Abstract
The study of flow and particle dynamics in microfluidic crossslot channels is of high relevance for labonachip applications. In this work, we investigate the dynamics of a rigid spherical particle in a crossslot junction for a channel heighttowidth ratio of 0.6 and at a Reynolds number of 120 for which a steady vortex exists in the junction area. Using an inhouse immersedboundarylatticeBoltzmann code, we analyse the effect of the entry position of the particle in the junction and the particle size on the dynamics and trajectory shape of the particle. We find that the dynamics of the particle depend strongly on its lateral entry position in the junction and weakly on its vertical entry position; particles that enter close to the centre show trajectory oscillations. Larger particles have longer residence times in the junction and tend to oscillate less due to their confinement. Our work contributes to the understanding of particle dynamics in intersecting flows and enables the design of optimised geometries for cytometry and particle manipulation.
Introduction
The deformability of blood cells is a valuable labelfree biomarker to indicate changes in the cell state^{1}. Changes in the nucleus or cytoskeleton of cells are associated with various diseases, including malaria^{2,3}, sickle cell disease^{4}, cancer^{5} and sepsis^{6}. For example, oral cancer cells have been found to be more deformable than healthy oral epithelial cells^{7}, thus facilitating metastasis^{8}. Leucocytes have been shown to become stiffer or slower, passing through pores upon stimulation with chemokines associated with infection, such as sepsis^{9}. Conventional methods for singlecell analysis, for instance, atomic force microscopy^{10}, optical stretching^{11} and micropipette aspiration^{12}, are useful for fundamental research in mechanobiology, but their low throughput limits the potential for clinical applications.
Fast and accurate analysis of blood samples is important for identifying diseased cells, thus accelerating clinical decisionmaking^{13}. Inertial microfluidics (IMF) has been widely used for cell manipulation, separation, and characterisation in biomedical applications^{14}. IMF operates in a flow regime where inertial forces influence the fluid and particle dynamics^{15}. In the 1960s, Segré and Silberberg^{16,17} observed radial migration of rigid spherical particles towards an annulus at around 60% of the radius of a tube when the Reynolds number is of the order 10–100. Since then, several studies investigated the nature of the inertial lift forces in straight channels^{18,19,20}. In the 2000s, it was recognised that the SegréSilberberg effect could be exploited in a microfluidic setting^{21} and in more complex channel geometries, such as curved^{22,23}, serpentine^{24,25} and spiral channels^{26,27,28}.
Since its inception, IMF has played an important role in imagebased deformability cytometry (DC). DC uses different channel geometries, such as linear channels^{29}, constrictions^{30} and crossslot channels^{31}. While traditional cytometry techniques, such as atomic force microscopy, micropipette aspiration and optical stretchers, are limited to low cell processing rates (10–100 cells/h)^{29}, DC yields significantly higher throughput (around 10–1000 cells/s)^{32}, making it more suitable for diagnostic applications where large quantities of cells should be measured to have reliable measurements.
Crossslot channels (Fig. 1) are particularly appealing for DC since the flow field in their junction features both high stresses, therefore deforming cells, and a relatively low flow speed, thus facilitating optical imaging. The fluid flow in crossslot junctions is complex and has been extensively investigated. For specific ranges of channel aspect ratio and Reynolds number, a spiral vortex can appear in the junction^{33}, which has been explained by a symmetrybreaking bifurcation above a critical Reynolds number for a given channel aspect ratio^{34}. Furthermore, Burshtein et al.^{35} determined the characteristics of a steadystate vortex in crossslot flows at higher Reynolds numbers, such as the vortex core structure and timeperiodic fluctuations around the stagnation point in the junction, by tuning the aspect ratio of the device. The design of the junction has been optimised to achieve more homogeneous strain rates^{36,37}, while Zhang et al.^{38} identified different flow regimes in the cross slot by changing the flow conditions. Recently, a crossslot device has aided the early detection of sepsis by using the deformability of leucocytes as a biomarker^{39}.
Despite earlier works focused on fluid dynamics and particle trapping^{40,41,42} in crossslot junctions, only a limited amount of studies have considered moderate inertia^{43,44} but focused mainly on cell deformation and not on the transient behaviour. The interaction of a particle and the vortex at moderate inertia has not been thoroughly investigated. It is currently unclear how flow field characteristics and particle properties correlate with observable metrics, such as the shape of the particle trajectory and the residence time in the junction. In this paper, we use immersedboundarylatticeBoltzmann (LB) simulations to explore the dynamics of a spherical, rigid particle in an incompressible Newtonian liquid in the steadyvortex regime (Section “Physical model and numerical methods”). We pay particular attention to the characterisation of the geometry, observables and simulation initialisation (Section “Crossslot geometry and simulation setup”). We investigate the effect of initial particle position and size on the trajectory and residence time of the particle in the crossslot junction at a Reynolds number of 120 and a channel aspect ratio of 0.6 (Section “Results and discussion”). Our results show that particle trajectory and residence time strongly depend on the initial lateral position. Under the investigated flow conditions and channel geometry, the vortex affects larger particles less compared to smaller ones. In Section “Conclusions”, we provide a concluding discussion. We hope that our results stimulate further work, for example, the investigation of deformable particles in the crossslot junction.
Crossslot geometry and simulation setup
The geometry consists of a crossslot channel, as shown in Fig. 1a. The origin of the Cartesian coordinate system is at the centre of the crossslot geometry. The flow in the inlets is along the yaxis (inlet axis), the flow in the outlets is along the xaxis (outlet axis), and the zaxis denotes the height direction (height axis). Depending on the combination of Reynolds number, heighttowidth aspect ratio and the ratio of the outlet flow rates, different flow regimes inside the crossslot junction can be observed, for example, a vortex which might be steady or unsteady^{35,38,45}. In this work, we are interested in the steadyvortex regime since the steady vortex leads to reproducible and nontrivial particle behaviour, which is relevant for flow cytometry applications. Therefore, we set the Reynolds number to Re = 120, the channel aspect ratio to α = 0.6, and the ratio of both outlet flow rates to 1, which leads to the formation of a steady vortex^{34}.
Since both inlet flow rates and both outlet flow rates are identical, the resulting flow field, including the vortex, is symmetric with respect to the inletheight plane (yzplane) and under a 180°rotation about the outlet axis (xaxis) in the absence of the particle. Therefore, it is sufficient to enter the particle only on one side of the inletheight plane, and all initial particle positions are x_{0} > 0. Furthermore, we only inject particles at one inlet (y_{0} < 0). Due to the presence of the vortex spiralling about the outlet axis, we need to explore the entire height range of initial particle positions. This procedure allows us to scan all physically unique initial particle positions without running multiple simulations that are physically equivalent.
The height H is constant within the entire domain, and both inlets and both outlets have the same width W. We use 36 grid points for H and 60 grid points for W, giving 45 grid points for the hydraulic diameter D_{h}. The length of the inlets and outlets is L_{in} = L_{out} = 4D_{h}, which we found to be sufficient to give results independent of the inlet and outlet lengths. The relaxation time is chosen as τ = 0.55Δt in all simulations, resulting in a viscosity of ν = 1/60Δx^{2}/Δt. The peak and average values of the inlet velocity profile are \({u}_{\max }=0.068\,{{\Delta }}x/{{\Delta }}t\) and \(\bar{u}=0.033\,{{\Delta }}x/{{\Delta }}t\), respectively. We consider particle sizes resulting in confinement values χ in the range [0.35, 0.65]. The particle is neutrally buoyant in all cases, ρ_{p}/ρ_{f} = 1.
To reduce the required time for the vortex to form, we introduce a force field that accelerates the fluid in the junction’s upper and lower halves in opposite directions along the inlet axis for a finite amount of time. Once the force has been switched off, the vortex flow is allowed to converge to a steady state with a convergence criterion of 10^{−6} for the relative difference of the flow field between consecutive time steps. It is reported that the right or lefthanded orientation of the spiral vortex occurs with the same probability^{34}. For consistency, we initialise the simulations in such a way that the vortex is always lefthanded, as shown in Fig. 1d. Eventually, the particle is injected at the desired initial position next to the inlet at y_{0} < 0.
We always initialise particles with a finite value of x_{0} since a particle initially on the midplane between the outlets would be caught in the vortex indefinitely in the absence of distortions. Since numerical noise in the simulation would eventually let the particle escape in one of the outlets, the residence time would be determined by numerical noise and not deterministic physics. In realworld experiments, channel and particle imperfections always lead to finite residence times in the geometry considered here.
Due to inertial forces, particles initialised near the inlet would migrate away from their prescribed lateral coordinates (x_{0}, z_{0}) while moving along the inlet, therefore losing control over the location where the particle enters the actual junction. To avoid this premature particle migration in the inlet, we lock the x and zcomponents of the particle position while allowing the particle to rotate and move along the inlet axis freely. The ‘guidance’ of the particle is switched off when the particle reaches a point located D_{h} upstream of the junction; the particle is allowed to move freely afterwards. Our simulations show that, at an upstream distance of about one hydraulic diameter from the junction, the velocity profile in the inlet starts to be affected by the junction. Therefore, the initial crosssectional position (x_{0}, z_{0}) defines at which point the particle enters the region of influence of the junction.
In the interest of using a concise notation, we define
Therefore, the accessible range of initial positions for a point particle in the inlet channel is \({\hat{x}}_{0}\in [1,+1]\) and \({\hat{z}}_{0}\in [1,+1]\).
Results and discussion
We present and discuss our simulation results in two main sections. Section “Effect of initial particle position” covers the effect of initial particle position (\({\hat{x}}_{0}\), \({\hat{z}}_{0}\)) on the trajectory and residence time of the particle. Section “Effect of particle confinement” focuses on the effect of particle size, including a neutrally buoyant point particle.
Effect of initial particle position
To study the effect of the initial particle position, we investigated a range of initial positions (\({\hat{x}}_{0},{\hat{z}}_{0}\)), as summarised in Table 1. A first simulation study revealed that the richest particle dynamics could be found for small values of \({\hat{x}}_{0}\). Hence, we ran more subsequent simulations for smaller values of \({\hat{x}}_{0}\). For the values of \({\hat{z}}_{0}\), we chose a starting position on the midplane (\({\hat{z}}_{0}=0\)) and the equilibrium positions of differently sized particles (χ between 0.35 and 0.75) in a straight duct with the same aspect ratio (α = 0.60) as the inlet of the crossslot geometry at Re = 120, see App. Equilibrium positions in a straight channel are equivalent to the inlet of the crossslot. The same confinement values are used in Section “Effect of particle confinement” to investigate the effect of particle size. The particle is neutrally buoyant and has a fixed confinement of χ = 0.4. The channel aspect ratio is α = 0.6, and the Reynolds number is 120.
Particle trajectories
Figure 2a, b shows some representative particle trajectories inside the crossslot junction for two different values of \({\hat{x}}_{0}\) and a range of values of \({\hat{z}}_{0}\). The trajectories of rigid particles found here are qualitatively similar to those presented by Gosset et al.^{31} in their supporting videos and Zhang et al.^{46}.
It is inconvenient to show the particle trajectories for all the simulations. Instead, we extract key characteristics from the particle trajectories that are also experimentally relevant. Figure 2c illustrates the chosen ‘macroscopic’ metrics of the inletoutlet projection (xyprojection, ‘top view’) of the particle trajectories: the number N of peaks observed in the central area of the crossslot junction, the amplitude A of each peak, and the lateral distance L between consecutive peaks. A peak is defined as a minimum or maximum of the inlet–outlet projection of the particle trajectory. The residence time \({T}_{{{{\rm{res}}}}}\) is the length of time the centre of mass of the particle spends in the central junction area, \(\hat{x}\in [1,+1]\) and \(\hat{y}\in [1,+1]\).
Particles initialised very close to the midplane of the inlet, \({\hat{x}}_{0}=0.0033\) in Fig. 2a, give rise to damped oscillations in the trajectories. For moderate distances from the midplane, \({\hat{x}}_{0}=0.15\) in Fig. 2b, the oscillations are either weak or completely absent. Figure 2 also shows that \({\hat{z}}_{0}\) has a stronger effect on the shape of the trajectories when \({\hat{x}}_{0}\) is small. We found that particles only show appreciable oscillations when the initial position \({\hat{x}}_{0}\) is smaller than 0.1. We will discuss these findings in the remainder of this section by providing further ‘microscopic’ information: instantaneous particle velocities (translational and angular) and forces (lift and drag) which we use to rationalise the behaviour of the macroscopic observables.
In the following, we will focus on an analysis of the macroscopic metrics defined in Fig. 2c (number of peaks, peak amplitude, interpeak distance) and the residence time to characterise the particle trajectories and their dependence on initial positions.
