19 August 2026 · correspondence feedback · v2.0
A five-page note reformulates the Riemann hypothesis geometrically: the critical line is the set equidistant from the two boundary lines of the critical strip; the zeros are discrete because ζ oscillates; after adding the point at infinity the zero set becomes closed; and only one vertical line can carry a zero sequence running off to infinity, so it must be the equidistant one. Its author asks for an opinion and for help putting it on arXiv. This is the opinion, at the standard we apply to our own work: what the picture gets right and why it is right for a better reason than the note gives, the four steps that do not carry, the one function that decides the matter — and we hold 386 verified zeros of it — and an honest answer about arXiv. v2.0 adds the review of his second note, written a day later, which retires the compactification argument, reaches this programme's own displacement coordinate, and proposes a quadratic "transverse energy" whose vanishing would give the hypothesis — a criterion that is correct, that we ran on the deciding function, and that returns 1.0076×10−4 there.
One day after this review was written, Kazushi Mizutani posted a second note: A Projective Zeta Geometry Interpretation of the Critical Line: Euler Products, Transverse Displacement, and a Proposed Arithmetic Confinement Principle — fifteen pages, twenty sections. It goes straight at the objection below, and it does so on its own: this review cannot be shown to have reached him before he wrote it. The steps that did not carry are dropped, this programme's own displacement coordinate is reached independently, and a hidden gap is replaced by a stated one. The review of it is Parts VIII–XII, and the short version is that his new criterion is correct — we ran it on the function that decides these things and it works — and that being correct is not the same as being able to forbid anything. Parts I–VII below are the original review of the first note, unchanged.
A Geometric Reinterpretation of the Riemann Hypothesis, Kazushi Mizutani, five pages, posted to Academia.edu with an open discussion attached. The abstract states its own ceiling in the first three lines — "We do not claim a proof" — and that sentence is worth saying first, because it is the sentence most notes in this area do not contain. We hold ourselves to the same rule on every page of this site, and the review below is written on the assumption that the author means it.
Discussion: A Geometric Reinterpretation of the Riemann Hypothesis · Academia.edu · 5 pp · abstract, §1 introduction, §2 Euclidean motivation, §3 the strip and equidistance, §4 oscillation, §5 zeros and the point at infinity, §6 the reinterpretation, §7 the reformulation, §8 conclusion.
The argument, restated as faithfully as we can in four steps:
Step 1 is correct and it is elementary — |σ| = |σ−1| has the single solution σ = ½ — but the reason it lands on ½ is not the metric, and seeing what it really is makes the observation better rather than worse. The set of real parts equidistant from 0 and 1 is exactly the fixed-point set of the reflection σ ↦ 1−σ, which is the real-part shadow of the map s ↦ 1−s: the involution the functional equation supplies. The equidistance construction is the functional equation's symmetry axis in metric clothing. It picks out ½ because the completed function is symmetric about ½, not because the boundary lines happen to sit at 0 and 1. Measured
Apply the note's own construction to ζ(2s). Its functional equation relates s to ½−s, its critical strip is 0 < σ < ½, and the line equidistant from those two boundaries is Re s = ¼. And that is exactly right: the non-trivial zeros of ζ(2s) are ζ's own zeros halved, so they sit on Re s = ¼ precisely to the extent that ζ's sit on ½. The construction transports correctly to a function whose critical line is not ½ — which is the strongest evidence that it is reading the functional equation rather than the arithmetic, and the author is entitled to that as a positive result about his own picture.
Checked here at thirty digits rather than asserted, on the first six zeros: every one sits on Re s = ¼ exactly, with |ζ(2s)| below 10−30 there. And the check discriminates — at those same six heights, on the line Re s = ½ that the construction would have returned had it been reading the numbers 0 and 1 rather than the functional equation, |ζ(2s)| never falls below 0.33. Measured
One thing has to be said about where L₀ and L₁ come from, because the note takes them as given. The strip is not a geometric datum: ζ has no zeros with σ > 1 because of the Euler product, no zeros on σ = 1 by Hadamard and de la Vallée Poussin, and no non-trivial zeros with σ < 0 because the functional equation reflects the first fact across. The two lines the equidistance argument measures from are therefore consequences of the two hardest classical inputs in the subject. A derivation that starts from L₀ and L₁ has already spent them. Conceded
Step 2's conclusion is true and its reason is not needed. ζ is holomorphic on the strip and is not identically zero, so its zeros are isolated — that is the identity theorem for holomorphic functions, two lines, and it holds whether or not anything oscillates. It is worth replacing the oscillation argument with this one, because the oscillation argument also has a well-formedness problem: "∂/∂t ζ(σ+it) changes sign infinitely often" is not defined for a complex-valued function, which has no sign.
There is a real object behind the intuition and it is worth knowing, because the whole computational side of this subject runs on it. On the critical line, Hardy's Z(t) = eiθ(t)ζ(½+it) is real-valued, and its sign changes are the zeros of ζ on the line. That is Turing's method, it is how every large-scale verification of RH has been done, and it is the instrument our own census below runs on. Its limit is instructive for this note: Z's sign changes see only the zeros on the line, which is exactly why detecting zeros off it requires a counting deficit rather than an oscillation argument. Conceded
Step 3 is true and empty. Any discrete unbounded subset of C has ∞ as its only accumulation point in the Riemann sphere, so adjoining ∞ makes it closed. This is true of the integers, of the zeros of sin, of the zeros of every Dirichlet series in this discussion, and of the zeros of ζ. §5 and §6 therefore cannot distinguish ζ from anything else, and nothing in the conclusion can rest on them.
Step 4 uses oscillation to prevent a zero sequence on a line σ ≠ ½. What oscillation classically gives is the opposite. For ½ < σ ≤ 1 the closure of the value set {ζ(σ+it) : t ∈ R} is the whole complex plane — Bohr and Courant in 1914 for denseness, Jessen and Wintner in 1935 in the form we hold first-hand (Trans. AMS 38, Thm 31, p. 82) — so
inft |ζ(σ+it)| = 0 on every vertical line inside the right half of the strip.
ζ comes arbitrarily close to zero on every one of those lines. The open question is not whether it gets close — it does, on all of them — but whether the infimum is ever attained. That distinction is not a technicality we are importing; it killed one of our own routes. This programme proved a pointwise lower bound of exactly the shape the note wants (|ζ(σ+it)| bounded below, diverging, for σ ≤ 0, unconditional, carried by the Euler product through the functional equation) and then found that its continuation into the strip cannot be what it looked like: a positive pointwise floor there is not open, it is false, by the theorem above. What survives is the rate at which the infimum is approached, and that is where every attempt of ours stalled. Conceded
This is the load-bearing gap. Grant step 4's uniqueness clause in full. Nothing so far ties the unique line it produces to the equidistant line of step 1. §7 writes "Combining this with the equidistance condition d(s,L₀) = d(s,L₁) yields σ = ½" — but no step establishes that the zeros satisfy the equidistance condition. That the zeros are equidistant from the two boundaries is the statement being proved. The geometry supplies a distinguished line and the uniqueness clause supplies a distinguished line, and the note joins them by naming them both σ.
