Stimulated Raman Scattering Microscopy: Real-Time In-Situ Physical and Chemical Characterization of Reverse Osmosis Desalination Membrane Scaling

Abstract

We introduce a stimulated Raman scattering (SRS) methodology designed for rapid, real-time, and in situ monitoring of RO membrane scaling adapted for bench-scale desalination flow cells. The methodology can provide new insights into membrane scaling dynamics by offering time-resolved reflection imaging of inorganic crystal growth, coupled with chemical identification from Raman spectral data. These capabilities allow for direct local measurement of the membrane surface area covered by different scalants as well as an approximation of the scalant volume using three-dimensional, integrated Raman intensity. The 2D and 3D SRS results obtained from CaSO₄ scaling experiments are compared to and are in reasonable agreement with those provided by confocal microscopy. The real-time physical and chemical characterization capabilities presented here could be extended to study combinations of inorganic, organic, and biological fouling. Overall, the SRS methodology represents an advancement in real-time sensing of membrane fouling that offers the potential for improved operation, lower cost, and more resilient RO membrane systems for sustainable water management.

1. Introduction

The increasing severity of global water scarcity presents one of humanity’s most pressing challenges, impacting everything from environmental resiliency, public health, food security, and economic stability. , In this critical context, effective water filtration stands as a cornerstone solution, enabling the sustainable use and reuse of precious water resources. − However, the reliability and efficiency of these filtration systems are dependent on robust, state-of-the-art monitoring techniques, which could transform how our water future is managed and secured. − Here, we describe a stimulated Raman scattering (SRS) methodology for rapid, real-time, and in situ monitoring of bench-scale reverse osmosis (RO) membrane desalination flow cells, offering a significant advancement in characterization of membrane scaling combining imaging and chemical identification.

Water scarcity is a rapidly escalating crisis that impacts environmental resiliency, public health, food security, economic stability and more. ,,− Currently, 55% of the world’s population experiences a deficit of clean water for at least one month per year. By 2100, this number is projected to increase to 66%. This challenge is amplified by climate change, population growth, urbanization, and pollution. ,,− In this critical context, effective water filtration stands as a cornerstone solution, enabling the sustainable use and reuse of precious water resources. − Over the last few decades, desalination − has emerged as a critical solution, offering a vital means to address freshwater shortages by converting saltwater into a usable resource. Desalination technologies generally fall into two main categories: thermal distillation and membrane-based processes. − Reverse osmosis is the dominant membrane-based technology used for seawater desalination ,,− representing about 80% of the world’s desalination plants. It is also a mainstay for ultrapure water preparation and industrial water reuse. , However, the reliability and efficiency of these filtration systems are dependent on robust, state-of-the-art monitoring techniques, which will transform how our water future is managed and secured. − In this work, we propose a stimulated Raman scattering (SRS) methodology for rapid, real-time, and in situ monitoring of bench-scale reverse osmosis (RO) membrane desalination flow cells that represents an important advancement in fouling detection.

The newest RO systems offer relatively low energy consumption, high efficiency, and straightforward operation. Despite these advantages, the accumulation of contaminants on the membrane surface, referred to as membrane fouling, remains a significant challenge in many practical applications. , Fouling is generally categorized into three types: organic, inorganic (scaling), and biofouling. ,,

Significant effort has focused on seawater desalination, where organic fouling presents the largest challenge. Brackish water sources such as estuaries and groundwater have received far less attention despite the need for their use in clean water generation, especially in regions far from coastal areas. RO filtration of these brackish sources is particularly impacted by crystal scaling from calcium sulfate, calcium carbonate, barium sulfate, and silica. ,, Scaling reduces operating efficiency, increases energy consumption, and decreases the lifetime of the RO membranes. , Furthermore, scaled membranes require chemical cleaning tailored to the composition of foulants.

Currently, industrial desalination plants use pressure, flow, and conductivity metrics to assess the state and efficiency of their RO operations. While such measurements enable an estimation of membrane fouling, they do not provide early detection, imaging capabilities, or chemical identification. Consequently, more effective measurement of RO membrane fouling under realistic operating conditions remains an important challenge. Ultrasonic time-domain reflectometry and optical coherence tomography can monitor fouling and are minimally invasive, but do not provide chemical information. While electrical impedance spectroscopy shows promise in identifying types of scaling and can provide sensitive detection, the technique requires electrode installation inside the RO cell, which can be problematic. Transmission stimulated Raman spectroscopy can provide detailed chemical information but cannot be performed in situ. , Thus, there exists a need for a measurement methodology that can provide early stage fouling detection with chemical specificity to quantify fouling growth dynamicswithout requiring significant modification to the RO process.

