1. Introduction

Liquid thin films on a surface are important for a myriad of industrial processes over a wide range of applications including tribology, thermal management, energy-conversion, water desalination, and fire-suppression. In all cases, optimization of the application requires characterization of the liquid film thickness, temperature, and/or chemical composition. A variety of optical diagnostics techniques have been reported including optical interference [1, 2], laser absorption [3–6], and laser-induced fluorescence of added tracers (LIF) [7–10]. The use of absorption to monitor liquid water thin films is especially attractive because of the overlap of overtone (2ν₁ and 2ν₃) and combination (ν₁ + ν₃) bands of liquid water with the wavelength range of telecommunications lasers and fibers in the near-infrared (NIR, 1.3–1.7 μm). An additional advantage compared with LIF techniques is that no tracers need to be added to the fluid.

Our previous work on NIR absorption sensing of liquid water films successfully investigated quasi-point transmission measurements of films supported on a transparent substrate. Four-color absorption sensing was used for measurements of liquid-film thickness and temperature for pure water [6, 11], and liquid-film thickness and solute concentration for aqueous solutions of NaCl [12] and urea [13]. More recently, we successfully demonstrated pointwise time-multiplexed two-color, retro-reflection absorption measurements of water films on non-transparent surfaces [14]. This single-ended configuration is more practical, as in most practical cases, liquid films are supported on opaque materials that can only be optically accessed from one side.

In the work reported here, we extend the sensor from quasi-point measurements with resolution limited by the diameter of the laser beam, to 2D image detection using a fast-framing infrared camera. Prior approaches towards water-film imaging included transmission measurements of flowing water constrained in a channel with broadband illumination using narrow-band filters to time-multiplex absorption measurements at three specific wavelength ranges near 2 μm from the output of a halogen lamp to determine temperature images from hot-wire heating [15]. In an experiment by Dupont et al. [16], back-reflected NIR from a tungsten lamp was detected with an InGaAs focal plan array (FPA) camera to image the thickness of falling water films.

In the work reported here, we significantly increase the precision and time-resolution of the absorption measurements using a pair of time-multiplexed distributed feedback (DFB) diode lasers and a fast framing NIR camera. One of the lasers was tuned to a relatively strong liquid water absorption feature and the other one was tuned to a nearby region with relatively weak liquid-water absorption. Beer’s law was then used to determine two unknowns: the absorbance of the liquid thin film and the light transmission efficiency. The laser wavelengths were chosen to generate absorbances that well match the range of water film thickness of interest (stronger features in the 2 µm range could monitor thinner films, but the surface tension of liquid water in most practical cases produces rather thick free-standing films in the millimeter range).

Measuring the thickness of liquid films supported on opaque surfaces via NIR absorption provides several challenges. When light reflected (or scattered) from the substrate is used for detecting the film thickness, it is required to accurately determine the path length through a given film and to suppress interference with spurious reflection such as reflection from the outer (liquid/air) surface. In the direct or specular reflection configuration that generates the strongest possible signal, the collection solid angle must be optimized to avoid unwanted interference from partial front-surface reflection of the incident laser beam, as this light will not be absorbed by the liquid thin film and thus can contribute a strong background signal previously noted and/or corrected [16, 17] but not suppressed. This problem can be circumvented by placing the film on a retro-reflecting surface that is illuminated at a non-normal angle. The back-reflected light that twice traverses the liquid film that is now geometrically separated from the front-surface reflection can then be collected with an off-axis parabolic mirror mounted close to the sending optics [14].

Since retroreflective substrates may be of limited practical use, here we investigate two different alignment configurations: Collection of light after either specular or diffuse reflection from the substrate of the target liquid thin film. In the specular reflection strategy we suppress the interfering surface reflection by exploiting the linear polarization of the laser where orienting the incident and detection optical path at Brewster’s angle avoids interference from front surface reflections. This configuration would – if the polarization plane of the incident light is parallel to the plane of incidence – completely eliminate the unwanted partial surface reflections from contributing to the signal. A downside is the strong deviation of the incoming and reflected light from the normal direction, which requires spatial separation of light source and receiving detector. The diffuse scattering configuration provides a more convenient undistorted view perpendicular to the film surface with sending and receiving optics near each other. Diffuse reflection, however, is weaker and wavy surfaces could lead to local interference that would need to be removed in image post processing.

In the present work, the optimization of the detection optics and camera (saturation, dynamic range, and spatial resolution) is also considered. The paper concludes with a discussion of potential improvements in the experimental design that could enhance the sensor performance for simultaneous measurements of multiple thin film variables (thickness, temperature, chemical composition).

2. Sensor setup

The optical setup of the two-wavelength laser absorption imaging sensor in reflection-mode is shown in Fig. 1 for two different alignment configurations of the imaging camera.

 

Fig. 1 Optical setup for a time-division multiplexed (TDM) thin-film absorption imaging sensor in reflection mode utilizing the specular reflected light from a polished substrate (a), or the diffusely back-scattered light from a matte-finish surface (b). Both cases only differ in the arrangement of the detector. FC: Fiber combiner, PM-SMF: Polarization maintaining single-mode fiber, A: Attenuator, C: Collimator, P: Polarizer, T: Telescope optics. An enlarged view of rays entering and exiting the probe volume is given below each setup. The quartz window is optional and used here to generate variable film thicknesses for calibration measurements.

Download Full Size | PDF

The light from two DFB diode lasers operating at 1353 and 1469 nm was combined and coupled into a polarization-maintaining single-mode fiber (Panda, PM-SMF). The two lasers were operated in a time-multiplexed mode; one laser was pulsed on for 1 ms and off for 4 ms then the other laser pulsed on/off and the laser pulse rate synchronized with the camera’s 200 Hz frame rate. The total laser power was varied for each alignment with fiber-based optical attenuators (Thorlabs FA10T and Thorlabs NENIRXXB) (A in Fig. 1) to adjust the illumination intensity and provide sufficient signal to the camera while avoiding saturation of the pixel array.

The light exiting the fiber was collimated, passed through a linear polarizer oriented to transmit light parallel to the plane of incidence, and then expanded by a telescope before impinging onto a 30 × 20 mm² elliptical area of the liquid film. Two types of substrates were used: aluminum plates with a polished finish for strong specular reflection and aluminum substrates with a matte-finish for diffuse reflection. The polished finish was prepared by mechanical polish of a smooth machined surface, and the matte finish was prepared by sand blasting with a ± 10 μm surface finish. For calibration measurements, the film was captured between the aluminum plate and a polished quartz plate (window) supported on spacers and clamped to a known thickness. Dynamic imaging of liquid film thickness was demonstrated for films with and without the cover plate. Note, that in principle all surfaces that scatter enough light towards the detector are suitable for this method. Besides aluminum this includes other practically relevant metals like nickel, titanium (spray cooling of turbines) and stainless steel (selective catalytic reduction in exhaust pipes).

Two different light collection alignments were used. Figure 1(a) shows the camera aligned near the specular reflection for measurements using polished aluminum substrates. Here, front surface reflections were minimized by aligning the incident beam with its polarization parallel to the plane of incidence to near Brewster’s angle for the quartz/film surface (55.4°). Such front-surface reflections do not pass through the liquid film and generate interfering signal. In measurements with the plane of laser polarization perpendicular to the plane of incidence, the front-surface reflection was suppressed by at least a factor of 10. The laser light reflected from the substrate (having passed through the liquid film twice) was collected onto the camera (Xenics Cheetah-640CL, LOT Quantum Design). Figure 1(b) shows an alternative detector arrangement used for measurements with diffuse surface scattering. Here the camera is aligned normal to the film in the top-view position above the substrate, thus avoiding collection of light from unwanted specular reflection.

2.1 Camera performance

The NIR camera has a 640 × 512 InGaAs pixel array sensitive to light from 0.9 to 1.7 μm. A 1 ms camera integration time was used and synchronized with the laser-on period. Light was collected f/1.4 with an f = 25 mm camera lens (Kowa LM25HC). The laser-illuminated portion of the thin film was imaged onto the central ~160 × 220 pixels. This limit is determined by the frame-grabbing speed of the computer. The camera was read out with a 12-bit digitizer, providing a dynamic range of 0–4096 counts/pixel. The image intensity was adjusted using the laser attenuators for each experimental configuration to be a maximum on any pixel <3900 counts. The minimum image intensity after absorption was >900 counts. The background intensity or the camera dark signal (camera counts without laser irradiation, ~600 counts) was measured before each experiment and subtracted. The frame-to-frame intensity fluctuation was ~3% for the background intensity and ~0.5% for a background-corrected laser intensity of 3300 counts.

The camera timing was synchronized with the single laser pulses resulting in a 200 Hz frame rate; a pair of images (one for each laser wavelength) is needed for liquid film thickness evaluation (see below), which results in a true measurement rate of 100 Hz. This measurement rate was limited in these experiments by the frame-grabbing speed of the computer via a camera link readout and not by the maximum camera framing rate (800 Hz, according to the specification sheet). The measurements used a camera integration time of 1 ms, and thus, with an appropriate computer the demonstration measurements reported here could have a 400 Hz measurement rate (at full pixel resolution) with the same signal-to-noise; we estimate that with a reduced pixel resolution the same camera with such a faster computer could be used with similar signal-to-noise at kHz measurement rates.

The camera’s spatial resolution was measured using a NBS 1963A resolution test target (Thorlabs) and calculated from a contrast transfer function (CTF), defined as:

2.2 Spectral database

A database of temperature-dependent spectral absorption coefficients was previously measured using a temperature-stabilized quartz cuvette (Hellma, 1 mm optical path length) and a FTIR spectrometer (Bruker, Vertex 80, spectral resolution 2 cm^–1) [12]. Absorption coefficients for liquid water in the temperature range 298–338 K and the wavelength range between 5500 and 8000 cm^–1 is shown in Fig. 2. The right y-axis shows the variation of the absorption coefficient with temperature. Note that for 6944 cm^–1 (1440 nm) there is a relatively large absorption coefficient and for 7570 cm^–1 (1321 nm) there is a relatively small absorption coefficient. In addition, the values of the absorption coefficients at these wavelength positions are independent of temperature. If lasers at these wavelengths had been used, the measurements would be independent of temperature [15]. Unfortunately, lasers at these wavelengths were not available for the demonstration measurements, and lasers at nearby wavelengths were used instead at 1469 nm (6807 cm^–1) for strong absorption and 1353 nm (7390 cm^–1) for weak absorption. For the 40 K change in temperature shown in Fig. 2, there is a 14% variation of the two absorption coefficients; for the measurements reported here, we estimate the film temperature was constant to better than 3 K resulting in less than a 2% error in evaluated film thickness.

 

Fig. 2 NIR absorption coefficient $k_{T_{\text{L}}}^{\lambda }$ of pure water for temperatures ranging from 298 to 338 K measured by FTIR in a heated cell with a 1 mm pathlength. Note the absorption coefficient is independent of temperature at 6944 and 7570 cm^–1 (1440 and 1321 nm). Also shown is the variation of the absorption cross-section with temperature (right y-axis)

Download Full Size | PDF

2.3 Measurement technique

The water-film thickness was determined from measuring the optical absorption with two distributed feedback (DFB) diode lasers using the Beer-Lambert law (Eq. (2)):

Finally, we must consider the orientation of the laser beam and the camera viewing angle with respect to the plane of the liquid film. The optical pathlength is directly related to the film thickness δ by:

3. Results

3.1 Imaging of known film thicknesses

To validate the sensor’s performance, liquid samples were confined in a cell of known, homogenous film thickness (achieved with spacers of 100–300 μm) at room temperature for polished and matte-finish film support substrates; the recovered 2D images were uniform to ± 5% and the expected water film thickness was recovered within ± 8%. The power of 2D imaging, however, is better demonstrated with structured thin films. An example measurement using the specular reflection alignment of a water film captured between a cover plate and a polished substrate is shown in Fig. 3(a) where the “structure” was produced by clamping the cell (composed of substrate, cover plate and spacers) with different height spacers at the top and bottom (260 and 100 µm, respectively). This produces a film whose thickness varies linearly from bottom to top of Fig. 3(a), and the trace on the right shows the near linear variation in film thickness between the two spacers. The trace on the top of Fig. 3(a) shows the roughly 200 μm thickness expected for a trace halfway between the top and bottom of the cell. Because the cell is round and clamped with three screws, when different thickness spacers are used to separate top/bottom, the clamping screws do not tighten properly, the cell leaks and there is some right-to-left structure in the cover-plate/substrate. However, the cell limitations are a feature in that the resulting film is structured in thickness that is captured in the image in Fig. 3(a).

 

Fig. 3 2D film-thickness images (color-coded thickness values) and profile plots along two orthogonal axes across the image (red and blue lines). Image areas not processed are colored in black. (a) Specular reflection mode (Fig. 1(a)) for a polished aluminum support surface; (b) orthogonal reflection mode (Fig. 1(b)) for a diffuse or matte-finish aluminum support surface. The white arrow in the left figure indicates the direction of incidence of the laser beams onto the targets.

Download Full Size | PDF

In Fig. 3(b), a matte-finish substrate is used with the camera positioned above the substrate. Here, the cover plate and substrate are separated by 300 μm spacers. Two “steps” in the film thickness are produced by placing a 150 μm spacer on the substrate about one quarter of the distance from the top of Fig. 3(b) and a 200 μm spacer about a quarter of the distance from the bottom of Fig. 3(b). Recall that the cell is round and thus the available spacers are curved as seen by the curved structure seen in the measured water film thickness image. The image captures the expected reduced thickness of the water film at these steps where the space between substrate and cover plate is reduced by the spacer. The spacers are bare stainless-steel pieces and the applied ratio method is rather immune with respect to the surface finish of the individual objects within the camera field of view. Note that the spacer in the lower half of Fig. 3(b) can be distorted as it is not clamped between support plate and cover glass plate resulting in variable range of liquid thickness between the cover plate and spacer. The uncertainty of the liquid film thickness measurement calculated by error propagation based on the parameters in Eq. (4) was about 8% for measurements with the polished substrate surface (configuration (a) in Fig. 1 and 8.9% for the measurements with the diffuse substrate surface (configuration (b) in Fig. 1). In the latter case, the additional uncertainty in the path lengths of diffusely back-scattered light reaching the camera must be considered (cf. sketch in Fig. 1(b)). This error results from light that is not reflected under β = 0°. The maximum angle of light the camera can capture is 18° with respect to the normal of the film surface, leading to an additional error of 0.9% in film thickness. The difference in the incident angle between film vs no film results in a spatial shift of maximum 15 pixels corresponding to 220 µm. This error is smaller than the current achievable spatial resolution of the camera system of 313 µm.

3.2 Time-resolved film thickness imaging (captured film)

Experiments of time-varying water films captured between substrate and a transparent cover plate were also conducted. The measurement cell with a matte-finish substrate was realized with two of the 200 μm spacers, and the uniform water film is captured in the image in Fig. 4(a). After starting the measurement at some later time, the water film was purged with an air flow from a syringe needle causing water to be expelled from the cell as captured in the sequence of images shown in Fig. 4(b) (t + 100 ms), Fig. 4(c) (t + 500 ms) and Fig. 4(d) (t + 700 ms). These images are frames from a movie recorded at a 100 frames/s rate, and each image in Fig. 4 is the result of a single 1 ms camera integration period. The displacement of water between the substrate and the cover plate is clearly demonstrated when the film is purged by the air flow; after 700 ms of purging Fig. 4(d) shows almost all water disappeared except for some remaining small droplets or puddles still attached to the surface of either the substrate or cover plate.

 

Fig. 4 Water-film thickness images between substrate and cover plate with 200 µm spacers. Each frame is from a 100 frame/second movie with 1 ms camera integration. (a) Water film between plates, (b) after 100 ms of air purge, water begins to flow out of the cell, (c) after 500 ms: more water expelled, (d) after 700 ms: nearly all the water displaced from the cell. The white arrow in the upper left figure indicates the direction of incidence of the laser beams onto the targets.

Download Full Size | PDF

3.3 Time-resolved film thickness imaging (free film)

Free-standing films of liquid water on a substrate have a thickness on the order of several millimeters because of the strong surface tension of water. For the wavelengths used for this demonstration such films are optically thick for the light at 6807 cm^–1 (1469 nm), and to image free-standing pure water films a laser with a lower absorption coefficient is required. For this purpose, the 6807 cm^–1 laser was replaced with a 7183 cm^–1 (1392 nm) laser, thereby reducing the absorption coefficient roughly 3 times compared to the former. The surface tension was also reduced by preparing a 50% by mass ethanol/water mixture. This produces an ~1 mm thick free-standing film. Ethanol does not absorb at the two wavelengths used for this demonstration, and thus the absorption coefficient of the mixture is roughly reduced proportionally to the water mixture fraction [18].

The dynamics of a flowing free-standing film of the liquid water/ethanol mixture captured by the imaging system are illustrated in the four frames from a 100 Hz movie in Fig. 5. The liquid impinges onto the matte-finish substrate with a syringe from the left side of Fig. 5; note the substrate is tilted downward from left to right so that gravity will contribute to the liquid film flow. Figure 5(a) shows the nearly dry substrate before injection; Fig. 5(b) is taken 370 ms after start of injection (aSOI) where the tip of the liquid is just entering the field of view (FOV) of the illuminated substrate. Figure 5(c) taken 1200 ms aSOI shows the liquid flow crossing the lower half of the substrate, while at 3230 ms aSOI, (Fig. 5(d)) a well-developed liquid flow approximately 1.5 mm thick nearly fills the width of the FOV.

 

Fig. 5 Frames with 1 ms camera integration time from a 100 Hz movie showing a 50% water/ethanol (by mass) mixture injected by a syringe (out of the image to the left) onto an inclined (gravity assisted flow from left to right) matte-finish substrate. (a) nearly dry substrate before liquid impingement, (b)−(d) liquid flow 370, 1200, and 3230 ms after impingement, respectively. The white arrow in the upper left figure indicates the direction of incidence of the laser beams onto the targets.

Download Full Size | PDF

3.4 Time-resolved film thickness imaging (droplet impingement)

The absorption imaging method to capture waves flowing across liquid thin films is demonstrated in Fig. 6. Here, the matte-finish substrate is covered with an approximately 1 mm thick film of the water/ethanol mixture and a syringe is used to have a droplet of the mixture dripped onto the film. Figure 6(a) depicts the instant where the droplet approaches the uniform film surface. The shadow of the liquid drop appears as a void in the liquid film (the incident laser light is inclined and nearly no laser light is transmitted by the droplet making a shadow on the film). Figure 6(b) is a frame of the 100 Hz movie taken 40 ms later where outgoing spherical waves are clearly observed in the film produced by the droplet impingement.

 

Fig. 6 Two frames of 1 ms camera integration time from a 100 Hz movie of a liquid drop of the water/ethanol mixture impinging onto a matte-finish substrate covered with a ~1-mm thick liquid film of the same mixture. (a) Shadow of a liquid droplet in the air above the surface of the film; (b) outgoing spherical waves after the droplet hit the surface. The white arrow in the left figure indicates the direction of incidence of the laser beams onto the targets.

Download Full Size | PDF

4. Conclusions

The imaging of thin liquid water films at constant temperature was demonstrated using optical absorption of light from two time-multiplexed NIR diode lasers (1469 and 1353 nm) detected with a fast frame rate InGaAs focal plane array camera. Thickness values were derived from absorbance ratios at the two wavelengths for each pixel. The films were supported on opaque substrates with two optical device schemes of illumination and detection. Precise measurements (uncertainty less than 9%) of the thickness of pure water films were demonstrated for known film thicknesses in the 100–300 µm range by capturing the liquid between a quartz glass cover plate mechanically separated by spacers from the substrate surface.

Movies of liquid film thickness taken at 100 Hz demonstrate the ability of this scheme to capture the dynamics of flowing films. Three different examples were reported: The air-purging of a liquid film captured between substrate and cover plate, the flow of a free-standing liquid film, and wave motion of a disturbed free-standing liquid film after droplet impingement.

These experiments demonstrate the applicability of NIR absorption imaging for a wide range of liquid thin-film applications. Following our previous work [13, 19] multiple unknowns can be determined by increasing the number of multiplexed lasers (e.g., thin-film thickness, temperature, and solute concentration). Other work [15, 20] has demonstrated that suitable choice of wavelengths (e.g., temperature insensitive isosbestic points) in the water spectrum can help making data interpretation more robust against varying environmental parameters (e.g., temperature).

Most important for practical application of this absorption-based characterization of liquid thin films is the capability of imaging the diffusely backscattered light from a substrate which does not require special surface finish. Significant signal for these imaging experiments with a 1 ms integration time was readily achieved from a film supported on a bare metal substrate. This suggests excellent potential for practical applications.