Wiki
Wiki

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

Updated


Source. The large-prime step in the Notes of proof claim 133. The cross-difference and reduced-fraction uniqueness arguments are supplied explicitly here.

For M≥2M\ge2, let

RM={(a,b):1≤a,b≤M, gcd⁡(a,b)=1}.\mathcal R_M=\{(a,b):1\le a,b\le M,\ \gcd(a,b)=1\}.

Statement. Let q>M2q>M^2 be prime, and suppose that the residues 1,…,M1,\ldots,M lie in a subgroup H≤Fq∗H\le\mathbf F_q^*. Then

ι:RM⟶H,ι(a,b)=ab−1(modq),\iota:\mathcal R_M\longrightarrow H,\qquad \iota(a,b)=ab^{-1}\pmod q,

is well defined and injective. Consequently ∣H∣≥C(M)|H|\ge C(M), where C(M)=∣RM∣C(M)=|\mathcal R_M|.

Complete proof. Since b≤M<qb\le M<q, its residue is nonzero and has an inverse. Both aa and bb belong to HH, so subgroup closure gives ab−1∈Hab^{-1}\in H.

Suppose ab−1=cd−1ab^{-1}=cd^{-1} in Fq\mathbf F_q for two members (a,b)(a,b) and (c,d)(c,d) of RM\mathcal R_M. Then q∣ad−bcq\mid ad-bc. Each product lies between 11 and M2M^2, so

∣ad−bc∣≤M2−1<q.|ad-bc|\le M^2-1<q.

The divisibility therefore forces ad=bcad=bc over the integers. Because gcd⁡(a,b)=1\gcd(a,b)=1, the equality implies a∣ca\mid c; reversing the two pairs gives c∣ac\mid a. Positivity yields a=ca=c, and then b=db=d. Thus ι\iota is injective.

Dependencies. Elementary modular arithmetic and uniqueness of a reduced positive fraction.

Bears on. The large-prime case in the partial threshold theorem.