Today I came across this problem:

For a given integer $q$, find the smallest natural number $n > 1$ such that sum of the $q$th powers of its digits is equal to $n$.

For example, we can't find any number for $q=2$, but we can do it for $q=3$, and it's $153$ because $153$ is the smallest number such that $$1^3+5^3+3^3 = 153.$$ For $q=4$, the smallest such $n$ is $1634$.

I tried to find any properties by writing very simple brute force to check every possible number. Moreover, OEIS doesn't know this sequence.

Is there any better and more interesting approach ?

  • $\begingroup$ I editted the post to try and make the question more clear. If you do not like my edits, feel free to revert back to an earlier version of your post. $\endgroup$
    – JavaMan
    Oct 9, 2011 at 22:18
  • $\begingroup$ It made my post more clear, thanks @DJC :) $\endgroup$ Oct 9, 2011 at 22:42
  • $\begingroup$ Also see en.wikipedia.org/wiki/Narcissistic_number $\endgroup$ Oct 21, 2011 at 21:00

2 Answers 2


OEIS does know this sequence; it's http://oeis.org/A014576.

EDIT: As DJC points out in a comment, I was somewhat mistaken. To make amends, I note that these numbers are known under many names: narcissistic numbers, Armstrong numbers, and, my favorite, pluperfect digital invariant numbers. Searching under these terms will turn up many documents. It appears to have been proved that there are exactly 88 such numbers, and that the largest permissible value of $q$ is 39.

  • 3
    $\begingroup$ Gerry, the sequence you link to requires the number $n$ to have $q$ digits. The OP does not add this criterion. Rather, the sequence I believe is oeis.org/A003321. This sequence makes no restriction on digit length. For example, $4150$ seems to be the smallest number such that $4,150 = 4^5 + 1^5 + 5^5 + 0^5$. However, the smallest $5$ digit number with this property is $54,748$. $\endgroup$
    – JavaMan
    Oct 10, 2011 at 1:53

This isn't really an answer just a few observations. So we're interesting in finding the smallest number $a$ such that

$$a=\sum_{i=0}^n a_i 10^i = \sum_{i=0}^n a_i^p,$$

or equivalently

$$\sum_{i=0}^n a_i(10^i-a_i^{p-1})=0.$$ It seems like we could use this formula to do a little bit better than brute force. In particular if we're doing a brute force once we've found $a_0,\dots,a_{n-1}$ we can directly compute $a_n$. There are also a few edge cases you could rule out based on when $(10^i-9^{p-1})>0$. You could also try a variety of approximation methods using the summation formula as your error function.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .