2020 Volume 37 Pages 42-63
Optical Coherence Tomography (OCT) is a light-based imaging method capable of simultaneously capturing the internal structure and motion (1D, 2D or 3D) of various opaque and turbid materials with a micron-level spatial resolution. Depending on the OCT technology, axial scanning rates can vary in a range of tens to hundreds of kHz. The actual imaging depth significantly depends on the optical properties of the material and can vary from micrometers to a few millimeters. From the viewpoint of industrial applications, OCT technology is very appealing. Due to its resolution, speed, and ability to deal with opaque materials, it fills an apparent gap in available measurement methods. Nonetheless, OCT has not to date seen widespread growth in the industrial field. This has been at least partly due to a lack of commercial devices compact and flexible enough to adapt to industrial needs. The recent emergence of more generic commercial OCT devices has considerably lowered the threshold for adapting the technique. The utilization of OCT for structural analysis, also outside the medical field, has been thoroughly discussed in scientific literature. Therefore, in this paper, we will mainly concentrate on applications of OCT that also utilize its capability of performing velocity measurements. The emphasis will be on industrially motivated problems such as rheology, microfluidics, fouling and turbulence.
Research, process development and operation as well as process quality control depend greatly upon our ability to collect information with measurements, through which engineers hope to get answers to the questions “what and why?”. The enormous developments that have taken place in the process industry during the last twenty years have been mainly related to improvements in accuracy, reliability, and speed of measurement techniques. The ultimate goal of process engineers is the ability to perform the measurements in situ, online and in real time. However, due to a lack of existing online techniques, one often has to settle for offline measurements.
There are a number of measurement technologies and approaches available today, each one with their respective advantages. Optical measurement systems, such as an interferometer, optical displacement meter, moiré inspection, stereoscopy, and holography, are some of them with several advantages. Firstly, they are noninvasive and nondestructive methods which afford flexibility in setting up measuring systems. Secondly, one is not limited to pointwise measurements, as full-field measurements are possible. This feature is essentially inherent, e.g. in CCD imaging. For many applications, the number one benefit of optical measurement methods is their capability for high-speed sampling. Optical sensors also have the advantage of high precision in comparison, e.g. with ultrasonic, electromagnetic, and electrostatic sensors. They are widely used in industry, e.g. for end-product quality, material surface, topography, defect detection, 3D dimension measurement, and color inspection.
Optical Coherence Tomography (OCT) is a rather new member of the family of optical sensors (Drexler and Fujimoto, 2008; Popescu et al., 2011). OCT is capable of measuring the internal structure of various semi-transparent materials with a micron-level spatial resolution, as illustrated in Fig. 1a. Depending on the OCT technology, axial scanning rates can vary in a range of tens to hundreds of kHz. The actual imaging depth significantly depends on the optical properties of the material and can vary from micrometers to a few millimeters. The OCT technique was invented in the late 1980s. It mainly evolved in the medical field, especially in ophthalmology in the 1990s. Later, it attracted attention for application in other fields as well. Applications of OCT—also outside of the biomedical field—have been reviewed in (Wiesauer et al., 2005a; Stifter, 2007; Song and Harding, 2012; Kawasaki, 2013; Yoshizawa, 2015).
a) 3D structural OCT image of microfibrillated cellulose (MFC) suspension in a pipe with a diameter of 8.6 mm. Image taken with permission from Haavisto et al. (2015a). b) Marangoni effect (spontaneous flow due to evaporation of alcohol) in a droplet of wine as imaged with Doppler OCT (unpublished).
While online sensors are still scarce, the applicability of OCT for structural analysis for industrial purposes has been demonstrated in numerous cases. These include crystallization processes in polymers (Hierzenberger, 2014), liquid sorption in paper (Fabritius, 2007), thermal degradation of epoxy resin composites (Awaja et al., 2009), automotive coatings (Lawman et al., 2017a), nanocomposites (Schneider et al., 2016), polymer-fiber composites (Stifter et al., 2008), quality control of film coating (Koller et al., 2011; Nemeth et al., 2012; Markl et al., 2014a; Markl et al., 2015a; Markl et al., 2015b), and LCD-display production (Shirazi et al., 2016; Wijesinghe et al., 2017). An important functional extension of OCT, namely polarization-sensitive OCT (Boer et al., 2017), has been used, e.g. for the analysis of stress fields of various materials (Stifter et al., 2003; Wiesauer et al., 2005b; Wiesauer et al., 2007; Heise et al., 2010; Stifter et al., 2010). Markl et al., (2014b), discuss general aspects related to the inline quality control of moving objects with OCT.
In addition to structural analysis, the capability of measuring flow velocities is often vital for industrial purposes such as process development and process control. While there are robust indicators for measuring average volumetric flows, e.g. in pipes, the availability of instruments for measuring local velocities is more limited. Various techniques that can provide information on velocity fields are presented, e.g. in Chaouki et al. (1997) and Powell (2008). If the medium is transparent and there is optical access, one can apply, e.g. PIV or LDA techniques. For turbid media, however, only Ultrasound Velocity Profiling (UVP), has gained broader use in industry (Takeda, 2012). Other methods, such as Nuclear Magnetic Resonance Imaging (MRI) and various tomographic methods are widely used on a lab scale, but their industrial use is still very limited. Another option for velocity field measurements is Doppler OCT (Drexler and Fujimoto, 2008). DOCT combines the Doppler principle with OCT to obtain velocity information simultaneously with structural imaging (see Fig. 1b). It was first used in ophthalmology for the quantitative imaging of blood flow in vessels in vivo. Recently, the use of DOCT has finally spread outside of biomedicine, but it is still in its infancy for industrial applications.
In principle, from the viewpoint of industrial applications, OCT technology should be very appealing. Due to its resolution, speed, and ability to deal with opaque materials, it fills an apparent gap in the available measurement methods. Nonetheless, OCT has not to date seen large growth in the industrial field. This has been at least partly due to a lack of commercial devices compact and flexible enough to adapt to industrial needs. The recent emergence of more generic commercial OCT devices such as Telesto SD-OCT Systems by Thorlabs have somewhat decreased this bottleneck. While they cannot beat the performance of state-of-the-art laboratory-built devices, the commercial OCT devices have lowered the threshold for adapting the technique considerably.
