Electromagnetism

The wire that melts where it can no longer balance

A fuse is a piece of wire chosen to melt before anything else in the circuit does. William Preece measured in 1884 that the current needed to melt a wire grows as the diameter to the power three-halves, not as the square — so a thin wire carries three times the current per square millimetre of a thick one. The reason is surface against volume. The reason a fuse is a good fuse rather than a slowly cooking wire is subtler: copper's resistance rises as it heats, and above a certain current no temperature balances the heat it makes. A quarter-millimetre copper wire melts at 10.1 amperes and has no steady state at all above 11.2. The fuse does not wear out; it falls off a cliff.

Assumes: The resistance that is a length · The insulation that makes a wire lose more heat

The oldest overcurrent protection is still among the most common: a short length of wire, thinner than the conductors it protects, placed in series with them so that it melts first. The household cartridge fuse, the blade fuse in a car, the glass tube in a power supply and the strip of tinned copper inside a battery pack all work by the same rule — a wire that cannot shed the heat a current makes in it gets hot until it melts, and the melting opens the circuit.

The rule sounds like it should give a simple answer, the current at which the wire reaches its melting point, and a smooth approach to it. It gives neither. The current that melts a wire does not grow in proportion to the copper in it, and the wire does not approach melting smoothly; it approaches a current above which it has nowhere to stop. Both facts come from writing down the balance between the heat a current makes and the heat a wire can lose, and both are what a fuse designer is actually using.

The surface against the volume

A current II through a wire of cross-section AA and resistivity ρ\rho makes heat at a rate I2ρ/AI^2 \rho / A per unit length — Joule heating, the same dissipation the resistance that is a length found concentrated in the few small spots where two conductors actually touch. A bare wire in air loses heat from its surface, by convection and by radiation, at a rate proportional to its circumference and to how much hotter it is than the air: hπd θh \pi d\, \theta per unit length, with θ\theta the temperature rise and hh a heat-transfer coefficient that bundles the convection and the radiation into one number. In a steady state the two are equal. With the resistivity held fixed, that gives

I=π2 d3/2h θρ.I = \frac{\pi}{2}\, d^{3/2} \sqrt{\frac{h\,\theta}{\rho}}.

The heat is made in the volume and lost through the surface. Doubling the diameter quadruples the copper but only doubles the surface, so the wire can carry only 23/22^{3/2}, about 2.8, times the current before reaching the same temperature — not four times. William Preece, the engineer-in-chief of the British Post Office, measured fusing currents for wires of many metals and diameters and published the three-halves power law in 1884; it is still printed in engineering handbooks as Preece’s formula, with a constant for each metal.

A thinner wire carries more current per square millimetre before it melts. The current that melts a bare copper wire in still air, against its diameter, on logarithmic axes. Heat is made throughout the cross-section, in proportion to d² for a given current density, and lost only through the surface, in proportion to d, so the balance gives a fusing current growing as the 3/2 power of the diameter — Preece's law of 1884 — rather than as d². A 1 mm wire fuses at 81 A, 103 A per square millimetre; a 0.1 mm wire at 2.55 A, 324 A per square millimetre. The dashed line is what the fusing current would be if every wire tolerated the 1 mm wire's current density.
Fig. 1 The current that melts a bare copper wire in still air, against its diameter, on logarithmic axes. Heat is made throughout the cross-section and lost only through the surface, so the fusing current grows as the 3/2 power of the diameter rather than as its square. A 1 mm wire fuses at 81 A, 103 A per square millimetre; a 0.1 mm wire at 2.55 A, 324 A per square millimetre. The dashed line is what the fusing current would be if every wire tolerated the 1 mm wire’s current density.

