Performance assessment of chemical mechanical planarization wastewater treatment in nano-electronics industries using membrane distillation

Wastewater from chemical mechanical planarization (CMP) processes in nano-electronics industries must be treated properly in order to fulfil local and international environmental regulations. This study is focused on a performance assessment of membrane distillation (MD) technology for CMP wastewater treatment. A new prototype of air gap membrane distillation (AGMD) module was utilized, with feed water consisting of CMP wastewater collected from imec, Belgium. The module was tested at different operating conditions (temperatures, flow rates and filtration time) and responses in terms of separation efficiency, permeate water quality, transmembrane flux, specific heat demand and exergy efficiency were determined. High quality permeate was produced in all trials, i.e. conductivity ~2.11 μS/cm, pH~5.4, TOC~1.13 ppm, IC~0.24 ppm, TDS~1.18 ppm and COD~1.9 ppm; for most of the contaminants the separation efficiency was> 99%. These findings clearly show that the resulting MD permeate does not exceed environmental regulations for release to recipient, and the permeate can even be considered for reuse. Moreover, the determined specific heat demand at different operating conditions was varying between 1390 and 2170 kWh/m whereas; the achievable exergy efficiency was ~19%.

With an ever-growing application of CMP technology in nanoelectronics industries, the amount of CMP wastewater has increased exponentially and is thus attaining considerable attention. Typically, the fresh water demand of a nano-electronics manufacturing plant ('fab') is approximately 1000 m 3 /day where 30-40% of total is accounted by CMP processes [12,13]. To observe EU Directive 91/271/ EEC concerning urban wastewater treatment [14], CMP-derived wastewater needs to be treated by removing nano-sized amorphous silica, alumina, ceria, and other chemical contaminants prior to discharge. Traditional chemical coagulation/flocculation treatment processes involve high dosage of chemicals (polyvalent electrolytes such as calcium, aluminum, and iron salts) to help solids content agglomerate and settle down for removal through subsequent filtration [15,16]. These processes lead to high operational costs due to high chemical demand and sludge disposal costs. Moreover, such operations are unable to ensure a high separation efficiency for the typical contaminant concentrations.
A number of alternate CMP wastewater treatment methods have been presented and analyzed by several researchers and include processes involving electro-chemical separation, membrane separation, and other methods. Within the first category, Belongia et al. [17] studied electro-decantation and electro-coagulation for removing and reusing alumina and silica from CMP waste. The findings illustrated that the coupled method has the potential to agglomerate and recover alumina and silica. Lia et al. [12,18] used an aluminium/iron electrode pair and observed 96.5% turbidity removal and 75-85% chemical oxygen demand (COD) reduction with an effluent COD of below 100 mg/l. Yang et al. [19] investigated electro-microfiltration considering pulsed mode, no electric field mode and continuous mode operations for treatment of CMP wastewater. The outcomes showed that the continuous mode operation displayed the optimized results i.e., high quality filtrate having turbidity as low as 0.39 NTU. Further, coupled electro-microfiltration and electro-dialysis has also been examined by these researchers [20]. Coupling the two methods yielded a permeate/filtrate suitable for high level recycling; obtained permeate exhibited turbidity < 1 NTU, TOC < 3 mg/l, and total dissolved solids (TDS) < 50 mg/l. Hu et al. [21] performed experiments with aluminium electrodes for electro-coagulation and flotation process. A turbidity reduction of 90% was observed while adding cationic surfactant cetyltrimethylammonium bromide (CTAB). Den et al. [22] presented the effect of hydraulic retention time and applied current density on turbidity reduction. Iron anodes and stainless steel cathodes were used in this study to reach removal efficiency of 95%. Liu et al. [23] studied electro-coagulation using iron electrodes for treating the CMP wastewater. The outcomes showed that the particles removal efficiency was~99% at a current density of 5.9 mA/cm 2 . Wang et al. [24] revealed that iron/aluminium electrode pair is relativity an efficient choice as compared to other typical electrode pairs for electro-coagulation in terms of energy demand. Finally, Chou et al. [25] investigated thermodynamic aspects of the electro-coagulation for oxide CMP wastewater treatment and demonstrated that the system operation was endothermic and spontaneous between 288 and 318 K.
Membrane based processes have also been investigated for CMP wastewater treatment. Brown et al. [26], Lin et al. [27] and Juang et al. [28] showed the performance of ultrafiltration (UF) and reverse osmosis (RO) in this regard. Brown et al. [26] demonstrated that metal and mixed oxide CMP wastewater can be treated and reused using ultrafiltration (UF). Lin et al. [27] considered chemical coagulation and reverse osmosis (RO) for CMP wastewater treatment for reuse. High quality permeate has been recovered after removing 99% of alumina and silica, and lowering the CMP wastewater COD < 100 mg/l. Juang et al. [28] investigated an arrangement of integrated UF and RO for CMP wastewater treatment for reuse. The results showed permeate having turbidity~0.01 NTU, conductivity~6 µS/cm and TOC~1.6 mg/L.
Other approaches for CMP wastewater treatment include use of magnetic seeds along with chemical coagulant to enhance aggregation and precipitation of alumina and silica. Wan et al. [29] indicated that turbidity of the CMP wastewater could be reduced from 1900-2500 NTU to 23 NTU with the action of 3.74 g L −1 magnetite (FeO*Fe 2 O 3 ) seeds using applied magnetic field of 1000 G. The coupling has significantly reduced the production of waste sludge as well. Kim et al. [30] tested the combined effect of magnetic separation and chemical coagulation on purification of CMP wastewater. The researchers employed magnetite (1.5 g/L) and ferric chloride (0.2 g/L) which displayed relatively better performance, reaching 0.94 NTU.
These processes have shown promise in certain applications, however there are significant challenges that prevent their widespread adaptation. Electrode-aided processes (electro-filtration/dialysis/coagulation) have the problem of reduced treatment efficiency due to electrode blockage. These processes are also cost-inefficient due to high electrical energy demand. Furthermore, microfiltration and ultrafiltration have the issues related to organic and inorganic fouling/scaling resulting in membrane blockage. Reverse osmosis is a pressure-driven separation technique and has a relatively high electrical energy demand. Additionally, high-pressure differences across the membrane require high mechanical strength of the membrane and induces biofouling (that needs to be treated using harsh chemicals). Moreover, treatment of large volume of CMP wastewater using magnetic seeding aggregation becomes unnecessarily expensive due to high cost of needed magnetic seeds. Thus, these practices are unreliable, energy inefficient, involve chemical treatments and are expensive.
Considering these limitations, membrane processes are judged to hold the most promise assuming that the following aspects can be addressed satisfactorily: reasonable pretreatment requirements; low fouling propensity; low chemical and electricity demands; and cost efficiency. Therefore, this study introduces membrane distillation as a promising method to treat CMP wastewater especially for removal of silica, alumina and copper. Membrane distillation (MD) is a thermally driven separation process utilizing a microporous hydrophobic membrane that only allows volatiles (i.e., water vapors etc.) to permeate through the membrane. The main driving force is a vapor pressure gradient across the membrane, which is developed due to temperature differences involved in the process [31][32][33]. The term MD originates from conventional distillation process modified with membrane separation [34]. The involved phase separation in MD process is based on the vapor-liquid equilibrium where latent heat of evaporation drives the change in phase from liquid to vapor [35]. The water transport through the membrane can be summarized in three steps: (1) formation of a vapor gap at the hot feed solution-membrane interface; (2) transport of the vapor phase through the microporous system; (3) condensation of the vapor at the cold side of the membrane-permeate solution interface [36]. As compared to other membrane technologies, MD operates at mild temperatures and atmospheric pressure. Moreover, it is relatively insensitive to pH and concentration fluctuations [34].
Furthermore, previous record of MD's successful applications for recovery of heavy metals [37,38], dehydration of organic compounds [39][40][41], concentration of acids [42][43][44], separation of pharmaceutical residues at very low to moderate concentrations [45,46], and wastewater treatment and water recovery [47][48][49][50][51][52] provides a strong argument to consider MD as a potential technology that can be successfully employed for treating wastewater streams contaminated with heavy metals, organic compounds, acids and nano-scale oxides i.e., CMP wastewater. We believe that this is the first work that shows the potential of MD technology for treatment of CMP wastewater. The present study is dedicated to a performance analysis of membrane distillation for CMP wastewater treatment. In this regard, separation efficiency of major contaminants is considered as the key performance factor whereas, conductivity, pH, TOC, COD and TDS are also determined in order to satisfy the water quality standards for reuse of the treated water in industrial utilities processes. Moreover, energy and exergy analyses have also been performed for a complete technical evaluation.

