Optimized design of obstacle sequences for microfluidic mixing in an inertial regime.

Mixing is a basic but challenging step to achieve in high throughput microfluidic applications such as organic synthesis or production of particles. A common approach to improve micromixer performance is to devise a single component that enhances mixing through optimal convection, and then sequence multiple such units back-to-back to enhance overall mixing at the end of the sequence. However, the mixing units are often optimized only for the initial non-mixed fluid composition, which is no longer the input condition for each subsequent unit. Thus, there is no guarantee that simply repeating a single mixing unit will achieve optimally mixed fluid flow at the end of the sequence. In this work, we analyzed sequences of 20 cylindrical obstacles, or pillars, to optimize the mixing in the inertial regime (where mixing is more difficult due to higher Péclet number) by managing their interdependent convection operations on the composition of the fluid. Exploiting a software for microfluidic design optimization called FlowSculpt, we predicted and optimized the interfacial stretching of two co-flowing fluids, neglecting diffusive effects. We were able to quickly design three different optimal pillar sequences through a space of 3220 possible combinations of pillars. As proof of concept, we tested the new passive mixer designs using confocal microscopy and full 3D CFD simulations for high Péclet numbers (Pe ≈ O(105-6)), observing fluid flow shape and mixing index at several cross-sections, reaching mixing efficiencies around 80%. Furthermore, we investigated the effect of the inter-pillar spacing on the most optimal design, quantifying the tradeoff between mixing performance and hydraulic resistance. These micromixer designs and the framework for the design in inertial regimes can be used for various applications, such as lipid nanoparticle fabrication which has been of great importance in vaccine scale up during the pandemic.

Scanning two-photon continuous flow lithography for the fabrication of multi-functional microparticles.

In this work, we demonstrate the high-throughput fabrication of 3D microparticles using a scanning two-photon continuous flow lithography (STP-CFL) technique in which microparticles are shaped by scanning the laser beam at the interface of laminar co-flows. The results demonstrate the ability of STP-CFL to manufacture high-resolution complex geometries of cell carriers that possess distinct regions with different functionalities. A new approach is presented for printing out-of-plane features on the microparticles. The approach eliminates the use of axial scanning stages, which are not favorable since they induce fluctuations in the flowing polymer media and their scanning speed is slower than the speed of galvanometer mirror scanners.

FlowSculpt: software for efficient design of inertial flow sculpting devices.

Flow sculpting is a powerful method for passive flow control that uses a sequence of bluff-body structures to engineer the structure of inertially flowing microfluidic streams. A variety of cross-sectional flow shapes can be created through this method, offering a new platform for flow manipulation or material fabrication useful in bioengineering, manufacturing, and chemistry applications. However, the inverse problem in flow sculpting - designing a device that produces a target fluid flow shape - remains challenging due to the complex, diverse, and enormous design space. Solutions to the inverse problem have been constrained to single-material fluid streams that are shaped into top-bottom symmetric shapes due to the bluff-body structures available in current libraries (pillars) that span the height of the channel. In this work, we introduce multi-material design and symmetry-breaking flow deformations enabled by half-height pillars, presented within an extremely fast simulation method for flow sculpting yielding a 34-fold reduction in runtime. The framework is deployed freely as a cross-platform application called "FlowSculpt". We detail its implementation and usage, and discuss the addition of enhanced search operations, which enable users to more easily design flow shapes that replicate their input drawings. With FlowSculpt, the microfluidics community can now quickly design flow shaping microfluidic devices on modest hardware, and easily integrate these complex physics into their research toolkit.

A deep learning framework for causal shape transformation.

Recurrent neural network (RNN) and Long Short-term Memory (LSTM) networks are the common go-to architecture for exploiting sequential information where the output is dependent on a sequence of inputs. However, in most considered problems, the dependencies typically lie in the latent domain which may not be suitable for applications involving the prediction of a step-wise transformation sequence that is dependent on the previous states only in the visible domain with a known terminal state. We propose a hybrid architecture of convolution neural networks (CNN) and stacked autoencoders (SAE) to learn a sequence of causal actions that nonlinearly transform an input visual pattern or distribution into a target visual pattern or distribution with the same support and demonstrated its practicality in a real-world engineering problem involving the physics of fluids. We solved a high-dimensional one-to-many inverse mapping problem concerning microfluidic flow sculpting, where the use of deep learning methods as an inverse map is very seldom explored. This work serves as a fruitful use-case to applied scientists and engineers in how deep learning can be beneficial as a solution for high-dimensional physical problems, and potentially opening doors to impactful advance in fields such as material sciences and medical biology where multistep topological transformations is a key element.

Shaped 3D microcarriers for adherent cell culture and analysis.

Standard tissue culture of adherent cells is known to poorly replicate physiology and often entails suspending cells in solution for analysis and sorting, which modulates protein expression and eliminates intercellular connections. To allow adherent culture and processing in flow, we present 3D-shaped hydrogel cell microcarriers, which are designed with a recessed nook in a first dimension to provide a tunable shear-stress shelter for cell growth, and a dumbbell shape in an orthogonal direction to allow for self-alignment in a confined flow, important for processing in flow and imaging flow cytometry. We designed a method to rapidly design, using the genetic algorithm, and manufacture the microcarriers at scale using a transient liquid molding optofluidic approach. The ability to precisely engineer the microcarriers solves fundamental challenges with shear-stress-induced cell damage during liquid-handling, and is poised to enable adherent cell culture, in-flow analysis, and sorting in a single format.

Deep Learning for Flow Sculpting: Insights into Efficient Learning using Scientific Simulation Data.

A new technique for shaping microfluid flow, known as flow sculpting, offers an unprecedented level of passive fluid flow control, with potential breakthrough applications in advancing manufacturing, biology, and chemistry research at the microscale. However, efficiently solving the inverse problem of designing a flow sculpting device for a desired fluid flow shape remains a challenge. Current approaches struggle with the many-to-one design space, requiring substantial user interaction and the necessity of building intuition, all of which are time and resource intensive. Deep learning has emerged as an efficient function approximation technique for high-dimensional spaces, and presents a fast solution to the inverse problem, yet the science of its implementation in similarly defined problems remains largely unexplored. We propose that deep learning methods can completely outpace current approaches for scientific inverse problems while delivering comparable designs. To this end, we show how intelligent sampling of the design space inputs can make deep learning methods more competitive in accuracy, while illustrating their generalization capability to out-of-sample predictions.

Micropillar sequence designs for fundamental inertial flow transformations.

The ability to control the shape of a flow in a passive microfluidic device enables potential applications in chemical reaction control, particle separation, and complex material fabrication. Recent work has demonstrated the concept of sculpting fluid streams in a microchannel using a set of pillars or other structures that individually deform a flow in a predictable pre-computed manner. These individual pillars are then placed in a defined sequence within the channel to yield the composition of the individual flow deformations - and ultimately complex user-defined flow shapes. In this way, an elegant mathematical operation can yield the final flow shape for a sequence without an experiment or additional numerical simulation. Although these approaches allow for programming complex flow shapes without understanding the detailed fluid mechanics, the design of an arbitrary flow shape of interest remains difficult, requiring significant design iteration. The development of intuitive basic operations (i.e. higher-level functions that consist of combinations of obstacles) that act on the flow field to create a basis for more complex transformations would be useful in systematically achieving a desired flow shape. Here, we show eight transformations that could serve as a partial basis for more complex transformations. We initially used in-house, freely available custom software (uFlow), which allowed us to arrive at these transformations that include making a fluid stream concave and convex, tilting, stretching, splitting, adding a vertex, shifting, and encapsulating another flow stream. The pillar sequences corresponding to these transformations were subsequently fabricated and optically analyzed using confocal imaging - yielding close agreement with uFlow-predicted shapes. We performed topological analysis on each transformation, characterizing potential sequences leading to these outputs and trends associated with changing diameter and placement of the pillars. We classify operations into four sets of sequence-building concatenations: stacking, recursion, mirroring, and shaping. The developed basis should help in the design of microfluidic systems that have a phenomenal variety of applications, such as optofluidic lensing, enhanced heat transfer, or new polymer fiber design.