Λ ≤ 0.1787854 — a new bound for the de Bruijn–Newman constant
A complete walkthrough of the proof
A new ceiling for _Λ_: the de Bruijn–Newman constant is at most _0.1787854_
I'm Jude Gomila and I've been exploring the zeta function in private since 2025. This post is part of a series of posts on discoveries obtained from human/ai collaboration. This post is about the de Bruijn–Newman constant Λ — a single real number with this property: the Riemann hypothesis holds exactly when Λ ≤ 0. Nobody can prove that yet, but its known ceiling can be lowered, and this is my computer-assisted proof taking it from 0.2 to 0.1787854, unconditionally, with no unproved conjecture anywhere in the chain. I'll walk you through the whole proof, step by step. Every claim links back to myaudit repository and the independent review record. Feedback, bugs and upgrade comments are welcome asGitHub issues.
Previous bound**Λ ≤ 0.2**This result**Λ ≤ 0.1787854**Method**Polymath 15 criterion + interval certificates**Special thanks**Dan Romik, Max Atkin**
Λ ≤ 0.1787854
= 129/800 + 87677/5,000,000: an exact rational, obtained by exact arithmetic from 3,149,013 + 883 + 1 machine-checked interval certificates.
1. 00 Why the Riemann hypothesis matters 2. 01 Λ, the constant whose value decides the Riemann hypothesis 3. 02 Heating the function pulls its zeros onto the real axis 4. 03 The bounds on Λ, and how the methods work 5. 04 Three finite checks that prove an upper bound on Λ 6. 05 Check one: RH is already machine-verified below the barrier 7. 06 Check two: 3.1 million windows certified zero-free 8. 07 Check two, continued: one lemma covers the rest to infinity 9. 08 Check three: a wall no zero can cross 10. 09 Combining the checks gives Λ ≤ 0.1787854 11. 10 How the proof was checked, four layers deep 12. 11 Why this method cannot reach Λ ≤ 0 13. 12 Provenance & links
Prologue
Why the Riemann hypothesis matters
The primes 2, 3, 5, 7, 11, 13, … are the atoms of arithmetic: every whole number factors into primes in exactly one way, so facts about primes become facts about all numbers. Individually they are irregular — no known rule produces the next prime from the ones before it. Counted in bulk, they obey a law: the number of primes up to x stays close to a single smooth curve (the prime number theorem, proved in 1896). The open question is the size of the error — how far the true count can stray from the curve. That error term is what the Riemann hypothesis governs, and it is why RH matters: sharpen the error term and you sharpen hundreds of results in number theory that depend on it.
**Fig. 1**The prime-counting staircase π(x) (the dark stepped line) climbs one step at each prime. The dashed blue curve is the smooth prediction Li(x). The gap between them is a sum of waves, one wave per zeta zero, and the Riemann hypothesis says every wave has the smallest possible amplitude.
In 1859Bernhard Riemannexplained where that hidden order comes from. He took Euler's identity, which connects the primes to a single function of one complex variable,
$$
\zeta \left(\right. s \left.\right) \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \underset{n \geq 1}{\sum} \frac{1}{n^{s}} \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \underset{p \textrm{ } p r i m e}{\prod} \left(\left(\right. 1 - p^{- s} \left.\right)\right)^{- 1} $$
hover or tap a colored term for what it does
extended it to the whole complex plane, and discovered that the wobble of the prime count around its smooth curve is governed — _exactly_, via an explicit formula— by the locations of the **zeros** of this function. Each zero contributes one wave to the error; the zero's height sets the wave's frequency and, crucially, its horizontal position sets the wave's _amplitude_. Riemann observed that every zero he could examine sat on one vertical line, Re s = ½, now called the _critical line_ — the position giving the _smallest possible_ amplitude — and remarked it was “very probable” all of them do. That remark is the **Riemann hypothesis**. Its concrete content: the prime-count error up to x never exceeds roughly √x, the same size as the wobble of a fair coin flipped x times. The primes are allowed to look random; RH says they are never allowed to drift with a bias.
The wave description is an actual formula, and you can run it below. The slate staircase counts prime powers (a cousin of the staircase above, weighted so the mathematics is exact), and the vermillion curve is Riemann's formula built from the smooth trend plus one wave per zeta zero. Drag the slider and watch thirty zeros carve the primes:
Try it — build the primes out of zeta zeros, one wave at a time
zero-waves included **0 / 30**no zeros: just the smooth trend x, which misses every step
every one of these zeros has real part exactly ½, which makes its wave swell like √x as x grows — the slowest growth the explicit formula allows. A zero off the line at real part β ≠ ½ would make its wave grow like x^β instead, out of step with all the others, and the prime count would drift off course.
It has now been open for 167 years. It is part of the eighth of Hilbert's problems (1900) and is one of the Clay Millennium Prize problems today; hundreds of theorems across number theory and beyond are proved _conditionally_, “assuming RH.” Its zeros have been checked by computer into the trillions — every one on the line — but a check is not a proof. The way forward is to turn the question into a number that can be moved — and that is exactly what Λ is.
Λ (defined properly in the next chapter) repackages the Riemann hypothesis as a statement about one real number: **RH holds if and only if Λ ≤ 0** (a proved equivalence, established in Chapter 1). That reformulation has three consequences. First, progress becomes _measurable_: a yes/no conjecture has no partial credit, but an upper bound on Λ can shrink: ½ → 0.22 → 0.2 → and now 0.1787854. Second, since 2018 we know Λ ≥ 0, so Λ is confined to the interval from 0 to the current ceiling, and RH is the statement that Λ sits at the left endpoint; every improvement to the ceiling is measured distance toward the answer. Third, bounds on Λ are _unconditional_ — nothing in them assumes RH itself. Lowering Λ is one of the few rigorous, quantifiable ways to make progress on the Riemann hypothesis.
