Mechanics

The pendulum that slows as it empties

Hang a can of sand from a rod, open a hole in its base and set it swinging. A pendulum's period does not depend on the mass of its bob, so it might be expected to keep time while the sand pours out. It does not: it slows for most of the drain, by a couple of per cent, and then speeds up again in the last moments, ending close to where it began. The mass is irrelevant; where the mass is, is not. Draining from the top of the heap lowers the centre of mass until it meets the falling surface of the sand, and then raises it.

Assumes: The pendulum, and the small lie that makes it simple · The length nobody has to measure

The first thing a student learns about a pendulum is that the mass of its bob does not matter. A heavy bob and a light one on strings of the same length swing in the same time, because gravity pulls harder on the heavy one in exactly the proportion that it resists being moved. The pendulum and its small lie put that in one formula, T=2πL/gT = 2\pi\sqrt{L/g}, with no mass in it.

So consider a bob that loses mass while it swings: a can of sand with a hole in its base, hung from a rod and set going. A pendulum like this was a classroom favourite for drawing curves on paper, the sand trickling out to trace the swing on a moving sheet underneath. If mass does not matter, the period ought to stay the same as the sand runs out. Timed carefully, it does not. The period lengthens, by about two per cent, over most of the drain, then shortens sharply at the end, finishing a little longer than it began.

The mass really does not matter. What the formula hides is the length. LL is the distance from the pivot to the bob’s centre of mass, more precisely a length built from how the mass is distributed, and the sand draining from a can changes that distribution all the time.

The bob is not a point

A can of sand ten centimetres wide and fifteen tall, hung below a rod a metre long, is not a point mass at the end of a string. It is a compound pendulum, an extended rigid body swinging about a fixed axis, and its period is

T=2πIMg d,T = 2\pi\sqrt{\frac{I}{M g\, d}},

where II is its moment of inertia about the pivot, MM its total mass and dd the distance from the pivot to its centre of mass. The ratio I/MdI/Md is the length of the simple pendulum that would swing in the same time, the equivalent length. The length nobody has to measure told how Kater found that length for a real bar by building a pendulum that could be swung from either of two points, rather than by trying to compute it. For the can it can be computed, because the shapes are simple: a thin cylindrical wall, a disc of a base, and a cylinder of sand filling the can from the bottom to some height hh.

Each part contributes its mass times the square of its distance from the pivot, plus its own moment of inertia about its centre. The sand moves with the can, its grains jammed against each other and the wall, so it swings as a rigid body and its own rotational inertia counts. For a can of these proportions hung a metre down, the own-inertia terms are a fraction of a per cent of the whole, and the equivalent length is very nearly the distance to the centre of mass. Almost everything in the period is that one distance.

A draining pendulum slows down and then speeds up again. The period of a pendulum whose bob is a can of sand, 10 cm wide and 15 cm tall, hung on a light rod 1 m long, as the sand drains out of the bottom. Full, it swings in 2.083 s with a 120 g can; empty, in 2.097 s. In between it is slower than either: the sand that remains sits ever lower, carrying the centre of mass down, until so little is left that the empty can's own mass takes over. The longest period, 2.128 s, comes with 17% left. A lighter can makes the rise larger and pushes it later.
Fig. 1 The period of a pendulum whose bob is a can of sand, 10 cm wide and 15 cm tall, on a light rod 1 m long, as the sand drains from a hole in its base, for cans of 40, 120 and 400 g. Full it swings in 2.083 s with the 120 g can, and empty in 2.097 s; in between it is slower than either, longest at 2.128 s with about a sixth of the sand left.

The curve for a 120-gram can begins at 2.083 seconds with the can full and ends at 2.097 seconds with it empty. Between those ends it rises to 2.128 seconds, two per cent longer than at the start, with about one sixth of the sand still in the can, and then drops steeply. The lighter the can, the higher and later the peak; a heavy can damps the whole effect. Nothing about the mass of the sand enters except through where it is.

Where the centre of mass goes

The sand drains from the bottom, but what disappears is the top of the heap. The grains near the hole leave and the grains above them move down to take their place, so the column of sand gets shorter from its upper surface while its base stays on the floor of the can. Removing material from the top of a column moves its centre of mass down. The centre of mass of the can and the sand together is pulled down with it, the equivalent length grows, and the pendulum slows.