Experimental apparatus
For all the experiments, a prototype air gap membrane distillation (AGMD) module supplied by Xzero AB has been employed, as presented in Fig. 1 [53]. The design of the Xzero AGMD module is based on the HVR AGMD module [54] with certain modifications (essential for MD application in nano-electronics fabs). The comparison shows that both of the AGMD modules consist of single-cassette (placed between two condensation plates) configurations employing two microporous hydrophobic polytetrafluoroethylene (PTFE) membranes. However, the way used for attaching the membranes to the polyethylene (PE) frame was thermal welding in case of HVR AGMD module and beam clamping in case of Xzero AGMD module. (The latter was used in order to avoid leakages and to cope with higher pressures on the membrane.) Flow spreaders were added to the cassette for improved heat and mass transfer. The air gap between the membrane and condensation plates was also reduced. Furthermore, in order to solve corrosion issues and for providing an inert environment, the condensation plates were covered with polyvinylidene fluoride (PVDF) on the permeate side in order to ensure high permeate quality.
The CMP wastewater (feed) is heated with a Teflon-coated immersion heater (capacity 2 kW) mounted in a 30 L PVDF feed storage tank. The heated water is then circulated towards AGMD module using an Iwaki MD-30 PVDF pump (magnetic coupled shaft) and controlled by an FIP FlowX3 paddlewheel flow sensor. The hot water is fed into the top of the MD module and the exiting (concentrated) feed recirculates back to the storage tank. Fresh water is cooled using a R1134a chiller (capacity: 1.8 kW) integrated with a 80 L PP cold-water tank. Using an Iwaki PP-70 pump, the cold water is circulated through the cooling plates. The flow rate of the cold-water has been controlled and measured with similar type of flowmeter as mentioned earlier. Permeate is collected at the base of the MD module and is measured with a graduated cylinder and stopwatch. The temperatures of hot and cold streams are measured with temperature sensors (pT100). All sensors and alarms are controlled by a Crouzet logic unit. A handheld conductivity meter and temperature sensor are used for checking the permeate conditions. In this unit, Donaldson® PTFE membrane is used considering its attractive cost-performance comparison. The characteristics of the used membrane are mentioned in Table 1 and the process flow diagram of the Xzero membrane distillation purification system is shown in Fig. 2.

Experimental procedure
A total of 100 L of CMP wastewater was collected in five 20 L samples from imec, Belgium during a ten day period (some variation in composition was inevitable owing to the changes in upstream trials). Samples 1, 2 and 3 were used to determine the separation efficiency of contaminants using the aforementioned Xzero AGMD module. Considering that concentration does not affect the parametric study significantly, the other two samples were considered to determine permeate yield and energy requirement.

