22

Today on this site I came across a user with the username $2^59^2$. It turns out this is equal to $2592$.

I would like to understand how rare this is. Are there other examples of the following?

$$ N = b_1^{e_1}\cdots b_n^{e_n} = \overline{b_1e_1\cdots b_ne_n} $$

In order of interest:

  1. $b_i,e_i$ are all base-$10$ digits
  2. $b_i,e_i$ are all base-$10$ numbers
  3. removing the base-$10$ constraints from the above

Note: this is a small subset of Printer's Errors (OEIS, WM) because deleting $e_i=1$ or having leading zeros is not allowed here. This generalises this question.

Bill Dubuque
  • 282,220
  • The answer to the related question mentioned by FishDrowned proves that 2592 is the only solution with 4 digits. I'm very curious about this generalization – lesath82 Feb 26 '25 at 10:31
  • 2
    The examples on WA include $3^4 425 = 34425$ which would instead have to be $34412151$. This question is separate from Printer's Error.

    However, it does include a solution to the third "order of interest":

    – FishDrowned Feb 26 '25 at 10:45
  • 1
    A related web page from WolframAlpha: https://mathworld.wolfram.com/PrintersErrors.html – Integreek Mar 01 '25 at 09:31

2 Answers2

13

According to Alex Bellos in his book Alex's Adventures in Numberland (p. 261) another example in base $10$ is $$24547284284866560000000000=2^45^47^28^42^84^86^65^60^00^00^00^00^0$$ discovered by Neil Sloane. (Here, using a multiplicative or set-theoretical definition of exponentiation with nonnegative-integer powers, $0^0=1$.) By the way, the original example $$2592=2^59^2$$ was discovered by Henry Ernest Dudeney who published it as puzzle 115 in his 1917 book Amusements in Mathematics (p. 261) and rediscovered by Raphael Robinson who proposed it as problem E69 in the American Mathematical Monthly, vol. 40, December 1933.

Thanks to Bill Dubuque for a link to Sloane's paper "Eight Hateful Sequences" where he writes

. . . and we conjecture that there are no others. Certainly there are no other fixed points below $10^{100}$.

and also for a link to the associated OEIS sequence A135385: Fixed points of the map $\,m\to{\rm powertrain}(m)$ (see A133500 for definition).

Oscar Lanzi
  • 48,208
bof
  • 82,298
  • 2
    Correction Sloane conjectured that those are the only nontrivial fixed points of Conway's powertrain map (i.e. case $1$ in OP: single digit $b_i,e_i$), see Eight Hateful Sequences, 2008 and OEIS A135385. $\ \ $ – Bill Dubuque Feb 27 '25 at 12:18
  • 2
    "Recall that $0^0=1$." I recall that different definitions of exponentiation become nonequivalent at $0^0$, and some give $1$ while others give an undefined result. Here the exponent domain is forced to nonnegative integers, which favors definitions that do give $0^0=1$. – Oscar Lanzi Feb 27 '25 at 21:59
  • 1
    It's nice to know that there is an existing sequence almost equivalent to (and slightly more general than) bullet 1 in my question. Thank you for pointing out these and giving sources. – Benjamin Wang Feb 28 '25 at 21:48
  • 2
    @Ben I found the Sloane and OEIS links via obvious searches, e.g. this one., which shows the $2592$ case was already discussed here way back in 2012 and 2013, with links to Mathworld and Dudeney. See also bof's 2017 answer and others in said search result (I did not peruse the entire list). It is always best to perform such background searches before posting a question. – Bill Dubuque Mar 01 '25 at 02:02
4

We're trying to solve this equation $$ (b_{0}^{e_o}b_{1}^{e_1}\dots b_{n}^{e_n})_k = b_{0} + ke_{0} + k^{2}b_{1} + k^{3}e_{1} + \dots + k^{2n - 1}b_{n} +k^{2n}e_{n}, $$

where $k$ denotes the numerical base.

Here's the program I wrote in C (I first wrote it in python but realized that C would be much better in terms of compile time)

#include <stdio.h>
#include <math.h>
#include <stdlib.h>

// Function to compute the product of digit pairs long long product_digits(int digits, int num_digits) { long long total = 1; for (int i = 0; i < num_digits; i += 2) { int x = digits[i]; int y = digits[i + 1]; total = (long long)pow(x, y); } return total; }

// Function to compute base conversion long long base_to_ten(int digits, int num_digits, int base) { long long sum = 0; for (int i = 0; i < num_digits; i++) { sum += digits[num_digits - 1 - i] (long long)pow(base, i); } return sum; }

// Recursive function to generate digit combinations and check the equation void generate_combinations(int base, int num_digits, int *digits, int index) { if (index == num_digits) { long long lhs = product_digits(digits, num_digits); long long rhs = base_to_ten(digits, num_digits, base);

    if (lhs == rhs) {
        printf(&quot;Solution found (base %d): &quot;, base);
        for (int i = 0; i &lt; num_digits; i++) {
            printf(&quot;%d &quot;, digits[i]);
        }
        printf(&quot;= %lld\n&quot;, lhs);
    }
    return;
}

// Generate all possible digits (1 to base-1)
for (int i = 1; i &lt; base; i++) {
    digits[index] = i;
    generate_combinations(base, num_digits, digits, index + 1);
}

}

// Function to start checking equations for a given base and digit count void check_equation(int base, int num_digits) { int digits = (int )malloc(num_digits * sizeof(int)); if (!digits) { fprintf(stderr, "Memory allocation failed\n"); return; } generate_combinations(base, num_digits, digits, 0); free(digits); }

int main() { int num_digits = 8; // Adjust as needed int base = 10;

// Iterate over even digit counts
for (int num_digit = 2; num_digit &lt;= num_digits; num_digit += 2) {
    check_equation(base, num_digit);
}

return 0;

}

Interest $1$

Checking up to $n = 4$ (8 digits) for $k = 10$ yields only $2^{5}9^{2}$ as an answer.