Chapter 1
Λ, the constant whose value decides the Riemann hypothesis
A single real number whose sign settles the question: the Riemann hypothesis holds exactly when Λ ≤ 0.
Start with Riemann's xi function, a repackaging of the zeta function:$\xi \left(\right. s \left.\right) = \frac{1}{2} s \left(\right. s - 1 \left.\right) \pi^{- s / 2} \Gamma \left(\right. s / 2 \left.\right) \zeta \left(\right. s \left.\right)$. The Riemann hypothesis says all its zeros lie on the critical line$R e \textrm{ } s = \frac{1}{2}$. Rotate coordinates so that line becomes the real axis (this proof uses the Polymath 15 normalization$H_{0} \left(\right. z \left.\right) = \frac{1}{8} \textrm{ } \xi \textrm{ } \left(\right. \frac{1}{2} + \frac{i z}{2} \left.\right)$) and RH becomes a single sentence:
RH, restated
Every zero of the entire function$H_{0}$ is a **real number**.
In 1950 de Bruijn had the idea of _deforming_ this function with a one-parameter flow — mathematically, running the heat equation on it:
$$ H_{t} \left(\right. z \left.\right) \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \int_{0}^{\infty} e^{t u^{2}} \textrm{ } \Phi \left(\right. u \left.\right) \textrm{ } cos \left(\right. z u \left.\right) \textrm{ } d u $$
where $\Phi$ is the fixed super-exponentially decaying kernel with $H_{0} = \frac{1}{8} \xi$. Positive t smooths the function and, as we'll see, herds its zeros toward the real axis; negative t roughens it and pushes zeros off. De Bruijn proved that once all zeros are real they stay real at every later time. So there is a single threshold, made precise by Newman in 1976:
$$ \Lambda \textrm{ }\textrm{ } = \textrm{ }\textrm{ } inf \left{\right. \textrm{ } t : \textrm{ } H_{t} \textrm{ }\text{has}\textrm{ }\text{only}\textrm{ }\text{real}\textrm{ }\text{zeros} \textrm{ } \left.\right} $$
That threshold is the **de Bruijn–Newman constant**, and it converts the Riemann hypothesis from a statement about infinitely many zeros into a statement about one real number:
$$ R H \textrm{ }\textrm{ } \Longleftrightarrow \textrm{ }\textrm{ } \Lambda \leq 0 $$
**Analogy.** Λ is a thermostat reading. The xi function is a room full of particles (its zeros), and t is heat: warm the room and the particles settle onto the floor (the real axis); chill it and some lift off. Λ is the exact temperature at which the last airborne particle lands. The Riemann hypothesis says the room as built — at temperature zero — already has everything on the floor. Since we can't yet check every particle, we do the next best thing: prove the landing temperature is low.
**Fig. 2**The state of knowledge about Λ. The Riemann hypothesis is equivalent to Λ ≤ 0; the equivalence is itself a theorem. Rodgers–Tao (2018) proved Λ ≥ 0, so RH, if true, is true with nothing to spare. This work moves the other wall: Λ is now known to be at most 0.1787854. The truth lives somewhere in the vermillion interval, and RH says it lives at its left endpoint.
One direction is now settled. Newman conjectured Λ ≥ 0, famously adding that if RH is true, it is _“only barely so”_ — and Rodgers and Tao proved this in 2018. So Λ is confined: $0 \leq \Lambda$, and progress can now come only from the upper side. This proof moves the ceiling to$\Lambda \leq 0.1787854$. With the floor at 0, this removes just over 10% of the interval that remained.
Chapter 2
Heating the function pulls its zeros onto the real axis
Complex zero pairs sink toward the real axis at a computable rate; the proof is a schedule for when the last of them arrives.
Under the flow, the zeros of $H_{t}$ move like interacting particles: real zeros repel each other along the axis, and each complex-conjugate pair gets pulled toward the axis. For an isolated pair at height y (that is, at$z = x \pm i y$), the dynamics are governed, up to smaller corrections, by
$$ \frac{d y}{d t} \textrm{ }\textrm{ } \approx \textrm{ }\textrm{ } - \frac{1}{y} \Longrightarrow \frac{d \left(\right. y^{2} \left.\right)}{d t} \textrm{ }\textrm{ } \approx \textrm{ }\textrm{ } - 2 $$
so the _squared_ height falls at a constant rate 2, and a pair starting at height y touches down at time about $y^{2} / 2$. The final bound has the same form: it will be$t_{0} + y_{0}^{2} / 2$: “run the flow to time$t_{0}$, show nothing flies higher than$y_{0}$, then wait $y_{0}^{2} / 2$ longer for the stragglers to land.” Polymath 15's Theorem 1.2 (introduced next chapter) is the rigorous, effective version of that sentence.
Try it — run the heat flow and watch the zeros land
time t **0.0000**still complex **5 pairs**rule **a pair at height y lands at t = y²/2**…heating
**Analogy.** Rain settling after a storm. Each complex pair is a droplet at altitude y, and gravity here acts on altitude-squared: every droplet loses y² at the same steady rate 2. You don't need to track any droplet individually — if you can certify that by time t₀ no droplet is above altitude y₀, physics finishes the argument: the sky is completely clear by t₀ + y₀²/2. The hard work is the certifying, and it splits into the three parts of the next chapter.