Separately, the uniqueness clause itself is not proved anywhere in the note, and as far as we can determine it is open — it is a consequence of the Riemann hypothesis, not an input available to a proof of it. Exploratory
Even granting everything, §7's final display is not equivalent to RH. RH forbids every off-line zero. The note's statement forbids only an infinite family of zeros sharing one vertical line. A function with one off-line zero at σ = ½ + 1/n for each n violates the hypothesis comprehensively and is invisible to the note's statement, which only ever speaks about families on a single line. So §7's "the Riemann Hypothesis is reinterpreted as" should read "is implied by the Riemann Hypothesis". That is not pedantry: a reformulation needs a proof of equivalence, and a referee will ask for it in the first paragraph.
The way to test a geometric criterion is to find a function that satisfies all of its hypotheses and violates its conclusion. One exists, it has been in print since 1936, and this programme holds its zeros to thirty digits.
Take the period-5 coefficient list (1, ξ, −ξ, −1, 0) with ξ = (√(10−2√5) − 2)/(√5 − 1) = 0.284079…, and let
f(s) = 5−s( ζ(s,1/5) + ξ·ζ(s,2/5) − ξ·ζ(s,3/5) − ζ(s,4/5) ).
This is the Davenport–Heilbronn function. It is a Dirichlet series of degree one with real coefficients; it satisfies a Riemann-type functional equation relating s to 1−s with conductor 5; its non-trivial zeros are discrete and their ordinates diverge to infinity; its critical strip has the same two boundary lines; and the completed function is symmetric about the same line Re s = ½, so the note's equidistance construction returns ½ for it, exactly as for ζ. Every geometric ingredient of the note is present. Verified
And it has zeros off the critical line. Their existence is Davenport and Heilbronn's, 1936; Saias and Weingartner (2009) upgrade it to a positive proportion — for any periodic coefficient sequence that is not a Dirichlet polynomial times a single L-function, there are ≍ T zeros in every substrip ½ < σ₁ < σ₂ < 1+η. Conceded (We hold Saias–Weingartner at statement level; the paper has not been read here first-hand.)
The zeros, located. 193 certified off-line quartets to height 4000 — 386 members, every one verified — cross-validated against the complete published record (Spira 1994, four zeros; Balanzario and Sánchez-Ortiz 2007, thirty) at 34 of 34. The first of them sits at
σ = 0.8085171824566374, t = 85.69934848537759
with its mirror partner at 1−σ, as the functional equation demands. That is a zero of a function carrying the note's entire geometry, sitting a third of the way across the strip. Verified
Both sides through one instrument. Over t ∈ [10, 10⁴], with identical code and identical sampling, five genuine L-functions with Euler products (ζ and the L-functions of the characters mod 3, 4, 5 and 7) return 60,318 zeros, every one on the critical line. The Davenport–Heilbronn function over the same range is missing 1,140 sign changes — 570 mirror pairs that left the line. Five of those pairs were then located individually by the argument principle with σ free, all five with their mirrors present; one of them reproduces a value banked in this project years of sessions earlier, by an unrelated route, to nine digits. Measured
Ceilings on our own numbers, because they bind here too. A census certifies a finite window above a finite detection floor and forbids nothing outside it. The on-line half is not new about these functions — that range sits inside Platt's rigorous verification — and what is ours is the comparative design, both classes measured by one instrument over one range. The deficit is overstated by under one per cent, one-sided, by a resolution correction we ran against our own favour.
One more thing the author can check for himself in five lines of arithmetic, because it is the exact property Davenport–Heilbronn lacks. If a period-q coefficient list has a residue r coprime to q with a(r² mod q) ≠ a(r)², then the series is not a Dirichlet polynomial times an L-function, so the theorem above applies to it and its off-line zeros are forced before a single one is computed. A genuine Dirichlet character reads 0 exactly at every residue. Davenport–Heilbronn fails at two residues, worst gap |a(4) − a(2)²| = |−1 − ξ²| = 1.0807 at r = 2. Verified (the certificate is ours; we have run no literature search on its novelty.)
The symmetry gives the axis. It does not give the confinement. Everything the note builds — the equidistant line, the discreteness, the divergence to infinity, the compactification — is a property of the functional equation, and Davenport–Heilbronn has the functional equation. So any argument assembled from those ingredients holds verbatim for a function with 570 pairs off the line below height 10⁴, and therefore cannot forbid an off-line zero of ζ. What separates the two objects is the Euler product, and nobody knows how to convert that into a statement about where a zero can be. Conceded — at degree one this is a 1999 classification (Kaczorowski and Perelli) together with a 2007 theorem (Kaczorowski and Kulas), not a discovery of ours; what is ours is the measurement.
The same theorem sharpens Part III's point 5, and this is worth stating because it is not obvious: for a degree-one series of this kind the off-line zeros have real parts dense in subintervals of (½,1). The off-line population is not a family sitting on one vertical line — it is spread across a continuum of them, which is precisely the shape §7's statement says nothing about. Conceded (Kaczorowski–Kulas, held here at statement level, pinned through two independent first-hand sources.)
Here is the single most useful thing we own, and it costs one line to apply:
If a proposed criterion never mentions the coefficients — if it uses only
analyticity, the functional equation and the strip — then it holds verbatim for
Davenport–Heilbronn, and it cannot forbid an off-line zero.
We did not start with that rule; we paid for it. Across eight papers this programme built roughly seventy exact objects — identities, closed forms, normal-form relations — scored each against requirements fixed in advance, and every single one failed the same requirement in the same way. The reason turned out to be structural rather than accidental: an equality inherited from the functional equation cannot separate two functions that share that functional equation. The note's construction is in that class, and knowing it in advance is worth more than any individual correction above. Measured
Step 4 contains a precise, unproved, and genuinely interesting statement, and the note passes over it as if it were a lemma:
Can a vertical line Re s = σ₀ ≠ ½ carry infinitely many zeros of ζ?
As far as we can determine, that is open. What is known bears on it without settling it: the zero-density theorems give, for each fixed σ > ½, that the number of zeros with real part ≥ σ up to height T is o(T), against roughly (T/2π)·log T zeros in total — so such a family would be a vanishing proportion of all zeros, but no theorem excludes one. A short note that states that question carefully, surveys what density estimates do and do not give, and says plainly that the answer is unknown, is a real expository contribution and a much stronger paper than the reinterpretation. It also needs none of the geometry. Exploratory (no literature search was run on this question here; it may well be known, and that is the first thing to check.)