That cannot go on to the end. When the can is nearly empty, the remaining sand is a thin layer on the floor, and the can’s own centre of mass, halfway up its wall, dominates the total. Removing that last layer lightens the bottom of the bob, and the combined centre of mass rises back towards the can’s own. The period follows.

The centre of mass is lowest when it reaches the sand's surface. How far below the pivot the centre of mass of can and sand lies, and where the surface of the sand is, as the sand drains (120 g can, 10 by 15 cm, sand of density 1.5 g/cm³). Draining from the top of the heap lowers the centre of mass as long as the sand being removed lies above it, and raises it once the surface has fallen below. So the lowest point, 112.4 cm below the pivot with 18% left, is exactly where the falling surface meets the centre of mass.
Fig. 2 The depth below the pivot of the combined centre of mass of can and sand (solid) and of the sand’s surface (dashed), as the sand drains, for the 120 g can. The centre of mass is lowest, 112.4 cm below the pivot with 18 per cent of the sand left, exactly where the falling surface reaches it.

There is an exact rule for where the turnaround comes, and it is one of the tidiest results in elementary mechanics. Take a thin slice of sand off the top. If the slice lies above the combined centre of mass, removing it lowers the centre of mass; if below, raising it. So the centre of mass falls as long as the surface is above it and rises once the surface is below it, and it is lowest at the moment the falling surface passes through it. The same rule says that a can of drink, emptied a sip at a time, is hardest to knock over when the level of the drink has fallen to the combined centre of mass of can and drink, which is why a part-empty can stands more steadily than a full one or an empty one. In the figure the two curves meet at the minimum, at 112.4 centimetres below the pivot.

The period’s maximum comes very slightly earlier than the centre of mass’s lowest point, at 17 per cent of the sand left rather than 18, because the period depends on the moment of inertia as well and the sand’s own rotational inertia shifts the balance a little. The difference is too small to see in a classroom and large enough that the two calculations should not be confused.

Where the turnaround falls, in one line

The rule that the centre of mass is lowest at the surface turns into a formula for when the turnaround comes. Treat the can as a mass mcm_c whose own centre of mass is halfway up its height HH, and the sand as a column with mass λ\lambda per unit height standing on the floor of the can to a height hh. Measure depths from the floor. The combined centre of mass is at height

z=mcH/2+λh2/2mc+λhz = \frac{m_c H/2 + \lambda h^2/2}{m_c + \lambda h}

above the floor, and setting z=hz = h, the height of the surface, gives

λh22+mch−mcH2=0,h∗=mc2+λmcH−mcλ.\frac{\lambda h^2}{2} + m_c h - \frac{m_c H}{2} = 0, \qquad h^\ast = \frac{\sqrt{m_c^2 + \lambda m_c H} - m_c}{\lambda}.

For the 120-gram can and its 1.77 kilograms of sand this gives a fifth of the sand left, close to the 18 per cent the fuller calculation finds once the can’s base is placed at its floor rather than spread up its wall. The formula has a clean limit. When the can is light, its own mass is negligible next to λh\lambda h except near the end, and the turnaround height becomes

h∗H≈mcMsand,\frac{h^\ast}{H} \approx \sqrt{\frac{m_c}{M_{\text{sand}}}},

the square root of the ratio of the can’s mass to the full load of sand. A can a hundredth the mass of its sand turns the period round with a tenth of the sand left; a can a quarter of its sand, with half. The square root is why even a light can matters: a mass ratio of one per cent sounds negligible, but it decides the last tenth of the drain, which is where the period changes fastest.

The same argument settles a question that sounds unrelated. The point that keeps moving treated the centre of mass as the single point whose motion external forces govern; here it is the point whose position sets a period, and the same weighted average answers both. Neither the pendulum nor the can of drink cares how much it holds, only how its contents are stacked.

How much a can’s own weight takes away

The size of the effect is a contest between the sand, which moves its centre of mass by up to half the can’s height as it drains, and the can, whose centre of mass stays fixed. A massless can would let the sand’s centre of mass carry the whole bob down to the floor of the can, the period rising until the last grain left and then jumping back. A heavy can holds the combined centre of mass near its own and the sand’s movement barely registers.

