betaFeedback

WorksPhysics and Astronomy

A discussion on the critical electric Rayleigh number for AC electrokinetic flow of binary fluids in a divergent microchannel

Abstract

Electrokinetic (EK) flow is a type of flow driven or manipulated by electric body forces, influenced by various factors such as electric field intensity, electric field form, frequency, electric permittivity/conductivity, fluid viscosity and etc. The diversity of dimensionless control parameters, such as the electric Rayleigh number, complicates the comparison of EK flow stability. Consequently, comparing the performance and cost of micromixers or reactors based on EK flow is challenging, posing an obstacle to their industrial and engineering applications. In this investigation, we theoretically derived a new electric Rayleigh number (RaeRa_e) that quantifies the relationship among electric body forces, fluid viscosity, and ion diffusivity, based on a tanh model of electric conductivity distribution. The calculation results indicate that the new RaeRa_e exhibits richer variation with the control parameters and better consistency with previous experimental reports. We further conducted experimental studies on the critical electric Rayleigh number (RaecRa_{ec}) of AC EK flow of binary fluids in a divergent microchannel. The experimental variations align well with the theoretical predictions, particularly the existence of an optimal AC frequency and electric conductivity ratio, demonstrating that the tanh model can better elucidate the underlying physics of EK flow. With the new electric Rayleigh number, we found that EK flow in the designed divergent microchannel has a much smaller RaecRa_{ec} than previously reported, indicating that EK flow is more unstable and thus more suitable for applications in micromixers or reactors in industry and engineering.

Introduction

Microfluidic device is a pivotal member of lab on a chip (LOC), facilitating the manipulation of fluids ranging from microliters to picoliters[1] in fields such as biomedical[2, 3, 4] and chemical engineering,[5, 6, 7, 8] including applications like polymerase chain reaction (PCR),[9] synthesis of nanoparticles[10] and DNA sequencing chips.[11]

Micromixer represents a significant category of microfluidic devices. According to the mechanism of mixing enhancement, micromixers are typically classified into two types: passive and active[12]. Passive micromixers enhance mixing through the folding and stretching of fluids within specially designed microchannel geometries, requiring no additional driving mechanisms beyond a pressure gradient to induce flow disturbance.[13, 14, 15, 16] While passive micromixers often have specific functions, they generally lack versatility and exhibit relatively low mixing efficiency. In 2002, Strootk et al.[17] introduced a staggered herringbone mixer, discovering that at a Péclet number of 2×1032 \times10^3, the length required for 90% mixing was 7 mm, with the channel length needed for mixing increasing logarithmically with the Péclet number.

Active micromixers can significantly enhance mixing by applying external forces, such as electrostatic force,[18] Lorentz force[19] and acoustic pressure.20 Electrokinetic (EK) micromixer is a specific type of active micromixer that enhances mixing by EK flow driven by electrostatic force, and experiences a rapid advancement since 1990s.21 In 1998, Baygents et al.22 investigated the neutral stability curves for EK flow under a DC electric field that parallel to the electric conductivity interface. In their investigation, the status of EK flow is characterized by an electric Rayleigh number (RaeRa_e, see Table I), which is determined by the electric intensity, the interface width and the gradient of electric conductivity. In @page 1/12 starts_mid=0 ends_mid=1

ResearchersBaygents et al (1998) 22Chen et al (2005) 23Posner et al (2006) 24Wang et al (2014) 25
RaeRa_eεE2Dμd2Δσσ1\frac{\varepsilon E^2}{D\mu}d^2\frac{\Delta\sigma}{\sigma_1}εE24Dμdh2Δy(λ−1)2(λ+1)2\frac{\varepsilon E^2}{4D\mu}\frac{dh^2}{\Delta y}\frac{(\lambda-1)^2}{(\lambda+1)^2}εE2Dμh2λ−1λ∇∗σ∗∣ma\frac{\varepsilon E^2}{D\mu}h^2\frac{\lambda-1}{\lambda}\left.\nabla^*\sigma^*\right|_{ma}εE2Dμw2Δσσ1\frac{\varepsilon E^2}{D\mu}w^2\frac{\Delta\sigma}{\sigma_1}
Diagram of the EK flowσ2\sigma_2, σ1\sigma_1, ddσ2\sigma_2, σ1\sigma_1, Δy\Delta y, hh, ddσ2\sigma_2, σ1\sigma_1, bb, Δy\Delta y, hh5∘5^\circ, 5∘5^\circ, ww
ResearchersWu et al (2015) 26Yoshikawa et al (2015) 27Hassen et al (2020) 28Nan et al (2022) 21
RaeRa_eεEdKμ\frac{\varepsilon Ed}{K\mu}eΔθl3μκ∇εrefE22\frac{e\Delta\theta l^3}{\mu\kappa}\nabla\frac{\varepsilon_{ref}E^2}{2}εEdKμ\frac{\varepsilon Ed}{K\mu}4εE2Dμw2λ−1λ+1(1−β2)\frac{4\varepsilon E^2}{D\mu}w^2\frac{\lambda-1}{\lambda+1}(1-\beta^2)
Diagram of the EK flowddThe circular Couette flow; R2R_2, R1R_1MWCNT-oil nanofluid; dd5∘5^\circ, 5∘5^\circ

Table I. Definitions of RaeRa_e in different EK flow model. The red arrow indicates the flow direction, while the green arrow indicates the direction of the electric field.

Introduction

Microfluidic device is a pivotal member of lab on a chip (LOC), facilitating the manipulation of fluids ranging from microliters to picoliters[1] in fields such as biomedical[2, 3, 4] and chemical engineering,[5, 6, 7, 8] including applications like polymerase chain reaction (PCR),[9] synthesis of nanoparticles[10] and DNA sequencing chips.[11]

Micromixer represents a significant category of microfluidic devices. According to the mechanism of mixing enhancement, micromixers are typically classified into two types: passive and active[12]. Passive micromixers enhance mixing through the folding and stretching of fluids within specially designed microchannel geometries, requiring no additional driving mechanisms beyond a pressure gradient to induce flow disturbance.[13, 14, 15, 16] While passive micromixers often have specific functions, they generally lack versatility and exhibit relatively low mixing efficiency. In 2002, Stroock et al.[17] introduced a staggered herringbone mixer, discovering that at a Péclet number of 2×1032 \times10^3, the length required for 90% mixing was 7 mm, with the channel length needed for mixing increasing logarithmically with the Péclet number.