Two structural facts make the schedule provable. First, zeros move _continuously_ in t: a zero cannot appear at a point without traveling there. Second, a zero's vertical speed is bounded by the dynamics, so a region that is empty stays empty unless a zero crosses its boundary. Together these reduce the problem to fencing a region and checking the fence — which is what the criterion asks.
Chapter 3
The bounds on Λ, and how the methods work
From de Bruijn's 1/2 to Polymath's 0.22 to 0.1787854.
De Bruijn proved$\Lambda \leq \frac{1}{2}$ in 1950. Fifty-nine years later, Ki, Kim and Lee improved it to the strict inequality $\Lambda < \frac{1}{2}$. The current method dates from 2018–2019, when Terence Tao's Polymath 15 project built an _effective_ theory of the heat flow: explicit, computable approximations to $H_{t}$ with rigorous error bounds, and a criterion (their Theorem 1.2) that converts three finite, checkable statements into an upper bound on Λ. Polymath 15 instantiated it to get Λ ≤ 0.22; plugging in Platt–Trudgian's 2020 record-height RH verification pushed the same machinery to 0.2.
**Fig. 3**Seventy-six years of squeezing from above. De Bruijn's 1/2 stood for six decades before Ki, Kim and Lee made it strict; Polymath 15's 2019 breakthrough — the criterion this proof instantiates — brought it to 0.22, and Platt–Trudgian's record RH verification pushed the same criterion to 0.2. Below the axis of battle: Newman conjectured Λ ≥ 0 in 1976 and Rodgers–Tao proved it in 2018, so the interval can never close from this side past zero.
**Analogy.** Theorem 1.2 works like a crane left standing on the site: anyone who can supply the three certified parts it needs — a verified height, a clean final-time region, and a barrier — can use it to lift the bound. This proof machines heavier parts and lifts with the same crane, near its rated capacity at today's verified height.
The squeeze on Λ has two independent sides, and they work by completely different mechanisms. Both live in the **upper half-plane**: since$H_{t}$ takes real values on the real axis, every off-axis zero at$x + i y$ has a mirror twin at $x - i y$, so it is enough to track the region above the axis — position x running along the axis, height y measuring how far a zero still has to fall. RH is the statement that this upper region is empty.
**Fig. 4**H t takes real values on the real axis, so its complex zeros come in conjugate pairs: x + iy above the axis, x − iy below. The lower half-plane is a perfect reflection — nothing happens there that the upper half doesn’t already show — so the whole proof watches one region: position x along the axis, height y above it. The Riemann hypothesis is the statement that the tinted region is empty.
1. #### The floor: why Λ ≥ 0 (Rodgers–Tao, 2018)
Recall from the last chapter what Λ is: the moment in flow time when the last zeros finish landing on the real axis — and t = 0 is the function as built. Saying Λ is negative would mean the landing finished _before_ t = 0 — so the zeros we actually observe would not be fresh arrivals, but particles that had already spent time settled on the axis, jostling each other. That extra time matters, because zeros on the axis repel each other, and the longer they jostle, the more the irregularity irons out: crowded neighbors push apart, wide gaps close up, and the spacings drift toward uniformity — the way a traffic jam evens out once cars can spread along the road. Rodgers and Tao showed the real zeta zeros are too unevenly spaced for that: their gaps fluctuate more than any configuration that had already relaxed this way could still allow. So the landing cannot have finished early, and Λ ≥ 0. Note what the argument does and does not do: it locates no zero off the line; it only shows the observed spacings are incompatible with earlier relaxation.
**Fig. 5**The floor argument compares two worlds. If the landing had finished before time 0 (Λ < 0), the zeros would have spent extra time repelling each other on the axis, ironing their gaps toward uniformity (top). The zeros we actually observe keep clusters and wide gaps (bottom) — too uneven to have relaxed. So the landing cannot have finished early, and Λ ≥ 0.
1. #### The ceiling machine: Tao and Polymath 15's barrier method (2019)
The upper bound must rule out zeros in the **canopy** — Polymath's name for the region at height $y_{0}$ and above at the final time $t_{0}$, stretching along the entire infinite axis. A theorem of de Bruijn says an empty canopy forces$\Lambda \leq t_{0} + y_{0}^{2} / 2$. Their method empties it in three zones. Left of a chosen abscissa X: the computer-verified stretch of RH keeps the region clean at t = 0, and a continuity analysis of each zero's velocity keeps it clean afterward — provided nothing crosses in from the right. Blocking that crossing is the **barrier construction**: a rectangle one unit wide at X, meshed along its boundary, where the effective approximation is evaluated at a ladder of time slices; certified bounds on how fast the values can move between mesh points and between slices, plus the argument principle (winding number 0), keep it zero-free at _every_ intermediate time. Right of the barrier, the canopy is checked at the final time only, using the approximation$H_{t} \approx B_{t} f_{t}$ and Euler-product mollifiers. Polymath 15 ran this at $t_{0} = 0.2$,$y_{0} = 0.2$,$X = 6 \times 10^{10} + 83952 - 0.5$, with the barrier parked at the limit of Platt's 2011 verification, the fractional shift chosen to make the small-prime Euler product large at the barrier — and obtained Λ ≤ 0.22. Their bottleneck was the final-time check (ii).
**Fig. 6**The ceiling machine. The _canopy_ is everything at height y₀ and above at the final time — if it is empty, de Bruijn’s theorem forces Λ ≤ t₀ + y₀²/2. It is emptied in three zones: left of X the machine-verified stretch of RH keeps it clean (velocity bounds keep zeros from wandering up); the barrier at X is a wall checked zero-free at _every_ intermediate time, so nothing crosses in from the right; and right of the barrier the canopy is checked at the final time only.
