3.3.1. Rejection of TrOCs via LPRO Operation
The removal values of TrOCs during LPRO operation are summarized in Table 7
(respective values for ACF processes are discussed in the next section). In most cases, permeate and retentate concentrations for the different pressure vessels were determined (see supplementary data, Table S5)
, and subsequently, the specific removal values of TrOCs for different vessels were calculated.
It was found that the rejection of group 1, 2 and 3 of TrOCs was not influenced by the variation in the raw water quality, as seen in the cases of BTA (62–65%), DMS (95–97%) and DTPA (87–92%); variations were within the measurement uncertainty (see Supplementary Data, Table S4
). In addition, higher apparent rejection rates for TrOCs were not measured when the flux was increased from 25 to 31 L/m²·h.
One of the major parameters influencing the retention of TrOCs via LPRO is the relation of the TrOCs molecular weight and MWCO of the RO membrane [10
]. For the employed LPRO membranes, MWCO is reported to be approximately 200 g/mol (Da) [30
]. Figure 4
presents the relation between molecular weight values of TrOCs and the measured rejection values. It was found that TrOCs exhibiting molecular weight ≥ 150 Da were completely retained, except DTPA (90%). Such superior rejection performance is mainly interpreted in terms of the size exclusion mechanism, which prevailed over any possible impact of surface charges and hydrophobicity [17
]. Concerning NMOR, MTBE, ETBE, caffeine, isoproturon, terbutaline, carbamazepine and sulfamethoxazole, the rejection rates are in good accordance with the work by Kegel et al., employing Trisep ACM5 membranes [31
]. Nevertheless, better rejection values were measured for NDMA (74%) and benzene (88%), compared to [31
], since the LPRO membrane employed here was tighter.
Furthermore, TrOCs exhibiting lower molecular weight values than RO membranes’ MWCO, i.e., DCM, DCE, TCM and TRI, were not properly retained. Samadi et al. measured rejection rates between 64% and 88% for TCM employing a Perma-Pure membrane (MWCO 300 Da) [32
]. Nevertheless, a direct correlation between the molecular weight of chlorinated compounds and the measured rejection rates was observed; the rejection rates for DCM < DCE < TCM < TRI increased with increasing molecular weight (cf. Figure 4
Concerning BTA and TCM, although they have different structures and physical properties, both substances showed comparable rejection rates of 63% and 67%, respectively. This proves the dominant influence of molecular weight. The rejection ratio for MBTA was 15% better than for BTA, which is related to the difference in molecular weight by 14 g/mol. This is in agreement with a study in which BTA was retained by 66% employing a UF/RO system [33
The polycarboxylic acid DTPA (393 Da) was retained by 87–89%, while EDTA (292 Da), a structurally comparable compound, was retained >7% more. Both substances are bivalent anions according to their pKa
values and can form complexes with calcium ions in the feed water. The different rejection ratios may be related to the complexation constants (EDTA/DTPA). The rejection rate for EDTA is in accordance with a study by Müller et al. in a pilot-scale treatment plant [34
Moreover, the unexpected very different retention rates measured for DCM and MTBE as well as DCE and ETBE, despite their small molecular weight difference, may imply the contribution of other parameters in the retention mechanism, i.e., size exclusion is not the only mechanism here. Klüpfel related the different rejection rates measured for TrOCs, having similar molecular weight values, to their LOG Kow
]; higher LOG Kow
values resulted in lower rejection rates. The data of the current study support, to a large extent, this hypothesis. Nevertheless, the low rejection ratios of BTA in relation to MTBE may not be explained on this basis. In addition, the rejection ratio for DMS was higher than MBTA (by 13%), although MBTA has a higher molecular weight than DMS that may be explained by the higher polarity of DMS as depicted by the LOG Kow
values. Likewise, the different rejection ratios for MBTA and metformin could be also correlated to their LOG Kow
values. An analogous concept might also apply to benzene, which was retained by ~82% [36
According to Xu et al. [18