Our honest reading, and it is about the destination rather than about the author. We should say first where we are standing: this programme is itself blocked at arXiv, for the same reason — we have no endorsement either. What follows is not advice from a height we occupy.
1. Look at the objects before theorising about them. This site carries a free explorer — the interactive explorer, desktop only — that draws zeta and any Dirichlet L-function in the strip, computes the coefficients in the browser, and locates the zeros by the sign change of the real-valued Hardy function while printing the functional-equation residual beside the count. It will not draw Davenport–Heilbronn, which is a combination of two characters rather than one, but it draws the geometry the note is reasoning about, and it costs nothing. We say this from experience rather than as advice: several of this programme's better observations, and more than one of its corrections, started at a picture on that screen and not at a sheet of paper.
2. Generalise the fixed-line reading properly — it is a correct short paper. Run the note's own construction across the family: for a Dirichlet L-function of conductor q the strip is still 0 < σ < 1 and the axis is still ½, with the conductor entering the completed function Λ(s) = (q/π)(s+κ)/2Γ((s+κ)/2)L(s) and changing the density of the zeros but not the line; for ζ(2s) the strip moves to 0 < σ < ½ and the axis with it, to ¼. Written out carefully, with the reflection each completed function satisfies, that is a correct, honest, self-contained note, and it teaches exactly the machinery any next step needs. Steuding's Value-Distribution of L-Functions (Lecture Notes in Mathematics 1877) is a good place to learn the normalisations from — we hold it first-hand and it is written for exactly this entry point.
3. And the place where a geometric instinct genuinely has purchase: how a zero leaves the line. A zero does not wander off the critical line; it collides. Along a continuous family of Dirichlet series carrying one functional equation, two zeros on the line meet and split off it as a mirror pair — the mechanism is Balanzario and Sánchez-Ortiz's (2007), the velocity of a zero along the family is the implicit function theorem, and we measured forty such departures with the generic square-root law, exponent 0.481 with interval [0.436, 0.503]. The local picture is thoroughly geometric and it is worth drawing: fix the height of an off-line pair and walk σ from one member across to its mirror. The completed function leaves the origin, travels out, and returns to the origin — a closed lens whose far tip is exactly σ = ½ and lies on the real axis (imaginary part 1.6×10⁻³⁰ of the modulus), whose two halves are mirror images of equal length (arc length 1.052012 each, difference exactly 0), and whose height equals the near-collision margin of the zeros on the line to ten digits. A zero on the critical line is that lens closed. Measured
The obvious next sentence — "an L-function never has such a lens" — is not an independent structural fact and will not become one. Unpacked, it says the real-valued Hardy function never dips toward the axis without crossing it, which is the Riemann hypothesis for that function, restated. It has the same shape as Speiser's criterion and as Li's: a faithful portrait of what an off-line zero is, never a condition forbidding one. This programme reached that wall from three directions before recognising it. Knowing it in advance is worth more than every correction above.
Write the falsifier first. Before a picture becomes an argument, write down — on paper, before computing anything — what measurement or what example would show it false, and then go and run it. The corollary costs nothing and settles most questions in an afternoon: run any new criterion against Davenport–Heilbronn before running it against zeta. If it cannot tell those two apart, it cannot tell you where a zero is. Most of what this programme knows was bought that way, and the rest was bought by neglecting the rule and paying for it afterwards.
You reached, on your own, a true statement about why the number is one half: the critical line is the fixed line of the reflection the functional equation supplies, and equidistance is a correct metric way of seeing it. That is not a small thing to arrive at unaided, and your own construction passes a test you can run yourself — it returns one quarter for ζ(2s), and one quarter is right.
What fails is the step from the axis to the confinement, and that step is not a lapse in your note. It is the entire problem. Over eight papers this programme built some seventy exact objects aimed at that step and every one failed in the same place for the same structural reason, which we only understood late: symmetry gives the axis, the arithmetic is what everyone believes gives the confinement, and nobody knows how to convert the second into a statement about where a zero can sit. You are stopped by the thing that stops everyone, and you are stopped at a well-chosen place.
So the note is worth rewriting rather than abandoning. Cut the claims back to what is proved, keep the picture, add the ζ(2s) check as your own evidence that it transports, and add Davenport–Heilbronn as the honest reason it stops there. That version is correct, it is useful to anyone meeting the subject, and it is a document you can put your name to without a caveat. The hardest habit in this area is not having ideas but pricing them, and you have already done the difficult half of that by writing "We do not claim a proof" into your own abstract. Keep it there, and keep going.
The picture is real, and it is right for a better reason than the note gives. The equidistant line is the fixed line of the functional equation's involution, and the construction proves it by transporting correctly to ζ(2s), where it returns ¼ and is right to. Anyone who arrives at that by himself has understood something true about why the number is ½ and not another number.
The argument does not reach the conclusion, and the gap is one sentence long. It is the step that joins the geometry's distinguished line to the uniqueness clause's distinguished line by calling them both σ. Below it sit three repairable items — discreteness is the identity theorem, the compactification is automatic, and the oscillation classically pushes the other way, bringing |ζ| arbitrarily close to zero on every line in the strip — and above it sits a scope correction: the final display is implied by the Riemann hypothesis rather than equivalent to it.
And there is a function that decides it. Davenport–Heilbronn satisfies every geometric hypothesis the note uses and has 386 verified zeros off the critical line, 570 pairs of them below height 10⁴ on our own instrument. That is not a defect in the note's execution; it is the wall the whole subject sits behind, and eight papers of ours are on the same side of it.
A Projective Zeta Geometry Interpretation of the Critical Line: Euler Products, Transverse Displacement, and a Proposed Arithmetic Confinement Principle, Kazushi Mizutani, fifteen pages, twenty sections, dated 18 August 2026 — one day after the review above. It keeps the abstract's own ceiling in the same form ("The present note does not claim a proof of the Riemann Hypothesis") and it goes at the objection directly.
The argument, restated as faithfully as we can. Define the transverse displacement of a non-trivial zero, δ(ρ) = Re(ρ) − ½, which is zero exactly when ρ is on the critical line. Observe that the functional equation makes the linear sum of displacements cancel in reflected pairs, so a linear quantity cannot separate the on-line case from the off-line one. Go quadratic instead: define the transverse energy
E⊥[W] = ∑ρ ( Re(ρ) − ½ )2 W(ρ), W ≥ 0
and note that if it vanishes for a strictly positive weight then every term vanishes and every zero is on the line. Then propose — as a conjecture, explicitly labelled as one — that the Euler product might be what forces the vanishing, with the logarithmic derivative −ζ′/ζ(s) = ∑ Λ(n) n−s as the bridge from the zeros to the prime powers, and an explicit formula as the mechanism.