1. #### Platt–Trudgian's lever (2020–21)
The criterion consumes exactly one external computation, the verified height: the barrier must stand below it, and a barrier further out allows better$\left(\right. t_{0} , y_{0} \left.\right)$. Platt and Trudgian raised the rigorously verified height a hundredfold, to exactly 3,000,175,332,800, itself an interval-arithmetic computation. Feeding it to the same criterion gives Λ ≤ 0.2, noted intheir own paper.
2. #### What I changed (2026)
Same criterion, same verified height; the differences are in the parameters and in how the result is certified. I parked the barrier so the height it consumes, X/2, sits 175,239,886.5 below the exact Platt–Trudgian record (0.006% from the edge), re-optimized the row in exact rationals to the criterion's boundary, and built the heavier certification the thinner row demands: mollifiers up to the prime 11 where the floors are thinnest, the window-freeze, native-binding and Dini-transfer lemmas, an all-N tail theorem in place of sampling, and a 883-prism barrier certificate that includes t = 0 exactly. That takes 0.2 to 0.1787854, about$2 \times 10^{- 7}$ below the most this criterion can give at this height. The rest of this post is the detail.
In numbers, my row is$X = 6,000,000,185,827$,$t_{0} = \frac{129}{800}$, and$y_{0}^{2} = \frac{87677}{2500000}$, supported by roughly 3.15 million interval-arithmetic certificates and an audit trail built so a reviewer can rebuild every piece from source. The next chapter turns the flow's machinery into three finite checks, taken one at a time.
**What’s new in this work.** The crane is Polymath 15’s and the verified height is Platt–Trudgian’s; here is exactly what this proof adds on top of them:
- A new exact parameter row — t₀ = 129/800, y₀² = 87677/2,500,000 — chosen at the edge of what the criterion admits at today’s verified height (Chapter 11 maps that edge).
- Three small theorems that turn a 3.1-million-window scan from numerics into proof: one evaluation per window, proved conservative across the whole window (the Window Freeze theorem); certified floors that bound the paper’s own quantity exactly, with no conversion factor left to trust (the native binding lemma); and one certified height carried to every height in the band (the Dini transfer) — Chapter 6.
- A contraction argument covering the infinite remainder — every window past the cutoff — in a single certified computation (Chapter 7).
- A barrier rebuilt at the modern verified height: 883 closed prisms, each held zero-free through time by winding-number certificates (Chapter 8).
- An audit trail built like flight certification: a hash-pinned repository, fail-closed checkers, replays on two toolchains at two precisions, an adversarial AI review panel, a full from-source recompute, and an independent human review (Chapter 10).
Chapter 4
Three finite checks that prove an upper bound on Λ
Theorem 1.2 of Polymath 15, instantiated at one exact row of parameters.
Pick three numbers: a barrier location$X$, a final time $t_{0}$, and a landing height $y_{0}$. Polymath 15's Theorem 1.2 says: if the three hypotheses below hold, then $\Lambda \leq t_{0} + y_{0}^{2} / 2$. (A theorem's hypotheses are its if-part: supply them, and the conclusion is yours. They are the three “checks” of the chapter titles ahead.) The colors below match the map and the chapter headings that follow: slate for (i), brass for (ii), vermillion for (iii). I chose the exact row
$$ X & = 6000000185827 , \\ t_{0} & = \frac{129}{800} = 0.16125 , y_{0}^{2} = \frac{87677}{2500000} = 0.0350708 , $$
chosen so that $t_{0} + \frac{y_{0}^{2}}{2} = \frac{893927}{5000000} = 0.1787854$exactly — every parameter is a rational number, and the final arithmetic is exact, not floating-point.
The three hypotheses (Polymath 15, Theorem 1.2)
**(i) Verified height.** No zeta zeros$\zeta \left(\right. \sigma + i T \left.\right) = 0$ with$\frac{1 + y_{0}}{2} \leq \sigma \leq 1$ and$0 \leq T \leq X / 2$.
**(ii) Final-time clearance.** No zeros of$H_{t_{0}} \left(\right. x + i y \left.\right)$ with$x \geq X + \sqrt{1 - y_{0}^{2}}$ and$y_{0} \leq y \leq \sqrt{1 - 2 t_{0}}$.
**(iii) The barrier.** No zeros of $H_{t} \left(\right. x + i y \left.\right)$ in a thin curved wall over $X \leq x \leq X + \sqrt{1 - y_{0}^{2}}$, for _every_ intermediate time $0 \leq t \leq t_{0}$.
**Fig. 7**The Polymath 15 criterion as a quarantine map (schematic — the barrier is 1 unit wide at x ≈ 6×10¹²). Hypothesis (i) clears everything left of the wall at time zero, using the finite RH verification. Hypothesis (iii) makes the wall itself zero-free at _every_ time 0 ≤ t ≤ t₀. Hypothesis (ii) clears the right half-line at the final time t₀, starting at x = X + √(1−y₀²) — a landmark just _inside_ the wall (since √(1−y₀²) ≈ 0.982 < 1), so regions (ii) and (iii) overlap and leave no gap. Zeros move continuously — so nothing can be above height y₀ at time t₀, and the last stragglers land within y₀²/2 more.