Separation efficiency
For CMP wastewater treatment tests, sample 1 (S1) was tested as MD feed without considering any pretreatment however, samples 2 and 3 (S2 and S3, respectively) were neutralized with 10 mL of 40% H 2 SO 4 per 20 L of CMP wastewater samples prior to introduction into the MD modules. The samples S1, S2 and S3 were used in Test 1, Test 2 and Test 3, respectively. The nominal operating conditions for CMP wastewater treatment tests were as follows: MD feed inlet flow rate 7.2 L/min; coldwater inlet flow rate 8.3 L/min; MD feed inlet temperatures 85°C for S1, 80°C for S2 and 75°C for S3; cold-water inlet temperatures 35°C for S1, 30°C for S2 and S3; and elapsed time of 3 h. Fluid flow conditions within the module indicate an average main flow channel velocity within a range of 0.025-0.055 m/s, assuming a U-type flow pattern from inlet to outlet. This condition corresponds to a Reynolds number range of 800-2500.
After achieving steady state, feed, retentate and distillate samples were taken in every 30 min, and the subsequent physico-chemical analysis included determination of cations (sodium, potassium and ammonium) concentration, anions (fluoride, chloride, nitrate, sulfate and phosphate) concentration, metals (aluminum, calcium, potassium, chromium, manganese, iron, nickel, cobalt, copper, zinc, gallium, strontium, phosphorous, germanium, tungsten, titanium, silicon, tantalum and zirconium) concentration, conductivity, pH, TOC, COD, IC (inorganic carbon), and TDS. For determining the mentioned water quality parameters, several analytical methods have been used. The pH was measured with an Orion Star Series meter with Orion Ross half-cell electrode and Ross reference electrode. The conductivity was determined using LF3000 with Pt-Cell (K = 0.1). The TOC (as well as IC and TC (total carbon)) was measured with Sievers 900 TOC-analyzer and COD was determined with Hach reagents LCI500. The anions and cations were measured with ion chromatography after adequate dilution on ThermoFischer ICS-5000 Capillary System. Metals were analyzed on ICP-OES from PerkinElmer.
The concentration factor of the contaminants in the MD feed over the elapsed time was also determined for each of the experiments. Concentration factor (CF i ) of the contaminants was defined by retentate to feed concentration ratio of contaminants as illustrated in Eq. (1): where C R,i depicts the retentate concentration of contaminant i and C f,i represents the concentration of contaminant i in MD feed water (wastewater). Moreover, feed volume reduction factor (VRF) was also calculated in order to determine the degree of concentration of the feed for the elapsed time using Eq. (2).
where V d defines the volume of permeate and V F shows volume of initial MD feed.

Energy analysis
Typically, flow rates and temperatures have been considered as the critical variables that influence the transmembrane flux and thermal energy demand of AGMD system. The considered levels of these parameters for energy analysis were as follows: feed flow rates (3.5, 4.6, 5.3, 6.7 and 7.2 L/min), feed temperatures (67, 70, 75, 80 and 85°C), coldwater flow rates (3.5, 6, 7.2 and 8.3 L/min) and cold-water temperatures (15, 20, 25, 30 and 35°C). Each experiment was performed for 30 min after approaching steady state conditions. The outlet temperatures of feed and cold-water, module surface temperature, permeate temperature and permeate flow rate were measured.
Transmembrane flux can be defined as collected distillate volume as function of experimentation time (V̇) d and active area (A m ) of the membrane and can be determined with Eq. (3).
Enthalpy changes of the feed (Q f ) and cold-water (Q c ) streams as shown in Eqs. (4) and (5) were used to determine thermal performance: where ṁand mḟ c are the mass flow rates of feed and cold-water streams, respectively; T f,in and T f,out are feed inlet and outlet temperatures while cold-water inlet and outlet temperatures are denoted by T c,in and T c,out . The heat capacity of water is termed as c p (4180 J/kg K). For the calculations, it was assumed that the permeate contribution could be neglected, i.e. ≫ ṁmḟ d and ≫ ṁmċ d , where ṁd is the permeate mass flow rate.
The specific thermal energy required (Q̇) f to operate MD system was determined in terms of heat demand per unit volume of the permeate (V̇) d . The specific heat demand can be calculated using Eq. (6).
Additionally, rate of heat transfer flow to the cold-water and heat transfer flow via convection and via permeate release were also determined using Eqs. (7)-(9).
Rate of heat transfer flow to cold-water (Q̇) c was calculated using Eq. (7).
Moreover, rate of heat transfer flow from the module surface to the surrounding can be determined in terms of free convective heat transfer rate (Q̇) cv that was calculated using Eq. (8).
where h defines the heat transfer coefficient for free convection (approximately 10 W/m 2 K for natural convective cooling by air) and A is the area of the module surface.T s defines the temperature at the surface of the module and ∞ T represents the atmospheric temperature at infinity point while taking module as the reference point.
The heat transfer rate of the distillate (Q̇) d was calculated with Eq. (9).
where T d is defined as measured temperature of distillate.

Exergy analysis
Apart from energetic analysis, exergy analysis has also been performed in order to measure the extent of ideality and reversibility of the process. Considering the basic exergy definition, the exergy of each stream in the system can be calculated using following relations in Eqs. (10)- (14). Since the key parameters which effect the performance of the MD system are mainly temperatures and composition, therefore the kinetic and potential exergies were not determined.
In the above expressions, Ex j represents total exergy of each stream , mj shows the mass flow rate of each stream involved in the process, T R and P R depict the reference temperature and pressure respectively, P is total pressure and C i defines the concentration of each contaminant/ solute. Moreover, m w,i and m w,solv show the molecular weight of each contaminant and solvent whereas n solv presents the solvent concentration.
Exergies associated with the component's inlet and outlet streams are referred as inlet and outlet exergies of the specific component, respectively. Moreover, exergy change across any component of the MD system can be calculated from the differences in inlet streams exergies (∑ Ex comp,in ) and outlet streams exergies (∑ Ex comp,out ) and shown as irreversible exergy of that component (Ex comp,ir ) in Eq. (15).
Furthermore, ratio of minimum exergy required to the total actual exergy input provides the exergy efficiency of the MD system (ƞ) which can be calculated using Eq. (16).
Additionally, irreversible exergies for each component and for the whole process were used to determine the contribution of each component of AGMD purification process towards total system irreversibility, mentioned as exergy destruction of the component (Ex comp,des ) in Eq. (17)

Results and discussion
The Xzero AGMD module performance was assessed mainly on the basis of the separation efficiency of the contaminants and permeate water quality. Transmembrane flux, thermal energy demand, energy distribution and exergy efficiency of the system are also presented in this section. Table 2 presents the concentration of MD feed samples and the resulting permeate samples after 3 h of operation. The analysis results show that in all the MD feed samples ammonium ions (NH 4 + ), potassium ions (K + ) and phosphate ions (PO 4 −3 ) were in high concentration, while except S1, the other two samples also have higher concentration of sulfate ions (SO 4

−2
). The reason is addition of 10 mL of 40% sulfuric acid as neutralization solvent in the pre-treatment process for S2 and S3. In CMP wastewaters, the key contaminants were silicon (Si), aluminum (Al) and copper (Cu) as expected. Other than these contaminants, phosphorus (P) was also in high concentration.
The MD permeate analysis from all the three runs with different composition of CMP wastewaters shows the metal concentrations under detection limit except calcium (Ca), which was also reasonably low.
The reduction of ammonium ions concentration in the permeate from S1 was not substantial (only~25%) due to the presence of highly volatile ammonia vapor in S1. However, the addition of sulfuric acid in feed samples (S2 and S3) played an important role in reducing the volatility of ammonia gas, resulting in < 0.05 ppm of ammonium ions in permeate. Thus, the pretreatment of feed samples shows three times better rejection performance in case of ammonium ions. For other contaminants, the outcomes show non-detectable concentration of sodium ions, potassium ions, nitrate ions, chloride ions and fluoride ions in the permeate along with very low levels of phosphate (0.03 ppm in S1). Moreover, in case of sulfate ions, the MD shows remarkable results in terms of separation efficiency i.e., < 0.1 ppm.
Since the permeate was released from the MD system, the volume reduction of the initial feed samples led to an increase in retenate concentration. The concentration levels of the contaminants in MD retentate samples are summarized in Fig. 3. The outcomes are for three tests of S1, S2 and S3, which were run for 3 h. The concentrations of ions and metals were increased in the initial feed over time, as expected. However, chloride ions and nickel shows a slightly different trend of concentration change compared to other contaminants. For both of them, the initial feed shows higher concentration as compared to the concentrated retentate. The probability could be that these contaminants might be adsorbed on the membrane surface. (Follow-on studies would be needed to study this effect in more detail.) It is clear from the outcomes presented in Table 3 that the D1 has high pH (~8) which corresponds to the presence of both NH 3 vapor (~10%) and NH 4 + ions (~90%) in the permeate [55,56]. Moreover, the conductivity reduction was only 37%, and TC and COD removal were 77% and 62%, respectively. While comparing the pre-neutralized CMP wastewater S2 with the resulting permeate D2, it is found that the reduction in CMP wastewater pH (~3) has clearly an impact on the reversible reaction of ammonia-water (Eq. (15)) which means the expected equilibrium shift of NH 3 towards NH 4 + ions might observed.
Since volatility of the NH 3 is highly dependent on pH, therefore the better MD performance was obtained (permeate pH~5.4 and conductivity 2.1 µS/cm). Furthermore, the TOC, TDS and COD were reduced up to 96%, 99.8% and 97.8% respectively. Considering sample S3, which was introduced in MD set up at relatively low temperature (75°C), permeate (D3) water quality is also quite satisfactory. For instance, the conductivity was decreased up to 98.8% and TC reduction was reached to 82%. Moreover, TDS and COD were reduced > 99.9%.
When comparing overall separation efficiency performance, MD shows very encouraging results for CMP wastewater treatment as compared to other available methods. Table 4 shows the comparison of MD with potential technologies including electro micro-filtration [19] and combination of electro-dialysis and RO [20]. For other related membrane based technologies i.e., integration of UF and RO was found to have comparable performance i.e., the reported conductivity was 5-6 µS/cm and TOC was 1.2-1.6 ppm [28]. Fig. 4 shows the effect of varying feed flow rates and cold-water flow rates on transmembrane flux. In these experiments, the feed and cold-water temperatures were considered constant i.e., 80°C and 25°C, respectively. However, the feed flow rates ranged between 3.5 L/min to 7.2 L/min while the cold-water flow rate was held constant at 8.3 L/min in the first set of experiments. In the second set, the cold-water flow rates were varied from 3.5 to 8.3 L/min while considering the constant feed flow rate of 7.2 L/min. The reported transmembrane fluxes were measured when MD system approached steady state i.e., approximately after 60 minutes of operation.

