Why is 5² = 25 more than just a multiplication?

Have you ever noticed that 5² = 25 seems to tell a story? The result 25 is literally the exponent “2” glued to the left of the base “5”. This phenomenon, called an *exponential‑prefix polymorphic number*, is more than a coincidence. In practice, we look for every pair (b, e) in a given base b that satisfies the equation b^e = concat(e, b).

The problem becomes a Diophantine equation: find positive integers b, e such that b^e = e·b^{k}+b, where k is the number of digits of e in the chosen base. Researchers have tackled it with several tools: Newton‑Raphson to approach real solutions, the arithmetic‑geometric mean inequality to bound possible values, and even Pell’s equation for special cases.

Among the most famous solutions are (5, 2) and (6, 2) in base 10, but the search doesn’t stop there. Changing the base reveals new surprising pairs, like (3, 3) in base 4. Moreover, Fermat’s little theorem helps prove that certain combinations can never work because the result cannot be a multiple of a specific prime.

The outcome is an almost complete classification of all numbers that display this “exponential prefix”. It’s not just a party trick: it bridges elementary arithmetic and advanced number theory, showing how mathematics hides surprises behind seemingly simple operations.

The core equation b^e = e·b^{k}+b can be rewritten as b^{e-1}=e·b^{k-1}+1. When e is small compared to b, the term e·b^{k-1} dominates, suggesting that b must grow quickly to keep the equality. To analyze solutions, we introduce the variable x = log_b(e), turning the equation into b^{e-1}=b^{k·x}+1. Applying Newton‑Raphson to f(b)=b^{e-1}-e·b^{k-1}-1 yields iterations b_{n+1}=b_n-\frac{f(b_n)}{f'(b_n)} that converge rapidly to real roots, providing initial estimates for integer solutions.

The arithmetic‑geometric mean inequality (AM‑GM) imposes b^{e-1} \ge (e·b^{k-1})^{\frac{e-1}{k}}. From this we derive a growth condition: e \le b^{(e-1)(1-\frac{1}{k})}. This bound eliminates many impossible (b, e) pairs, shrinking the search space.

For cases where k=1 (e has a single digit), the equation simplifies to b^{e-1}=e·b+1. Here Pell’s equation x^2-Db^2=1, with D linked to e, comes into play. Fundamental Pell solutions generate infinite families of (b, e) that satisfy the condition, e.g., (5,2), (29,2), (169,2), etc.

Fermat’s little theorem (a^{p-1}≡1 mod p) helps prove the absence of solutions when p is prime and divides b. If p|b, then b^e ≡0 (mod p), but concat(e,b) ≡ e·b^{k}+b ≡ b (mod p), which cannot be zero unless p|e, contradicting the choice of e.

Finally, the complete classification combines these approaches: enumerate all k=1 solutions via Pell, discard impossible cases with AM‑GM and Fermat, and verify the remaining candidates with Newton‑Raphson. The outcome is a finite list of parametric families plus a few isolated solutions, describing all known exponential‑prefix polymorphic numbers.