Discussion: A Projective Zeta Geometry Interpretation of the Critical Line · Academia.edu · 15 pp, 20 sections · abstract, §1 introduction, §2 the compactified zero space, §3 transverse displacement, §4 why symmetry alone is not sufficient, §5 a proposed transverse energy, §6 the Euler product as arithmetic structure, §7 the logarithmic derivative and the von Mangoldt function, §8 an Euler-induced potential, §9 zeros and the logarithmic derivative, §10 from transverse geometry to arithmetic geometry, §11 the proposed arithmetic confinement principle, §12 relation to the one-point compactification, §13 the "diagonal escape" problem, §14 a possible PZG energy interpretation, §15 why the quadratic form is important, §16 connection with explicit formulae, §17 a possible hierarchy of the research program, §18 limitations and open problems, §19 comparison with the original geometric picture, §20 conclusion.
1. He accepted the corrections and states them himself. The compactification argument is not defended — it is retired, in his own words, in his own §1:
However, this observation by itself does not distinguish between a sequence lying on the critical line and a hypothetical sequence lying away from the critical line. In particular, one-point compactification alone does not imply the Riemann Hypothesis.
He says it again at §12 ("Thus topology alone does not distinguish these possibilities") and a third time in his own list of limitations. The "second point at infinity" is gone, and the "diagonal escape" is now labelled a problem rather than a mechanism. Part III's second item above therefore now stands by his agreement rather than against him, and there is nothing left to argue about there. Conceded
2. He reached this programme's own coordinate, unaided. His Definition 3.1, δ(ρ) = Re(ρ) − ½, is the displacement coordinate this project has used throughout its census work under the name dσ = σ* − ½ — the same object, the same normalisation, arrived at independently — and on the dated record there is nothing of this programme's that he can be shown to have had in front of him when he did it. That is a real convergence and it is worth saying plainly. Measured (ours: the distribution of dσ·log t on ζ's 1-points over seven windows, pooled skew −0.713, and −0.443 after the boundary trim. One free hint that comes with it: the coordinate is only scale-free after multiplying by log t, because the mean spacing of the zeros at height t is 2π/log(t/2π). Any weight W will have to respect that scale. Exploratory — our normalisation is measured on 1-points, not on zeros, and nothing about it transports automatically.)
3. His §4 and §15 ask exactly the right question, and he got there on his own. A quantity inherited from the functional equation cancels in reflected pairs, so an unweighted linear invariant cannot separate the two cases. That is the same shape as the screen in Part V above — an equality inherited from the functional equation cannot separate two functions that share it — reached from the other direction and without being told. The inference he then draws, that one must therefore go quadratic, is the one clause in the note we would tighten, and the reason is more interesting than the clause: a weighted linear sum already works, and has for years. Going quadratic turns out to buy something else, and something real.
4. The gap was hidden; now it is stated. Part III's fourth item called the combining step the load-bearing gap and noted it was one sentence long and unmarked. In the second note there is no combining step to hide: §18 is his own numbered list of the five things that are missing, and the fourth of them is exactly the one carrying the whole weight ("the equality E⊥[W] = 0 has not been established in this note"), repeated in the closing acknowledgment. That is a real improvement in hygiene even where the mathematics has not advanced, and it is the habit that separates a research note from a claim.
The first note's final display was weaker than the Riemann hypothesis: it forbade only an infinite family of zeros sharing one vertical line, and off-line zeros need not share one. That scope problem is gone. For any weight strictly positive on the zeros, E⊥[W] = 0 holds if and only if every zero is on the critical line — term by term, in two lines, in both directions. The new statement is equivalent to the hypothesis rather than implied by it, and that is exactly the repair Part III's fifth item asked for. Verified
And because it is an equivalence, it can be tested. The natural test is the one this page has been recommending since its first paragraph: run it against Davenport–Heilbronn before running it against zeta.
His §18 lists convergence as the first difficulty. It is the easy one and it costs two lines: |Re(ρ) − ½| < ½ at every non-trivial zero, so E⊥[W] ≤ ¼·∑ρ W(ρ), and it is enough that ∑ W converges. W(ρ) = |ρ|−2 does, and it is not an arbitrary choice — it is the weight class the classical theory already runs on (see Part X). So we took that weight and evaluated his quantity on the certified Davenport–Heilbronn ledger this project holds.
Weight W(ρ) = |ρ|−2, over 193 certified off-line quartets to height 4000 — 386 members, every one verified:
| quantity | value |
|---|---|
| certified off-line members read (Im > 0) | 386 |
| highest ordinate in the ledger | 3992.648801 |
| E⊥[W] over those members | 5.03810655514×10−5 |
| with their conjugates at −t | 1.00762131103×10−4 |
| the first off-line quartet alone, t = 85.699… | 5.18374051×10−5 |
| — its share of the whole ledger's energy | 51.4453 % |
| CONTROL — the same functional over the same 386 ordinates, with every real part moved onto the line | 0.0 exactly |
The control is what makes it a test rather than an assertion: a functional that returns a positive number on everything is not measuring anything. Measured (exact arithmetic on a verified ledger; script and log behind every figure.)
This certifies a finite window — 193 quartets to height 4000, above that census's detection floor — and forbids nothing outside it. It is a strict lower bound on that function's transverse energy at that weight, not its value. It says nothing whatever about where a zero of ζ is.
Davenport–Heilbronn carries every ingredient his Stage I uses — the same strip, the same s ↦ 1−s completed symmetry, the same discrete zero set running off to infinity — and it has no Euler product. His observable is defined for it verbatim, and it returns a number that is not zero. That is not a defect in his criterion. It is the criterion working. It is also precisely why the criterion cannot, on its own, forbid anything: a correct equivalence tells you what an off-line zero would look like; it does not tell you that there isn't one.
Applied to his six-stage plan, the same test says something sharper and more useful. Stage I goes through for Davenport–Heilbronn. Stage III goes through, and returns 1.0076×10−4. Stage II is the only stage that cannot be executed, because that function has no Euler product. So the entire weight of the programme rests on the one step that does not yet exist — which is a thing he can confirm for himself in an afternoon, and the concrete witness is one zero:
σ = 0.8085171824566374, t = 85.69934848537759
That single quartet carries 51.4453 % of the whole ledger's transverse energy, so a number of the right order is reachable from one zero and a pocket calculator. Verified
E⊥[W] is a sum of non-negative terms, so E⊥ ≥ 0 is free and carries no information; and E⊥ = 0 is, by his own §5, the Riemann hypothesis. His Stage V asks a positivity principle to deliver vanishing — but positivity supplies a lower bound, and vanishing needs an upper one. No positivity principle supplies one.
This is the same wall Part VI above flagged in advance and this programme paid for three times: a non-negative quantity whose vanishing is the theorem. It is worth stating as the one transferable sentence in this half of the page:
The sign condition has to live on the side you can compute without already knowing the answer. In the classical criteria the functional sits on the prime and archimedean side, where it is computable from the primes alone, and the hypothesis is the statement that this computable thing has a definite sign. Here it sits on the zero side, where the sign is automatic and the content is nil.