As discussed above, the utilization of OCT for structural analysis, also outside the medical field, has been thoroughly reviewed in scientific literature. Therefore, in this paper, we will mainly concentrate on OCT applications that utilize its capability of performing velocity measurements. The emphasis will be on industrially motivated problems such as rheology, microfluidics, fouling, and turbulence. We hope this paper will encourage more people to adopt this emerging field of measurement technology for solving additional R&D problems.
OCT employs low-coherence spectroscopy and interferometry to determine the optical properties of the sample as a depth-dependent reflectivity profile (Drexler and Fujimoto 2008; Popescu et al., 2011; Leitgeb et al., 2014; Hong, 2018). The short coherence length of the employed broad bandwidth light source is utilized for axial sectioning. 2D slice images (B scans) or 3D volumes can be constructed by employing scanners to acquire a series of axial scans (A scans) at different lateral locations.
In the first realization of OCT, i.e. time domain OCT (TD-OCT), the light from a low-coherence source is split into two paths by a coupler, directing it along two different arms of an interferometer (see Fig. 2). One arm is designated as the reference arm, while the other is the sample arm. When the light exits the fiber end of either arm, it is controlled by various optical components (mirrors, lenses, etc.) to control specific beam parameters such as shape, depth of focus, and the intensity distribution of the light. In the reference arm, the light is back-reflected by a reference mirror, and it returns into the interference system, propagating along the same path it came from, only in the opposite direction. The same process applies for the light in the sample arm, the only difference being that the beam is backscattered by the sample. In an inhomogeneous sample, different structures within the sample will have different indices of refraction, and light will be backscattered when it encounters an interface between materials with different refractive indices. The returning light from both arms recombines at the coupler generating an interference pattern, which is recorded by the detector. For a particular position of the reference mirror, the light propagating in the reference arm travels a certain optical distance and forms the corresponding interference pattern only with light that traveled the same optical distance along the sample arm, including the portion of the distance traveled inside the sample. Therefore, when the reference mirror is translated along the propagation direction of light for different positions of the mirror, the returning reference generates interference patterns with light backscattered from corresponding depths within the sample. In this way, the dependence on the depth of the intensity of light backscattered from beneath the sample surface can be measured.
In time domain OCT (TD-OCT), the reference mirror moves during the measurement, and at a given moment the signal is obtained from a unique position. The different interfaces within the sample lead to a peak in the degree of interference represented as a peak in the signal intensity graph. In Fourier domain OCT (FD-OCT), the reference mirror is fixed, and the detector arm of the interferometer uses a spectrometer instead of a single detector. The spectrometer measures spectral modulations produced by interference between the sample and reference reflections.
In Fourier (or spectral) domain OCT (FD-OCT), the light echoes are collected simultaneously from all axial depths and are detected as modulations in the source spectrum with all the spectral components captured simultaneously (see Fig. 2). The technique is called Fourier domain OCT because the interference pattern is recorded in the spectral domain for each lateral position, and the final information is reconstructed via Fourier transform. The main difference between these two technologies is that the reference arm in an FD-OCT system has a static mirror instead of a moving one as in TD-OCT. This feature eliminates the moving mirror and the limitations imposed by the inertia of that mechanical device. Due to the elimination of the mechanical translation, FD-OCT systems are capable of higher data acquisition speeds than TD-OCT systems.
Due to the high imaging speed of OCT, structural 2D/3D imaging can be used for obtaining velocity information through a PIV-type analysis to determine two/three velocity components (Shen et al., 2011; Chen et al., 2013; Mujat et al., 2013; Zhou et al., 2016). It is also possible to obtain velocity information from the speckle noise of the structural data with autocorrelation analysis (Pretto et al., 2016). When the velocity information is retrieved simultaneously with structural imaging, the method is referred to as Doppler OCT (DOCT), or OCT velocimetry. Conventional OCT is based on measurement of the amplitude of the backscattered light. In DOCT, depth localized phase shift is also measured.
The early approaches of measuring the OCT phase shift were based on time domain techniques (Wang et al., 1995; Chen et al., 1997a; Chen et al., 1997b; Proskurin et al., 2003a). Interference fringes are then detected if the sample arm distance in an interferometer matches the reference arm length within the temporal coherence length. This defines the axial resolution, which is inversely related to the spectral bandwidth of the light source. A simple method to detect the Doppler-shifted time domain OCT signal is to record the full fringe signal and calculate the local frequency shift within a small window that is slid across an A-scan. This approach is, however, computationally very expensive, and cannot be used for real-time flow measurements. In addition, the spatial resolution is inversely dependent on the frequency resolution of the Doppler system. An alternative method uses phase-sensitive Doppler analysis, offering the advantage of decoupling of the spatial and frequency resolution. In this case, the phase changes are measured between two consecutive A-scans.
Nowadays, Fourier domain DOCT has largely replaced time domain DOCT due to its much higher imaging speed and better sensitivity (Liu and Chen, 2013; Leitgeb et al., 2014). In Fourier domain DOCT, the phase changes are measured between two A-scans, and the full depth structure is encoded in parallel by the recorded spectral interference pattern, and reference arm scanning is not needed. There are two variants of Fourier domain DOCT: Spectrometer-based Fourier domain DOCT (FD-DOCT) and swept-source DOCT (SS-DOCT). In the first case, the spectral interference pattern is recorded with a spectrometer in a parallel way, whereas in the second, the interferogram is acquired as a function of time using a wavelength-tuning source at the interferometer entrance (Hendargo et al., 2011; Leitgeb et al., 2014). The SS-DOCT method has certain advantages in the near infrared region above 1 μm, where detector arrays are expensive (Lexer et al., 1997). In addition, SS-DOCT systems may improve acquisition speed. The introduction of Fourier domain mode-locked (FDML) swept sources made A-scan rates of several hundreds of kHz available for the first time (Huber et al., 2006, 2007). A recent review of SS-OCT can be found in Alibhai et al. (2018).
In a DOCT measurement, the measured target velocity in a y-directional A-scan is
| (1) |
where f is the scanning rate, Δϕ(y) is the measured phase shift at position y, λ is the central wavelength, n is the medium refractive index (n = 1 for air), which is included due to the light velocity depending on it, and α is the angle between the measurement beam and the velocity vector (see Fig. 3a). The values of the phase shift are restricted between −π and π. The theoretical maximum measurable velocity is thus
a) The DOCT measurement gives the component v of the velocity vector U in the direction of the measurement beam. b) The DOCT measurement arrangement for a pipe flow. The average flow direction is parallel to the pipe axis. c) A closer look at the measurement optics. Here U is the real velocity, v is the component of velocity in the direction of the measurement beam, v’ is the velocity value given by the DOCT device, and nf is the refractive index of the medium. The angle α is obtained by applying Snell’s law of refraction to the two interfaces.
| (2) |
The maximum measurable velocity can be increased by decreasing the measurement angle α between the DOCT beam and the velocity vector. The relative error due to the angle is then in proportion to Δ tan
Due to inherent noise (and possible velocity fluctuations) a series of phase-shift measurements gives a histogram of phase shifts (see Fig. 4a and 4b). The width of the histogram has been analyzed in detail in Szkulmowska et al. (2008). If the phase shifts exceed the boundaries they are wrapped by −2π (see Fig. 4c). The situation is similar to data sampling where frequencies exceeding the Nyquist frequency (half of the sampling frequency) are aliased among the measured low-frequency data.
Measured histograms of phase shifts (gray). The solid lines are Gaussian fits to the histograms. a) Static sample. b) Moving sample. c) Sample is moving close to the theoretical maximum velocity. The phase-shift histogram is wrapped.
The phase shift Δϕ used in Eq. (1) is usually a temporal average of instantaneous phase shifts to decrease the signal-to-noise ratio, resulting in a higher sensitivity to velocity. The average can be calculated either by averaging directly over the phase-shift histogram or (preferably) by fitting a Gaussian distribution to it. Before averaging, an unwrapping procedure can be applied to correct the histogram aliasing at large phase shifts (Ghiglia and Pritt, 1998; Werkmeister et al., 2012). This makes it possible to measure velocities that exceed the theoretical maximal velocity given by Eq. (2). Phase unwrapping is an essential process for accurate phase profile reconstruction. Various methods for improving the accuracy of OCT phase wrapping have been discussed in Xia et al. (2017).
Just like Doppler ultrasound methods, in its basic form, DOCT is only capable of measuring a single velocity component of the flow velocity vector in the beam direction. Two or three velocity components can naturally be obtained by using dual/triple-beam methods (Werkmeister et al., 2012; Trasischker et al., 2013; Kumar et al., 2014; Leitgeb et al., 2014). The absolute value of the velocity vector can be determined by monitoring the bandwidth broadening of the Doppler spectrum (Ren et al., 2002; Piao et al., 2003; Proskurin et al., 2003a), combining spectral and time domain methods (Grulkowski et al., 2010), autocorrelation function of the backscattered light (Wang and Wang, 2010; Weiss et al., 2013), or characterization of the time scale of random fluctuations of the dynamic scattering component (Srinivasan et al., 2012). Moreover, all three velocity components can be obtained by introducing a variable scan bias in the OCT system (Huang and Choma, 2014).
The methods for obtaining more than just the beam directional component of the flow velocity described above are often technically challenging or require investment in several DOCT devices. Fortunately, the main direction of the flow is known a priori in many industrially relevant geometries. In such cases, the absolute values of the velocity vectors can be obtained quantitatively. Indeed, one of the advantages of the OCT when compared, e.g. to ultrasound technologies, is the possibility to measure the measurement angle α directly from 2D or 3D structural images (Michaely et al., 2007; You et al., 2014; Haavisto et al., 2015b; You et al., 2017).
Fig. 3b shows as an example a measurement setup of pipe flow. From Fig. 3c, we get the real axial velocity in the pipe to be
| (3) |
where β is the DOCT measurement angle (the angle between the camera axis and pipe surface normal), v’ is the velocity value given by the DOCT device, and nf is the refractive index of the monitored medium. Similar reasoning also works, e.g., for plate-plate and bob and cup rheometer geometries.
The resolution and scanning speed of state-of-the-art lab devices are currently below one micrometer and hundreds of kilohertz, respectively. The imaging depth in optimal conditions can vary a lot depending on the technology used. Typically, it is a couple of millimeters. The performance of available commercial devices is naturally more modest. In our work, e.g. we usually used a commercial FD-DOCT (Telesto I, Thorlabs, Inc.). Fig. 14 shows a set-up where this device was combined with a rheometer. The central wavelength λ of the device is 1,325 nm, and the A-scan rate can be either 5.5, 28 or 91 kHz. The theoretical maximum velocity Eq. (2) in beam direction (α = 0), measurable with Telesto I, is thus 3.1 cm/s. As an example, in our pipe flow studies, the measurement angle β varied typically between 2°–5° (see Fig. 3). With these measurement angles, the maximum measurable velocity Umax varies in the direction of flow, e.g. for water (nf = 1.33) between 0.2–0.5 m/s. The axial resolution of the device is approximately 5 μm in air and thus ca. 3.7 μm in water; the lateral resolution is 7–20 μm, depending on the optics used. The maximum pixel amount of 512 per A-scan provides, in optimal conditions, a maximum imaging depth of 1.9 mm in water.
Fig. 5 shows a laminar velocity profile of water in a pipe with a diameter of 8.6 mm measured with Telesto I SD-OCT (the exact wall position was determined by fitting a second-order polynomial to the mean velocity profile at the wall). Typically for velocity profiling, sequential axial A-scans were recorded over a certain period in a single physical position. The result is compared, in Fig. 5, with a corresponding theoretical profile calculated from the flow rate measurement. The measurement angle β (see Fig. 3) was determined directly from a B-scan structural image composed of 4,096 A-scans in a spatial range of 5.00 mm, and there was no fitting between the two data sets. The agreement between the two data sets is excellent—DOCT can obviously be used for quantitative velocity measurements out-of-the-box without the need for calibration (see also Haavisto et al., 2017).
Laminar velocity profile of water in a pipe measured with Telesto I SD-OCT (solid line). The dashed line indicates the theoretical laminar profile calculated from the flow rate measurement. The mean velocity profile is based on an ensemble of 50,000 A scans. Image taken with permission from Haavisto et al. (2015b).
The real measurement depth depends on the material properties. If the medium scatters too much, the beam weakens quickly when it travels in the medium, and the intensity of backscattering light is low. On the other hand, if the scattering level is too low, the amount of observed backscattered light also drops quickly when the distance from the probe increases. For transparent materials such as pure water, one needs to add extra light-scattering tracer particles. For this purpose, we used a coffee creamer, as it disperses easily in water and the size of the fat droplets is typically 0.2–1 μm (Kirk and Othmer, 2008) and they follow the flow faithfully. An intralipid solution can also be used (Lauri et al., 2011a). The amount of creamer added (ca. 0.5–1 %) was adjusted to balance between the DOCT signal intensity and its penetration depth. In Haavisto et al. (2017) the measurement depth of Telesto I was about 1.2 mm for 1 % creamer and 0.01 % xanthan gum water solutions, and 0.4 % softwood and 0.4 % microfibrillated cellulose (MFC) suspensions. Obviously, the maximum measurable velocity can limit the measurement depth considerably.