Raman spectroscopy is a good candidate for addressing this problem. Spontaneous Raman scattering is a process in which a single laser illuminates a sample and is inelastically scattered by molecules. The scattered light loses energy to a vibrational (or rotational) mode of the molecule, shifting its frequency. The resulting Raman spectrum provides chemical information and identification of the scattering molecule. Unfortunately, spontaneous Raman scattering is weak; only ∼one in a million photons will undergo inelastic scattering. , To enhance the process and increase data acquisition speed, stimulated Raman scattering (SRS) uses two lasers separated in energy by the Raman shift of a target molecule. The higher frequency pump beam excites the molecule and the second beam, known as the Stokes, stimulates the emission of a photon. The strong and highly directional stimulated emission, which is automatically phase-matched in the forward direction, greatly increases signal strength and acquisition speed as compared to spontaneous Raman spectroscopy. Taking advantage of this speed, SRS can be used as a label-free imaging technique for inorganic and organic molecules. , Its chemical specificity makes SRS an attractive technique for identifying foulants on membrane substrates. Previous work using SRS for membrane fouling detection reported on batch measurements of the organics found in seawater. Others have used surface enhanced Raman scattering (SERS) to evaluate chemical variation or biofouling on membranes. − Additionally, Raman spectroscopy has been used to image chemical foulants on membranes after desalination experiments have been performed. These important works demonstrate the capability of a Raman-based technique for identifying and removing foulants, but none of these studies provided label-free real-time monitoring. Recent work utilizing spontaneous Raman spectroscopy for real-time, nondestructive measurements of membrane scaling lacked imaging capabilities, instead relying on Raman intensity measured at a single point at a specific time. − This approach makes large-area scans and scaling quantification difficult.

Here, we present the first stimulated Raman scattering (SRS) methodology for rapid, real-time, and in situ monitoring of bench-scale reverse osmosis (RO) membrane desalination flow cells, offering a significant advancement in characterization of membrane scaling combining imaging and chemical identification. In particular, we demonstrate this novel technique for inorganic crystal growth on a commercial RO membrane using stimulated Raman scattering microscopy. Unlike other techniques, our methodology provides for real-time membrane scaling imaging in a reflection mode, alongside Raman spectral data. This allows for the direct measurements of the scaled area and calculation of the approximate scalant volume using the three-dimensional, integrated Raman intensity. The methodology has the potential to provide unique insight into multicomponent membrane scaling dynamics. Moreover, such real-time physical and chemical characterization capability, if adapted for large-scale operation, could enable improved process efficiency with better optimized cleaning strategies, thus reducing maintenance and operational costs.

2. Materials and Methods

2.1. Stimulated Raman Spectroscopy and Microscope Design

The stimulated Raman system consists of an SRS spectrometer and laser scanning microscope. The SRS spectrometer, shown in Figure (a), uses an optical parametric oscillator (OPO) (Coherent Chameleon Compact OPO VIS) with a repetition rate of 80 MHz generating 200 fs pulses. The output is used for both the pump and Stokes beams. A Ti/sapphire mode-locked laser at 810 nm pumps the OPO, which has a signal tunable from 1100 to 1500 nm. To reach the Raman wavenumber range required for the scalants, a second harmonic generation (SHG) module frequency doubled the OPO signal output, producing light ranging from 550 to 750 nm. Since the pump wavelength is tunable, the fixed wavelength Stokes is modulated for a Raman loss measurement. Ti:sapphire light from a pick-off mirror before the OPO cavity is used as the Stokes beam. The maximum average power output of the pump was 70 mW and the Stokes was 460 mW.

1.

Diagram of the SRS microscopy system. (a) The SRS spectrometer system is used to obtain the Raman spectra needed for chemical identification of membrane scalants. (b) The laser scanning microscope portion of the system is used to produce three-dimensional images of the scalants on the surface of the RO membrane that is contained within the flow cell.

Spectral focusing is used to improve the Raman spectral resolution. , The instantaneous spectral bandwidth is narrowed by chirping the pulse. − To chirp the pulses, both beams are propagated through 1.2 m of dense flint glass (CDGM Glass, H-ZF52GT, n = 1.825, group velocity dispersion GVD = 228.14 fs²mm^–1 at 780 nm), stretching the pulses from 200 fs² to ∼3.3 ps (see the Supporting Information (SI) Figure S1). An acousto-optic modulator (AOM) (Optoelectronic AA.MT.80-B30A1-IR) modulates the Stokes beam with a 4 MHz sinusoid. A motorized translation stage (Newport XPS-RLD) controls the temporal overlap of the two pulses. Power control in both arms is provided with half-wave plates (HWP) and polarizing beam splitters (PBS). To maximize the numerical aperture (NA) of the imaging objective, both beams are expanded with telescopes to a beam diameter of 4 mm, limited by the scan lens entrance pupil. A dichroic mirror (Thorlabs DMLP805) combines the beams in a collinear geometry.

The copropagating beams are sent through a custom-built laser scanning microscope (LSM) (Figure (b)). At the start both beams travel through a PBS and quarter-wave plate (QWP). The LSM scan optics consist of galvanometer mirrors (Thorlabs GVS002), a scan lens (Thorlabs SL502P2), and a tube lens (Thorlabs TTL 200MP). The microscope objective is chosen for the imaging application. When samples are measured in ambient conditions, under a coverslip with Si oil for thermal protection, a commercial 20× objective (Olympus UPlanSAPO, 0.75NA) is used. For real-time in situ membrane scaling experiments, a custom objective (Figure S4) is employed. The custom objective is necessary to eliminate chromatic aberration between the pump and Stokes wavelengths and to focus light through the flow cell’s 1.65 mm fused quartz window and 4.2 mm depth of feed solution. The objective consists of eight off-the-shelf lenses with a NA of ∼0.45, field of view of 700 μm × 700 μm, and a working distance of 10.65 mm. Course axial positioning of the objective is controlled by a motorized stage (Thorlabs PLSZ) and fine positioning for z-stack acquisition is performed by a piezo stage (PI P-721.17). A manual micrometer XY stage is used to provide transverse movement of the membrane flow cell transverse movement of the membrane relative to the objective.

2.2. Raman Data Collection Procedures

To achieve in situ imaging in the RO filtration system, a reflection (epi) modality is used, a technique where the excitation and the emitted light travel through the same objective lens. This geometry is in distinct contrast to the transmission operation that is more common in SRS microscopy (SRM). ,− To measure a spectrum in this geometry, the laser scanning microscope first situates the beams on a region of interest. Short-pass and band-pass filters (Thorlabs DMSP805, Semrock FF01–775/SP) remove excess Stokes light before the Raman loss light is detected with a balanced photodiode (Si, Thorlabs PDB450A-AC). The spectral focusing employed in our system allows the targeted Raman shift to be tuned over a range of ∼150 cm^–1 using only pump and Stokes beam delay. To achieve the highest spectral Raman resolution via spectral focusing in our system, we ensure the pump and Stokes pulses have near equal chirp values, confirmed through frequency-resolved optical gating (FROG) measurements (Figure S2). A pick-off from the pump beam before the microscope is used to provide common-mode noise rejection. The balanced photodiode signal is sent through high-and low-pass filters (1 and 4 MHz cutoff, respectively) before amplification by a preamplifier (Stanford Research Systems SR445A). A lock-in amplifier (Stanford Research Systems SR844) with a typical integration time of 30 μs per point demodulates the SRS loss signal, and a computer records the x and y lock-in components via a DAQ (National Instruments PCIe-6351). The unbalanced photodiode monitor output is also recorded, creating an image of the scattered light from the membrane.

2.3. Flow Cell

Using this SRS setup, we investigate the scaling of RO membrane coupons (Sterlitech TriSep ACM4, polyamide thin-film composite) in a custom-built pressurized RO cross-flow cell. The flow cell consists of two machined aluminum halves. The top side includes an integrated fused quartz optical window (1.65 mm thick) as well as an inlet (feed) and outlet (retentate) port. An O-ring beneath the glass window seals the fluid in the chamber. The bottom half holds a stainless steel metal mesh to support the RO membrane and a permeate port below the mesh. The distance from the optical window to the membrane is 4.2 mm, owing to the need to accommodate the custom objective as noted above. The cell is designed to withstand operating pressures of 1.24 MPa (180 psi). Detailed views of the machined cell are provided in Figure S5.

