---

Subject and independence

Role: independent reviewer in a fresh context, commissioned with a refutation charge and given only the assignment. The reviewer took no part in writing the reviewed page, the library cards it cites or any other page of the Problem 416 research folder, and read no other review. Independence facts: every step was re-derived from the preprint's pages before the page's text was compared with it, and the sanity computation reported below is the reviewer's own.

Subject: path wiki/research/erdos_416/zeraoulia_theorem_1_1_reconstruction.md (the reconstruction page) as it stood on 2026-09-28T05:03:27Z, read in full, line by line.

Artifacts read:

• The preprint PDF under the Zeraoulia card, fifteen pages, its SHA-256 equal to the card's provenance line: physical pages 1–5 and 8 at page-image depth (rendered at 120 dpi; every displayed formula of Theorem 1.1, §2, §3 and Proposition 6.1 was checked on the images); pages 6–7 and 9–15 in text extraction only, for the result labels listed under "What is not reconstructed", the §8 figure and the reference list. Physical and printed page numbers coincide.
• The Ford PDF under the Ford card, 43 pages: physical pages 1–2 at page-image depth (110 dpi) for Theorem 1 in §1.1; page 3 (the Remark after Theorem 3) and page 42 (reference [14]) in text extraction, to identify the version of the held text.
• The Zeraoulia card's provenance paragraph, and the Ford card's provenance line and its Theorem 1 sentence.
• The Statement paragraph of the E0416 problem page.
• docs/verification.md ("Whole-claim report" and "Audit checklist"), docs/evidence.md ("Source fidelity") and docs/math_authoring.md.

Not read: the folder's _index.md; the three Kruer–Kohlmeyer reconstruction pages of the folder (the page cites none of them as an input); the Kruer–Kohlmeyer card and its result pages (the page's Source paragraph does not link them); every evidence folder; other reviews; the web.

Exposures: (1) the first sentence of the E0416 page's Status paragraph, on the doubling question, appeared in a heading search made to locate the Statement paragraph; (2) the Zeraoulia card printed whole, so its "Bears on" and "Relation to E416" paragraphs, which mention the acceptance of another proof, were seen with the provenance paragraph; (3) a list of recent commit subjects unrelated to this page was seen; (4) the file names, not the contents, of three other review files in this directory were seen. None of these affected a mathematical verdict; item (2) bears only on finding F6.

Restatement

Conventions. $\varphi$ is Euler's function; for real $x\ge1$, $V(x)$ is the number of integers in $[1,x]$ that are values of $\varphi$; $c>1$ is a fixed real number and $R_c(x)=V(cx)/V(x)$ for real $x\ge1$; $\log_k$ is the $k$-fold iterated natural logarithm; a cluster set is a set of subsequential limits.

Claim (the preprint's Theorem 1.1 as the page states it). For every fixed real $c>1$:

1. $\liminf_{n\to\infty}|R_c(n)-c|=0$, the limit inferior over the integers $n$.
2. For every function $L$ with $L(X)\to\infty$ there is a function $\omega(X)\to0$, allowed to depend on $c$ and $L$, such that for every large real $X$ some integer $n\in[X,cXL(X)]$ has $|R_c(n)-c|\le\omega(X)$.
3. The cluster set of $(R_c(n))_{n\in\mathbb N}$ is $[\alpha_c,\beta_c]$ with $\alpha_c=\liminf_nR_c(n)$ and $\beta_c=\limsup_nR_c(n)$, both finite, and $\alpha_c\le c\le\beta_c$; the cluster set of $R_c(x)$ as $x\to\infty$ through the reals is the same interval.

Consequence: exactly one of "$R_c(x)\to c$ as $x\to\infty$" and "the cluster set is a nondegenerate closed interval containing $c$, hence uncountable" holds. No bound on $\beta_c-\alpha_c$ is claimed. Imported: Ford's Theorem 1 (the order of $V$ up to a bounded factor) and Chebyshev's lower bound for $\pi$. The added Proposition 6.1: with $P(x)$ the number of totient values $v\le x$ such that $v/2$ is not a totient value, $V(x)=\sum_{k\ge0}P(x/2^k)$ and $V(2x)-V(x)=P(2x)$ for every real $x\ge1$.