**Analogy.** A quarantine protocol. You want to certify a city district is virus-free at time t₀ without inspecting the whole world. So: (i) at time zero, door-to-door inspect everything inside the district — that's the brute-force RH verification up to height X/2. (iii) Close the border for the entire duration — a wall one unit thick, checked zero-free at every moment: nothing sneaks in from outside while the flow runs. (ii) At the final time, scan the region just beyond the wall to confirm nothing is pressed up against it. Continuity of zero trajectories is the epidemiology: infection can't teleport. Conclusion: at time t₀, nothing in the protected zone flies above y₀ — so everything lands by t₀ + y₀²/2.
Why _these_ three numbers and not others? Try it yourself — the bound, the criterion's validity condition, and the feasibility of actually certifying a row all move as you drag:
Try it — pick your own row (t₀, y₀²) and see what bound it would buy
final time t₀ = 0.16125 landing height² y₀² = 0.03507
bound Λ ≤ t₀ + y₀²/2 =**0.1787854**canopy y₀² + 2t₀ = **0.3576< 1 ✓**
provable — and essentially at the wall (this is the proof's territory)
the proof's row: t₀ = 0.16125, y₀² = 0.0350708 → 0.1787854
The three hypotheses are supplied by three completely different technologies — a cited world-record computation, a 3.15-million-row certified scan with a tail theorem, and an argument-principle certificate. The next four chapters take them in order.
Chapter 5
Check one: RH is already machine-verified below the barrier
Platt and Trudgian's verification covers the height the criterion needs, with margin 175,239,886.5.
The first hypothesis is the one part I didn't prove myself but licensed from the literature:Platt and Trudgian's record verification(2020) that every nontrivial zeta zero up to height
$$ T_{P T} = 3,000,175,332,800 $$
lies exactly on the critical line. The criterion at our row consumes height$X / 2 = 3,000,000,092,913.5$, which sits inside the verified range with an exact margin of $175,239,886.5$. The zero-height endpoint $T = 0$ is closed classically: on the real segment$0 < \sigma < 1$ a two-line alternating-series argument shows$\zeta \left(\right. \sigma \left.\right) < 0$ (strictly negative, so no zero hides at the bottom), and at $s = 1$ zeta has a pole, not a zero. Nothing conjectural enters: RH is used only as a _finite, machine-verified fact_ up to a specific height, plus the functional equation to map $H_{0}$-zeros to zeta zeros by the exact change of variables$x = 2 T$.
A detail the audit surfaced: the margin lives entirely inside Platt–Trudgian's _precise_ Theorem-1 height 3,000,175,332,800. With the rounded “3×10¹²” of their abstract, hypothesis (i) would fail by 92,913.5. The height the barrier consumes, X/2, sits less than one part in 10⁴ below the exact record; it is the single thinnest external dependency in the whole chain, and it is cited to the digit.
**Analogy.** The verified height is a foundation poured by others — a twelve-trillion-zero computation, published and checkable — that this proof builds on rather than repeats. The house goes up at the property line: a margin of 175 million against a depth of three trillion puts the fence 0.006% from the edge. Chapter 11 explains why building there is the right choice: at this height, that is where the value lies.
Check (i) secured.
Chapter 6
Check two: 3.1 million windows certified zero-free
3,149,013 certified inequalities, one for every window from the barrier to the tail.
Hypothesis (ii) demands that at the final time $t_{0}$, the function $H_{t_{0}}$ has no zeros to the right of the barrier in the band $y_{0} \leq y \leq \sqrt{1 - 2 t_{0}}$, an _infinite_ region. Polymath 15's Theorem 1.3 makes this tractable: it supplies an explicit approximation
$$
H_{t} \textrm{ }\textrm{ } \approx \textrm{ }\textrm{ } B_{t} \cdot f_{t} , f_{t} \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \text{a}\textrm{ }\text{finite},\textrm{ }\text{computable}\textrm{ }\text{Dirichlet}-\text{type}\textrm{ }\text{sum} $$
where $B_{t}$ is an explicit nonvanishing factor, together with hard error bounds whose certified total we call$E_{max}$. So it suffices to prove$\mid f_{t_{0}} \mid > E_{max}$ everywhere: if the computable sum always beats the worst possible approximation error, the true function cannot vanish — a strict inequality between computable quantities. The region is cut into natural “Riemann–Siegel windows” $W_{N} = \left[\right. x_{N} , x_{N + 1} \left.\right)$ with$x_{N} = 4 \pi \left(\right. N^{2} - t_{0} / 16 \left.\right)$, indexed by the length N of the finite sum — the natural coordinate system of the Riemann–Siegel approximation. The scan runs from the window containing the barrier,$N = 690,988$, to the tail cutoff$N = 3,840,000$: one certified row per window,**3,149,013 rows** — each a strict interval-arithmetic inequality — with no gaps, duplicates, or uncertain records.
Most of this proof consists of certificates, so the word needs a precise meaning. A certificate is not an approximate computation. It is a guaranteed enclosure: an interval the true value provably cannot escape, which narrows with more work. A claim is proved only when the whole interval lands on the required side. Try it yourself:
Try it — certify a fact about π the way this proof certifies 3,149,896 facts
computation effort (series terms) = 2
interval width **4.27e-2**UNDECIDED — the interval still straddles the claim, so the checker will not rule in either direction; add effort to shrink it
this toy uses Machin's series for π with guaranteed truncation directions; the proof's certificates play the same game with exact rationals at 256-bit precision. Note the third claim: it is false by 2.7 millionths, and no amount of effort ever certifies it — the interval just lands on the wrong side. A fail-closed checker has three possible outcomes: proved, disproved, or no answer.
