Exosomes are nanoscale, membrane-bound particles released by cells. In experiments at Dartmouth, researchers attached magnetic beads to exosomes so they could collect them on small magnetic structures, or micromagnets. Under flow, those structures have to hold onto the bead-exosome complexes against the drag of the passing liquid. The team tested three micromagnet patterns to find out where capture was likely and which shape to take forward into a microfluidic device.
In their 2024 papers, the team combined magnetic-field simulations in FEATool Multiphysics with fluid-flow simulations in CFDTool. Their MATLAB model used microscope images of fabricated micromagnets to compare capture regions around triangular, hexagonal, and Y-shaped patterns. The Y-shaped arrangement had larger predicted capture regions. Together with initial microwell experiments, this result led the team to select it for microfluidic testing. The simulations used normalized fields and flow to compare the patterns, so the predicted capture regions do not give an absolute capture efficiency.
Micromagnet geometry comparison, reproduced from Figure 3 of Molinski et al.. Panels A-C show fabricated unit-cell geometries; D-F combine FEATool magnetostatic and CFDTool flow results; G-I show the vector-based motion analysis; J-L identify predicted capture zones where magnetic force dominates drag. All scale bars are 50 µm.
Micromagnet geometry and field models
For the magnets, the team used a mixture of iron oxide nanoparticles and poly(dimethylsiloxane) (PDMS). They patterned this magnetic elastomer in a layer about 35 µm thick, making the triangles, hexagons, and Y-shaped structures used in the calculations and capture tests.
Manufacturing left differences between the intended shapes and the finished magnets. To include those differences in the model, the team started with microscope images of the fabricated patterns. Supplementary Figure 8 follows that preparation step by step: conversion from JPEG to DXF, cleaning and smoothing the outlines of one unit cell, and creation of a three-dimensional model saved as an STL file. This was the geometry imported into MATLAB.
Within MATLAB, the model had two branches. The flow branch used CFDTool for computational fluid dynamics (CFD), giving the velocity components around the magnets in two dimensions. For the magnetic branch, the team used FEATool to solve the three-dimensional magnetostatic problem. They justified the simpler, planar flow calculation by the pattern’s symmetry along the micromagnet height.
A permanent magnet beneath the substrate supplied the external, or bias, field and magnetized the structures perpendicular to their patterned surface. For the unit-cell calculation, the authors treated this field as uniform and strong enough to saturate the micromagnets. Each micromagnet therefore had uniform magnetization in the model.
MATLAB modeling workflow, reproduced from Supplementary Figure 8 of Molinski et al.. The diagram shows geometry preparation, the separate CFDTool and FEATool calculations, and the steps used to align their results and identify capture regions.
Particle motion and capture regions
Combining these results took several steps. First, the team averaged the three-dimensional magnetic field over the height of the magnets. That gave them a planar field map to use with the two-dimensional flow calculation.
The flow results already described velocity, but the magnetic field still had to be converted into particle motion. From its spatial variation and the bead’s magnetic properties, the team calculated the force on the bead-exosome complex, then the drift velocity that force would produce in the liquid. This is the magnetic part of the motion, usually called magnetophoresis; the fluid carries the particle along as well.
After interpolation and alignment of the datasets, the MATLAB code added the magnetic drift and fluid velocity components at each position and calculated particle trajectories. The capture regions in Figure 3 mark where the magnetic force was greater than drag.
Both field and flow were normalized for the geometry comparison. A scaling factor on the magnetic contribution let the team vary its strength relative to drag and examine the resulting capture regions.
Of the three patterns, the Y-shaped arrangement had the largest predicted capture regions. The flow slowed between the arms of the Y. Around it, small triangles split the stream toward either the central Y or neighboring unit cells. The authors linked this pattern of slow flow and divided streams to the larger capture regions. Their result applies to the whole layout, including those triangles.
Microwell and microchannel experiments
For the first experiments, the team used open microwells above a permanent magnet. Antibody-coated magnetic beads were attached to exosome standards, and a fluorescent membrane stain made the exosomes visible under a confocal microscope. Most of the fluorescence appeared at the micromagnets, with little elsewhere, consistent with the predicted capture locations. Separate controls checked whether the stain itself was sticking to the beads.
With the simulation and microwell results in hand, the team chose the Y-shaped pattern for continuous-flow tests. The magnets were enclosed in a serpentine microchannel with nine sections. A syringe pump drove about 200 µL of labeled sample through the device at 3 mL/h before the captured material was imaged.
Fluorescence was uneven along the channel. Average image intensity peaked near the inlet and outlet, with a smaller peak in the middle. The gradient of the external magnet’s field followed a similar pattern. The authors interpreted this as suggesting that both the external field gradient and the local magnetization of the micromagnets could affect capture. These intensity measurements included the image background; they were used to compare trends, not to measure the fraction of exosomes recovered.
Model assumptions and limits
The simulations helped select a pattern for testing, but a larger predicted capture region does not by itself establish a higher recovery rate. The preprint gives no quantitative comparison between capture-region area and measured capture efficiency. Bead properties, the applied field, flow conditions, and device geometry also affect capture.
The uniform bias field assumed around a unit cell is another limit. Across the full device, the external field varies, and the channel observations suggest that this variation influences where exosomes collect. Averaging over the micromagnet height also replaces the field’s vertical variation with a single planar map. Combined with a planar flow model, it leaves particle transport through the full three-dimensional channel unresolved.
For a reproduction of the model, the preprint provides the sequence of calculations and points to an earlier modeling study. It does not give mesh-convergence results, software versions, or all the magnetic material and boundary-condition values needed to reproduce the absolute fields from this document alone.
Related FEATool examples address individual parts of this workflow. The Magnetic Field Around a Horseshoe Magnet tutorial illustrates magnetization and static-field visualization in two dimensions. The Flow Around a Cylinder tutorial covers incompressible flow around a solid obstruction, including velocity and pressure fields. These examples introduce related field calculations; reproducing the research model would additionally require its image-derived geometry, magnetic-to-particle conversion, and registration of the separate solutions.
References
J. H. Molinski, S. Parwal, and J. X.J. Zhang. Laser-Manufactured Magnetic Microchips for Exosome Isolation and Patterning, IEEE 37th International Conference on Micro Electro Mechanical Systems (MEMS), pp. 336-33 2024, doi: 10.1109/MEMS58180.2024.10439482.
J. H. Molinski, S. Parwal, and J. X.J. Zhang. Laser-patterning of micromagnets for magnetophoretic biomolecule capture, Research Square preprint, posted March 26, 2024, doi: 10.21203/rs.3.rs-4011875/v1.