Number of peaks of particle trajectories
Figure 3a shows the number of peaks, N, for all investigated initial positions (\({\hat{x}}_{0},{\hat{z}}_{0}\)). Only for the narrow range \({\hat{x}}_{0} < 0.05\), some trajectories have two or more peaks. Increasing \({\hat{x}}_{0}\) from 0.0033 to 0.01 reduces N by one for all the investigated values of \({\hat{z}}_{0}\). Upon a further increase of \({\hat{x}}_{0}\), the number of peaks reduces to one or zero, depending on \({\hat{z}}_{0}\). No peaks are observed for any trajectory with \({\hat{x}}_{0}\ge 0.25\).
The initial height of the particle, \({\hat{z}}_{0}\), has a less pronounced effect on N. Trajectories with an initial position within the range \(0.22\le {\hat{z}}_{0}\le +0.22\) have the same number of peaks for identical values of \( {\hat{z}}_{0}\) although the trajectories look different (cf. Fig. 2). For \( {\hat{z}}_{0} > 0.22\), trajectories with \({\hat{z}}_{0} > 0\) tend to have fewer peaks than those with \({\hat{z}}_{0} < 0\) for the same magnitude of \({\hat{z}}_{0}\). For the entire range of \({\hat{x}}_{0}\), particles initially located at \({\hat{z}}_{0}=0.5\) follow trajectories without any peaks.
The peak amplitude of particle trajectories
The normalised peak amplitude, \(\hat{A}=A/(W/2)\), gives finer information about the shape of the trajectories. The data in Fig. 3b show the general trend that, for trajectories with more than a single peak, the amplitude of consecutive peaks is smaller than the amplitude of the previous peak, indicating that the spiralling motion of the particle is damped towards the outlet. The peak amplitudes are larger when particles are released at \({\hat{z}}_{0}\ge 0\), compared to trajectories with \({\hat{z}}_{0} < 0\) highlighting the influence of the vortex handedness on the particle’s motion in the junction under the investigated conditions.
Figure 3c shows the behaviour of the normalised amplitude of the first peak, \({\hat{A}}_{1}\), for all studied initial positions. For a given value of \({\hat{z}}_{0}\), there is little change of \({\hat{A}}_{1}\) with \({\hat{x}}_{0}\), evidenced by the collapse of data for different values of \({\hat{x}}_{0}\), as long as the first peak exists. Changing \({\hat{x}}_{0}\) merely affects the number of peaks, but not the amplitude of the first peak. The amplitude \({\hat{A}}_{1}\) increases with \({\hat{z}}_{0}\) for negative and positive values of \({\hat{z}}_{0}\), respectively, while the amplitude for \({\hat{z}}_{0}=0\) escapes the trend and defines a local maximum.
Axial interpeak distance
Like the amplitude of the peaks, \(\hat{A}\), the axial distance between consecutive peaks, \(\hat{L}=L/(W/2)\), serves as a quantitative measure of the particle trajectories. We define \({\hat{L}}_{i}\) as the normalised axial distance (along the outlet axis) between the peaks i − 1 and i, as long as both peaks exist. The axial distance \({\hat{L}}_{1}\) denotes the distance of the first peak from the inlet centreline as illustrated in Fig. 2c.
Figure 3d, e shows results for \({\hat{x}}_{0}=0.0033\) and 0.01, respectively. In both panels, only those curves are shown for which at least two peaks exist such that \({\hat{L}}_{2}\) is defined. Since trajectories with \({\hat{z}}_{0} > 0\) tend to have fewer peaks than those with \({\hat{z}}_{0} < 0\), there are fewer data points for the former. In all observed trajectories, the interpeak distance grows between consecutive pairs of peaks, i.e. \({\hat{L}}_{i+1} > {\hat{L}}_{i}\). Furthermore, \({\hat{L}}_{i}\) for a fixed i increases with \( {\hat{z}}_{0}\): the farther away from the horizontal midplane a particle is located initially, the longer its spiralling trajectory is stretched toward the outlet. The data also show that \(\hat{L}\) is larger for positive z_{0} than for negative z_{0} with the same magnitude \( {\hat{z}}_{0}\). Comparing data from Fig. 3d, e, we see that \(\hat{L}\) increases with \({\hat{x}}_{0}\): particles initially farther away from the vertical midplane have trajectories with larger interpeak distances. Specifically, Fig. 3(f) shows the increase of \({\hat{L}}_{2}\) with \({\hat{x}}_{0}\).
Residence time
The residence time of a particle in the junction (cf. Section “Particle trajectories”) is an important parameter to consider. Due to the symmetry of the problem, a particle with \({\hat{x}}_{0}=0\) would in principle remain in the junction forever, \({T}_{{{{\rm{res}}}}}\to \infty\). For any finite value of \({\hat{x}}_{0}\), the residence time should be finite since the particle is located closer to one outlet and can leave the junction area more easily. In fact, Fig. 4a shows that \({T}_{{{{\rm{res}}}}}\) sharply decreases with \({\hat{x}}_{0}\) for small \({\hat{x}}_{0}\) and drops further when \({\hat{x}}_{0}\) is increased.
Although \({T}_{{{{\rm{res}}}}}\) is largely determined by \({\hat{x}}_{0}\), there is also a clear dependency on \({\hat{z}}_{0}\) (Fig. 4b). For all studied values of \({\hat{x}}_{0}\), there is a mild increase of \({T}_{{{{\rm{res}}}}}\) with \({\hat{z}}_{0}\): the residence time is shorter when the particle is initialised in the lower channel half (\({\hat{z}}_{0} < 0\)) compared to initial positions in the upper channel half.
Discussion
In the previous sections, we identified four clear trends upon increasing the initial position \({\hat{x}}_{0}\): (1) the number of peaks, N, decreases, (2) the peak amplitude, \({\hat{A}}_{1}\), is largely independent of \({\hat{x}}_{0}\), (3) the interpeak distance, \(\hat{L}\), increases and (4) the residence time, \({T}_{{{{\rm{res}}}}}\), decreases (cf. Figs. 3a, c–e, and 4a, respectively). When changing the initial position \({\hat{z}}_{0}\) instead, we observed that (1) the number of peaks is smaller for \({\hat{z}}_{0} > 0\) than for \({\hat{z}}_{0} < 0\) for the same magnitude of \( {\hat{z}}_{0}\), (2) the peak amplitude increases with \({\hat{z}}_{0}\), (3) the interpeak distance is smaller for \({\hat{z}}_{0} > 0\) than for \({\hat{z}}_{0} < 0\) for the same magnitude of \( {\hat{z}}_{0}\) and (4) the residence time mildly increases with \({\hat{z}}_{0}\) (cf. Figs. 3a, c, d and 4b, respectively).