**Fig. 8**Hypothesis (ii), lane by lane (the N-axis is stretched for legibility). Each window W_N gets one certified interval-arithmetic row proving |f| stays above a stored floor; richer Euler mollifiers (more auxiliary primes) are spent exactly where the problem is hardest, near the barrier. The tail lemma is a single theorem covering every N ≥ 3,840,000 at once — and the two lanes overlap by construction on the full window N = 3,840,000, so no seam is left uncovered.
Three ideas make the scan rigorous rather than merely numerical:
1. #### Freeze conservatively, prove monotonicity
Within each window, the certificate evaluates the bound at the window's left edge and proves — by elementary monotonicity in x, including through the kink of a positive-part term — that this frozen value is conservative across the whole window (_the Window Freeze theorem_). Exact rational Machin brackets for π and exact square-root brackets pin the very first window:$x_{690988} < X + \sqrt{1 - y_{0}^{2}} < x_{690989}$ with millions to spare on both sides.
2. #### Mollify with small primes
Raw $\mid f_{t} \mid$ can dip close to zero. Multiplying by a short Euler product $E$ (four families of auxiliary primes:{2,3,5,7,11} near the barrier, stepping down through {2,3,5,7} and{2,3,5} to {2,3} far out) flattens the dips — a damping factor of this kind is called a _mollifier_, from the Latin for “to soften”. The _native binding lemma_ proves by exact Dirichlet convolution that the stored floors bound the paper's normalized $\mid f_{t} \mid$ directly — the certified numbers are in the right units, with no extra conversion factor left to trust.
3. #### Certify one height, transfer to all heights
Each row certifies a floor at one height; a separate _Dini-derivative_ theorem proves the relevant mass is monotone in y — with the scan's thinnest certified ratio, $0.99999860767275095 < 1$, so the floor transfers to the entire band $y_{0} \leq y \leq \sqrt{1 - 2 t_{0}}$. The margin is thin, but it is a proven strict inequality, checked at two precisions that agree to 17 digits.
The payoff: the binding floor across all 3,149,013 rows is$\mid f_{t_{0}} \mid \geq 7.91366 \times 10^{- 7}$, against a certified total error $E_{max} \leq 2.33495 \times 10^{- 7}$: nonvanishing with a 3.4× cushion.
**Analogy.** Mowing an infinite lawn to prove there are no snakes in the first three million meters: you don't photograph every blade. You cut the lawn into mower widths (windows), prove the grass in each strip is never taller than at its left edge (freeze), wear polarized glasses that make snakes stand out against grass (the mollifier), and prove that if a strip is clear at eye level it's clear at every level (Dini transfer). What remains — the lawn past three million — needs a different tool.
Chapter 7
Check two, continued: one lemma covers the rest to infinity
One theorem, one contraction, and every window past three million folds into a single computation.
Past the cutoff $N_{*} = 3,840,000$ no finite list of certificates can help — there are infinitely many windows left. This infinite remainder is called the**tail**, and the **tail lemma** handles all of it at once. Its engine is a contraction: the mollified sum is written as its leading term 1 plus a correction mass$D$, and every ingredient of D is bounded by quantities that _decrease_ as N grows. A capping lemma — each ingredient is bounded by its value at the cutoff, which is its worst case — plus exact finite Dirichlet convolution reduce all $N \geq N_{*}$ to one interval computation at the cutoff itself:
$$
D \textrm{ }\textrm{ } < \textrm{ }\textrm{ } 0.999721 \textrm{ }\textrm{ } < \textrm{ }\textrm{ } 1 \Longrightarrow \mid f_{t_{0}} \mid \textrm{ }\textrm{ } \geq \textrm{ }\textrm{ } \frac{1 - D}{M_{max}} \textrm{ }\textrm{ } > \textrm{ }\textrm{ } 1.734 \times 10^{- 4} $$
after subtracting all error terms — uniformly for every window past the cutoff, every height in the band (in fact an extended band), and a full closed box of times around $t_{0}$ — with no sampling in N and no assumed monotonicity in t. The computation is a single standalone FLINT/Arb program, run at 256-bit and 512-bit precision (agreeing), with a separate Python interval implementation as corroboration.
The certified 0.999721 overstates how close D is to 1: it includes a provably redundant padding term. Removing it (the audit checked this) gives a true contraction around 0.96, so the tail has real headroom: about four orders of magnitude above the interval widths. And the two proof lanes overlap on the complete window N = 3,840,000: the finite scan proves it, the tail proves it again, and no seam between them is left to trust.
**Analogy.** Proving every domino past the three-millionth falls without touching any of them: show the push each domino receives always exceeds its resistance by a fixed factor, and show that factor only improves down the line. Then checking the single worst domino — the one at the cutoff — settles infinitely many at a stroke.
Checks (i) and (ii) secured: the final time is clean everywhere right of the barrier.
Chapter 8
Check three: a wall no zero can cross
A wall one unit thick, held zero-free at every instant by 883 interval certificates.
