Abstract
A novel optical method was applied to measure the binary liquid diffusion coefficient (D) quickly. Equipped with an asymmetric liquid-core cylindrical lens (ALCL), the spatially resolving ability of the ALCL in measuring refractive index of liquid was utilized to obtain the gradient distribution of the liquid concentration along diffusive direction. Based on Fick’s second law, the D value was then calculated by analyzing diffusion images. It was worth mentioning that only one instantaneous diffusive image was required to measure D value by the method, reducing the measurement time greatly from several hours in traditional methods to a few seconds. The diffusion coefficients of ethylene glycol diffusing in pure water, at temperatures from 288.15 to 308.15 K, were measured by analyzing instantaneous diffusion images, the results were consistent well with the values measured by using holographic interferometry and Taylor dispersion methods. The method is characterized by faster measurement, direct observation of diffusive process, and easy operation, which provides a new method in measuring diffusion coefficient of liquids rapidly.
© 2015 Optical Society of America
1. Introduction
Diffusion is a process by which matter is transported from one part of a system to another part due to molecules’ thermal energy under the influence of a concentration gradient [1, 2]. The diffusion coefficient (D) is a basic physical property data in chemical engineering and knowledge about it is needed for mass transfer studies and in many other fields such as crystal growth, biological systems, pollution control and separation of isotopes and so on [3–7]. Diffusion phenomena can appear in all gaseous, liquid and solid materials. To measure the D values of liquids is more difficult than the values of gases and solids [8, 9]. Many physical and chemical methods are employed for finding the liquid diffusivity experimentally, for instance, holographic interferometry [10], Taylor dispersion method [11] and the fluorescent molecule tracing [12], each with their characteristic advantages and disadvantages. The several common disadvantages in these methods include (1) take long time (>2h) to obtain the experimental data, (2) cannot observe the diffusion process directly, (3) need very stringent optical conditions.
Some emerging methods can solve these problems partly. For example, the experimental equipment of Moiré deflectometry method [13] is simple and it is low sensitive to external disturbances. To overcome these disadvantages, we have designed and fabricated an asymmetric liquid-core cylindrical lens (ALCL), which is able to measure refractive index (RI) of filled liquids in the way of spatial resolution along the axis of ALCL. ALCL is used as both the diffusion pool and the core imaging device, in which a dynamic gradient distribution of RI forms along the diffusion direction because of molecule diffusion. Once a thin liquid layer with a fixed RI (nc) is selected to be imaged in measurements, a shifting sharp focal point related to the layer will appear on CCD. Based on Fick’s second law [1], only one instantaneous image in the diffusion process is required to figure out the binary liquid D value, because the gradient distribution of RI, n(Z), is implied by the width of the diffusion image. It takes only 20 micro-seconds to capture one instantaneous image, less than 10 seconds to get information about n(Z), and work out D values by using an analytic program designed by ourselves. Compared with the traditional methods [10–13] of which several hours are needed for measuring a single D value, the method is characterized by faster measurement, direct observation of diffusive process, and easy operation.
Ethylene glycol (EG), featured by non-corrosive and non-crystalline, has the effect of absorbing carbon dioxide and dehumidification, researches find that EG is of important application in the field of environmental protection in recent years [14]. Accurate D values of EG in water at five different temperatures from 288.15 K to 308.15 K were measured by analyzing instantaneous diffusion images in this paper, and the results were consistent well with the values measured by using holographic interferometry [15] and Taylor dispersion method [16].
