MathematicsPreprintTheory3 min read

1729 IS THE ONLY CARMICHAEL NUMBER OF ITS KIND

A Carmichael number is a composite number N (a whole number greater than 1 that is not prime) with a curious property: for every whole number a, the number aᴺ − a is divisible by N. A test from 1899, Korselt’s criterion, makes them easy to check: N must be built from distinct primes, with no prime repeated, and for each of its primes q, q − 1 must divide N − 1. Every Carmichael number is odd and has at least three distinct prime factors.

Take 1729, the Hardy–Ramanujan taxicab number. It is 7 × 13 × 19, three distinct primes. Subtract 1 from each: 6, 12 and 18. All three divide 1728. So 1729 is a Carmichael number.

A family of numbers built on powers of 2

Mathematicians have hunted Carmichael numbers inside sequences of the form 2ⁿ × k + 1, with k odd. Javier Cilleruelo, Florian Luca and Amalia Pizarro-Madariaga showed that each such sequence contains only finitely many of them, and that the smallest k that works at all is 27 — with n = 6, which gives 1729 = 2⁶ × 27 + 1.

Then the pieces fell one by one. Earlier work showed that k cannot be a prime, nor the square of a prime. The cube case is different, since 27 = 3³ does work, and Luca proved that if 2ⁿ × p³ + 1 is a Carmichael number for some prime p, then p must be 3. One question remained: which values of n work in the sequence 27 × 2ⁿ + 1?

Only n = 6

Jean Lucien Randrianantenaina and Ali K. Uncu, at the University of Bath in the United Kingdom, answer it: 27 × 2ⁿ + 1 is a Carmichael number only for n = 6. Combined with Luca’s theorem, 1729 is the only Carmichael number of the form 2ⁿ × p³ + 1 with p prime.

Their proof works in four steps:

  1. By Korselt’s criterion, since q − 1 must divide N − 1 = 27 × 2ⁿ, every prime factor of such a number must look like 2ᵃ × 3ᵇ + 1, with b between 0 and 3.
  2. At most one of these factors can be “dependent” on the number: some power of q − 1 equals some power of N − 1. That happens only when n is a multiple of 3. In 1729, it is 13 = 3 × 2² + 1: 13 − 1 = 12, and 1729 − 1 = 1728 = 12³. Earlier general results required p to be larger than the exponent 3 in p³, so they did not cover p = 3.
  3. When b = 0, a factor has the form 2ᵃ + 1, a Fermat prime. A sharper bound rules these out altogether: for k = 27, only 3, 5 and 17 were possible, and none of them can divide such a number if it is a Carmichael number.
  4. A counting argument then caps n at 2,742. For each remaining n, a sieve multiplies all the allowed candidate factors and checks, with exact arithmetic, whether the number divides the product. A Carmichael number is the product of its distinct prime factors, all on the candidate list, so it must divide that product; if it does not, that n is out. No factorisation, no primality test.

Only n = 6 survives, with the three factors 7, 13 and 19. A second run with older, looser bounds, up to n = 4,350, gives the same answer.

In memory of Florian Luca

The paper is dedicated to Luca, “a great inspiration and teacher”, whose theorem it completes. The authors also point to another number with the same cube structure, 46657 = 2⁶ × 3⁶ + 1 = 36³ + 1 = 13 × 37 × 97, also a Carmichael number. Whether it is the only Carmichael number of the form 2ᵃ × 3⁶ + 1 is not yet known: the authors plan to study such cases next.