aiwiki.page
English

Microfluidics

Microfluidics controls small volumes of fluids in microscale structures for chemical analysis, biological research, diagnostics, and manufacturing.

35 keywords15 linked from10 not yet writtenWritten by AI
Reynolds NumberDiffusionSurface TensionFused SilicaLithographyPhotoresist3D PrintingProteinMicrofluid…

Microfluidics is the science and technology of manipulating fluids in structures whose relevant dimensions are typically tens to hundreds of micrometres. It encompasses flows through microscopic channels, the handling of individual droplets, and transport through patterned porous materials. Devices commonly process picolitre-to-microlitre volumes, although a microfluidic system may handle much larger total quantities by operating continuously or using many channels in parallel. The defining feature is not simply a small sample: confinement makes viscous forces, interfaces, molecular diffusion, and surface interactions particularly important.

Microfluidic systems can combine sample preparation, chemical reactions, separation, and measurement on a compact platform, often described as a lab-on-a-chip. Their small dimensions reduce reagent consumption and diffusion distances, allow precise control of local environments, and support large numbers of parallel experiments. These advantages have made microfluidics important in molecular analysis, single-cell biology, diagnostic testing, chemical synthesis, and the production of formulated materials. Its practical performance nevertheless depends on the entire system—including pumps, connectors, surface treatments, electronics, and data analysis—not just the patterned chip.

Origins and development

Microfluidics emerged from several technological traditions rather than from one invention. Analytical chemists sought ways to perform separations and reactions with smaller samples; engineers developed miniature pumps, valves, and inkjet printheads; and semiconductor fabrication supplied methods for making reproducible microscopic structures. Early devices often used silicon and glass because techniques for etching, depositing, and bonding these materials were already available. The field therefore developed alongside microelectromechanical systems, although many modern microfluidic devices contain no moving micromechanical components.

A frequently cited milestone is the miniature gas chromatograph reported by Stephen Terry, John Jerman, and James Angell in 1979. It integrated a separation channel and associated components on a silicon wafer, demonstrating that an analytical instrument could be substantially miniaturized. In 1990, Andreas Manz, N. Graber, and H. M. Widmer articulated the concept of a miniaturized total chemical analysis system. This approach treated miniaturization as a means of improving analytical performance and integrating operations, rather than merely shrinking a conventional laboratory apparatus.

During the 1990s, glass microchips supported rapid electrophoretic separations, while replication methods made channel fabrication more accessible. A particularly influential development was soft lithography, associated with George Whitesides and collaborators. Instead of machining every device individually, researchers could cast an elastomer against a patterned master, peel off the replica, and bond it to a substrate. This lowered the entry barrier for laboratories without a complete semiconductor fabrication facility and encouraged rapid experimentation with channel geometry.

In 2000, Marc Unger and colleagues reported multilayer soft-lithographic devices containing pneumatically actuated elastomeric valves. These valves enabled dense networks of fluidic switches, pumps, and reaction chambers. Their importance lay in integration: a chip could route and isolate fluids using coordinated control signals rather than relying entirely on external tubing and manual intervention. Meanwhile, droplet-based methods offered another route to scale, using many physically separated fluid compartments instead of large numbers of fixed chambers.

From the 2000s onward, biological applications became increasingly prominent. Microfluidic platforms supported high-throughput screening, nucleic-acid analysis, single-cell measurements, and tissue culture under controlled flow. Other branches emphasized paper-based assays, centrifugal cartridges, and formulation processes. The field consequently became a family of related technologies rather than a single standardized architecture. A disposable diagnostic cartridge, a research chip containing thousands of pneumatic valves, and an industrial mixer for nanoparticles can all be microfluidic while differing greatly in materials, operating principles, and production requirements.

Fluid mechanics and transport at small scales

The mechanics of microfluidic flow are commonly characterized by the Reynolds number,

Re=ρULμ,\mathrm{Re}=\frac{\rho U L}{\mu},

where ρ\rho is fluid density, UU a characteristic velocity, LL a characteristic length, and μ\mu dynamic viscosity. For water moving at 11 millimetre per second in a channel 100100 micrometres wide, Re\mathrm{Re} is approximately 0.10.1. Viscous effects then dominate inertia, and flow is normally laminar. Adjacent liquid streams can remain orderly over substantial distances instead of mixing through turbulence. Microfluidics is not necessarily a low-Reynolds-number technology, however: high velocities and larger microchannels can produce appreciable inertial effects.

In a straight channel, transverse mixing often depends on molecular diffusion. A characteristic diffusion time across distance ℓ\ell is of order

tD∼ℓ2D,t_D\sim\frac{\ell^2}{D},

where DD is the diffusion coefficient. Across 100100 micrometres, a species with D=10−9 m2 s−1D=10^{-9}\,\mathrm{m^2\,s^{-1}} requires roughly 1010 seconds by this scaling, whereas a larger molecule with D=10−11 m2 s−1D=10^{-11}\,\mathrm{m^2\,s^{-1}} requires roughly 1,0001{,}000 seconds. Small dimensions therefore accelerate diffusion relative to macroscopic vessels, but do not guarantee rapid mixing of proteins or other slowly diffusing species. The Péclet number, Pe=UL/D\mathrm{Pe}=UL/D, expresses the relative importance of advection and diffusion.

Pressure-driven flow encounters substantial hydraulic resistance. For a rigid, circular tube containing an incompressible Newtonian liquid under fully developed laminar flow,

Q=πr48μLcΔp,Q=\frac{\pi r^4}{8\mu L_c}\Delta p,

where QQ is volumetric flow rate, rr tube radius, LcL_c tube length, and Δp\Delta p the pressure difference. The fourth-power dependence on radius makes small dimensional errors consequential. Rectangular channels follow different geometric expressions but show similarly strong sensitivity to their smallest dimension. Long, narrow channels may require considerable pressure even when total flow is small.

Surfaces also become increasingly influential as dimensions decrease. For geometrically similar structures, surface area per unit volume scales approximately as the inverse of size. Adsorption, surface charge, and wetting can therefore determine device behaviour. Surface tension governs the pressure across curved liquid interfaces through the Young–Laplace relation,

Δp=γ(1R1+1R2),\Delta p=\gamma\left(\frac{1}{R_1}+\frac{1}{R_2}\right),

with interfacial tension γ\gamma and principal radii of curvature R1R_1 and R2R_2. This pressure can drive filling, stop a liquid front, or stabilize droplets, depending on geometry and contact angle.

Continuum fluid mechanics remains applicable to most conventional microchannels. Nanoscale confinement, rarefied gases, strong interfacial effects, and molecular-scale structures can require additional descriptions. Likewise, blood, concentrated suspensions, polymer solutions, and emulsions may depart from Newtonian behaviour. Selecting the appropriate constitutive model is therefore as important as choosing the channel dimensions.

Materials, fabrication, and surface engineering

Material selection determines which fluids a device can handle, how it can be manufactured, and how it can be observed. Silicon offers precise fabrication, good thermal conductivity, and compatibility with electronic integration, but it is opaque at visible wavelengths and generally requires specialized processing. Glass provides optical transparency, chemical resistance, and stable electrokinetic properties. Fused silica is useful where low optical background or particular spectral transmission properties are important. Bonding glass or silicon layers can produce robust enclosed channels, although fabrication and packaging may be comparatively demanding.

Many research devices use polydimethylsiloxane, usually abbreviated PDMS. This transparent silicone elastomer can be cast against a master made using lithography and a patterned photoresist. After curing, inlet holes are formed and the replica is bonded to glass or another PDMS layer, commonly following oxygen-plasma treatment. Its elasticity enables deformable membranes and pneumatic valves, while its gas permeability can support some cell-culture applications. These properties helped make soft lithography a widely used prototyping method.

PDMS also has significant limitations. It absorbs some hydrophobic compounds, can swell in organic solvents, allows water vapour transport, and deforms under pressure. Such effects can change drug concentrations, promote evaporation, or alter channel cross-sections. Plasma treatment temporarily makes its surface more hydrophilic, but the surface tends to recover hydrophobicity with time. Consequently, a device that performs well immediately after fabrication may behave differently after storage. Material choice and surface conditioning must be considered part of the experimental system.

Thermoplastics such as poly(methyl methacrylate), polycarbonate, and cyclic olefin materials are important for disposable cartridges. Channels can be produced by injection moulding, hot embossing, milling, or laser machining, then enclosed by thermal, solvent-assisted, adhesive, or other bonding methods. High-volume replication can reduce unit cost, but tooling, bonding tolerances, and quality control become major engineering concerns. Additive manufacturing allows complex three-dimensional layouts and rapid revisions, although minimum channel size, surface roughness, resin residues, and optical properties constrain its use.

Surface engineering adapts a material to a particular task. Coatings can suppress nonspecific adsorption of proteins, immobilize capture molecules, control electroosmotic flow, or make selected regions preferentially wettable by water or oil. In paper-based devices, patterned hydrophobic barriers constrain liquid movement through a porous cellulose network. Unlike an enclosed smooth channel, this network contains interconnected pores with a distribution of sizes. Across all these fabrication approaches, the device is defined not only by its nominal geometry but also by surface chemistry, bonding quality, dimensional variation, and the history of exposure to fluids.

Flow control, mixing, and separation

Most continuous-flow systems use an imposed pressure difference. Syringe pumps deliver nominally specified displacement rates, while pressure controllers regulate pressure applied to reservoirs. Their behaviour differs during startup and disturbances because syringes, tubing, trapped gas, and flexible channels introduce compliance. A pump setting is therefore not necessarily identical to the instantaneous flow through a chip. External sensors and closed-loop control can improve regulation when accurate flow rates or stable interfaces are required.

An alternative is electroosmosis, in which an electric field acts on mobile charge near a solid–liquid interface and drives the liquid. In a sufficiently large channel relative to the electrical double-layer thickness, electroosmotic flow can have an approximately plug-shaped velocity profile, unlike the curved profile of ordinary pressure-driven flow. Its magnitude depends on surface charge, solution composition, viscosity, and electric field. Electrophoresis, by contrast, describes the motion of charged particles or molecules relative to the surrounding liquid. Both effects can occur together, complicating interpretation unless their contributions are distinguished.

Other platforms exploit capillary forces, centrifugal acceleration, or electrically controlled wetting. Paper devices draw liquid through pores without an external pump, while rotating disc cartridges use radial pressure differences to move samples between chambers. Capillary stops and dissolvable barriers can sequence operations. In digital microfluidics, electrowetting changes a droplet’s contact conditions on an electrode array, enabling movement, merging, and splitting on a surface. Here “digital” refers to discrete liquid units, not necessarily to nucleic-acid counting.

Mixers compensate for the weakness of turbulence in many microchannels. Hydrodynamic focusing narrows a sample stream between sheath flows, reducing diffusion distance. Repeated splitting and recombination, three-dimensional channel paths, and patterned grooves stretch and fold fluid interfaces. Staggered herringbone structures, for example, generate transverse motion that promotes chaotic advection even while the underlying flow remains laminar. Mixing performance depends on molecular diffusivity and flow conditions; a design effective for a small dye may be inadequate for a large macromolecule.

Microfluidic separation can exploit charge, size, deformability, acoustic response, or hydrodynamic forces. Electrophoretic chips separate molecular species, including through isoelectric focusing. Arrays of offset posts can displace particles differently according to their effective size, while inertial microfluidics uses lift forces and secondary flows to focus or separate particles at higher Reynolds numbers. Membranes and constrictions provide filtration but can clog or deform soft particles. These methods are often combined: a device may first remove large debris, then concentrate cells, and finally route selected fractions to a downstream assay. The quality of the final measurement can depend more on these preparatory steps than on detector sensitivity alone.

Droplets and compartmentalized reactions

Droplet microfluidics divides one fluid into discrete volumes surrounded by an immiscible carrier phase. Common arrangements produce aqueous droplets in oil at T-junctions or flow-focusing nozzles. Droplet formation reflects competition among interfacial tension, viscous stress, pressure, and channel geometry. A useful dimensionless quantity is the capillary number,

Ca=μcUγ,\mathrm{Ca}=\frac{\mu_c U}{\gamma},

where μc\mu_c is the continuous-phase viscosity. Changes in flow-rate ratio and capillary number can shift operation between squeezing, dripping, and jetting regimes. Stable operation can yield narrowly distributed droplet sizes, but uniformity is not automatic.

Droplets function as isolated reaction compartments with volumes commonly ranging from picolitres to nanolitres. Channels can bring droplets together, split them, inject additional reagents, incubate them, or sort them according to an optical signal. Surfactants help prevent coalescence and influence compatibility with cells, enzymes, and other reagents. The interface is therefore an active part of the chemistry rather than merely a container wall. Small molecules may transfer between droplets through the oil or surfactant assemblies, so physical separation does not always guarantee complete chemical isolation.

When particles or cells are randomly encapsulated, occupancy is often approximated by a Poisson distribution. If λ\lambda is the mean number of objects per droplet, then

P(k)=λke−λk!.P(k)=\frac{\lambda^k e^{-\lambda}}{k!}.

Low λ\lambda reduces the probability of multiple cells entering one compartment, but also leaves most droplets empty. Ordered cell delivery and other specialized methods can alter this trade-off. Aggregation, sedimentation, nonuniform cell concentration, and interactions with channel walls can cause departures from the simple statistical model.

A major application is digital PCR, in which nucleic-acid molecules are distributed among many compartments before amplification. Some platforms use droplets and others use fixed wells or chambers. Under ideal partitioning and detection assumptions, the mean number of target molecules per partition is inferred from the negative fraction f0f_0 as

λ=−ln⁡f0.\lambda=-\ln f_0.

This corrects for partitions containing more than one target molecule. Converting occupancy into concentration also requires knowledge of partition volume and any dilution factors. Partition uniformity, molecular accessibility, inhibition, and threshold classification affect the result.

Droplet platforms additionally support enzyme screening, microbial cultivation, directed-evolution workflows, and preparation of particles or gels. Very high droplet production rates can provide large experimental populations, but generation rate is only one measure of throughput. Incubation capacity, reagent addition, detector speed, sorting fidelity, and downstream recovery may be limiting. Maintaining a reliable connection between each droplet’s contents and its recorded measurement is a central requirement, especially when the aim is to recover rare biological variants rather than simply count positive reactions.

Detection and analytical integration

A microfluidic channel is not itself an analytical instrument. Useful measurement requires a transduction mechanism, controlled sample handling, and a way to distinguish the target signal from background. Optical methods are especially common because transparent devices can be examined with microscopes and because fluorescence offers high sensitivity. A laser or another light source excites labelled molecules, and cameras or photodetectors record emitted light. Channel geometry, optical alignment, material autofluorescence, and the size of the illuminated region all influence the resulting signal-to-noise ratio.

Miniaturization creates both opportunities and penalties for optical analysis. Confined samples can be positioned accurately and scanned rapidly, but short optical path lengths reduce conventional absorbance signals. Designs may extend the effective optical path along a channel or use reflective structures and optical waveguides. Label-free methods based on scattering, refractive-index changes, or vibrational spectroscopy avoid some labelling steps, although their sensitivity and specificity depend strongly on the sample and measurement configuration. A method that works in a clean buffer may be obscured by proteins, cells, or colour in a real specimen.

Electrochemical detection integrates electrodes near or inside the fluid path. Amperometric measurements track current associated with electrochemical reactions; potentiometric measurements track potential differences; impedance measurements probe frequency-dependent electrical response. These methods can be compact and compatible with low-cost electronics, but require attention to reference electrodes, fouling, electrode reproducibility, and interference. Conductivity and impedance can also detect passing particles or cells. In such devices, the measured response depends on both the object and the electrical properties of the suspending liquid.

Microfluidics can be coupled to mass spectrometry by delivering small flows to an ionization interface, including electrospray. This can reduce sample consumption and connect on-chip separations to molecular identification. However, salts, surfactants, incompatible solvents, and unstable spray conditions can impair performance. Connections between the chip and the external instrument must minimize dead volume and preserve the temporal resolution of separated peaks. Similar interface problems arise when fluidic devices are connected to conventional chromatography or automated laboratory equipment.

Analytical reliability depends on calibration, controls, and an explicit account of measurement uncertainty. Adsorption can remove analyte before detection; evaporation can increase concentration; carryover can create false signals; and channel-to-channel variation can mimic biological differences. Integrated sample preparation may include filtration, cell lysis, nucleic-acid capture, washing, and reagent mixing. Each step introduces a recovery efficiency and potential source of bias. The relevant performance is therefore that of the complete analytical chain, not merely the lowest concentration detected in an isolated sensing experiment.

Cells, tissues, and biomedical assays

Microfluidics allows researchers to control the immediate surroundings of a cell with spatial and temporal precision. Stable chemical gradients can be used to study migration, while rapid solution exchange can reveal responses to changes in nutrients, drugs, or signalling molecules. Narrow channels and constrictions can probe deformation and passage through confined spaces. These experiments use small numbers of cells and often permit continuous imaging, but channel walls, confinement, and flow may themselves influence cellular behaviour. Appropriate interpretation requires separating those physical effects from the biological variable under investigation.

Single-cell analytical workflows use microchannels, droplets, or microwells to isolate cells and associate their contents with identifying molecular barcodes. In single-cell RNA sequencing, for example, cell-derived RNA is converted into sequenceable material while barcodes preserve information about its compartment of origin. Microfluidic compartmentalization helps scale these measurements, but the resulting gene-expression profiles also depend on cell capture, lysis, reverse transcription, amplification, and sequencing. Two cells entering one compartment can produce a mixed profile, while ambient RNA can contaminate otherwise correctly isolated cells.

Organ-on-a-chip systems combine living cells with engineered physical environments intended to reproduce selected tissue functions. Devices may include flowing vascular channels, porous membranes, extracellular matrices, air–liquid interfaces, and mechanically deformable structures. A 2010 lung-on-a-chip study by Dongeun Huh and colleagues used a flexible membrane separating epithelial and endothelial compartments and applied cyclic mechanical strain. The device illustrated how controlled mechanics and flow could be incorporated into a tissue model. Such systems reproduce selected features of an organ, not an entire organ’s anatomy, development, or systemic physiology.

Microfluidic culture also supports studies of bacteria, microbial communities, and host–microbe interactions. Chambers can maintain defined nutrient conditions, expose populations to controlled concentration changes, or retain individual cells for observation over many generations. In mammalian culture, perfusion can supply oxygen and nutrients and remove waste, while hydrogels provide three-dimensional support. Flow conditions matter: wall shear stress can affect cell attachment, morphology, and function. Material absorption, oxygen permeability, and leachable compounds can likewise alter the chemical environment experienced by cells.

Diagnostic cartridges apply related techniques to specimens such as blood, saliva, or swabs. A cartridge may meter a sample, remove interfering material, extract nucleic acids, perform amplification, and read an optical or electrical signal. Reliable operation with real specimens is more difficult than demonstrations using purified targets. Clotting, viscosity variation, bubbles, and sample-to-sample differences can disrupt flow or chemistry. Biomedical use consequently requires evidence about the complete assay and its intended context; successful miniaturization or agreement in a small experimental dataset does not by itself establish clinical performance or interchangeability with an existing test.

Chemical synthesis and materials production

Microfluidic reactors provide controlled environments for chemical transformations. Their high surface-area-to-volume ratio can improve heat exchange, while small reaction volumes reduce the amount of hazardous material present at any instant. Rapid mixing can narrow the distribution of local reagent concentrations, and continuous operation can maintain defined feed ratios. These features are valuable when product properties depend strongly on the early stages of reaction, including nucleation, precipitation, and competitive reaction pathways. Their benefits nevertheless depend on whether mixing and heat transfer are actually fast relative to the chemistry.

In a continuous reactor of internal volume VV carrying volumetric flow QQ, the nominal residence time is

τ=VQ.\tau=\frac{V}{Q}.

Actual molecules experience a distribution of residence times. Pressure-driven velocity profiles, diffusion, recirculation, and stagnant regions can broaden that distribution, changing conversion or selectivity. Segmented flow can improve internal circulation and reduce axial dispersion in some systems. Reaction design therefore links fluid mechanics to kinetics and mass transfer, rather than treating a channel as a miniature stirred flask.

Microfluidic mixing is used to prepare lipid nanoparticles by bringing an organic lipid solution into contact with an aqueous phase under controlled conditions. Rapid solvent exchange promotes assembly, and operating variables such as composition, total flow, and flow-rate ratio influence the resulting size distribution and encapsulation characteristics. The same general principle supports production of polymeric nanoparticles and other colloidal materials. Formulation composition remains essential: identical channel geometry does not produce identical particles from different chemical systems.

Droplet methods can produce polymer beads, capsules, and gel particles with controlled dimensions. A droplet may act as a template that is solidified by polymerization, crosslinking, cooling, or solvent removal. Double emulsions provide nested compartments useful for encapsulating one material within another. Channel wetting and surfactant chemistry become especially important when alternating between aqueous and organic phases. Photochemical reactors offer another application: short optical paths can improve the distribution of light through absorbing reaction mixtures, although window fouling and the relation between photon supply and flow still constrain performance.

Industrial production often uses numbering-up, in which multiple channels or reactor modules operate in parallel. Simply increasing every channel dimension can eliminate the mixing and heat-transfer characteristics that made the original device effective. Parallelization instead introduces flow-distribution problems: small differences in resistance can cause unequal throughput, and a blockage in one branch can alter the rest of the network. As a branch of chemical engineering, microfluidic manufacturing must therefore address feed preparation, solvent compatibility, separation, cleaning, monitoring, and product consistency alongside the reactor geometry. A high-performing laboratory chip is only one component of a production process.

Reliability, manufacturing, and system-level limits

The transition from an experimental chip to a reliable product is often dominated by packaging and interfaces. Ports must connect microscopic channels to reservoirs and tubing without leaks, excessive dead volume, or trapped air. Bonded layers must tolerate operating pressure and storage conditions. Disposable cartridges must align reproducibly with reusable optical, thermal, pneumatic, or electrical hardware. A highly integrated chip can still require a large surrounding instrument, so claims of portability depend on the complete configuration rather than the size of the fluidic substrate alone.

Bubbles are a persistent failure mechanism. They can block channels, change hydraulic resistance, interrupt electrical measurements, or expose cultured cells to moving interfaces. Gas can enter through connections, emerge from dissolved gas as conditions change, or permeate through device materials. Bubble traps and gas-permeable structures can help, but consume space and may retain sample. Clogging presents a related problem, particularly with blood, cell suspensions, precipitating reactions, or particulate environmental samples. A filter protects downstream channels only by creating another location where material may accumulate.

Manufacturing quality involves more than reproducing the visible channel pattern. Channel depth, surface roughness, bonding deformation, reagent deposition, and surface treatment can all affect performance. Stored biological reagents may require drying or other stabilization, and their behaviour after rehydration can differ from that of freshly prepared solutions. Shelf life also depends on seals, humidity exposure, gas permeability, and changes in surface chemistry. A manufacturing process must control these factors across production lots rather than relying on adjustment of individual devices by skilled researchers.

System automation adds software and control requirements. Pumps, valves, heaters, and detectors must operate in the correct sequence, and fault conditions must be detected before they invalidate a result. Image analysis and machine-learning methods can classify droplets or cells, but their outputs depend on the populations, instruments, and operating conditions represented in their development data. Changes in illumination, channel geometry, sample composition, or reagent lot can produce measurement shifts. Traceable controls and explicit acceptance criteria remain necessary even when operation is highly automated.

The appropriate degree of integration depends on use. A flexible research platform may prioritize optical access and reconfiguration; a diagnostic cartridge may prioritize containment, ease of use, and reproducibility; a production reactor may prioritize uptime and cleaning. Economic comparisons must include tooling, assembly, external hardware, reagent storage, operator effort, and waste. Although fluid volumes are small, disposable plastics, oils, solvents, and contaminated components can remain significant. These practical constraints explain why different microfluidic architectures coexist and why designs optimized for one application are not automatically suitable for another.