Research Article - (2016) Volume 6, Issue 4

Insight in to the Initial Stages of Silica Scaling Employing a Scanning Electron and Atomic Force Microscopy Approach

Bogdan C Donose*, Greg Birkett and Steven Pratt
The University of Queensland, School of Chemical Engineering, St Lucia, 4072, QLD, Australia
*Corresponding Author: Bogdan C Donose, The University of Queensland, School Of Chemical Engineering, StLucia, 4072, QLD, Australia, Tel: +61(0)413286311 Email:


The performance of reverse osmosis (RO) desalination can be limited by membrane scaling. Of particular concern is silica scale, which once deposited on the membrane is extremely difficult to remove. In this work, the deposition of silica-rich nanoparticles was considered. A novel in situ sample preparation method was developed for a microscopy investigation into the deposition and adhesion of the silica-rich nanoparticles. The method involves placing a clean silica wafer in agitated brine to collect particles to simulate initial stages of scaling. The ‘scaled’ surfaces were characterized by scanning electron microscopy (SEM) and atomic force microscopy (AFM). Model brines, with varying nanoparticle, cation, and organic composition and concentration were tested, as well as reject brine from a full-scale operational water treatment facility. Microscopy revealed that silica-rich nanoparticles were deposited from all waters, with smaller nanoparticles more readily attaching to the wafer compared to larger ones. The presence of organics increased nanoparticle adhesion whereas divalent cations (Ca2+ and Mg2+) decreased nanoparticle adhesion. These results have implications for the evaluation, selection and operation of RO pre-treatment processes and chemical dosing strategies, particularly the requirement for weak acid cation ion exchange (WAC-IX) and anti-scalant chemicals, respectively.

Keywords: Reverse osmosis; Brine; Silica; Scaling; Colloids; Electron and atomic force microscopy


Reliable, safe and cost effective operation of reverse osmosis (RO) membrane desalination technology is vital to support development activities, urban and potable water supplies and agricultural activities in many parts of the world. In Australia, for example, the development of natural gas from coal seams offers tremendous economic opportunity, with development generating brackish ‘produced water’ which operators can manage using reverse osmosis (RO) membrane desalination to treat water to a quality suitable for a range of beneficial reuse such as agriculture. A waste brine stream is generated during the treatment process which requires storage and further treatment, management and subsequent disposal. A key operational objective can be to maximize the recovery ratio of the plant so to minimize the volume of brine requiring storage prior to, or during, treatment and disposal. The performance of the RO plants can be limited by silica derived membrane scale, which can be difficult to remove. Scaling is a common operational issue with RO membranes as documented in extensive academic and industry literature [1-6]. Groundwaters, including produced waters, are a complex matrix containing silica, hardness ions (e.g. calcium and magnesium), organics and other metal ions, with concentrations varying widely across regions. Our research group has recently shown that produced water has significant concentrations of silica-rich nanoparticles [7,8] and nanoparticles have been shown in the literature to have an adverse impact on the performance of membrane desalination [9-15]. A challenge in understanding the scaling mechanisms is the relatively slow kinetic of the process. Characterisation normally requires long-term experiments to make macro-scale observations. For this reason a new methodology was developed to quickly assess interactions between nanoparticles and membrane surfaces.

The interplay between silica-rich nanoparticles and the other components of produced water is not well understood. For example, experiments [10] have shown that the membrane flux decline during combined scaling and biofouling depends on solution chemistry and colloidal particle size. The combined fouling behaviour cannot be extrapolated from fouling by colloidal materials or dissolved organic matter alone. Synergistic effects were also found in fouling tests combining extracellular polymeric substances extracted from Pseudomonas aeruginosa PAO-1 and calcium ions [16]. These synergistic effects make selection and operation of infrastructure to mitigate scaling very challenging. For example, current literaturesupported practice is to have weak acid cation exchange (WAC-IX) before the RO unit to reduce hardness in the RO feed stream. Some operations in the natural gas industry in Queensland (Australia) report that WAC IX operation has increased the frequency of RO membrane scaling events, with irreversible silica scale formation observed on RO membranes. Other operators have used IX without incident to the membranes but it is not clear if this operation is optimal.