In order to understand the behaviour of the residence time, we first analyse the velocity and acceleration of the particle along the outlet axis (xaxis) inside the junction (Fig. 4c, d). Particles that are initially close to the vertical midplane (small \({\hat{x}}_{0}\)) accelerate more slowly than a particle that enters the junction farther away (larger \({\hat{x}}_{0}\)). The slow acceleration of a particle with small \({\hat{x}}_{0}\) is caused by the particle covering nearly identical portions of volume on either side of the midplane; hence the net force accelerating the particle towards one of the outlets is nearly zero. Once the particle is slowly moving towards one outlet, it gets increasingly exposed to the drag force exerted by the flow in that outlet channel, and the particle acceleration, and therefore velocity, increases. A particle that enters the junction far away from the midplane, on the one hand, is closer to the outlet and, on the other hand, experiences a larger drag force early on and leaves the junction sooner; hence, the residence time is shorter. The residence time also depends on \({\hat{z}}_{0}\) (Fig. 4b), but to a lesser extent than it depends on \({\hat{x}}_{0}\). This dependency stems from the particle acceleration \({\dot{u}}_{x}\) being smaller for particles entering the junction above the horizontal midplane (\({\hat{z}}_{0} > 0\)). The mechanism behind this effect will become obvious at the end of this section. Interestingly, the acceleration increases exponentially with time, as revealed in Fig. 4d.
The behaviour of the number of peaks, N, and the interpeak distance, \(\hat{L}\), is determined by an interplay of the residence time and the ability of the particle to follow the vortex and revolve about the outlet axis (xaxis). The slower the particle moves towards the outlet and the faster it revolves about the outlet axis, the larger the number of peaks and the smaller the interpeak distance. Figure 5a depicts the distribution of the vorticity magnitude of the unperturbed flow field in the junction on the horizontal midplane (\(\hat{z}=0\)). The vortex can be recognised by a region of high vorticity stretched along the outlet axis. The vortex causes the particle to revolve about the outlet axis with an angular velocity Ω_{x}, thus creating the spiralling trajectories illustrated in Fig. 2a. Figure 5b shows the angular velocity, averaged over the time the particle is located inside the junction area, \({\bar{{{\Omega }}}}_{x}\). The number of peaks N, residence time \({T}_{{{{\rm{res}}}}}\) and the average angular velocity \({\bar{{{\Omega }}}}_{x}\) are related according to
remembering that a full revolution of 2π radians leads to two peaks as defined in Fig. 2c. Particles initially located closer to the vertical midplane (small \({\hat{x}}_{0}\)) experience a higher angular velocity than particles entering the junction farther away from the midplane (large \({\hat{x}}_{0}\)). Combining the faster revolution \({\bar{{{\Omega }}}}_{x}\) with the smaller linear acceleration \({\dot{u}}_{x}\) (and therefore larger residence time) for particles with \({\hat{x}}_{0}\to 0\) explains why these particles have the largest number of peaks, N, according to Eq. (2). Since peaks are only counted in the junction area, which has a fixed size, a trajectory with more peaks has smaller interpeak distances \(\hat{L}\), which explains the trend of \(\hat{L}\) becoming smaller for \({\hat{x}}_{0}\to 0\). Looking at particles with small \({\hat{x}}_{0}\), Fig. 5b reveals that \({\bar{{{\Omega }}}}_{x}\) is significantly smaller for particles with \({\hat{z}}_{0} > 0\), while the residence time in Fig. 4(b) shows only a moderate increase for these particles, compared to particles with \({\hat{z}}_{0} < 0\). Consequently, particles with \({\hat{z}}_{0} > 0\) show fewer peaks than particles with \({\hat{z}}_{0} < 0\) as long as \({\hat{x}}_{0}\) is small. The smaller revolution frequency of particles with \({\hat{z}}_{0} > 0\) can be understood as follows: Fig. 5c shows that, upon entering the junction, these particles move over the vortex, rather than getting trapped in the vortex, then reach a region with relatively small vorticity before finally entering the vortex. Averaged over the time they spend in the junction, these particles, therefore, do not revolve as much as the particles that are able to enter the vortex immediately.
In order to understand the variation of the peak amplitude with the initial positions \({\hat{x}}_{0}\) and \({\hat{z}}_{0}\), we first inspect the particle trajectories as seen from the outlet (projection on the inletheight plane, Fig. 5c). A particle that is released at \({\hat{z}}_{0} < 0\) experiences an immediate upward lift force and is drawn into the vortex where the particle revolves with small amplitude. Contrarily, a particle with larger \({\hat{z}}_{0}\) features a flatter trajectory tends to avoid the vortex and approaches lowvelocity zone B. Only after the encounter with zone B the particle is pulled into the vortex, resulting in a larger peak amplitude \({\hat{A}}_{1}\). None of the particles simulated here is able to reach zone A; this would only be possible for particles with very small confinement χ released at \({\hat{z}}_{0}\to 1\). The different shapes of particle trajectories can be understood by inspecting the streamlines in Fig. 1d: streamlines with slightly negative \({\hat{z}}_{0}\) turn upwards and reach the vortex immediately, while streamlines with slightly positive \({\hat{z}}_{0}\) tend to avoid the vortex and approach zone B; only streamlines near the bottom of the channel are drawn towards zone A. The independence of \({\hat{A}}_{1}\) from \({\hat{x}}_{0}\) can be explained by the shape of the highvorticity region in Fig. 5a. Since the vorticity profile is largely constant along the outlet axis inside the entire junction area, the initial interaction between the particle and the vortex is not strongly sensitive to \({\hat{x}}_{0}\).
Finally, Fig. 5c explains the trend of the residence time to increase with \({\hat{z}}_{0}\) (Fig. 4b). Particles with larger \({\hat{z}}_{0}\) spend some time in the lowvelocity zone B, which delays the particle’s acceleration towards the outlet, therefore increasing the residence time in the junction.
Effect of particle confinement
In this section, we analyse the effect of the size of the particle on its dynamics in the crossslot junction. We investigate five different confinement values, χ ∈ [0.35, 0.65] and the limiting case of a massless point particle, χ → 0. In order to keep the number of free parameters manageable, we focus on \({\hat{x}}_{0}=0.0033\), which leads to the largest number of peaks and \({\hat{x}}_{0}=0.15\), which is the lowest value of \({\hat{x}}_{0}\) for which nearly no peaks have been observed in Fig. 3(a). Additionally, we fix the initial positions at \({\hat{z}}_{0}=\pm 0.31\) which is the lateral equilibrium position of a particle with χ = 0.65 in a straight channel with the same dimensions and Reynolds number as the inlet channel of the crossslot geometry. The hydrodynamic conditions and channel aspect ratio are kept constant at Re = 120 and α = 0.6, respectively.
