4
$\begingroup$

In Error-Correcting Codes: A Mathematical Introduction (Chapman & Hall, 1998), John Baylis wrote (p.109)

Moving on to 2-error correcting linear codes, the condition for perfection of linear codes of dimension $k$ is $$M = 2^k = \frac{2^n}{1 + n + \binom{n}{2}} = \frac{2^{n+1}}{2 + n + n^2}$$ so $2+n+n^2$ must be a power of 2. It was shown in 1930 that $n = 1, 2, 5$ and 90 are the only positive integers for which this is true.

without any citation. Any ideas whose 1930 number theory result he had in mind? Of course, a reference would be wonderful.

This came up in a MathOverflow question about when $\sum_{k=0}^n \binom{N}{n}$ is a power of 2.


Edit: As markvs explains in his answer (and as discussed in my answer to the MO question), this is equivalent to what is now known as the Ramanujan--Nagell equation $2^n + 7 = x^2$. This only deepens the question of what Baylis was referring to:

  • Is there a 1930 result that addresses $2^n + 7 = x^2$, predating Nagell's 1948 solution (and Ljunggren's 1943 posing of the problem, both presumably unaware of Ramanujan's 1913 question)?

  • Was there a 1930 result about when $n^2 + n + 2$ is a power of 2 with no connection made to Ramanujan's question?

$\endgroup$
4
  • $\begingroup$ Who is the first (so far) known person to state and prove that $n^2 + n + 2$ or the sum of the first three numbers in a row of Pascal's triangle (as Golay put it in 1949) can be a power of two iff $n=1,2,5,90$? Nagell does not seem to do it. Was it Cohen 1964? $\endgroup$
    – Conifold
    Jan 9 at 5:24
  • 1
    $\begingroup$ @Conifold There are two possibilities the Cohen mentions. (1) Shapiro & Slotnick come very close (On the mathematical theory of error correcting codes, IBM Journal, 1959): they invoke an additional condition from coding theory for one case of their Theorem 5 since "the calculation is reduced somewhat," but they claim it could be done without it. (continued in next comment) $\endgroup$ Jan 9 at 16:56
  • 1
    $\begingroup$ @Conifold (2) Browkin & Schinzel (Sur les nombres de Mersenne qui sont triangulaires, C. R. Acad. Sci. Paris, 1956) find all solutions to $2^x - 1 = y(y+1)/2$, which is closer than the form of Ramanujan's question. By the way, based on the title of a 1960 work of the same two authors, On the equation $2^n−D=y^2$ (Bull. Acad. Polon. Sci. Sér. Sci. Math. Astronom. Phys.), they probably made the connection to the Ramanujan-Nagell equation, although I have not been able to access this latter article to verify their references. $\endgroup$ Jan 9 at 17:13
  • 1
    $\begingroup$ At least, the MR reviewer (S. Chowla) of their 1960 paper did make the connection in MR0130215, according to MathSciNet. $\endgroup$
    – Conifold
    Jan 10 at 6:47

1 Answer 1

4
$\begingroup$

See this oeis entry. This can be easily reduced to Ramanujan-Nagell problem. It was conjectured by Ramanujan in 1913 and resolved by Nagell in 1948 (where the year 1930 come from is not clear).

$\endgroup$
6
  • $\begingroup$ Thanks. I'll add this background to the question; I had found these connections in the MO question. Perhaps Baylis was referring to an earlier solution of Ramanujan's "conjecture," or something more directly in the context he mentioned. $\endgroup$ Jan 8 at 15:16
  • $\begingroup$ The solution of R. conjecture was in 1948. The fact that it implies your statement was noticed in the 21st century ($\sim$ 2012). I am not sure adding an answer to a question is appropriate, but do as you wish. $\endgroup$
    – markvs
    Jan 8 at 15:43
  • $\begingroup$ The connection between Nagell's solution and the statement about binomial sums was made by 1964 at the latest: Edward Cohen, A note on perfect double error-correcting codes on $q$ symbols, Information and Control 7 (1964) 381-384. $\endgroup$ Jan 8 at 16:23
  • $\begingroup$ What I wrote is taken from OEIS. Ask them to correct the error. Anyway Nagell's solution is dated 1944, not 1930. 1964 is also not 1930. $\endgroup$
    – markvs
    Jan 8 at 16:35
  • 1
    $\begingroup$ The most reasonable explanation of 1930 is that it is a misprint. $\endgroup$
    – markvs
    Jan 8 at 18:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.