A liquid class feels like a property of the liquid, and mostly we talk about it that way: this is how you move DMSO, this is how you move a protein solution. But a class is validated against a whole context, and one of the quietest parts of that context is the shape of the vessel you are aspirating from and dispensing into. Change the well from a flat bottom to a cone and you have not touched a single parameter, yet the same class can start missing, because the geometry has moved the surface, the depth, and the residual volume out from under it.
The point of this piece is that well geometry is not labware trivia to be sorted out once during deck setup. It is physics that reaches directly into how a transfer behaves, and a class that ignores it is a class that works until the plate changes.
The shapes and what they are for
The common well bottoms are not arbitrary. Each shape solves a problem, and each imposes a cost on the transfer in exchange.
- Flat or F-bottom: a flat floor that gives a clean optical path for reading absorbance, which is why it dominates assay plates, at the cost of spreading the last of the liquid into a thin, hard-to-reach layer.
- Round or U-bottom: a curved floor that pools liquid gently toward the center, common for cell work and mixing, and kinder to a tip that has to reach the bottom.
- V-bottom: a shallow cone that funnels the sample to a point, prized for recovering as much of a small volume as possible.
- Conical PCR: a steep, deep cone that concentrates a tiny volume at the tip, standard for thermal cycling where every microliter counts.
The moment you list them this way, the through-line is obvious. The flatter the floor, the better you can read it and the worse you can drain it. The steeper the cone, the better you recover the sample and the more precisely your tip has to be placed to do it. The shape is a trade, and the class has to honor whichever side of the trade the plate landed on.
The meniscus is not a flat line
It is tempting to picture the liquid surface as a flat disc sitting at some height in the well. It is not. Liquid climbs the wall it wets, so the real surface is a curved meniscus, higher at the edges and lower in the middle, or the reverse for a liquid that beads. In a wide flat well the curvature is a small correction. In a narrow well it is a large fraction of the surface, and the notion of a single surface height starts to blur.
This matters because so much of a transfer is referenced to that surface. Where you start an aspiration, where you position a dispense, where a level sensor decides it has found liquid: all of these assume a surface, and the surface is curved and moves with the liquid property and the well width. A class tuned where the meniscus was a rounding error can be caught out where it is not.
Depth per microliter changes with shape
Here is the effect that surprises people most. In a straight-walled well, adding or removing a microliter changes the liquid height by the same small amount whether the well is full or nearly empty, because the cross-section is constant. In a V or conical well it is nothing like constant. Near the top, where the cone is wide, a microliter barely moves the level. Near the bottom, where the cone pinches to a point, that same microliter moves the level a great deal, because the cross-section has shrunk toward zero.
So the depth-to-volume relationship is not a fixed slope you can set once. It is a curve that steepens fast as you approach the bottom of a cone, and any tip position that assumes a linear relationship will be wrong exactly where wrong is most dangerous, down near the floor where there is no margin left.
Immersion depth has to track the geometry
A tip aspirating should sit just below the surface: deep enough to stay submerged as the level falls, shallow enough not to scrape the bottom or bury itself in liquid it then drags out on its walls. In a flat well that is a forgiving target with a wide band of acceptable depths. In a V or conical well the band narrows dramatically, because the surface drops fast per microliter near the bottom and the bottom itself is a point the tip can jam into.
This is why level following, the tip tracking the surface downward as it draws, goes from a refinement to a requirement as the wells get steeper and smaller. A fixed immersion depth that works on a full flat well will either break the surface early or crash into the cone on a V-bottom, and the same touch-off or dispense-to-surface position that was comfortable on one plate can be impossible on another. The class has to place the tip against the geometry, not against a habit formed on a different plate.
Dead volume and the sample you cannot reach
Every well has a residual you cannot recover, the dead volume that sits below where the tip can safely draw. Shape governs it almost entirely. A flat well spreads its last liquid into a broad shallow film with a large unreachable residual. A V or conical well funnels its last liquid into a point the tip can approach closely, so the dead volume is far lower. When the sample is precious, a patient aliquot, a purified protein, a scarce compound, that difference decides how much of it you actually get to use, and it is a reason cones exist at all.
Small deep wells and the rise of capillarity
Push toward 384 and 1536 formats and the wells become small and deep relative to their width, and effects that were negligible in a 96-well grow teeth. Surface tension and capillary action, which barely register in a wide well, become a meaningful part of how the liquid sits and moves in a narrow one. The meniscus curvature is now a large share of the geometry, and the tolerances on tip position shrink with the well. A class that was comfortable in a roomy flat 96-well is operating in a different physical regime by the time it reaches a small deep 384-well, even before you change the liquid.
Level detection depends on all of it
Every one of these effects converges on the same operation: finding the liquid. Level detection has to locate a surface that is curved by the meniscus, that moves nonlinearly with volume in a cone, and that offers less and less clean area to sense as the wells shrink. The detection method that was reliable in a wide flat well can start throwing false or late triggers in a narrow deep one, not because the sensor changed but because the surface it is hunting for did.
Put the pieces together and the failure is easy to picture. A class validated in a flat 96-well moves to a V-bottom 384-well, and three things shift at once: the immersion depth that kept the tip submerged is now wrong because depth per microliter changed, the dead volume assumptions no longer hold, and the meniscus is a larger part of a smaller surface. No parameter was edited. The plate did the editing.
A liquid class is not validated against a liquid alone. It is validated against a liquid in a shape, and when the shape changes the class has quietly changed with it.