], the rejection rates for TCM and TRI decreased dramatically (from > 80% to < 5%) after 24 h of operation when the dosing concentrations were 80 µg/L; this was attributed to strong solute–membrane interactions. Here, the first sampling was made after 5 days; however, the results did not reveal the same phenomenon. Instead, it was found that the rejection rates for di- and trichloro compounds, i.e., DCM, DCE, TCM and TRI at a 5 µg/L-level, were found to be improved upon increasing the operation time, e.g., in the case of DCM of 23%. This might be related to the formation of a combined fouling layer, comprising organic and inorganic substances, time or filtration progression, a phenomenon called cake-reduced concentration polarization [37
In addition, low rejection values were measured for NDMA (74 Da): 21% at 25 L/m²·h and 35% at 31 L/m²·h. Rejection ratios for NDMA in the literature, by ESPA2, LFC3, TFC-HR, 70LW and NF-90 membranes, were reported to be in the range of 37% to 52% [7
]. Takeuchi et al. reported variable removal of NDMA from 20% to 88% by the RO processes [39
]. Fujioka et al. studied and compared the rejection of NDMA and boron (as boric acid at pH 6–8) [27
]. These small and neutral molecules (74 and 62 Da, respectively) were found to be very hydrophilic [27
], and the size exclusion mechanism (based on free-volume hole-size) was considered to be the main retention mechanism [26
]. Therefore, comparable rejection ratios of NDMA and boron were found for different commercial RO and NF membranes [27
]. Similarly, the same behavior could be, in principle, confirmed in the current study; see Table 5
and Table 7
. Nevertheless, the influence of increasing the flux was relatively not the same; the rejection of NDMA was increased from 21% to 35%, while the rejection of boron was decreased from 30.6% to 27.7%.
3.3.2. Removal of TrOCs via ACF Process
The removal values of TrOCs during the ACF process are introduced in Table 7
, the data being sorted according to the increase in specific loading of ACF. In general, ACF showed significant removal performance for most TrOCs, including substances that were not properly retained by LPRO, e.g., DCM, DCE, BTA and TCM. On the other hand, ACF was not able to remove salts as expected, which is reflected in the comparable electric conductivities of raw water and processed/treated water. Nevertheless, in general, ACF operation is less energy demanding than LPRO; the pressure drop of ACF was below <1 bar.
Removal of TrOCs by ACF was emphasized to depend on the specific throughput (Table 7
). The highest removal was always measured at the beginning of the process for a filter filled with virgin GAC; then, it decreased during the operation in some cases. This effect was only obvious for DMS, metformin, PFBA, DCM, NDMA and EDTA because of the limited operation time in this study. Whereas excellent removal of all other TrOCs by ACF was found, removal of NMOR, MTBE, ETBE, caffeine, isoproturon, terbutaline, carbamazepine and sulfamethoxazole was in good accordance with Kegel et al. [31
], where Norit Row Supra 0.8 carbon was employed. Nevertheless, higher removal of NDMA was measured here. Metformin, EDTA and DMS are rather polar substances (i.e., LOG Kow
< −1), and consequently, the breakthrough occurred already at specific load of 8.1 m³/kg.
DCM having a LOG KOW of 1.96 was one of two TrOCs whose removal values varied by changing the GAC filling. By increasing the operation time of ACF, a reduction in the removal from 81% to 50% (2nd filling) and 70% to 93% (1st filling) was observed. As a consequence, the differences might be related to the differences between the two AC charges, as well as to the errors of sampling and analytics, which are documented to some extent by the double measurement of the first filling. For metformin, a non-systematic variation of removal from 63% to 97% with specific loading was found.
Moreover, it was found that the risk of breakthrough for TrOCs, especially those with high polarity, was increased for ACFs with high specific throughputs. In the case of riverbank filtrates, this may also occur due to accidental pollution of the corresponding river.
Using virgin GAC, decreasing concentration profiles (i.e., increasing removal trends) were observed for most of the TrOCs along the filter depth (cf. Figure 5
). It was also found that the mass transfer zone (MTZ) is sharp for PFBA and EDTA but wide for DMS and DCM; see Section 3.4.2
for further information. At a specific throughput of 14–18 m³/kg, the upper layer was saturated by PFBA and EDTA occurred. DMS, a low-molecular and polar compound with LOG Kow
of −1.16, was no more retained by ACF. In contrast, DCM, also a small and polar compound with a LOG Kow
of 1.5, was completely adsorbed, as shown in Figure 5