betaFeedback

WorksMathematics

Balanced additive complements with bounded representation functions

Abstract

We construct infinite sets A={a1<a2<⋯ }A=\{a_1<a_2<\cdots\} and B={b1<b2<⋯ }B=\{b_1<b_2<\cdots\} of positive integers such that ∣ai−bi∣≤5bi|a_i-b_i|\leq5\sqrt{b_i} for every ii, every integer at least 1638416384 belongs to A+BA+B, and the number of ordered representations of any integer as a+ba+b with a∈Aa\in A and b∈Bb\in B is at most 4700547005. In particular, ai/bi→1a_i/b_i\to1, giving a negative answer to a conjecture of Erdős and Sárközy. The construction uses two nested prime-power scales, a uniform residue bound, and a local cardinality matching.

Introduction

For sets A,B⊆N={1,2,…}A, B \subseteq\mathbb{N} = \{1,2,\ldots\}, define their sumset and ordered representation function by

A+B={a+b:a∈A, b∈B},rA,B(n)=#{(a,b)∈A×B:a+b=n}.A+B = \{a+b : a \in A,\ b \in B\}, \qquad r_{A,B}(n) = \#\{(a,b) \in A \times B : a+b=n\}.

Erdős and Sárközy asked whether, for infinite A,BA, B and some N≥1N \ge1,

aibi⟶1,A+B⊇[N,∞)∩Z⟹lim sup⁡n→∞rA,B(n)=∞,(1)\frac{a_i}{b_i} \longrightarrow1,\qquad A+B \supseteq[N,\infty) \cap\mathbb{Z} \quad\Longrightarrow\quad\limsup_{n\to\infty} r_{A,B}(n)=\infty, \tag*{(1)}

where aia_i and bib_i are the increasing enumerations of AA and BB. The question appears as Problem 1.17 in the 1999 collection [4] and as Problem 1145 in [1].

Theorem 1.1. There are infinite sets A={a1<a2<⋯ }A = \{a_1 < a_2 < \cdots\} and B={b1<b2<⋯ }B = \{b_1 < b_2 < \cdots\} of positive integers satisfying

∣ai−bi∣≤5bi(i≥1),(2)|a_i-b_i| \le5\sqrt{b_i} \qquad(i \ge1), \tag*{(2)}
[16384,∞)∩Z⊆A+B,(3)[16384,\infty) \cap\mathbb{Z} \subseteq A+B, \tag*{(3)}
rA,B(n)≤47005(n∈Z).(4)r_{A,B}(n) \le47005 \qquad(n \in\mathbb{Z}). \tag*{(4)}

Consequently the implication in (1) is false.

The construction proceeds through finite blocks whose elements have size comparable to the square of their scale. Geometric block constructions and matched local configurations for this problem were considered in the public notes [2, 3]. Here we maintain a residue-class bound for the full union of preceding blocks. Powers of two control one component and powers of three control the other. The resulting invariant bounds all mixed representations at a fixed maximal scale simultaneously.

Scales and residue bounds

Fix

H=260,L=H+1=261,M=3L=783,K=2M+1=1567.(5)H = 260,\qquad L = H + 1 = 261,\qquad M = 3L = 783,\qquad K = 2M + 1 = 1567. \tag*{(5)}

For k≥3k \ge3, put qk=3kq_k = 3^k and let pkp_k be the least power of two greater than qkq_k. Since distinct positive powers of two and three are unequal,

qk<pk<2qk,qk+1=3qk,pk+1=dkpk,dk∈{2,4}.(6)q_k < p_k < 2q_k,\qquad q_{k+1} = 3q_k,\qquad p_{k+1} = d_kp_k,\qquad d_k \in\{2,4\}. \tag*{(6)}

Indeed, pk<3qk<4pkp_k < 3q_k < 4p_k, so the next selected power of two is either 2pk2p_k or 4pk4p_k. In particular, p3=32p_3 = 32, q3=27q_3 = 27, and gcd⁡(pi,qj)=1\gcd(p_i,q_j) = 1 for all i,j≥3i,j \ge3. Each of the two modulus sequences is nested under divisibility.

For a finite set S⊆ZS \subseteq\mathbb{Z} and a positive integer mm, write

ρm(S;r)=#{x∈S:x≡r(modm)}.\rho_m(S;r) = \#\{x \in S : x \equiv r \pmod{m}\}.