Transmembrane flux
The results obtained from first set of experiments show that with increasing feed flow rate, the transmembrane flux increases and ' < ' indicates a value below the respective detection limit.  presents the positive linear trend because the higher bulk temperature is maintained along the feed flow path and due to decrease in the boundary layer resistance. The increase in the transmembrane flux was observed from 9.7 to 11.7 L/m 2 h while almost doubling the feed flow rate (from 3.5 L/min to 7.2 L/min) in line with the results published by Baaklini [57]. Moreover, it was observed from the second set of experiments that the reduction in cold-water flow rate provides the lower transmembrane flux at the constant feed flow rates. The lower cold-water flow rate indicates the lower heat recovering capacity from the distillate water vapors. This leads to lower extent of condensation happing in the air gap of the MD system. Fig. 5 demonstrates the effect of the feed temperature on transmembrane flux while considering constant feed and cold-water flow rates (7.2 L/min and 8.3 L/min, respectively) and varying coldwater temperature (15-40°C). The outcomes indicate that the transmembrane flux adopts linear to weakly exponential increasing trend with increasing feed inlet temperature. The observed trends can be linked to lower viscosity of water at higher temperatures along with changes in the relationship between the vapor pressure and feed temperature, as defined by the Antoine equation illustrated in Eq. (16).
where A, B and C are the regression constants for the specific compounds, p v and T f represent vapor pressure and temperature of feed, respectively. For water, A = 23.238; B = 3841; C = 45.
The same feed to cold-water temperature difference (defined by different feed and cold-water temperatures) results in different values of vapor pressure difference which effects directly on the transmembrane flux. The transmembrane flux was maximum at feed temperature of 85°C and cold-water temperature of 15°C. These temperatures provide the highest extent of driving force as compared to other scenarios.
Since CMP wastewater samples were quite diluted, therefore, fouling phenomenon was not observed as expected during the elapsed time. However, follow-on studies would be needed to investigate this phenomenon in more detail.