2.4. Cross-Flow RO System

Experiments were conducted to both scale and clean the membranes in the pressurized flow cell. A pressure head pump and rotary valve pump generate the desired pressure. The solution from the reservoir enters the flow cell on the side of the optical window through the feed port. A bypass valve upstream controls the flow rate by diverting water to the reservoir, bypassing the cell. A back pressure regulator downstream is used to restrict flow and maintain pressure in the system.

In addition to analog gauges, continuous digital logging is performed during the experiments. The pressure on the membrane and the pressure drop through the cell are monitored with digital pressure meters in both upstream and downstream locations. Flow rate through the cell is measured with a digital flow meter, and the permeate flow rate is quantified with a sensitive flow meter. Initially, the downstream pressure and upstream flow rate are set to 1 MPa (145 psi) and 25L h^–1. Solution temperature is maintained at 25 °C by a liquid chiller in the feed reservoir. The full flow system is shown in Figure S6.

2.5. System Cleaning Methods

Before each experiment, the system is cleaned by passing a 50:50 deionized (DI) and hydrogen peroxide (3%) solution from the reservoir through the flow system. The cell and tubing are then flushed with DI water.

2.6. Real-Time Imaging Procedures

With the cell and flow system fully assembled, the pumps and chiller are turned on and the salt solution begins to flow through the system. Immediately following the initiation of solution flow, the RO cell is positioned on the optical table. The cell’s observation window is centered above the microscope objective, and the assembly is leveled and secured prior to data acquisition. To obtain imaging results, the cell is moved in the transverse direction with respect to the laser beam until one or more crystals can be captured in the objective’s field of view. The travel distance is 1000 μm in each direction, limited by the opening diameter of the window. The crystal, or crystals, are centered in the field of view and the membrane surface is axially located. For each Raman shift, z-stack images are collected with an axial step size of 7 μm. The full z-stack of images are taken beginning at the membrane surface in intervals of 7 μm up to a maximum height of 105 μm above the membrane surface. These images capture both the lateral and vertical growth of the crystals in the field of view. The acquisition of each z-stack takes 300 to 600 s, and sets are taken at time intervals such that a full three-dimensional image of the growth as a function of time could be compiled.

2.7. Solution Preparation

Solutions are prepared by mixing single or multiple salts into deionized water in a large container at desired chemical concentrations for each experiment. The four experiments are performed with one or a combination of CaSO₄H₂O (Sigma-Aldrich C3771), CaCl₂ (Sigma-Aldrich 383147), and NaHCO₃ (Mallinckrodt Chemicals 7412–12). The solutions are mixed overnight using an overhead impeller at a rate of 600 rpm. Additionally, a magnetic stir bar is used to mix the solution while the flow is engaged during data collection.

2.8. Membrane Preparation

TriSep ACM4 membrane coupons are cut to size from a large flat sheet (as-received condition). The dimensions are 6.5 × 11 cm, such that the edges are sandwiched between the two o-rings of the cell. Each coupon is soaked in a 50:50 solution of deionized water and isopropyl alcohol for a minimum of 1 day and is rinsed before being placed in the cell. Since compaction does not affect the ability of the methodology to image and identify scalants, experiments were conducted without an initial membrane compaction phase for the sake of simplicity and time efficiency.

2.9. Experimental Plan

The objective of the experimental plan was to confirm the potential of our SRS methodology to provide unique real-time physical and chemical measurements of membrane scaling. Based on this focused research objective, we utilized a set of real-time and post-mortem experiments. Since previous work had demonstrated that scale removal could be successfully monitored using spontaneous Raman, a DI water cleaning step was added at the end of one of the experiments to confirm that SRS also had this capability.

Given the novel features of this SRS methodology, the scaling experiments were designed to be as operationally simple as possible. Specifically, the duration of the real-time experiments was minimized by eliminating an initial compaction step and using high salt concentrations in the feed solution. Thus, the experiments were conducted to provide proof-of-concept rather than a comprehensive study of membrane scaling.