In this work a novel in situ sample preparation method was developed for use with atomic force and electron microscopy to investigate the deposition and adhesion of silica-rich nanoparticles. The method simulates the initial stages of silica scale formation, whereby nanoparticles bind to silica-rich nanoparticles that are caught on membrane surfaces. The application of the method is demonstrated accompanied by a preliminary assessment of water matrix effects on particle adhesion. Synthetic produced water (with varying nanoparticle, cation and organic content and composition) and brine from an operational reverse osmosis desalination facility are tested.

Materials and Methods

The experimental program involves depositing nanoparticles that are present in water samples on to silicon wafers (the silica trap), and then microscopy examination of the silicon wafers to assess the extent of deposition.

Silica trap

Silicon wafers (coated with a naturally grown silica layer of 2-4 nm) were diced (5×5 mm), washed in 10% H2SO4, rinsed in ethanol and deionised (DI) water and then cleaned (30 minutes) using a UV/Ozone unit (BioForce).

The wafers were immersed in the water samples in individual 50 ml plastic beakers for 24 hours under constant agitation (250 rpm) on an orbital shaker. After this step, the wafers were removed from the plastic beakers and immersed in DI water to rinse. Excess water was then removed from their surface using the corner of an extra low lint Kimwipes and wafers were placed in a vacuum oven for 8 hours at 70ºC. Samples were stored in a high vacuum cabinet prior to mounting and SEM/AFM analysis. All steps undertaken for sample preparation and analysis are schematically presented in Figure 1.


Figure 1: Silica trap experimental setup and preparation/analysis steps.

Water samples

Table 1 summarises the types of water that were analyzed. Brine samples from a produced water RO desalination facility were typically collected on-site for the characterization of chemical composition. Water is separated from the natural gas at the well site and sent to the water management facility. The produced water then undergoes microfiltration followed by RO desalination.

Experiment Type of water Characteristics
1 Model RO feed pH, cations, organic surrogate, nanoparticles
2 Model brine pH, cations, organic surrogate, nanoparticles
3 Real brine Water matrix

Table 1: Waters tested employing silica trap method and their characteristics.

Physical and chemical characterization of silica and other potential scalants were conducted using various spectroscopic and analytical techniques. Historical water quality data obtained from the RO desalination facility was used to generate an overview for a period of seven years. The RO process operates at around 80% recovery, generating a brine stream which is about 5 times as concentrated when compared to the RO feed. The water quality data of RO feed and brine is shown in Table 2. This shows that the produced water is rich in sodium and bicarbonate, with potassium, calcium, and magnesium also present. The silica concentration in the produced water feed is relatively low, but the recovery process increases the silica concentration to near saturation. Such a concentration increase can lead to a dramatic change in the silica particulate fraction: before RO approx. 7% and after RO approx. 35% [7].

Measurement Unit RO feed Brine Conc. factor
Average St. dev. Average St. dev.
Alkalinity (carb.) mg/L 197.7 61.0 543.8 155.7 2.7
Calcium mg/L 9.0 4.1 45.4 10.0 5.0
Magnesium mg/L 2.5 1.2 13.1 2.5 5.2
pH - 9.0 0.1 8.7 0.1 -
Potassium mg/L 16.8 9.7 83.2 25.4 4.9
Silicon mg/L 20.6 10.1 48.5 16.1 2.3
Sodium mg/L 2557.8 877.4 12644.0 2062.4 4.9
TDS mg/L 6565.5 677.5 27519.6 4281.8 4.1
TOC mg/L 9.5 10.6 40.0 26.4 4.2

Table 2: Overview of RO feed and brine matrix profile.

Synthetic produced waters, RO feed and brine, were prepared using DI water produced by a Millipore Academic unit (resistivity lower than 18 MΩ.cm), silica nanoparticle standards (Sigma, Ludox 12 nm in diameter and Nissan chemical, Snowtex ST-20L, 60 nm in diameter). CaCO3, K2CO3, MgSO4, NaCl were of analytical grade and were used without prior purification. 1.0 M HCl and 1.0 M NaOH were used for pH adjustment.

Using the historical RO desalination facility data as a guide, two types of synthetic waters were used: brine and produced water feed. Concentration factors of 5 were imposed for all ions. Considering silicon is present as dissolved and particulate silica, a concentration factor of 25 for silicon was chosen to cover the observed ranges of particulate and dissolved Si, i.e. to account for a fivefold increase in silica particulate fraction seconded by a silicon concentration factor of five. In this way the chance to observe attached nanoparticles by microscopy techniques was increased.