Standing as stated: an author-recorded reconstruction of a self-published, unreviewed preprint; no tier and no status change.

Checklist

Canonical failure modes:

• "Almost all" upgraded to "all": absent. The page's quantifiers (every fixed $c>1$; every $L\to\infty$; all large $X$; all large $N$ and every integer $H\ge1$) are the preprint's (pp. 2–5), and every largeness threshold in the proof is named ($x_1$, $t_0(c)$, $x_2(c)$) except the two recorded in F2 and F5.
• Induction presupposing termination: no induction is used. The halving process in the dyadic identity stops within $\log_2v$ steps, which the page states.
• Averaging heuristic presented as proof: absent. The block mean (P) is an exact telescoping identity plus bounds, and it is turned into a statement about single values either by "the minimum is at most the geometric mean" or by an explicit crossing; nothing is inferred from an average alone.
• Circular use of an equivalent statement: absent. Nothing in Steps 1–7 assumes the limit; the equivalence "doubling law if and only if $P(x)\sim V(x)/2$" is labeled as a rephrasing and used for nothing.
• Exceptional sets dropped from density arguments: inapplicable; there is no density argument.
• Finite verification cited beyond base cases: absent. The page quotes the preprint's $V(10^{10})$ as unverified and uses no computation; the reviewer's sanity computation below is likewise not evidence.
• A relaxed or averaged system standing in for the actual objects: absent. The averaged statement (P) is converted into statements about the actual sequence $R_c(n)$, and the page claims no convergence.

Named patterns:

• Model-class transport: inapplicable; no axiom system or certificate class is involved.
• Uniformity asserted from finitely many instances: absent. Every constant's dependence is stated: $K$, $K'$, $K_1$, $K_2$ absolute; $K_c$, $K_3$, $K_4$, $K_5$, $C_c$ depending on $c$; $\omega$ depending on $c$ and $L$; the thresholds $t_0(c)$ and $x_2(c)$ depending on $c$.
• Extremal claims audited in the claim's own units: pass. Clause 1 is checked in its own units, by an explicit sequence $n_k\in[k,ck^2]$ with $|R_c(n_k)-c|\le\omega(k)\to0$.
• Consequence sentences as claim surfaces: pass. "Exactly one of the following holds" was attacked on its own (a bounded sequence whose cluster set is one point converges; a nondegenerate interval is uncountable; the two cases exclude each other), and the dyadic rephrasing was checked in both directions (Strongest attack).
• Carry hypotheses actually used: two thresholds are used without being stated at the point of use (F2, F5); both are available and harmless.
• A composition inherits its unproved premises: pass. The imported premises are Ford's Theorem 1, held and matching, and Chebyshev's bound, standard and not held; the page names both as external and records the preprint as unreviewed.
• Reproducibility notes are claims: inapplicable; the page carries no rerun line and no check count.
• Verifier quotations are claims: the page quotes no ruling on itself. Its Standing sentence about an accepted Lean proof for $c=2$ reports the problem page's record, which lies outside this review's read set (F6).
• Verdict words spelled in full: the page carries none; this report's verdict is refutation-failed, written in full.
• Certified-bracket functions, harness legs without a failing input, and gates that read caches: inapplicable; the page has no numerics, harness or gate.

Weakest steps

Step 5, the crossing case (preprint pp. 4–5). Let $s_j=R_c(n_j)$ for $N\le j\le N+H-1$ with $n_j=\lceil c^j\rceil$, and suppose no $s_j$ equals $c$ while some $s_j<c$ and some $s_k>c$. Walking from $j$ toward $k$, the first index at which the side of $c$ changes gives consecutive $i,i+1$ in the block with $s_i$ and $s_{i+1}$ on opposite sides. Since $c^{i+1}-c^i=(c-1)c^i\ge(c-1)c^N>1$ once $N$ is large, $n_i<c^i+1<c^{i+1}\le n_{i+1}$. Take $s_i<c<s_{i+1}$; the other case is the mirror image. The set of integers $m\in[n_i,n_{i+1})$ with $R_c(m)<c$ contains $n_i$; let $m^\ast$ be its largest element. Then $m^\ast+1\le n_{i+1}$, and either $m^\ast+1=n_{i+1}$, where $R_c>c$, or $m^\ast+1<n_{i+1}$ and maximality gives $R_c(m^\ast+1)\ge c$. So $R_c(m^\ast)<c\le R_c(m^\ast+1)$ and $|R_c(m^\ast)-c|\le R_c(m^\ast+1)-R_c(m^\ast)$. By (7) at $x=m^\ast$ and $h=1$, legitimate because $m^\ast\ge n_N\ge c^N\ge\max(x_2(c),2)$, this is at most $K_3(\log m^\ast)/m^\ast\le K_3(\log c^N)/c^N$, the last step because $t\mapsto(\log t)/t$ decreases for $t\ge e$ and $m^\ast\ge c^N\ge e$. Composition: with the two one-sided cases (the minimum is at most the geometric mean, the maximum at least it) and the exact-hit case, every block $\{n_N,\dots,n_{N+H-1}\}$ with $N$ large contains an integer $x$ with $|R_c(x)-c|\le C_c(\Delta(N,H)+N/c^N)$, which is (8); the minimum in (8) ranges over all integers of the block, not only the sample points, and $m^\ast$ lies in it.

Step 3, the boundedness of $R_c$ and the unit-interval bound (7) (preprint p. 4). The preprint writes "Ford's estimate gives $V(cx)\asymp_cV(x)$"; the page derives this from Theorem 1 alone. For $x\ge x_1$, (F) at $x$ and at $cx$ gives $R_c(x)=(M(cx)/M(x))\,e^{E(cx)-E(x)}\le e^{2K}M(cx)/M(x)$, and $M(cx)/M(x)=c\cdot(\log x/\log(cx))\cdot e^{\Psi(cx)-\Psi(x)}$. With $x=c^t$, $\Psi(cx)-\Psi(x)=\psi_c(t+1)-\psi_c(t)$, whose absolute value is at most $\int_t^{t+1}K'/s\,ds\le K'/t$ by Step 1, so $M(cx)/M(x)\to c$ and $R_c(x)\le K_c:=e^{2K}(c+1)$ for $x\ge x_2(c)$. For (7): with $A=V(cx)$, $B=V(x)$, $\delta=V(c(x+h))-V(cx)\in[0,\lceil c\rceil]$ and $\epsilon=V(x+h)-V(x)\in\{0,1\}$, the difference $(A+\delta)/(B+\epsilon)-A/B=(B\delta-A\epsilon)/(B(B+\epsilon))$ has absolute value at most $\delta/B+(A/B)(\epsilon/B)\le(\lceil c\rceil+K_c)/V(x)$, and $V(x)\ge\pi(x)\ge c_0x/\log x$ for $x\ge2$ because $p\mapsto p-1$ injects the primes $p\le x+1$ into the totient values in $[1,x]$. The count bounds on $\delta$ and $\epsilon$ hold because a half-open interval $(a,a+\ell]$ contains exactly $\lfloor a+\ell\rfloor-\lfloor a\rfloor\le\lceil\ell\rceil$ integers: exactly $\ell$ when $\ell$ is an integer, at most $\lfloor\ell\rfloor+1$ otherwise. Composition: (7) is the only input to Steps 4 and 7 and to the crossing case of Step 5, and the bound $K_c$ is what makes $\alpha_c$ and $\beta_c$ finite; the page's Standing sentence that Theorem 1 and Chebyshev's bound are the only external inputs rests on this derivation, which is correct.

Step 1, the derivative bound (preprint p. 3, Lemma 2.1). With $u=t\log c$, $\log_3(c^t)=\log\log u$ and $\log_4(c^t)=\log\log\log u$, so their $t$-derivatives are $1/(t\log u)$ and $1/(t\log u\log\log u)$. For $u\ge e^e$ one has $\log u\ge e$ and $\log\log u\ge1$, so both derivatives lie in $(0,1/(t\log u)]$ and $0\le\log_4(c^t)\le\log_3(c^t)=\log\log u$. Differentiating $\Psi(c^t)$ term by term, $|\psi_c'(t)|\le(2C\log\log u+D+|D+\tfrac12-2C|)/(t\log u)\le K'/t$ with $K'=2C+D+|D+\tfrac12-2C|$, using $\log\log u\le\log u$ and $\log u\ge1$; with the derivative $-1/t$ of $-\log(t\log c)$ this gives $|\eta_c'(t)|\le K_1/t$, $K_1=K'+1$ an absolute constant, for $t\ge t_0(c)$. Integrating over $[N,N+H]$ gives $|\eta_c(N+H)-\eta_c(N)|\le K_1\log(1+H/N)$. Composition: with the telescoping identity and (F) at $c^N$ and $c^{N+H}$ this is exactly (6); the constant's independence of $c$ is stronger than the preprint's $O_c$, but it is what the page proves and states.

Strongest attack

The attack aimed at the crossing argument of Step 5 and its second use in Step 7, since the theorem rests on converting an averaged bound into a bound at a single integer.

• Blocks with one sample ($H=1$): the mixed case cannot occur, and the one-sided bound $|s_N-c|\le|S-c|$ needs only $|\log S-\log c|$ bounded, which (P) gives with $\Delta(N,1)=1+\log(1+1/N)\le1+\log2$; so $K_5$ exists and the exponentiation step needs no smallness. Failed.
• Coinciding samples ($n_i=n_{i+1}$), which would leave no integer to walk over: excluded by $(c-1)c^N>1$. Failed.
• A jump of $R_c$ across $c$ larger than the unit-interval bound: impossible, since (7) holds at every real $x\ge\max(x_2(c),2)$ with $h=1$ and $m^\ast$ is at least $c^N$. Failed.
• A real-variable cluster point missing from the integer cluster set: for real $x$ with $n=\lfloor x\rfloor$, (7) with $h=x-n<1$ gives $|R_c(x)-R_c(n)|\le K_3(\log n)/n\to0$, so the two cluster sets coincide. Failed.
• A slowly growing window, say $L(X)=\log\log X$: $H=\lfloor\log_cL(X)\rfloor$ still tends to infinity and the block still sits inside $[X,cXL(X)]$ once $L(X)\ge c$ and $(c-1)XL(X)\ge1$, by $n_N\ge c^N\ge X$ and $n_{N+H-1}<c^Nc^{H-1}+1<XL(X)+1\le cXL(X)$; and $\Delta(N,H)\le(1+\log(1+H))/H\to0$ needs no relation between $N$ and $H$. Failed.
• The dichotomy sentence: if $\alpha_c=\beta_c$ then $R_c(n)$ converges to the common value, which is $c$ because $c\in[\alpha_c,\beta_c]$, and the real variable follows from the previous item; if $\alpha_c<\beta_c$ the cluster set is a nondegenerate interval, hence uncountable. Failed.
• The dyadic rephrasing: if $V(2x)/V(x)\to2$ then $P(2x)=V(2x)-V(x)\sim V(x)\sim V(2x)/2$, that is $P(y)\sim V(y)/2$; conversely $P(x)\sim V(x)/2$ gives $V(2x)-V(x)\sim V(2x)/2$, that is $V(x)\sim V(2x)/2$. Both directions hold. Failed. A sanity computation of $V(x)=\sum_{k\ge0}P(x/2^k)$ and $V(2x)-V(x)=P(2x)$ for every integer $x\le2\cdot10^5$ found no failure; for instance $V(10)=6=P(10)+P(5)+P(2.5)+P(1.25)=3+1+1+1$ and $V(20)-V(10)=4=P(20)$, the dyadically primitive values up to $20$ being $1,6,10,18$.

The attack found no defect in the mathematics. What it found is recorded in the findings: a misdescription of the version of the held Ford text (F1), an omitted largeness qualifier in Step 7 (F2) and unlabeled supplied steps (F3).

Premises

• Ford's Theorem 1. Interface as used: there are $K>0$ and $x_1$ with $V(x)=M(x)e^{E(x)}$, $|E(x)|\le K$, for $x\ge x_1$, where $M(x)=(x/\log x)\exp\Psi(x)$ and $\Psi$ is the displayed quadratic in $\log_3x$ and $\log_4x$ with constants $C=0.8178\ldots$ and $D=2.1769\ldots$. Source held: the Ford PDF, Theorem 1 in §1.1, physical page 2, read on the page image; the display agrees with the page symbol for symbol. The held text is a corrected revision of the 1998 journal paper, not the journal print (F1). Hypothesis: $x$ large; met wherever applied ($x\ge x_1$, $c^N\ge x_1$). Standing: imported, named as external on the page. The values of $C$ and $D$ are not used.
• Chebyshev's lower bound. Interface: an absolute $c_0>0$ with $\pi(y)\ge c_0y/\log y$ for all real $y\ge2$. Not held; a standard theorem. Applied through $V(x)\ge\pi(x+1)\ge\pi(x)$ (F7). Standing: imported, named as external on the page.
• Elementary facts proved on the page. Unit jumps of $V$ on intervals of length at most one and at most $\lceil c\rceil$ new values on $(cx,c(x+h)]$; closure of the totient values under doubling ($\varphi(2m)=2\varphi(m)$ for even $m$, $\varphi(4m)=2\varphi(m)$ for odd $m$). Re-derived above and in the dyadic attack.
• The preprint. Self-published and unreviewed, held with SHA-256 equal to the card's; its §§2–3 proofs are one to four lines each and the page expands them. No local claim is consumed. Explicit assumptions: none beyond $c>1$ fixed. No batch acceptance order.

Findings

F1. Severity: suggested. Location: "The preprint cites the same theorem from Ford's revised arXiv version; the statement is identical." Defect: the sentence reads as if the held Ford text were the 1998 journal print and the preprint's source a different, revised text compared with it. The held PDF is itself a later corrected revision: its Remark after Theorem 3 (physical page 3) says the proof of "[14, Theorem 3]" contains an error and gives a corrected proof, its reference [14] (physical page 42) is the 1998 Ramanujan Journal paper, and its metadata date is 2012. The comparison the held material supports is between the held revision's Theorem 1 (page 2) and the preprint's display (2) (page 3), which agree. Proposed replacement: "The held Ford PDF is the author's later corrected text (its Remark after Theorem 3 corrects the 1998 journal print, cited there as [14]); the preprint's reference [5] is that revision, and its display (2) on p. 3 agrees with the statement above. The 1998 journal print is not held."

F2. Severity: suggested. Location: Step 7, "let $n_0$ be given" and the display "$\le K_3\,\frac{\log n_0}{n_0}$". Defect: the display uses (7) at $n>n_0$, which needs $n\ge\max(x_2(c),2)$, and the monotonicity of $(\log t)/t$ beyond $n_0$, which needs $n_0\ge e$; neither is stated, whereas Step 5 states the parallel condition "$m^\ast\ge n_N\ge c^N\ge e$". Witness: for $n_0=1$ the display reads $|R_c(n)-y|\le0$. The conclusion is unaffected, since $n_0$ is then sent to infinity. Proposed replacement: "let $n_0\ge\max(x_2(c),3)$ be given".

F3. Severity: suggested. Location: the Source paragraph, "proved through Lemma 2.1 and Theorem 2.2". Defect: the page labels one supplied item (the $P(x)\sim V(x)/2$ rephrasing) as its own but not the others. Supplied without a label: the derivation of $1\le R_c\le K_c$ from Theorem 1 and Step 1 in Step 3 (the preprint, p. 4, writes "Ford's estimate gives $V(cx)\asymp_cV(x)$"); the passage from $|\log S-\log c|$ to $|S-c|$ through $K_5$ in Step 5 (the preprint, p. 4, writes that Proposition 3.2 gives the bound on $|S-c|$ directly); the explicit constants $K'$, $K_1$ to $K_5$ and $C_c$; the explicit $m^\ast$ and $n$ constructions of Steps 5 and 7 and the containment arithmetic of Step 6. All are correct and routine. Proposed replacement: add to the Source paragraph "The preprint's proofs are one to four lines each; the explicit constants, the derivation of the boundedness of $R_c$ from Theorem 1 alone, the exponentiation step of Step 5 and the explicit crossing constructions of Steps 5 and 7 are supplied here."

F4. Severity: note. Location: Steps 2, 4 and 5, "$H\ge1$". Defect: the sums and products run over $j=N,\dots,N+H-1$, so $H$ is an integer; the preprint (Theorem 2.2, p. 3; Theorem 3.3, p. 4) has the same implicit convention. Step 6's $H=\lfloor\log_cL(X)\rfloor$ is an integer, so nothing breaks. Proposed replacement: "integer $H\ge1$" at the first occurrence in Step 2.

F5. Severity: note. Location: Step 4, "take $N$ so large that $c^N\ge\max(x_2(c),2)$ and $c^N\ge e$". Defect: (6), combined at the end of the step, also needs $N\ge t_0(c)$ (an integer with $c^N\ge x_1$); Step 5 then says "for all large $N$", so the hypothesis is carried, but not at the point of use. Proposed replacement: "take an integer $N\ge t_0(c)$ so large that".

F6. Severity: note. Location: Standing paragraph, "for $c=2$ the accepted Lean proof recorded on the problem page collapses the interval to the point $2$". Defect: the sentence reports another record's acceptance without the qualification the library card uses ("if that acceptance stands"); the problem page's status text lies outside this review's read set, so the report is unchecked here. The sentence claims nothing about this page's own standing. Proposed replacement: "for $c=2$ the problem page records an accepted Lean proof of the limit, which, if that acceptance stands, collapses the interval to the point $2$".

F7. Severity: note. Location: "Chebyshev's lower bound", the display "$V(x)\ge\pi(x+1)\ge c_0\,\frac{x}{\log x}$". Defect: the second inequality follows from $\pi(x+1)\ge\pi(x)$ and the stated bound at $y=x$, not from the bound at $y=x+1$; for $2\le x<e$ one has $(x+1)/\log(x+1)<x/\log x$ (at $x=2$: $2.73<2.89$). Proposed replacement: "$V(x)\ge\pi(x+1)\ge\pi(x)\ge c_0\,\frac{x}{\log x}$".

Verdict

Source fidelity: faithful with corrections. The statement's hypotheses, conclusion, quantifiers and conventions match Theorem 1.1 (p. 2), Theorems 3.3 and 3.4 and Corollary 3.5 (pp. 4–5) and Proposition 6.1 (p. 8); every locator on the page (physical pages, section numbers, result labels, the fifteen-page count and the coincidence of physical and printed pages) is correct; the imported Ford statement matches the held text symbol for symbol. The corrections are F1 (the version of the held Ford text), F2 and F3; none is required for the mathematics.

The argument as reconstructed: sound. Every deduction of Steps 1–7 and of the dyadic identity was re-derived; the constants are consistent with each other and their dependence on $c$ is stated; nothing the source proves is altered or strengthened beyond the absolute constants $K_1$ and $K_2$, which the page proves. Verdict of the refutation charge: refutation-failed.

Limitations: the review covers the unconditional argument and the dyadic identity only; §§4–8 of the preprint were read in text extraction only and are not assessed; Chebyshev's bound is accepted as a standard theorem without a held source; the sanity computation of the dyadic identity is not evidence. Required corrections: none; suggested: three; notes: four. This focused review assigns no tier and changes no status.