3. Results and Discussion

3.1. Stimulated Raman Spectra for CaSO₄ and CaCO₃

To first confirm the accuracy of SRS, we used membrane coupons from a previous experiment that utilized spontaneous Raman used for simultaneous detection of CaSO₄ and CaCO₃. Spectral measurements of the scaled membranes were used to identify the chemical scalants, for which gypsum is the most common polymorph of CaSO₄ and calcite is the most common polymorph of CaCO₃. Comparison of the SRS and spontaneous Raman spectra of CaSO₄, CaCO₃, and the RO membrane are shown in Figure . Spontaneous Raman measurements were obtained with a Renishaw InViaRaman Microscope as an independent verification of our SRS results. A peak at 1015 cm^–1 is observed in both the spontaneous and the stimulated Raman spectra, which corresponds to the anticipated peak for CaSO₄. , Additionally, the three peaks at 1075, 1117 and 1150 cm^–1 are consistent with the Raman spectra of the membrane itself. The SRS spectra are consistent with the spontaneous spectra, with the SRS providing a spectral full-width at half-maximum (FWHM) of 7 cm^–1, limited by the degree of spectral focusing, and the spontaneous providing a FWHM of 4 cm^–1 for the CaSO₄ and CaCO₃ peaks. The experimental SRS spectra confirmed that the CaSO₄ formations were gypsum (1008 cm^–1) rather than anhydrite (1017 cm^–1). Distinguishing CaCO₃ polymorphs by their primary peaks, calcite (1087 cm^–1) and aragonite (1083 cm^–1), was not possible with the system’s current spectral resolution. However, doing so by the secondary peaks (716 and 701 cm^–1, respectively) would be possible in future work by expanding the range of the SRS spectrometer.

2.

Spontaneous and stimulated Raman scattering spectra of CaCO₃, CaSO₄, and the reverse osmosis membrane (polyamide thin-film composite). The stimulated Raman spectra peaks possess a full-width at half-maximum (FWHM) of 7 cm^–1 and the spontaneous a FWHM of 4 cm^–1. These are sufficient spectral resolutions to distinguish the CaCO₃ peak from the nearby membrane peaks.

3.2. Ambient 3D Imaging and Chemical Identification of CaSO₄ and CaCO₃

We next evaluated the potential of SRS to provide 3D imaging. Using a 20X objective, these previously tested membranes that were scaled with CaSO₄ and CaCO₃ were measured under ambient conditions. With the higher NA objective, z-stacks were collected with a finer z-step size of 2 μm. As shown in Figure (a), large CaSO₄ rosettes and smaller CaCO₃ crystals can be clearly identified in the z-projection. In addition to the “top-down” view, the individual slices can also be evaluated. Figure (b) shows slices from the z-stack over a depth of 80 μm. Previous spontaneous Raman studies of membrane scalants had difficulty distinguishing CaCO₃ from the membrane due to the close location of the spectral peaks. In contrast, in this work, optical sectioning provides a solution. When the beams are focused above the membrane surfaceeven just microns awaythe membrane does not contribute to the Raman signal, thus isolating the CaCO₃ crystals.

3.

3D imaging results. (a) Stimulated Raman microscopy (SRM) z-projection of axial slices showing a large CaSO₄ rosette in yellow along with smaller, spherical CaCO₃ crystals in blue. (b) Z-stack layers of (a). Each layer is a z-projection of 10 axial slices collected at a 2 μm spacing.

3.3. Real-Time In-Situ Monitoring of CaSO₄ Scaling

Three real-time in situ experiments were performed to observe CaSO₄ scaling (Figure ). Experiments 1 and 2 were prepared with a concentration of 1.8 g L^–1 of CaSO₄ H₂O. Experiment 3 used a supersaturated solution of 3.6 g L^–1 of CaSO₄ H₂O, chosen to obtain more rapid and uniform scaling over the field of view. Discussion of solution chemistry, saturation, and permeation can be found in prior work. ,, At this higher concentration, crystal morphology is changed and much of the scaling can be attributed to bulk deposition on the membrane rather than crystal growth. − Given the random nature of early stage scaling, imaging was performed at different locations on the membrane until a growing crystal was observed, at which point a time-series of z-stacks was collected. Microscope imaging parameters are listed in Table and include the field of view, the acquisition time to image one CaSO₄ stack, the total test duration, and the imaging duration. During the experiments, a decrease in the permeate flow rate was also observed due to membrane compaction and progression of scaling.

4.