At stage kk, the preceding blocks will have unions

PkA=⋃3≤j<kAj,PkB=⋃3≤j<kBj.P_k^A = \bigcup_{3 \le j < k} A_j,\qquad P_k^B = \bigcup_{3 \le j < k} B_j.

The induction maintains

ρqk(PkA;r)≤K,ρpk(PkB;r)≤K(r∈Z).(7)\rho_{q_k}(P_k^A;r) \le K,\qquad\rho_{p_k}(P_k^B;r) \le K\qquad(r \in\mathbb{Z}). \tag*{(7)}

Lemma 2.1. Let m,m′m,m' be positive integers with m∣m′m \mid m', and suppose ρm(S;r)≤K\rho_m(S;r) \le K for every rr. Call a residue class modulo m′m' safe if it contains at most K−M=784K-M=784 elements of SS. Each residue class modulo mm contains at most one unsafe class modulo m′m'. Moreover, if a finite set CC uses only safe classes and ρm′(C;r)≤M\rho_{m'}(C;r) \le M for every rr, then ρm′(S∪C;r)≤K\rho_{m'}(S \cup C;r) \le K for every rr.

Proof. Two unsafe classes in the same class modulo mm would contain at least

2(K−M+1)=1570>K2(K-M+1) = 1570 > K

elements of SS. In a safe class, the union contains at most (K−M)+M=K(K-M)+M=K elements. An unsafe class receives no element of CC, and its old count is bounded by the count of its parent class modulo mm. □

At stage kk, safety for the AA-component is defined from PkAP_k^A modulo qk+1q_{k+1}. Safety for the BB-component is defined from PkBP_k^B modulo pk+1p_{k+1}. Thus every class modulo qkq_k has at most one unsafe child among its three classes modulo qk+1q_{k+1}, and every class modulo pkp_k has at most one unsafe child among its dkd_k classes modulo pk+1p_{k+1}.

Construction of a block

Assume that the earlier blocks have been constructed, satisfy (7), and are pairwise disjoint within each component. Throughout this section abbreviate p=pkp=p_k and q=qkq=q_k.

Separating the blocks

We choose integers

Uk=p2+αk,0≤αk<p,Vk=p2+βk,0≤βk<q.(8)U_k = p^2 + \alpha_k,\qquad0 \le\alpha_k < p,\qquad V_k = p^2 + \beta_k,\qquad0 \le\beta_k < q. \tag*{(8)}

Write U=UkU=U_k and V=VkV=V_k. The candidate sets are

Ak={U+pX+sq:0≤X<Lq, 0≤s<3},(9)A_k = \{U + pX + sq : 0 \le X < Lq,\ 0 \le s < 3\}, \tag*{(9)}
Bk={V+qY+up:0≤Y<Lp, 0≤u<2}.(10)B_k = \{V + qY + up : 0 \le Y < Lp,\ 0 \le u < 2\}. \tag*{(10)}

All parameters in these formulas are integers.

Lemma 3.1. The anchors in (8) can be chosen so that Ak∩Aj=∅A_k \cap A_j = \varnothing and Bk∩Bj=∅B_k \cap B_j = \varnothing for every 3≤j<k3 \le j < k.

Proof. For each earlier jj, impose

U+sq≢Uj+s′qj(modpj)(0≤s,s′<3).(11)U + sq \not\equiv U_j + s'q_j \pmod{p_j} \qquad(0 \le s,s' < 3). \tag*{(11)}
V+up≢Vj+u′pj(modqj)(0≤u,u′<2).(12)V + up \not\equiv V_j + u'p_j \pmod{q_j} \qquad(0 \le u,u' < 2). \tag*{(12)}

Each condition in (11) excludes one residue class for αk\alpha_k modulo pjp_j. Because pj∣pp_j \mid p, this removes exactly a fraction 1/pj1/p_j of its pp possible values. Also,

9∑j=3k−11pj≤932∑h=0∞2−h=916<1.9\sum_{j=3}^{k-1}\frac{1}{p_j} \le\frac{9}{32}\sum_{h=0}^{\infty}2^{-h} = \frac{9}{16} < 1.

The union of the excluded choices therefore has size less than pp. Similarly, the excluded fraction of the qq possible values of βk\beta_k is at most

4∑j=3k−11qj≤4∑j=3∞3−j=29<1.4\sum_{j=3}^{k-1}\frac{1}{q_j} \le4\sum_{j=3}^{\infty}3^{-j} = \frac{2}{9} < 1.

Both anchors exist.

A point of AkA_k is congruent to U+sqU+sq modulo pjp_j, since pj∣pp_j \mid p. Every point of Aj⊆AjA_j \subseteq A_j is congruent to Uj+s′qjU_j+s'q_j for some s′s'. Thus (11) gives the required disjointness on the AA-side. The divisibility qj∣qq_j \mid q and (12) give the same conclusion on the BB-side. □\square

The parameters within each candidate set are unique. Indeed, an equality between two expressions in (9), reduced modulo pp, gives p∣(s−s′)qp \mid(s-s')q. Coprimality and ∣s−s′∣≤2<p|s-s'| \le2 < p imply s=s′s=s', and then X=X′X=X'. On the BB-side, reduction modulo qq gives q∣(u−u′)pq \mid(u-u')p, whence u=u′u=u' and Y=Y′Y=Y'.

Every candidate satisfies

p2≤z≤263p2.(13)p^2 \le z \le263p^2. \tag*{(13)}

For the upper bound in (9), use q<pq < p and p≥32p \ge32 to obtain

U+pX+sq<p2+p+Lpq+2q<(L+1)p2+3p≤(L+2)p2=263p2.U+pX+sq < p^2+p+Lpq+2q < (L+1)p^2+3p \le(L+2)p^2 = 263p^2.

The BB-bound follows from V+qY+up<p2+q+Lpq+p<(L+1)p2+2p≤263p2V+qY+up < p^2+q+Lpq+p < (L+1)p^2+2p \le263p^2.

Selection and matching

Only safe candidates will be retained. This suffices to preserve the next instance of (7). For a fixed ss and a fixed residue modulo qq, exactly LL indices 0≤X<Lq0 \le X < Lq give that residue, because pp is invertible modulo qq. Consequently

ρq(Ak;r)≤3L=M.\rho_q(A_k;r) \le3L=M.

Every class modulo qk+1q_{k+1} is contained in a class modulo qq, so its candidate count is also at most MM. Likewise ρp(Bk;r)≤2L≤M\rho_p(B_k;r) \le2L \le M, which bounds every candidate count modulo pk+1p_{k+1}. Lemma 2.1 now applies to any subsets of the safe candidates.

The choice of those subsets is arranged to give a close matching. For an integer 0≤x<q0 \le x < q, put

Yx={y∈Z:0≤y<p, ⌊qy/p⌋=x},dx=∣Yx∣.Y_x=\{y\in\mathbb{Z}:0\le y<p,\ \lfloor qy/p\rfloor=x\}, \qquad d_x=|Y_x|.

The interval description

pxq≤y<p(x+1)q(14)\frac{px}{q} \le y < \frac{p(x+1)}{q} \tag*{(14)}

shows that

1≤dx≤2.(15)1 \le d_x \le2. \tag*{(15)}

Indeed, the half-open interval in (14) has length p/q∈(1,2)p/q \in(1,2) and lies in [0,p][0,p].

For integers 0≤t≤H0 \le t \le H and 0≤x<q0 \le x < q, form the two stocks

Sx,tA={U+p(x+qt)+sq:0≤s<3},S^A_{x,t}=\{U+p(x+qt)+sq:0\le s<3\},
Sx,tB={V+q(y+pt)+up:y∈Yx, 0≤u<2}.S^B_{x,t}=\{V+q(y+pt)+up:y\in Y_x,\ 0\le u<2\}.

Let Tx,tAT^A_{x,t} and Tx,tBT^B_{x,t} denote their safe subsets. The three AA-points occupy the three different children of one class modulo qq. Hence ∣Tx,tA∣≥2|T^A_{x,t}| \ge2. For each y∈Yxy \in Y_x, the two BB-points occupy different children of one class modulo pp, because pk+1≥2pp^{k+1} \ge2p. At least one is safe. Parameter uniqueness gives

∣Tx,tA∣≥2≥dx,∣Tx,tB∣≥dx.(16)|T^A_{x,t}| \ge2 \ge d_x,\qquad|T^B_{x,t}| \ge d_x. \tag*{(16)}

The mandatory points in each stock are its safe unshifted points, obtained by setting s=0s=0 or u=0u=0. Their counts, say mAm_A and mBm_B, satisfy mA≤1m_A \le1 and mB≤dxm_B \le d_x. Set r=max⁡(mA,mB)r=\max(m_A,m_B). By (16), on each side we can choose exactly rr safe points containing all its mandatory points. Match the two chosen subsets by any bijection.