2. Principle and method
2.1 Imaging principle
ALCL is composed by sticking two different cylindrical lenses together in order to reduce spherical aberration to the minimum. Acted as both a diffusion pool and a key imaging device in the D value measurement, the imaging principle for measuring D with ALCL is shown in Fig. 1. When collimated light beams passing though the ALCL, if the ALCL is filled with one uniform liquid, the focal lengths at different heights (Zi) of ALCL are the same, a CCD chip positioned at the focal plane will receive a sharp line image, as shown in Fig. 1(a). If two kinds of liquids with different RIs are loaded into the ALCL in successively, the refractive capabilities of the upper and lower parts of ALCL are different. When collimated light beams in upper part is imaged clearly on the CCD, the lower part of the beams is a diffused image, as shown in Fig. 1(b). Once the two liquids are contacted each other, the diffusion process commences. A dynamic gradient distribution of concentration for the two solutions is formed gradually along the Z-axis, corresponding to a measureable RI gradient distribution. If the CCD is fixed at a specific position, where the collimated light beams will be imaged clearly for a thin liquid layer with an explicit RI value (say n = nc), an image with “beam waist” pattern shall appear on the CCD chip, as shown in Fig. 1(c). The narrowest width of the image is determined mainly by diffraction limit, however, the width of other area in the image is a function of RI value of liquid layer, which will be described in next part of this paper. Since liquid diffusion is a dynamic process, the “beam waist” pattern will drift slowly along the Z-axis, which is recorded visually as shown in Visualization 1, when the chemical EG diffusing in pure water.
2.2 Formula derivation for measuring RI
The top view of ALCL is showed as Fig. 2. The RI of liquid filled in the ALCL must be known before the measurement of D value. Let the focal length and RI of liquid filled in the ALCL be fi and ni, respectively, based on paraxial imaging Gaussian formula [17], the iterative relations between fi and ni can be written as,
Where R1 = 20.0 mm, R2 = R3 = 17.0 mm, and R4 = 37.6 mm are the curvature radii of the four refraction surfaces, respectively; d1 = d2 = d3 = 3 mm are the wall thickness and cavity thickness, respectively; material of cylindrical lens is K9 glass (n0 = 1.5163 at λ = 589 nm). With the help of a digital translation stage, the RI = ni of liquid filled in ALCL can be figured out by measuring the focal length fi. The standard deviation in measuring RI is less than 0.0001 and in [18], Q. Li et al have described the measurement in detail.Let h be the width of collimated light beams, fc be the focal length of the ALCL filled with liquid of RI = nc, fi be the focal length of the ALCL filled with liquid of RI = ni. If the CCD is fixed at the focal plane of the ALCL filled with liquid of RI = nc, as shown in Fig. 2, the collimated beams passing the ALCL filled with liquid of RI = ni>nc (or ni<nc) will project a width Wi on the CCD plane. From the view of geometrical optics, the image width (Wi) and focal length (fi) satisfies with
Where h is a known value; fc can be measured accurately by using the method mentioned before; Wi is the image width that can be read out straightforwardly by a software built in the CCD camera; thus, fi can be calculated by Eq. (2). Inserting fi to Eq. (1), ni, the RI of liquid filled in ALCL can be figured out. Since Wi is varied with Z-axis as shown in Fig. 1(c), Wi = Wi(Z). After fi = fi(Z) is determined by measuring Wi = Wi(Z), ni = ni(Z), that is, the spatial distribution of RI is determined completely by only one diffusion image.2.3 Calculation theory for diffusion coefficient
Assuming binary solutions involved in diffusion to be A and B, diffusion direction to be Z-axis, and C(Z, t) to be the mass fraction of A in B at diffusion time t and position Z. Let D be the diffusion coefficient, based on Fick’s second law [1], C (Z, t) satisfies with
Assuming the initial concentrations to be C1 and C2 on each side of the contact interface (Z = 0) before the diffusion beginning. According to the boundary and initial conditions, the solution of Eq. (3) is the form of Gauss error function [19],
Where, . Let “erfinv” be the inverse error function, and suppose the function relation between solution concentration and its RI to be, Eq. (4) can be rewritten asThe concrete form of is pre-determined by experimental method.In actual measurements, due to the intermolecular attraction, the contact interface cannot be an exact horizontal plane which leads to a difficulty in positioning the contact interface exactly, i.e. the position of Z = 0. So the position Z, a value relative to the contact interface, will have a deviation ΔZ. Equation (5) can be revised to
In the method of instantaneous analysis, a diffusion image is recorded by a CCD chip at any time t. The spatial distribution of liquid concentration C(Z, t) caused by diffusion can be worked out by analyzing the spatial distribution of RI, n(Z, t), which is implied by the recorded image, as described in section 2.2. Let be a constant, taking linear regress analysis between Zi and, supposing the linear coefficient to be a, thus . The diffusion coefficient to be measured can be expressed by,
which is independent on the deviation ΔZ.3. Experimental setup
Figure 3 is the experimental setup. A 2-mW green laser beam (λ = 543.5 nm) was attenuated by a neutral density filter, and then the attenuated beam was expanded and collimated by a microscope objective lens ( × 10) and a double convex lens (f = 150 mm). An adjustable slit was used to control the width of the collimated beam. After passing though the ALCL filled with liquid, the light reached a CCD with 2448 × 2058 pixels and 3.45 × 3.45 µm2 per pixel size. The CCD, connected to a PC via a USB, was driven by a translation stage with 0.001 mm minimum scale to adjust the imaging plane position slightly, which made the observation for a diffusion image convenient. The coordinates of arbitrary pixel on an image can be gotten and shown on PC screen by the software built in the CCD camera.
The cross section of the ALCL was showed in Fig. 2 and its height was 50 mm. The bottom of the ALCL was sealed off with rubber membrane. The denser solution EG (chemical purity grade) was injected to ALCL’s lower half part using a digital syringe (RSP02-C, produced by Ruichuang Electronic Technology Co., Ltd) firstly, waiting for 5-10 minutes to damp liquid convection, and then the distilled water was introduced to the ALCL’s upper half part along the inner wall of ALCL by the same digital syringe with a slow speed (2.5ml/ min), making sure no obvious convection current in the two liquids. The injection of distilled water was taken about 2 minutes, the time of two solutions starting to contact was defined as the onset of diffusion (t = 0).
The CCD was set along Y-axis at a position between two focal planes where the ALCL was filled with water and EG, respectively. The CCD was also set along Z-axis at the position where the contacting interface of two solutions can be imaged on the bottom of the CCD. To reduce the interface turbulence caused by liquid injection, diffusion images were taken 20 minutes after the diffusion process beginning.
4. Experimental results and discussion
4.1 Preparatory works
4.1.1 Experimental relationship between the solution concentration and its RI
In order to calculate the diffusion coefficient D by Eq. (5) at different temperatures, the experimental relation between the concentration and RI of EG aqueous solution at different temperatures, must be known in advance. To do it, all of experiments were carried out in an air-conditioned room. EG aqueous solutions were put to use when their temperatures were equal with ambient temperatures. Eight groups of EG aqueous solutions with different mass fractions were prepared using a precision electronic auto-balance with 0.0001g precision, and their RIs were measured by an Abbe Refractometer at different temperatures. The results were shown in Table 1, and the fitting relations between solution concentrations and related RIs were listed at the last row of Table 1, showing a good linear relationships whose correlation indexes were better than 0.9998.
4.1.2 Relationship between the image width and the RI filled in ALCL
Both the experimental and theoretical methods have been used to find the relationship between the image width and the RI of liquid filled in the ALCL. For the experimental method, distilled pure water of RI = 1.3334, measured by the Abbe Refractometer at 298.15 K, was injected into ALCL to calibrate the imaging system. A CCD was fixed at the position where the image appeared on the CCD was of the narrowest width, when the ALCL was filled with the liquid of RI = nc = 1.3391. Groups of alcohol aqueous solutions were prepared and their RIs were measured by the Abbe Refractometer. When aqueous solutions with different RIs were filled in the ALCL in sequence, the images recorded by the CCD were shown in Fig. 4, which indicated that the image width decreased with the rise of RI, if RI is smaller than nc = 1.3391, otherwise the image width increased with the rise of RI. The image widths (Wi) were measured and plotted in Fig. 5 by the dots.
For the theoretical method, the RI of liquid in the Eq. (1) is substituted by ni = nc = 1.3391, the calculated focal length is fc = 52.475mm. Setting the entrance slit be h = 7mm (experimental condition), according to Eq. (1) and Eq. (2), the image width (Wi) varied with liquid RI (ni) is figured out, the result is shown in Fig. 5 by the solid line, which is in good consistent with experimental data. The experimental data is fitted linearly by the Least Square method, the fitting result is expressed by the conditional function as
4.2 Measurement of diffusion
In the whole experiment of EG diffusing in pure water, the CCD was fixed at the position where the width of image was the narrowest when ALCL was filled with liquid of RI = nc = 1.3391. Both ambient and solution temperatures were 298.15 K in the first experiment. The whole diffusion process was shown in Visualization 1, while a typical diffusion image at the time of t = 1800s was shown in Fig. 6.
Data pairs of (Zi, Wi)s were read out automatically by a software built in the CCD camera. The data pairs (Zi, Wi)s were converted into (Zi, ni) based on Eq. (8), and then (Zi, Ci) based on the formula of C = 9.8478n −13.130 in the 4th column of Table 1, so the spatial distributions of RI and concentration were obtained, those were shown in Fig. 7.
Let C1 = 0, C2 = 1, t = 1800s, and = 9.8478*n(Zi) −13.130, Eq. (6) can be expressed as
Zi, and are fitted linearly by the least square method, the linear coefficient a = 0.2882 is obtained, and then D = 1.1536 × 10−5cm2/s is figured out by Eq. (7), which is a litter bit smaller than the literature value [15], 1.189 × 10−5cm2/s, measured by the method of holographic interferometry. It is worth noting that the total time cost on measuring D value is about 7 second, including 20ms for taking an instantaneous image, and 6.297 second for data-processing that can be shorten further. Therefore, fast measurement is the significant feature of the presented method.Besides the instantaneous image shown in Fig. 6, other eight diffusion images, recorded by an interval of 5 minutes, are shown in Fig. 8. Using the same techniques in processing of Fig. 6, we have calculated D values based on the eight diffusion images, and the calculated results are listed in Table 2.
It is clear that the calculated D values are stable basically in the period of 1200 to 3600s, the mean and standard deviations of D values are 1.1378 and 0.0353 × 10−5cm2/s, respectively. A larger value at 1200s, D = 1.1932 × 10−5cm2/s may be caused by the interface turbulence, while a smaller value at 3600s, D = 1.0866 × 10−5cm2/s is probably caused by the reading error of the image width.
4.3 Verification
For the same diffusion images shown in the Fig. 8, the diffusion coefficient (D) can also be figured out by different method. The narrowest positions (Zi) of instantaneous images, marked by arrows in Fig. 8, drift with time (ti) due to dynamic diffusion process. Since the narrowest position corresponds to a liquid layer with a fixed RI, nc = 1.3391 in the experiment, g [nc (Z, t)] in Eq. (6) is a constant that is g[nc (Z, ti)] = 9.8478*n(Zi) −13.130 = 0.057. Let C1 = 0, C2 = 1, Eq. (6) can be expressed as
The experimental data of Zi varied with ti are listed in the Table 3, where N represents the number of pixels between the “waist” and the diffusion interface, which are read out by the software built in the used CCD, and value Zi is obtained by multiplying N with a pixel size. Zi and are fitted linearly by the least square method, the fitting result can be expressed as
The linear correlation coefficient is 0.998. Comparing Eq. (10) with Eq. (11), we obtain the D value of EG diffusing in water at temperature 298.15 K, that is D = 1.1087 × 10−5 cm2/s. The value is closed to the mean value calculated by analyzing an instantaneous diffusion image, the reliability of the instantaneous image analysis is thus verified.4.4 D values at different temperatures
Liquid diffusion process is very sensitive to temperature. For the EG diffusing in pure water, the dependence of D value on temperature has been measured by using the instantaneous image analysis, and the result is shown in Table 4. Similar to our expectation, the measured D value increases with temperature, and the D values obtained by us are almost the same as that of infinite dilution diffusion coefficient measured by using Taylor dispersion method [16] at temperatures of 298.15, 303.15 and 308.15K. The functional dependence of the infinite dilution D value on temperature is given by the Arrhenius equation [20, 21] expressed as
Where E is the diffusion activation energy, R is the gas constant and A is a constant value. Our measured D values are well consistent with the Arrhenius equation as showed in Fig. 9, the correlation index is 0.9939. The measured activation energy is 2.098 Kcal/mol.4.5 Error analysis and discussion
According to Eq. (7), the deviation in measuring D value can be analyzed by the error transfer formula expressed as
The deviation of is caused by the error in measuring diffusion time Δt. The distilled water is injected to the ALCL slowly in the diffusion experiment to avoid the convection current, and the moment that the first drop of water meet with EG solution is defined as the onset of diffusion process. But the diffusion mechanism when the amount of water is rare is not the same as that when the water is filled with the upper part cell. Therefore, there is a deviation in determining the diffusion onset time which is assumed to be Δt = 30s. Let t = 1200s and a = 0.2393, the calculated deviation of D value is = 0.0298 × 10−5cm2/s, which is the largestfor the D values in Table 2. The deviation of is composed by two parts, the error in measuring RI of liquid and the error in measuring the image width ,. The accuracy of Abbe Refractometer, which is used to measure the RIs of experimental solutions, is Δn = 0.0002. Based on the experimental formula listed on the 4th column of Table 1, the concentration deviation caused by the value of Δn = 0.0002 is ΔC = 9.8478Δn = 0.0019, thus Δa = 0.0061, leading to a calculated deviation of D value = 0.0489 × 10−5 cm2/s. The widths (Wi) of images are read out automatically by a software built in the CCD camera with one pixel reading error, so the error (ΔW) of width is 3.45 µm which produces a RI deviation Δn = 0.00009 according to Eq. (8), leading to a calculated deviation of D value = 0.0224 × 10−5cm2/s. Therefore, the deviation of D value caused by the value a is about = 0.0713 × 10−5 cm2/s.Combining the above error analysis, the total deviation of D value is = 0.0773 × 10−5 cm2/s, and the main deviation is from the error in measuring the concentration of diffusion liquid. A higher precision measuring tool for RIs is necessary and useful to reduce, and a selected diffusion image with a longer diffusion time may decrease to the minimum.
This method is based on the determination of the gradient distribution of RI, and the standard deviation in measuring RI is about 0.0001, therefore, the method does not work for two fluids with similar refractive indices. The gradient distribution of RI decreases as the diffusion process goes on, which also reduces measurement accuracy, so the diffusion images after 60 minutes are not collected and analyzed. The sensitivity of this method at present is inferior to holographic interferometry method [22] and Taylor dispersion method [23], and is approximate with Moiré deflectometry method [13]. However, the characteristics of the method, such as quick measurement, direct observation of diffusive process and easy operation, are irreplaceable by other measurement methods.
5. Conclusion
We introduce a new method for measuring liquid diffusion coefficient rapidly in the paper. The key optical device of the method is an ALCL that is used as both the diffusion cell and imaging lens. Owing to the spatially resolved ability of the ALCL in measuring liquid RI, the distribution of the liquid concentration caused by diffusion process can be recorded by an instantaneous diffusion image, and then the D value is calculated on the base of Fick’s second law. The experimental time cost on measuring a D value is shortened greatly from several hours in traditional method to a few seconds in the new method. The experiments of EG diffusing in pure water have been carried out, which demonstrates that the new method is feasible, and it opens a door for measuring liquid diffusion coefficient rapidly.
Acknowledgments
This work was supported by the National Science Foundation of China (Grant NO. 11164033 and 61465014).