Energy analysis
An energy analysis of the AGMD unit was performed by considering the specific thermal energy demand as the key parameter. Fig. 6 presents specific thermal energy demand values for varying feed and coldwater inlet temperatures while considering constant feed and coldwater flow rates i.e., 7.2 L/min and 8.3 L/min, respectively. Specific Table 3 Water quality parameters for the three MD feed water samples and corresponding distilled water and concentrated streams (elapsed time of 3 h). C and D represent concentrate/retentate and distillate/permeate, respectively.    heat demand values show a positive linear trend while increasing feed temperature and keeping cold-water temperature constant. The fact behind is higher heat transfer rate from the feed to permeate region across the membrane which results in higher temperature difference between feed inlet and outlet streams (T f,in -T f,out ) at elevated feed temperatures. While comparing different cold-water temperatures at constant feed temperature, a similar trend can be observed i.e., an increase in cold-water temperatures results in higher specific thermal energy demand. The reason is the lower transmembrane flux across the membrane at higher cold-water temperature associated with lower feed to coolant temperature difference. Thus, specific heat demand significantly increases with higher feed inlet temperature and higher coldwater inlet temperature i.e., 1390-2170 kWh/m 3 in agreement with the values presented by Baaklini [57] and Woldemariam & Martin [54]. In the present and referred studies, relatively higher specific heat demand was observed due to absence of heat recovery systems in laboratory based units [58]. The specific heat demand can be reduced in largescale equipment when internal heat recovery concept is introduced while adding stages in series [46,59], using thermal storage to recover surplus energy or considering concentrate recycle loop [60]. Furthermore, specific heat transfer flow rates to cold-water and via permeate and convection were also determined and presented in Table 5 for constant feed inlet temperature of 80°C and varying coldwater temperature between 15°C and 35°C. The calculations show that approximately 90% of the total specific thermal energy was transferred indirectly to the cold-water circulating through the cooling plates. The rest of the total specific thermal energy was accounted for energy stored in the distillate, lost due to convection and lost through the pipe walls, valves and joints.
Although higher driving force at lower cold-water temperatures (i.e., 15°C) was associated with higher transmembrane flux, however, the permeate temperature was lower as compared to when the coldwater temperature was higher i.e., 35°C. Therefore, with increasing cold-water temperature, specific heat transfer flow rate via permeate release is relatively higher. Moreover, the similar trend can be observed for convective heat transfer that indicates higher specific heat transfer flow rate from the module surfaces at elevated cold-water temperatures. The specific heat transfer flow rate to the cold-water also increases at higher cold-water temperature due to lower transmembrane flux.

Exergy efficiency
Along with energy analysis, exergy efficiency was also determined in this study. The considered operating conditions include MD feed inlet flow rate of 7.2 L/min, cold-water inlet flow rate of 8.3 L/min, MD feed inlet temperature of 80°C and cold-water inlet temperature of 30°C. Moreover, the chemical composition and concentration of sample S2 has been considered for calculating total exergy flow rates. The total exergy flow rates are shown in Table 6 for each component.
Furthermore, it was found that the exergy efficiency of the whole unit was 19%, which is comparable to the published results [61]. Each component in the unit is typically accountable for certain percentage of the total irreversibility produced. The results show that recirculation tank is responsible for~32% of total exergy destruction. Heat losses through the recirculation tank walls and evaporation through the tank cover openings are responsible for the exergy destruction in the hot recirculation tank. The cold-water tank share was~48% of total exergy destruction, which was comparatively higher since the cold-water tank was uncovered. MD module was accountable of~20% exergy destruction due to heat losses through condensation walls and the heat transfer through conduction, convection and permeate release. These results indicate the need of optimized MD unit in terms of its membrane material, insulation and condensation plates design. Moreover, the performance of recirculation tank and cooling water tank can be improved using proper insulation in order to reduce evaporative and conductive losses.

Concluding remarks
The study presents the potential of membrane distillation (MD) technology for treatment of chemical mechanical planarization wastewater from nano-electronics industries. Case study of imec, Belgium has been selected for the purpose and Xzero MD prototype was used for experimental studies. Considering the performance of MD unit in terms of treated water quality, different parameters have been reported including the compositional analysis, concentration, conductivity, pH, TOC, TDS and COD however, in terms of technical assessment of the methods transmembrane flux, specific heat demand, energy distribution and exergy efficiency were determined while varying different operating parameters (feed and cold-water flow rates and temperatures). The outcomes depict that high quality permeate was recovered having major contaminants (silicon, aluminum and copper) concentration below the detection limit, conductivity~2.11µS/cm, pH~5.4, TOC~1.13 ppm, IC~0.24 ppm, TDS~1.1 ppm and COD ppm~1.9 while considering neutralization prior to membrane distillation at MD feed flow rate of 7.2 L/min and temperature of 80 • C and cold-water flow rate of 8.3 L/min and temperature of 30°C. From the   parametric analysis, the maximum flux achieved was 14.8 L/m 2 h at the feed to cold-water temperature difference of 70°C. The specific heat demand was varied between 1390 and 2170 kWh/m 3 depending on the feed temperature and feed to cold-water temperature difference. Moreover, the estimated exergy efficiency of Xzero AGMD prototype was~19%.