The consequence that matters is in the current density. Per square millimetre of copper, a thin wire tolerates more current than a thick one, in proportion to one over the square root of the diameter: a tenth of a millimetre of copper melts at over three hundred amperes per square millimetre, a millimetre at about a hundred. That is why a fuse element is thin, and why a printed-circuit trace — a film tens of micrometres thick, whose resistance the sheet resistance that forgets the shape showed depends only on its count of squares — can carry far more current per area than a busbar. It is also why the current ratings of building wiring grow more slowly than the conductor’s area: a cable twice the cross-section is not rated for twice the current, because it does not have twice the surface.

The heat-transfer coefficient used here, 220 watts per square metre per kelvin, is chosen so that the 1 millimetre wire matches Preece’s measured eighty amperes; nothing else is fitted. It is large because at the temperatures that matter — hundreds of degrees — a bare wire loses as much by glowing as by warming the air, and both grow steeply with temperature. A constant hh is a simplification, and it is the simplification that makes the exponent exactly three-halves. Real fusing currents follow the law over a factor of twenty in diameter well enough to have stayed in the handbooks for a hundred and forty years; thin wires lose heat by convection somewhat better than a constant coefficient says, which bends the curve slightly at the thin end. Insulation complicates it further, and not in the obvious direction: the insulation that makes a wire lose more heat found that a thin sheath round a thin wire increases its heat loss, because it adds more surface than resistance.

A resistance that rises with the heat

The balance above treats the resistivity as constant, and copper’s is not. It rises by about 0.39 per cent for every kelvin, so copper at a dull red heat conducts only a third as well as at room temperature, and the heating at a given current rises with the temperature it causes. Put that in the balance and the steady temperature rise becomes

θ=P0H−αP0,P0=I2ρ0A,H=hπd,\theta = \frac{P_0}{H - \alpha P_0}, \qquad P_0 = \frac{I^2\rho_0}{A},\quad H = h\pi d,

with α\alpha the temperature coefficient. The denominator is the whole story. As the current grows, αP0\alpha P_0 approaches HH — the extra heating per degree approaches the extra loss per degree — and the temperature rise grows without limit. Above that current there is no solution at all: whatever temperature the wire reaches, it makes more heat than it loses, and it keeps heating until it melts.

The fuse is a cliff, not a slope. The steady temperature rise of a bare 0.25 mm copper wire in still air against the current through it. Because copper's resistance grows with temperature, the heating grows faster than the loss, and above 11.20 A there is no temperature at which they balance: the wire runs away. It reaches copper's melting point at 10.06 A, only 10% below the runaway, so the last tenth of the current takes the wire from barely glowing to molten. With the resistance held at its room value the same wire would survive to 22.9 A and approach melting gradually.
Fig. 2 The steady temperature rise of a bare 0.25 mm copper wire in still air against the current through it. Because copper’s resistance grows with temperature, the heating grows faster than the loss, and above 11.20 A there is no temperature at which they balance. The wire reaches copper’s melting point at 10.06 A, only 10% below the runaway. With the resistance held at its room value the same wire would survive to 22.9 A and approach melting gradually.

For a quarter-millimetre copper wire the runaway is at 11.20 amperes and the melting point is reached at 10.06, a tenth below it. At nine amperes the wire is at about five hundred degrees, just short of the faint red that any hot solid begins to show near 525 °C whatever it is made of — the point the glow that says nothing about the surface explains; at ten it is at a thousand; just above ten it melts. The wire’s last tenth of current takes it across most of its temperature range, which is what makes it a switch rather than a heater. And the rising resistance has more than halved the current the wire can carry: with its room-temperature resistance fixed, the same wire would reach melting only at 22.9 amperes, and would approach it along a gentle parabola.

This is the shape of every thermal runaway, and it is the same shape the heap that sets itself alight found in a haystack: a source of heat that grows with temperature, set against a loss that grows only in proportion to it. The haystack’s source grows exponentially, through an Arrhenius rate, and the wire’s only linearly, but the condition is the same in both — the heat made per extra degree must stay below the heat lost per extra degree — and both have a critical size or current at which a warm steady state stops existing. In the haystack it is a catastrophe; in the fuse it is the design.

Driven by a current, not by a voltage

The runaway depends on something outside the wire: what holds the current fixed. A fuse sits in series with a load, and the current through it is set by the load and the supply, not by the fuse’s own few milliohms; as the fuse heats and its resistance rises, the current hardly changes, and the heat it makes, I2RI^2R, rises with RR. A lamp filament is the opposite case. It is connected across a fixed voltage, so the heat it makes is V2/RV^2/R, and as it heats and its resistance rises tenfold its power falls: the rising resistance is a brake rather than an accelerator, and a tungsten filament settles at a steady white heat that the same tungsten, carrying a fixed current, could never hold. The material is the same and the temperature coefficient is the same; the stability belongs to the circuit. The pan a better conductor will not heat met the same reversal in an induction hob, where whether a higher resistivity means more heat or less depends on whether the field drives a current into the metal or a voltage round it.

It explains why fuses are rated in a way that sounds cautious. A fuse is specified by a current it will carry indefinitely and a larger one at which it must open within a stated time, and the two are typically a factor of about one and a half to two apart. Between them is the range in which the element is hot but stable — the left-hand part of the curve, where small changes in air flow or ambient temperature move the temperature but not the outcome — and above the runaway the outcome is certain.

How long it takes

A current above the fusing current does not melt the wire instantly; the wire has to be heated, and its heat capacity sets how long that takes. The heat balance with a heat capacity in it is a linear equation in the temperature, because the heating is linear in temperature through the resistivity and the loss is linear through hh, and it has an exact solution:

t=−ρcAsln⁡ ⁣(1−θmsP0),s=H−αP0,t = -\frac{\rho c A}{s}\ln\!\left(1 - \frac{\theta_m s}{P_0}\right), \qquad s = H - \alpha P_0,

where θm\theta_m is the rise to the melting point. When ss is negative — above the runaway — the logarithm still has a meaning, and the formula gives the time to an exponential climb.

Below one current never, above it sooner the harder it is pushed. The time a bare copper wire takes to melt against the current through it, for diameters of 0.1, 0.25 and 1 mm, from the lumped heat balance with copper's rising resistance. Each curve rises vertically at that wire's fusing current — below it the wire never melts — and at large currents falls with a slope of −2, the adiabatic regime in which the heat has no time to leave and the melting energy I²Rt is fixed. The 0.25 mm wire melts in 0.60 s at twice its fusing current. Every diameter has the same shape, shifted: the fusing current grows as the 3/2 power of the diameter and the adiabatic times as its fourth power.
Fig. 3 The time a bare copper wire takes to melt against the current through it, for diameters of 0.1, 0.25 and 1 mm. Each curve rises vertically at that wire’s fusing current — below it the wire never melts — and at large currents falls with a slope of −2, the regime in which the heat has no time to leave. The 0.25 mm wire melts in 0.60 s at twice its fusing current.

This is the fuse maker’s time–current characteristic, the curve printed on every data sheet, and its shape is fixed by the two limits it joins. Near the fusing current the time grows without bound: the wire approaches its melting point more and more slowly, and below the fusing current it never gets there. At large currents the wire has no time to lose anything to the air, all the Joule heat goes into raising its temperature, and the time falls as the inverse square of the current. A quarter-millimetre wire at twice its fusing current, twenty amperes, melts in six-tenths of a second; at a hundred amperes, in a few milliseconds.

The curves for different diameters are the same shape shifted, because both limits scale with the diameter: the fusing current as d3/2d^{3/2}, as above, and the fast-fault times as the heat capacity per length, which goes as d2d^2, divided by the heating per length, which at fixed current goes as 1/d21/d^2 — so as d4d^4. A wire ten times thicker takes ten thousand times longer to melt at the same current.