The lighter the can, the bigger the slowing. The largest rise in the period of a draining sand pendulum, compared with its period full, against the mass of the can as a fraction of the sand it holds. A massless can would let the period rise until the last grain left, by the full shift of the sand's centre of mass; a heavy can holds the centre of mass in place and the rise vanishes. For a can a fifteenth of its sand, the period rises by about two per cent, and the maximum comes when about a fifth of the sand is left.
Fig. 3 The largest rise in period over the full can’s period, against the mass of the empty can as a fraction of the sand it holds. A can a fifteenth of its sand gives about two per cent, with the maximum at about a fifth of the sand left; a can as heavy as half its sand gives a fraction of a per cent.

For a light can, a few per cent of its sand by mass, the rise approaches the geometric limit: the centre of mass moves from the middle of the can to near its floor, seven and a half centimetres in a metre and a bit, and the period, which goes as the square root of the length, rises by a little over three per cent. For a can weighing half as much as its sand, the rise is under half a per cent. The peak also moves: the heavier the can, the earlier in the drain the falling surface meets the centre of mass, because the can’s weight holds the centre of mass high.

This is the sense in which the mass of the bob matters after all. It matters not because heavier things fall differently but because the bob is two things, a container and its contents, and their masses decide how far the moving one can drag their common centre.

Sand keeps time, water does not

Sand has one more property that makes it the right material for this pendulum. A granular material flowing out of a hole in the bottom of a container leaves at a rate that does not depend on how much is above it. The hourglass that keeps time traced that to the walls carrying the weight of the grains, so the pressure at the hole stops growing once the column is a little taller than the container is wide. The level in the can therefore falls at a constant speed, and the period traced against time is the same curve as the period against the sand remaining.

Water behaves differently. The vessel whose level falls evenly followed Torricelli’s law, in which water leaves a hole at a speed set by the depth above it, so a cylinder empties quickly at first and slowly at the end.

Sand and water drain the same can on different clocks. The period of the same pendulum against time when the can, emptying in two minutes, holds sand or water. A granular flow through a hole runs at a constant rate whatever the depth above it, so the sand's level falls evenly and the period traces the same curve as against the fraction left. Water runs out faster when deep, so the level falls quickly at first and slowly at the end; the same curve is squeezed into the early part of the run and the peak comes at 68 s instead of 100 s. (The water is treated as if it moved with the can, which a sloshing liquid does not quite do.)
Fig. 4 The period of the same pendulum over a two-minute drain, with the can holding sand, which flows out at a constant rate, and water, which flows out faster when deep. The sand’s curve is the period-against-fill curve laid out evenly in time, with its peak at 100 s; the water’s is squeezed into the early part of the run, with its peak at 68 s.

With water in the can, the same rise and fall is compressed into the first part of the run and stretched out at the end, and the peak comes earlier, at 68 seconds rather than 100 in a drain lasting two minutes. Water also swings differently inside a moving can. It does not move as a rigid body: it sloshes, and a liquid in a container swinging about a distant axis lags and leads the container in a way grains held by friction do not. For a can a metre below the pivot the difference to the period is small, but it adds a second, faster oscillation to the motion that a sand pendulum does not have. The water curve in the figure treats the water as if it moved with the can, which is why it is labelled as a comparison of draining laws rather than a prediction for a can of water.

The sand that leaves takes nothing with it

There is a tempting second explanation for the slowing, and it is wrong. Sand streaming out of a moving can might be thought to push on it, like the exhaust of a rocket, and so to brake or drive the swing. The push that needs nothing to push against measured a rocket’s thrust as the exhaust’s speed relative to the rocket times the rate at which mass leaves. Sand falling through a hole in the base leaves with the can’s own velocity: it has no speed relative to the can at the moment it leaves, so it carries away exactly its own share of the momentum and exerts no thrust. Its departure changes the mass of the bob and, through the redistribution of what remains, the period. It does not push.

What the departing sand does carry away is angular momentum about the pivot, its mass times the square of its distance from the pivot times the angular velocity. Because it leaves from the bottom of the can, the furthest point from the pivot, it takes away a little more than its average share. That changes the amplitude of the swing slightly over a long drain, but not the period, which is set at each instant by the distribution of the mass that stays.