Active micromixers can significantly enhance mixing by applying external forces, such as electrostatic force,[18] Lorentz force[19] and acoustic pressure.20 Electrokinetic (EK) micromixer is a specific type of active micromixer that enhances mixing by EK flow driven by electrostatic force, and experiences a rapid advancement since 1990s.21 In 1998, Baygents et al.22 investigated the neutral stability curves for EK flow under a DC electric field that parallel to the electric conductivity interface. In their investigation, the status of EK flow is characterized by an electric Rayleigh number (RaeRa_e, see Table I), which is determined by the electric intensity, the interface width and the gradient of electric conductivity. In

EE: Applied electric field; dd: Channel width; σ1\sigma_1: Lower electric conductivity; σ2\sigma_2: Higher electric conductivity; Δσ=σ2−σ1\Delta\sigma=\sigma_2-\sigma_1; DD: Effective diffusivity; ε\varepsilon: Electric permittivity; μ\mu: Dynamic viscosity; κ\kappa: Thermal diffusivity; Δy\Delta y: Interface width; hh: Channel height; λ=σ2/σ1\lambda=\sigma_2/\sigma_1: Electric conductivity ratio; KK: Ionic mobility; bb: Initial interface width; ww: Initial width at the inlet; ll: Characteristic length scale of the flow; θ\theta: Temperature deviation; εref\varepsilon_{ref}: Electric permittivity at reference temperature; ee: Coefficient; β=2ωε(σ1+σ0)\beta=\frac{2\omega\varepsilon}{(\sigma_1+\sigma_0)}: Dimensionless number to evaluate the influence of frequency;

∇∗σ∗∣max≈1−e−(bΔy)2−e−(d−bΔy)2\left.\nabla^*\sigma^*\right|_{max}\approx1-e^{-\left(\frac{b}{\Delta y}\right)^2}-e^{-\left(\frac{d-b}{\Delta y}\right)^2}: Normalized maximum conductivity gradient;

2001, Oddy et al.29 designed a delicate micromixer to promote mixing through EK instability induced by sinusoidal electroosmotic flow. Subsequently, Lin et al.30 showed a plot of the growth rates of the most unstable eigenfunction in the wavenumber (kk)—RaeRa_e space. They found the critical electric field required for mixing decreases as the electric conductivity ratio increases. In 2005, Chen et al.23 defined a RaeRa_e (refer to Table I) for DC EK flow with electric conductivity difference in a T-shaped microchannel with a high aspect ratio. By experimental and analytical studies, they found the presence of both convective and absolute instabilities of EK flow. When the applied RaeRa_e is beyond a certain critical threshold, say Raec≈10Ra_{ec}\approx10, the flow becomes convective unstable where the disturbance only grows downstream. When RaeRa_e is beyond a second Raec≈160Ra_{ec}\approx160, the flow becomes absolute unstable, with the disturbances

propagating both upstream and downstream simultaneously. In 2006, Posner et al.31 studied the EK instability in a cross-shaped microchannel flow diagram. By defining a modified local electric Rayleigh number, characterized by the charge density as a function of the maximum electric conductivity gradients, they determined that the critical electric Rayleigh number Raec=205Ra_{ec}=205.

In 2014, Wang et al. designed an elaborate Y-type EK micromixer[5] [25, 32] and demonstrated that micro electrokinetic (μ\muEK) turbulence could be generated under an alternating-current (AC) electric field with 100 kHz frequency. However, the additional control parameter, e.g. AC frequency, makes the comparison of working conditions with the previous reports unattainable. Further investigations by Nan et al.21 demonstrate that the conditions of inducing the most unstable mode of EK flow is inconsistent to that of generating highlydeveloped EK turbulence. This further indicates the difficulty to advance a universal control parameter for EK turbulence, e.g. electric Rayleigh number (RaeRa_e). Although the RaeRa_e deduced from the Zhao-Wang model33 can evaluate the flow status in a highly-developed AC EK turbulence, they are not appropriate indicators to show how the EK flow evolves from its initial status, which could be more crucial for engineering applications.

The strong diversity of dimensionless control parameters, e.g. RaeRa_e, could be attributed to the broad factors that influencing the electric body force in EK flow, including the form, intensity and frequency of electric field, the geometry of the flow field, the distribution of electric conductivity and permittivity, pH value, electrode property and etc. The entangled dynamics and interfacial phenomenon make the problem more complicated to reach a common sense on the format of RaeRa_e.

In this investigation, a novel RaeRa_e that describing the transition of AC EK flow of binary fluids from stable to unstable states is theoretically analyzed and experimentally evaluated in a divergent microchannel. We focus on how RaeRa_e is influenced by the physical and chemical conditions, e.g. electric field intensity, AC frequency, electric conductivity ratio and interface width. A theoretical analysis on electric Rayleigh number is conduct first. Then, the theory is adopted to predict the critical electric Rayleigh number RaecRa_{ec} of AC EK flow in a divergent microchannel. This study aims to show the newly developed electric Rayleigh number can better elucidate the underlying physics of how the electric body force perturbs the interface of the binary fluids. With the electric Rayleigh number, people can more precisely predict RaeRa_e, facilitating the comparison of physical and chemical conditions of generating EK flow. This is important for enhancing our understanding of the mechanisms of EK turbulence and promoteing its application in microfluidics.

Theory and dimensionless parameters

Control equations

The EK flow can be described by Navier-Stokes equations as shown below34, 35

ρ(∂∂tu⃗+u⃗⋅∇u⃗)=−∇p+F⃗e+μ∇2u⃗(1)\rho\left(\frac{\partial}{\partial t}\vec{u}+\vec{u}\cdot\nabla\vec{u}\right)=-\nabla p+\vec{F}_e+\mu\nabla^2\vec{u} \tag*{(1)}
F⃗e=ρeE⃗−12(E⃗⋅E⃗)∇ε+12∇[ρE⃗⋅E⃗(∂ε∂ρ)T](2)\vec{F}_e=\rho_e\vec{E}-\frac{1}{2}(\vec{E}\cdot\vec{E})\nabla\varepsilon+\frac{1}{2}\nabla\left[\rho\vec{E}\cdot\vec{E}\left(\frac{\partial\varepsilon}{\partial\rho}\right)_T\right] \tag*{(2)}
ρe=∇⋅εE⃗(3)\rho_e=\nabla\cdot\varepsilon\vec{E} \tag*{(3)}
∂∂tσ+u⃗⋅∇σ=Dσ∇2σ(4)\frac{\partial}{\partial t}\sigma+\vec{u}\cdot\nabla\sigma=D_\sigma\nabla^2\sigma \tag*{(4)}

where ρ\rho is fluid density, μ\mu is dynamic viscosity, F⃗e\vec{F}_e is electric body force (EBF). ε\varepsilon and σ\sigma are electric permittivity and electric conductivity of the solution respectively, while ρe\rho_e is the charge density. The instantaneous velocity vector, u⃗=ui⃗+vj⃗+wk⃗\vec{u}=u\vec{i}+v\vec{j}+w\vec{k}, consist of uu, vv, ww as the velocity components in streamwise, spanwise and vertical directions respectively, where i⃗\vec{i}, j⃗\vec{j} and k⃗\vec{k} are the unit vectors in these corresponding directions. DσD_\sigma stands for the diffusion coefficient. Furthermore, we also assume the fluid to be incompressible, i.e. ∇⋅u⃗=0\nabla\cdot\vec{u}=0.

In this model, we consider that the electric field is perpendicular to the interface of the binary fluids, as diagramed in Figure 1, with the AC frequency (fff_f) much higher than the response frequency (frf_r) of the flow.36 The fluids share identical viscosity, temperature and electric permittivity, yet differ in electric conductivity, denoted as σ1\sigma_1 and σ2\sigma_2 (σ2≥σ1\sigma_2\geq\sigma_1).

Figure 1. Schematic diagram of electric conductivity distribution along y direction in initial stage. y1y_1 denotes the width with electric conductivity σ1\sigma_1. Δy\Delta y represents the interface width with electric conductivity gradient. d−y2d-y_2 is the width with electric conductivity σ2\sigma_2. Note that, given the flow rate is the same, here y1=d−y2y_1=d-y_2.

Since the fluid is incompressible, 12∇[ρE⃗⋅E⃗(∂ε∂ρ)T]=0\frac{1}{2}\nabla\left[\rho\vec{E}\cdot\vec{E}\left(\frac{\partial\varepsilon}{\partial\rho}\right)_T\right]=0.

Accordingly, (2) becomes

F⃗e=ρeE⃗−12(E⃗⋅E⃗)∇ε(5)\vec{F}_e=\rho_e\vec{E}-\frac{1}{2}(\vec{E}\cdot\vec{E})\nabla\varepsilon \tag*{(5)}

When an electric field is applied, initially, the interface between the binary fluids is slightly disturbed by the EBF. The ultrasmall flow disturbance has three consequences. First, the distribution of electric conductivity has negligible difference from the case without EBF. Second, the convective transport of charge is much weaker than that of the electric field. Therefore, the charge conservation equation37 becomes

∂ρe∂t+∇⋅σE⃗=0(6)\frac{\partial\rho_e}{\partial t}+\nabla\cdot\sigma\vec{E}=0 \tag*{(6)}

The third is that the disturbance of electric conductivity is negligible as well. Consequently, the electric field can be approximated as a 1D model in the yy-direction, EyE_y. Thus, Eq. (6) can be rewritten in the following as

∂∂yσ∗Ey=0(7)\frac{\partial}{\partial y}\sigma^{*}E_y=0 \tag*{(7)}

where σ∗=iωε+σ\sigma^{*}=i\omega\varepsilon+\sigma is the complex electric conductivity, with ω=2πff\omega=2\pi f_f denoting the angular frequency. So considering the complex Ohmic law, σ∗Ey=J∗\sigma^{*}E_y=J^{*}, it is constant for yy where J∗J^{*} is complex electric current density.

Electric body force

In EK flow, EBF is the most important quantities that determines the state of flow, however, influenced by various factors. At the initial stage, the time-averaged AC EBF per unit volume can be rewritten as34

F⃗e,ac=12Re[(∇ε⋅E⃗+ε∇⋅E⃗)⋅E⃗−12E2∇ε]\vec{F}_{e,ac}=\frac{1}{2}\mathop{\mathrm{Re}}\left[(\nabla\varepsilon\cdot\vec{E}+\varepsilon\nabla\cdot\vec{E})\cdot\vec{E}-\frac{1}{2}E^2\nabla\varepsilon\right]

where E⃗~\widetilde{\vec{E}} is the complex conjugation of electric field intensity and Re\mathop{\mathrm{Re}} is the real part. We are unable to use a local F⃗e,ac\vec{F}_{e,ac} to evaluate the influence of EBF on EK flow. To characterize the influence of EBF, especially aiding to engineering applications, a spatially averaged EBF36 on the interface with a width of Δy\Delta y is applied as

Fe,Δy=⟨Fe,ac⟩Δy=ε4Δy∣J∗∣2∣σ∗∣2∣y1y1+Δy(8)F_{e,\Delta y}=\langle F_{e,ac}\rangle_{\Delta y}=\frac{\varepsilon}{4\Delta y}\frac{|J^{*}|^2}{|\sigma^{*}|^2}\bigg|_{y_1}^{y_1+\Delta y} \tag*{(8)}

where ⟨⋅⟩Δy\langle\cdot\rangle_{\Delta y} indicates the spatial averaging on Δy\Delta y in the yy direction. The complex current density is conserved in space as

J∗=σΔy∗Ey,Δy=σd∗Ey,d(9)J^{*}=\sigma_{\Delta y}^{*}E_{y,\Delta y}=\sigma_d^{*}E_{y,d} \tag*{(9)}

where σΔy∗=1/⟨1/σ∗⟩Δy\sigma_{\Delta y}^{*}=1/\langle1/\sigma^{*}\rangle_{\Delta y} and Ey,Δy=⟨Ey⟩ΔyE_{y,\Delta y}=\langle E_y\rangle_{\Delta y} are the complex electric conductivity and electric field intensity averaged on the interface with Δy\Delta y respectively. Similarly, σd∗=1/⟨1/σ∗⟩d\sigma_d^{*}=1/\langle1/\sigma^{*}\rangle_d and Ey,d=EampeiωtE_{y,d}=E_{amp}e^{i\omega t} are the complex electric conductivity and electric field intensity averaged across the channel width dd respectively, where Eamp=Vamp/dE_{amp}=V_{amp}/d is the amplitude of the AC electric field and VampV_{amp} is the voltage amplitude between the two electrodes as shown in the Figure 1. Thus, ∣J∗∣2=J∗⋅J=σdσd∗Ey,dEy,d∗=σdσd∗Eamp2|J^{*}|^2=J^{*}\cdot J=\sigma_d\sigma_d^{*}E_{y,d}E_{y,d}^{*}=\sigma_d\sigma_d^{*}E_{amp}^2. σd∗\sigma_d^{*} is strictly dependent on ω\omega, ε\varepsilon and the spatial distribution of σ\sigma.

Here, we consider an initial case of the EK flow, where the EBF is too weak to induce a perturbation of σ\sigma. Thus, the distribution of σ\sigma can be approximated to be steady. Furthermore, we utilize a tanh distribution (Fig. 1) of electric conductivity to demonstrate how the EBF and the corresponding dimensionless parameters is influenced by the physical quantities. If the electric conductivity has a tanh distribution as below

σ(y)=σ1+e(2y−d)/yse(2y−d)/ys+1(σ2−σ1)(10)\sigma(y)=\sigma_1+\frac{e^{(2y-d)/y_s}}{e^{(2y-d)/y_s}+1}(\sigma_2-\sigma_1) \tag*{(10)}

where ysy_s is a shape factor to characterize the distribution of σ\sigma. In the tanh model, to simulate an interface width of Δy\Delta y, approximately we have ys≈Δy/10y_s\approx\Delta y/10. According to Eq. (10), σd∗\sigma_d^{*} can be calculated as

σd∗=11⟨1/σ∗⟩d=2dσ1∗σ2∗ys(σ1−σ2)ln⁡σ1∗+σ2∗ed/ysσ1∗+σ2∗e−d/ys+2dσ2∗(11)\sigma_d^{*}=\frac{1}{\frac{1}{\langle1/\sigma^{*}\rangle_d}}=\frac{2d\sigma_1^{*}\sigma_2^{*}}{y_s(\sigma_1-\sigma_2)\ln\frac{\sigma_1^{*}+\sigma_2^{*}e^{d/y_s}}{\sigma_1^{*}+\sigma_2^{*}e^{-d/y_s}}+2d\sigma_2^{*}} \tag*{(11)}

Furthermore, according to σ∗=σ+iωε\sigma^{*}=\sigma+i\omega\varepsilon and Eq. (10), we have

1∣σ∗∣2∣y1y1+Δy=16(1−λ2)Btanhσ12Atanh(12)\frac{1}{|\sigma^{*}|^2}\bigg|_{y_1}^{y_1+\Delta y}=16\frac{(1-\lambda^2)B_{tanh}}{\sigma_1^2A_{tanh}} \tag*{(12)}

where λ=σ2/σ1\lambda=\sigma_2/\sigma_1 is electric conductivity ratio, χ=2ωε/(σ1+σ2)\chi=2\omega\varepsilon/(\sigma_1+\sigma_2) is a dimensionless frequency, m=χ(1+λ)m=\chi(1+\lambda), Atanh=[(4eΔy/ysλ+1)2+m2(eΔy/ys+1)2][(eΔy/ys+λ)2+m2(eΔy/ys+1)2]A_{tanh}=[(4e^{\Delta y/y_s}\lambda+1)^2+m^2(e^{\Delta y/y_s}+1)^2][(e^{\Delta y/y_s}+\lambda)^2+m^2(e^{\Delta y/y_s}+1)^2] and Btanh=(e2Δy/ys+1)(eΔy/ys+1)2B_{tanh}=(e^{2\Delta y/y_s}+1)(e^{\Delta y/y_s}+1)^2. By substituting Eq. (12) into (8), with binomial expansion, we have

Fe,Δy=εEamp2Btanh(1−λ2)(4λ2+m2)(4+m2)ΔyAtanhCtanh(13)F_{e,\Delta y}=\frac{\varepsilon E_{amp}^2B_{tanh}(1-\lambda^2)(4\lambda^2+m^2)(4+m^2)}{\Delta y A_{tanh}C_{tanh}} \tag*{(13)}

where Ctanh=(1−λ)2(ys/d)2qtanhptanh+(ys/d)(1−λ)(2λ−im)qtanh+(ys/d)(1−λ)(2λ+im)ptanh+4λ2m2C_{tanh}=(1-\lambda)^2(y_s/d)^2q_{tanh}p_{tanh}+(y_s/d)(1-\lambda)(2\lambda-im)q_{tanh}+(y_s/d)(1-\lambda)(2\lambda+im)p_{tanh}+4\lambda^2m^2, with qtanh=ln⁡2+im+(2λ+im)ed/ys2+im+(2λ+im)e−d/ysq_{tanh}=\ln\frac{2+im+(2\lambda+im)e^{d/y_s}}{2+im+(2\lambda+im)e^{-d/y_s}} and ptanh=ln⁡2−im+(2λ−im)ed/ys2−im+(2λ−im)e−d/ysp_{tanh}=\ln\frac{2-im+(2\lambda-im)e^{d/y_s}}{2-im+(2\lambda-im)e^{-d/y_s}}

Linear modelTanh model
Schematic diagram of electric conductivity distribution[Schematic diagram of electric conductivity distribution][Schematic diagram of electric conductivity distribution]
Electric conductivity distribution σ\sigma{σ1,0≤y<y1σ2−σ1Δy(y−y1)+σ1,y1≤y<y2σ2,y2≤y≤d\begin{cases} \sigma_1, & 0 \le y < y_1 \\ \dfrac{\sigma_2-\sigma_1}{\Delta y}(y-y_1)+\sigma_1, & y_1 \le y < y_2 \\ \sigma_2, & y_2 \le y \le d \end{cases}σ1+e(2y−d)/yse(2y−d)/ys+1(σ2−σ1)\sigma_1+\dfrac{e^{(2y-d)/y_s}}{e^{(2y-d)/y_s}+1}(\sigma_2-\sigma_1)
Fe,ΔyF_{e,\Delta y}4εEamp2Δy(1−λ)2(1−λ2)Alinear\dfrac{4\varepsilon E_{amp}^{2}}{\Delta y}\dfrac{(1-\lambda)^2(1-\lambda^2)}{A_{linear}}εEamp2ΔyBtanh(1−λ2)(4λ2+m2)(4+m2)AtanhCtanh\dfrac{\varepsilon E_{amp}^{2}}{\Delta y}\dfrac{B_{tanh}(1-\lambda^2)(4\lambda^2+m^2)(4+m^2)}{A_{tanh}C_{tanh}}
RaeRa_e4εEamp2Δy2ρνDe(1−λ)2(1−λ2)Alinear\dfrac{4\varepsilon E_{amp}^{2}\Delta y^2}{\rho\nu D_e}\dfrac{(1-\lambda)^2(1-\lambda^2)}{A_{linear}}εEamp2Δy2ρνDeBtanh(1−λ2)(4λ2+m2)(4+m2)AtanhCtanh\dfrac{\varepsilon E_{amp}^{2}\Delta y^2}{\rho\nu D_e}\dfrac{B_{tanh}(1-\lambda^2)(4\lambda^2+m^2)(4+m^2)}{A_{tanh}C_{tanh}}
Alinear=4(1−Δy/d)2(λ−1)2(λ2+2λ+1+m2)+2Δy/d(1−Δy/d)(λ−1)[2(qlinear+plinear)+im(qlinear−plinear)](m2+2+2λ2)+(Δy/d)2qlinearplinear(16λ2+4m2+4m2λ2+m4)A_{linear}=4(1-\Delta y/d)^2(\lambda-1)^2(\lambda^2+2\lambda+1+m^2)+2\Delta y/d(1-\Delta y/d)(\lambda-1)[2(q_{linear}+p_{linear})+im(q_{linear}-p_{linear})](m^2+2+2\lambda^2)+(\Delta y/d)^2q_{linear}p_{linear}(16\lambda^2+4m^2+4m^2\lambda^2+m^4)<br>qlinear=ln⁡2λ+im2+imq_{linear}=\ln\dfrac{2\lambda+im}{2+im}<br>plinear=ln⁡2λ−im2−imp_{linear}=\ln\dfrac{2\lambda-im}{2-im}Atanh=[4(eΔy/ysλ+1)2+m2(eΔy/ys+1)2][4(eΔy/ys+λ)2+m2(eΔy/ys+1)2]A_{tanh}=[4(e^{\Delta y/y_s}\lambda+1)^2+m^2(e^{\Delta y/y_s}+1)^2][4(e^{\Delta y/y_s}+\lambda)^2+m^2(e^{\Delta y/y_s}+1)^2]<br>Btanh=(e2Δy/ys−1)(eΔy/ys+1)2B_{tanh}=(e^{2\Delta y/y_s}-1)(e^{\Delta y/y_s}+1)^2<br>Ctanh=(1−λ)2(ys/d)2qtanhptanh+(ys/d)(1−λ)(2λ−im)qtanh+(ys/d)(1−λ)(2λ+im)ptanh+4λ2C_{tanh}=(1-\lambda)^2(y_s/d)^2q_{tanh}p_{tanh}+(y_s/d)(1-\lambda)(2\lambda-im)q_{tanh}+(y_s/d)(1-\lambda)(2\lambda+im)p_{tanh}+4\lambda^2<br>qtanh=ln⁡2+im+(2λ+im)ed/ys2+im+(2λ+im)e−d/ysq_{tanh}=\ln\dfrac{2+im+(2\lambda+im)e^{d/y_s}}{2+im+(2\lambda+im)e^{-d/y_s}}<br>ptanh=ln⁡2−im+(2λ−im)ed/ys2−im+(2λ−im)e−d/ysp_{tanh}=\ln\dfrac{2-im+(2\lambda-im)e^{d/y_s}}{2-im+(2\lambda-im)e^{-d/y_s}}

