Accessing greater thickness and new morphology features in polyamide active layers of thin-film composite membranes by reducing restrictions in amine monomer supply

Abstract

Polyamide formation, via interfacial polymerization (IP) during thin-film composite (TFC) membrane fabrication, is regarded as self-limiting—in the sense that the polyamide film limits its own growth as it forms. During IP, trimesoyl chloride (TMC) and m-phenylenediamine (MPD) react rapidly to form an incipient polyamide film that densifies and slows the diffusion of the more permeable monomer (MPD), thereby limiting polyamide growth and yielding films that typically exhibit thicknesses <350 nm. The morphology of these polyamide films is characterized by a basal layer of void nodular and leaf-like features that is sometimes overlaid by a secondary layer of overlapping flat features. Here, we present evidence showing that polyamide active layers are substantially permeable to MPD, and that minimizing certain restrictions in the MPD supply conditions during IP can result in polyamide active layers of thicknesses several times greater (>1 μm) than those typically reported in the literature. In addition to the basal layer of void nodular features and secondary layer of overlapping flat features that characterize typical polyamide active layers, the thicker films also exhibited three additional morphological features: blanket-like layers atop the basal layer or other void features, multi-layer void structures, and/or void mega-nodules (up to over a micron in diameter). Overall, the results indicate that reducing restrictions in the MPD supply conditions during IP: (1) overcomes the limited polyamide growth observed in conventional TFC membrane fabrication and (2) leads to film morphologies with a more prominent void structure. This latter observation is consistent with recent literature describing the role of CO₂ degassing and nanobubble confinement in the development of polyamide active layer morphology. Future studies could vary MPD supply conditions as a new tool to expand the range of achievable thicknesses in active layer casting, regulate active layer morphology and optimize nanobubble confinement conditions independently of MPD supply. Such capabilities could aid in the development of novel supports and TFC structures.

Graphical Abstract

1. Introduction

Thin-film composite membranes (TFCs) are widely employed in drinking water production because of their ability to generate potable water from a wide variety of sources [1–4]. TFCs are composed of three distinct layers, but it is the top polyamide ‘active layer’ that is ultimately responsible for purifying water [5]. This ultrathin film rejects contaminants based on size exclusion, charge repulsion, and physicochemical interactions between the membrane, contaminant, and water [1,2].

The TFC active layer is fabricated directly upon a porous support via interfacial polymerization (IP) [5,6]. This reaction commences when the support layer surface, soaked in an aqueous amine solution, is brought into contact with acid chloride in an organic solvent. The amine and acid chloride react at the aqueous/organic interface, where the amine (typically m-phenylenediamine, MPD) diffuses into the organic phase to react with the acid chloride (typically trimesoyl chloride, TMC), [5,7,8] forming the polyamide active layer [5–11]. Polyamide film growth comprises multiple kinetically distinct stages. Initially, rapid formation of an incipient film occurs (tens of nanometers within seconds) followed by sharply slower growth (tens of nanometers in minutes), limited by the diffusion of the amine through the increasingly thicker and denser polyamide film [7,12–15]. A survey of available studies evaluating the effect of fabrication conditions (e.g., monomer concentration and IP time) on TFC characteristics revealed that typical active layer thicknesses are less than 350 nm [8,13,16–18].This is the case even with relatively high monomer concentrations and reaction times (e.g., up to 10 wt% MPD, 1.0 wt% TMC, and five minutes IP) [13]. As such, the membrane community regards polyamide film growth as self-limiting [7,12,14,15].

However, results from studies of unsupported polyamide IP (i.e., freestanding polyamide formed independent of a support layer) [19,20] provide evidence that IP conditions during TFC fabrication can be modified to minimize limitations in polyamide growth. Lee et al. [19] and Ukrainsky and Ramon [20] formed freestanding polyamide films orders of magnitude thicker than those commonly observed in TFCs. Both groups of researchers used monomer concentrations typical of conventional TFC fabrication (i.e., ~2.0 wt% MPD and ~0.1 wt% TMC). Lee et al. achieved thicknesses over 10 μm after one day of IP time using fixed-volume reservoirs of MPD and TMC, showing that an unsupported reaction with long IP times resulted in substantially greater polyamide growth than the conventional (supported) IP approach to TFC fabrication. Ukrainsky and Ramon also achieved polyamide thicknesses of tens of microns using a continuously flowing MPD solution to the reaction zone paired with a 60-s IP time, showing that–during unsupported IP–polyamide film growth limitations were overcome within practical IP times by continuously replenishing the MPD supply.

In contrast to unsupported IP, the conventional IP approach to TFC fabrication uses porous support layers that act as the MPD reservoir [15]. After the support layer is soaked with MPD solution, excess is removed (e.g., with squeegees, air knives, rollers, blot drying) [8,13,18,21,22] before the support layer surface is put in contact with the TMC solution. Therefore, during IP, two sources of limitations in the MPD supply are how much MPD is (1) absorbed initially by the porous support and (2) removed during the “excess removal” step. As such, the Lee et al. [19] and Ukrainsky and Ramon [20] studies provide evidence that polyamide growth in TFC fabrication could be promoted further by modifying the MPD supply conditions.

Therefore, the objective of this work was to experimentally assess whether polyamide growth during TFC membrane fabrication can be enhanced by modifying MPD supply conditions. We first evaluated if polyamide active layers of various TFCs were sufficiently permeable to MPD to allow for further polyamide growth within practical IP times (minutes). We then employed an alternative IP procedure that reduces restrictions in the MPD supply compared with conventional TFC fabrication to evaluate (1) the extent of additional polyamide formation atop existing TFC active layers and (2) whether thicker active layers formed when fabricating hand-cast TFCs. We characterized the morphologies of the resulting active layers and discuss the formation mechanisms behind specific polyamide active layer morphologies as they relate to MPD supply conditions and the role of carbon dioxide degassing and nanobubble confinement in film morphology [23–27]. Lastly, we assessed whether membrane modification with additional active layer formation affected membrane performance.