Let AkA_k and BkB_k be the unions of these choices over all (x,t)(x,t). The uniqueness of X=x+qtX=x+qt and Y=y+ptY=y+pt, together with parameter uniqueness, makes the stocks pairwise disjoint on each side. We have therefore constructed finite sets and a bijection

fk:Ak⟶Bk.(17)f_k:A_k\longrightarrow B_k. \tag*{(17)}

Every safe unshifted point in (9) or (10) belongs to the corresponding selected set.

Proposition 3.2. The sets Ak,BkA_k,B_k are disjoint from the preceding blocks in their respective components. They satisfy (13), and PkA∪AkP^A_k\cup A_k, PkB∪BkP^B_k\cup B_k satisfy (7) with kk replaced by k+1k+1. Moreover,

∣a−fk(a)∣≤5pk(a∈Ak).(18)|a-f_k(a)|\le5p_k\qquad(a\in A_k). \tag*{(18)}

Proof. Disjointness follows from Lemma 3.1, and the range bound holds for all candidates. The residue bounds follow from the safety argument above.

For points in one matched group, the term pqtpqt cancels and their difference is

(U−V)+(px−qy)+sq−up.(U-V)+(px-qy)+sq-up.

We have ∣U−V∣<p|U-V|<p, 0≤qy−px<p0\le qy-px<p, sq≤2q<2psq\le2q<2p, and up≤pup\le p. The triangle inequality proves (18). □

Coverage of each scale interval

Lemma 4.1. Suppose U,VU,V satisfy (8) and nn is an integer with 16p2≤n≤256p216p^2\le n\le256p^2. There are integers 0≤x<q0\le x<q, 0≤y<p0\le y<p, and 12≤R≤50712\le R\le507 such that

n=U+V+px+qy+pqR.(19)n=U+V+px+qy+pqR. \tag*{(19)}

Proof. Coprimality gives unique xx and yy in the stated ranges satisfying

px≡n−U−V(modq),qy≡n−U−V(modp).px\equiv n-U-V\pmod q,\qquad qy\equiv n-U-V\pmod p.

Thus n−U−V−px−qyn-U-V-px-qy is divisible by pqpq. Put S=U+V+px+qyS=U+V+px+qy. The parameter ranges imply

2p2≤S≤2p2+2pq−2.(20)2p^2\le S\le2p^2+2pq-2. \tag*{(20)}

In particular, S<nS<n, so R=(n−S)/(pq)R=(n-S)/(pq) is a nonnegative integer. If R≤11R\le11, then

n≤2p2+13pq−2<15p2,n\le2p^2+13pq-2<15p^2,

contrary to n≥16p2n\ge16p^2. If R≥508R\ge508, then p<2qp<2q gives

n≥2p2+508pq>256p2.n\ge2p^2+508pq>256p^2.

This proves the bounds on RR. □

Proposition 4.2. Every integer in [16pk2,256pk2][16p_k^2,256p_k^2] belongs to Ak+BkA_k+B_k.

Proof. Use Lemma 4.1, and consider

a(t)=U+p(x+qt),b(t)=V+q(y+p(R−t)).a(t)=U+p(x+qt), \qquad b(t)=V+q(y+p(R-t)).

where tt ranges over the integer interval

I=[max⁡(0,R−H),min⁡(H,R)]∩Z.(21)I=[\max(0,R-H),\min(H,R)]\cap\mathbb{Z}. \tag*{(21)}

Both tt and R−tR-t lie between 00 and HH. The bounds 12≤R≤50712\le R\le507 show that II contains at least thirteen consecutive integers.

All the a(t)a(t) lie in a single class modulo qq. Modulo 3q3q, changing tt by one changes a(t)a(t) by pqpq. Since pp is relatively prime to 33, the three residues of tt modulo 33 give the three different child classes. Lemma 2.1 implies that at most one residue of tt modulo 33 makes a(t)a(t) unsafe.

Write pk+12=dkpk2p_{k+1}^2=d_kp_k^2 with dk∈{2,4}d_k\in\{2,4\}. All the b(t)b(t) lie in one class modulo pp. Changing tt by one changes b(t)b(t) by −pq-pq, and qq is relatively prime to dkd_k. The residues of tt modulo dkd_k therefore give the different child classes. At most one of those residues makes b(t)b(t) unsafe.

In any thirteen consecutive integers, a fixed residue modulo 33 occurs at most five times, and a fixed residue modulo dkd_k occurs at most seven times. At most twelve positions are excluded. Choose a remaining t∈It\in I. Then a(t)a(t) and b(t)b(t) are safe unshifted candidates, so both were retained. Finally (19) gives a(t)+b(t)=na(t)+b(t)=n. □

The infinite sets and their representation function

Start at k=3k=3 with empty preceding unions, which satisfy (7). Lemma 3.1 and Proposition 3.2 allow the construction to continue for every k≥3k\ge3. Define

A=⋃k≥3Ak,B=⋃k≥3Bk.(22)A=\bigcup_{k\ge3} A_k,\qquad B=\bigcup_{k\ge3} B_k. \tag*{(22)}

Proposition 4.2 makes each block nonempty. Disjointness within each component therefore makes both unions infinite. All their elements are positive by (13).

The intervals covered in Proposition 4.2 overlap, since

16pk+12≤16(4pk)2=256pk2.16p_{k+1}^2\le16(4p_k)^2=256p_k^2.

Their endpoints tend to infinity. Their first lower endpoint is 16p32=16⋅322=1638416p_3^2=16\cdot32^2=16384, which proves (3).

Lemma 5.1. For any positive integer nn, at most five indices j≥3j\ge3 satisfy

pj2≤n≤526pj2.(23)p_j^2\le n\le526p_j^2. \tag*{(23)}

Proof. Successive values of pj2p_j^2 increase by a factor of at least four. If six indices j0<⋯<j5j_0<\cdots<j_5 satisfied (23), then

1024pj02=45pj02≤pj52≤n≤526pj02,1024p_{j_0}^2=4^5p_{j_0}^2\le p_{j_5}^2\le n\le526p_{j_0}^2,

which is impossible. □

Proposition 5.2. The sets in (22) satisfy rA,B(n)≤47005r_{A,B}(n)\le47005 for every n∈Zn\in\mathbb{Z}.

Proof. There are no representations when n≤0n\le0, so let n>0n>0. Each representation n=a+bn=a+b has unique block indices i,ji,j with a∈Aia\in A_i and b∈Bjb\in B_j. Put ℓ=max⁡(i,j)\ell=\max(i,j). Monotonicity of the scales and (13) give

pℓ2≤n≤526pℓ2.p_\ell^2\le n\le526p_\ell^2.

Lemma 5.1 leaves at most five possibilities for ℓ\ell. Fix one such ℓ\ell, and abbreviate p=pℓp=p_\ell, q=qℓq=q_\ell, U=UℓU=U_\ell, V=VℓV=V_\ell. Representations with maximal index ℓ\ell fall into the following three disjoint cases.

The AA-point lies in AℓA_\ell and the BB-point lies in PℓBP^B_\ell. The AA-point belongs to one of the three residues U+sqU+sq modulo pp. For a fixed ss, its partner belongs to the residue n−U−sqn-U-sq modulo pp. The invariant (7) bounds the number of such partners in the entire set PℓBP^B_\ell by KK. Each partner determines the AA-point uniquely, giving at most 3K3K representations.

The BB-point lies in BℓB_\ell and the AA-point lies in PℓAP^A_\ell. There are two possible residues V+upV+up modulo qq for the BB-point. The same argument, using the AA-side of (7), gives at most 2K2K representations.

Both points lie in block ℓ\ell. Fix their shifts 0≤s<30\le s<3 and 0≤u<20\le u<2. The remaining equation is

pX+qY=n−U−V−sq−up.pX+qY=n-U-V-sq-up.

Modulo qq, this determines XX uniquely as a residue class. Among the LqLq indices 0≤X<Lq0\le X<Lq, that class contains LL choices. Each choice determines at most one YY. The six shift pairs therefore contribute at most 6L6L representations.

Summing first over the three cases and then over the eligible maximal indices yields

rA,B(n)≤5(3K+2K+6L)=5(5⋅1567+6⋅261)=47005.r_{A,B}(n)\le5(3K+2K+6L)=5(5\cdot1567+6\cdot261)=47005.

Increasing enumerations

The block bijections (17) combine to a bijection f:A→Bf:A\to B, because the blocks are disjoint within each component. For a∈Aka\in A_k, both aa and f(a)f(a) are at least pk2p_k^2. Equation (18) thus gives

∣a−f(a)∣≤5a,∣a−f(a)∣≤5f(a).(24)\lvert a-f(a)\rvert\le5\sqrt{a},\qquad\lvert a-f(a)\rvert\le5\sqrt{f(a)}. \tag*{(24)}

Lemma 6.1. Suppose infinite sets AA, B⊆NB\subseteq\mathbb{N} admit a bijection satisfying (24). Their increasing enumerations satisfy ∣ai−bi∣≤5bi\lvert a_i-b_i\rvert\le5\sqrt{b_i} for every i≥1i\ge1.

Proof. The images of a1,…,aia_1,\ldots,a_i are ii distinct points of BB. Each is at most ai+5aia_i+5\sqrt{a_i}, since the function x↦x+5xx\mapsto x+5\sqrt{x} is increasing. Hence

bi≤ai+5ai.b_i\le a_i+5\sqrt{a_i}.

Apply the same argument to f−1f^{-1} to obtain ai≤bi+5bia_i\le b_i+5\sqrt{b_i}. If ai≤bia_i\le b_i, the first inequality gives bi−ai≤5ai≤5bib_i-a_i\le5\sqrt{a_i}\le5\sqrt{b_i}. If ai>bia_i>b_i, the second gives the required bound directly. □

Proof of Theorem 1.1. The construction in Section 5 gives infinite positive sets satisfying (3). Proposition 5.2 proves (4), and Lemma 6.1 proves (2). Since bi→∞b_i\to\infty,

∣aibi−1∣≤5bi⟶0.\left\lvert\frac{a_i}{b_i}-1\right\rvert\le\frac{5}{\sqrt{b_i}}\longrightarrow0.

The representation function has finite upper limit, which disproves (1). □

The everywhere-covering formulation

The same construction also gives a counterexample when the covering condition is imposed on every nonnegative integer, with zero allowed in the summand sets.

Corollary 7.1. There are infinite sets A∗,B∗⊆N0A^*, B^* \subseteq\mathbb{N}_0 with A∗+B∗=N0A^* + B^* = \mathbb{N}_0, whose increasing enumerations satisfy ai∗/bi∗→1a_i^*/b_i^* \to1 and whose ordered representation function is uniformly bounded.

Proof. Let A,BA, B be the sets in Theorem 1.1 and put N=16384N = 16384. The complement of BB in N0\mathbb{N}_0 is infinite. Otherwise, for any 47006 distinct elements aa of AA, every sufficiently large nn would have n−a∈Bn-a \in B for all of them. This would give rA,B(n)≥47006r_{A,B}(n) \ge47006, contradicting (4).

Set F={0,…,N−1}∖AF = \{0,\ldots,N-1\} \setminus A and m=∣F∣m = |F|. We have m≥1m \ge1 because 0∉A0 \notin A. Choose a set G⊆N0∖BG \subseteq\mathbb{N}_0 \setminus B with 0∈G0 \in G and ∣G∣=m|G| = m, and put A∗=A∪FA^* = A \cup F, B∗=B∪GB^* = B \cup G. Every n<Nn < N is represented by n+0n+0, and every n≥Nn \ge N retains an original representation.

For a fixed target, the representations using an element of FF number at most mm, and those using an element of GG number at most mm. Thus rA∗,B∗(n)≤47005+2mr_{A^*,B^*}(n) \le47005 + 2m. Beyond the finitely many added elements, the increasing enumerations are ai∗=ai−ma_i^* = a_{i-m} and bi∗=bi−mb_i^* = b_{i-m}. Their ratio therefore tends to one.

Acknowledgments

The authors acknowledge assistance from OpenAI’s artificial intelligence systems in developing the construction and preparing and checking the exposition.

References

  1. [1]Erdős Problems, Problem 1145, problem statement and discussion, accessed November 2026. https://www.erdosproblems.com/1145.
  2. [2]Plasma AI, Proof of the annular square-complement limits, public research notes, 2026. https://erdosproblems.ai/wiki/theory/additive_bases/L3_square_complement_nine_halves/_proof.
  3. [3]Plasma AI, Monotone path shells for Erdős Problem 1145, public research notes, 2026. https://github.com/plasma-ai/erdos/blob/main/wiki/research/erdos_1145/monotone_path_shells.md.
  4. [4]Various authors, Some of Paul’s favorite problems, booklet circulated at the conference “Paul Erdős and his mathematics,” Budapest, July 1999, Problem 1.17. https://web.math.pmf.unizg.hr/~vjekovac/EP/Some_of_Pauls_favorite_problems.pdf.

Paper details

vlnc:W_jpbta528
Version
v1 · latest
Published
Fresh — published within the last 30 days
License
CC BY 4.0
Usage

views · 9

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

OpenAI’s artificial intelligence systems

View statement
Looking for related papers…
Contents