Channels, not wires

You are trying to make a tiny ditch that a drop of liquid can travel, the way a hose carries water. The drawing in an editor looks like electronics: a stripe, a couple of ports, a slab. Electrons do not live here. Liquid does. That ditch is a microfluidic channel. A channel drawn at 50 µm that prints at 35 µm after bonding crush, at an assumed height of 20 µm, raises the resistance to flow by about 1.66×. A height drop from 20 to 15.5 µm is what doubles it. The picture at the top is the object: a clear chip, blue in one network, red in the other, wells where a pipette lands. Width is hydraulic. A sharp corner is a place a bubble can sit.

What the liquid actually does

If you treat the stripe as a metal wire, you will add square jogs to dodge a label, neck the width to squeeze past a hole, and leave a chamber with no vent. Liquid then stalls. A bubble parks in the corner and the assay reads empty. A sudden widening holds reagents you wanted one after another; the flow is laminar, so they meet only by diffusion. A lid that misses a port seals a well, and the pipette finds no opening. The drawing can still satisfy a minimum-width rule copied from a chip manual. The failure is a clog, a leak, or a dead volume, not a short.

The second picture is that failure in a single corner. Blue liquid meets a square turn and an air bubble sits in the pocket. The corner was easy to draw. It is a trap. A fillet, or a bend with a radius, gives the bubble somewhere to leave. Dead volume — liquid that sits and never moves on — is the parasitic of this craft. Every sudden chamber adds some.

The drawing is still a layout

A channel is often a path: a centerline plus a width. The factory still needs a closed shape on a layer that means “this mold,” “this etch,” or “this hole through the lid.” A port that looks like a via on a board is here a hole through a cover, not a metal plug. The hole has to sit inside the chamber with margin. That margin is enclosure for a biologist.

Bonding, baking, and plasma are the recipe, the same way a MEMS release is a recipe. The pump you sized on the drawn width may no longer push once that crush has happened. The photograph of red and blue in glass is the recipe succeeding. The photograph of the bubble is the drawing forgetting that liquid has a surface.

Do not send a chip metal stack and call it a fluidic chip because the colors were pretty. Send the channel layer, the port layer, the outline of the slab, and the sentence that says what each one is made of.

Bond one chip, fill a syringe with a volume of dyed water you have written down, and push that volume into a single inlet. Collect what leaves the outlet in a second marked tube. If the outlet returns less than you pushed, the missing liquid is sitting in the chip, in a chamber that never moves on. That shortfall is dead volume you can hold up to the light: the second tube is short, and a clear pocket or a parked bubble is still inside the slab.

Find that pocket under a loupe while the dye is still in the chip. It will sit where a wall turns square, the same trap as the bubble in the second photograph, or in a sudden widening you drew to dodge a hole. Mark the spot on a print of the channel layer. Replace the square with a radius, or delete the chamber, and cut the next mold from that file. A second bonded chip should return the volume you pushed. If it does not, you marked the wrong pocket, and the dye will show you the next one.

Do this before you trust a flow number calculated from the width on the screen. The clear chip in the first photograph is already past bonding. The syringe is how you learn whether that chip still contains the ditch you drew, or a ditch the crush and the corners have quietly narrowed and stalled. Start a clock when the plunger moves and stop it when the outlet tube reaches the volume you pushed. Write that time on the print of the channel layer, next to the drawn width. A short ditch that takes several times longer than a neighbor of the same drawn width has already been crushed, and the time belongs in the note that travels with the mold. Run the same clock on a straight neighbor so the comparison is a time, not a hunch about how slow the dye looked.

A 50 µm ditch, then the crushed one

Draw the channel at 50 µm and the neighbor at 35 µm, both straight, both with the same inlet. Bond one chip and push the dyed volume you wrote down into the 50 µm inlet first. Time it. Then push the same volume into the 35 µm inlet and time that. The narrower ditch should take longer. If the 50 µm ditch takes as long as the 35 µm one, or longer, the lid crushed the wider channel and the width on the screen is not the width the liquid sees.

Find the stall under the loupe while the dye is still in. A square corner holds a bubble the way the second photograph does. Mark that corner on the print of the channel layer, replace it with a radius, and cut the next mold from that file. The clear chip in the first photograph is already past bonding. The two times, written next to 50 µm and 35 µm, tell you whether you are measuring the ditch you drew.