Work by our group at IBM Research Europe in Zurich has led to a new method for the rapid implementation of microfluidic operations. By tailoring the potential landscape inside a flow cell, we form so-called “virtual channels” on demand to perform high-precision guiding and transport, splitting, merging and mixing of microfluidic flows. This allows to bypass traditional fabrication of microfluidic chips and provides a whole new range in dynamic functionality.
Need for reconfigurability
So far, the function of a microfluidic system has been intimately linked to its geometrical properties – often requiring researchers to fabricate and test several design iterations to achieve a desired performance. Especially heterogeneous samples, for example tissue sections which have variations from sample to sample, are challenging to address using hard-coded microchannel structures. While several demonstrations of rapid prototyping technologies for microfluidic structures have been shown, no method allows to implement truly adjustable and reconfigurable microfluidic channel structures
Our approach – virtual channels
Figure 1: Virtual channels are formed by tailoring the potential landscape within a flow cell to match a user-defined channel geometry.
Our process of forming custom microfluidic patterns starts with the definition of a desired channel geometry on a computer interface. An analytical framework on a standard computer or laptop then computes the potential landscape required to reproduce the defined pattern within a microfluidic flow cell. We can picture this potential landscape as a slope in a sandbox, with small hills and pits in it, guiding a creek downhill along a specific path. While in the sandbox liquid moves due to differences in gravitational potential, in microfluidic flow cells liquid moves due to differences in the so-called velocity potential: an opening from which liquid is injected would form a local hill in the velocity potential landscape, whereas withdrawing liquid would form a pit. By leaving one end of our flow cell open, which corresponds to enforcing a velocity potential of 0, and injecting buffer liquid from aperture A (see Fig. 1) from the opposite closed end, we tilt our velocity potential landscape in the flow cell in analogy to the slope in the sandbox. The ability to now manipulate this potential landscape in such a way that the flow of a liquid injected from aperture B exactly matches a user-defined channel geometry mainly depends on the number of degrees of freedoms available to alter the potential: each additional aperture allows to add an inflection point to the produced flow path. We found that using a 4×4 array of apertures, we can very well reproduce a wide range of standard channel geometries.
Having computed, which flow rates need to be applied to the apertures in the 4×4 array to form the desired pattern, our software now sends corresponding commands to a set of syringe pumps. After about 3 seconds, all flow rates have stabilized, and a reagent injected from aperture B will flow along the custom shaped virtual channel. A full transition between two completely different channel patterns takes about 30 seconds in our demonstrator setup.
This work on a new approach to realize reconfigurable microfluidics was featured on the front cover in a recent issue of the Lab on a Chip journal, the leading scientific journal in the field of microfluidics and bioanalysis published by the Royal Society of Chemistry.
Type of microfluidic features possible with virtual channels
Having set up the system, our goal then was to prove that virtual channels can perform standard microfluidic tasks in an on-demand and dynamic fashion. These key functions are the transport, splitting, merging and mixing of microscale flows. Virtual channels can be used to implement all these functions dynamically, allowing to precisely tune the function until a desired performance is reached. Especially mixing is a challenge in classical microfluidics, as it usually is a slow, diffusion-driven process due to the laminar nature of flow in. Unlike in traditional channels, in our system, mixing can be implemented as a two-step process: two liquids to be mixed can be aspirated into a thin capillary attached to one of the apertures in the array, where the interface between the two liquids is significantly prolonged to promote diffusive mixing. This approach enables efficient and complete mixing of liquids, which can then be re-injected on demand and guided to desired target locations using corresponding virtual channels.
Figure 2: Microfluidic key functions implemented by means of virtual channels.
Virtual channels are readily compatible with most experimental settings: any kind of sample surface could be flipped and used as the top cover of the flow cell (see Fig. 2 bottom). Further, the concept of virtual channels is highly scalable: while our channels have widths on the order of a few hundred micrometers, they could be scaled up to meters or down to few tens of micrometer in width by applying corresponding flow boundary conditions.
In the field of electronics, field programable gate arrays (FPGA) have proven that the ability to directly implement customized functions and being able to adjust them is game-changing in several ways and enables novel applications. We see our platform as the microfluidic equivalent to FPGAs We think that virtual channels have the potential to revolutionize the way in which microfluidic functions are implemented. We especially see great potential in combining virtual channels with machine learning algorithms to enable fast adaptations to rapidly changing conditions, for example for the in-line synthesis of chemical species.
This work was supported by the European Research Council Starting Grant BioProbe (311122) and by the European Research Council’s Proof-of-Concept Grant CellProbe (842790). and a Doc.Mobility grant from the Swiss National Science Foundation to David Taylor (P1ELP2_181841).
Our team of researchers recently published paper “Fine-Grained Visual Recognition in Mobile Augmented Reality for Technical Support,” in IEEE ISMAR 2020, which outlines an augmented reality (AR) solution that our colleagues in IBM Technology Support Services use to increase the rate of first-time fixes and reduce the mean time to recovery from a hardware disruption.
A team formed by IBM Research scientist Dr. Leo Gross, University Regensburg professor Dr. Jascha Repp, and University Santiago de Compostela assistant professor Dr. Diego Peña Gil has received a European Research Center (ERC) Synergy Grant for their project “Single Molecular Devices by Atom Manipulation” (MolDAM).
A new AI model, developed by IBM Research and Pfizer, has used short, non-invasive and standardized speech tests to help predict the eventual onset of Alzheimer’s disease within healthy people with an accuracy of 0.7 and an AUC of 0.74 (area under the curve).