References and links
1. J. Crank, The Mathematics of Diffusion (Oxford University, 1975).
2. R. Taylor and R. Krishna, Multicomponent Mass Transfer (Wiley & Sons Inc., 1993).
3. E. L. Cussler, Diffusion: Mass Transfer in Fluid Systems, 3rd ed. (Cambridge University, 2009).
4. N. Weiss, T. G. van Leeuwen, and J. Kalkman, “Simultaneous and localized measurement of diffusion and flow using optical coherence tomography,” Opt. Express 23(3), 3448–3459 (2015). [CrossRef] [PubMed]
5. C. Angstmann and G. P. Morriss, “An approximate formula for the diffusion coefficient for the periodic Lorentz gas,” Phys. Lett. A 376(23), 1819–1822 (2012). [CrossRef]
6. S. Lang, T. J. Kazdal, F. Kühl, and M. J. Hampe, “Diffusion coefficients and VLE data of aqueous phosphoric acid,” J. Chem. Thermodyn. 68, 75–81 (2014). [CrossRef]
7. Y. Zou, Z. Shen, X. Chen, Z. Di, and X. Chen, “An integrated tunable interferometer controlled by liquid diffusion in polydimethylsiloxane,” Opt. Express 20(17), 18931–18936 (2012). [CrossRef] [PubMed]
8. X. Liu, S. K. Schnell, J. Simon, P. Krüger, D. Bedeaux, S. Kjelstrup, A. Bardow, and T. J. H. Vlugt, “Diffusion coefficients from molecular dynamics simulations in binary and ternary mixtures,” Int. J. Thermophys. 34(7), 1169–1196 (2013). [CrossRef]
9. X. Liu, A. Bardow, and T. J. H. Vlugt, “Maxwell-Stefan diffusivities and velocity cross-correlations in dilute ternary systems,” Principles Diffusion Theory 16, 81 (2011).
10. V. K. Chhaniwal, A. Anand, S. Girhe, D. Patil, N. Subrahmanyam, and C. S. Narayanamurthy, “New optical techniques for diffusion studies in transparent liquid solutions,” J. Opt. A, Pure Appl. Opt. 5(5), S329–S337 (2003). [CrossRef]
11. T. C. Chan and W. K. Tang, “Diffusion of aromatic compounds in nonaqueous solvents: a study of solute, solvent, and temperature dependences,” J. Chem. Phys. 138(22), 224503 (2013). [CrossRef] [PubMed]
12. C. T. Culbertson, S. C. Jacobson, and J. Michael Ramsey, “Diffusion coefficient measurements in microfluidic devices,” Talanta 56(2), 365–373 (2002). [CrossRef] [PubMed]
13. K. Jamshidi-Ghaleh, M. T. Tavassoly, and N. Mansour, “Diffusion coefficient measurements of transparent liquid solutions using Moiré deflectometry,” J. Phys. D Appl. Phys. 37(14), 1993–1997 (2004). [CrossRef]
14. K. Y. Suh and R. Langer, “Microstructures of poly (ethylene glycol) by molding and dewetting,” Appl. Phys. Lett. 83(8), 1668–1671 (2003). [CrossRef]
15. J. Fernández-Sempere, F. Ruiz-Beviá, J. Colom-Valiente, and F. Más-Pérez, “Determination of diffusion coefficients of glycols,” J. Chem. Eng. Data 41(1), 47–48 (1996). [CrossRef]
16. M. H. Wang, A. N. Soriano, A. R. Caparanga, and M. H. Li, “Mutual diffusion coefficients of aqueous solutions of some glycols,” Fluid Phase Equilib. 285(1-2), 44–49 (2009). [CrossRef]
17. M. Born and E. Wolf, Principles of Optics (Cambridge University, 2003).
18. Q. Li, X.-Y. Pu, R.-F. Yang, and Y. Zhai, “Measurement of diffusion coefficient of liquids by using an asymmetric liquid-core cylindrical lens: observing the diffusion process directly,” Chin. Phys. Lett. 31(5), 054203 (2014). [CrossRef]
19. W. Jost, Diffusion in Solids, Liquids, Gases (Academic, 1960).
20. D. F. Othmer and M. S. Thakar, “Correlating diffusion coefficients in liquid,” Ind. Eng. Chem. 45(3), 589–593 (1953). [CrossRef]
21. E. D. Snijder, M. J. M. te Riele, G. F. Versteeg, and W. P. M. van Swaaij, “Diffusion coefficients of several aqueous alkanolamine solutions,” J. Chem. Eng. Data 38(3), 475–480 (1993). [CrossRef]
22. A. Anand, V. K. Chhaniwal, and C. S. Narayanamurthy, “Diffusivity studies of transparent liquid solutions by use of digital holographic interferometry,” Appl. Opt. 45(5), 904–909 (2006). [CrossRef] [PubMed]
23. G. D’Errico, O. Ortona, F. Capuano, and V. Vitagliano, “Diffusion coefficients for the binary system glycerol + water at 25°C. a velocity correlation study,” J. Chem. Eng. Data 49(6), 1665–1670 (2004). [CrossRef]