The three-prime floor-function set formula

From papers

Let G(x)G(x) denote the number of primes in the set {x/n:1nx}\left\{\left\lfloor x/n\right\rfloor:1\le n\le x\right\}. Suppose that x=p1p2p3x=p_1p_2p_3, where p1,p2,p3p_1,p_2,p_3 are primes satisfying 2<p1<p2<p32<p_1<p_2<p_3. The three-prime floor-function set formula. Then

G(x)={G(x1)if p1p2>p3,G(x1)+1if p1p2<p3.G(x)=\begin{cases} G(x-1)&\text{if }p_1p_2>p_3,\\ G(x-1)+1&\text{if }p_1p_2<p_3. \end{cases}

This statement is presented as a generalisation suggested by a limited computational investigation; unlike the preceding prime and semiprime cases, no proof or resolution is supplied in the source.

Progress summary

Open

The proposed three-prime rule remains an unproved conjecture, with no public counterexample or verification found.

M. A. Heyman proposed the statement as Conjecture 4 in 2021, based on a limited Maple investigation. It predicts the change in G(x)G(x) when x=p1p2p3x=p_1p_2p_3 according to whether p3p_3 is larger or smaller than p1p2p_1p_2.

Known results

  • For prime p3p\ne3, G(p)=G(p1)+1G(p)=G(p-1)+1 (Adams–Watters attribution; proved by Heyman).
  • For x=pqx=pq with odd primes p,qp,q, G(x)=G(x1)+1G(x)=G(x-1)+1 (Heyman).

Current status (as of August 2026): The three-prime formula remains open; no proof, counterexample, or independent verification is recorded in the retrieved sources.

Sources
Sources & referencesView supporting material

Primary source

Randell Heyman, “Primes in floor function sets”, arXiv:2111.00408 (2021).

Solutions 1

Proof

In fact, a stronger statement holds for every odd integer N5N\ge5:

G(N)G(N1)=#{q prime:qN, q2N},G(N)-G(N-1) = \#\{q\text{ prime}:q\mid N,\ q^2\ge N\},

where G(M)G(M) counts the distinct primes in

S(M)={Mj:1jM}.S(M)=\left\{\left\lfloor\frac Mj\right\rfloor: 1\le j\le M\right\}.

For every integer q2q\ge2, the number of indices jj for which M/j=q\lfloor M/j\rfloor=q is

DM(q)=MqMq+1.D_M(q) = \left\lfloor\frac Mq\right\rfloor - \left\lfloor\frac M{q+1}\right\rfloor.

Thus qS(M)q\in S(M) if and only if DM(q)>0D_M(q)>0.

If NN and qq are both odd, then q+1q+1 is even and cannot divide NN. Consequently

DN(q)DN1(q)=1{qN}.D_N(q)-D_{N-1}(q)=\mathbf1_{\{q\mid N\}}.

Therefore no odd prime disappears when passing from N1N-1 to NN, and a new odd prime must divide NN.

Write N=qmN=qm for such a prime divisor. Then

DN1(q)=m1qmq+1=mq+11.D_{N-1}(q) = m-1-\left\lfloor\frac{qm}{q+1}\right\rfloor = \left\lceil\frac m{q+1}\right\rceil-1.

Hence qq appears for the first time precisely when

mq+1.m\le q+1.

Since mm and qq are odd, this is equivalent to mqm\le q, or

q2N.q^2\ge N.

Finally, 2S(M)2\in S(M) for every M4M\ge4, so the prime 22 contributes no change when N5N\ge5. This proves the general formula.

Now suppose

N=p1p2p3,2<p1<p2<p3.N=p_1p_2p_3,\qquad 2<p_1<p_2<p_3.

Neither p1p_1 nor p2p_2 can satisfy q2Nq^2\ge N. The only possible new prime is p3p_3, and

p32Np3>p1p2.p_3^2\ge N \quad\Longleftrightarrow\quad p_3>p_1p_2.

Equality is impossible because p3p_3 is prime. Therefore

G(N)G(N1)={1,p3>p1p2,0,p3<p1p2,G(N)-G(N-1) = \begin{cases} 1,&p_3>p_1p_2,\\ 0,&p_3<p_1p_2, \end{cases}

which proves both asserted cases.

0 endorsements
Shivam Patel ·