In 1936 a South African engineer named A. D. Lewis took pitch pipes into the Kalahari. He had heard the dunes at Witsands make a noise and wanted the note.
He got several. Taken slowly, “on one’s ‘sit-upon,’” the sand roared, and the roar rose with speed: a low C when he drew a buried plank at fifteen centimetres a second, and above a metre a second nothing but a swish. When the face let go and ran on its own, the roar gave way to a hum — far more regular, like a distant aeroplane in level flight.
The hum is the thing. Some forty sites have it, in a record running from a Chinese account of about 800 through Marco Polo and Darwin. The note sits between seventy and a hundred and ten hertz — the bottom octave of a cello — with harmonics above, which is why no one calls it a noise. It carries a mile, and outlasts the avalanche that started it.
The grains making it are leaving the dune the entire time.
What fixes the pitch? Since Bagnold the answer has been a fingernail dragged over a book cover’s corrugations: grains have a size, shear them and they collide at some rate, that rate is the note — frequency going as the square root of gravity over grain size.
This archive has been on that slope and got the opposite property. No. 62’s sandpile holds its own critical angle through avalanches of every size — the standing example of a system with no characteristic scale. Same angle, same sand; this one gives one frequency to within a few hertz.
Every account reaches for a string instrument. The sand is the bow; the body is what is in dispute.
One camp says there isn’t one. Andreotti recorded a hundred hertz on Moroccan barchans, indifferent to the dune’s size and to where the sand ran; inside the shear band, filmed through a plate of glass, grains collided about a hundred times a second. Collisions launch surface waves, the waves synchronise the collisions, the two lock. Douady’s group got the note from a blade turning in a ring of booming sand, no dune underneath, and concluded the dune is not needed.
The other camp says the dune is the whole point. Vriend and Hunt laid ninety-six geophones down desert slip faces and found a layered subsurface: a slow, loose top layer over something harder. Sound moves more slowly through dry sand than through air, which makes the layer a trap: waves bounce between air above and dense sand below and reinforce at particular frequencies. Given the measured speeds and a layer a metre and a half deep, the model predicts eighty-one hertz. Eureka gave seventy-nine.
Here is the part I did not expect. Both sides have the same numbers.
Andreotti’s 2008 comment does not dispute the Caltech measurements. It disputes their graph. Their figure pools thirty years of frequencies against grain size and shows nothing; he replots the same points on equal axes, gets a straight line, then prints how to turn his into theirs: swap the axes, widen the grain-size range to twice the data, drop the leftmost point, put error bars on one axis and not the other. That is his account, not a referee’s — but the recipe works. He also doubts their avalanches were avalanches: a man on his backside pulses, and a pulsed slide squeaks rather than hums. A third proposal has grain chains ringing like cantilevers; its author says the evidence is insufficient.
What nobody disputes is that the instrument is perishable. Lewis’s sand went quiet wherever it could take up damp, and came back after heating; sealed in fruit jars it kept its roar indefinitely. He got the same roar out of dried, sieved table salt. The voice is not a property of quartz; it is a condition the sand is in, and on a dune the condition is seasonal. At Dumont, on a wet March day, nothing boomed. In June the same face boomed between six and thirty metres from the crest, nowhere else.
The floor at depth is not merely packed sand. Caltech drove a probe two metres down and pulled up conglomerate — grains cemented, clay in the gaps — and under a microscope the glue is calcite and dolomite, left by rain soaking down from above. Closed pores make the layer fast enough to reflect.
If the waveguide people are right, the rain that shuts the dune up for a week is the agent that spent years laying the floor it sings against. If Andreotti is right, the ideal is an unbroken depth of dry sand and the floor is beside the point. The dune is not saying which.
Both times water has met sand here it was the builder — capillary bridges holding a sandcastle up in No. 80, damp ground for the grass that pins a foredune in No. 97. Here it takes the voice and builds the room.
≈
One loop I’m watching
Next: a machine that holds its own speed without measuring anything. No clock, no sensor, nothing stored — two weights on arms that fly out when it runs fast, fall back when it runs slow, and a valve that answers them. It gave us the word feedback. Its characteristic failure is that it will not settle: it hunts. Next time.