This supplies the irrational case of the ordinary-prime omission theorem. It also gives a count for pairs of integers on the two sides of an irrational dilation.
Theorem. Suppose is a set of relative density zero among the primes. Let be the positive integers whose prime factors lie outside , and put . For a fixed irrational and the denominators of its continued-fraction convergents, along we have
In particular, the pair with violates unit separation if a multiplicative semigroup contains and .
Lattice counts with congruences
At use the area-preserving map
The region , maps to a square of side , with the indicated half-open boundary. For convergents and of consecutive indices, the vectors and give a basis of . Both vectors have coordinates bounded by 1 in absolute value. The bound on the second coordinate uses and . Their determinant has absolute value 1. Thus the shortest length in is bounded below by an absolute constant. In particular the number of nonzero lattice vectors in a ball of radius is whenever the ball contains a nonzero vector.
We will also use for the sublattice with , and for the one with . Either has covolume . For any one of these lattices , write for the shortest length of a nonzero vector. A translate of meets the square in
points, by the two-dimensional lattice-counting estimate in terms of the first successive minimum. The estimate applies to the translated region as well. Adding or dropping a side of length changes the bound by .
For use integer coordinates on the lattice, and for use . At each sifting prime , erase one coordinate’s zero residue or both coordinates’ zero residues. For a squarefree product of these primes, the simultaneous erased set occupies residue pairs modulo , by the Chinese remainder theorem. Each residue pair corresponds to one translate of . The count for these congruences is
The local retained densities are or , so Mertens’ theorem gives the uniform dimension-two product bound. With sieve level and range , the fundamental lemma bounds the sifted total above and below by
where is the product of the local retained densities. We will use the displayed error as a two-sided bound on the difference from the main term.
Small and medium omitted primes
The prime number theorem and partial summation give and for each fixed . Fix an error tolerance . On use level and range with sufficiently small depending on . Both coordinates are sifted by . The count passing this first sieve is along .
Fix . A failure at an omitted prime in the range lies on or . In either case we may use upper sieve level and a range with a sufficiently small fixed . Both original coordinates can be sifted by the primes in up to , because is invertible at each of those primes. The main terms summed over are at most a constant times
For the remainders, consider the primes in from this range. For , each prime satisfying divides the nonzero coordinate of some lattice vector in the radius- ball. The coordinate is nonzero because every nonzero vector with has length at least , while . Moreover . Each such has at most prime divisors above . There are vectors, so there are primes satisfying this short-vector condition. The same bound holds for , since implies length at least for a nonzero vector and otherwise.
Dyadically summing the inverse lengths up to , then using primes for the lengths above , gives
with the sum restricted to the medium range. Thus the total sieve remainder for medium primes is bounded by a constant times
Large omitted primes
Suppose for a prime . Then , and every survivor of the first sieve has . On , sieve the coordinate by all primes up to , using sufficiently small for level . This upper bounds the requirement that is prime. Sift by up to the same range. The retained product is
The lattice count gives the upper bound plus the remainder. If it is that has a large prime factor, use with and exchange the two coordinates. Both lattice minima are bounded below by that of . The sum of the sieve remainders for in this range is , which is .
For , Euler’s product gives . Since for each fixed , the large primes altogether remove at most . Choose small enough after fixing and that this contribution is . The upper and lower estimates then differ from the claimed main term by . Taking arbitrarily small proves the theorem.