A fast fault is judged by its energy

The fast end of the curve has a cleaner description than a time. When the heat has no time to leave, the wire melts when it has absorbed a fixed amount of energy per unit volume, and the current enters only through the Joule heat delivered per ohm, the integral of I2I^2 over time. For copper, with its rising resistivity folded in, this gives

I2t=ρc A2ρ0αln⁡(1+αθm),I^2 t = \frac{\rho c\, A^2}{\rho_0 \alpha}\ln(1 + \alpha\theta_m),

which is Onderdonk’s formula for the fusing of conductors under short-circuit currents, still used to size the wiring that has to survive a fault until the protection opens.

A fast fault is judged by its energy, a slow one by its current. The quantity I²t — the Joule heat delivered per ohm — needed to melt a bare 0.25 mm copper wire, against the time the melting takes. For faults lasting less than about 2.0 s the wire cannot shed heat in time, and I²t settles at a floor of 202.4 A²s fixed by the copper's heat capacity and how fast its resistance rises: any current delivering that much melts it. For slow overloads the wire sheds almost everything and only the current matters, and I²t grows in proportion to the time. The crossover time grows in proportion to the diameter.
Fig. 4 The quantity I²t needed to melt a bare 0.25 mm copper wire, against the time the melting takes. For faults lasting less than about 2.0 s the wire cannot shed heat in time, and I²t settles at a floor of 202 A²s fixed by the copper’s heat capacity and how fast its resistance rises. For slow overloads the wire sheds almost everything and only the current matters, and I²t grows in proportion to the time.

For the quarter-millimetre wire the floor is about two hundred ampere-squared-seconds, and the crossover between the two regimes is about two seconds — of the order of the wire’s thermal time constant, the time it would take to heat by the rise it settles at, and growing in proportion to the diameter. This divides the problems a fuse solves. A slow overload, a motor stalled or too many heaters on one circuit, is judged by its current: below the fusing current it can go on for ever, above it the wire eventually melts. A short circuit, with a current a hundred times the rating limited only by the wiring’s resistance, is judged by its energy: what matters is how much I²t passes before the element melts and the arc it leaves is extinguished, and the downstream wiring has to survive that much. Fuse data sheets give both a time–current curve and an I²t, and they are two ends of the one heat balance.

The floor counts only the heat needed to bring the copper to its melting point. Melting it takes more: copper’s latent heat of fusion is about half the heat needed to raise it from room temperature to 1,085 °C, and the heat that changes no temperature has to be supplied before the solid becomes a liquid that can part. So the energy to open a fault is somewhat above the energy to reach the melting point, by an amount that depends on how much of the element has to melt before a neck breaks — which is part of why manufacturers give a melting I²t and a larger clearing I²t, and test for both rather than computing either.

The terminals cool a short element

A fuse element is not an infinitely long wire. It is a short one, held at each end by brass caps or blades that are much larger and much cooler, and heat flows along the copper into those ends as well as out through its surface. The temperature along the element is then highest in the middle and held down near the terminals, and the distance over which the terminals matter is a thermal length — the distance at which conduction along the wire is worth as much as loss from its surface — of about ten millimetres for a quarter-millimetre copper wire, growing as the current approaches the runaway.

The terminals cool a short fuse, and the length is a design choice. The current that melts the middle of a 0.25 mm copper element held between terminals kept at room temperature, against the element's length. Heat conducted along the copper to the cold ends is worth as much as heat lost to the air over a distance of about 10 mm, growing as the current approaches the runaway. A 10 mm element needs 33.5 A, a 50 mm one 11.7 A, and a long wire 10.06 A. Above the long wire's runaway current, 11.20 A, only the conduction to the terminals holds a balance at all, and only for elements shorter than a length that shrinks as the current grows. Fuse makers set the rating with the element's length and its notches as much as with its diameter.
Fig. 5 The current that melts the middle of a 0.25 mm copper element held between terminals kept at room temperature, against the element’s length. A 10 mm element needs 33.5 A, a 50 mm one 11.75 A, and a long wire 10.06 A. Above the long wire’s runaway current, 11.20 A, only conduction to the terminals holds a balance at all, and only for elements shorter than a length that shrinks as the current grows.