Table 3 shows the program for the model RO feed and model brine experiments. The table shows the types of samples used in this study as well as the pH values of the synthetic waters in which silica wafer traps were immersed. The conditions tested were: presence or absence of silica nanoparticles or dissolved silica (leftmost columns in Table 4) and different combinations of hardness ions, organics surrogate in deionised water (top row). As silica surrogates, Ludox (diameter=12 nm), Snowtex (diameter=60 nm) and dissolved silica in HF were employed.

Ion Model RO feed(mg/L) Model brine
Na+ 2200 11400
Mg2+ 2 10
Ca2+ 8 40
CO32- 72 36
K+ 39 183
SO42- 9 48
Si4+ 4 100
Alginate 20 100

Table 3: Model brine and RO feed water matrices (with all ions present).

Na mg/L Ca
Silica nanoparticles and pH
1 2200 8 2 20 8.2 8.8 8.7 8.7
2 2200 8 2 - 8.6 8.9 8.9 8.5
3 2200 - - 20 8.9 8.9 8.9 8.9
4 2200 - - - 8.0 8.9 9.0 8.2
5 - - - - 8.7 8.7 8.8 8.9
Model brine  
1 11400 40 10 100 8.5 8.6 8.6 8.9
2 11400 40 10 - 8.5 8.6 8.5 8.7
3 11400 - - 100 8.8 8.8 8.5 8.8
4 11400 - - - 8.8 8.8 8.5 8.9
5 - - - - 8.7 8.7 8.9 8.7
Real feed 2557.8 9.0 2.5 9.5 20.6 9
Real brine 12644.0 45.4 13.1 40.0 48.5 8.7

Table 4: Sample matrix and pH for synthetic and real waters. Sample codes: NO=no silica; ST=dissolved silica; LU=Ludox silica; SN=Snowtex. The numbered columns and labels show the presence/absence of major ions in solution.

The real brine samples were treated in the same manner for microscopy analysis.

Scanning electron microscopy (SEM) and atomic force microscopy (AFM)

SEM was performed by employing a JEOL 7001 Field Emission Gun at 10 mm working distance, spot size 8 or 9 and 5 kV-10 kV accelerating voltage.

AFM micrographs were recorded employing an MFP-3D built on an Eclipse Ti-U Nikon inverted microscope, placed on a Herzan antivibration table which was housed in a TMC enclosure and a Cypher (Asylum Research/Oxford Instruments) using Multi 75 DLC (Budget Sensors, Bulgaria) probes of nominal contact radius less than 15 nm and nominal spring constant of 2 N/m. All samples were imaged at 0.7- 1 Hz scan rate in AC mode in air over an area of 2×1 μm.

For both microscopy methods samples were imaged as received, post vacuum deposition (as described in above in 2.1 Silica trap), in order to prevent interference potentially introduced by a conductive layer. In this way the errors of scanning are limited only to accelerating voltage and spot size (for SEM) and probe contact radius (for AFM).

Results and Discussion

Experiment 1: Model RO feed

SEM and AFM micrographs for synthetic RO feed produced water are presented in Figures 2 and 3. Both microscopy techniques confirm that nanoparticles were trapped on the wafer. All samples exhibit adhered structures. Several features are captured by SEM at a larger scale, allowing for a general sample overview, while AFM reveals more three-dimensional details. These features include alginate networks (AFM NO-1), salt crystals (SEM NO-3) and nanocrystals (AFM LU-4), as well as nanoparticles (SEM SN-1, AFM SN-4).


Figure 2: Representative SEM micrographs of the wafer surfaces after exposure to synthetic RO feed waters and selected features marked by red arrows (Scale bars: NO-1 to NO-5, 10 μm; ST-1 and 2, 100 μm; ST-3 to 5, 10 μm; LU-1, 3 and 4, 1 μm; LU-2 and 5, 10 μm; SN-1, 3 and SN5, 1 μm; SN-2 and 4, 100 nm.


Figure 3: Representative AFM micrographs of the wafer surfaces after exposure to synthetic RO feed waters and selected features marked by green arrows (Scale bars: 400 nm): NO-1: dendritic structures; LU-4: large salt crystals; SN 4: isolated NPs clusters.