Time series images of CaSO₄ scaling on a reverse osmosis membrane in a benchtop flow cell. Imaging data for three separate experiments are shown. Coupons from experiments 1 and 2 were saved and dried for postexperiment imaging and the membrane in experiment 3 was cleaned at the end of the scaling process. Experiment 3 at 227 min shows the results of the membrane cleaning procedure.

1. Imaging Parameters, Including the Field of View Observed, the Time to Acquire one CaSO₄ z-Stack, Background, and Monitor Image, the Total Duration of the Filtration Campaign, and the Imaging Duration .

experiment	field of view (μm × μm)	acquisition rate (/min)	experiment duration (min)	imaging duration (min)
1	500 × 500	16	210	180
2 (0–25 min)	500 × 500	9	90	60
2 (25–214 min)	650 × 650	14	180	180
3	500 × 500	9	240	270

^aImaging began when a growing crystal site was observed on the membrane. The field of view was increased in experiment 2 in order to continue capturing the crystal growth since it exceeded the field of view beginning at 25 min.

Images collected from the SRS system (Figure ) clearly capture crystal growth in two distinct cases: experiments 1 and 2 used solution concentrations consistent with prior work whereas experiment 3 used a super saturated solution. Selected images as a function of time of the three experiments are shown in the figure. Images from experiments 1 and 2 reveal the rosette structure that is characteristic of CaSO₄ scaling. Experiment 1 contains a set of four rosettes growing into the field of view over time, with overlap occurring near the end of the time series. Experiment 2 shows an isolated rosette structure nearly centered in the field of view. Here, we note that the field of view was increased during experiment 2 after 25 min to capture more of the crystal growth. We also draw attention to the fact that the imaged crystal in experiment 2 formed on a ‘ridge’ in the membrane stemming from the periodic pattern of the metal support mesh. This resulted in a slight blurring on one side of the image due to this part of the crystal being out of the focal plane while the rest of the crystal remained in focus (Figure S7). Experiment 3 shows a more uniform scaling morphology consisting of small needle-shaped platelets of CaSO₄ crystals, rather than the distinct rosettes seen in experiments 1 and 2. Scale removal from the cleaning step at the end of experiment 3 results in the absence of crystals from the bottom right panel of Figure .

3.4. Post-Mortem Confocal Microscopy

After the SRS measurements were complete, the membrane coupons were removed from the flow cell and set to dry for post-mortem imaging.With precise location control, confocal laser microscopy (Keyence VK-X 3D Surface Profiler) images were obtained from the same crystals as analyzed via SRS. As indicated in Figure , there is a strong correspondence in the rosette details between the SRS and confocal images.

5.

Confocal laser scanning microscope CaSO₄ images. Images (a–c) are of experiment 1 (Figure ) and (d–f) are of experiment 2. (a) and (d) are the final SRS images acquired in the real-time measurements. (b) and (e) are laser confocal microscopy images of the same regions of interest post-mortem in ambient conditions. (c) and (f) are the height maps produced from the laser confocal microscope profiler. These height profiles can be used to segment the membrane and CaSO₄, which is used for the analysis in Figure .

3.5. Calcium Sulfate Growth Rates

The real-time in situ volumetric data provide insight regarding the rate and extent of membrane scaling. From the z-stack SRM images, the scalant surface area can be calculated by summing the pixels above a threshold intensity value and multiplying this value by the pixel size (∼0.5 μm). The threshold is chosen using background SRM images acquired at each time point, which target a Raman shift away from CaSO₄ (∼1050 cm^–1). The results of the calculation are shown in Figure (a), where the data indicate that the surface area increases as a function of time for both experiments 1 and 2. Experiment 3 showed evidence of a similar monotonic increase, but reflects the expected decrease in area due to the cleaning step initiated at 3.5 h.

6.

Surface area and integrated volumetric Raman intensity for each experiment. (a) calculates the total surface area covered by scaling in the field of view, from a top-down projection. (b) Due to complexities of calibrating Stimulated Raman volumetric measurements, the Raman intensity, integrated over every slice of the z-stacks, is reported. The cleaning procedure employed in experiment 3 is highlighted with a dotted box.

To validate our surface area measurements, we analyzed experiments 1 and 2 using a laser scanning surface profiler (VK-X3000) after their removal from the flow cell. We obtained confocal images and corresponding height maps of the same crystal structures depicted in Figure . The height maps, shown in Figure (c) and (f), enabled us to segment the membrane from the crystals, allowing for precise measurement of the scaled surface area on the membrane. For experiment 1, the surface roughness-measured (SRS) area and confocal measurements showed good agreement with a difference of ∼14% (Table ). Although the SRS slightly overestimated the area, this discrepancy likely stems from image noise.