2. Materials and methods

2.1. Membranes

We used fully aromatic [28,29] TFCs that spanned a broad range of performance: seawater SWC4+ [30] and brackish water ESPA3 [31] (Hydranautics, Oceanside, CA); nanofiltration NF90 [32] (Dupont Water Solutions, Minneapolis, MN); and a hand-cast (HC) membrane. The HC membrane was fabricated in-house via conventional IP [8,21] between TMC (0.15 wt% in Isopar G) and MPD (3.5 wt% in laboratory-grade water, LGW) atop a porous polysulfone support (PS20, Nanostone Water Inc., Carlsbad, CA) using 1 min of polymerization time similarly as described elsewhere [22] (Section S2). These HC membranes and the original commercial membranes are referred to as “unmodified” membranes throughout this study.

2.2. Alternative interfacial polymerization

We used an alternative IP procedure to modify membranes by growing additional polyamide atop existing active layers of SWC4+, ESPA3, NF90, and HC TFCs. We also used this alternative IP procedure to form new polyamide atop polysulfone supports. The alternative IP procedure reduced restrictions in the MPD supply compared with conventional TFC fabrication. The yet unmodified membrane coupon (7.55 cm² effective area) divided a custom-built diffusion cell into two 16-mL compartments (Fig. S1): Compartments A and B. Membranes were oriented with support layers facing Compartment A, which contained an aqueous 2.0 or 3.5 wt% MPD solution to ensure a consistent MPD supply to the membrane support side. Two MPD delivery methods were evaluated: (1) a continuously replenished method where the MPD solution was continually pumped through Compartment A; and (2) a stirred fixed-volume method where Compartment A was filled with a fixed-volume of the MPD solution and stirred. Once Compartment A was filled with the MPD solution, 10 mL of a 0.15 wt% TMC/Isopar G solution were added to Compartment B—at which point IP commenced and progressed from 0.5 to 15 min. The resulting films were rinsed with 10 mL of n-hexane, dried vertically for 60 s, rinsed with LGW, and stored in LGW at 4.5°C. Two sets of control experiments were performed: (1) 15-min exposure to a 3.5 wt% MPD solution in Compartment A and Isopar G in Compartment B, and (2) 15-min exposure to LGW in Compartment A and a 0.15 wt% TMC/Isopar G solution in Compartment B. A larger custom-built diffusion cell was used to modify membrane coupons (35.6 cm² effective area) for testing in a cross-flow filtration system; a 3.5 wt% MPD solution, stirred fixed-volume (350 mL) MPD supply configuration, and 60 s polymerization time were used to fabricate these larger coupons.

2.3. Membrane imaging

Active layer morphology was assessed by transmission electron microscopy (TEM), scanning electron microscopy (SEM), and visual inspection captured by a digital camera (Section S4). ImageJ v1.53a [33] was used with TEM images to measure average total film thickness (inclusive of voids), average polymer thickness (excluding voids), and void fraction, similarly as described in our previous work [29] (Section S4).

2.4. Solute permeation in a diffusion cell

The diffusion cell described in Section 2.2 (7.55 cm² effective membrane area) was used to evaluate MPD permeation through membranes under conditions mimicking IP during TFC fabrication (i.e., no applied pressure and 3.5 wt% MPD solution), but with a consistent MPD supply. Membranes were oriented with support layers facing Compartment A, ensuring contact with a continuously replenished MPD solution. LGW (10 mL) was added to Compartment B at the start of each sampling interval, which ranged from 0.25–60 min. LGW samples were collected from Compartment B to assess MPD permeation. Control experiments using LGW instead of MPD solution were run before MPD permeation experiments to account for any MPD leaching from membranes.

To evaluate NaCl permeation, a 3.5 wt% NaCl solution was used instead of the MPD solution, and membrane samples were oriented with the active layer facing Compartment A. LGW samples were collected from Compartment B at 2 hrs to determine NaCl permeation.

MPD and NaCl concentrations in solution were determined by UV-Vis spectroscopy and conductivity measurements, respectively (Sections S5–S6). Solute mass permeation per unit membrane area (M_areal, g/cm²) was determined by multiplying solute concentrations in LGW samples (C_B, g/L) by the solution volume in Compartment B (V_B, L) and dividing by membrane area (a, cm²) as

$$M_{\mathit{\text{areal}}}=\frac{C_B{V}_B}{a}.$$ (1)

2.5. Solute passage and water permeance in cross-flow filtration

MPD and NaCl passage/rejection and water permeance were evaluated using a cross-flow filtration system operated in recycle mode after system performance stabilized (see Sections S7–S9 for system description and equations). MPD and NaCl concentrations in feed waters were approximately 200 mg/L and 500 mg/L, respectively.

3. Results and discussion

3.1. MPD permeation through unmodified membranes

We evaluated whether the polyamide active layers of various TFCs were permeable to MPD by testing MPD passage through unmodified membranes under cross-flow filtration conditions (Fig. 1a). We analyzed four membranes with a range of performance: SWC4+ is a ‘tight’ desalination membrane; NF90 is a ‘loose’ nanofiltration membrane; ESPA3 is a brackish-water membrane with performance between ‘tight’ and ‘loose’; and HC is a membrane with a similar performance to SWC4+. Consistent with the different performance levels of the membranes, their active layers had differences in morphologies and thickness (see Table 1, Fig. 2, and further discussions throughout the results and discussion). Membrane integrity was confirmed by evaluating NaCl passage. Results indicate that MPD could permeate the active layers of all membranes tested, with NF90 exhibiting the most MPD and NaCl passage (22.3% and 7.4%, respectively), and the HC membrane exhibiting the least (2.8% and 0.7%, respectively). Results correspond to the different levels of performance inherent to these membranes, with solute passage being higher in ‘looser’ membranes (i.e., NF90) and lower in ‘tighter’ membranes (i.e., HC and SWC4+ membranes). We cannot say how the different original morphologies (i.e., geometrical architecture) of the unmodified membranes affected their modification. However, we know that their original morphology affects the MPD permeation rate, which affects the rate of polyamide formation, rate of CO₂ degassing, and extent of nanobubble confinement. Therefore, we focused our discussion on comparing the initial MPD permeation rates of each membrane and relating those permeation rates to the rates of polyamide growths, final polyamide layer thickness, and evolution of polyamide layer morphology.

