The Kelvin Water Dropper: Lightning From Falling Water

Two streams, two rings, two cans — and one crucial cross-wire

The apparatus is deceptively plain. A raised reservoir feeds two thin nozzles, so that two vertical streams of water fall side by side and break into drops. Each stream passes through a metal ring — the inductor — and the drops that clear the ring land in a metal can below, the receiver. Both cans, and both rings, are mounted on insulating supports so charge cannot leak away to ground.

The whole trick lives in the wiring. The left ring is connected to the right can, and the right ring to the left can — the connections are crossed. Nothing is plugged into a battery; the only input is water falling under gravity. That single crossed pair of wires is what converts a symmetric, do-nothing arrangement into a runaway high-voltage generator. Remove the cross and each side just sits at zero; keep it, and a chance imbalance on either can is fed back to grow the imbalance on the other.

Kelvin himself called it a water-dropping condenser; it belongs to a family of “influence machines” he devised in the 1860s — alongside the separate rotary replenisher — to build up and maintain electric charge for his sensitive electrometers.

How a falling drop steals a charge: electrostatic induction

Water is not an insulator. Tap water carries dissolved ions, so its charge-relaxation time — the time for internal charges to rearrange, roughly τ = ε/σ — is on the order of 10⁻⁷ s (with permittivity ε ≈ 80ε₀ and conductivity σ ∼ 0.01 S/m). A drop takes milliseconds to form and pinch off, so on that timescale water behaves like a conductor: its free charge fully responds to any external field.

Now suppose the right ring is held at a positive potential V (we will see why in a moment). As a drop grows on the right stream inside that positive ring, the ring’s field drives the drop’s free charge apart by electrostatic induction: negative charge is pulled toward the ring’s field, positive charge is pushed up the column back toward the grounded reservoir. While the drop is still attached, it stays near ground potential and the excess positive charge escapes upward. The instant the neck breaks — a Plateau–Rayleigh pinch-off driven by surface tension — the drop is severed from that escape route and flies off carrying a net negative charge, opposite in sign to the ring.

Quantitatively, the detached charge is roughly q ≈ −C_mV, where C_m is the small mutual capacitance between drop and ring. The magnitude is tiny — picocoulombs early on — but its sign is guaranteed: every drop through a positive ring is negative, every drop through a negative ring is positive. That deterministic sign flip is the engine of the whole device.

Positive feedback and exponential runaway

Follow the loop. A negative drop through the right ring falls into the right can, making it more negative. But the right can is wired to the left ring, so the left ring becomes more negative — and a negative ring induces positive charge on the left stream’s drops, which fall into the left can and make it more positive. The left can is wired to the right ring, pushing the right ring more positive, which makes the right drops more negative… and around it goes. Each step reinforces the last. This is textbook positive feedback.

Write it as equations. Let Q_L, Q_R be the can charges and C each can’s self-capacitance, so a can’s ring sits at V = Q/C. With n drops per second each depositing −αV_ring (with α a geometric capacitance) and defining k = nα/C:

• dQ_L/dt = −k Q_R
• dQ_R/dt = −k Q_L

The difference D = Q_L − Q_R then obeys dD/dt = k D, so D(t) = D₀ e^t/τ with time constant τ = C/(nα). The charge grows exponentially. The state Q_L = Q_R = 0 is an unstable fixed point; any seed — thermal noise, a stray ion, a cosmic ray, contact potentials in the metal — is enough to tip it. Which can ends up positive is essentially random from run to run: a small, everyday example of spontaneous symmetry breaking. For growth the loop gain must exceed one, so the rings must couple tightly to the streams and leakage must stay low.

Where the energy comes from: gravity against the field

A high-voltage generator with no battery raises an obvious question: what does the work? The answer is gravity. Consider a negative drop falling toward the already-negative right can. Like charges repel, so the can’s field pushes up on the drop, opposing its fall. The drop lands anyway because gravity still wins — and in doing so it is forced onto a like-charged conductor, which costs energy qV. That energy is drawn from the drop’s gravitational potential, mgh. In effect the machine is a pump that trades the potential energy of falling water for electrostatic potential energy stored in the charged cans.

This also explains the device’s built-in ceiling. As the voltage climbs, the electric force opposing each drop grows until it becomes comparable to the drop’s weight and inertia. Then charged drops are deflected sideways, repelled away from their target can, or the stream sprays into a fine charged mist that misses entirely. Charge delivery stalls. The overall efficiency is modest — most of the water’s mgh ends up as kinetic energy and heat — but the physics is exact: the field never does net work to build itself, and gravity always pays the bill for separating charge against it.

The ceiling: breakdown, corona, and why humidity kills it

If drop deflection does not cap the voltage first, dielectric breakdown of air does. Air ionizes and conducts once the field exceeds roughly 3 MV/m (30 kV/cm) at ordinary pressure — the regime described by Paschen’s law. With the two oppositely charged cans (or a discharge point on each) separated by about a centimetre, breakdown arrives at roughly 10–30 kV, and a spark cracks across. The spark neutralizes the accumulated charge, the potential collapses, and the exponential build-up begins again — giving the characteristic rhythm of a click every few seconds.

Two leak paths compete with the charging current and set a steady-state maximum before any spark. First, corona discharge: at sharp edges the local field is enhanced enough to ionize nearby air, bleeding charge away; a well-made dropper avoids sharp points to push the ceiling higher. Second, and most practically, humidity. A film of moisture on the insulating supports, or simply damper air, provides a surface-conduction path that drains the cans faster than the drops can charge them. This pulls the loop gain below one and the device does nothing. It is a classic lecture-demo heartbreak: the Kelvin dropper works beautifully on a dry winter day and refuses to spark on a muggy summer afternoon.

From Kelvin’s water dropper to modern charged droplets

William Thomson (Lord Kelvin) described the water dropper in 1867, in work on self-acting apparatus for “multiplying and maintaining” electric charge. It belongs to a lineage of induction (influence) machines — Volta’s electrophorus (1775), the Wimshurst machine (~1883), later the Van de Graaff generator (1929) — all of which separate charge mechanically rather than chemically. The dropper is distinctive for using nothing but gravity and geometry. Its bootstrapping is exactly analogous to a self-excited dynamo, which grows a large current from residual magnetism through positive feedback; the water dropper grows a large voltage from residual charge the same way.

A modern cousin worth distinguishing is electrospray ionization, for which John Fenn shared the 2002 Nobel Prize in Chemistry. Electrospray also produces charged droplets, but there an external applied field does the charging — there is no self-amplifying feedback loop. The Kelvin dropper, by contrast, generates its own field from scratch. It remains a favourite teaching demonstration precisely because it packs several deep ideas into a bucket of water: electrostatic induction, capacitance, the difference between conductors and insulators, dielectric breakdown, energy accounting — and, most elegantly, an unstable equilibrium amplified by positive feedback into something that spits sparks.

Electrostatic generators compared: separating charge without a chemical battery.

Generator	Charge-separation mechanism	Energy source	Typical voltage
Kelvin water dropper (1867)	Electrostatic induction + positive feedback	Gravity (falling water)	~10–20 kV
Wimshurst machine (~1883)	Induction on counter-rotating sectored disks	Hand crank (mechanical)	~tens of kV
Van de Graaff (1929)	Corona spray + tribocharging onto a moving belt	Motor (mechanical)	~0.1–25 MV
Electrophorus (Volta, 1775)	Induction + manual charge transfer	Manual lifting	~10 kV per lift
Thunderstorm (natural)	Graupel–ice collisions in convective updrafts	Gravity + convection	~10⁸–10⁹ V