2. Comparison of the Measurement of Scaled Surface Area for Experiments 1 and 2 .

experiment	SRS surface area (×105 μm2)	confocal area (×105 μm2)
1	2.14	1.88
2	3.06	3.35

^aFrom the stimulated Raman scattering (SRS) microscopy (SRM) images, area is calculated as the total area covered by CaSO₄ in a z-projection. With the confocal data (Figure ), the height map is used to segment the crystals and membrane, at which point the area is calculated. Surface area values for both SRS and confocal are affected by imaging angle and curvature of the membrane (Figure ).

While results from Experiment 2 indicated a lower area value for the SRS as compared to the confocal measurement, they differed by only ∼9%. The lower SRS value prompted further investigation using the confocal image and height map. The confocal image revealed portions of the rosette structure near the crystal’s edge that were not visible in the SRS image. Examination of the height map provided crucial insight: the membrane surface itself was curved. Details of the degree of curvature are shown in Figure S7 which confirms the presence of an approximately 75 μm deep trough. This curvature is an artifact of the RO membrane deforming into the grooves of the supporting metal mesh when under pressure, leading to a periodic surface undulation. This tilting of the crystal plane increases its total axial extent, making it challenging to capture the entire crystal without manual intervention or specialized software for error detection. Specifically, a section of the experiment 2 crystal was absent in the SRS image (lower right corner), which the height map confirmed by showing a corresponding trough in that region, thus explaining its absence in the SRS data.

Despite the aforementioned curvature in experiment 2, the confocal and SRM images are in reasonable agreement. All confocal imaging was done in ambient conditions. Additional imaging with the profiler was performed for membrane coupons under a small layer of Si oil (index of refraction n = 1.41, close to seawater), capped by a coverslip. Adding silicon oil reduces the index contrast of the crystals, rendering confocal and wide-field methods an ineffective replacement for SRS for real-time imaging.

In addition to evaluating the surface area, we investigated measuring the scalant volume. This required our SRS system to generate signal at different heights, essentially performing optical sectioning, which requires appropriate axial resolution. , SRS is well-suited for these measurements as its nonlinear nature provides inherent optical sectioning. − However, achieving high-quality z resolution (∼1.3 μm for 0.75 NA) critically depends on a sufficiently high NA objective. The inability to differentiate in-focus and out-of-focus signals then compromises the accuracy of axially dependent measurements, such as crystal volume. While our custom 0.45 NA objective was adequate for quantifying the scaled surface area, a higher NA objective is critical to accurately quantify scalant volume with our methodology. Additionally, all SRS-based volume and concentration measurements require rigorous calibration alongside a high NA objective. , Given our objective’s limitations, we instead use integrated Raman intensity as a proxy for scalant volume. As shown in Figure (b), the trends of the integrated SRS intensity for each experiment qualitatively aligned with the surface area results. Notably, experiment 3 data clearly showed the effect of the cleaning cycle when DI water replaced the saline solution.

The inherent variability in RO membrane scaling, particularly in the early stages, present a significant challenge for generalizing specific growth-rate values. However, the confocal microscopy employed as a reference for the SRS methodology clearly demonstrates the potential and robustness of our measurement approach.

3.6. Implications of This Work

This work demonstrates the potential of the SRS methodology for real-time, in situ physical and chemical measurement of scalant growth on the surface of an RO membrane. This could be important for future drug development and personalized medicine. Efficient monitoring systems are critical for increasing efficiency in membrane-based desalination. ,,, Additionally, as new membrane technologies emerge, − a monitoring system that can simultaneously distinguish the membrane chemical composition in addition to that of the foulants may be particularly impactful. The SRS methodology highlights the potential for such monitoring.

The primary obstacle to a direct, in situ application of SRS is the geometry of commercial RO systems. These systems overwhelmingly rely on spiral-wound modules, in which the membranes are multilayered and encased in a composite shell. This complex, three-dimensional structure prevents the direct access required for an external, high-resolution optical probe like the SRS microscope. To overcome this geometric limitation, we propose using flat membrane surrogates, often referred to in the industry as ‘canary cells.’

As well described in Sim et al., a “canary cell” permits monitoring the fouling process in an RO membrane module via measurements on a side stream that employs sensors for in situ, real-time and nondestructive detection. In addition to being representative of a spiral wound RO module in terms of materials, spatial dimensions, and hydraulics, a canary cell is characterized by its accuracy and reproducibility under conditions that can simulate realistic operation at different locations in a spiral wound module.

Since the flat geometry allows for easy spectroscopic access, the SRS method could be applied to these surrogate cells to quantify the extent and chemical nature of the scaling that has developed, thereby providing a high-resolution diagnostic map of the scaling severity at desired locations. Correlating the scaling results from these surrogates with their position in the RO train could enable effective mapping of spatial variation and total extent of scaling.

SRS sensing is a particularly valuable tool that addresses the need for real-time, noninvasive chemical monitoring across multiple fields. For instance, SRS microscopy will transform future drug development and personalized medicine by providing noninvasive, real-time pharmacokinetic (PK) analysis in vivo. The critical advantage is that the signal is linearly proportional to concentration, enabling quantitative chemical imaging that is vital for accurate PK modeling, predicting therapeutic efficacy, and optimizing treatment dosage. Another key application is the monitoring of continuous flow reactors, where chemical ingredients are continuously pumped through a system to react and generate a finished product. Here, SRS provides immediate, highly detailed feedback on intermediate species, reactant consumption, and product formation rates without the need for physically sampling or halting the flow.

Given such capabilities, the adaptation of the SRS methodology to a suitable canary cell could provide improved understanding regarding the complex mechanisms that drive scaling under realistic operating conditions, moving beyond traditional ex-situ analyses that often fail to capture the transient nature of crystal nucleation and growth. Future studies can achieve this goal by employing lower concentration feed solutions to slow the crystal formation process and limit the imaging volume to the membrane surface to increase the temporal resolution of the measurements. Importantly, optical contrast is often too weak to detect nanometer-scale nucleation sites, whereas the strong vibrational signal of SRS allows for the detection of deposits based solely on their chemical identity. The direct translation of these insights into practical, scalable solutions for real-world desalination challenges, including the long-term potential of this approach in demanding industrial environments, will require further development to improve and quantify the precision and accuracy of the volumetric measurements. By providing volumetric data, SRS offers a more comprehensive understanding of scale morphology and its impact on membrane performance as well as the dynamics of crystal growth. Furthermore, future studies can leverage the chemical specificity afforded by SRS offers the potential for the identification and quantification of multiple feed solution components, limited only by possible spectral overlap, which can be mitigated with a-priori information about the contaminants. In addition to the detection of inorganic scalant species, measurement of organic foulants by SRS is also possible.

Beyond volumetric insights, we have also developed a complementary method for estimating the membrane area occupied by scalant deposition as a function of time. This additional metric can provide a critical link between the three-dimensional growth observed and the two-dimensional impact on membrane flux and overall system performance. The ability to acquire these diverse data sets in situ and in real-time also provides potential for future work in developing more effective antiscaling strategies, optimizing cleaning protocols, and designing more resilient RO membrane systems. This integrated approach not only enhances the capacity to study crystal growth dynamics in a relevant operating environment but also provides powerful tools that have the potential for improving operational control and predictive modeling in industrial desalination processes.

In summary, the SRS methodology described in this work offers the possibility to advance the study of membrane scaling as well as sensing needs in other applications, such as fouling in microfluidic systems. Our integrated approach not only enhances the capacity to study crystal growth dynamics in a relevant operating environment, but it also provides power tools that have the potential for improving operational control and predictive modeling in industrial desalination processes. Indeed, the extension of the methodology to more complex scenarios that include simultaneous scaling, organic, and biological fouling represents a next logical step.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.5c10405.

• Descriptions, schematics and data plots of our bandwidth and pulse characterization, methods and data for our Raman spectra calibration, calculations for our custom objective design, images of the pressurized cell used in experiments, and a description and schematic of our cross-flow system (PDF)

Y.L.S., J.M.A., and M.Z: Methodology, investigation, visualization, writing original draft, writing review and editing. Y.L.S.: Software, formal analysis. V.M.B, A.R.G, and J.T.G: Conceptualization, resources, writing review and editing. V.M.B., A.R.G, and J.T.G: Funding acquisition.

The authors declare no competing financial interest.