Fig. 1.

Solute permeation through SWC4+, ESPA3, NF90, and hand-cast (HC) membranes. (a) Percent passage of MPD (circles) and NaCl (triangles) and pure water permeance (A, L/m².hr.bar) from cross-flow filtration experiments. (b) MPD permeation per unit membrane area as a function of permeation time from diffusion cell experiments. Reported values correspond to the average of duplicate water permeance, MPD passage, and MPD concentration measurements, and triplicate NaCl passage measurements for each of duplicate membrane samples. Error bars represent the standard error of the mean.

Table 1.

Estimates of the average total film thickness, average polyamide thickness, average void fraction, and average additional film growth rate of membranes used, modified, or fabricated in this study.

Membrane	IDa	IP Time (min)	MPD Supply Conditionb	MPD Concentration (wt%)	Average Total Film Thickness (nm)	Average Polyamide Thickness (nm)	Average Void Fraction (−)	Average Additional Film Growth Ratec (nm/min)	Figure
HCd	1	0	A	3.5	110±14	100±11	0.09±0.01	N/Af	2
	2	0.5	B	3.5	155±9	135±5	0.13±0.02	90	2, S8
	3	15	B	3.5	1082±185	761±167	0.30±0.03	65	2, S8
	5	15	B	2.0	443±32	267±1	0.39±0.05	22	S8
SWC4+	6	0	A	N/Af	156±11	139±13	0.11±0.02	N/Af	2
	7	0.5	B	3.5	185±10	156±12	0.16±0.02	59	2, S8
	8	15	B	3.5	774±127	410±24	0.46±0.06	41	2, S8
	10	15	B	2.0	463±128	354±109	0.24±0.02	20	S8
ESPA3	11	0	A	N/Af	142±7	113±2	0.20±0.06	N/Af	2, S7
	12	0.5	B	3.5	309±4	244±11	0.21±0.02	335	S7
	13	15	B	3.5	351±27	257±16	0.27±0.01	14	2, S7
NF90	14	0	A	N/Af	239±16	187±10	0.21±0.01	N/Af	2, S7
	15	0.5	B	3.5	324±3	240±6	0.26±0.02	171	S7
	16	15	B	3.5	1206±218	727±146	0.40±0.01	65	2, S7
New TFCe	17	0.5	B	3.5	116±13	99±9	0.14±0.02	232	3
	18	1	B	3.5	192±44	150±25	0.21±0.05	192	3
	19	5	B	3.5	295±33	227±14	0.22±0.04	59	3
	20	15	B	3.5	1282±20	678±1	0.47±0.01	85	3, 5
	21	15	C	3.5	529±34	375±11	0.29±0.07	35	5
	22	15	B	2.0	610±112	494±138	0.20±0.08	41	5
	23	15	C	2.0	361±102	233±42	0.33±0.07	24	5
	24	15	A	3.5	282±25	227±21	0.19±0.00	19	5
	25	15	A	2.0	211±22	182±20	0.14±0.01	14	5

^aIDs correspond to the original IDs of the membrane samples in our experimental records.

^bA: Conventional, B: Continuously replenished, C: Stirred, fixed-volume.

^cRefers to the film thickness grown during the fabrication/modification procedure divided by the IP time (e.g., for ID#2, 45 nm/0.5 min = 90 nm/min).

^dHand-cast (HC) samples at IP = 0 min correspond to membranes prepared with the conventional TFC fabrication procedure using porous polysulfone supports, 3.5 wt% MPD, 0.15 wt% TMC, and 1 min polymerization time.

^eAll “new TFC” samples were fabricated with the alternative IP procedure using porous polysulfone supports, 3.5 wt% MPD, and 0.15 wt% TMC.

^fNot applicable.

Fig. 2.

TEM images (A-M) and schematic (N) of polyamide active layers before (A–D) and after (E-M) modification with an alternative IP procedure that reduces restrictions in the MDP supply compared with conventional TFC fabrication. Images in panels (E–M) are cross-sections of membranes modified with the least restricted MPD supply conditions employed in this study (not accounting for the rate of MPD permeation inherent to each membrane): continuously replenished MPD supply, 3.5 wt% MPD concentration, and 15 min IP time. Images in (A–H) are representative cross-sections of the samples, and images (I–M) provide additional examples of recurrent morphology features identified in the cross-sections of modified active layers. The schematic (N) depicts the five morphological features recurrent in modified membranes: 1) the basal layer present in unmodified membranes comprising a single layer of void nodular and leaf-like features that are densely packed (A–D); 2) a secondary layer of overlapping flat features covering a substantial fraction of the basal layer (C, G–I); 3) blanket-like layers of void-free polyamide covering a substantial fraction of the basal layer or other collection of void features (I–M); 4) multi-layer structures of densely packed void nodular and leaf-like features (G, H); and 5) void mega-nodules identified as single features with sizes (i.e., >350 nm) greater than traditional active layer thicknesses (E–J). Features 2 and 3 differ in that overlapping flat features exist in 2 (C, G–I), whereas a single void-free film can be discerned in 3 (J–M).

Given that MPD permeated all membranes in cross-flow filtration experiments, we proceeded to evaluate whether MPD could permeate polyamide active layers under conditions representative of IP in TFC membrane fabrication (e.g., no applied pressure) but with a continuously replenished MPD supply. Results from diffusion cell permeation experiments (Fig. 1b) show that MPD permeated all membranes and that permeation rates remained constant for 60 minutes. MPD permeation at 60 min ranged from 0.11 mg/cm² (SWC4+) to 0.71 mg/cm² (HC). MPD permeation trends differed between diffusion cell and cross-flow filtration experiments. For example, the HC membrane had the lowest MPD passage in cross-flow filtration but the highest MPD permeation in diffusion cell experiments. We speculate that these differences arose from a combination of variations in support compressibility and internal concentration polarization among membranes (Section S10).

