Molecular layer deposition for the fabrication of desalination membranes with tunable metrics

Highlights

• MLD added an ultrathin MPD-TMC polyamide active layer to commercial NF membranes.

• Progressive MLD systematically altered flux/rejection metrics from NF to RO values.

• Controlled increase of MLD layer thickness increased rejection but decreased flux.

• Atomic force microscopy revealed that MLD removed small-scale roughness features.

Abstract

The recent advancement of semiconductor devices to the near-atomic scale necessitated the development of atomic layer processing methods, including molecular layer deposition (MLD). This gas-phase deposition technique creates semipermeable polymer films with precise control of composition and thickness. Herein, MLD was used to produce thin-film composite reverse osmosis membranes. Aromatic polyamide films as thin as 0.5 nm were applied to NF270 nanofiltration membranes using m-phenylenediamine and trimesoyl chloride. Within two molecular layers, desalination performance was affected. As film thickness increased to 15 nm (48 MLD cycles), performance progressed from nanofiltration to reverse osmosis metrics in terms of salt rejection and water permeance. With film thickness > 5 nm, rejection values exceeded a small sampling of commercial membranes. In all cases, a tradeoff between rejection and permeance was observed. Atomic force microscopy measurements indicate that MLD enhancement led to removal of small-scale roughness features and resulted in a root mean square roughness difference of <0.1 nm from the substrate. These initial MLD studies represent a novel processing approach that offers a potential pathway for the fabrication of membranes with finely tailored properties.

Introduction

Conventional thin-film composite (TFC) reverse osmosis (RO) membranes made with interfacial polymerization (IP) require control over a number of important processing parameters. Variations in concentration [1], [2], substrate [3], [4], [5], solvent [6] and synthesis procedure [7] impact the density, morphology, homogeneity, and thickness of the IP polyamide layer [8]. Such variability in turn can affect desalination performance, operation lifetime and fouling characteristics. While many aspects of the nanometer-scale mechanics of the solution-based film growth have been characterized, its complex nature prohibits comprehensive understanding and control over this decades-old technique [7]. As a result, current methods lean on empirical approaches rather than molecular-level design [2].

Despite these challenges, advances in fabrication have enabled the creation of RO membranes which incorporate ultrathin, chemically and morphologically controlled polyamide active layers. Such characteristics are desirable for minimal transport resistance and improved anti-fouling behavior [8], [9], [10]. The molecular layer-by-layer (mLbL) method [11], [12], [13], [14], [15], [16] has proven effective in generating RO membranes with a dense layer as thin as 4 nm and with a root mean square (RMS) roughness of ~1 nm [16]. Free standing film growth techniques have been leveraged to create RO polyamide film thicknesses below 10 nm [1], [5], [9]. Electrospray processes have also been used to fabricate polyamide RO films as thin as 20 nm and with a RMS roughness that is indistinguishable from the substrate [17]. These complex procedures reiterate the challenge of controlling the liquid phase chemistry at a nanometer scale. While achieving impressive thickness/roughness dimensions, each of these fabrication routes still faces the limitations of IP including incomplete crosslinking and limited consistency in film thickness and density. Additionally, the mLbL and free standing film techniques are difficult to scale up for commercial production and produce films which are prone to delamination, relying on physical adhesion or interlayer materials which are susceptible at extreme pH and salt concentration [14], [15], [18]. Thus, the need remains for a scalable and robust TFC membrane synthesis technique that enables fine control of film nanostructure and thickness.

Molecular layer deposition (MLD) is a scalable, polymer thin-film deposition technique that can provide extremely thin, smooth, chemically consistent films with nanometer precision [19]. The MLD process is analogous to atomic layer deposition (ALD), differing by its use of organic precursors to incorporate molecular fragments [20]. Whereas ALD precursors typically contribute one atom into a metal or ceramic film, MLD precursors incorporate an entire organic linkage. These techniques have become instrumental in achieving nanoscale transistors and enabling previously unachievable length scales for gate oxides and other components of integrated circuits and memory devices [21], [22]. They have also been used in energy storage [23], [24] and conversion [25]. ALD and MLD offer exciting possibilities for improved membrane materials and have already been utilized for tuning pore sizes as well as affecting hydrophilicity and fouling characteristics [26].

A schematic of the MLD process is shown in Fig. 1. All-organic MLD is performed under vacuum conditions by introducing a substrate to two gas-phase organic precursors in an isolated and sequential order. Upon exposure to the initial reactant, functional groups of the precursor (e.g. acyl chlorides) react with the functional groups of the surface (e.g. amines). A layer of monomers is added to the surface by a single step of step-growth polymerization. The substrate is now terminated with a new functional end group (acyl chlorides). The precursors are chosen such that self-polymerization will not occur, ensuring self-limited growth at each exposure. Once the reaction environment is cleared of the first precursor and reaction byproducts, the surface is exposed to a second precursor with functional groups which are reactive with those of the first precursor (e.g. amines). The polymer chain-ends are again extended by one monomer in a self-limiting fashion. The reaction space is once again purged. This two-step process constitutes one MLD cycle which may be repeated for controlled growth of polymer films [20]. The formation of polymer films containing crosslinks may be achieved by use of a trifunctional precursor [27]. This sequence of self-limited, step-wise surface saturation provides many advantages. Film thickness is highly controlled, scaling linearly with increasing cycles. Films are smooth and conformal to the substrate surface [28].

MLD is capable of producing semipermeable polymer films for water purification without the use of environmentally harmful solvents; this includes the often-used aromatic polyamide made from m-phenylenediamine (MPD) and trimesoyl chloride (TMC) [30], [31], [32]. Homogenous films grown with this controlled, layer-by-layer technique are ideal in view of recent work by Culp et al. which concluded that membrane resistance is the result of nanoscale variations in thickness and density [8].

In principle, MLD is very much akin to the mLbL technique: both are layer-by-layer processes which grow polymer films with monomer precision. Each method creates smooth, conformal films, but each has its own advantages. MLD exhibits more control over film thickness with a deposition rate around 0.4 nm/cycle compared to the mLbL rate of ~1 nm/cycle [13], [30]. The liquid phase mLbL may be done at ambient temperatures but requires heavy solvent usage and long cycle times. The mLbL solution chemistry is simple to perform on the lab bench but is difficult to scale for commercial production. Gas-phase MLD requires temperatures sufficient to vaporize the reactants, but the solvent-free technique can be performed with sub-second cycles times [30]. MLD is compatible with commercial scale roll-to-roll (R2R) processing and has a clear path to manufacture [33], [34]. Despite its advantages, fabrication of a robust, freestanding MLD film remains a difficult objective due to the conformal, isotropic manner of film growth. MLD precursors diffuse into pore openings and deposit material within, not across pore openings [32]. Thus, fabrication of desalination membranes using MLD has been a significant technical challenge.

This paper demonstrates that the advantages of MLD can be realized by utilizing nanofiltration (NF) TFC membranes as a suitable substrate. The existing films atop NF membranes provide sufficient transport resistance to MLD precursors which would otherwise deposit onto pore walls. FilmTec NF270 membranes were chosen as the substrate for subsequent MLD processing given their ability to withstand high temperature exposure in a vacuum. This study explores the ability to control the tradeoff between salt rejection and water permeance by altering the MLD film thickness. To the authors’ knowledge, this study represents the first demonstration of active membrane films made by gas-phase MLD.