Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Erdős and Obláth restrict to coprime x,yx,y and prove three theorems. Their Satz 1: apart from 2!=22!=2, no factorial is a sum or a difference of the ppth powers of two coprime numbers when p≥3p\ge3 is not a power of 22. Their Satz 2: a difference of the eighth powers of two coprime integers is never a factorial, which excludes differences for every p=2αp=2^\alpha with α≥3\alpha\ge3. Their Satz 3, proved with the prime number theorem for the progressions 4k+14k+1 and 4k+34k+3: for sufficiently large nn, n!n! is not a difference of the fourth powers of two coprime integers; no threshold is given. For sums, the introduction reduces the equation to prime exponents and shows that n!n! is a sum of two squares for no n≥7n\ge7, since some prime q≡3(mod4)q\equiv3\pmod4 with n/2<q≤nn/2<q\le n divides n!n! exactly once; 6!=122+2426!=12^2+24^2. The exponents not covered by Satz 1 are the powers of 22, so for sums with k>2k>2 even this reduction applies, coprime or not, once the small cases are checked: 720720 is not a sum of two fourth powers.

Covers. No solution of n!=xk±ykn!=x^k\pm y^k in Problem 399 with gcd⁡(x,y)=1\gcd(x,y)=1, xy>1xy>1 and k>2k>2, except possibly a difference n!=x4−y4n!=x^4-y^4 with nn below the unspecified threshold of Satz 3. The site records the result as the coprime case with k≠4k\ne4. Nothing here constrains the case gcd⁡(x,y)>1\gcd(x,y)>1 with an odd exponent or a difference, where the solution 10!=484−36410!=48^4-36^4 lies.

Depends on. No page of this wiki.

Acceptance. Refereed: P. Erdős and R. Obláth, Über diophantische Gleichungen der Form n!=xp±ypn!=x^p\pm y^p und n!±m!=xpn!\pm m!=x^p, Acta Litt. Sci. Szeged 8 (1937), 241–255; library card erdos_1937_uber_diophantische_gleichungen_der_form_und. The site's curator credits the coprime theorem with k≠4k\ne4 to this paper in the commentary, but the site's label settles the problem through Barfield's counterexample and not through this result, so reviewed is not listed. The formal-conjectures file states the site's version of the theorem as erdos_399.variants.erdos_oblath with sorry, which is not a formalization.