PDMS & soft lithography

How casting a silicone rubber against a micro-patterned master turned microfluidic prototyping from a cleanroom project into a benchtop afternoon — and what that convenience costs.

If you only remember three things

PDMS is a transparent, rubbery silicone. You make a chip by pouring it over a micro-patterned master, curing it, peeling it off, and sealing it to a flat surface — a process called soft lithography.
It won academia because it is fast, cheap and forgiving: once a master exists, a new chip takes hours on a benchtop rather than a full cleanroom run. Its transparency, elasticity and gas permeability also enable imaging, valves and cell culture.
The expensive, cleanroom part — the master — is made once by photolithography (typically an SU-8 photoresist pattern on a silicon wafer) and then reused to cast many PDMS replicas.
The very properties that make PDMS convenient are also its limits: it absorbs small hydrophobic molecules, swells in many solvents, recovers hydrophobicity after surface treatment, deforms under pressure, and is hard to mass-manufacture — a superb prototyping material but often a poor production one.

What you’ll learn

What PDMS is, why it became the workhorse of academic microfluidics, and its key material properties
What soft lithography is, and how a silicon/SU-8 master becomes a finished PDMS chip
The full workflow — from CAD design through photolithography, casting, bonding and testing — and why each parameter matters
The real advantages of PDMS, and the limitations that constrain where it should be used
When PDMS is the right choice, when it is not, and how it compares with glass, silicon and thermoplastics

Intuition

Soft lithography is moulding — the same idea as a jelly mould or a rubber stamp, shrunk to the micron scale. You make one detailed mould (the master), then cast a soft material against it as many times as you like; every replica carries a faithful negative of the pattern.

The trick is a division of labour: the hard, precise, expensive work happens once, in a cleanroom, to make the master. After that, copying the pattern into PDMS is cheap, quick and needs no cleanroom — which is why a student can iterate a design in an afternoon.

PDMS itself is best pictured as a clear, springy silicone rubber — transparent like glass, flexible like an eraser, and slightly permeable to gases like a sponge is to water. Those three traits (clear, springy, breathable) explain most of both its powers and its problems.

The concept

PDMS (polydimethylsiloxane) is a silicone elastomer — a polymer with a flexible –Si–O–Si– backbone and methyl (–CH₃) side groups. It is supplied as two liquids, a base and a curing (cross-linking) agent; when they are mixed and heated, the curing agent links the chains into a soft, transparent, rubbery solid. Because it begins as a pourable liquid and sets into an elastic solid, it can be cast against a mould and then peeled away intact.

Before PDMS, microchannels were etched into glass or silicon — accurate but slow, costly and cleanroom-bound. Casting PDMS against a reusable master (Duffy et al., 1998; Whitesides, 2006) cut the turnaround from weeks to hours and moved most steps out of the cleanroom, so ordinary labs could design, build and test a chip in a day. That accessibility, more than any single property, is why PDMS came to dominate academic microfluidics (McDonald & Whitesides, 2002).

The properties that matter most are its advantages: optical transparency across the visible and into the near-UV, good for microscopy and optical detection; a very low elastic modulus (on the order of a few MPa — roughly 1–3 MPa for common formulations — thousands of times softer than glass), so it is elastic enough to make flexible membranes, valves and pumps; easy moulding, since the liquid flows into micron- and sub-micron features and cures at modest temperatures; high gas permeability, so oxygen and CO₂ diffuse through it, useful for oxygenating cells on-chip; low cost and simple processing, giving cheap and rapid prototyping; and high replication fidelity, faithfully reproducing fine features from the master (Xia & Whitesides, 1998; McDonald & Whitesides, 2002).

Soft lithography is a family of techniques that use a soft elastomer — usually PDMS — cast or stamped against a patterned master to transfer micro- and nanoscale features, rather than etching each device directly (Xia & Whitesides, 1998). In microfluidics its dominant form is replica moulding: cast PDMS on the master, cure, peel, and bond to a substrate to close the channels. The master is patterned once by conventional photolithography; the elastomer does the copying.

The workflow therefore splits into two halves: a hard, one-time master fabrication (photolithography, in a cleanroom) and a soft, repeatable replication (casting PDMS, on the bench). Understanding the process means understanding both halves — and why the parameters in each shape the final chip.

Why it matters

The soft-lithography route is a large part of why microfluidics grew into a broad academic field: it let thousands of labs prototype devices cheaply and quickly. Knowing how it works — and where PDMS's convenient properties become liabilities — lets you judge whether a result will translate beyond a research chip, and when to reach for a different material.

How it works

Design (CAD). The channel network is drawn as a 2D layout in CAD. Feature sizes, spacing and the resolution of the eventual mask set the smallest channels you can make, and the design fixes where inlets, outlets and functional elements sit. This step is essentially free to iterate — much of PDMS's appeal.
Photomask. The layout is printed as a photomask: a transparent film or glass plate with opaque regions defining the pattern. High-resolution transparency masks are inexpensive and adequate down to roughly the tens-of-microns range; chrome-on-glass masks are used for finer features. Mask quality directly limits edge sharpness and the minimum feature size.
Photolithography. In a cleanroom, a light-sensitive polymer (photoresist) is spin-coated onto a silicon wafer to a controlled thickness, then exposed to UV light through the mask. Exposure changes the resist's solubility where light strikes it; developing then washes away either the exposed or unexposed regions, leaving a patterned relief. The spin speed sets the resist thickness — and therefore the eventual channel height — so it is a key parameter.
SU-8 and the master. For microfluidic masters the resist is usually SU-8, a negative epoxy photoresist prized for tall, straight-walled, high-aspect-ratio features (del Campo & Greiner, 2007). After exposure and development, the SU-8 relief standing proud of the wafer is the master — a positive relief of the channels. Because the master is reused for many castings, its accuracy is worth the cleanroom cost; feature height equals the SU-8 thickness, and sidewall quality sets channel-wall quality.
Mixing PDMS. The base and curing agent are mixed — a widely used example is a 10:1 base-to-curing-agent mass ratio for Sylgard 184 (Duffy et al., 1998; McDonald & Whitesides, 2002); this is an example process condition, not a universal rule. The ratio matters because it sets the cross-link density: more curing agent gives a stiffer, less deformable chip, less gives a softer one. Mixing entrains air, so the blend is degassed under vacuum to remove bubbles that would otherwise become defects.
Casting (replica moulding). The liquid PDMS is poured over the master in a holder and, being a low-viscosity liquid, flows into and fills every feature — the origin of its high replication fidelity. The pour depth sets the slab thickness.
Curing. Heating cross-links the PDMS into a solid; higher temperature cures faster (room temperature over many hours, or roughly an hour in a warm oven, are commonly reported ranges — treat any specific schedule as an example, since it depends on the formulation and the target stiffness). Under-curing leaves sticky, leachable uncross-linked material; over-curing can embrittle the surface — so the cure schedule is a genuine quality parameter, not a formality.
Demolding. The cured PDMS is peeled off the master. Because it is elastic and does not adhere strongly to silicon/SU-8, it releases cleanly and the master survives for reuse, so one master yields many chips. Aggressive peeling, or fragile tall features on the master, can tear either part, so geometry and release behaviour matter.
Port creation (punching). Access holes for tubing are punched through the PDMS at the inlets and outlets, usually with a coring punch, connecting the channels to the outside world. The hole diameter must match the tubing or connector for a leak-free fit, and ragged punches shed debris that can clog channels.
Surface treatment. The PDMS surface is natively hydrophobic. Exposure to oxygen plasma (or UV-ozone) oxidises the surface methyl groups into silanol (–OH) groups, making it temporarily hydrophilic and, crucially, chemically activating it for bonding. Plasma power, time and pressure are instrument-specific tunables — over-treatment can crack the surface, under-treatment gives weak bonds — so they are optimised per system rather than copied as fixed numbers.
Bonding. Pressing a freshly plasma-activated PDMS surface against another activated PDMS or glass surface lets the silanol groups condense into covalent –Si–O–Si– bonds, sealing the channels with a strong, irreversible bond (Duffy et al., 1998). Any layer alignment happens here; the surfaces must be clean and dust-free, since a single particle prevents contact and causes leaks, and bonding must be done promptly, before the surface reverts.
Assembly and testing. Tubing is inserted into the ports and the device is tested — typically by flowing liquid through and checking for leaks, correct filling and bond integrity at the intended pressure. Only a chip that seals and fills as designed is ready for use; failures here usually trace back to dust, weak bonding, or an over- or under-cured slab.

Microfluidic example

Imaging and detection (transparency). Because PDMS is optically clear, a bonded PDMS-on-glass chip sits straight on a microscope stage, so flow, cells and fluorescence are observed directly through the device — a major reason PDMS suits biology.

Valves and pumps (elasticity). PDMS's softness is exploited in the classic multilayer 'Quake' microvalve: pressurising a channel in one layer deflects a thin PDMS membrane to pinch off a channel below it. Chaining such valves builds on-chip pumps and large-scale integrated fluidic circuits — something a rigid material cannot easily do.

Cell culture (gas permeability). Oxygen and CO₂ diffuse through PDMS, so cells in a sealed channel can still breathe, which underpins many organ-on-chip and long-term culture devices — though the same permeability lets water vapour escape and concentrate the medium over time.

Rapid iteration (prototyping + fidelity). Because one master casts many faithful replicas cheaply, a lab can build several design variants in parallel and refine quickly — the practical engine behind microfluidics' rapid growth.

Practical design implications

Small-molecule absorption: PDMS soaks up hydrophobic small molecules (many drugs and dyes), depleting them from solution and skewing concentrations — a serious problem for quantitative assays and drug studies (Toepke & Beebe, 2006). Account for it, coat the surface, or choose another material.
Hydrophobic recovery: the hydrophilicity from plasma treatment is temporary — low-molecular-weight chains migrate to the surface and it reverts to hydrophobic over hours to days (McDonald & Whitesides, 2002). Bond and fill promptly, and do not rely on lasting wettability without a stabilising treatment.
Solvent compatibility: PDMS swells in, and is degraded by, many organic solvents (e.g. toluene, hexane, chloroform), distorting channels and leaching oligomers (Lee, Park & Whitesides, 2003). It suits aqueous work best; check solvent compatibility before use.
Gas permeability (double-edged): excellent for oxygenating cells, but it also lets bubbles form and pass and lets water vapour escape, so aqueous samples evaporate and concentrate over long runs. Plan humidification or sealing for long experiments.
Mechanical deformation: being soft, PDMS channels bulge under pressure, changing their cross-section and hence hydraulic resistance and flow rate. At higher pressures this is significant — factor it into flow-control designs, or use a stiffer formulation.
Scalability: soft lithography is a manual, low-throughput process that does not translate to high-volume manufacturing, where thermoplastic methods such as injection moulding and hot embossing are used instead (Becker & Gärtner, 2008). A PDMS prototype rarely becomes a product unchanged.
Batch variability: hand mixing, degassing, curing and plasma treatment introduce device-to-device variation in stiffness, surface chemistry and dimensions — a reproducibility concern to control for, especially across labs.

Common mistakes

Watch out for:

Assuming a plasma-treated surface stays hydrophilic. It reverts (hydrophobic recovery), so any wettability-dependent step must happen before it does.
Using PDMS for quantitative small-molecule work without checking absorption. Concentrations can drift as the molecule partitions into the walls, invalidating dose–response data.
Running incompatible solvents. Non-aqueous solvents can swell or dissolve PDMS, deforming channels and contaminating samples.
Ignoring channel deformation under pressure. Treating soft PDMS channels as rigid overestimates flow rates and misreads resistance.
Skipping degassing, or working with dust. Trapped bubbles and single particles are the most common causes of failed casts and leaky bonds.
Treating a PDMS prototype as a manufacturable product. What works as a one-off cast usually needs a different material and process to scale.

Researcher notes(advanced)

When PDMS is a good choice: rapid prototyping and design iteration; devices needing optical access, elastic membranes and valves, or gas-permeable walls (cell culture, organ-on-chip); and low-volume, aqueous, academic work where turnaround and cost matter more than manufacturability.
When PDMS is a poor choice: quantitative small-molecule or drug assays (absorption); work with organic solvents (swelling); applications needing rigid, dimensionally stable channels or precise pressure–flow relationships; and anything destined for mass production or regulatory-grade reproducibility (Berthier et al., 2012).
Versus glass: glass is rigid, chemically inert, non-absorbing, solvent-resistant and optically excellent, but its fabrication (etching, bonding) is slow, expensive and cleanroom-intensive, and it cannot form elastic valves. Choose it when chemical robustness and dimensional stability outweigh turnaround.
Versus silicon: silicon enables the highest-resolution features and integrates with electronics and standard IC processing, and is rigid and chemically robust, but it is opaque to visible light (no simple transmitted-light imaging), electrically conductive and expensive — best where precision and integration outweigh optical access and cost.
Versus thermoplastics (PMMA, COC, polystyrene, PC): rigid, low-absorbing materials that, unlike PDMS, scale to mass manufacturing via injection moulding and hot embossing (Becker & Gärtner, 2008), and polystyrene matches the validated surface of conventional cultureware (Berthier et al., 2012). The trade-off is less nimble prototyping and the need for moulding tooling, so they suit production more than early exploration.
The trade-off in one line: PDMS optimises for speed, cost and its unique elastomer and gas-transport properties, at the expense of chemical inertness, dimensional stiffness, quantitative fidelity and scalability — so the best material depends on whether you are exploring or deploying (Berthier et al., 2012).

Related concepts