The Universe of Disco

Archimedes and the square root of 3, revisited
Back in 2006 I discussed Archimedes' calculation of the approximate value of π. In the calculation, he needed rational approximations to several irrational quantities, such as √3, and pulled approximations like 265/153 apparently out of thin air.

I pointed out that although the approximations seem to come out of thin air, a little thought reveals where they probably did come from; it's not very hard. Briefly, you tabulate a² and 3b², and look for numbers from one column that are close to numbers from the other; see the previous article for details. But Dr. Chuck Lindsey, the author of a superb explanation of Archimedes' methods, and a professor at Florida Gulf Coast University, seemed mystified by the appearance of the fraction 265/153:

Throughout this proof, Archimedes uses several rational approximations to various square roots. Nowhere does he say how he got those approximations—they are simply stated without any explanation—so how he came up with some of these is anybody's guess.

This was news to me. I did not know anyone else had been puzzled by the 265/153. I had assumed that nearly everyone else saw it the same way that M. Nockolds and I did. But M. Nockolds provided me with a link to an extensive discussion of the matter, which included quotations from several noted mathematicians and historians of mathematics:

It would seem...that [Archimedes] had some (at present unknown) method of extracting the square root of numbers approximately.
—W.W Rouse Ball, Short Account of The History of Mathematics, 1908

...the calculation [of π] starts from a greater and lesser limit to the value of √3, which Archimedes assumes without remark as known, namely 265/153 < √3 < 1351/780. How did Archimedes arrive at this particular approximation? No puzzle has exercised more fascination upon writers interested in the history of mathematics... The simplest supposition is certainly [the "Babylonian method"; see Kline below]. Another suggestion...is that the successive solutions in integers of the equations x²-3y²=1 and x²-3y²=-2 may have been found...in a similar way to...the Pythagoreans. The rest of the suggestions amount for the most part to the use of the method of continued fractions more or less disguised.
—T. Heath, A History of Greek Mathematics, 1921

He also obtained an excellent approximation to √3, namely 1351/780 > √3 > 265/153, but does not explain how he got this result. Among the many conjectures in the historical literature concerning its derivation the following is very plausible. Given a number A, if one writes it as a² ± b where a² is the rational square nearest to A, larger or smaller, and b is the remainder, then a ± b/2a > √A > a ± b/(2a±1). Several applications of this procedure do produce Archimedes' result.
—M. Kline, Mathematical Thought From Ancient To Modern Times, 1972

Archimedes approximated √3 by the slightly smaller value 265/153... How he managed to extract his square roots with such accuracy...is one of the puzzles that this extraordinary man has bequeathed to us.
—P. Beckmann, A History of π, 1977

No, I'm mystified. Even working with craptastic Greek numerals, it would not take Archimedes very long to tabulate kn² far enough to discover that 3·780² = 1351² - 1. Or, if you don't like that theory, try this one: He tabulated n² and 3n² far enough to discover the following approximations:

2	/	1
5	/	3
7	/	4
19	/	11
26	/	15
71	/	41
97	/	56

And the pattern is obvious. In the left column, we have 2+5=7, 5+2·7=19, 7+19=26, 19+2·26=71, 26+71=97. In the right column we have 1+3=4, 3+2·4=11, 4+11=15, 11+2·15=41, 15+41=56. It would be trivial to conjecture that the next entries should be 71+2·97 = 265 and 41+2·56 = 153 and then to check 265² and 3·153² to see that yes, they are close together. Another couple of iterations will get you to 1351/780, which you can check similarly.

I know someone wants to claim that this is nothing more than the Babylonian method. But this is missing an important point. Although this sort of numeric tinkering might well lead you to discover the Babylonian method, especially if you were Archimedes, it is not the Babylonian method, and it can be done in complete ignorance of the Babylonian method. But it yields the required approximations anyway.

So I will echo Nockolds' puzzlement here. There are a lot of things that Archimedes did that were complex and puzzling, but this is not one of them. You do not need sophisticated algebraic technique to find approximations to surds. You only need to do (at most) a few hours of integer calculation. The puzzle is why people like Rouse Ball and Heath think it is puzzling.

There's an explanation I'm groping for but can't quite articulate, but which goes something like this: Perhaps mathematicians of the late Victorian age lent too much weight to theory and analysis, and not enough to heuristic and simple technique. As a lifelong computer programmer, I have a great appreciation for what can be accomplished by just grinding out the numbers. See my anecdote about the square root algorithm used by the ENIAC, for example. I guessed then that perhaps computer science professors know more about mathematics than I expect, but less about computation. I can imagine the same thing of Victorian mathematicians—but not of Archimedes.

One thing you often hear about pre-19th-century mathematicians is that they were great calculators. I wonder if appreciation of simple arithmetic technique might not have been sometimes lost to the mathematicans from the very end of the pre-computation age, say 1880–1940.

Then again, perhaps I'm not giving them enough credit. Maybe there's something going on that I missed. I haven't checked the original sources to see what they actually say, so who knows? Perhaps Heath discusses the technique I suggested, and then rejects it for some fascinating reason that I, not being an expert in Greek mathematics, can't imagine. If I find out anything else, I will report further.

[Other articles in category /math] permanent link