Microfluidic device coupled with total internal reflection microscopy for in situ observation of precipitation

The quasi-2D microfluidic device was fabricated via standard photolithography process followed by wet etching as shown in Fig. 1. Borofloat glass wafer (81% SiO₂, 13% B₂O₂, 4% Na₂O/K₂O, 2% Al₂O₃. Borofloat wafer was first cleaned with piranha solution (3 : 1 = H₂SO₄ : H₂O₂) and rinsed by de-ionized water. Chrome (20 $nm$) and gold (150 $nm$) were sputtered on the surface of the Borofloat sequentially as masking layer. Photoresist (AZ1529) was spin coated on the metal-masked substrate, which was baked for 1 $min$. The substrate was exposed under UV light through a photo mask with pre-designed side channel pattern on it. The exposed patterned photoresist was removed by developer. Gold and chrome layers were then etched by gold and chrome etchant. The Borofloat substrate was then etched by hydrogen fluoride solution to create 240 $\mu m$ side channel. After etching of side channel, blue tape was used to protect both ends of main channel. Photoresist on main channel was then manually removed by acetone wetted cotton swab. Gold, chrome and Borofloat etching steps were repeated to get 20 $\mu m$ etch of the main channel. Borofloat was cleaned by de-ionized water and the remaining gold and chrome layer were stripped by etchants. The patterned Borofloat was then bonded with a glass cover slip (24 $mm$ $\times$ 50 $mm$ $\times$ 170 $\pm$ 5 $\mu m$, Azer Scientific) by epoxy. Similar to the structure of microchamber used in our previous work,lu2017 the detailed dimensions of the quasi-2D microfluidic device are shown in Fig. 2a. The narrow main channel is   20 $\mu m$ in depth and 6 $mm$ in width, flanked with two deep side channels with   260 $\mu m$ in depth and 3 $mm$ in width. The lengths of both the main channel and side channel are about 3 $cm$.

All of the experiments were conducted at room temperature ($\sim$19 - 21 ${}^{o}C$). To confirm the diffusive mixing inside the microchamber, the mixing process between toluene and n-pentane was followed by using a fluorescence microscope (Nikon ECLIPSE Ni) equipped with a camera (Nikon DS-Fi3). Toluene (ACS grade, Fisher Scientific, 99.9+%), doped with a trace amount of Nile Red (Fisher Scientific), was injected to fill the entire chamber. Then n-pentane (Fisher Scientific, 98%) was injected from one end of the side channel at a flow rate 5 $mL/h$ by a syringe pump (NE-1000, Pumpsystems Inc.) as sketched in Fig. 2. As flowing in the side channel along y-axis, n-pentane diffused transversely into the narrow quasi-2D main chamber to mix with toluene along x-axis. The location of the field of view was set at a distance of 80 $\mu m$ from the side channel through the transparent top glass of the main chamber.

Green laser (559 $nm$) was used to excite Nile Red and the emission was monitored at 635 $nm$. The change in the fluorescence intensity of the dye in the liquid was recorded at 2 frames per second by the fluorescence microscope (Nikon ECLIPSE Ni) equipped with a camera (Nikon DS-Fi3). The fluorescence intensity was converted to gray scale value and read using MATLAB software (The MathWorks Inc.). The fluorescence intensity as a function of time is governed by the mixing process, which is vital dataset to confirm whether diffusion mixing is dominant and to estimate the chemical composition of solution at given location.

The microfluidic device was initially filled with asphaltene solution in toluene (solution A). Then, solution B – n-pentane : toluene mixture at various ratios – was injected through the deep side channel with flow rate 5 $mL/h$ by a syringe pump where n-pentane is an antisolvent of asphaltene. Part of solution B diffuses transversely into the main channel and mixes with solution A. Asphaltene gradually precipitated with n-pentane diffusion in the main channel and can be detected by DeltaVision OMX Super-resolution microscope (TIRF 60$\times$/1.49 NA objective lens) (GE Healthcare UK limited, UK). TIRF was used to detect asphaltene precipitates in the microchamber. The principle of TIRF detection is sketched in Fig. 2b. Light refraction follows Snell’s Law, as shown in Equation (1):

n₁ and n₂ are incident and refracted index of a given pair of media, separately, $\theta_{1}$ and $\theta_{2}$ are incident and refracted angle of light in the two media, separately. When the light propagates in a denser medium (e.g the glass substrate) and meets the interface with a less dense medium (e.g the sample on the substrate) above a critical angle, rays of light are no longer refracted but totally reflected in the denser medium. Total internal reflection (TIR) occurs, generating an evanescent field adjacent to the interface of the media. The electromagnetic wave in the evanescent field can excite the fluorophores in the sample with penetration depth around 100-200 $nm$, depending on the optical properties of the media and the incident angle.fish

In our measurements of asphaltene precipitation, the refractive index of asphaltene (RI = 1.72) was used base on the literaturebuckley1998 ; wattana2003 , which is different to toluene (RI = 1.50)kedenburg2012 and n-pentane (RI = 1.36)kerl1995 . An excitation laser with wavelength $\lambda$ = 488 $nm$ was used to excite fluorophores and the emission wavelength detected by the detector was up to 576 $nm$. The TIRF images were taken on an 82.5 $\mu m$ $\times$ 82.5 $\mu m$ from the field of view in the area of the narrow main channel by a pco.edge sCMOS camera. Within this excitation and emission wavelength range, the asphaltene in n-pentane and toluene solution has much stronger signal than precipitated asphaltene particles, resulting in good contrast of image with the asphaltene particles shown as black particles.