Initial position at \({\hat{x}}_{0}=0.0033\)
In Section “Effect of initial particle position”, we found that the vortex has a more profound influence on the particle dynamics when \({\hat{x}}_{0}=0.0033\) than for the other investigated initial positions. Figure 6 shows the particle trajectories in the junction for the investigated confinement values for \({\hat{z}}_{0}=0.31\) in (a, e) and \({\hat{z}}_{0}=0.31\) in (b, f). As the confinement χ increases, the characteristic spiral trajectories seen for particles with low confinement are less pronounced or no longer observed at all. For \({\hat{z}}_{0}=0.31\), the particle with χ = 0.65 leaves the junction without any oscillation. For \({\hat{z}}_{0}=0.31\), no particle with χ ≥0.5 oscillates in the junction. A massless point particle, which follows the fluid streamlines, shows the most oscillatory behaviour of all particles for \({\hat{z}}_{0}=0.31\); contrarily, when initialised at \({\hat{z}}_{0}=0.31\), the point particle avoids the vortex and hardly oscillates.
The revolution frequency of the particle about the outlet axis, Ω_{x}, provides a complementary view. When particles enter the junction below the horizontal midplane, \({\hat{z}}_{0}=0.31\) (Fig. 7a), Ω_{x} is highest for particles with χ < 0.5, including the point particle, and Ω_{x} decays quickly for larger confinement. For \({\hat{z}}_{0}=0.31\) (Fig. 7b), particles with 0.35 ≤ χ < 0.5 show a behaviour similar to that for \({\hat{z}}_{0}=0.31\). In contrast, both the point particle and the particles with χ ≥ 0.5 hardly oscillate when released above the horizontal midplane, \({\hat{z}}_{0}=0.31\). While the oscillation of the point particle is suppressed by the particle reaching the lowvelocity zone B, larger particles are strongly confined when they approach the top wall of the junction and, hence, are unable to revolve freely.
Figure 7c displays the residence time, \({T}_{{{{\rm{res}}}}}\), for all investigated cases. For particles that are released below the horizontal midplane, \({\hat{z}}_{0}=0.31\), confinement has a negligible effect on the residence time: for all particles, including the point particle, \({T}_{{{{\rm{res}}}}}\approx 2{t}_{{{{\rm{ad}}}}}\). The insensitivity of the residence time to the confinement can be explained by the observation that all particles remain in the vortical zone throughout their passage (cf. Fig. 6e), where particles are efficiently advected towards the outlet. The situation is different for \({\hat{z}}_{0}=0.31\) where the residence time strongly depends on particle size. While particles with 0.35 ≤ χ ≤ 0.5 leave the junction nearly as quickly as their respective counterparts at \({\hat{z}}_{0}=0.31\), both the point particle and the larger particles (χ > 0.5) require more time to leave the junction area. The increased residence time of the point particle is caused by the point particle reaching and remaining in the lowvelocity zone B, as seen in Fig. 6f. The large particles with χ > 0.5 do not reach zone B; instead, they are nearly in contact with the top wall, which reduces their ability to follow the flow around the vortex or towards the outlet.
Initial position at \({\hat{x}}_{0}=0.15\)
In Section “Effect of initial particle position”, we demonstrated that the vortex has little effect on the motion of the particle when the particle enters the junction farther away from the vertical midplane, for example at \({\hat{x}}_{0}=0.15\). Here, we show that, even if particle confinement is varied, the vortex continues to have little influence on the motion of the particle at \({\hat{x}}_{0}=0.15\). Figure 6 depicts the particle trajectories in the junction for the investigated confinement values for \({\hat{z}}_{0}=0.31\) in (c, g) and \({\hat{z}}_{0}=0.31\) in (d, h). For both \({\hat{z}}_{0}=0.31\) and 0.31, the particle experiences a strong acceleration towards the outlet once it enters the junction area since most of the particle volume is located in the outletfacing half of the junction. During its passage, the particle does not reach the vortex core and, therefore, does not undergo a strong revolution about the outlet axis (Ω_{x} not shown for \({\hat{x}}_{0}=0.15\)). Consequently, in nearly all situations, the particle leaves the junction without displaying a peak in its trajectory. The only exception is the point particle for \({\hat{z}}_{0}=0.31\) with a single peak, although the point particle essentially follows a similar trajectory as the finitesize particles.
The residence time \({T}_{{{{\rm{res}}}}}\) is consistently lower for \({\hat{x}}_{0}=0.15\) compared to \({\hat{x}}_{0}=0.0033\) (Fig. 7c). The shorter residence time is caused by the particle with \({\hat{x}}_{0}=0.15\) being accelerated towards the outlet earlier and, at least to some extent, the shorter distance to travel to the outlet. The residence time is essentially independent of confinement χ, and it is nearly the same for \({\hat{z}}_{0}=0.31\) and \({\hat{z}}_{0}=0.31\): all particles follow roughly the same trajectory and therefore experience a similar acceleration along the outlet axis (Fig. 6).
Physical model and numerical methods
We briefly outline the physical model (Section “Physical model, parameters and nondimensional groups”) and the numerical methods (Section “Numerical methods”) used in this work.
Physical model, parameters and nondimensional groups
We consider a single rigid, spherical particle immersed in an incompressible, isothermal, viscous Newtonian liquid. The incompressible NavierStokes equations and Newton’s laws of motion determine the dynamics of the liquid and the particle, respectively. The noslip condition is assumed at all solidliquid boundaries.
The relevant parameters characterising the fluid flow are the liquid density ρ_{f}, the kinematic viscosity ν, and the mean velocity \(\bar{u}\) in the inlet of the crossslot geometry. The particle is fully characterised by its radius a and density ρ_{p}. As will be described in more detail in Section “Crossslot geometry and simulation setup”, the geometry is fully determined by the width W and height H of the inlet and outlet channels of the crossslot geometry.
Following^{34}, we define the channel Reynolds number as
The nondimensional confinement of the particle in the crossslot geometry is
The aspect ratio of the crosssection of the inlet and outlet channels is given by
Note that we chose α < 1 in our study, justifying the use of H in the definition of the confinement, Eq. (4). The hydraulic diameter
is important for the characterisation of the inflow behaviour of the particle. We further define the advection time scale
to nondimensionalise time as \(\hat{t}=t/{t}_{{{{\rm{ad}}}}}\).
Numerical methods
We employ the latticeBoltzmann (LB) method with the D3Q19 lattice^{47}, the BGK collision operator^{48} and the Guo forcing scheme^{49} to solve the NavierStokes equations. The relaxation time τ of the BGK collision operator is linked to the kinematic viscosity according to
where \({c}_{{{{\rm{s}}}}}=\sqrt{1/3}{{\Delta }}x/{{\Delta }}t\) is the lattice speed of sound, Δx is the lattice spacing and Δt is the time step.
The multidirectforcing immersedboundary method with a threepoint stencil is used to satisfy the noslip boundary condition on the surface of the moving rigid particle^{50,51}. The halfway bounceback scheme^{52} is employed to satisfy the noslip condition on the stationary walls of the crossslot geometry. Additionally, due to nonnegligible particle inertia, a coefficient of restitution is included to model perfectly elastic collisions to avoid any overlap between the particle and the wall.
We impose a fully developed rectangular Poiseuille flow with the desired mean velocity \(\bar{u}\) at both inlets^{53}. The boundary condition at both outlets is based on an extrapolation approach^{54} modified in such a way that the average pressure on the outlet plane is constant in time. Therefore, the inlet velocity and the outlet pressure are imposed, while the outlets permit the swirling flow caused by the vortex at the centre of the crossslot geometry to leave the domain without undesired reflections at the outlet planes.
To ensure that our simulation code accurately captures the fluid dynamics and the fluidparticle interaction, we compared the flow structure in the junction with available data^{34} (App. Flow structure in the crossslot junction) and the lateral equilibrium positions in a square crosssection with published results^{55} (App. Test case: effect of initial position and confinement on lateral migration in the square duct).
Conclusions
We used the immersedboundarylatticeBoltzmann method to numerically investigate the dynamics of a rigid neutrallybuoyant spherical particle in a crossslot junction for a channel heighttowidth ratio of 0.6 and a Reynolds number of 120 for which a steady vortex exists in the junction. We characterised the effects of the initial position and the particle confinement χ = 2a/H (where a is the particle radius and H is the height of the channel) on the particle trajectories in the junction. Particles generally show a swirling motion within the crossslot. Thus, observed trajectories were described in terms of macroscopic observables, such as the number of peaks of each trajectory (N), the peak amplitudes (A), the axial interpeak distances (L), and the residence time within the junction area (\({T}_{{{{\rm{res}}}}}\)). The simulations also provide access to microscopic observables, such as instantaneous linear and angular velocity and acceleration, which were used to rationalise the particle’s behaviour.
There are several key observations when varying the location at which the particle with fixed size enters the junction:

The residence time \({T}_{{{{\rm{res}}}}}\) strongly depends on the initial position of the particle. A particle entering the junction closer to the vertical midplane is accelerated more slowly towards its closest outlet, thus leading to a larger residence time. There is a mild effect of the initial height of the particle on the residence time. The behaviour of the residence time can be explained by the interaction of the particle with lowvelocity areas in the flow field.

Two competing mechanisms determine the number of peaks N and the axial interpeak distance L seen in the particle trajectories: particles that are able to follow the vortex and revolve about the outlet axis more easily show a larger number of peaks and shorter interpeak distances. Contrarily, particles with a shorter residence time have a trajectory with fewer peaks and larger interpeak distances. As a net effect, the number of peaks and interpeak distance depend strongly on the lateral position and weakly on the height position of the particle when entering the junction.

The amplitude A of consecutive peaks of a given trajectory generally decrease as the particle follows the dissipating vortex along the outlet channels. The initial height of the particles has a strong effect on the amplitude: depending on the initial position, the particle either moves along the vortex, resulting in a larger amplitude, or it moves against the vortex, resulting in fast migration towards the centre of the junction and a smaller amplitude.
The properties of the particle trajectories also depend on the particle confinement:

When entering the junction close to the vertical midplane, larger particles have a slightly higher residence time \({T}_{{{{\rm{res}}}}}\) since a large fraction of the particle is initially located in the region drawn into the second outlet.

The trajectories of particles with small confinement show more peaks N with smaller interpeak distance L because smaller particles are able to revolve faster about the outlet axis.

Particles with larger confinement show a higher amplitude A of the first peak since these particles tend to pass the vortex first before being drawn into the vortex.
The crossslot geometry has been successfully employed in flow cytometry to deform cells^{30,31,43}, where cell deformation is used as a biomarker for the cell state^{39}. In deformability cytometry using a junction, cells or particles are usually inertially focused before entering the junction^{31}. Therefore, cells or particles are expected to enter and traverse the junction with typical trajectories, which can be quantified using the metrics used in the present study. The videos of Gossett et al.^{31} display polystyrene beads, droplets and cells performing an oscillating trajectory in the crossslot junction as seen from above (along the zaxis), similar to the ones we observed in Section “Effect of initial particle position”. Zhang et al.^{46} reported oscillations in the trajectory of glass beads in a crossslot channel with a square crosssection at Re = 80. The trajectories of the rigid particle in our study show a good qualitative agreement with those trajectories seen under steadyvortex conditions published previously^{31,46}.
Our study provides a deeper understanding of the dynamics of a rigid particle interacting with the vortex in a crossslot junction at moderate inertia. The macroscopic observables (N, A, L, \({T}_{{{{\rm{res}}}}}\)) used in our work might help distinguish between particles with different properties in experimental crossslot flows. To this end, a future investigation of the effect of channel geometry, in particular aspect ratio and Reynolds number, appears beneficial. Since most inertial microfluidic applications involve deformable cells, a study of deformable particles in crossslot flows would provide further insight into the particle dynamics of realworld applications.
Data availability
The data that support the findings of this study are openly available in the Edinburgh DataShare repository at https://hdl.handle.net/10283/4780.
References
Darling, E. M. & Di Carlo, D. Highthroughput assessment of cellular mechanical properties. Annu. Rev. Biomed. Eng. 17, 35–62 (2015).
DiezSilva, M., Dao, M., Han, J., Lim, C.T. & Suresh, S. Shape and biomechanical characteristics of human red blood cells in health and disease. MRS Bull. 35, 382–388 (2010).
Bow, H. et al. A microfabricated deformabilitybased flow cytometer with application to malaria. Lab Chip 11, 1065–1073 (2011).
Barabino, G. A., Platt, M. O. & Kaul, D. K. Sickle cell biomechanics. Annu. Rev. Biomed. Eng. 12, 345–367 (2010).
Cross, S. E., Jin, Y.S., Rao, J. & Gimzewski, J. K. Nanomechanical analysis of cells from cancer patients. Nat. Nanotechnol. 2, 780–783 (2007).