Counting the lost swings

A change of two per cent in a period of two seconds is forty milliseconds a swing, which is hard to see by eye on a single swing and easy to see by comparison.

The slowing adds up to most of a swing. How far the draining sand pendulum falls behind an identical pendulum that stays full, counted in swings, over a two-minute drain (120 g can, 1 m rod). A change of a per cent or two in the period is hard to see swing by swing, but it accumulates: by the time the can is empty the draining pendulum is 0.75 of a swing behind, 1.6 s, which a second pendulum swinging beside it makes obvious.
Fig. 5 How many swings the draining pendulum falls behind an identical one that stays full, over a two-minute drain with a 120 g can. The lag accumulates to 0.75 of a swing, 1.6 s, by the time the can is empty.

Hang two identical cans side by side, both full, one with its hole sealed, and start them swinging together. Over a two-minute drain the draining one drifts behind, until by the time it is empty it is three-quarters of a swing out of step, nearly in opposition. Nothing about either pendulum individually looks different; the sum of many small differences in period is what shows. It is the same principle by which a clock’s rate is measured, and the same one that the pendulum that must not grow when it warms used to show how a fraction of a millimetre of thermal expansion costs a clock half a second a day. A pendulum clock compensated against temperature carries its mass in a container whose centre of mass is designed to stay put; Graham’s mercury pendulum did it by letting mercury in a jar rise as the rod expanded, which is the sand pendulum’s mechanism run deliberately in the opposite direction.

What the model leaves out

The can is treated as a rigid body swinging in a plane on a light rigid rod. On a string rather than a rod, a bob this size swings as a double pendulum, the can tipping relative to the string, and its period is not exactly that of the compound pendulum. The sand is treated as moving rigidly with the can and as draining evenly from the top of a level surface. Real sand draining through a hole forms a funnel-shaped depression above the hole and leaves a sloping heap against the walls, so its surface is not flat and its centre of mass is not exactly where a level surface would put it; the shape of the heap depends on the grain’s angle of repose and the hole’s size, and it shifts the numbers by a few millimetres.

The swing itself disturbs the flow. A granular flow’s rate depends on the effective gravity at the hole, and in a swinging can the effective gravity varies through the swing, largest at the bottom where the bob moves fastest. For small swings the variation is small and averages out; for large ones the sand pours faster at the bottom of each swing and the drain rate becomes slightly uneven. The period is also computed for small swings. The period that depends on the swing showed that a larger amplitude lengthens it, by about four tenths of a per cent at fifteen degrees, which is a correction of the same order as the effects here and would have to be held constant in any careful measurement.

The domain in which the figures are right is therefore a can on a rod, swinging gently, with dry free-flowing sand. Within it, the shape of the curve, its rise, its maximum where the surface meets the centre of mass and its fall, is exact.

Still open: what the drain does to the amplitude

A pendulum whose length changes slowly keeps a quantity called the adiabatic invariant, its energy divided by its frequency, and so a lengthening pendulum swings through a smaller angle, as a lengthening string in the swing that is pumped, not pushed changes the energy of its swing. A draining sand pendulum combines a slowly changing length with a slowly changing mass whose departing parts carry away more than their average share of angular momentum. Whether its amplitude rises or falls over a drain, and by how much, depends on the balance of those two effects and on how the flow varies through each swing. Careful measurements of the amplitude of draining pendulums are rare; the drift is a fraction of a per cent and easily masked by air drag, which damps the swing over the same two minutes by more.

The habit worth carrying away is to ask where the mass is, not how much there is. A pendulum’s period ignores the mass of its bob but not its distribution: a can of sand draining from its base loses its upper layers, so the centre of mass sinks until it meets the falling surface and then rises, and a 1 m sand pendulum slows from 2.083 s to 2.128 s with a sixth of its sand left before ending at 2.097 s, falling three-quarters of a swing behind a full twin over a two-minute drain. A bob that loses mass changes nothing; a bob that loses mass from one end changes the length.

Part 11 of 11

This essay is one argument about Pendulum. 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.

Centre of massCompound pendulumGranular flowMoment of inertiaPendulumVariable mass