2CO3 + CO2 + H2O: temperature and water activity. The change in water activity is sensitive to both temperature and the composition of the liquid. The variation in the purity of the crystal can be explained by the membrane system's energy losses while heating the osmotic solution. This produces a slight variation along the membrane that can induce decomposition of NaHCO3 to Na2CO3.

#### **4. Conclusions**

Osmotic membrane distillation crystallization (OMDC) is a novel technology considered as an alternative to conventional crystallizers. OMDC has been studied for crystallization of sodium bicarbonate, due to its advantages such as lower energy and material consumption, control over the operational parameters, and larger evaporation surface area, among others. Several parameters such as the feed and osmotic velocities, feed concentration, and feed temperature were optimized. Regarding the feed and osmotic velocities, as the velocity increases the possibility of membrane wetting increases significantly. Therefore, a feed velocity of 0.078 m/s and an osmotic velocity of 0.01 m/s were chosen as the optimal conditions, which resulted in obtaining a Kov of 5.4 × <sup>10</sup> − 11 m3/Pa·m2·s. In addition, since the driving force in OMDC is the difference in concentration, an increase in feed concentration reduces the driving force and results in a reduction in the flux. However, when studying Kov and removing the driving force effect, the process was found to be affected by concentration polarization, and Kov still decreased by 23.6%. Finally, the effect of the temperature on water evaporation showed that the driving force of the system increased with temperature, as the flux increased from 2.45 × <sup>10</sup>−<sup>8</sup> to 7.49 × <sup>10</sup>−<sup>8</sup> <sup>m</sup>3/m2·s, but a great deal of energy was lost via the heat exchange between the feed and osmotic solutions. It was also observed that the size and the purity of the crystals were affected by the temperature, with larger sizes and higher purities obtained at higher temperatures.

#### **5. Patents**

The process presented here is registered under the patent application EP 2021163.

**Author Contributions:** Conceptualization, M.G.A. and V.S.S.; methodology, M.G.A. and V.S.S.; formal analysis, M.G.A. and V.S.S.; investigation, M.G.A., V.S.S., M.B. and A.C.; writing—original draft preparation, M.G.A. and V.S.S.; writing—review and editing M.G.A., V.S.S., R.B.S. and P.L.; visualization, P.L.; supervision, P.L.; funding acquisition P.L. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the European Research Council (ERC) as part of the European Union's Horizon 2020 research and innovation program (grant agreement ERC Starting Grant UE H2020 CO2LIFE 759630).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data available upon request to the authors.

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

**Appendix A**

**Figure A1.** Feed and osmotic temperature evolution with time during the thermal evaluation of membrane distillation crystallization: (**a**) 15 ◦C; (**b**) 20 ◦C; (**c**) 35 ◦C; (**d**) 40 ◦C.

**Figure A2.** XRD quantitative analysis for crystals produced at 15 ◦C.

**Figure A3.** XRD quantitative analysis for crystals produced at 20 ◦C.

**Figure A4.** XRD quantitative analysis for crystals produced at 30 ◦C.

**Figure A5.** XRD quantitative analysis for crystals produced at 35 ◦C.

**Figure A6.** XRD quantitative analysis for crystals produced at 40 ◦C.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Energies* Editorial Office E-mail: energies@mdpi.com www.mdpi.com/journal/energies

Academic Open Access Publishing

www.mdpi.com ISBN 978-3-0365-7769-2