A ten-millimetre element of the same wire melts at thirty-three amperes, more than three times what a long wire needs, because its middle is never far from a cold end. A fifty-millimetre one is within a sixth of the long-wire value. Short elements can even be held stable above the long wire’s runaway current, because conduction to the ends is a loss that does not care about the resistivity, and it rescues a short enough element at any current — up to a point, since the length it can rescue shrinks as the current grows.

Fuse makers use this directly. The rating of an element is set by its length as much as by its diameter, and most fuse elements are not plain wires but strips with notches or holes cut in them: the narrow necks are where the current crowds — the same crowding the charge that steers a current round a bend followed at a corner — and so where the heat is made fastest and the element melts first, placed away from the terminals so that the melting happens in a known place and the arc that follows is confined to a known gap. Some elements carry a small bead of tin or a tin–lead alloy on the copper, the so-called M-effect, which melts at a few hundred degrees and dissolves the copper beneath it, so that a long, slow overload opens the fuse at a temperature far below copper’s melting point while a fast fault is still judged by the copper’s I²t.

What the single balance leaves out

Everything here is one wire, one temperature across its cross-section, and one heat-transfer coefficient. The single temperature is a good approximation for thin copper, which conducts heat across its own diameter in a fraction of a millisecond. The constant coefficient is not: convection from a thin hot wire depends on the diameter and on the temperature, and radiation grows as the fourth power of the absolute temperature, so a fitted hh is right at one condition and approximate elsewhere. The coefficient here is fitted to Preece’s one-millimetre value and the other fusing currents follow from the scaling; measured values for thin wires run somewhat above the line, because convection from a thin cylinder is more efficient per unit area than from a thick one. The melting itself is treated as a temperature, where in practice a fuse element first sags, may oxidise, and opens when a molten neck breaks under its own surface tension. And the most important part of a real fuse’s job, extinguishing the arc that forms across the gap when the element parts — which is what the sand in a cartridge fuse is for — is a different subject altogether.

Still open: what happens after the wire parts

The physics of the arc in a breaking fuse is where the hard questions now are. When the element melts, current continues through a plasma channel across the gap, and the fuse has to force that current to zero by lengthening and cooling the arc faster than the circuit can sustain it; in a sand-filled fuse the arc melts the sand into a glassy tube, a fulgurite, that absorbs its energy. How fast the arc voltage rises, and how much energy the fuse lets through before it clears, depends on the element’s geometry, the sand’s grain size and packing, and the circuit’s inductance — which, as the circuit that fights its own change showed, keeps the current flowing with whatever voltage it takes, and so drives the arc — and it is still modelled with heavy reliance on testing — especially for the high-voltage direct-current fuses that electric vehicles and solar installations now need, where there is no natural zero in the current to help the arc go out.

A wire makes heat in its volume and loses it from its surface, so its fusing current grows as the 3/2 power of its diameter — 81 A for 1 mm of copper, 2.55 A for 0.1 mm, three times the current density — and because copper’s resistance rises with temperature the heating outruns the loss: a 0.25 mm wire melts at 10.06 A, a tenth below a runaway at 11.20 A above which no temperature balances, where with a fixed resistance it would last to 22.9 A. A fast fault is then judged by its energy, a floor of 202 A²s, a slow one by its current, and a short element is cooled by its terminals — 33.5 A for 10 mm — which is why a fuse is designed by its length and its notches as much as by its wire. The fuse is the instability a haystack suffers, put to work.

Part 13 of 13

This essay is one argument about Conductors. The others:

The objects named here

The third axis, after the field and the reading path: the things themselves, and every essay that touches each one.

Heat conductionJoule heatingPositive feedbackResistivityThermal runawayTime constant