Continuously Sorting Particles According to Shape

There are numerous filters to separate particles in liquid based on their size, which can be enough to isolate them; however, particle shape can be more important, as it distinguishes healthy red blood cells from those affected by sickle-cell disease or malaria. Shape can also be used to determine what stage a cell is in of the cell cycle, which would benefit researchers looking for dividing cells. Recent research by Dino Di Carlo of UCLA looks to separate particles of differing aspect ratios continuously, using inertial fluid-dynamics. His work, “Continuous Inertial Focusing and Separation of Particles by Shape,” featured in Physical Review X reminds me of his previous work to use inertial fluid-dynamics to continuously filter particles according to size.

Particle Separation

Existing methods to separate particles according to shape include hydro-dynamic filtration (HDF) and deterministic lateral displacement (DLD), along with a few others. DLD involves a grid of posts in a channel that is arranged in a way that effectively separates particles according to size. This can be enhanced by controlling particle orientation in order to filter by particle shape or shear stress to separate by particle deformability. HDF uses highly branched channels to separate particles by size according to the fluidic resistance of the side channels. Both of these methods are passive and continuous, but they have highly complex structures and require low flow rates (no greater than a few µL/min). Di Carlo’s method uses fluid inertia to focus particles in channels at high flow rates from 40 to 80 µL/min. This utilizes a shear-gradient lift force and a wall-effect lift force in order to shift particles across streamlines when the Reynolds number of the particle is on the magnitude of 1 or greater. The shear-gradient lift force is directed down the shear gradient and toward the wall, while the wall-effect lift force is caused by the wake of a particle near the wall and is directed away from the wall. With a Reynolds numbers on the order of 1, inertial lift forces dominate the particle behavior while viscous interactions dominate when the particle Reynolds number is << 1. As the particle Reynolds number increases, migration across streamlines, away from the center line, is observed.

Di Carlo attempted to separate spheres and ellipsoids of varying aspect ratios, while conserving volume. Spheres of 3 µm and 6 µm in diameter, and ellipsoids that conserved the volume at 1:3 and 1:5 aspect ratios were used.


Spheres and ellipsoids are sorted using inertial fluid-dynamics. Particles with larger aspect radios reach equilibrium positions (Xeq) 4 cm downstream.

Spheres and ellipsoids are sorted using inertial fluid-dynamics. Particles with larger aspect radios reach equilibrium positions (Xeq) 4 cm downstream.

This work demonstrates that rod-like particles find equilibrium positions close to the center of the channel, while spheres of the same volume end up in streamlines close to the wall. When the major axis of the particles rotates perpendicular to the plane of the wall, the wall-effect lift increases and the particles is pushed away from the wall. Once the major axis has realigned with the direction of the flow, the wall-effect lift decreases and the particles move towards the wall one again. However, particles with higher aspect ratios experience a wall-effect lift greater than the shear-gradient lift and find equilibrium positions closer to the center of the channel. As the Reynolds number of the particle increases (increasing flow rate is one way to increase Reynolds number), the shear-gradient lift force increases faster than the wall-effect lift force. But the particles with higher aspect ratios rotate and experience a greater wall-effect lift force and return to the center. This relationship with increasing particle Reynolds number allows this method to scale with flow rate, while the previous methods do not.

Four different shape-activated particle-sorting (SAPS) devices were designed with varying numbers of outlets, outlet resistances, channel aspect ratios and flow rates. Three of the devices used 6 µm spheres and their derived ellipsoids, while the final device used the 3 µm particles. All of the devices had varying performances, but the researchers selected device C, which had 7 outlets and isolated 88% of the spheres with 87% purity, 49% of 1:5 rods with 78% purity and 77% of 1:3 rods with 80% purity, for sorting yeast cells. Yeast cells are normally spherical, but form a bispherical twin or aggregate when budding. This change in shape is similar to the varying aspect ratios examined previously. It is useful to synchronize cell cycle stages, but this may be achieved with chemicals that alter cell physiology, changes in temperature or size filtration. SAPS C was able to extract nondividing singles with high yield and purity up to 94% and 54% of budded yeast cells were recovered at 31% purity, which increased from 6.6% purity at the inlet. According to my rough measurements from the paper’s figures, the aspect ratio of budding yeast is less than 2:1 which may explain the difference in performance compared to the original ellipsoids. Previously, Sugaya et al. used a 5 outlet HDF system to separate budding yeast cells. In comparison, Sugaya achieved up to 69.4% purity of budding cells in one outlet, up from 39.4% at the inlet. This same outlet recovered 28.8% of budding cells while another outlet recovered 65.2%. There is still room for improvement of Di Carlo’s budding yeast yield, but this operated at 1500 cell/s, compared to 100 cell/s from previous work that utilized dielectrophoretic forces according to the opacity of dividing yeast.

Future Work

Di Carlo has proposed that this work be used to sort shaped particles in other areas to improve cytometry that operates on spherical particles, alignment of barcoded particles, and identification of microalgae that vary in size and shape. Interestingly, he also introduced the capability of this setup for a non-biological process: improving cement. Cement strength and stability are affected by particle shape and size and could benefit from shape based separation. According to Dr. Di Carlo, “… [Cement] particles that are too large may not react completely in internal regions of the particle, while smaller particles with very high surface area to volume ratios can react too quickly and may not be stable.” I’m excited to see microfluidics expand into more established industries and further demonstrating real-world potential to be a more cost effective, accessible technology.




Masaeli, M., Sollier, E., Amini, H., Mao, W., Camacho, K., Doshi, N., Mitragotri, S., Alexeev, A., & Di Carlo, D. (2012). Continuous Inertial Focusing and Separation of Particles by Shape Physical Review X, 2 (3) DOI: 10.1103/PhysRevX.2.031017

Di Carlo, D., Irimia, D., Tompkins, R., & Toner, M. (2007). Continuous inertial focusing, ordering, and separation of particles in microchannels Proceedings of the National Academy of Sciences, 104 (48), 18892-18897 DOI: 10.1073/pnas.0704958104

Sugaya, S., Yamada, M., & Seki, M. (2011). Observation of nonspherical particle behaviors for continuous shape-based separation using hydrodynamic filtration Biomicrofluidics, 5 (2) DOI: 10.1063/1.3580757

Studying the Effects of Confinement on Cell Division

On Microfluidic Future I like reviewing advancements in therapeutic or diagnostic devices because I’m really drawn to those areas of research. Every once in a while, however, I take interest in research for the sake for knowledge, like the Root Chip. I recently came across an article from Dino Di Carlo of UCLA that describes a microfluidic device used to study cancer cells. The article, “Increased Asymmetric and Multi-Daughter Cell Division in Mechanically Confined Microenvironments” appeared in PLoS ONE, which is an open access journal (very cool!).

Specifically, Di Carlo’s device is used to study the effects of the mechanical environment on cancer cells during division. It’s commonly known that the course of the cell cycle is affected by soluble factors, but the cell’s mechanical interaction with the environment also affects its morphology, differentiation and cell cycle. Changes in confinement and substrate elasticity were tested using the HeLa cervical cancer cell line in this study. The authors looked for several deviations from standard cell division including delayed mitosis, multi-daughter mitosis events, unevenly sized daughter cells and induction of cell death.

Device Design

A. The device features posts to confine cells with continuous perfusion. B. The device is deformable, bringing the two layers into contact. C. The device has variable elasticity and confinement height.

A. The device features posts to confine cells with continuous perfusion. B. The device is deformable, bringing the two layers into contact. C. The device has variable elasticity and confinement height.

Di Carlo’s device has a bottom layer and an elevated PDMS layer supported by posts with varying height for control over cell confinement. In a relaxed state, there is a 15 µm clearance between the posts and bottom layer. When pressure is applied to the device, the two layers meet which confines the cells between posts and reduces the clearance to 3 µm or 7 µm. Additionally, the top layer has an elasticity of 130 KPa or 1 MPa. The device is designed to allow media to flow throughout all the confining chambers, eliminating the possibility of cell death due to a toxic environment.


Induced multi-daughter division resulted in 3, 4 and even 5 daughters

Induced multi-daughter division resulted in 3, 4 and even 5 daughters

In an unconfined environment, a HeLa cell would normally ball up into a sphere during mitosis, which would take no longer than 140 minutes. But with increased confinement and stiffness, the authors witnessed multi-daughter mitosis (one cell dividing into three or four daughter cells), unbalanced daughter sizes, prolonged mitosis and cell death. Resulting control cells from division would often be spheres with a diameter of 20 µm, while the confined cells would be highly asymmetric with diameters 40-80 µm. Increases in stiffness and confinement generally increased the odds of abnormal cell division, with some clear observed patterns. Under low compression of 7µm and stiffness of 130 KPa, 90% of multi-daughter divisions resulted in three cells. When a stiffness of 1 MPa was applied to the same low compression, 85% of multi-daughter divisions resulted in four cells. The authors believe that the cells aren’t able to effectively deform the stiffer substrate and are limited in how spherical they can be before mitosis. The confined shape may also affect chromosomes lining up at the metaphase plane(s), resulting in a bias toward multi-daughter divisions. The multi-daughter cell divisions can produce viable cells, which subsequently can undergo their own multi-daughter division, and also generate daughters that re-fuse after division.

The authors also hypothesized that when the cells are forced to divide in a discoid shape, signaling and regulation may be affected. Diffusion or active transport of signals would take much longer to traverse the large cross-section of the cell, and the force of cytoskeletal elements might be diminished across the same large distance.


This work has produced some findings that may not be totally surprising, but are definitely peculiar. A follow-up to the findings generated here would surely add to the increasing knowledge base of cancer cell behavior. In its current form, this information wouldn’t lead to any new treatments, although under high confinement 70% of cell cycles resulted in cell death, which holds potential in therapeutic applications. Studying diseases help us learn more about healthy cells because we can see what goes wrong when specific elements fail, but I’m also interested in seeing how healthy cells react under the same mechanical conditions. The microfluidic device itself also has potential beyond the study of cellular life cycles: One area in particular includes investigating the effects of mechanical strain in osteocyte and chondrocyte differentiation.


Henry Tat Kwong Tse, Westbrook McConnell Weaver, & Dino Di Carlo (2012). Increased Asymmetric and Multi-Daughter Cell Division in Mechanically Confined Microenvironments PLoS ONE, 7 (6)

Vortices, not Vortexing: Replacing the Centrifuge with a Lab-on-a-Chip


In case you didn’t get the first part of my title, let me tell you a little about centrifugation. Centrifugation is a very common research technique. A solution is centrifuged to isolate suspended particles by spinning it around at high speeds. Depending on the weight of the particles and the force of the centrifuge, the heavier particles will form a pellet at the bottom of the container. The rest will still be suspended in fluid. Depending on which particles you’re after, you can continue doing this by removing the fluid and changing the forces in order to manipulate which particles form the next pellet. But in between each step the pellet must be resuspended and is vortexed (with a vortex mixer) to break up the pellet. Now, keep the definition of vortexing and the process of centrifugation in mind.

Researchers from the University of California, Los Angeles have proposed a ‘Centrifuge-on-a-Chip’ that would replace its bench top counterpart. The article, “Automated cellular sample preparation using a Centrifuge-on-a-Chip” by Di Carlo et al. was featured on the cover of the 2011 Issue 17 of Lab on a Chip. The article proposes a streamlined system that requires no external forces which, like I said before, makes it much more powerful. In order to evaluate the competency of new technology we need to compare it to a golden standard that accomplishes its task the best, no matter how long it may take. In this case, it is the centrifuge and we’ll monitor its three main jobs throughout this post:

  • Separation of cells by size or density
  • Concentration of cells
  • Labeling of cells via solution exchange

The authors’ forceless Centrifuge-on-a-Chip is defined by its structure. It features a simple channel that suddenly widens. It is this widening that creates vortexes, enabling filtration. When the particles were flowing through the plain channel, they were held in place by a force that pushes them towards the wall and a force by the wall that pushes them away. When the channel widens (the wall effectively disappears) they succumb to the force pushing them away from the center of the channel. The rate that particles move out of the channel area, into the vortex region, is proportional to their size (greater than the square of the diameter). Therefore only particles above a certain threshold can be captured. The particles are able to be selectively released from their orbits by decreasing the flow rate.

Beyond simply organizing different beads, the authors validate the filtration process by separating and concentrating circulating tumor cells (CTCs) from dilute blood. Given that CTCs can be 33% to 900% greater than size than blood cells, they should be able to be separated using the vortices. This was confirmed by processing 10 ml samples with ~500 CTCs and 2.5 billion blood cells. The end result was a volume of 200 µl that recovered about 20% of the CTCs, which constituted 40% of the volume.

Now, so far we’ve been able to see the device filter cells by size and concentrate the cells, both things on our list. The only thing left to do is see if this is able to label the cells like a normal centrifugation process can. In order to traditionally label cells, they must first be incubated with the labels and centrifuged into a pellet. The fluid with extra labels must be removed and then the cells are resuspended. With the Centrifuge-On-A-Chip, the cells merely have to be caught in a vortex, exposed to the labels and rinsed. The traditional and new techniques were used to label intracellular proteins, cell surface proteins, enzyme substrates and DNA. I’m not a labeling expert, but the comparison between the two techniques seems pretty similar. There seems to be more of a difference for the labeling of the cell surface proteins, but I’m not sure if it is significant. They also demonstrated the ability to tag with primary and secondary antibodies, as well as microbeads. After 5 minutes the vortexed cells were labeled with the same number of microbeads as the traditional cells after 30 min (the same amount of labels was used for both techniques in all experiments). Further, after 30 minutes, the vortexed cells had twice the number of microbeads as the traditional cells. Safe to say, this technique can definitely label cells at least as well as the traditional method.


The Centrifuge-On-A-Chip is clearly a viable contender against the traditional bench top centrifuge and its techniques. While there is no given price for each device, or how many devices would be necessary to provide the same output as a centrifuge, it will certainly cost less than thousands of dollars. Additionally, it is portable and can process the tasks in less time. When compared to the Centrifuge-On-A-Chip’s competitors, it still comes off fairly well. Some others are only able to filter cells and not concentrate them. Some immobilize the cells on membranes, which may prevent them from performing assays, or may clog the membranes. There was also no change in viability before and after the vortexing. However, it is necessary to dilute the blood before it can be processed, but I doubt that this would be negative enough to prevent someone from using this entirely. I suppose that we’ll just have to wait and see how quickly this can enter the market, and how cheaply it can be produced.


Mach, A., Kim, J., Arshi, A., Hur, S., & Di Carlo, D. (2011). Automated cellular sample preparation using a Centrifuge-on-a-Chip Lab on a Chip, 11 (17), 2827-2834 DOI: 10.1039/c1lc20330d