For the visualization of $\beta$-alanine crystallization, the device was prefilled with a ternary mixture comprising of 6% $\beta$-alanine (Acros organics, 99%), 31.5% isopropyl alcohol (Fisher Scientific, 99.9%), and 62.5% water by weight. Pure isopropyl alcohol was injected through the side channel at 12 $mL/h$ to drive the crystallization. The process was monitored with the microscope in bright field imaging mode through a 10$\times$ magnification objective.

Fig. 3a shows the snapshots of fluorescence images of the diffusion of n-pentane in toluene solution with Nile Red in the microchamber. At time of 0 $s$, the fluorescence signal shown as red is strongest as n-pentane just starts to reach the shallow main channel from the deep side channel. With time, n-pentane moves along the x-direction into the main channel as revealed by expansion of the dark region along x-direction in the field of view because Nile Red does not fluoresce in n-pentane within the selected excitation. After 40 $s$, the entire field of view (691.2 $\mu m$ $\times$ 491.5 $\mu m$) turns dark, suggesting that toluene is entirely displaced by n-pentane from the examined area. As n-pentane diffuses into the main channel, toluene also diffuses into the side channel. Toluene that diffuses into the side channel is pushed away by the flow in the side channel.

The fluorescence intensity at four locations is plotted as a function of time shown in Fig. 3b. The locations are at 80-140 $\mu m$ away from the junction of the side channel and the main channel. The decay of the intensity (i.e. fluorescence signal vs. time) at any given location was fitted well by a sigmoid function with Equation (2), derived by the same method for fluorescence in polymerase chain reaction (PCR)rutledge2004 ; campa2015 . As shown in Equation (2), time (t) is used to replace the number of cycles in PCR.

I is normalized intensity, t is time and A is a constant depending on the location. The delay in the intensity drop of location correlated well with the distance leading to a steady increase of 0.3 per 20 $\mu m$ distance increment for parameter A, as shown in Fig. S1, suggesting that the mixing conditions are identical at different locations in our observation area.

In Fig. 3c, the front layer moving distance d normalized by the square root of time t^1/2 is plotted as a function of the distance d from the side channel. For a constant concentration source Fick’s diffusion in 1-dimension scenario, the diffusion length d and time t follows:

D is diffusion coefficient, which is assumed to be constant in our system. The results of constant $\frac{d}{\sqrt{t}}$ suggest that mixing at a critical distance (i.e. 80 $\mu m$ to the side channel) is well described by a diffusion process. At a distance from the side channel is less than 80 $\mu m$, the mixing is not well described by diffusion, which could be caused by the influence of the side channel. On the other hand, a longer distance means a longer diffusion time, which is undesirable as the maximal fluorescence intensity is reached in a shorter duration. Therefore, our observation area is kept at 80 $\mu m$ to the side channel in the following experiments.

The duration time for n-pentane replacing toluene is about 40 $s$ over 491.5 $\mu m$ distance. Since n-pentane diffusion rate follows Equation (3), $\frac{d}{\sqrt{t}}$ is constant, at around 60 $\mu m/s^{1/2}$, as shown in Fig. 3c. At 40 $s$, the calculated travel distance for front layer is 379 $\mu m$, which is comparable with the actual moving distance, i.e. 491.5 $\mu m$. The difference between calculated and actual values may be from the non-diffusive behaviour below the critical distance of 80 $\mu m$.

Given that n-pentane diffusion from side channel into the main channel is a one-dimension diffusion in our system, the diffusion follows:

$c$ is concentration of n-pentane, $t$ is time, $D$ is diffusivity of n-pentane in toluene and $x$ is the distance to the side channel. Solving Equation (4), n-pentane concentration distribution along x-direction is:

For a given time, n-pentane concentration distribution along x-direction follows error function. The error function distribution is proved in Fig. S2.

Assuming the ratio of the fluorescent dye Nile Red and toluene remains constant, from the intensity with time, we can obtain r_{solvent/toluene} in the diffusive mixing process. As shown in Fig. 4a and b, the intensity of fluorescence is quantified to obtain the change in the chemical composition in space and in time. Toluene concentration is estimated by $c$ $\sim$ $I$. Here $c$ is toluene concentration, and $I$ is normalized fluorescence intensity. Toluene concentration decreases with diffusion time, while n-pentane concentration, reverse of the toluene concentration, increases with time.

The mixing device was used to observe precipitation of asphaltene from toluene by n-pentane. As shown in Fig. 5, the precipitation of asphaltene can be observed in situ. Images were taken from 80 $\mu m$ which is the critical distance need to achieve diffusive mixing as also shown in Fig. 3. Up to a distance of 80 $\mu m$, mixing may be influenced by the bulk flow of solution B. The high amount of precipitates observed at 80 $\mu m$ is likely due to the convective mixing. However, at longer distances, the diffusive mixing is established and the amount of precipitates remain similar with distance. With the high resolution of TIRF – i.e. 200 $nm$ – observation of individual asphaltene precipitates may also be possible.

The proposed device is not limited to the in situ asphaltene precipitation, but it can also be applied for studying dilution induced crystallization from a liquid mixture. The dilution induced crystallization occurs due to the reduced solubility of the solute in the liquid mixture by the addition of an antisolvent. Using the diffusive mixing device, we can visualize the crystallization process. As an example, we demonstrate in situ growth of $\beta$-alanine crystals by dilution of a ternary solution consisting of $\beta$-alanine, isopropyl alcohol, and water (solution A) by isopropyl alcohol (solution B). As solution B is diffused through the main channel, $\beta$-alanine becomes oversaturated and crystallizes on the surface of the channel as shown in Fig. 6. Using the device shown here, it is possible to control the crystallization by tuning the mixing between antisolvent and the liquid mixture.