Most of what you know about tuning a liquid class was learned in the range where air-displacement pipetting behaves itself, somewhere between a couple of microliters and a milliliter. In that band the mental model is stable: you set an aspiration speed, a dispense speed, an air gap, a blowout, and a settling delay, and you tune those numbers against a balance until the volume lands. The trouble starts when the target volume keeps dropping. Somewhere around a microliter, and unmistakably by a few hundred nanoliters, the parameters you have relied on stop being the ones that matter. The physics has changed underneath you, and the class you carry over from the higher range quietly falls apart.
This is not a best-practices article about squeezing a bit more accuracy out of a contact tip at low volume. That is a real and separate problem. This is about the regime below it, where the honest answer is often to stop touching the well at all. Understanding why classic classes run out of room is the first step to knowing when to reach for a non-contact head instead.
Why the classic class runs out of room
The core issue is that at sub-microliter scale the errors that were once rounding noise become a large fraction of the target. When you aspirate two hundred microliters, the thin film of liquid that wets the outside and inside of the tip is negligible against the volume you moved. When you aspirate two hundred nanoliters, that same wetting film and the dead volume in the tip are suddenly comparable to the dose itself. Surface tension, which barely registered before, now dominates how the liquid forms, hangs, and releases at the tip orifice. A droplet that will not let go, or lets go a beat too late, is no longer a small percentage error. It is the difference between delivering your target and delivering half of it.
You can push a positive-displacement mechanism further down than air displacement, because it removes the compressible air cushion that smears out small movements, and that buys you real range. But it does not repeal surface tension or the dead-volume problem. Eventually the volume you are trying to place is small enough that the act of touching it to a surface, or hanging it off a tip, introduces more variability than the mechanism can control. That is the wall. It is not a tuning failure you can parameter your way out of.
What non-contact dispensing changes
Non-contact dispensing sidesteps the wall by firing the liquid rather than placing it. Instead of forming a droplet and touching it off against a well wall or a pool, the head ejects droplets across an air gap. Two broad families cover most of the field, and they solve the problem in different ways.
- Pressure and solenoid jetting: a pressurized fluid line and a fast valve, or a piezo actuator, fire discrete droplets from a nozzle so the liquid leaves the head under its own momentum and never depends on a tip touching the target.
- Acoustic droplet ejection: a focused acoustic pulse from below the source well launches a small droplet up off the free surface of the liquid, transferring it to a plate held above with no nozzle, no tip, and no physical contact with the fluid at all.
The payoffs are the reason these methods exist. Because nothing touches the liquid in the acoustic case, there is no tip-contact carryover and no per-transfer tip to consume, which matters enormously when you are moving one compound into hundreds of destination wells. Jetting heads are fast, firing many droplets per second, and both approaches make aggressively miniaturized assays practical, so a screen that once burned through milliliters of precious reagent can run on a fraction of that. When the assay volume shrinks, the reagent bill shrinks with it.
The parameter set is genuinely different
It helps to keep the word class in your head, because the concept survives. You still have a named, reusable description of how a particular liquid should be handled, and you still tune it and version it. What changes is that almost none of the knobs are the ones you knew. Aspiration speed and dispense speed, the twin dials of the contact world, largely disappear. In their place you tune the properties of firing.
- Droplet volume: the quantum of transfer, since a non-contact head builds a dose out of discrete droplets rather than a single metered aspiration, so the achievable resolution and the minimum transfer are set here.
- Firing frequency and count: how fast and how many droplets are ejected to build the target volume, which trades throughput against the risk of satellite droplets and mid-air coalescence.
- Fluid type by acoustic property: for acoustic ejection in particular, the class is keyed to how the liquid carries sound, so an aqueous buffer, a high-glycerol solution, and a DMSO stock are three distinct calibrations rather than one class with a tweak.
That last point is the one people underestimate. An acoustic instrument has to know the composition of what it is firing, because the acoustic coupling depends on it, and a droplet calibrated for aqueous buffer will misfire from a glycerol-heavy well. The class is no longer a loose recommendation. It is a declaration of what the fluid actually is.
Calibrate by what lands, not by the balance
Gravimetry is the honest arbiter in the microliter world, but at the nanoliter scale it becomes very hard to do well. Weighing a two-hundred-nanoliter dispense against evaporation, air currents, and balance drift is a fight you will usually lose. So the calibration philosophy shifts from weighing what left the head to measuring what actually arrived in the destination. Fluorometric readouts of a dye standard, absorbance, and image-based droplet counts all answer the real question, which is how much liquid is now in the well, not how much the mechanism believes it fired.
This is a healthier way to think even when gravimetry is available, because it measures the outcome you care about. The head can fire a perfect train of droplets and still under-deliver if half of them scatter or evaporate in flight. Trust the destination.
The new failure modes
Leaving the contact world does not leave trouble behind. It swaps one set of failure modes for another, and the new ones bite differently.
- Clogging: viscous fluids and particulates that a wide tip would tolerate will block a fine jetting nozzle, so anything with suspended matter or a heavy viscosity is a poor fit for a non-contact head.
- Composition sensitivity: acoustic ejection needs a known and compatible fluid, and a well whose contents are unexpectedly concentrated, layered, or contaminated will eject wrong or not at all.
- Evaporation: at nanoliter scale the ratio of surface to volume is brutal, so an open source well or a freshly dispensed destination droplet can lose a meaningful fraction of its volume before the run is even finished.
Evaporation deserves the most respect, because unlike a clog it gives you no error. Nothing jams, nothing alarms, and the plate simply reads low. It is the quiet failure of the small-volume world, and it is the one that will fool you into distrusting a head that was firing perfectly.
A liquid class does not vanish when you go non-contact. It sheds every parameter you knew and grows a new set, and the discipline is recognizing that you are calibrating a different machine to answer the same question.