I'd like to note that for larger $n$'s, the product (LHS) will result in a number with a number of digits $\geq2n$, because of this I believe it's possible to prove that there exist finite solutions for base $10$. Particularly, some minimum value, say, $k =\underbrace{2^{4}2^{4}\dots} _\text{$n$ times}$, such that the number of digits in $k$ is greater than $2n$. I'm not exactly sure how to prove this, so I'll leave it as a conjecture for now.

So, to answer the question, not likely. It's hard to compute at the moment for higher values of $n$.

Interest $2$

Checking up to $n = 6$, $k = 10$, and increasing the possible digits to $[1, 100]$, I couldn't find a solution. And this makes sense. Take $2^{10} = 1024$, for example, this already attributes $4$ digits. But we can sort of "cheat the system". Here's the method: Since $1^{n}$ for $n \geq 1$ is $1$, we just need to find a number that starts with $1$, has some numbers $n$ between, then ends with $xy$ so that we can write it as $1^{n}x^{y}$.

I searched up to $x, y \in [1, 1000]$ and found: $$ 1^{844674407370955}16^{16} = 18446744073709551616, $$ pretty neat. We can do something similar for four digits and an interval of $[1, 100]$ to obtain the following: $$ 1^{A}6^{88}8^{96} = \overline{A688896} $$ where $A=$ 49261837930424181019782306980112000220529917757320182269028725210958882068540501080877333683282833234761504686305466752126923092015623395616828168020 and $$ 1^{B}81^{1}7^{83} = \overline{B811783} $$ where $B=$ 126217677935462499355285958106810248209716299799868722720464961928 and $$ 1^{C}4^{42}6^{496} = \overline{C4426496} $$ where $C=$ 77639616149170183337496905014553824675277932849536369386649820645372361449190489637626678015518085702298093149509480620671632214451835574765888159311459082678179264350547076482449615727891548893204840176012914249489156305948368663946858037186326014142014080031067989541411905654690680896047802285331626672703551588831423345725677690111832053407287254594706993093724057899428747007045660445690532017662674 and $$ 1^{D}8^{7} 2^{192} = \overline{D872192} $$ where $D=$ 3164036458569648337239753460458804039861886925068638906788

Given this trick, Interest $2$ has a lot of intriguing solutions.

The program for this is separate:

def check_power_condition(x, y):
    power_str = str(x ** y)
    xy_str = str(x) + str(y)
# Check if the first digit is '1'
if power_str[0] == '1' and power_str.endswith(xy_str):
    return True
return False

for x in range(1, 1000): for y in range(1, 1000): if check_power_condition(x, y): print(f"{x}^{y} meets the condition!!!!!!!!!!!!!!!")

Interest $3$

Compared to Interest $3$, there does appear to be an abundance of solutions across $k$ but fewer across $n$. Here are some solutions I found given different $n$'s (I'll be using $[d]$ for non-base$10$ numbers):

$n = 1$: $$ \begin{align} 2^{4} = 24_{6} \quad & 2^{8} = 28_{124}\\ 3^{3} = 33_{8} \quad & 3^{6} = 36_{241}\\ 2^{6} = 26_{29} \quad & 2^{[10]} = 2[10]_{507}\\ 4^{4} = 44_{63} \quad & 5^{5} = 55_{624}\\ \end{align} $$

$n = 2$: $$ \begin{align} 2^{5}9^{2} = 2592 &\\ 2^{9}9^{2} = 2992_{26} &\\ 1^{[21]}[23]^{4} = 1[21][23]4_{59}& \end{align} $$

$n = 3$: No small base solutions for $n =3$ and searching any higher requires much more computation time. EDIT More solutions found: $$ 5^{[12]}[23]^{[20]}[10]^{[21]} = 5[12][23][20][10][21]_{34} \\ [27]^{[17]}[36]^{6}[37]^{[26]} = [27][17][36]6[37][26]_{38} $$

$n = 4$: $$ 5^{1}2^{3}2^{8}7^{4} = 51232874_{9} $$

As you can see, the depth of which we can search for base-$k$ gets smaller for larger $n$'s.

Final

For the specific case of base 10 numbers, $2^{5}9^{2}$ seems to be incredibly rare.

I hope this helped.

FishDrowned
  • 1,019
  • 1
    Wow, it is nice to know that for base 10 digits it might be the only one. It does seem difficult to prove. And nice trick to get the second part working. Maybe there will be too many solutions there then. Also for part 3, perhaps you meant n=1,2,3,4 rather than 0,1,2,3. I do wonder if there is a provable upper bound when it comes to digits. – Benjamin Wang Feb 27 '25 at 10:30