Overall, the results in Fig. 1 show that MPD can permeate the active layer of various types of TFCs, both under filtration conditions and conditions that mimic IP in TFC membrane fabrication. These findings are significant to note—they show that with a consistent MPD supply, MPD permeates pre-existing polyamide layers at substantial rates, indicating that the limited polyamide growth observed in conventional TFC fabrication may result from limited MPD supply. We calculated the thickness of additional polyamide that would form atop the membranes depicted in Fig. 1b if the permeated MPD mass reacted with TMC (Section S11). Our results show that additional polyamide thickness would be similar to or greater than original active layer thicknesses. Therefore, we experimentally evaluated whether an alternative IP procedure that used a continuously replenished MPD supply would result in additional polyamide formation atop existing membrane active layers.

3.2. Additional polyamide formation via alternative interfacial polymerization

Polyamide film thickness.

Fig. 2 contains representative cross-sectional TEM images of unmodified active layers and active layers modified with the alternative IP procedure using the least restricted MPD supply conditions employed in this study (not accounting for the rate of MPD permeation inherent to each membrane): continuously replenished MPD supply, 3.5 wt% MPD concentration, and 15 min IP time. The low magnification (30,000x) in panels (A)-(H) enables visualization of relatively large active layer segments (~7–8 μm). We provide additional TEM images in panels (I)-(M) to facilitate discussion of morphological features of modified active layers (Section 3.3). The TEM images show that although some modified membranes had greater additional polyamide formation (HC and NF90) than others (SWC4+ and ESPA3), they all had total film thicknesses of approximately 1 μm and greater, either consistently across the active layer segment or interspersed with regions of lesser active layer growth. Therefore, the TEM images demonstrate that substantial additional polyamide growth atop pre-existing active layers is possible with a continuously replenished MPD supply.

To enable comparisons of active layer thickness and morphology with IP time, we modified membrane samples using 30-s IP under otherwise the same conditions as in Fig. 2 (i.e., continuously replenished MPD supply, 3.5 wt% MPD concentration). Fig. 3 contains representative digital photos, TEM images, and SEM images of the 30-s and 15-min modified SWC4+ and HC membranes, along with their unmodified counterparts. Additional membrane images, including for NF90 and ESPA3, are in Section S12. Upon visual inspection of membrane surfaces in the digital photos, modified membranes appeared coarse and uneven at the millimeter scale, contrasting the smooth surfaces observed in unmodified and control membranes (Figs. 3, S6). We attribute the increased roughness in modified membranes to additional polyamide formation. In the digital photos, it also appears that additional polyamide formation increased with IP time, as mound formations are more pronounced with longer IP times (Figs. 3, S7). Consistent with the digital photos, cross-sectional TEM images (Figs. 2–3, S7) confirmed additional polyamide formation in 30-s and 15-min modified membranes, with the 15-min modified membranes having the greatest polyamide formation.

Fig. 3.

Conventional photos, cross-sectional TEM images, and SEM images of unmodified (left column) and modified SWC4+ and hand-cast (HC) membrane samples. Modified samples underwent 30 s (middle column) and 15 min (right column) of additional polyamide growth.

Estimates of average total film thickness (i.e., including voids) and average polymer thickness (i.e., excluding voids) from TEM image analysis (Table 1) provided further evidence that additional polyamide growth increased with IP time. Modified membranes had substantially thicker active layers than their unmodified counterparts. Specifically, estimated average total film (polymer) thicknesses were 110–239 nm (100–187 nm) for unmodified membranes, 155–324 nm (135–244 nm) for 30-s modified membranes, and 351–1,206 nm (257–761 nm) for 15-min modified membranes. Over the first 30 s of IP, these thickness values correspond to a total film (polymer) thickness increase of 19–118 % (12–117 %) and an average rate of total film (polymer) growth of 59–335 nm/min (34–263 nm/min). By comparison, Matthews et al. [13] reported an active layer growth rate of approximately <5 nm/min after the first minute of IP time for TFCs fabricated using the conventional IP procedure and monomer concentrations similar to or greater than those used in this study. Our results therefore indicate that substantial additional polyamide growth atop pre-existing active layers is possible within practical IP times when the MPD supply is continuously replenished. As expected, the film growth rate was slower at greater (15-min) IP time but was still substantial, with 14–65 nm/min (10–44 nm/min) of average total film (polymer) growth, corresponding to a 148–886% (128–662%) increase in total film (polymer) thickness.

Comparing 30-s IP time results across membranes, active layer growth was least pronounced for the SWC4+ membrane, which had the lowest MPD permeation rate (i.e., the most restricted MPD supply accounting for the rate of MPD permeation inherent to each membrane, Fig. 1b). The order of active layer growth for the other membranes did not follow that of their MPD permeation rates but was in all cases substantially greater than for the SWC4+ membrane, consistent with their substantially greater MPD permeation rates (i.e., less restricted MPD supply accounting for the rate of MPD permeation inherent to each membrane). We hypothesize that the lack of correlation between MPD permeation rate (Fig. 1b) and film growth rate over 15 min of IP (Table 1) may be due to changes in the MPD permeation rate as additional polyamide formed during membrane modification. For example, the growth rate of additional film over 15-min of IP time for the ESPA3 membrane was substantially slower than for the other membranes. For the HC, NF90, and SWC4+ membranes, the 15-min film growth rate was 34–63% of the 30-s growth rate compared with 4% for ESPA3. This result indicates that there was a sharp decline in MPD permeance (i.e., a sharp increase in MPD supply restriction) in the first 30 s of IP time for the ESPA3 membrane.

3.3. Overall polyamide film morphology

The mounds observed on the surface of modified membranes remained on the membranes after (i) rinsing the surfaces with hexane and water and (ii) cross-flow filtration tests, indicating that the mounds were an integral part of the film structure. The mounds are notable because they may have formed from preferential MPD permeation locations on the active layer surface. On a micrometer scale, SEM and TEM images (Figs. 3, S7) revealed that surface roughness was the least pronounced for 30-s modified membranes, suggesting that the newly formed polyamide may have preferentially formed in the valleys of the pre-existing ridge-and-valley structure. Previous studies have postulated locations of increased permeation corresponding to the valleys in the polyamide surface, pores of the support, honeycomb-like pores on the backside of the polyamide film, and spots corresponding to the co-location of polyamide valleys and support pores [24,34–36].

TEM images (Figs. 2, 3, S7) also show that unmodified and modified polyamide films had numerous voids in them. The presence of voids in polyamide active layers has been reported and studied in detail in recent years [18,29,36–41]. Recent studies [23–27] have demonstrated that the void structure of polyamide films in TFCs results from interfacial degassing and confinement of CO₂ nanobubbles during IP. The same studies have shown that the void structure significantly determines surface roughness. Comparing our TEM images reveals that the void structure and surface roughness was more prominent in modified films than in unmodified films. Void fractions generally increased with IP time and were up to 35 percentage points greater in the 15-min modified membranes than in unmodified membranes (Table 1). This observation is consistent with recent studies [24,26] reporting that the support plays an important role in determining polyamide film morphology, with smaller support pore sizes leading to increased film heterogeneity due to greater nanobubble confinement. During TFC modification, IP occurs with the pre-existing, non-porous polyamide film (not the porous support) as the layer confining CO₂ nanobubbles. Therefore, the greater void content and roughness of the modified films compared with the unmodified films is consistent with the mechanism of CO₂ nanobubble confinement as a determinant of polyamide film morphology.

3.4. Recurring morphological features in modified active layers

We first focus on unmodified active layers (Fig. 2A–D) and active layers modified with the least restricted MPD supply conditions employed in this study (Fig. 2E–M). Unmodified active layers (Fig. 2A–D) had cross-sectional morphologies consistent with the void structure of polyamide active layers reported recently [18,29,36–41]. Specifically, all unmodified active layers had a basal layer consisting of a single-layer structure in which void nodular and leaf-like features are densely packed side-by-side (Feature 1).

There were differences in morphologies among unmodified membranes. For example, they covered a range of void fractions (0.09–0.21) and thicknesses (110–239 nm). Further, one of the unmodified active layers (NF90) also had a secondary layer consisting of overlapping flat features (Feature 2) covering a substantial fraction of the basal layer. This secondary layer had maximum thicknesses between approximately 20–100 nm accounting for no more than half of the active layer thickness at any location in the active layer segment imaged. Feature 2 in the NF90 membrane became more apparent in the 15-min modified samples and was also present in the 15-min modified HC membrane atop the basal layer. Song et al. [36] reported a similar secondary structure covering the active layer of unmodified commercial NF90 and XLE membranes. Song et al. posited that the flat features comprising this secondary layer are extensions of the void nodules in the basal layer and form under conditions where rigorous degassing leads to a forceful escape of the gas bubbles from the forming nodules. The forceful escape of gas results in an elongated shape of the reaction frontier instead of the balloon-like shape that gives origin to the void nodules. Such explanation only applies if a fast CO₂ degassing rate is accompanied by extensive nanobubble confinement, a confluence of conditions that likely occurred during HC and NF90 modification (see Feature 2 in Fig. 2G–I). First, all samples would have had extensive nanobubble confinement during membrane modification because of the pre-existing polyamide active layer (although the relative extents of the confinement conditions are unknown). Further, the HC and NF90 membranes also had high MPD permeation rates (Fig. 1), which translated into fast average film growth rates over 15-min of IP (44 and 36 nm/min, respectively) that should correspond with fast reaction and degassing rates. By contrast, Feature 2 was not predominant in modified SWC4+ and ESPA3 membranes, consistent with the slower average film growth rates over 15-min IP (18 and 10 nm/min, respectively) of these two membranes compared with those of HC and NF90.

Instead of overlapping flat features (Feature 2) covering the basal layer (Feature 1), 15-min modified SWC4+ and ESPA3 active layers predominantly had a blanket-like layer of void-free polyamide (Feature 3) atop the basal layer (Fig. 2L, M). This void-free polyamide layer was approximately 20–50 nm thick and appeared to be relatively separated from the basal layer. Though less predominant than Feature 2, Feature 3 was also present in NF90 and HC membranes at certain locations atop the basal layer and at the surface of the modified active layers (Fig. 2J–K). Xu et al [18], Ma et al [25], and Song et al. [26] reported fabricating hand-cast active layers consisting only of void-free polyamide when using conventional IP with low monomer concentrations (0.2 wt% MPD and 0.01 wt% TMC), a reduced carbonate concentration in monomer solutions, and unsupported IP, respectively. For these three void-free polyamide films, CO₂ nanobubble confinement was minimized directly (i.e., using unsupported IP) or indirectly (i.e., inhibiting CO₂ degassing). Although the initial MPD permeation rate of ESPA3 was high, the MPD permeation rate likely slowed after 30-s of IP, as evidenced by the drastically different film growth rates over 30 s and 15 min of IP (335 and 14 nm/min, respectively). Therefore, we attribute the predominance of Feature 3 atop modified SWC4+ and ESPA3 membranes to their slower film growth rates (compared with those for HC and NF90 membranes), which should correspond to a slower CO₂ nanobubble degassing rate and fewer confined nanobubbles.

Fig. 2 shows that the morphology of the 15-min modified HC membrane (Fig. 2H) was dominated by a multi-layer structure of densely packed void nodular and leaf-like features (Feature 4; similar to Feature 1 but multi-layered). Feature 4 was also observed in the modified NF90 active layers (Fig. 2G) but less predominantly than in the modified HC active layers. We did not observe Feature 4 in modified ESPA3 or SWC4+ membranes. Hand-cast supported polyamide active layers with a multi-layer morphology similar to Feature 4 were reported by Xu et al. [18] when using conventional IP with monomer concentrations of 20 wt% MPD and 1.0 wt% TMC. These concentrations are approximately 5–10 times greater than typical in conventional IP, resulting in much less restricted monomer supply conditions than conventional IP. Similarly, in our study, the HC membrane had the least restricted MPD supply overall during modification: all membrane types had the same external MPD supply conditions but the control HC (ID #1) membrane had the highest MPD permeation rate (Fig. 1b), as evidenced by the HC membrane exhibiting the most polymer growth over 15 min (Table 1). Accordingly, we posit that maximizing MPD supply (without taking steps to minimize CO₂ nanobubble confinement) results in active layer morphologies dominated by Feature 4.

The last recurring morphological feature in modified active layers is void ‘mega’-nodules that can be identified as single features with sizes greater than typical active layer thicknesses (i.e., thicker than ~350 nm), some of them spanning a few microns (Feature 5, see Fig. 2E–J). Feature 5 was predominant in modified SWC4+ and ESPA3 active layers, in which–as discussed above–Feature 4 was not observed. Feature 5 was less predominant in modified NF90 and HC active layers than in SWC4+ and ESPA3. It appears that these void mega-nodules occurred mainly as the most exterior feature in modified active layers, above Features 2–4. Based on these observations, Feature 5 forms when relatively slow CO₂ degassing rates (i.e., SWC4+ and ESPA3 had the least polymer growth, Table 1) occur with relatively extensive nanobubble confinement (i.e., the already thickened polyamide film underlies Feature 5) and consistent MPD supply (i.e., as per our experimental setup). We speculate forming void nodules continued to grow into void mega-nodules with the right balance between reduced degassing rate and nanobubble confinement—the opposite of the rigorous degassing that results in a forceful escape of gas from the forming nodules and corresponding formation of flat features (Feature 2) [36].

Fig. 2N summarizes our observations regarding the thickness and recurring morphological features in modified active layers. At relatively high MPD permeation rates and fast film growth rates (HC membrane as a prime example), a secondary layer of overlapping flat features (Feature 2) and a multi-layer structure of densely packed void nodular and leaf-like features (Feature 4) form. At relatively low MPD permeation rates and slow film growth rates (SWC4+ as a prime example), a blanket-like layer of void-free polyamide (Feature 3) covering a substantial fraction of the basal layer forms. The formation of the blanket-like layer may be followed by the formation of void mega-nodules (Feature 5) that can be identified as single features with sizes greater than typical active layer thicknesses (i.e., thicker than ~350 nm). At intermediate MPD permeation rates and film growth rates (NF90 as a prime example), mixed morphologies are obtained where Features 2–5 may all be observed.

3.5. Polyamide active layer formation via alternative interfacial polymerization

The alternative IP procedure used to modify TFCs was also used to fabricate new TFCs on porous polysulfone supports (Fig. 4) to compare their thickness and morphology with those of the modified membranes. Like the modified TFCs, thickness increased with IP time and, at long reaction times (5 and 15 min), mound formation was observed across the entire membrane surface. Total film (polyamide) thicknesses ranged from 116 nm (99 nm) at 30-s IP time to 1,282 nm (678 nm) at 15-min IP time (Table 1). For reference, typical TFC active layers prepared using the conventional IP procedure have thicknesses of <350 nm [8,13,17,18,42–47].

Fig. 4.

Digital photos and corresponding cross-sectional TEM images of new TFC membranes fabricated by forming polyamide films atop polysulfone supports using an alternative interfacial polymerization procedure in which the support layer is exposed to a continuously replenished supply of MPD. IP times ranged from 30 s (left) to 15 min (right). The numbers refer to recurrent morphological features described in Fig. 2 and Section 3.4.

We compared the active layer thicknesses of the control HC membrane (110 nm, ID#1 in Table 1) and the new TFC fabricated with a 60-s IP time (192 nm, ID#18 in Table 1) to evaluate the effect of the MPD delivery method on active layer thickness. The only difference in the fabrication of these two membranes was that ID#1 was prepared using the conventional TFC fabrication procedure while ID#18 was fabricated using the alternative IP procedure with a continuously replenished MPD supply. For both membranes, we used 3.5 wt% MPD, 0.15 wt% TMC, and 60-s IP time. Results show that using a continuously replenished MPD supply resulted in a total film thickness 74% greater than that generated using the more limited MPD supply of the conventional IP procedure. This supports our conclusion drawn from the membrane modification results (Figs. 2 and 3)—that restrictions in the MPD supply contribute to the limited active layer growth observed in conventional TFC fabrication.

Matthews et al. [13] provide results of membrane thicknesses achieved with conventional TFC fabrication at longer reaction times. Film thicknesses of their membranes did not exceed ~110 nm after 5 min of reaction time even when they used monomer concentrations (10 wt% MPD and 1.5 wt% TMC) substantially greater than is typical. In general, after the first minute of IP time, the active layer thickness increased only an additional ~20 nm or less in the next 4 min. If the reaction had proceeded to 15 min with this growth rate, the film thickness would have been <170 nm. By contrast, the films we produced with the alternative IP procedure at typical monomer concentrations (3.5 wt% MPD and 0.15 wt% TMC) were substantially thicker than theirs with thicknesses of 192 nm, 295 nm, and 1282 nm at 1 min, 5 min, and 15 min of IP time, respectively (Table 1). This observation further supports that substantially thicker films are achievable compared with those obtained by conventional TFC fabrication by reducing restrictions in the MPD supply.

There is an instance in the literature where an unusually thick (2.7 μm) polyamide active layer prepared using conventional IP was reported [18]. Xu et al. [18] achieved this result using a high MPD concentration (20 wt%), typical TMC concentration (0.1 wt%), and standard reaction time (1 min). The greater thickness of this membrane could be attributed to the atypically high MPD-to-TMC ratio (200:1) compared with the typical range of 10:1 to 20:1 [48]. The other MPD-to-TMC ratios used by Xu et al. were 20:1 or lower for which all other potential differences in experimental details between their study and others in the literature (i.e., membrane preparation, excess MPD removal method, support properties) did not result in thick (>350 nm) polyamide films. Both MPD-to-TMC ratios are nominal because the actual MPD-to-TMC ratio at the reaction zone is affected by various factors such as excess MPD removal protocols, support properties, and monomer diffusivities. Even so, more MPD would have infiltrated the support during the MPD soaking step with a 20 wt% MPD solution than a 2.0 wt% MPD solution, and a 200:1 MPD-to-TMC ratio would have provided an unusually high driving force for the MPD to permeate the incipient polyamide film. Since both conditions had the same support, the MPD reservoir (i.e., support porosity) and initial nanobubble confinement conditions were the same. Therefore, the high MPD concentration paired with the high MPD-to-TMC ratio likely resulted in the thick active layer. This instance of a thick membrane reported in the literature is evidence that thicker films can be achieved by changing the MPD supply conditions in TFC fabrication.

Similar to the TFCs modified with the alternative IP procedure, the active layer morphology of new TFCs was rough and heterogeneous with noticeable voids. The void structure was more prominent at longer IP times (Fig. 4), with void fractions increasing from 0.14 at 30-s IP time to 0.47 at 15-min IP time (Table 1). This observation is consistent with our previous discussion on the effect of the support (and pre-existing polyamide for modified membranes) on polyamide film morphology. Given that smaller pore sizes lead to exacerbated nanobubble confinement [24,36], greater void formation occurred at longer IP times, when the growing polyamide film (not the porous support) played a role in nanobubble confinement. The morphological features observed for modified membranes (Fig. 2) were also present in the new TFCs fabricated with the alternative IP procedure (Fig. 4). Specifically, at short IP times (30 s), the active layer comprised a single layer of densely packed void nodular and leaf-like features (Feature 1). At longer IP times, a layer of overlapping flat features (Feature 2) or a blanket-like layer of void-free polyamide (Feature 3) covered segments of the basal layer. Finally, the longest IP time (15 min) resulted in the formation of multi-layer structures of densely packed void nodular and leaf-like features (Feature 4) and void mega-nodules (Feature 5). The void mega-nodules were located towards the outermost region of the active layer, as was also observed in the modified membranes. Because the same morphological features were observed in both modified TFCs and new TFCs, these morphological features are likely inherent to interfacially polymerized polyamide under a consistent MPD supply.

In conventional TFC fabrication, the properties of the support layer (e.g., porosity) determine the amount of MPD that can be supplied to the reaction zone during IP [15] and the confinement conditions [24,36], such that a trade-off exists between the MPD storage capacity [49] and confinement conditions. Low porosity results in low MPD storage capacity and highly confined nanobubbles; high porosity results in high MPD storage capacity and less confined nanobubbles [24]. Our results show that the alternative IP procedure may provide the means to optimize MPD supply conditions and nanobubble confinement conditions (i.e., tune support porosity) independently. In addition to the conditions employed in the alternative IP procedure to reduce restrictions in MPD supply, one could use an interlayer with high MPD affinity to reduce restrictions to the MPD supply [50].

3.6. Effect of MPD delivery method and MPD concentration on films made or modified with the alternative IP procedure

For comparison with modified membranes, we evaluated the effect that the MPD delivery method (i.e., continuously replenished MPD supply, stirred, fixed-volume MPD supply, and conventional MPD supply) and MPD concentration (2.0 and 3.5 wt%) had on the active layer thickness and morphology of new TFCs fabricated using a 15-min polymerization time (Table 1, Fig. 5, Section S13). Membranes fabricated with different MPD supply conditions had substantially different morphologies and thicknesses. However, results were consistent with observations for modified membranes in previous sections: reducing restrictions in MPD supply conditions resulted in thicker active layers with a higher prevalence of voids. For example, the new TFC membrane prepared with the least restricted MPD supply conditions (i.e., 3.5 wt% continuously replenished MPD supply, corresponding to Fig. 5a, ID#20 in Table 1) had the thickest active layer (1,282 nm), the fastest film growth rate over 15 min (85 nm/min), and the highest prevalence of multi-layer void structures (Feature 4). By contrast, the new TFC prepared with the most restricted MPD supply conditions (i.e., 2.0 wt% conventional MPD supply, corresponding to Fig. 5f, ID#25 in Table 1) had the thinnest active layer (211 nm), slowest film growth rate over 15 min (14 nm/min), and absence of multi-layer void structures (Feature 4). New TFCs prepared with intermediate MPD supply conditions (i.e., Fig. 5b (ID#22), 5d (ID#21), and 5e (ID#23)) resulted in intermediate thicknesses (610 nm, 529 nm, and 361 nm), intermediate film growth rates over 15 min (41, 35, and 24 nm/min), and mixed morphologies. Also, notably, the MPD delivery method had a stronger effect on thickness and active layer morphology than MPD concentration. Using a continuously replenished MPD supply, regardless of MPD concentration, resulted in (1) the occurrence of multi-layer void structures (Feature 4; such structures were not observed in any of the samples prepared using a fixed-volume or conventional MPD supply) and (2) thicker membranes than those prepared using a fixed-volume or conventional MPD supply.

Fig. 5.

Cross-sectional TEM images and surface SEM images of new TFC membranes fabricated with 15 min of IP time. Supports were exposed to a (a) continuously replenished supply of 3.5 wt% MPD, (b) stirred, fixed-volume supply of 3.5 wt% MPD, (c) conventional MPD supply of 3.5 wt% MPD, (d) continuously replenished supply of 2.0 wt% MPD, (e) stirred, fixed-volume supply of 2.0 wt% MPD, and (f) conventional MPD supply of 2.0 wt% MPD. Labels are Features 2, 4, and 5 depicted in Fig. 2.

We also evaluated the effect of MPD concentration (2.0 and 3.5 wt%) on the thickness and morphology of HC and SWC4+ membranes modified with a continuously replenished MPD supply and 15 min of polymerization time (Table 1, Fig.S8). Results were consistent with the observations above on the effect of restrictions in MPD supply on active layer thickness and morphology. Specifically, for HC and SWC4+ membranes, a higher MPD concentration resulted in thicker modified active layers: 2.4 and 1.7 times greater for HC (ID#3 vs ID#5) and SWC4+ (ID#8 vs ID#10), respectively. Additionally, for the HC membrane (which had the highest MPD permeation rate, Fig. 1b), the lower MPD concentration resulted in membranes lacking the multi-layer void structures (Feature 4) that characterized the samples modified with the higher MPD concentration. For the SWC4+ membrane (which had the lowest MPD permeation rate, Fig. 1b), the lower MPD concentration resulted in active layers lacking the void mega-nodules (Feature 5) that characterized the samples modified with the higher MPD concentration.

3.7. Solute permeation through modified membranes

We compared solute permeation through modified (30-s IP time) and unmodified TFCs. Results (Fig. 6) show that MPD permeated both modified and unmodified membranes in a diffusion cell setup, but the permeation rate was substantially lower for modified membranes than for their unmodified counterparts. Relative decreases in MPD permeation at 60 min ranged from 16% (HC) to 73% (SWC4+). We observed similar results for NaCl permeation (Fig. S9); modified membranes displayed substantially lower NaCl permeation than unmodified counterparts, confirming their sustained integrity. Relative decreases in NaCl permeation ranged from 58.3% (ESPA3) to 88.4% (SWC4+).

Fig. 6.

Permeation of MPD per unit membrane area through (a) SWC4+, (b) ESPA3, (c) NF90, and (d) hand-cast (HC) membranes that underwent 30 s of additional polyamide formation (dark symbols) compared to MPD permeation through the same membrane coupons prior to being modified (light symbols). Note that the y-axis scale is different in each panel to clearly show changes in MPD permeation in membranes modified with additional polyamide formation.

The MPD and NaCl results indicate that the additional polyamide thickness observed in modified membranes resulted in a decreased solute permeation rate. This is consistent with an increase in polyamide thickness, as permeation rate is inversely proportional to active layer thickness [51]. The observed decreases in permeance, combined with observations from our visual inspections, indicate that the additional polyamide is an integral part of the modified active layer. These results also indicate that polyamide growth promoted via a consistent supply of MPD to the support layer can be used to modify membrane performance, which we further verified by evaluating water permeance and salt rejection for unmodified and modified membranes in cross-flow filtration tests (Section S15). The results clearly show that membrane performance is affected by the additional polyamide formation. In particular, the water flux of modified membranes decreased between 56.3% and 86.2% compared with unmodified counterparts. Similarly, the water flux was 31.1% lower for a TFC membrane fabricated using the alternative IP procedure than for a TFC membrane fabricated using the conventional IP procedure, both with the same polymerization time (1 min).

In general, membrane scientists are not interested in making membranes with thicker active layers because doing so comes at the detriment of water flux. Although we devised the alternative IP procedure to gain fundamental insights into the IP reaction, not to enhance membrane performance compared with existing technology, the alternative IP procedure could potentially be revised or optimized improve membrane performance. For example, one could target blocking the permeation pathways of contaminants of interest, such as boron, while minimizing increased active layer thickness. Even if the optimized procedure also blocks some of the permeation pathways of water, the improvement of boron rejection would be a triumph for desalination processes [52].

4. Conclusions

We compared the thickness and morphology of polyamide active layers in TFCs fabricated via conventional TFC fabrication with those fabricated or modified using an alternative IP procedure that reduced restrictions in MPD supply. We used a range of MPD supply conditions and a set of membranes with a range of water permeance and salt rejection performance levels. The results support the following conclusions:

• MPD can permeate TFC polyamide active layers at substantial rates.

• Limitations in polyamide film growth (<350 nm) observed in conventional TFC fabrication can be overcome by reducing restrictions in the MPD supply conditions.

• The least restricted MPD supply conditions used in this study resulted in a TFC with a polyamide active layer having a thickness of 1282 nm.

• Five distinct recurrent morphological features were identified in membranes modified or fabricated with the alternative IP procedure (Fig. 2); Feature 1 (basal single layer of tightly-packed void nodular and leaf-like features) and Feature 2 (overlapping flat features atop the basal layer) were observed in unmodified commercial membranes and have been described in the literature.

• Reduced restrictions in the external MPD supply conditions paired with high MPD permeation rates through membranes promoted the development of multi-layer void structures (Feature 4).

• Reduced restrictions in the external MPD supply conditions paired with low MPD permeation rates through membranes promoted the development of void mega-nodules (Feature 5).

• The formation of Features 4 and 5 appeared to be preceded by the formation of Feature 2 (overlapping flat features) or a blanket-like void-free film (Feature 3) atop the basal layer (Feature 1).

• The observed morphologies were consistent with the reported mechanisms of formation of void structures in interfacially polymerized polyamide (i.e., CO₂ degassing and nanobubble confinement).

These insights aid our understanding of the formation mechanisms of polyamide active layer thickness and morphology. Our new understanding of the relationship between MPD supply restrictions (i.e., MPD permeation rate, MPD concentration, MPD delivery method, IP time) and active layer thickness and morphology could be used to optimize nanobubble confinement without restriction to the MPD supply and develop new procedures for achieving novel TFC structures. Future work could focus on optimizing the alternative IP procedure by using an interlayer with high MPD affinity to further reduce restrictions in MPD supply or targeting enhanced rejection of contaminants known to be rejected poorly (e.g., boron, trivalent arsenic, and N-nitrosodimethylamine) by membranes.

Highlights.

• MPD permeates TFC polyamide layers substantially

• Reducing restrictions in MPD supply during IP generates polyamide layers >350 nm

• The least restricted MPD supply conditions resulted in polyamide thicknesses >1 μm

• High MPD permeation rates promoted development of multi-layer void structures

• Low MPD permeation rates promoted development of void mega-nodules