The full conjecture remains open. This note extends the dominant-largest-prime argument to a family where the second-largest prime can be smaller than its cofactor. Work is tracked in issue #47.
Theorem. Let consist of with prime, , and . For every set of natural density zero,
In particular, implies when , so the displayed primes are the largest three and are coprime to . The result adds a family of density about to the previously covered inputs.
Reductions
Work on . We can impose fixed positive margins on the bounds for and after discarding an arbitrarily small upper proportion of inputs. To see this, the number of possible for each is . If , sum the reciprocal condition over the two primes. For it costs at most . The sum of these weights against for is by integration. The case costs .
Requiring removes only inputs by the reciprocal sum over between constant multiples of and . For , the interval has prime reciprocal sum uniformly in , by Mertens' theorem. The reciprocal sum over is now bounded in terms of . Thus we can require as well. Finally, for a sufficiently small fixed , we can require that have a prime factor of exponent one. The inputs failing this because is -smooth cost by the average logarithm of the -smooth part. The inputs divisible by the square of any prime at least cost .
Apply the small-prime and normality reductions in the main note. Take with fixed smaller than the power lower bounds on . For the common -smooth part is , and we require and for each prime and . Also require , , , and the reciprocal-divisor condition for . In particular . Davenport's theorem and the iterate theorem from the original paper give the normality bounds. Exceptional sets in transfer by summing over and the reciprocal prime sums over , which are bounded since have fixed power lower bounds.
Partition the allowed logarithmic exponents of into finitely many boxes of width at most , centered at . The margins give and , uniformly away from equality. We will take sufficiently small in terms of the margins and . It suffices to prove a second moment of size within each box. No collisions between different boxes are needed for the preimage bound.
Small and intermediate collision moduli
For distinct cofactors in a collision, put . The two-prime sieve for in the main note applies since up to constants. For each such pair of cofactors it gives at most
prime pairs. In a collision and . We use the quadratic congruence in from the main note for , with any fixed between half a power lower bound for and a power lower bound for . That argument only requires these prime power bounds and the stated normality conditions. It gives after summing over .
Split further into dyadic intervals with lower endpoints , and use primes on these endpoints for . For , where will be chosen small, fix . The congruences determine and modulo since the coefficients and are units. The permitted lie in at most residue classes modulo a modulus at least , by the quadratic root cover in the main note. For each , the prime lies in a single class modulo . Even ignoring primality, the product of the reciprocal sums over , multiplied by , is at most
This saves a power of since , , and has a fixed power lower bound. The number of divisors of and are . All other sums, including the reciprocal sums over and the sieve singular factor, cost only . This range gives .
Exact relations for the largest collision moduli
It remains to bound . In this range replace in the sieve by . The identity gives
Comparing this congruence with the corresponding congruence for yields . Both sides are , whereas . Since , take small enough that the congruence is an equality. Define
The equality now says . For fixed , it determines . Together with this gives at most choices of : the product is strictly increasing in the smaller prime when their sum is fixed, and lies in an interval of length . The inverse of modulo exists because .
When , the sums are equal and for distinct . Thus . After multiplying by and summing , the resulting bound is for each . This gives such collisions after the prime-pair sieve.
When , put
This integer divides . For each , the sum over choices of of the sieve weight is at most a constant times
Here the first term follows by summing over the divisors of , and the second follows from the term in the number of choices of . The retained bounds the first term. Since and has exponent , the sum of over the eligible for each in their dyadic boxes is , by counting in one residue class. Including the outer prime-pair sieve and summing the dyadic boxes costs at most
We show that the displayed reciprocal sum saves a fixed power of .
Rational ratios of two cofactors
Write and . The gcd in the definition of is because . Also , by the retained gcd condition. Take and later take sufficiently small in terms of .
First suppose . For each fixed , there are possible choices of the divisors and such that , , and . They are pairwise coprime. The pair is a primitive integer point in the lattice specified by these two divisibilities and by . The lattice has determinant , by the Chinese remainder theorem. In a centered box of side length a planar lattice of determinant has primitive integer points. Indeed, if the points lie on one line through the origin, there are at most two. Otherwise a shortest nonzero lattice vector has length and belongs to a basis. The lattice lies on parallel lines in that direction at spacing , since there is a point off the line through the origin. Counting lines and then points on each gives .
Here , and with . Consequently the inner sum in this range is at most , uniformly in . The outer reciprocal sum over costs at most .
Now suppose . The numerator and denominator of each reduced ratio and are . Write where is a prime coprime to , and let . For each fixed , the identity gives
This polynomial in has discriminant . If , we have , so by monotonicity. Both fractions are reduced, giving , which has already been treated.
Otherwise the square condition on the right gives at most choices of for fixed . For completeness, clear denominators by multiplying both sides by a square, giving with integer coefficients of polynomial size in , , and nonzero discriminant . Completing the square gives . If is a square, factor and use the divisor bound. Otherwise, remove the square part of and work in the ring of integers of the real quadratic field with the resulting squarefree radicand. An algebraic integer of norm generates an ideal dividing , with only possibilities by unique ideal factorization. Within one ideal, generators differ by units. Their absolute values at one real embedding lie between and because both embeddings are polynomially bounded and the product is the nonzero norm. A unit of absolute value greater than one has absolute value at least the golden ratio, by its integral trace and norm . Thus only generators in the given range occur per ideal. This is the usual quadratic norm argument for generalized Pell equations, with the bound uniform since the radicand is of polynomial size.
There are choices of and at most choices for the two ratios. Each and the ratios determine uniquely, since is increasing in positive with fixed. Thus the sum over these pairs, using and , is at most . This completes the required power saving in both ranges.
The diagonal and small-modulus terms give , and the intermediate and large-modulus terms give , even after summing over . The second moment in each box is therefore . Cauchy–Schwarz proves the preimage statement after the discarded upper proportion is taken arbitrarily small.
Added density
The family overlaps whenever . The added inputs satisfy and , so they are also disjoint from the covered family below the square root. For the density on use for the logarithmic exponents of with respect to . The region is
Away from the boundary, the prime number theorem gives possible for each . Partial summation in the integer and the two primes gives the density
The bounds in the reduction control the strip near ; away from that strip, the integrand is bounded and the other boundary strips cost a vanishing proportion. Summing the dyadic intervals gives the natural density. Taken together, the proved families have natural density approximately .