Table 2. Comparison between the linear model and the tanh model.

Electric Grashof number and electric Rayleigh number

The electric Grashof number and the electric Rayleigh number are two commonly used dimensionless parameters to describe the state of EK flow. The former represents the ratio of EBF relative to viscous force, while the latter further considers the influence of ion diffusion. According to Eq. (13), the electric Grashof number38 (GreGr_e) can be quantified by

Gre=Fe,ΔyΔy3ρν2=εEamp2Δy2Btanh(1−λ2)(4λ2+m2)(4+m2)ρν2AtanhCtanh(14)\begin{aligned} Gr_e=\frac{F_{e,\Delta y}\Delta y^3}{\rho\nu^2} \\ &=\frac{\varepsilon E_{amp}^2\Delta y^2B_{tanh}(1-\lambda^2)(4\lambda^2+m^2)(4+m^2)}{\rho\nu^2A_{tanh}C_{tanh}} \tag*{(14)} \end{aligned}

The electric Rayleigh number (RaeRa_e) is further obtained according to Rae=GreScRa_e=Gr_eSc, where Sc=ν/DσSc=\nu/D_\sigma is Schmidt number. Then, we have

Rae=εEamp2Δy2Btanh⁡(1−λ2)(4λ2+m2)(4+m2)ρνDeAtanh⁡Ctanh⁡(15)Ra_e=\frac{\varepsilon E_{\mathrm{amp}}^{2}\Delta y^{2}B_{\tanh}(1-\lambda^{2})(4\lambda^{2}+m^{2})(4+m^{2})}{\rho\nu D_e A_{\tanh}C_{\tanh}} \tag*{(15)}

RaeRa_e in Eq. (15) is evaluated from the overall distribution of electric conductivity. It complexly interplays with λ\lambda, χ\chi and the initial distribution (through y1y_1, y2y_2 and Δy\Delta y) of electric conductivity, in a relatively nonintuitive manner. This is apparently distant from the previous investigations (see Table I), where the electric Rayleigh number are simply evaluated through either simple local quantities or a linear distribution of control scalars. For comparison, a linear model for electric conductivity distribution is also advanced in this investigation, as summarized in Table II.

Numerical calculation of RaeRa_e in both linear and tanh model

In this section, as indicated by RaeRa_e in the Table II, we demonstrate how the following parameters, e.g. EampE_{\mathrm{amp}}, Δy\Delta y, λ\lambda and χ\chi, contribute to the variations of RaeRa_e. Both the linear mode and tanh model are theoretically compared to reveal their difference on calculating RaeRa_e. To be consistent to experiments, some relevant parameters are given as: d=650d=650 μm, ε\varepsilon is 7.1×10−107.1\times10^{-10} F/m, ρ\rho is 10310^{3} kg/m3^{3}, ν\nu is 10−610^{-6} m2^{2}/s, σ1+σ2\sigma_1+\sigma_2 is 2.395×10−12.395\times10^{-1} S/m and DσD_\sigma is 5×10−105\times10^{-10} m2^{2}/s.

Electric conductivity ratio

First, the influence of λ\lambda on RaeRa_e in the initial stage of EK flow is analyzed, as shown in Figure 2. Under the linear model, as depicted in Figure 2(a), RaeRa_e increases rapidly and monotonically with increasing λ\lambda, eventually reaching a saturation point at λ=O(102)\lambda=O(10^{2}). This indicates a higher probability of reaching an unstable state at higher λ\lambda, accordingly, easier to get into a transition. There is no optimal λ\lambda related to the highest RaeRa_e in the linear model.

Plots of $Ra_e$ versus $\lambda$ under linear and tanh models

Figure 2. RaeRa_e calculated from both linear and tanh models under different χ\chi, where d=650d=650 μm, Δy=4Dσt=11.4\Delta y=\sqrt{4D_\sigma t}=11.4 μm and Eamp=3.1×104E_{\mathrm{amp}}=3.1\times10^{4} V/m, where t=x/Uet=x/U_e is the diffusion time of the interface. Here, UeU_e is the mean flow velocity at the centerline of the microchannel. x=0.2dx=0.2d is the measured position. All these data are consistent to the experimental conditions in the following. (a) RaeRa_e calculated by the linear model vs λ\lambda. (b) RaeRa_e calculated by the tanh model vs λ\lambda.

However, under the tanh model, RaeRa_e first increases with λ\lambda and then decreases at relatively smaller χ\chi, as depicted in Figure 2(b). This model provides us with the possibility of identifying an optimal λ\lambda to reach the highest RaeRa_e, for effectively perturbing flow and improving momentum and scalar transport. Relative to the linear model without optimal λ\lambda, the tanh model that predicts optimal λ\lambda can provide better consistency with previous experiments. For instance, Wang et al.38 found when the applied AC frequency was 100 kHz, i.e. χ=3.7×10−3\chi=3.7\times10^{-3}, an ultrafast EK mixing can be achieved with an optimal λ=5000\lambda=5000. Nan et al.21 also observed different optimal λ\lambda for perturbating the electric conductivity interface under different χ\chi.

