Ivars Peterson's MathTrek

# Medieval Harmony

The history of mathematics contains many gems of mathematical reasoning and thought.

At a Mathematical Sciences Research Institute meeting last October, number theorist Hendrik W. Lenstra Jr. of the University of California at Berkeley provided a glimpse of one of these gems, dating back to the 14th century.

Philippe de Vitry (1291-1361) was one of the most prominent figures in medieval music. He was the author of an important music theory text, Ars Nova, which introduced new rhythmic schemes and musical notation. He had a deep knowledge of philosophy, rhetoric, and mathematics.

In many ways, de Vitry's interests and accomplishments reflected the Pythagorean view that music is a subdivision of arithmetic, as shown, for example, in the simple mathematical relations between pitch and length of a string (see Circles of Dissonance, Nov. 24, 1997). His work honored the dictum of the Roman philosopher Boethius (480-524) that "music is number made audible."

One of de Vitry's observations concerned numbers that can be written as a power of 2 multiplied by a power of 3. In modern terms, such numbers would be expressed as 2n x 3m, where n and m are integers. He called them harmonic numbers.

Here's a table of all harmonic numbers less than 1000.

 1 3 9 27 81 243 729 2 6 18 54 162 486 4 12 36 108 324 972 8 24 72 216 648 16 48 144 432 32 96 288 864 64 192 576 128 384 256 768 512

De Vitry focused on pairs of harmonic numbers that differ by 1: 1, 2; 2, 3; 3, 4; and 8, 9. As it happens, those particular pairs correspond to musically significant ratios, representing an octave, fifth, fourth, and whole tone. He wondered whether there are any other such pairs of harmonic numbers.

In 1342, Levi ben Gerson, also known as Gersonides (1288-1344), proved that the original four pairs of harmonic numbers are the only ones that differ by 1. Here's how his proof went (using modern notation).

If two harmonic numbers differ by 1, one must be odd and the other even. The only odd numbers are powers of 3 (top row of chart, above). Hence, one of the two harmonic numbers must be a power of 3 and the other a power of 2 (first row and first column of chart). The basic task then involves solving two equations:

2n = 3m + 1 and 2n = 3m - 1.

Gersonides had the idea of looking at remainders after division of powers of 3 by 8 and powers of 2 by 8. For example, 27 divided by 8 gives a remainder of 3. For powers of 3, the remainders all turn out to be 1 or 3, depending on whether the power is even or odd. The remainders for powers of 2 are 1, 2, 4, then 0 for all powers higher than 2.

For 2n = 3m + 1, when m is odd, 3m has remainder 3, and 2n = 3m + 1 must then have the remainder 4, so n = 2 and m = 1. That gives the consecutive harmonic numbers 3, 4. When m is odd, the equation gives the consecutive harmonic numbers 1, 2.

For 2n = 3m - 1, when m is odd, 3m has remainder 3, so 2n = 3m - 1 has remainder 2; as a result, n = 1 and m = 1, to give the consecutive harmonic numbers 2, 3. The final case, when m is even, is a little trickier and requires substituting 2k for m, then solving the equation 2n = 32k - 1 = (3k - 1)(3k + 1). That gives the consecutive harmonic numbers 8, 9. QED.

Interestingly, the theorem about harmonic numbers proved by Gersonides has connections with Fermat's last theorem and with the ABC conjecture, a topic presently of considerable interest among number theorists (see The Amazing ABC Conjecture, Dec. 8, 1997).

Copyright 1999 by Ivars Peterson

References:

Grattan-Guinness, I. 1998. The Norton History of the Mathematical Sciences: The Rainbow of Mathematics. New York: W.W. Norton.

Hendrik Lenstra's lecture notes on harmonic numbers and a video of his talk are available at the MSRI Web site (http://www.msri.org/publications/video/contents.html). For a directory listing of pages corresponding to Lenstra's overhead transparencies, go to http://www.msri.org/publications/ln/msri/1998/mandm/lenstra/1/slides/ .

Comments are welcome. Please send messages to Ivars Peterson at ipeterson@maa.org.