Biosensor Microfluidic Channel Mold: A Machining Guide

A biosensor microfluidic channel mold is a metal master that carries the inverse of the fluid network you want on the finished chip. Cast PDMS against it, or hot emboss COC and PMMA over it, and the polymer picks up the channel pattern. The mold is not the product. It is the tool that decides whether every later copy has the same channel width, the same floor roughness and the same edge geometry.

Typical biosensor layouts use channels from 20 µm to 500 µm wide and 10 µm to 200 µm deep. Two features matter more than the rest. The first is the cross-section. A rectangular channel with sharp top corners gives steady flow and predictable fill volume; a rounded one shifts the flow profile and changes how reagents meet. The second is the floor finish, because light scattering and non-specific protein binding both rise with roughness.

We machine these masters from aluminium, stainless steel, tool steel, brass or copper on 3-axis, 4-axis and 5-axis centers. The choice of base material follows the replication process, not the other way round: hot embossing at 150 °C or above pushes you toward steel, while low-pressure PDMS casting is often fine on aluminium.

One point worth stating early. A mold is a wear part. It sees thermal cycles, demolding force and cleaning chemicals, so the surface that looks perfect on delivery can degrade after a few hundred shots. Design the master with that in mind, and you avoid a mid-project surprise.

Our general machining tolerance is ±0.005 mm, and we hold Ra 0.8–1.6 μm as a normal machined finish with Ra 0.2–0.8 μm available where the toolpath and geometry allow. Those numbers do not translate automatically to every micro channel. What matters is the ratio between channel width, depth and the length of the cut.

A 200 µm wide channel, 50 µm deep and 30 mm long is a comfortable job. A 30 µm wide channel at the same length is not, because the tool that can enter the slot is too slender to survive the cut without deflection. Deflection shows up as a tapered wall, a bowed floor or a broken cutter halfway through. When a design sits at that edge, we say so in the DFM review rather than quoting a number we cannot repeat across a production run.

Aspect ratio is the other constraint. Below roughly 2:1 depth-to-width, milling is stable. Above 3:1, chip evacuation becomes the problem. Swarf packs into the slot, rubs the wall and leaves a finish that no polishing step can fix without rounding the corners.

Sharp internal corners have a hard limit too. The corner radius can never be smaller than the tool radius, so a 100 µm wide channel cannot have a true 90° corner. If your design needs one, plan for a radius or move to a process that can produce it.

• 1
  Comfortable rangeChannel width 100–500 µm, depth under 100 µm, aspect ratio below 2:1.
• 2
  Achievable with careWidth 50–100 µm, shallow depths, short flow length, generous corner radii.
• 3
  Rework the designWidth under 30 µm, aspect ratio above 3:1, or long narrow slots with sharp corners.

Micro channel work lives or dies on the cutter. We use small-diameter end mills and micro ball cutters in carbide, and on hardened steels we run graphite electrodes for micro EDM instead. The cutter must be short enough to stay stiff yet long enough to reach the floor of the pocket, so we often step the job across several tools with decreasing diameter rather than trying to finish with one.

Roughing removes most of the volume with a larger tool, leaving 20–40 µm of stock on the floor and walls. Finishing then takes light radial cuts at higher spindle speed. Climb milling on the finishing pass keeps the cutting force pressing the workpiece into the tool rather than lifting it, which reduces edge chipping on brittle materials and improves wall straightness.

Burrs are the classic failure mode. On a 50 µm channel, a 5 µm burr at the top edge is not a cosmetic issue; it changes the sealing surface and can block the channel after bonding. We control it with sharp tooling, a finishing pass that exits the material cleanly, and where geometry permits, a light deburring that does not touch the channel floor.

Deep slots trap chips. We program pecking cycles with air blast or through-coolant, and we inspect between operations. Once a chip is welded to a channel wall by friction heat, cleaning it out usually costs more than re-cutting the feature.

Optical detection and fluorescence readout are sensitive to the channel floor. A rougher floor scatters more light and gives a noisier baseline, so biosensor masters usually call for a finer finish than a general machined part. We can reach Ra 0.2–0.8 μm on flat floors and accessible walls, but the method matters.

Mechanical polishing with abrasive media rounds edges. On a 100 µm channel, a polishing pass that removes 3 µm from the floor can take 5 µm off the corner radius. That single change alters the cross-section and can change fill behavior. When sharp corners are essential, we keep the polished area away from the edge and rely on the finishing toolpath for the rest.

Electropolishing and electroless nickel plating are alternatives. Electropolishing smooths stainless without mechanical contact, which preserves geometry better, but it removes material uniformly and will not fix a wavy floor. Nickel plating builds a hard, corrosion-resistant layer that extends mold life, at the cost of a thin added layer that must be accounted for in the dimensions.

Tell us the measurement method you will use to accept the mold. If your incoming inspection is a profilometer scan across the channel, we will report the same way. If it is a microscope image, say so, because edge definition reads differently under each method.

We inspect 100% of parts before shipment, with raw material checks, in-process monitoring and a final dimensional inspection. For microfluidic masters the final report typically includes channel width and depth at several positions along the flow path, floor roughness, and a note on corner radii. Reports are available on request.

The useful check is not one measurement at the center of the mold. It is a set of measurements at the inlet, the middle and the outlet of the longest channel. That is where thermal drift during a long finishing pass shows up. If the width drifts more than your allowance across the length, the mold will produce chips with a gradient in fill volume, and that gradient is hard to find later.

GreatLight works to ISO 9001:2015, IATF 16949:2016, ISO 13485:2016 and ISO 27001:2022. The medical device standard matters here because biosensor tooling usually sits inside a regulated development program. Drawings and process data stay confidential, and we sign an NDA on request before files move.

Quotation and a free DFM analysis come back within 12 hours. Production can start within 24 hours of approval, and parts ship in 3–5 days. From one prototype mold to a 10,000+ part run, there is no minimum order quantity.