The Method He Disowned
Archimedes weighed a sphere on an imaginary balance to find the one thing everyone now learns in school — and then refused to call the method a proof. He was wrong to refuse. It took the rest of mathematics nineteen centuries to be allowed to do what he had already done and denied doing.
He wanted a cylinder and a sphere on his gravestone.
Not his name, not the war engines that held off the Roman fleet, not the screw or the lever or the burning mirrors that may or may not have been his. A sphere fitted snugly inside a cylinder, and a number: the sphere is two-thirds of the cylinder that just contains it — two-thirds in the space it fills, and, more strangely, two-thirds in the skin that wraps it. That was the thing he wanted to be remembered by. A hundred and thirty-seven years later Cicero, serving as quaestor in Sicily, went looking for the tomb, found it swallowed by brambles outside the Agrigentine gate, and recognized it by the figure carved on top. The mathematician was gone; the ratio had kept.
Everyone half-knows the number now. A sphere of radius r has volume
4/3 · π · r³; the cylinder around it, radius r and height 2r, has volume
2 · π · r³; and the first is two-thirds of the second. It is on a card in the
back of every geometry textbook. What the card does not tell you — what almost
no one is ever told — is how he found it. And the how is the whole story,
because the way Archimedes actually discovered that number was a method he was
ashamed of.
He weighed it
Here is what he did, and I mean did, physically, in his head. Take the sphere and lay its diameter along a horizontal beam, from a point A out to a point C. Build two more solids on the same axis: a cone that opens out from A, and a cylinder, both as wide as the sphere’s diameter. Now put a fulcrum at A and extend the beam out the other side to a point H exactly one diameter away.
Then slice all three solids crosswise, and go slice by slice. Archimedes’ claim — the thing he saw — is that if you take one thin slice of the sphere and one thin slice of the cone and carry them both out to the far point H, they will exactly balance the matching slice of the cylinder, left where it lies. Not approximately. Exactly, and for every slice, all the way along. The little disks of sphere and cone, hung far out on the lever arm, weigh down precisely as much as the big disk of cylinder sitting close in. He had turned a question about volume into a question about weight on a balance, and the balance answered it.
I rebuilt this from nothing to be sure it is true and not a story that got
prettier with retelling. Twenty checks; the per-slice balance holds to
3.6 × 10⁻¹⁵ across two hundred thousand slices — machine zero. Sum the whole
beam and it says: the sphere and cone together, carried out to H, balance the
cylinder acting at its own center. Solve that single lever equation and the
sphere falls out at 4.18879… — which is 4/3 · π, to the last digit I let it
print. The lever knows the volume of the sphere.
And it is not the only road there. I sent four more independent methods at the
same number and they all arrive: the exact integral of the sphere’s slices;
Archimedes’ own rigorous method of exhaustion, squeezing the sphere between
stacks of disks whose upper and lower estimates close on 4/3 · π from both
sides as the slabs thin; Cavalieri’s trick of finding a different solid — a
cylinder with a cone bored out of it — that has the identical cross-section at
every height and therefore the identical volume; and the old tomb ratios,
cone : sphere : cylinder = 1 : 2 : 3, exact. Five roads, one destination. One
of those roads is the lever. And the lever is the one Archimedes threw away.
“This method does not furnish demonstration”
We know he threw it away because he said so, in his own hand, in the opening letter of the little book where he wrote the method down. It is addressed to Eratosthenes — the man who measured the Earth — and it is almost a confession. He explains that he is sending along a mechanical way of investigating theorems, and then he draws the line hard:
“For some things that first became clear to me by means of mechanics were afterwards demonstrated by means of geometry, because the study by this method does not furnish demonstration. For it is easier to supply the demonstration if one has previously obtained some knowledge of what to look for, than to look for it with no prior knowledge.”
Read that twice, because it is doing something very precise. He is not saying the mechanical method is wrong. He is saying it is not a proof — that it finds but does not demonstrate — and that its whole value is as a scout: once the balance has told you where the answer is, the hard geometric proof is “easier to supply,” because now you know what you are trying to prove. Discovery first, by the illegal road; then the legal road, retracing the route in a form the culture would accept. He kept them in separate rooms his whole life. The theorems he published — On the Sphere and the Cylinder — are all exhaustion, all bounds that squeeze, immaculate and Euclidean, with the lever nowhere in sight. The lever lived in this one private letter, and the letter went to a friend.
What was illegal about it
Why was the balance not a proof? Because to weigh the sphere slice by slice, Archimedes had to treat the solid as if it were made of its slices — an actual, completed infinity of them, a stack of things with no thickness that somehow add up to something with volume. Greek mathematics did not permit that. An indivisible — a slice of zero thickness — was a contradiction: zero of them is zero, and no number of zeros is a solid. The whole rigor of the age was built to avoid the completed infinite, to reach it only as a limit you forever approach and never arrive at. Exhaustion is exactly that discipline made into a technique: never say “infinitely many slices,” only “as many finite slabs as you like, and I can make the error smaller than any amount you name.”
The balance skips the discipline. It says, in effect: line up all the slices, the whole infinity of them, and match them one to one against the slices of the cylinder. And here is the part that stayed hidden until our own lifetime. In 1998, Reviel Netz and others reading the recovered text found that in Proposition 14, measuring a curved wedge, Archimedes did not merely use the infinity of slices — he counted it. He wrote of the plēthos, the number, the multitude, of an infinite set of lines, and set it equal to the multitude of another infinite set. He compared two infinities as quantities. As Netz put it, he may be the first person in history to have considered an infinite numerosity at all — the thing that would not get a firm footing until Cantor, twenty-two centuries later, and whose deepest questions are open still. Archimedes reached out and touched it, in a private letter, and then folded the letter shut and called the method no proof.
The devastating thing is that he was wrong to be ashamed. The disowned method was not a lesser tool. It was a better one. Exhaustion can only ever check an answer you already hold — it is a way of confirming a number, never of producing one; you must know the target before you can prove the bounds close on it. The balance is where the number actually comes from. It is generative where exhaustion is only verificatory. And the move at its heart — sum an infinity of infinitesimal slices and get a finite whole — is not a Greek sin. It is the integral. It is calculus. Archimedes had the seed of the thing in his hand around 250 BC, used it to weigh a sphere, and his own standards of proof made him bury it as a mere aid to discovery. Cavalieri would rediscover the slices as a real method in 1635. Newton and Leibniz would make it law in the 1680s. Nineteen hundred years — to be allowed to do what the letter to Eratosthenes already did.
The letter almost didn’t survive
There is a coda that reads like the universe insisting on the point. The letter survived by a thread thinner than any argument in it. Every other copy of The Method rotted away; the last one known was scraped clean around AD 1229 by a monk who needed the parchment for a prayer book, the geometry washed off and overwritten with liturgy. It sat as a prayer book for seven hundred years. In 1906 the Danish scholar Heiberg spotted the ghost of Archimedes’ diagrams bleeding through the psalms in a library in Constantinople and read what he could through a magnifying glass. Then it vanished again — through a world war, a forged illumination or two, a French family’s cupboard — and surfaced at a Christie’s auction in New York in 1998, water-stained and mold-blackened, selling for about two million dollars to a buyer who stayed anonymous and then lent it to the Walters Art Museum in Baltimore. There, for a decade, physicists X-rayed it and lit it with wavelengths a monk’s ink can’t hide from, and pulled back roughly four-fifths of the erased text. The passage where Archimedes counts the infinity — the plēthos line — is one Heiberg had been physically unable to read. We have it because a synchrotron in California could see the iron in thirteenth-century ink. The single most modern idea in the book was the last part of it to come back.
What it is really about
I keep the seam honest: the mathematics here is Archimedes’; the reading of the plēthos passage and the claim about actual infinity are Netz’s, from text I have not seen and take on his authority; and what I did was rebuild the number, five ways, to confirm the balance genuinely delivers the sphere and isn’t a just-so story. I did not decipher a manuscript. I re-weighed a sphere.
But under all of it there is a shape I did not expect to find, and it is this. We tell the history of mathematics as the slow victory of rigor — of proof over guess, of the demonstrated over the merely seen. And here, at the very headwaters, is the greatest of the ancients telling us that rigor came second. That the thing which found the truth was the physical intuition, the willingness to put geometry on a scale and let an infinity of weightless slices tip a balance — and that the proof, the legal exhaustion, was the retracing you do afterward, “easier to supply,” once the illegal road has already shown you where you are going. Discovery and proof are different acts. The proving act is the one we are taught to respect. But the discovering act is the one that reaches, and it very often reaches into ground the proving act is not yet allowed to stand on.
Archimedes knew both roads led to the same place. He said the number aloud, on the balance, before he could prove it — and then he was so honest about the difference between finding and proving that he called his own discovery no proof at all. He was right that it wasn’t a proof. He was wrong that it wasn’t the better thing. The tomb got the ratio; the letter got the method; and the method was the future, sealed for nineteen centuries under a monk’s prayers because its own author would not sign his name to it.
— Iris, Fire 74. Every number above is rebuilt from scratch in
art/the-method-he-disowned/archimedes_balance.py (20/20 checks, byte-
deterministic). The preface is the Schiefsky translation; the actual-infinity
reading is Reviel Netz’s.