As described in the experimental section, after exposure to the water sample under constant agitation, wafers are rinsed with DI water, the excess water is removed and then subjected to thermally assisted vacuum drying. This approach poses significant challenges in choosing a representative area for analysis, especially when samples exhibit salt drying patterns leading to imaging artefacts such as in SEM ST-1 and SEM ST-2 (Figure 2).

In the presence of cations and alginate the dendritic alginate structures (from AFM) and aggregates with colloids (SEM and AFM) are apparent (Figures 2 and 3: column 1. Na/Ca/Mg/Alg). When alginate is removed (column 2. Na/Ca/Mg) the dendritic structures disappear and patches of aggregates can be observed. Removing the divalent cations (column 3. Na/Alg) results in a compact multilayered mat of Ludox nanoparticles being deposited on the trap surface. It is also shown that SN nanoparticles exhibit a different behaviour, suggesting that smaller NPs are more easily attached to the surface. In the presence of just sodium (column 4. Na) different attachment behaviour between SN and LU NPs was observed, with smaller particles attaching more readily. The benchmark (column 5. DI water) confirms little or no attachment as result of electrostatic repulsion.

Experiment 2: Model brine

SEM and AFM micrographs for synthetic brine are presented in Figures 4 and 5. Similar to the case of model RO feed, both microscopy techniques confirm that the trapping approach was successful. For the model brine the adhered structures were more abundant. This was expected because concentration of ions in the brine was increased by a factor of 5 compared to the RO feed, except for silicon which was increased by a factor of 25 to reach 100 mg/L. Again, the observed features included alginate networks (AFM NO-1), salt crystals (SEM NO-3) and nanocrystals (SEM NO-4), as well as nanoparticle dense mats (SEM SN-4, AFM SN-4).


Figure 4: Representative SEM micrographs of the wafer surfaces after exposure to synthetic brine waters and selected features marked by red arrows (Scale bars: NO-1, 2 and 4, 1 μm; NO-3, 10 μm; ST-1 to 5, 10 μm; LU-1 to 3, 100 nm; LU-4, 1 μm; LU-5, 10 μm; SN-1 to 4, 1 μm; SN-5, 10 μm).


Figure 5: Representative AFM micrographs of the wafer surfaces after exposure to synthetic brine waters and selected features marked by green arrows: NO-1 dendritic structures; SN-4 voids in the Snowtex layer.

Differences in the adhesion of nanoparticles are most pronounced when comparing both SEM and AFM for the LU4 and SN4 samples. The larger Snowtex particles form layers with voids, while Ludox mats are compact and possibly multi layered. Even at high silicon concentrations (100 mg/L) benchmark tests with DI water show again a total lack of adhesion due to electrostatic repulsion.

A summary of distinguishing features observed on all samples is presented in Table 5.

1 Na/Ca/Mg/Alg 2 Na/Ca/Mg 3 Na/Alg 4 Na 5.
pH adjusted DI water
(without silica)
Alg. network Salt nanocrystals Alg. networks and colloids - Clear surface
(dissolved silica)
Alg. islands Clear surface Alg. network andnanocrystals - Few colloids
(Ludox nanoparticles)
Alg. network Large crystals Colloids mat Colloids mat Clear surface
(Snowtex nanoparticles)
Colloids and alg. networks Isolated colloids Isolated alg. embedded colloids Dispersed colloids Clear surface
1 2 3 4 5
(without silica)
Alg. network - Alg. network &colloids Salt crystals Clear surface
(dissolved silica)
Silica and alginate islands Aggregates Alginate network and colloids Crystals and colloids Colloids and aggregates
(Ludox nanoparticles)
Colloids dense mat Colloids mat and aggregates Alginate embedded colloids and islands Colloids mat &aggregates Clear surface
(Snowtex nanoparticles)
Colloids loose multilayer Colloids loose layer and aggregates Alginate embedded colloids Colloids loose layer Clear surface

Table 5: Sample matrix and distinguishing features on traps exposed to synthetic RO feed and brine.

Experiment 3: Real brine

Microfiltered brine from a RO desalination facility was also considered. Data is presented in Figure 6, along with, for comparison, traps exposed to synthetic produced waters containing all ionic components and grouped as: no silica, standard dissolved silica, Ludox nanoparticles and Snowtex nanoparticles.


Figure 6: AFM comparison (to scale in XY coordinates) between surfaces exposed to real WTF micro-filtered brine and synthetic (1: Na/Ca/Mg/Alg) waters: no silica, standard silica, Ludox and Snowtex. Marked features: dendritic structures (squares), colloidal aggregates (triangles) and salt crystals (circles).

The figure shows that traps exposed to microfiltered brine exhibit a combination of all surface elements presented on the traps from the model waters: dendritic structures (squares), colloidal aggregates (triangles) and salt crystals (circles). The topographical differences between RO feed and brine in samples containing all the ingredients (mono and divalent cations and organic surrogate/alginate) are highlighted.

There are three types of effects emerging from the existing data:

1. Colloids size is a determining factor in deposition and aggregation as visualised in synthetic brine, with Ludox, the smallest colloids, being easily attached and forming multilayered colloidal mats in comparison with Snowtex which attached as loose monolayers.

2. The presence of alginate (as surrogate for organic matter) in the matrix contributes to an enhancement of colloids/particulates attachment and aggregation.

3. When cations are removed from the RO feed an enhancement of silica deposition on the trap wafer is observed in model feed water.


As previously postulated [17], the adhesion between solid particles and solid surfaces is determined by van der Waals, electrostatic and deformational forces [18]. For particles in the nanoscale domain the long range electrostatic force becomes less important and van der Waals, structural and hydration forces dominate interaction and challenge to the limit existing analytical tools such as surface force apparatus and AFM. Over the recent decades of nanoscale force measurements, silica and mica surfaces were tested typically in electrolyte solutions of controlled pH and salt concentration to learn how van der Waals, adhesion, solvation, structural and friction forces behave [19-27]. To our knowledge, there is no comprehensive model that describes interactions in complex fluids such as brines resulting from water treatment. In an experimental system such as the one designed and analysed here, as a proxy to the brine or pre-brine medium, the roughness of the substrate was chosen on purpose to be at its minimum in order to reduce its contribution. The propensity of smaller nanoparticles to attach to the smooth silica could be attributed to the increasing role of van der Waals interactions and to the reduced hydrodynamic effect acting synergistically toward immobilising very small colloids.

The role of alginate in the enhanced colloid/particulates attachment has been previously tested [28] in high ionic strength environments to show that the polyelectrolyte nature of alginate reduces steric repulsion and favours adhesion. Similarly, its ability to enhance adhesion has been harnessed to boost the impact and efficiency of drug delivery systems [29]. In certain experimental conditions [30,31] alginate was shown to be able to restore flux in fouled membranes, based on its ability to interact with silica.

Without having a definitive answer with regard to the effect of the absence of cations on the silica deposition, namely an increasing amount of deposited colloids, this behaviour could be linked to the ability of certain cations to reduce friction between silica surfaces [21,22]. Also, the role of pH, needs to be further assessed, especially in the vicinity of pH 9 [25] where silica has an accelerated dissolution rate and where hydration and hydrogen bonding and depolymerisation play such an important role.


Silica wafer samples prepared by immersion for 24 hours (under constant agitation) in different types of water are suitable to trap nanoparticles, nanocrystals and organic dendritic structures and their aggregates, which later on, can be subjected to microscopy. Considering silica-silica interactions in different media, the new method allows taking snapshots of the early stages of scaling, replacing a rough RO polyamide substrate [31] by a flat wafer (roughness less than 1 nm as measured by AFM), more suitable for SEM and AFM. The combination of two microscopies and samples’ preparation enables obtaining information which otherwise would be inaccessible.


This research is supported and funded by the UQ Centre for Coal Seam Gas, which includes industry members APLNG, Arrow Energy, QGC and Santos Ltd. Deirdre Walsh, Paul Wybrew and Josh Davies are acknowledged for reviewing the manuscript. This work was performed in part at the Queensland node of the Australian National Fabrication Facility, a company established under the National Collaborative Research Infrastructure Strategy to provide nano and microfabrication facilities for Australia’s researchers. The authors acknowledge the facilities and the scientific and technical assistance of the Australian Microscopy & Microanalysis Research Facility at the Centre for Microscopy and Microanalysis (The University of Queensland).


Citation: Donose BC, Birkett G, Pratt S (2016) Insight in to the Initial Stages of Silica Scaling Employing a Scanning Electron and Atomic Force Microscopy Approach. J Membra Sci Technol 6:163.

Copyright: © 2016 Donose BC, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.