The barrier is where the proof must control _every intermediate time_, not just the endpoints: no zero of $H_{t}$ may enter the closed box
$$ R = \left[\right. X , \textrm{ } X + 1 \left]\right. \textrm{ }\textrm{ } + \textrm{ }\textrm{ } i \textrm{ } \left[\right. 0.1809 , \textrm{ } 1 \left]\right. , 0 \leq t \leq t_{0} = \frac{129}{800} $$
which strictly contains the curved wall Theorem 1.2 actually requires (the floor clearance is exact: $y_{0}^{2} - 0.1809^{2} = \frac{234599}{10^{8}} > 0$). The tool is the**argument principle**, the classical way to count zeros inside a region: walk the boundary of R, watch the function's value wind around the origin; winding number zero means no zeros inside. The certificate makes this rigorous under the flow by slicing time into**883 consecutive closed prisms**and, on each prism, gating a strict interval inequality:
$$
\underset{\text{min}\textrm{ }\text{boundary}\textrm{ }\text{modulus}}{\underbrace{M_{i}}} \textrm{ }\textrm{ } > \textrm{ }\textrm{ } \underset{\text{mesh}\textrm{ }\text{interpolation}}{\underbrace{\frac{D_{z , i}}{2 \left(\right. n u m - 1 \left.\right)}}} \textrm{ }\textrm{ } + \textrm{ }\textrm{ } \underset{\text{motion}\textrm{ }\text{in}\textrm{ }\text{time}}{\underbrace{D_{t , i} \textrm{ } \Delta t_{i}}} \textrm{ }\textrm{ } + \textrm{ }\textrm{ } \underset{\text{approximation}}{\underbrace{0.00125}} $$
In words: the boundary values stay so far from zero that neither the gaps between mesh points, nor the motion of the function during the prism's time slice, nor the approximation error of using $f_{t}$ for$H_{t} / B_{t}$, could possibly let a zero slip through. The certified total approximation error, $3.565 \times 10^{- 4}$, fits the 0.00125 allowance 3.5 times over; the worst prism margin is 0.5198.
**Fig. 9**Hypothesis (iii), the barrier certificate. Left: time is cut into 883 consecutive closed prisms from t = 0 (included exactly — not as a limit) to t₀ = 129/800, with byte-identical seams. Right: on each prism, the function's values around the boundary of the barrier rectangle trace a loop; the certificate proves the loop never comes near the origin (minimum certified margin 0.5198) and winds zero times around it — by the argument principle, no zeros inside. All 883 gates are strict interval inequalities: any overlap or indeterminacy is failure.
The function's values along the rectangle's boundary trace a closed loop in the complex plane, and a zero inside the rectangle forces that loop to wrap around the origin. The certificate proves the loop stays far away:
Try it — the argument principle as a zero detector
min |f| on the boundary **0.47**winding number 0 verdict**origin avoided — no zeros inside; this prism passes**
the real certificate proves min |f| ≥ 0.5198 on every one of the 883 prisms — the loop never gets close
The corners were the risky parts, so I gave each one its own lemma. The$t = 0$ endpoint is included _exactly_ — the first prism's seam is required to be literally zero, where the boundary modulus is 4.278, and the derivative majorants provably do not blow up as$t \rightarrow 0$. The _derivative box lemma_ proves the uniform bounds $D_{z} , D_{t}$ on whole prisms (never sampled at a single endpoint). The 7,688 stored Taylor-series coefficient components were independently regenerated — all contained in their stored balls — and the truncated Taylor tail is bounded by $1.96 \times 10^{- 22}$, carried explicitly through every value. Every gate is fail-closed: an interval that straddles it counts as failure.
**Analogy.** Guarding a gate for a fixed shift, with a camera that takes one frame per prism. A frame alone proves nothing — an intruder could cross between frames. So for each frame the guard also proves a speed limit (the derivative bounds): nothing can move far enough between this frame and the next to reach the gate. 883 frames, each with its certified speed limit, and the shift is covered — from the exact first second to the exact last.
All three checks secured: Theorem 1.2's hypotheses are met.
Chapter 9
Combining the checks gives Λ ≤ 0.1787854
Three checks, one exact rational, and margins that are proved inequalities.
By design, nothing new happens in the final step. A fail-closed assembly checker confirms that hypothesis (i) is licensed by the cited Platt–Trudgian height (with its exact margin of 175,239,886.5), that the finite lane and the tail jointly cover hypothesis (ii) with an overlapped seam, that the 883-prism certificate covers hypothesis (iii) down to exactly $t = 0$ and up to an endpoint enclosing$129 / 800$, and then performs the one line of mathematics left, in exact rational arithmetic:
$$
\Lambda \textrm{ }\textrm{ } \leq \textrm{ }\textrm{ } t_{0} + \frac{y_{0}^{2}}{2} \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \frac{129}{800} + \frac{87677}{5000000} \textrm{ }\textrm{ } = \textrm{ }\textrm{ } \boxed{\frac{893927}{5000000} \textrm{ }\textrm{ } = \textrm{ }\textrm{ } 0.1787854} $$
**Fig. 10**The whole proof on one page. Three independently certified inputs — colored as throughout this post — feed Polymath 15's Theorem 1.2, which converts them into the bound by exact rational arithmetic. Remove any input and the theorem concludes nothing; all three held.
**Fig. 11**The three decisive gates, drawn to per-gate scale: the full bar is the largest error each leg could tolerate; the filled portion is the error the interval arithmetic actually certifies. The tightest gate clears by a factor of 3.4. Every number shown is the rounded value of a directed interval inequality.
None of these margins is a floating-point comparison. Every gate is a _directed interval inequality_: outward-rounded enclosures on both sides, strict comparison, overlap treated as failure. The three cited inputs that code cannot prove (the two Polymath theorems and Platt–Trudgian's verification) are printed by the checker as**cited theorem inputs** and left ungated, marking the boundary between what the machine certifies and what the literature supplies. The audit also confirmed there is no circularity: the three hypotheses are logically independent, and no certificate consumes a value derived from the bound it proves.
Chapter 10
How the proof was checked, four layers deep
Checksums, independent replays, an adversarial review, and an independent human review.
A proof with this much computation is only as good as its audit trail, so I engineered the trail as carefully as the mathematics. Four layers, in escalating order of independence:
1. #### Checksums and fail-closed checkers
Every artifact — certificates, logs, sources, references — is pinned by SHA-256 and re-attested by `verify.sh`. The parsers are fail-closed: 54 strict checks on the barrier alone; any gap, duplicate, overlap, or indeterminate interval anywhere is a hard failure.
2. #### Cross-implementation and cross-toolchain replays
Every interval computation the bound rests on — the finite scan's error budget, the tail contraction, and the 883-prism barrier — is certified by FLINT/Arb ball arithmetic, with Python interval code as cross-checks. The finite error budget and the margin above it are certified by a standalone C program run at 256-bit and 512-bit precision inside the pinned container, gated as prerequisite P17 of the 40-gate fail-closed assembly (the original Python interval computation is retained as same-answer corroboration); the tail agrees across 256 and 512 bits plus a separate Python implementation; and the barrier replays end-to-end on two different OS/compiler/library stacks (Linux/FLINT 3.0.1 and macOS/FLINT 3.6.0), emitting identical verdicts on all 883 prisms.
3. #### An adversarial review, then a full recompute
I then set a four-agent adversarial AI review panel loose on each proof leg separately (July 2026), with instructions to break it. Verdict: no fatal or bound-invalidating defect; it surfaced — and fixed — one real gap (the thinnest inequality, the Dini transfer, wasn't wired into the assembly gate) and flagged the precise-height citation subtlety of Chapter 5. Every heavy computation was then rebuilt fresh from source: all 7,688 coefficients, all 883 prisms, all 3,149,013 finite rows regenerated and matched line-for-line.
4. #### Independent human review: Dan Romik
Dan Romik (author of the standard modern exposition of the ξ function's orthogonal-polynomial theory) reviewed my proof independently: he worked through the full package, verified the cited literature against the arXiv sources, re-proved every analytic lemma, and rewrote the mathematics as two journal-grade manuscripts — an expository account and a research paper — with the computer-assisted components isolated as seven standalone, self-contained verification programs.
**Analogy.** The trust model is an aircraft's: redundant independent systems (two toolchains, two implementations, two precisions), black-box recorders (hash-pinned logs), a hostile test pilot (the adversarial panel), and finally certification by an outside inspector who takes the engine apart himself (the independent reviewer). No single layer is asked to be infallible; the design is that they cannot all fail the same way.
The repository describes itself as a
_computer-assisted proof awaiting journal review_. The adversarial pass found no defect in the bound and the independent review has produced the journal manuscripts; formal publication is the remaining step. The chain is unconditional: it assumes neither the Riemann hypothesis, nor
GUE, nor any other conjecture.
Chapter 11
Why this method cannot reach Λ ≤ 0
Each certificate lane has its own wall, set by mollifier design; later research in the repository reaches 5.5×10⁻⁴ lower at the same verified height. Reaching Λ ≤ 0 still requires new mathematics.
This bound will not be improved by re-running the same machinery with more computing time: I chose the row at the edge of what this certificate lane admits — an early companion analysis placed the wall for these parameters about$2 \times 10^{- 7}$ below 0.1787854. Later research in the repository showed that wall belongs to the lane, not to the verified height: the mollifier is tuned to $t_{0}$ (lowering the time while holding it fixed fails structurally, at any precision), and a lower-time lane at the _same_ height reaches 0.1782354 — 5.5×10⁻⁴ lower, its arithmetic and archive machine-checked though its mathematics is not yet independently reviewed. Getting below that needs a mollifier retuned to the lower time, not a bigger dial.
**Fig. 12**A zoom of the number line near the new bound (full width here: three millionths). An early companion analysis placed the wall of what this lane's certificates can achieve about 0.0000002 below the bound — the parameters sit at the edge of what this mollifier admits. Later research showed the wall belongs to the lane, not the height: a lane retuned to a lower time reaches 0.1782354 at the same verified height. The long-run lever is verified height, which pays only inverse-logarithmically: reaching Λ ≤ 0 this way would require verification without end.
In the long run the binding lever is verified height — and it pays inverse-logarithmically: the achievable bound shrinks roughly like 1/log(height). Concretely, halving the bound requires roughly _squaring_ the verified height, from 3×10¹² to about 10²⁵ (far beyond any conceivable computation), and driving the bound all the way to 0 would require verifying RH to infinite height, which is just RH again. Together the two bounds say the following about$0 \leq \Lambda \leq 0.1787854$: Rodgers–Tao's floor says RH has no safety margin; the wall says brute force alone will never finish the job. Closing the remaining interval will take new mathematical ideas, most likely alongside heavier computation — each idea still has to be instantiated and certified. Working the bound downward is itself one way those ideas get found: every improvement forces a sharper structural understanding of how the zeros move under the flow, and of what the primes do and do not permit.
Provenance & links
The proof, its parameters, certificates, and audit apparatus are my own work (**Jude Gomila**, 2026). The mathematics instantiates**Polymath 15**'s Theorems 1.2 and 1.3 (D.H.J. Polymath, led by Terence Tao) at a new exact parameter row, and consumes **Platt–Trudgian**'s record RH verification as its only external computation. The independent reviewer is**Dan Romik**, who verified the analytic lemmas and reworked the material into journal-grade manuscripts; an adversarial AI audit with full from-source recomputation preceded his review. The result is unconditional in logical form; journal publication is the remaining step.