Stages III through VI are the Weil–Li programme, and it is not a sketch. Weil's criterion, in Bombieri's official Clay problem description for the hypothesis, is the statement that RH is equivalent to a definite sign of the arithmetic side of the explicit formula for every test function of the form f(x) = ∫0∞ g(xy)g(y) dy with two vanishing moments on g. Conceded (Bombieri, "Problems of the Millennium: the Riemann Hypothesis", Clay Mathematics Institute — read here first-hand. Note the sign in that arrangement is negativity, because the sum over zeros sits on the left of his identity.)
Li's criterion is the same idea one level down and it is concrete. With
λn = ∑ρ [ 1 − (1 − 1/ρ)n ]
the hypothesis holds if and only if λn ≥ 0 for every n ≥ 1. And Bombieri and Lagarias then proved the arithmetic half that his Stage IV asks for: their Theorem 2 converts λn into an explicit sum over Λ(n), which is exactly zero-side quantity expressed on the prime side. Stage IV, for that family, was done in 1999. Conceded (Bombieri and Lagarias, "Complements to Li's criterion for the Riemann hypothesis", J. Number Theory 77 (1999) 274–287 — read here first-hand from the author's own reprint. Li's 1997 paper itself we could not obtain; we hold its statement at statement level, pinned through Bombieri and Lagarias's verbatim quotation of it and through Lagarias's later survey, and we say so rather than implying we read it.)
His instinct, in other words, is the instinct the subject already had, and it was right. What v2.0 can add is where the wall then is.
Bombieri and Lagarias state their own result more sharply than a summary would: their positivity theorem is proved for an arbitrary multiset of complex numbers, with two hypotheses — that the multiset omits 1, and one convergence sum — and nothing else. Their abstract says it in their own words: the criterion "is not specific to zeta functions". The version that gives the critical line needs one thing more, and it is a symmetry rather than an arithmetic input: that the multiset is closed under ρ ↦ 1−ρ and under conjugation. Conceded
Which means the screen of Part V applies to the classical criterion exactly as it applies to the new note's. Davenport–Heilbronn's non-trivial zeros satisfy those hypotheses — its functional equation supplies the reflection, its real coefficients supply the conjugation — so Li's criterion holds verbatim for it, and there it correctly reports that some λn is negative. Exactly as his transverse energy correctly reports 1.0076×10−4. Exploratory (the applicability is checked against that function's symmetry and counting law, not against a printed theorem for it. No λn was computed here and none could be: λn is a sum over the whole zero set, and what we hold certified is the off-line part of it. The negativity is what Bombieri and Lagarias's theorem gives on a multiset with an off-line member, not a measurement of ours.)
That is not a criticism of Li's criterion and it should not be read as one. It is the observation that in both schemes the equivalence is free, so the arithmetic has to enter at the evaluation — and in Li's case it does, in Bombieri and Lagarias's Theorem 2, and then the programme stops. Proving the arithmetic positivity is open, for everybody, and has been for thirty years. Being stopped there is a much better thing to be told than being told the note is wrong.
His §10 asks for a relation of the form E⊥[W] = P[W] + A[W], with P a prime contribution in Λ(n) and A an archimedean one. That is exactly the shape of the Weil–Guinand explicit formula, so Stage IV is not virgin ground. But there is an obstruction to getting his particular summand out of it, and it is worth stating carefully because it points at the repair.
In every form of the explicit formula, the zero side is a sum over zeros of a test function that must be holomorphic on a region containing the closed critical strip — Bombieri's Clay description evaluates the Mellin transform at the zero itself, analytic for −δ < Re(s) < 1+δ; the Fourier-variable form evaluates a function analytic in a horizontal strip of half-width above ½ at (ρ−½)/i. The half-width above ½ is there precisely so that possible off-line zeros are covered. Conceded (held first-hand: Bombieri's Clay problem description; Lagarias, arXiv:math/0404394 §3, "if the Riemann hypothesis is not assumed, the set of test functions must be further restricted to functions analytic in a region containing the closed critical strip".)
Now (Re ρ − ½)2 = ((ρ + ρ − 1)/2)2 involves the conjugate, so it is not a holomorphic function of ρ anywhere, and no decay condition repairs that. (Two obstructions live near each other and a referee will separate them: the non-holomorphy above is what bites the displacement; a plain t2 on the line is entire, and what stops that one is decay — the sum diverges at zero density (T/2π)log T. Different obstruction, same verdict.)
The structural reason underneath is the useful one. The transverse energy is a quadratic functional of the zero set, and the explicit formula is a linear functional of one holomorphic test function. A quadratic reach exists, and it is exactly Weil's pairing, which is sesquilinear rather than linear:
⟨ F, G ⟩ = ∑ρ F(ρ) · conj[ G(1 − ρ) ]
On the critical line 1 − ρ = ρ, so ⟨F,F⟩ = ∑ |F(ρ)|2 is automatically non-negative; off the line 1 − ρ ≠ ρ and the terms are no longer non-negative. That is why Weil's criterion is a positive quadratic form and not an energy that vanishes — and it is the repair: the object he wants exists, it is quadratic in the test function rather than quadratic in the displacement, and its positivity rather than its vanishing is what is equivalent to the hypothesis. Conceded (Lagarias, arXiv:math/0404394 §3, first-hand.)
Write ρ = ½ + iw, so Re(ρ) − ½ = −Im(w), and the quadruple {ρ, 1−ρ, ρ, 1−ρ} becomes {w, −w, w, −w}. An even holomorphic test function, real on the reals, therefore reads that quadruple as 4·Re h(w) — a symmetric analytic function and nothing else. Take the simplest quadratic, h(w) = w2, and evaluate it on the first Davenport–Heilbronn off-line quartet:
| reading | value |
|---|---|
| 4·Re(w2), off the line | 29377.1325919 |
| the same at the same height, on the line | 29377.5133233 |
| difference, off minus on | −0.3807314075 |
| the quantity actually wanted, 4(Im w)2 | +0.3807314075 |
The best available analytic quadratic is off by exactly minus the transverse energy: an off-line zero makes it smaller. It rewards the departure instead of penalising it, which is the precise sense in which Stage IV cannot be done as written. Measured (we have run no literature search on whether this linear-versus-quadratic obstruction is stated anywhere; treat it as folklore made explicit, not as a discovery of ours.)
Reorder the difficulty list. §18's first item, convergence, is free: the bound above discharges it, and W(ρ) = |ρ|−2 works. It is also the right weight for a second reason — Bombieri and Lagarias's own convergence hypothesis is of exactly that class, with their remark that it relaxes to ∑ 1/(1+|ρ|)2 < ∞. So taking that weight puts Stage III inside the standard framework rather than beside it, and it costs one line. The hard items are his second, third and fourth. Verified
One clause to tighten — and it turns out to be the most interesting thing in this part. §4 and §15 say the linear sum cancels under σ ↦ 1−σ. That needs the weight to be reflection-invariant, W(ρ) = W(1−ρ) — and |ρ|−2 is not. So at the very weight that discharges the convergence problem, the linear sum does not cancel.
And it does not merely fail to cancel. It is an exact criterion, and a classical one. Pair a zero with its mirror: if Re ρ > ½ then (Re ρ)2 > (1−Re ρ)2, so the mirror sits closer to the origin and therefore carries the larger weight, and the pair contributes a strictly negative amount. Hence
∑ρ ( Re(ρ) − ½ ) |ρ|−2 ≤ 0 always, = 0 if and only if every zero is on the line.
For zeta the unconditional half of that is known in closed form — ∑ρ Re(1/ρ) = 1 + γ/2 − ½log(4π) = 0.023095708966121034 — so the criterion reads: the hypothesis holds if and only if ∑ρ |ρ|−2 = 2 + γ − log 4π, and if it fails the sum is strictly larger. Conceded (Gun, Murty and Rath, "Transcendental sums related to the zeros of zeta functions", arXiv:1807.11201, eqs. 2–4 — read here first-hand, proof included; the sign of the subtracted form re-derived here.)
So the quadratic was not forced. A linear transverse quantity at the weight that makes his own Stage III converge is already an exact two-sided criterion, and it has been on the record for years. That is worth knowing before building six stages on the premise that only a square will do.
But it is the same wall, which is the honest half. That classical criterion has exactly the structure of his transverse energy — the sign is automatic on the zero side, and the vanishing is the hypothesis. It forbids nothing either, for precisely the reason Part X gives. It is not an escape route; it is independent confirmation that his instinct landed on a real classical object, sitting behind the same wall as everything else.
Both criteria are exact, so the fair comparison is how loudly each one speaks. On the same certified Davenport–Heilbronn ledger, at the same weight |ρ|−2:
| observable | value on Davenport–Heilbronn | control, on the line |
|---|---|---|
| classical weighted linear sum | −4.04982244779×10−9 | 0.0 exactly |
| his transverse energy | 5.03810655514×10−5 | 0.0 exactly |
| ratio | 12,440 | |
The linear criterion fires by a near-cancellation between a zero and its mirror; the quadratic adds the two contributions instead of subtracting them. At this weight the transverse energy is the better-conditioned observable by four orders of magnitude, even though both are exact. That is a point in the note's favour and it should be said as one. Measured
The potential of §8 lives in the wrong half-plane. V(σ,t) = Re log ζ(σ+it) is given by the prime-power series only for σ > 1, which is exactly the half-plane containing no zeros. To reach the strip it has to be continued analytically, and past σ = 1 it is no longer the sum of the arithmetic contributions the picture draws. The distance between Stage II and Stage IV is that continuation, and the continuation is the classical difficulty rather than a notational one. Verified
§6 says the critical line corresponds to the distinguished amplitude scale p−1/2. The Euler product converges for σ > 1 and says nothing at ½; nothing inside it distinguishes one half. But the instinct is half right, and the half that is right is already on this page: the Euler product is what puts the right-hand edge of the strip at σ = 1, and one half is the reflection of that edge under s ↦ 1−s. So arithmetic does place one of the two boundaries; the functional equation does the rest. That is Part II's point about the boundary lines, arriving from the other side. Conceded
The note's Stage II consumes "ζ has an Euler product". That is not enough on its own, and the cleanest way to see it is that the two ingredients fail separately:
| object | functional equation | Euler product | zeros |
|---|---|---|---|
| Davenport–Heilbronn | yes | no | 386 verified off the line, ours |
| Beurling generalized primes (Diamond–Montgomery–Vorhauer) | no | yes | infinitely many at σ = 1 − a/log t |
| ζ | yes | yes | open |
A Beurling system is a zeta function defined by an Euler product over an arbitrary unbounded multiplicative system; what it lacks is a functional equation, any additive structure among its "integers", and analytic continuation except as bought by a regularity hypothesis. Diamond, Montgomery and Vorhauer built one whose integers are very regularly distributed — N(x) = κx + O(xθ) with ½ < θ < 1 — and whose zeta function nevertheless has infinitely many zeros on the curve σ = 1 − a/log t. Conceded (Diamond, Montgomery and Vorhauer, "Beurling primes with large oscillation", Math. Ann. 334 (2006) 1–36; we hold the theorem at statement level from the authors' own abstract, pinned through two independent mirrors and restated as a zero result by three later papers we did read. The 36-page paper itself we have not read.)
But "an Euler product is inert" would be false, and we are not saying it. The Euler product is what forbids zeros with σ > 1 in the first place. And at degree one the constraint is much stronger than that: without the Euler product axiom, the degree-one elements of the extended Selberg class are the periodic-coefficient series, a class that contains Davenport–Heilbronn; add the Euler product axiom and the class collapses to a single shifted Dirichlet L-function. At degree one, the Euler product removes every known counterexample. Conceded (Kaczorowski and Perelli, the degree-one classification, arXiv:math/0306300, read here first-hand.)
So the honest statement is that an Euler product without a functional equation and without the additive rigidity of the ordinary integers does not confine zeros to a line — and that Stage II therefore has to consume something sharper than the phrase. Which property of Λ does Stage V need? Its non-negativity? Its support on prime powers? The multiplicativity underneath it? Naming that property is the whole job, and it is a concrete next step rather than a verdict.
The criterion is correct, and that is a real repair. The first note's final display was strictly weaker than the hypothesis; the transverse energy is equivalent to it, term by term, in both directions. The compactification argument has been retired by its own author, the displacement coordinate is the right one, and the missing step is now written down instead of hidden. Four things got better in one day.
The conjecture, unpacked, is the hypothesis. For a strictly positive weight, E⊥[W] = 0 is the statement that every zero is on the line. So Conjecture 11.1 reads, once unpacked, "the hypothesis is true and follows from the Euler product". That is the same shape as the trap this page printed in advance in Part VI — a non-negative quantity whose vanishing is the theorem — and the note has walked into a well-marked place rather than a careless one.
And the plan itself is thirty years old, which is the good news. Stages III to VI are the Weil–Li programme; Stage IV exists for Li's coefficients, published in 1999; and the equivalence there is arithmetic-free in the same way his is, so the whole difficulty concentrates in Stage V. That is where everybody is, and it is a much better place to be stopped than the one the first note stopped at.
You turned a review round in a day, dropped the two steps that did not carry, kept the one that did, and reached the same displacement coordinate this project uses. That is the right way to take feedback and it is rarer than it should be.
The most useful thing here is probably the smallest: your Stage III, run on Davenport–Heilbronn, returns 1.0076×10−4 instead of zero, and you can reproduce a number of that order from the single zero printed above. Stage I goes through for that function and Stage III goes through; Stage II is the only one that does not, because it has no Euler product. So the whole programme rests on the one step that is not built yet — and knowing that before building the rest is worth more than any verdict on this page.
If you want the sharpest version of your own question, it is this: which property of Λ(n) does Stage V need — that it is non-negative, that it is supported on prime powers, or the multiplicativity underneath both — and does that property fail for Davenport–Heilbronn? That question is answerable, it needs none of the geometry, and nobody has to grant you anything to let you work on it.
On 21 August 2026 the author posted a revision, A Projective-Zeta-Geometric Approach to the Riemann Hypothesis: From the Euler Product to a Transverse Energy Functional — twelve pages, eighteen sections, and nine references, where the note reviewed in Parts VIII–XII carried none. On 8 September he posted Projective Zeta Geometry (PZG): A Geometric and Logical Framework for Algebraic and Transcendental Numbers — three pages, seven sections, no references. Both were read here in full from the author's own PDFs.
The two notes are in different subjects, and this review treats them differently. The August note is about the zeros of ζ, and the certified counterexample of Part IV applies to it directly. The September note is about algebraic and transcendental numbers, where that ledger has nothing to say — and aiming it there would be the exact error this page has spent two versions warning against. Part XVI is written without it.
It cites. Nine references, including Weil, Li, Bombieri and Montgomery. That is the difference between a private document and one a reader can follow.
It separates the established from the conjectural, explicitly. Its §14.1 lists seven standard ingredients, §14.2 five proposed interpretations, §14.3 six things not proved. Most notes in this area do not contain that table, and writing it is how an author finds out what the paper is resting on.
It states the right distinction, and states it correctly. Its §15 is titled A Critical Distinction: Positivity versus Vanishing: that a non-negative transverse energy does not imply the hypothesis, that its vanishing would, and that the second task is substantially stronger than the first. That is exactly right, and Part XIV is about how much stronger.
And an argument was withdrawn rather than patched. The claim of the previous note that a linear transverse sum must cancel under s ↦ 1−s is simply gone; the revision says only that the displacement changes sign and its square is reflection-invariant, both of which are true. A correction had been written here for that argument and is no longer needed. That is the second time this author has removed a step instead of defending it, and it is the habit that matters most.
The August note's chain runs from the Euler product through the logarithmic derivative, a contour
functional, a Weil-type quadratic form, a proposed transverse projection P, and finally
the transverse energy. Its central open problem — its Conjecture 16.1 and its Research Questions
17.1 to 17.3 — is the construction of P.
Each step was put to one question: does it consume a property of the coefficients Λ(n) — their values, their multiplicativity, their positivity, their support on prime powers — or does it use only analyticity, the functional equation and the strip?
| step | what it consumes | a property of the coefficients? |
|---|---|---|
| §2 the transverse coordinate | the number one half | no |
| §3 the logarithmic derivative | the Euler product identity | names Λ, and no later step uses a property of it |
| §4 the contour functional | holomorphy, a contour shift | no |
| §5 passage to a quadratic form | the explicit formula | no |
| §6 the transverse energy | the displacement and the test function | no |
| §8 Mellin differentiation | that n is a real number greater than one | no |
| §9 the stated obstruction | algebra | no |
| §10 the two symmetrisations | s ↦ 1−s and conjugation | no |
| §11 a self-adjoint operator | aspiration | no |
| §13 the longitudinal / transverse split | a decomposition | no |
Every operation in that chain — the transverse coordinate, the Mellin differentiation, both symmetrisation operators, the passage to a quadratic form, the longitudinal/transverse split — is defined word for word for the Davenport–Heilbronn function of Part IV. It has the strip, the completed symmetry about one half, complex conjugation, a logarithmic derivative, an explicit formula, and no Euler product. And it has zeros off the line.
Therefore, if Conjecture 16.1 can be proved for ζ by those means, the same proof
constructs P for Davenport–Heilbronn — where it would return a
non-zero transverse energy, correctly, and forbid nothing. Part IX already measured what that number
is: 1.00762131103×10−4, against a control of exactly zero when
every real part is moved onto the line.
That figure certifies a finite window — 193 certified quartets to height 4000, above that census's detection floor. It is a strict lower bound on that function's transverse energy at that weight, not its value, it forbids nothing, and it says nothing whatever about where a zero of ζ is.
The consequence sorts the note's own open-problem list into two groups that are not
comparable. Its §14.3 lists six. The first four — a rigorous definition of
P, the identity it must satisfy, deriving the squared displacement rather than the full
complex one, and the archimedean, pole and trivial-zero terms — are all about constructing
P. They are real mathematics, they may well be achievable, and every one of them
is blind to the difference between ζ and a function with off-line zeros. The remaining
two — that the transverse energy vanishes, and that every displacement is zero — are the
Riemann hypothesis.
The note's own §15 says the second task is substantially stronger. The measurement says something sharper: the first task is not merely weaker, it carries no information about ζ at all. Research Questions 17.1 to 17.3 can be answered completely and leave 17.4 exactly where it stands.
The August note's §9, which it labels central rather than technical, observes that differentiating in the real direction produces the full complex displacement rather than its transverse part alone. That is the obstruction this page stated and measured at Part XI — and it is worth saying plainly that he arrived at it himself, in his own notation, and marked it as the thing that blocks the construction.
The reason, which Part XI gives in full, is that the squared transverse displacement involves the conjugate and so is nowhere holomorphic, while the explicit formula is a linear functional of one test function required to be holomorphic on a region containing the closed critical strip. No choice of test function repairs that. The best analytic substitute, measured at Part XI, reads the first off-line quartet smaller by exactly the transverse energy — it rewards an off-line zero instead of penalising it.
The repair is unchanged and is the most useful thing this page can offer that note: the object that reaches quadratically is Weil's pairing, which is quadratic in the test function rather than in the displacement. Its term at a zero pairs that zero with its mirror across the critical line; on the line the two coincide and the term is a squared modulus, non-negative automatically, and off the line they do not and the terms stop being non-negative. The transverse information is already inside Weil's form — it is the failure of the pairing to be diagonal. The note asks instead for a diagonal positive weight, which depends on the test function at one point per zero, where Weil's form depends on it at two.
The PZG note proposes that algebraic numbers form a geometric boundary and transcendental numbers an interior, and asks for a non-circular logical distance whose zero set is exactly the algebraic numbers — defined without referring to the predicate "is algebraic". Its §2 observes, correctly and on its own, that the algebraic numbers are dense, so no ordinary continuous representation separates them. Its §3 proposes building the distance from information layers, double-negation structure, rational computability, or descriptive complexity. Its §7 asks for criticism, counterexamples and related literature, and says the programme's goal is falsifiable: either construct a non-circular core, or identify a rigorous obstruction to it.
