Evaluation of Membrane Integrity Monitoring Methods for Hollow Fiber Nanofiltration Membranes: Applicability in Gray Water Reclamation Systems

Source-separated gray water reclamation using nanofiltration as an advanced post-treatment option has received substantial interest in meeting the growing water demand. During reclamation, membrane integrity is crucial to ensure the water’s safety. This study evaluated several chemical and novel microbial indicators as indirect membrane integrity-monitoring methods for hollow fiber nanofiltration membranes in reclamation schemes. Under normal conditions, high retention of divalent ions and organic matter and near-complete removal of Escherichia coli (E. coli) were observed. Limited removal of the antibiotic gene (ARG) tetO was observed due to low feed concentrations and a higher detection limit (LOD). While 16S rRNA and ARG sul1 were not limited by their LODs, lower removals were observed, most likely due to free-floating DNA passing through the membranes. A broken fiber in a pilot-scale module reduced organic matter and microorganism removal substantially, while flux and ion rejection remained similar. Predictions made using the observed results and a previously proposed model allowed for the evaluation of the selected methods in upscaled reclamation systems. Based on these results, it was concluded that microorganisms could be employed as indicators in indirect membrane integrity-monitoring methods in large-scale reclamation schemes, while UV254nm absorbance (used in organic matter determination) could be a viable solution in pilot-scale systems.


S2. Predictive model
A simple model predicting the change in TOC concentration due to a compromised fiber was proposed previously (Lidén et al., 2016).Within this study's scope, this model's applicability in predicting retention changes for all indirect monitoring compounds was evaluated.While an extensive description of the hydraulic model is provided in Lidén et al. (2016), a summarised version is provided in the following paragraphs.
In the case of an intact membrane, flow through the membrane, Q p (L.h -1 ), can easily be determined when the water permeability, P (L.m -2 .h - .Bar -1 ), is known.
Where A mem represents the known membrane surface area (m 2 ), and ΔTMP and Δ∏ represent the pressure drop over the membrane and the osmotic pressure in Bar.Since membrane area and the water permeability are constants, the flow through the membrane is, in this case, only dependent on the nominal pressure applied to the membrane.During this process, the retention (R) of solutes can be determined using equation 2, where C p and C f are the permeate and feed concentrations, respectively.
When a fiber is breached, the flow to the permeate side and the permeate concentration will change.Normally, the flow through a membrane depends on the membrane's resistance and the pressure difference over the membrane surface.During a breach in a fiber, the membrane's resistance will disappear, and an open flow from the feed to the permeate, dependent on the pressure drop between the feed and permeate side, will occur.Since this pressure drop simulates the flow through a capillary channel, the flow through the broken fiber can be calculated using a modified version of the Hagen-Poiseuille flow through a laminar pipe formula (Equation 3).
In this, Q B is the flow through the broken fiber from one side, A fiber is the cross-sectional area of one fiber, ρ the water density, α a kinetic energy correction coefficient, f the friction factor, s the point of breakage downstream from the inlet, and d the inner diameter of the broken fiber.
α was determined based on the flow regime, in which laminar flow (Reynold <2000), α was determined to be 2, and in turbulent flow (Reynolds >4000), α was 1.In the transition zone, α was changed linearly from 2 to 1.The friction factor (f) was similarly dependent on the flow regime and determined using either equations 3, 4 or 5.

𝑓 = 64
< 2000 (4) Since both sides of the broken fiber will develop a pressure drop, flows coming from both the feed and the concentrate side need to be considered when determining the effect of a break on the flow.Therefore, Q B will need to be calculated for both the feed and concentrate side of the module and added to a total flow from a leak, Q L .An example of the dependency of the flow on the position of the break and the transmembrane pressure is provided in Figure S2.In regular operation, the flow through the membrane and the concentrations of the feed and permeate are known.When fiber breakage occurs, the permeate concentration will change following a simple mass balance.
Where C p,b is the mixed concentration of the permeate with the flow of a broken fiber.The number of broken fibers can be determined since Q L for one breach can be approximated, and C m can be determined by analysis.

S3. Chemical analysis -Ion chromatography
Cations were determined in isocratic mode using a Metrohm Metrosep C-4 column with a Methrohm Metrosep RP 2 Guard pre-column and a 3 mM nitric acid aqueous mobile phase.
Anions were separated by a Metrohm Metrosep A Supp 5 with a Methrohm Metrosep A Supp 4/5 pre-column.Anions were separated in isocratic mode using a mobile phase containing 1% acetone, 3.2 mM sodium carbonate, and 1 mM sodium bicarbonate.Both cation and anion concentrations were determined using a built-in conductivity detector.If required, samples were diluted to fit within the detection range (Table S1).The qPCR thermal cycling conditions for E. coli detection were as follows: 5 min at 94 °C(initial polymerase activation and denaturation), followed by 40 cycles of one min 94 °C, a minute at 55 0 C, and a minute at 72 0 C, afterward, an extension step of 5 min at 72 °C (Walker et al.,2019).
For 16S rRNA gene detection, the qPCR thermal cycling conditions were as follows 15 min at 95 °C (initial polymerase activation and denaturation), followed by 40 cycles of one min 95 °C, a minute at 55 °C, 30 seconds at 72 °C and final polymerization at 72 °C for 8 mins.For the detection of tetO, the qPCR thermal cycling conditions were as follows 15 min at 95 °C (initial polymerase activation and denaturation), followed by 40 cycles of one min 95 °C, a minute at 60 °C, 30 seconds at 72 °C and final polymerization at 72 °C for 8 mins.To ensure the quality of the qPCR process, the PCR cycle threshold (CT) was set to 28 and defined as the limit of detection.For a gene to be said to be detected in a sample, both technical duplicates had to be amplified.

S5. Pressure dependent retention under normal conditions
Furthermore, the water recovery on the lab-scale and pilot scale was determined to assess the effect of concentration build-up.The water recovery is based on the fraction of produced permeate (Q permeate ) and the supplied feed flow (Q feed ) and is determined by equation 10.

Figure S1 .
Figure S1.Schematic overview of the pilot-scale mexperience system.

Figure S3 .
Figure S3.Pressure-dependent retention of the selected indicators by a lab-scale dNF40 module under normal conditions.

Figure S4 .
Figure S4.Pressure-dependent retention of the selected indicators by a pilot-scale dNF40 module under normal conditions.

Figure S7 .
Figure S7.Retention of indicator contaminants by lab-scale dNF40 membranes (A mem =0.065 m 2 ) with and without a breach.Error bars represent the standard deviation.

Figure S8 .
Figure S8.LRVs the microbial inidcators determined by plate counting and qPCR using a lab-scale dNF40 membrane (A mem =0.065 m 2 ) with and without a breach.Error bars represent the standard deviation.

Table S1 .
Detection Limits Ion chromatography Fragments (integrated DNA Technologies) were used as control templates, with template genes separated by repetitive T sequences..The concentration of the standards was confirmed using fluorometry.

Table S2 .
Oligonucleotide primer sequences used in the current study