Pierrakos, C., Velissaris, D., Bisdorff, M., Marshall, J. C. & Vincent, J.L. Biomarkers of sepsis: time for a reappraisal. Crit. Care 24, 287 (2020).
Remmerbach, T. W. et al. Oral cancer diagnosis by mechanical phenotyping. Cancer Res. 69, 1728–1732 (2009).
Suresh, S. Biomechanics and biophysics of cancer cells. Acta Biomater. 3, 413–438 (2007).
Nishino, M. et al. Serial changes in leukocyte deformability and whole blood rheology in patients with sepsis or trauma. J. Trauma Acute Care Surg. 59, 1425–1431 (2005).
Krieg, M. et al. Atomic force microscopybased mechanobiology. Nat. Rev. Phys. 1, 41–57 (2019).
Guck, J. et al. The optical stretcher: A novel laser tool to micromanipulate cells. Biophys. J. 81, 767–784 (2001).
Hochmuth, R. M. Micropipette aspiration of living cells. J. Biomech. 33, 15–22 (2000).
Toner, M. & Irimia, D. Bloodonachip. Annu. Rev. Biomed. Eng. 7, 77–103 (2005).
Kalyan, S. et al. Inertial microfluidics enabling clinical research. Micromachines 12, 1–41 (2021).
Amini, H., Lee, W. & Di Carlo, D. Inertial microfluidic physics. Lab Chip 14, 2739–2761 (2014).
Segré, G. & Silberberg, A. Radial particle displacements in poiseuille flow of suspensions. Nature 189, 209–210 (1961).
Segré, G. & Silberberg, A. Behaviour of macroscopic rigid spheres in poiseuille flow part 2. Experimental results and interpretation. J. Fluid Mech. 14, 136–157 (1962).
Ho, B. P. & Leal, L. G. Inertial migration of rigid spheres in twodimensional unidirectional flows. J. Fluid Mech. 65, 365–400 (1974).
Vasseur, P. & Cox, R. G. The lateral migration of a spherical particle in twodimensional shear flows. J. Fluid Mech. 78, 385–413 (1976).
Asmolov, E. S. The inertial lift on a spherical particle in a plane poiseuille flow at large channel Reynolds number. J. Fluid Mech. 381, 63–87 (1999).
Carlo, D. D., Irimia, D., Tompkins, R. G. & Toner, M. Continuous inertial focusing, ordering, and separation of particles in microchannels. Proc. Natl Acad. Sci. USA 104, 18892–18897 (2007).
Gossett, D. R. & Di Carlo, D. Particle focusing mechanisms in curving confined flows. Anal. Chem. 81, 8459–8465 (2009).
Oakey, J. et al. Particle focusing in staged inertial microfluidic devices for flow cytometry. Anal. Chem. 82, 3862–3867 (2010).
Jiang, D., Tang, W., Xiang, N. & Ni, Z. Numerical simulation of particle focusing in a symmetrical serpentine microchannel. RSC Adv. 6, 57647–57657 (2016).
Ren, H. et al. Multiplexed serpentine microchannels for highthroughput sorting of disseminated tumor cells from malignant pleural effusion. Sens. Actuators B 337, 129758 (2021).
Kuntaegowdanahalli, S. S., Bhagat, A. A. S., Kumar, G. & Papautsky, I. Inertial microfluidics for continuous particle separation in spiral microchannels. Lab Chip 9, 2973–2980 (2009).
Martel, J. M. & Toner, M. Inertial focusing dynamics in spiral microchannels. Phys. Fluids 24, 32001 (2012).
Warkiani, M. E. et al. Slanted spiral microfluidics for the ultrafast, labelfree isolation of circulating tumor cells. Lab Chip 14, 128–137 (2014).
Otto, O. et al. Realtime deformability cytometry: onthefly cell mechanical phenotyping. Nat. Methods 12, 199–202 (2015).
Nyberg, K. D. et al. Quantitative deformability cytometry: Rapid, calibrated measurements of cell mechanical properties. Biophysical J. 113, 1574–1584 (2017).
Gossett, D. R. et al. Hydrodynamic stretching of single cells for large population mechanical phenotyping. Proc. Natl Acad. Sci. USA 109, 7630–7635 (2012).
Ahmmed, S. M. et al. Multisample deformability cytometry of cancer cells. APL Bioeng. 2, 032002 (2018).
Ait Mouheb, N., Malsch, D., Montillet, A., Solliec, C. & Henkel, T. Numerical and experimental investigations of mixing in tshaped and crossshaped micromixers. Chem. Eng. Sci. 68, 278–289 (2012).
Haward, S. J. et al. Tricritical spiral vortex instability in crossslot flow. Phys. Rev. E 93, 031101 (2016).
Burshtein, N., Zografos, K., Shen, A. Q., Poole, R. J. & Haward, S. J. Periodic fluctuations of streamwise vortices in inertiadominated intersecting flows. Phys. Fluids 33, 014106 (2021).
Haward, S. J., Jaishankar, A., Oliveira, M. S. N., Alves, M. A. & McKinley, G. H. Extensional flow of hyaluronic acid solutions in an optimized microfluidic crossslot device. Biomicrofluidics 7, 044108 (2013).
GalindoRosales, F. J., Oliveira, M. S. N. & Alves, M. A. Optimized crossslot microdevices for homogeneous extension. RSC Adv. 4, 7799–7804 (2014).
Zhang, J.W., Li, W.F., Xu, X.L., Liu, H.F. & Wang, F.C. Experimental investigation of threedimensional flow regimes in a crossshaped reactor. Phys. Fluids 31, 034105 (2019).
Guillou, L. et al. Development and validation of a cellular host response test as an early diagnostic for sepsis. PLoS ONE 16, 1–17 (2021).
Tanyeri, M., JohnsonChavarria, E. M. & Schroeder, C. M. Hydrodynamic trap for single particles and cells. Appl. Phys. Lett. 96, 224101 (2010).
Tanyeri, M., Ranka, M., Sittipolkul, N. & Schroeder, C. M. A microfluidicbased hydrodynamic trap: design and implementation. Lab Chip 11, 1786–1794 (2011).
Tanyeri, M. & Schroeder, C. M. Manipulation and confinement of single particles using fluid flow. Nano Lett. 13, 2357–2364 (2013).
Armistead, F. J., Gala De Pablo, J., Gadêlha, H., Peyman, S. A. & Evans, S. D. Cells Under Stress: An InertialShear Microfluidic Determination of Cell Behavior. Biophys. J. 116, 1127–1135 (2019).
Hymel, S. J., Lan, H. & Khismatullin, D. B. Elongation index as a sensitive measure of cell deformation in highthroughput microfluidic systems. Biophys. J. 119, 493–501 (2020).
Zhang, W., Li, W.f, Liu, H.f & Wang, F.c Effects of outlets flow rates ratios on flow patterns in crossshaped channels. Phys. Fluids 33, 063601 (2021).
Zhang, W., ding Wang, Y., feng Li, W., feng Liu, H. & chen Wang, F. Capturing of particles in suspension flow through a micro crossshaped channel. Int. J. Multiph. Flow. 150, 104030 (2022).
Qian, Y. H., D’Humières, D. & Lallemand, P. Lattice BGK models for NavierStokes equation. Europhys. Lett. 17, 479–484 (1992).
Bhatnagar, P. L., Gross, E. P. & Krook, M. A model for collision processes in gases. i. small amplitude processes in charged and neutral onecomponent systems. Phys. Rev. 94, 511–525 (1954).
Guo, Z., Zheng, C. & Shi, B. Discrete lattice effects on the forcing term in the lattice Boltzmann method. Phys. Rev. E 65, 046308 (2002).
Wang, Z., Fan, J. & Luo, K. Combined multidirect forcing and immersed boundary method for simulating flows with moving particles. Int. J. Multiph. Flow. 34, 283–302 (2008).
Falagkaris, E., Ingram, D., Markakis, K. & Viola, I. Proteus: A coupled iterative forcecorrection immersedboundary cascaded lattice Boltzmann solver for moving and deformable boundary applications. Comput. Math. Appl. 75, 1330–1354 (2018).
Ladd, A. J. C. Numerical simulations of particulate suspensions via a discretized Boltzmann equation. Part 1. Theoretical foundation. J. Fluid Mech. 271, 285–309 (1994).
Bruus, H. Theoretical Microfluidics Oxford Master Series in Physics (Oxford University Press, 2008).
Geier, M., Schönherr, M., Pasquali, A. & Krafczyk, M. The cumulant lattice Boltzmann equation in three dimensions: theory and validation. Comput. Math. Appl. 70, 507–547 (2015).
Lashgari, I., Ardekani, M. N., Banerjee, I., Russom, A. & Brandt, L. Inertial migration of spherical and oblate particles in straight ducts. J. Fluid Mech. 819, 540–561 (2017).
Owen, B. & Krüger, T. Numerical investigation of the formation and stability of homogeneous pairs of soft particles in inertial microfluidics. J. Fluid Mech. 937, 1–31 (2022).
Prohm, C. & Stark, H. Feedback control of inertial microfluidics using axial control forces. Lab Chip 14, 2115–2123 (2014).
Acknowledgements
T.K. received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (803553). This work used the Cirrus UK National Tier2 HPC Service at EPCC (https://www.cirrus.ac.uk). The authors thank Emmanouil Falagkaris for their contributions to code development. For the purpose of open access, the authors have applied a Creative Commons Attribution (CC BY) licence to any Author Accepted Manuscript version arising from this submission.
Author information
Authors and Affiliations
Contributions
T.K. acquired funding, conceptualised and oversaw the study. B.O. and K.K. validated the code. K.K. generated the data, carried out the analysis and wrote the original paper. All authors discussed the results and edited the paper.
Corresponding author
Ethics declarations
Conflict of interest
H.T. is cofounder and CTO of Cytovale Inc., a company which is commercialising a deformability cytometry technology based on crossslot flows. D.D. is a cofounder and on the Board of Directors of Cytovale Inc. L.G. was employed by Cytovale Inc. during the preparation of the manuscript. T.K. has received financial support from Cytovale to partially fund a PhD student who is not a coauthor of this paper. The remaining authors have no conflict of interest to declare.
Appendix
Appendix
Flow structure in the crossslot junction
We determine the flow structure for an incompressible Newtonian liquid in a crossslot channel with a square crosssection at various values of the Reynolds number. The crossslot junction has a width and height of W = 60Δx with an inlet and outlet length equal to 4D_{h}. Haward et al.^{34} explained the mechanism of vortex formation in the crossslot junction under a range of Reynolds numbers and channel aspect ratios. Figure 8 compares our results with available experimental and numerical data from their work. Figure 8a–h corresponds to experimentally and numerically obtained data, respectively^{34}. Figure 8i–l is the generated streamlines from our simulations after the flow has converged. Our results show excellent agreement with published data.
Lateral equilibrium positions in straight channels
Test case: effect of initial position and confinement on lateral migration in square duct. We determine the lateral equilibrium position of a rigid neutrally buoyant spherical particle in a Poiseuille flow through a straight duct with a square crosssection at Re = 100. The duct has a width and height W = 50Δx. The particle radius is varied such that confinement values χ = 0.1, 0.2 and 0.2857 are obtained. Periodic boundary conditions are used at the inlet and outlet of the duct. With L = 6W, the domain length is sufficiently long to ensure that periodic images of the particle do not interact. The viscosity is set to ν = 1/30Δx^{2}/Δt.
We compare the equilibrium positions found using our IB–LB solver with those reported by Lashgari et al.^{55} who originally proposed this case. Figure 9 shows excellent agreement of the equilibrium positions. As expected, equilibrium positions are independent of the initial position. Larger particles assume equilibrium positions closer to the duct centre at the midpoint of a duct edge.
Equilibrium positions in a straight channel equivalent to the inlet of the crossslot. We perform simulations of a rigid neutrally buoyant spherical particle flowing through a straight channel (along the xaxis) with the same crosssectional shape as the inlet of the crossslot geometry to identify the lateral equilibrium positions for a given particle size. The simulation parameters are the same as in Section “Crossslot geometry and simulation setup”; in particular, the Reynolds number is Re = 120, and the channel aspect ratio is α = 0.6. Periodic boundary conditions are used at the inlet and outlet of the duct. With L = 4W, the domain length is sufficiently long to ensure that periodic images of the particle do not interact^{56}. The schematic in Fig. 10a illustrates the initial configuration of the particle in the channel.
Figure 10b shows the identified equilibrium positions z_{eq} that we used as the initial positions for our main study in Section “Results and discussion”. Our simulations confirm the known observation that, for the studied parameter values, the equilibrium position of the particle is located midway along the long crosssectional axis (y = 0)^{57}; the particle remains on this midplane during its lateral migration.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Kechagidis, K., Owen, B., Guillou, L. et al. Numerical investigation of the dynamics of a rigid spherical particle in a vortical crossslot flow at moderate inertia. Microsyst Nanoeng 9, 100 (2023). https://doi.org/10.1038/s4137802300541z
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1038/s4137802300541z