Mahler's classification of the complex numbers is a real-valued function whose zero set is exactly the algebraic numbers, and it is non-circular in precisely the sense the note asks for. For a complex number and integers n and H, one takes the least non-zero absolute value of an integer polynomial of degree at most n and height at most H evaluated there, and builds from its growth a single quantity. Mahler partitioned the complex numbers into four classes by its behaviour, and the class on which it vanishes is exactly the algebraic numbers. Koksma's 1939 classification, by approximation by algebraic numbers of bounded degree and height, agrees with it.
The definition quantifies only over integer polynomials of bounded degree and height. It never asks whether the number is algebraic. It is a logical distance whose vanishing characterises algebraicity without using algebraicity as an oracle, and it has existed for ninety years.
Tier. Held here at statement level, from the standard reference — Bugeaud, Approximation by Algebraic Numbers, Cambridge 2004, chapter 3 — not from Mahler's paper first-hand.
It works because it consumes integrality and height: the coefficients are integers and H is their height. That is not special to Mahler. Liouville consumes a degree-and-denominator bound; Hermite–Lindemann, Gelfond–Schneider and Baker an auxiliary function with integer coefficients plus vanishing-order and size estimates; Apéry a denominator estimate. If no step consumes integrality, degree, height or a denominator, no transcendence conclusion follows, whatever the geometry looks like. That is the exact counterpart, in the new subject, of the screen this page has been applying in the old one.
Two of the note's four proposed substrates can be closed in two lines each, and they are offered in the spirit in which its §7 asks for them. Computability cannot separate algebraic from transcendental: π is a computable number, so is e, so is every algebraic number, and the computable reals form a field properly containing the algebraic numbers — any distance built from computability vanishes on π. Descriptive complexity cannot either, and more sharply: the Kolmogorov complexity of the first n digits of π is of order log n, the same as for the square root of two, because a short program generates both. Complexity separates computable from non-computable, and that line runs straight through the transcendental numbers.
A third obstruction is more general and is the exact analogue of the compactification ruling of Part III, which this author has already accepted in his own text: assuming the axiom of choice, the complex field has automorphisms fixing every algebraic number and moving every transcendental one, so no construction invariant under the field automorphisms of the complex numbers can decide transcendence. An invariant that everything in the class shares cannot distinguish one member of it — the same sentence, in a second subject.
The note joins "algebraic and transcendental numbers" to zeta geometry through the word projective. There is a real join, it is ninety years old, and it descends from the same two authors as the function in Part IV.
For the Hurwitz zeta-function ζ(s,α), Davenport and Heilbronn proved in 1936 that there are infinitely many zeros to the right of the line of absolute convergence when α is transcendental, or rational and not one or one half; Cassels extended this in 1961 to algebraic irrational α — and that original proof contained an error, corrected later. Inside the strip the question is much harder, and to this day zeros there are known only for α transcendental or rational and not one or one half. The algebraic irrational case is open; that gap is essentially Gonek's conjecture, from his 1979 thesis.
The mechanism is the point. If α is transcendental, the logarithms of n+α over distinct non-negative integers n are linearly independent over the rationals, which makes the Kronecker–Weyl theorem available and supplies an equidistribution step that the proof needs. That is a worked example of transcendence of a parameter doing structural work in a zeta-function geometry, through a specific and checkable mechanism — not an analogy, and a live open problem.
ζ(s,1) = ζ(s) is precisely the excluded parameter of that whole line. Nothing in the Hurwitz material above says anything about the Riemann zeta function, and nothing in it bears on the Riemann hypothesis. It is a statement about the Hurwitz family away from that parameter, and it is offered because it is the honest home of the note's title — not because it is a route to anything.
Tier. The survey by Sh. Mine, New developments toward the Gonek conjecture on the Hurwitz zeta-function, arXiv:2305.01262, was read first-hand. Everything that survey reports stays at statement level, pinned through it — Davenport–Heilbronn 1936, Cassels 1961, Gonek's thesis, Garunkštis 2005, Sourmelidis–Steuding 2022 — and none of them has been read here. Further reading is named in that order.
Part VI gave one rule for the zeros: run any new criterion against Davenport–Heilbronn before running it against zeta; if it cannot tell those two apart, it cannot tell you where a zero is. Here is the counterpart for transcendence. Before believing any construction, run it on the register of known answers.
| object | status |
|---|---|
| π, e | transcendental — Hermite 1873, Lindemann 1882 |
| ζ at even positive integers | transcendental, and only via Lindemann |
| two to the power root two | transcendental — Gelfond–Schneider |
| π, e to the π, Γ(1/4) | algebraically independent — Nesterenko 1996 |
| ζ(3) | irrational (Apéry 1978); transcendence OPEN |
| ζ(5), ζ(7), … | irrationality of each one OPEN |
| Euler's constant | irrationality OPEN |
| π+e, πe, π to the e | ALL OPEN |
If a construction delivers the bottom four rows by the same move that delivers the top ones, the move is wrong, or it is conditional and the condition is the whole content. A framework must reproduce the theorems; it must not settle the open problems for free. For π+e and πe what is actually known is only that at least one of the two is transcendental — a two-line argument from the fact that π and e are the roots of a quadratic with those two as coefficients.
On the August note. The hygiene improved again, the missing step is now named precisely, and the author has correctly identified both the logical distinction his programme turns on and the obstruction that blocks its construction. The verdict of Part IX is unchanged and is now sharper: the criterion is correct, it works on a function where the hypothesis fails, and therefore it forbids nothing by itself. What is new is that the construction he has set as his target is blind to that distinction too. The distance to the hypothesis has not shortened.
On the September note. It is an honest research announcement that states its own status, asks for criticism and names its own falsifiable goal. The object it asks for exists and is ninety years old; two of its four proposed substrates fail against π; and the place where its two subjects genuinely meet is a named open problem it does not cite. None of that is a refutation — it is the literature the note asked for.
And the observation that spans both, which is worth more than either: in August the equivalence at the heart of the programme turned out to be arithmetic-free and thirty years old, with all the content in the evaluation. In September the distance turned out to exist since 1932, with all the content in evaluating it — the value of that distance at ζ(3) is exactly as unknown as the transcendence of ζ(3). Two subjects, two rounds, the same structure: the correct object keeps being found, and the difficulty keeps living one step past it.
No zero is located, excluded or constrained by anything on this page, and nothing here bears on whether the Riemann hypothesis is true. No RH claim is made or implied — not by us, and not by the note under review, which says so itself in its own abstract. Our own census numbers certify finite windows above finite detection floors and forbid nothing; the Davenport–Heilbronn results quoted here decide what a symmetry argument cannot do, and say nothing whatever about where a zero of ζ is.