Fig. 6 shows an example of measured DOCT velocity profiles together with the corresponding DOCT amplitude signals for three consistencies of MFC in a pipe flow. From Fig. 6, we see that the penetration of the amplitude signal decreases rapidly with increasing MFC consistency. We also see that while the increasing consistency decreases the measurement depth of the velocity profile, the deterioration is not as dramatic as for the amplitude signal. Note the rapid decrease of the amplitude signal when approaching the wall. This is probably due to a combination of device optics and a real depletion layer created at the wall.
Examples of measured DOCT velocity profiles (dashed lines) in a pipe flow of MFC (consistency 0.4 %—blue, 1 %—red, 1.6 %—green) together with corresponding DOCT amplitude signals (solid lines). The DOCT amplification was tuned for each consistency separately to optimize the quality of the velocity signal. The measurement was performed with Telesto I SD-OCT (Haavisto, 2018).
The most popular target for DOCT measurements by far has been the flow of blood in real and artificial blood vessels (Bonesi et al., 2010; Drexler et al., 2014; Leitgeb et al., 2014; You et al., 2014; Carlo et al., 2015; Kim et al., 2015; Chen et al., 2017; Kashani et al., 2017; Spaide et al., 2018). Other biomedical applications of DOCT have been, e.g. cilia-driven flow (Jonas et al., 2011; Huang and Choma, 2015) and temporal deformation of tissues (Lawman et al., 2017b). The development of OCT technology is driven by medical applications, and new modifications and algorithms are regularly published. These works often also have high relevance outside biomedicine—in addition to developing the methods, they give advice on how experiments should be performed and data analyzed, and give a good general picture of the possibilities OCT offers. However, in this paper, we will concentrate on DOCT applications that are related to industrial flows. Here, the most important applications have so far been microfluidics, fouling, and rheology. Moreover, a few studies on turbulent flows can also be found.
Microfluidics is the science and technology of systems that process or manipulate small amounts of fluids using channels with dimensions of tens to hundreds of micrometers. In addition to using continuous flow, microfluidic devices can also operate with droplets. The use of microfluidics offers a number of useful capabilities: Small size of samples, high sensitivity, low cost, and short analysis time. Among the most important applications of microfluidic technologies is analysis. The goal is the development of a lab-on-a-chip for various applications. Here, all the functionalities of an entire laboratory are integrated within a single microfluidic chip (Whitesides, 2006). Microfluidic devices have also proven promising in certain types of manufacturing processes (Schoenitz et al., 2015). Droplet-based microfluidics (Shang et al., 2017), e.g. can be used for synthesizing microcapsules, microparticles, and microfibers applicable, e.g. to pharmaceuticals, cosmetics, and foods.
Accurate control of microfluidic processes requires resolving flow dynamics with micron-scale spatial resolution. Flow characterization in microfluidic channels and lab-on-a-chip devices is important not only for optimizing the efficiency and design of devices but also for validating models of computational fluid dynamics (CFD). Although several techniques exist for both the direct and indirect measurement of fluid velocity fields, there are only a few measurement methods available that are non-invasive, give both structural and velocity information, are able to give 1D, 2D or even 3D data, have micron-scale spatial resolution, are applicable to opaque heterogeneous fluids, and are able to provide reliable data very close to solid boundaries. DOCT is currently the only measurement method which fulfills all of these requirements (Ahn et al., 2008; Huang and Choma, 2015).
The usefulness of DOCT velocity measurements has been demonstrated for pressure-driven flows in microchannels of various sizes and geometries (Proskurin et al., 2003b; Bukowska et al., 2013; Lauri et. al., 2015). In Cito et al. (2012), capillary-driven flow, which is used widely in healthcare and diagnostic applications, was studied in a circular microchannel (see Fig. 7). So far, the flow behind the meniscus has been difficult to study experimentally. In that paper, DOCT was used to confirm the presence of a recirculation pattern in a moving meniscus predicted by numerical simulations. Such a measurement would be practically impossible with other measurement methods.
a) Schematic of capillary-driven flow in a microchannel. b) Contours of the y-component of the velocity vector obtained with DOCT of a capillary-driven flow of an aqueous suspension of polystyrene beads in a glass microchannel with a diameter of 200 μm. Notice that the pipe is slightly tilted to eliminate refraction of the DOCT beam on the pipe surface. Image taken with permission from Cito et al. (2012).
The capability of OCT to provide both structural and velocity information makes it very attractive for analyzing the mixing of components. The microfluidic flows are usually laminar, and typically in the Stokes regime (Reynolds number Re << 1). In this regime, fluids mix only via diffusion, which is a rather slow mechanism and makes reactions within microfluidic devices harder to achieve. In Ahn et al. (2008), the effect of secondary flows on mixing was studied with OCT inside a meandering square microchannel micromixer (see Fig. 8). Here, the flow fields and structural patterns of two-liquid mixing were measured, and it was found that the efficiency of two-liquid mixing was rather low. Mixing could be clearly increased when air bubbles were injected into the system. OCT imaging revealed that air bubbles provoked an alternating pair of counter-rotating and toroidal vortices which improved the two-liquid mixing. Those vortices were very similar to vortices created behind a moving meniscus (see Fig. 7). Xi et al. (2004) used OCT to study three representative microfluidic mixers: A Y channel mixer, a 3D serpentine mixer, and a vortex mixer. It was found that light-microscopy images lead to an overestimation of the mixing efficiency, an effect that was eliminated with OCT imaging. Overall, OCT was found to significantly improve the characterization of 3D microfluidic device structure and function.
a) A meandering square microchannel. The cross-sectional dimensions of the channel are 600 μm × 600 μm. b) On the left: Cross-sectional structural OCT image before mixing of water (black) and polystyrene suspension (bright). In the middle and on the right: Cross-sectional mixing patterns after the first loop for two Reynolds numbers. The OCT images were taken at the position shown with the short dashed line in Figure c). The mixing efficiency is seen to be rather low. c) The measured flow velocity in the cross-plane direction on a symmetric plane cutting the loop into two halves. The cross-sectional velocity field shows a pair of counter-rotating vortices indicated in the inset. The OCT image was taken at the dashed line seen at the bottom of the figure. Image taken with permission from Ahn et al. (2008).
Aqueous two-phase systems (ATPS) provide a high-yield and high-purity process for the separation and purification of biomolecules. Since ATPS droplets can be obtained by phase separation between two immiscible aqueous solutions of distinct polymers, monitoring the phase separation is critical to permit evaluation of the performance of an ATPS droplet-based microfluidic system. In Lee et al. (2016), OCT was used successfully to investigate real-time changes of the volumetric droplet morphology under various fluidic and rehydration conditions of an ATPS system. The analysis was performed with 3D OCT images of the droplets (see Fig. 9).
Time-lapse OCT images of a rehydrating droplet. Cross-sectional (top) and volumetric (bottom) images show the process of rehydration. Image taken with permission from Lee et al. (2016).
Manukyan et al. (2013) studied internal flows in a drying sessile polymer dispersion droplet on hydrophilic and hydrophobic surfaces. Trantum et al. (2013) studied the “coffee ring effect” by tracking particle motion and accumulation in evaporating droplets. We have successfully performed similar measurements, as can be seen in Fig. 1b.
Despite its apparent advantages, the number of OCT-related publications is still rather low in microfluidics, but we expect this to change in the near future. This view is supported by the fact that 3D printing is currently emerging strongly in microfluidics (Bhattacharjee et al., 2016). 3D printing enables the production of very complex microfluidic devices. Their development will necessitate the availability of highly detailed experimental data on their microscopic flow conditions, which currently only OCT can provide.
The fouling of plant surfaces and the subsequent cleaning needed is a significant and often also a costly problem in various fields of the process industry (Bott, 1995; Dürr and Thomason, 2010). Filters, membranes and heat exchangers are especially sensitive to it, but the accumulation of deposits can take place on any surface when the conditions are favorable. Fouling is also a problem in microstructured devices (Schoenitz et al., 2015).
Fouling begins at the instant a material comes into contact with a fluid. During this period, the conditioning film forms. The makeup and formation of this conditioning film directly relates to the further development of fouling. Conditioning film development is followed by a rapid accumulation of deposit growth. Next, a pseudo steady-state period takes place when accumulation is almost constant. Finally, the surface can become fouled to the point that it can no longer be used effectively. Fouling is a function controlled by a variety of parameters, including the geometry of the system, surface material, interface temperature, deposit temperature, free stream velocity, and fluid characteristics. While fouling is often a purely physicochemical process, biofilm formation is also a common reason behind it (Kukulka et al., 2004).
At present, the mechanisms of fouling are not yet well understood. Monitoring the development of the conditioning film and the subsequent deposition process is thus very important in order to understand and ultimately control fouling. However, deposit thickness, which is typically only a few tens of micrometers, is challenging to measure in situ. A review of the applications and principles of different methods of visual observation that have been applied to fouling (excluding OCT) can be found in Shirazi et al. (2010).
Just as for microfluidics, OCT offers many advantages for the analysis of fouling. It has a high spatial resolution, and in addition to providing structural information, it also offers information on flow conditions, which is highly important for a deeper understanding of the fouling process. Moreover, as fouling processes are rather slow, the ability of OCT to provide 3D data can be fully utilized. For these reasons, OCT has been used in several works on fouling. We will present a few examples below—a comprehensive review on the use of OCT in biofilm research can be found in Wagner and Horn (2017).
The potential of OCT for fouling analysis is well manifested in Wagner et al. (2010), where the development of biofilms in a funnel was studied with varying flow conditions (see Fig. 10). A heterogeneous structure was detected for the biofilm cultivated in laminar conditions. In turbulent conditions, the biofilm structure was found to be more homogeneous, and the porosity was clearly lower. The authors also noticed that confocal laser scanning microscopy (CLMS) does not necessarily provide an accurate representation of the biofilm structure at the mesoscale. Additionally, the typical characteristic parameters obtained from CLMS image stacks can largely differ from those calculated from OCT images.
OCT-images of biofilms developed on a funnel wall in laminar (a and b) and turbulent (c and d) flow conditions. The dimensions of the 3D images are 4 mm × 4 mm × 1.6 mm. The two red orthogonal slices underline the porous biofilm structure. Images b) and d) represent xz-planes in the middle of the images. Spatial resolution is ca. 20 μm. Image taken with permission from Wagner et al. (2010).
Biofilm rheology was studied in situ in real time in Blauert et al. (2015). The biofilms were grown in a flow cell set-up at low shear stress and their deformation was studied under high shear stress (see Fig. 11). The measurement set-up allowed the calculation of many rheological parameters such as shear modulus, Young’s modulus, and strain; the values obtained were on a similar level to those in other studies. Moreover, viscoelasticity was analyzed in stress–strain experiments and was in good agreement with values reported in the literature.
OCT images of a biofilm in a shear stress experiment. Flow is from left to right. a) The initial structure of the biofilm. b) The structure two seconds after the flow started. The white line represents the initial structure. Scale bar equals 250 mm. The spatial resolution is ca. 7 μm. Image taken with permission from Blauert et al. (2015).
Weiss et al. (2016) studied the localized and simultaneous measurement of biofilm growth and local hydrodynamics in a microfluidic channel (see Fig. 12). In addition to the fouling film, they measured the longitudinal flow velocity component parallel to the imaging beam, and the transverse flow velocity component perpendicular to the imaging beam with the DOCT autocorrelation function. Based on the measured velocities, they were able to calculate the shear rates inside the flow channel. They found a clear relation between the measured biofilm structure and flow conditions as the biofilm growth progressed.
Cross-sectional image of channel morphology and corresponding transverse and longitudinal flow velocities in a (y, z)-plane. Flow direction is in the positive × direction. The first row shows the channel morphology, the second row shows the transverse flow velocity and the third row shows the longitudinal flow velocity. The columns show the data 24 and 48 hours after starting the experiment. Image taken with permission from Weiss et al. (2016).
Membrane technology has been applied in numerous fields such as water/wastewater treatment and seawater desalination due to its capability for producing a vast amount of water of high quality. However, membrane filtration is inevitably associated with membrane fouling. Gao et al. (2013) used OCT to characterize the velocity profiles normal to the membrane surfaces, inside a unit cell of the spacer in a modified membrane filtration module. The orientation of the spacer was varied with respect to the bulk flow direction. A series of DOCT images demonstrated the subtle interactions between the fluid and the spacer filaments; these characterization results were then used to interpret the performance variation of the reverse osmosis process with the same spacer configurations. In Gao et al. (2014), DOCT was applied to the analysis of fouling in a laboratory-scale membrane filtration system. Both the growth of the fouling layer and the velocity profiles of the fluid field were measured as illustrated in Fig. 13. The characterization results revealed for the first time the evolution of the morphology of the cake layer under different microhydrodynamic environments. This study demonstrated that DOCT-based characterization is a powerful tool for investigating the dynamic processes during membrane fouling. More recent membrane fouling studies utilizing OCT can be found in Han et al. (2018) and Park et al. (2018).
DOCT images of fouling (2 g/l bentonite microparticles) in a channel with a spacer. The primary cross-section scanned by DOCT is parallel to the direction of the bulk flow, as indicated by the red area in the upper left schematic. The images obtained in the primary cross-section include the following: a) y-component of the velocity vectors in the channel at t = 0 min, b) the structural image in the channel at t = 0 min, and c) the structural image in the channel at t = 60 min. The fouling layer can be seen clearly on the top of the membrane. The direction of the bulk flow is denoted by the arrows. The membrane surface is indicated by the dashed-dotted curve, and the filament cross-sections are circled with dashed lines. Image taken with permission from Gao et al. (2014).
OCT is the emerging imaging technique of the last decade in fouling. It enables a major step forward in biofilm research as it allows the monitoring of the fouling film structure in real time in situ under operational conditions. As OCT—unlike any competing measurement technique—also gives information on the flow, it facilitates the analysis of the fluid-structure interaction of fouling systems on a micron scale. Furthermore, the OCT data can be combined with numerical modeling by extracting structural templates from the biofilm images to be utilized as boundary conditions to solve Navier-Stokes and mass-transport equations (Li et al., 2016).
The success of a wide range of commercial products and industrial processes depends on meeting specific flow requirements. Architectural and industrial coatings, molded plastics, adhesives, personal care products and cosmetics, inks, cement, drilling muds, ceramic slips, solder pastes, papermaking suspensions, foodstuffs, and medicines are examples of complex fluids whose successful processing and commercial viability depends on having the “right” rheological properties (Eley, 2005).
Rheology is defined as the study of the deformation of matter, primarily in the liquid state, but also as ‘soft solids’ or solids under conditions in which they respond with a plastic flow rather than deforming elastically in response to an applied force. Rheology generally accounts for the behavior of fluids by characterizing relations between stresses, strains and strain rates either by experiments or by theoretical analysis. The experiments are typically performed either in oscillatory or continuous shearing conditions. In the former case, one obtains the viscoelastic properties of the medium, such as dynamic moduli, as a function of frequency and amplitude, and in the latter case, one obtains the viscosity as a function of shear stress or shear rate (Barrat and Hansen, 2003).
The rheology of simple fluids is rather straightforward and well understood from a practical point of view. Their flow behavior can be characterized either by a single temperature-dependent coefficient of viscosity (Newtonian fluids) or by relatively simple relations between the stress and the strain rate (non-Newtonian fluids). Furthermore, these material properties can be directly measured using conventional rheological methods where the interpretation of the measured data is usually based on an assumed flow behavior of the fluid. These simplifying assumptions, such as Couette-type flow and no-slip boundary condition, are reasonable for simple Newtonian and non-Newtonian fluids. Unfortunately, only a small group of fluids exhibit such simple behavior.
The actual observed flow behavior of many practically relevant fluids is very complicated and shows, e.g. hysteresis, thixotropy, shear banding, unstable behavior, and spontaneous formation of a lubrication/depletion layer close to solid boundaries (Barnes, 1995). The mechanisms underlying these complicated and poorly understood phenomena in these complex fluids are related to the presence of a mesoscopic length scale which is caused by the internal structure of the material. This is the case, e.g. for polymer/fiber suspensions for which the consistency, orientation, elongation, and flocculation of the polymers/fibers considerably affect their bulk and boundary layer behavior.
The intricate rheology of complex fluids, coupled with the optical turbidity and measurement difficulties with conventional rheometric techniques, render them challenging to study in conditions relevant to real processing. To date, this has been mainly due to the lack of experimental techniques that would allow the direct measurement of flows and internal structures of complex, opaque fluids especially in the immediate vicinity of the wall (typically, within a few tens of micrometers from the wall). OCT now provides a remedy for this long-standing grievance (Haavisto et al., 2014; Malm, 2015).
Rheo-OCT is a measurement set-up (Fig. 14) where OCT has been combined with a conventional rheological device such as a rotational rheometer (Harvey and Waigh, 2011), a capillary/pipe viscometer (Lauri et al., 2011b), or an elongational rheometer (Dufour et al., 2005). It allows the use of truthful flow behavior based on direct experimental observation instead of the often unrealistic and fallacious assumed flow behavior as the basis for interpreting and analyzing the results from rheometric measurements. Rheo-OCT captures various perturbative effects encountered in rheological experiments such as defective sample loading, wall depletion, slip flow, and shear banding (Harvey and Waigh, 2011; Jaradat et al., 2012; Saarinen et al., 2014; Malm et al., 2014). For examples see Figs. 15, 16 and 17.
a) A set-up of Rheo-OCT—Telesto I SD-OCT combined with a TA instruments rheometer with a plate-plate geometry. The OCT imaging direction can be changed by using a transparent upper/lower plate. b) A schematic of a Rheo-OCT set-up for a concentric-cylinder geometry (bob and cup) with transparent outer geometry.
a) Opaque margarine with wall slip. b) Transparent polyacrylamide particles in water (0.1 % w/w) with shear banding. In both cases, a rotational plate-plate geometry was used. Image taken with permission from Harvey and Waigh (2011).
Different flow regimes for 1 % microfibrillated cellulose suspension in a concentric-cylinder rotational rheometer: a) Complete slip on the stationary wall. b) Slip and rolling flocs on both walls. c) Slip flow and rolling flocs on the moving wall. The gap is 1.0 mm. The average velocity profiles are marked with a white line. Image taken with permission from Haavisto et al. (2015a).
a) Rheogram of an oscillatory experiment for a 0.5 % MFC + 0.11 % CMC suspension in a concentric-cylinder rheometer geometry (gap is 1.0 mm). b) Velocity profile in the gap in the linear viscoelastic (LVE) region (strain 0.78 %). c) Velocity profile in the gap outside the LVE region (strain 12 %). The frequency is 1.0 Hz, and the maximum speeds of the surface of the moving cylinder are thus 0.05 and 0.75 mm/s. In both cases, the velocity profile was measured during the cycle when the absolute value of the shear rate was highest. The stationary wall is at 0.0 mm and the moving wall at 1.0 mm. With the lower strain, the velocity profiles are approximately linear close to the stationary wall, but they are somewhat asymmetric. There is some shear banding and slip flow at the moving wall. With the higher strain, the velocity profiles are approximately symmetric. There is shear banding and strong slip flow on both walls. The velocity profile is linear in the middle of the gap. Notice that the velocity profiles were calculated from successive structural 2D images (Haavisto, 2018).
An important possibility afforded by Rheo-OCT is local rheological measurements, i.e. calculating the local viscosities in the fluid using the formula
| (4) |
where
| (5) |
where L is the pipe length, D is the pipe diameter, ΔP is the pressure loss, and
Viscosity of water measured with a combination of pressure loss measurement and DOCT. The dash-dotted line shows the theoretical viscosity of the water. The dashed line shows the onset point of turbulence, Re ~ 2300. Image taken with permission from Haavisto et al. (2015b).
When the size of the flow geometry exceeds the maximum measuring distance of OCT, the measurements can also be performed by hybrid multi-scale velocimetry. Here OCT measurements are complemented (see Fig. 19) by measuring the outer flow velocity profile with, e.g. UVP or MRI techniques (Salmela et al., 2013; Haavisto et al., 2017; Kataja et al., 2017). The hybrid multi-scale velocimetry can provide very detailed experimental information on the rheology of various complex fluids in a wide range of flow rates. It enables not only the analysis of the macroscopic (bulk) behavior of the working fluids, but also gives simultaneous information on their wall layer dynamics, both of which are needed for analyzing and solving practical fluid-flow-related problems (see Fig. 20). In particular, wall layer dynamics, inaccessible with UVP/MRI but obtained with DOCT, have great importance for understanding the flow behavior of complex fluids.
a) Velocity profiles of 1.0 % microfibrillated suspensions in a pipe flow combined from DOCT and UVP measurements. The labels of different profiles give the total low rate [ml/s]. The vertical dashed line indicates the centerline of the tube. b) The velocity values from DOCT, appearing in a) as dark symbols near the wall, shown in more detail. The labels of different profiles give the wall shear stress [Pa]. Image taken with permission from Kataja et al. (2017).
a)–b) Measured viscosity μ(y) in a pipe (D = 19 mm) flow as a function of the distance y from the tube wall for a) 0.4 % bleached softwood kraft (BSK), calculated using DOCT and UVP velocity profiles, and b) for 0.4 % microfibrillated cellulose (MFC) calculated using DOCT and MRI velocity profiles. The wall layer, where there appears to be a consistency profile, has been indicated with a vertical dashed line. c)–d) Viscosity vs. shear rate for the c) BSK as measured by UVP (circles) and DOCT (crosses), and d) for MFC as measured by MRI (circles) and DOCT (crosses). The DOCT data points originate from the wall layer, and the UVP/MRI data from the inner parts of the pipe, i.e. from the yielding and plug regions. Additionally, the power law fits to the UVP/MRI data, and reference data for BSK obtained from Silveira et al. (2002) are shown. Image taken with permission from Haavisto et al. (2017).
Microrheology is a relatively new rheology technique where the local and bulk mechanical properties of a complex fluid are extracted from the motion of probe particles embedded within it (Waigh, 2016). In passive microrheology, particles are forced by thermal fluctuations and linear viscoelasticity is probed, whereas active microrheology involves forcing probes externally and can be extended out of equilibrium to the nonlinear regime. The advantages of microrheology are: The very small size of the samples, the capability to measure rheological properties with high frequencies, and the ability to study materials such as polymer solutions with probes spanning some of the characteristic microscopic length scales (e.g. approaching the inter-chain separation or mesh size of gels). Due to its ability to analyze opaque materials, the wide frequency range, and relatively high resolution, DOCT is a very promising tool for microrheology (Sharma et al., 2008; Blackmon et al., 2016; Chu et al., 2016).
DOCT obviously offers great potential for rheological analysis. Moreover, combining DOCT data with structural information, also given by DOCT, opens truly unprecedented research prospects for the study of the rheology of opaque fluids (Koponen et al., 2018). The extension to polarization sensitivity makes it possible for OCT to also study birefringent materials (Park et al., 2003; Ju et al., 2013).
Turbulence is a flow regime in fluid dynamics characterized by chaotic changes in pressure and flow velocity. It is in contrast to a laminar flow regime, which occurs when a fluid flows in parallel layers, with only diffusive disruption between those layers. In general terms, in turbulent flow, unsteady vortices of many sizes appear and interact with each other. This increases the viscous shear stresses and the energy consumption, e.g. in pipe flow. Turbulence is involved in many processes in the process industry. It is not only a nuisance, but turbulence also has the ability to mix and transport species, momentum, and energy much faster than is done by molecular diffusion and is therefore employed in, for example, chemical reactors to make them perform better. It is of great importance to have a detailed understanding of the turbulence when industrial processes and devices are developed. Typically, this work is done with a combination of computational fluid dynamics and experiments (Batchelor, 2000; Hjertager et al., 2003).
DOCT is in principle capable of performing measurements with a high enough temporal and spatial resolution to visualize the turbulent eddies. However, the measurable velocity region is rather limited for the most popular realization of DOCT, the FD-OCT, and phase wrapping is often a problem. As discussed above, in laminar flows, even then it is possible to retrieve the actual velocity value by phase-unwrapping methods, if the phase wrapping is not too big. This is, however, problematic in turbulent flows, as the velocity fluctuations can exceed the measurable velocity region considerably. Moreover, the basic measurement set-ups of DOCT see only one velocity component, whereas turbulence is always a 3D phenomenon. While there are various ways to improve the method to measure two or even three velocity components simultaneously (as discussed above), to this day they appear not to have been applied for the quantitative analysis of turbulence.
Despite its current limitations, a few studies can be found where DOCT has been applied for turbulent flow analysis—mainly for obtaining temporally averaged profiles. Bonesi et al. (2007) used DOCT for monitoring turbulent velocity profiles in micro-channels. They observed, e.g. the transition from laminar to turbulent flow velocity profiles in a T-shaped micro-vessel junction (see Fig. 21). Villey et al. (2010) used TD-OCT and a zero crossing algorithm to obtain velocity maps in a 1.0-mm capillary free from phase aliasing or other common OCT artifacts. In Haavisto et al. (2017), DOCT was combined with MRI to measure the mean velocity profiles of turbulent flow of pure water and a xanthan gum solution in a 19-mm pipe. The combined velocity profiles obtained were in excellent agreement with their theoretical counterparts (see Fig. 22).
Stationary velocity profiles taken across a T-shaped junction. a) Initial flow distribution at the end of the inlet arm. b) Velocity flow distribution across the T junction. c), d) Turbulent flow profiles in the left and right outlet arms, respectively. The red and blue colors introduce minimum and maximum rates of the velocity of flow, respectively. e) The T-shaped micro vessel is 1.8 mm in diameter. Image taken with permission from Bonesi et al. (2007).
Turbulent velocity profiles of a) pure water and b) xanthan gum solution measured with DOCT and MRI in a pipe with a diameter of 19 mm in the dimensionless variables u+ and y+. The dashed line and the dash-dotted line show the theoretical linear viscous sublayer profile and the logarithmic law of the wall, respectively. Due to the drag reduction of the xanthan gum, the MRI data exceed the logarithmic law of the wall at y+ ≅ 15 before becoming parallel to it. Image taken with permission from Haavisto et al. (2017).
The flow instabilities of solutions of high molecular weight DNA were investigated with DOCT in a parallel plate rheometer in Malm and Waigh (2017). At lower DNA concentrations and low shear rates, the velocity fluctuations were well described by Gaussian functions, and the velocity gradient was uniform across the rheometer gap, which is expected for Newtonian flows. As the DNA concentration and shear rate were increased, there was a stable wall slip regime followed by an evolving wall slip regime which was finally followed by the onset of elastic turbulence (see Fig. 23).
a) Power spectral density (PSD) of the velocity fluctuations of elastic turbulence measured in a plate-plate rheometer for three different concentrations of low salt DNA solutions, with an applied shear rate of 60 s−1. Two separate power law behaviors are observed. b) Color map of shear-normalized velocity as a function of the normalized distance across the rheometer gap and time after the startup of shear with a shear rate of 60 s−1. Intermittent fluctuations in the velocity are observed. Note that elastic turbulence is a fresh topic of research, and a theoretical explanation for the observed peculiar behavior of PSD—it is clearly lowest at the intermediate concentration—does not yet exist. Image taken with permission from Malm and Waigh (2017).
Just like all measurement methods, OCT also has its limitations. The penetration depth, even in optimal conditions, may be too low for some applications and especially in highly scattering materials, it can be very limited. In Aigner et al. (2014), e.g. the flow characteristics of glass-fiber-reinforced polymers were directly analyzed in a real production process for an elongational rheometry. Because of air bubbles and the high consistency of scattering glass fibers, DOCT measurements turned out to be very difficult. In its basic form, DOCT gives only one component of the velocity vector. While it is possible to get more components with special techniques, these are not necessarily easy to perform. Fortunately, one velocity component is often enough for a meaningful flow analysis. Moreover, flow field analysis can often be performed for structural images to obtain two or even three velocity components. OCT needs optical access to the measured target, and the medium has to carry light-scattering particles. Obviously, this may limit its use for certain applications. Particles that are optically too dense can even block the penetration of the OCT signal deeper into the medium. This produces straight dark lines in the OCT images.
Multiple scattering in the studied medium reduces image contrast and resolution, and distorts the measured velocity profiles, especially in highly scattering media (Kalkman et al., 2010; Lauri et al., 2011a). This effect can be decreased by using complex averaging. The conventional OCT averaging method (“magnitude averaging”) calculates the magnitude-squared Fourier-transformed spectral fringes before averaging and thus is insensitive to changes in phase. Another option is to average the complex-valued, Fourier-transformed, spectral-fringe signals before calculating the magnitude. Complex averaging increases the dynamic range by reducing the noise floor while maintaining similar signal values when compared to magnitude averaging (Thrane et al., 2017).
OCT is obviously a very attractive tool for structural and flow analysis. However, the number of industrial applications of DOCT in structural and flow analysis is still rather limited. This is probably due to the thus far small number of DOCT devices available outside the medical field. There are now commercial, general-purpose DOCT devices available, but their prices are still rather high, which limits the spread of DOCT technology to other applications. Nevertheless, the many advantages of DOCT technology, i.e. high resolution, high speed, ability to work with opaque fluids, and simultaneous detection of velocities and structure are a combination that cannot be obtained through other measurement methods. Moreover, the existing commercial devices are very compact and portable, which makes their application easy for new problems. So, although DOCT activities are still dominated by various biomedical applications, we expect the use of DOCT to increase in numerous industrial fields in the near future.
The Academy of Finland (project 288694) is gratefully acknowledged for supporting this work. This project has received funding also from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 713475 (Spinnova Ltd). We also want to thank Mr. Juha Salmela and Dr. Markko Myllys for taking the OCT image Fig. 1b), and Mr. Roope Lehto for its visualization.
Antti Koponen
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/37_2020003_24.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Antti Koponen prepared a Ph.D. thesis on simulations of fluid flow in porous media and obtained his Ph.D. degree in 1998 from the University of Jyväskylä (Finland). In 2003, Koponen obtained a docentship in applied physics at the University of Jyväskylä. Principal scientist Antti Koponen has worked at VTT Technical Research Centre of Finland Ltd. since 2001. His work focuses on the rheology of complex fluids and flows in complex geometries. Recently, Koponen has worked extensively with the novel OCT-based rheological analysis of complex fluids (Rheo-OCT).
Sanna Haavisto
;%0A%09%09%09newWindow.document.open();%0A%09%09%09newWindow.document.write('<img src=%22./Graphics/37_2020003_25.jpg%22>');%0A%09%09%09newWindow.document.close();%0A%09%09)
Sanna Haavisto is currently a Product Development Specialist at Spinnova Ltd., developing environmentally friendly fibre technology. She is also finalizing her PhD at the University of Jyväskylä where she obtained her master’s degree (industrial physics) in 2004. Until 2014, Sanna Haavisto was employed at VTT Technical Research Centre of Finland as a senior scientist. Her work was mainly focused on experimental research of the flow properties and rheology of complex fluids, including novel OCT-based rheological measurements (Rheo-OCT).
https://ror.org/04b181w54 