From Figure 2(b), it should be noted that only when χ≥3.7×10−3\chi\geq3.7\times10^{-3} (i.e. ff=100f_f=100 kHz), RaeRa_e increases monotonically with λ\lambda. To avoid the violation of τrex≪1/ω\tau_{\mathrm{rex}}\ll1/\omega (where τrex=1.7×10−5\tau_{\mathrm{rex}}=1.7\times10^{-5} is the charge relaxation time)36, a larger χ\chi exceeding 3.7×10−23.7\times10^{-2} is not applied in this study. In the range 1.1×10−6≪χ≪3.7×10−21.1\times10^{-6}\ll\chi\ll3.7\times10^{-2}, we surprisingly find a special value of λc≈200\lambda_c\approx200 by the tanh model. When λ≤λc\lambda\leq\lambda_c, RaeRa_e is insensitive to χ\chi.

AC frequency

The insensitivity of RaeRa_e to χ\chi at λ≤λc\lambda\leq\lambda_c can be more visible in Figure 3(b). All the plots at λ≤λc\lambda\leq\lambda_c are exactly flat lines in the wide range of χ\chi. However, when λ>λc\lambda>\lambda_c, e.g. 2000 and 5400, RaeRa_e changes significantly with χ\chi. It stays almost unchanged with χ\chi first, then rapidly increases until reaching another plateau.

Plots of $Ra_e$ calculated from linear and tanh models versus $\chi$ for different $\lambda$

Figure 3. RaeRa_e calculated from both linear and tanh models under different λ\lambda, where d=650d = 650 μm, Δy=11.4\Delta y = 11.4 μm and Eamp=3.1×104E_{amp} = 3.1 \times10^{4} V/m. (a) RaeRa_e calculated by the linear model vs χ\chi. (b) RaeRa_e calculated by the tanh model vs χ\chi.

In contrast, the variation of RaeRa_e with χ\chi is not observed in Figure 3(a), indicating the linear model has no capability to reveal the influence of χ\chi (or AC frequency) on RaeRa_e. This accounts for why in previous experiments21, even though ultrafast EK mixing can be observed at χ>3.7×10−4\chi> 3.7 \times10^{-4} (or ff>10f_f > 10 kHz), we still cannot get an apparently larger RaeRa_e through the linear model.

Interface width

As illustrated by a series of investigations21, 22, 38, a larger gradient of electric conductivity is crucial for generating a larger EBF, accordingly a smaller Δy\Delta y is preferred. Thus, a smaller Δy\Delta y is corresponding to a larger RaeRa_e. For this purpose, we theoretically studied the effect of Δy\Delta y on RaeRa_e. As depicted in Figure 4, both models predict a larger RaeRa_e with increasing Δy\Delta y as anticipated. The distinction lies in the fact that the tanh model can better distinguish the influence of Δy\Delta y, as can be observed in Figure 4(a, Figure 4b).

Figure 4. RaeRa_e calculated from both linear and tanh models under different χ\chi, where d=650d = 650 μm, λ=5400\lambda= 5400 and Eamp=3.1×104E_{amp} = 3.1 \times10^{4} V/m. (a) RaeRa_e calculated by the linear model vs Δy\Delta y. (b) RaeRa_e calculated by the tanh model vs Δy\Delta y.

Electric field intensity

In contrast to λ\lambda, χ\chi and Δy\Delta y which have complex relationship with RaeRa_e, the quadratic effect of electric field

intensity is quite straightforward, i.e. Rae∼Eamp2Ra_e \sim E_{amp}^{2} according to Eq. (15). This is also observable from Figure 5(a, Figure 5b). The difference between the linear and tanh model is that the latter is distinguishable for different χ\chi, while the former is not.

Plots of $Ra_e$ calculated from linear and tanh models versus $E$ for different $\chi$

Figure 5. RaeRa_e calculated from both linear and tanh models under different Δy\Delta y, where d=650d = 650 μm, χ=3.7×10−3\chi= 3.7 \times10^{-3} and λ=5400\lambda= 5400. (a) RaeRa_e calculated by the linear model vs EE. (b) RaeRa_e calculated by the tanh model vs EE.

Experimental setup and results

Experimental setup

In this section, relying on the electric Rayleigh number RaeRa_e under the tanh model, we experimentally investigate the conditions when EK flow becomes unstable under AC electric field. To this end, the EK flow velocity is measured by laser induced fluorescence photobleaching anemometer (LIFPA), which has ultra-high spatial (~203 nm) and temporal (~4 μs) resolution, as diagramed in Figure 6(a). The system uses a 405nm continuous wave (CW) laser, with a laser power of 11 mW, as light source. An acousto-optic modulator (AOM) is applied to realize temporal control of the beam. The beam in turn passes through a special light filter (SLF) and a dichroic mirror (DM), and is then reflected into an objective lens (OL) to excite the fluorescent solution in the EK flow microchip. The fluorescent solution is Coumarin 102. The fluorescence emitted by the solution is filtered out by two bandpass filters (BP, 470/100 and 470/10 nm), collected by an optical fiber (OF), measured by a photon counter (PT) and finally processed by a computer (PC).

Experimental system

Figure 6. Experimental system. (a) The diagram of LIFPA system. Laser: 405 nm continuous wave laser, AOM: Acousto-optic modulator, SLF: Spatial light filter, M1 and M2: mirrors, DM: Dichroic mirror, BP1 and BP2: Bandpass filters, FM: flipping mount mirror, OF: Photon counter, PT: Photon counter, PC: computer, FG: Function generator, HVA: High voltage amplifier, SP: Syringe pump, OL: Objective lens, CMOS: camera. (b) Structure diagram of the chip. (c) Flow visualization of the pressure-driven basic flow without electric field. x0x_0 is the measurement position. x1x_1 is the velocity calibration position. (d) The spanwise distribution of velocity in x1x_1 when QQ = 5 μL/min. (e) The velocity calibration in the center of channel. (f) The velocity time trace of χ\chi = 1.8 × 10−410^{-4}, λ\lambda = 5400 and Δy\Delta y = 11.4 μm.

The concentration of Coumarin102 solution is 100 μmol/L. It is prepared by anhydrous ethanol and deionized water. Phosphate buffer solution (PBS) is applied to adjust the electric conductivity of the two solutions to reach λ\lambda =1:5400, 1:2000, 1:200 and 1:2 respectively.

Figure 6(b) illustrates the schematic diagram of the EK flow microchip. The microchannel is 100 μm in high (hh), 18 mm in long and 650 μm in width between two electrodes. The two streams with different electric conductivity are injected separately from inlets 1 and 2 by a syringe pump. Consequently, a pronounced electrical conductivity gradient is established immediately following the trailing edge.

The measurement point (x0x_0 in Figure 6(c)) is 150 μm downstream of the trailing edge. x1x_1 serves as the velocity calibration location, which is about 2 mm downstream of the trailing edge. At this position, the pressure-driven basic flow

is stable and fully developed. The velocity profile at the velocity calibration point is numerically simulated by COMSOL Multiphysics, as shown in Figure 6(d), to reveal the relationship between fluorescent intensity and velocity39. A velocity calibration curve is plotted in Figure 6(e). Figure 6(f) shows a typical velocity time trace.

Through the analysis and comparison of the above two models, it is clear that the tanh model contains richer and more complex information about the influence of λ\lambda, χ\chi and Δy\Delta y on RaeRa_e. To demonstrate the effectiveness of the tanh model, we studied the critical electric Rayleigh number (RaecRa_{ec}), when the EK flow becomes unstable, under different control parameters. RaecRa_{ec} is determined from the power spectrum (E(f)E(f)) of the velocity fluctuations, as shown in Figure 7(a). When no electric field was applied, a peak in E(f)E(f), referred to as EpeakE_{peak}, is observed attributing to the perturbation of the wake after the trailing edge. When a smaller AC electric field was applied, i.e. Rae<RaecRa_e < Ra_{ec}, EpeakE_{peak} is equivalent to that observed without electric field. While AC electric field is over some threshold value, Rae>RaecRa_e > Ra_{ec}, EpeakE_{peak} increases rapidly with RaeRa_e, as plotted in Figure 7(b), indicating an amplification of the flow perturbation induced by EBF. Through this way, RaecRa_{ec} can be determined (Figure 7(b)).

Figure 7. Velocity power spectrum, EpeakE_{peak} and the determination of RaecRa_{ec}. (a) Velocity spectrum E(f)E(f) at χ=1.1×10−6\chi= 1.1 \times10^{-6}, λ=5400\lambda= 5400 and Δy=11.4 μm\Delta y = 11.4\ \mu\mathrm{m}. (b) EpeakE_{peak} vs RaeRa_e showing the position of RaecRa_{ec} at χ=1.1×10−6\chi= 1.1 \times10^{-6}.

Experimental results

The relevant experimental (the symbols) and theoretical (the dotted lines) results on RaecRa_{ec} are plotted in Figure 8, Figure 9 and Figure 10 respectively. First, we analyze the influence of λ\lambda on RaecRa_{ec}, as plotted in Figure 8. When χ=1.1×10−6\chi= 1.1 \times10^{-6}, RaecRa_{ec} increases with λ\lambda first before subsequently decreasing in both experimental and theoretical results. Although the experimental one exhibits stronger bending relative to the theoretical one, their variation shows consistent trending. The consistency between experimental and theoretical results increases as χ\chi increases, e.g. at χ=1.8×10−5\chi= 1.8 \times10^{-5}. When χ=3.7×10−3\chi= 3.7 \times10^{-3}, RaecRa_{ec} is monotonically increasing with λ\lambda. All the experimental trends align with the theoretical results predicted by the tanh model. In contrast, the most used model, i.e. linear distribution of σ\sigma, fails to provide even a qualitatively consistent prediction.

Figure 8. RaecRa_{ec} vs λ\lambda under different χ\chi with Δy=9.4 μm\Delta y = 9.4\ \mu\mathrm{m}.

Figure 9. RaecRa_{ec} vs χ\chi under different λ\lambda with Δy=10.4 μm\Delta y = 10.4\ \mu\mathrm{m}.

Figure 10. (a) RaecRa_{ec} vs Δy\Delta y under different λ\lambda with χ=1.1×10−6\chi= 1.1 \times10^{-6}. (b) RaecRa_{ec} vs ReRe under different λ\lambda with χ=1.1×10−6\chi= 1.1 \times10^{-6}.

Then, we further demonstrate the tanh model can effectively predict the influence of χ\chi on RaecRa_{ec}. Both the experimental and theoretical results are plotted in Figure 9. The experimental results indicate, when λ≤200\lambda\leq200, RaecRa_{ec} is almost unchanged with χ\chi, which is consistent to the theoretical results. When λ>200\lambda> 200, the tanh model further reveals the complex variation of RaecRa_{ec} with χ\chi, such that RaecRa_{ec} keeps constant first, then increases until reaching a plateau. The results again support the effectiveness of the tanh model.

The parameter Δy\Delta y is crucial to the gradient of electric conductivity, which in turn, determines EBF and the corresponding electric Rayleigh number. From Figure 10(a), the experimental results of RaecRa_{ec} are monotonically increasing with Δy\Delta y, which is qualitatively consistent with the theoretical prediction by Eq. (15). Considering a larger Δy\Delta y is corresponding to a smaller electric conductivity gradient and smaller EBF, it indicates that the flow is more difficult to be disturbed at a larger Δy\Delta y under certain Reynolds number.

Finally, we also studied the influence of flow rate on RaecRa_{ec}, through bulk flow Reynolds number Re=UL/νRe = UL/\nu, where L=2dh/(d+h)L = 2dh/(d+h) is the hydraulic diameter of the EK micromixers and UU is the bulk flow velocity. From Figure 10(b), it can be seen that RaecRa_{ec} has an inverse relationship with ReRe, indicating a decreasing requirement of AC electric field. Under a certain electric field, it is easier to generate unstable EK flow at a higher ReRe.

Discussion

Dimensionless parameters are vital for engineering applications where complex relations among multiple control parameters are interplayed. An effective dimensionless parameter that provides insight into the underlying relationship among the control parameters can also enhance our understanding of the complex physical mechanisms in an engineering system. Unfortunately, when more control parameters are taken into account, it is significantly difficult to reach an effective dimensionless parameter. Part of the researchers believe that the pursuit of universal dimensionless parameter could be meaningless.

From this investigation, we aspire to show that, if a complex dimensionless parameter is appropriately expressed, it not only evaluates the status of the system, but also predicts (even though qualitatively) how the system changes. By using a more accurate distribution of electric conductivity, i.e. tanh model, the existence of optimal λ\lambda and χ\chi can be predicted through the electric Rayleigh number. The AC EK flow in a divergent microchannel can be unstable with Raec=O(10−1)Ra_{ec} = O(10^{-1}). In contrast, the critical electric Rayleigh numbers in previous reports, e.g. Chen et al23 and Nan et al21, are estimated to be 4.6×1034.6 \times10^3 and 92.692.6 respectively, if using the tanh model in this investigation.

On the other hand, it should be noted that the current experimental and theoretical results, e.g. on the influence of Δy\Delta y, still remain apparent discrepancies. This could be attributed to that the experimental environment is more complex relative to the theoretical model. Δy\Delta y of the electric conductivity interface varies along streamwise, making the electric conductivity distribution in experiments intrinsically 2D or more. To change Δy\Delta y, we change the flow rates of the two streams, which leads to the change of ReRe as well. It is almost impossible for us to isolate Δy\Delta y and ReRe from each other in experiments. Thus, discrepancies become inevitable.

Conclusion

In this investigation, we theoretically derive the expression of electric Rayleigh number RaeRa_e for AC electrokinetic flow of binary fluids, on the basis of both linear and tanh distribution model of electric conductivity. The influence of electric conductivity ratio, AC frequency, interface width and electric field intensity on RaeRa_e is systematically investigated in the tanh model, and compared with the linear model. Through comparison, it can be found that the tanh model provides more information about the variation of RaeRa_e with the multiple parameters. In the tanh model, RaeRa_e is more sensitive to electric conductivity ratio, AC frequency and interface width.

Furthermore, we conduct an experimental investigation on the critical electric Rayleigh number of the EK flow evaluated by the tanh model, through velocity measurement by LIFPA. The experimental findings exhibit an acceptable agreement with the theoretical predictions, thereby validating the efficacy of the electric Rayleigh number within the context of the tanh model.

An appropriate dimensionless parameter is not just a value, but an indicator of the underlying physical mechanism to guide engineering applications. In this paper, we propose a more accurate electric Rayleigh number to evaluate the state of AC EK flow, for designation of micromixers/reactors by electrokinetic approaches.

ACKNOWLEDGEMENTS

This research was funded by National Natural Science Foundation of China, grant number 51927804.

References

  1. [1]¹K. Karthikeyan, S. K. Kandasamy, P. Saravanan and A. Alodhayb, "Numerical simulation and parameter optimization of micromixer device using fuzzy logic technique," RSC Adv. 13, 4504-4522 (2023).DOI
  2. [2]²D. S. Kim, S. H. Lee, C. H. Ahn, J. Y. Lee and T. H. Kwon, "Disposable integrated microfluidic biochip for blood typing by plastic microinjection moulding," Lab Chip 6, 794-802 (2006).pubmed.ncbi.nlm.nih.gov/16738733
  3. [3]³V. E. Papadopoulos, I. N. Kefala, G. Kaprou, G. Kokkoris, D. Moschou, G. Papadakis, E. Gizeli and A. Tserepi, "A passive micromixer for enzymatic digestion of DNA," Microelectron. Eng. 124, 42-46 (2014).DOI
  4. [4]⁴Z. Geng, S. L. Li, L. X. Zhu, Z. Cheng, M. Jin, B. X. Liu, Y. Guo and P. Liu, "\"Sample-to-Answer\" Detection of Rare ctDNA Mutation from 2 mL Plasma with a Fully Integrated DNA Extraction and Digital Droplet PCR Microdevice for Liquid Biopsy," Anal. Chem. 92, 7240-7248 (2020).DOI
  5. [5]⁵S. S. Wang, Z. J. Li, S. Sheng, F. A. Wu and J. Wang, "Microfluidic biocatalysis enhances the esterification of caffeic acid and methanol under continuous-flow conditions," J. Chem. Technol. Biotechnol. 91, 555-562 (2016).DOI
  6. [6]⁶H. Z. Sun, Y. K. Ren, Y. Tao, T. Y. Jiang and H. Y. Jiang, "Three-Fluid Sequential Micromixing-Assisted Nanoparticle Synthesis Utilizing Alternating Current Electrothermal Flow," Ind. Eng. Chem. Res. 59, 12514-12524 (2020).DOI
  7. [7]⁷P. A. Auroux, D. Iossifidis, D. R. Reyes and A. Manz, "Micro total analysis systems. 2. Analytical standard operations and applications," Anal. Chem. 74, 2637-2652 (2002).pubmed.ncbi.nlm.nih.gov/12090654
  8. [8]⁸H. S. Santana, J. L. Silva, A. G. P. da Silva, A. C. Rodrigues, R. D. Amaral, D. Noriler and O. P. Taranto, "Development of a New Micromixer \"Elis\" for Fluid Mixing and Organic Reactions in Millidevices," Ind. Eng. Chem. Res. 60, 9216-9230 (2021).DOI
  9. [9]⁹C. Y. Lee and L. M. Fu, "Recent advances and applications of micromixers," Sens. Actuator B-Chem. 259, 677-702 (2018).DOI
  10. [10]¹⁰K. Illath, S. Kar, P. Gupta, A. Shinde, S. Wankhar, F. G. Tseng, K. T. Lim, M. Nagai and T. S. Santra, "Microfluidic nanomaterials: From synthesis to biomedical applications," Biomaterials 280, 33 (2022).DOI
  11. [11]¹¹G. Dutta, J. Rainbow, U. Zupancic, S. Papamathiou, P. Estrela and D. Moschou, "Microfluidic Devices for Label-Free DNA Detection," Chemosensors 6, 20 (2018).DOI
  12. [12]¹²X. Y. Chen and T. C. Li, "A novel passive micromixer designed by applying an optimization algorithm to the zigzag microchannel," Chem. Eng. J. 313, 1406-1414 (2017).DOI
  13. [13]¹³K. Karthikeyan, L. Sujatha and N. M. Sudharsan, "Numerical Modeling and Parametric Optimization of Micromixer for Low Diffusivity Fluids," Int. J. Chem. React. Eng. 16, 11 (2018).DOI
  14. [14]¹⁴M. Juraeva and D. J. Kang, "Mixing Enhancement of a Passive Micromixer with Submerged Structures," Micromachines 13, 15 (2022).DOI
  15. [15]¹⁵G. J. Liu, M. Wang, L. T. Dong, D. Y. Zhu, C. H. Wang, Y. H. Jia, X. B. Li and J. B. Wang, "A novel design for split-and-recombine micromixer with double-layer Y-shaped mixing units," Sens. Actuator A-Phys. 341, 8 (2022).DOI
  16. [16]¹⁶K. Karthikeyan and L. Sujatha, "Study of Permissible Flow Rate and Mixing Efficiency of the Micromixer Devices," Int. J. Chem. React. Eng. 17, 15 (2019).DOI
  17. [17]¹⁷A. D. Stroock, S. K. W. Dertinger, A. Ajdari, I. Mezic, H. A. Stone and G. M. Whitesides, "Chaotic mixer for microchannels," Science 295, 647-651 (2002).pubmed.ncbi.nlm.nih.gov/11809963
  18. [18]¹⁸H. Jalili, M. Raad and D. A. Fallah, "Numerical study on the mixing quality of an electroosmotic micromixer under periodic potential," Proc. Inst. Mech. Eng. Part C-J. Eng. Mech. Eng. Sci. 234, 2113-2125 (2020).
  19. [19]¹⁹D. Bahrami, A. A. Nadooshan and M. Bayareh, "Numerical study on the effect of planar normal and Halbach magnet arrays on micromixing," Int. J. Chem. React. Eng. 18, 11 (2020).DOI

Paper details

vlnc:W_q57y4a8c
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents