Summary · Dxtra Inc. (dxtra.com)

Looking for the behavior of a source, not for a transmitter

Twenty-nine searches of public solar and heliospheric archives for a technosignature carried not on a chosen channel but in a relationship among observables. Twenty-five are null, three void, and one recovers a known signal as a positive control.

Robert Griffin · Dxtra Inc. (dxtra.com) · draft of 13 September 2026

The premise

Technosignature searches begin by naming a carrier — the 21 cm line, an optical pulse, a waste-heat excess — and then searching it as deeply as instruments allow. Every one of them presupposes the channel. Coverage is measured in sky position, frequency and sensitivity: a map of how much radio has been searched, on the assumption that the answer is radio.

This work takes the case left over: no beam, no beacon hardware, no carrier specified in advance. Structure is carried by emission a star already produces, imposed by redistributing that flux rather than generating it. The receiver's problem is then not sensitivity at all. It is not knowing which series to examine.

Two constraints make that finite. A sender sharing no units with the receiver can use only dimensionless quantities — ratios and normalized combinations of measured observables. And a sender minimizing transmitted energy will pose a proof-of-work gate: cheap to set, expensive to solve, trivial to verify, so no energy is spent on receivers that could not act on the message.

Together those define a combination space that can be enumerated, searched, and — unlike a channel — have its coverage measured.

Where this differs from the NASA taxonomy

The 2018 NASA Technosignatures Workshop is the field's reference organization of what a technosignature is, and a reader arriving from it is entitled to know which of its categories this paper occupies. The answer is none of them, and the reason is worth stating as the categories themselves state it — by what each one assumes exists.

CategoryAssumesHere
radio — narrowband or pulseda dedicated transmitter, powered and pointednot assumed
optical / infrared — laser pulses, beaconsa dedicated transmitter, at higher powernot assumed
megastructures, waste heata structure large enough to occult or re-radiate a detectable fraction of the starnot assumed; the modulator is ~10−6 of the disc and invisible as an occulter
atmospheric — industrial speciesan industrial byproductnot assumed; nothing here is a byproduct
solar-system artifactsa physical object to image or encounternot the carrier here
planetary surface — city lightsa modified planetnot assumed
The premise of this paper is that none of those is necessary. The carrier is emission the star already produces, redistributed rather than generated: nothing to build, nothing to see, no byproduct to accumulate, no object to encounter.

Three differences follow. The carrier is a relationship among observables rather than a channel to point at. The gate is opened by the receiver's capability rather than paid for by the sender. And the target is our own star, from archives already collected — where the workshop's emphasis is on new instruments and on piggybacking planned missions. That premise does not appear in the workshop's taxonomy, and the archives it points at — solar and heliospheric records collected for space weather — do not appear in its inventory of what has been searched.

The gating argument, in twelve steps

The framework is not a claim about what a sender wants. It is an engineering derivation: what would a communications designer, facing an unknown number of unknown recipients over geological time, be forced to build? Four constraints — energy, parsimony, longevity, gating — each delete something, and what survives is a chain.

StepGate it passes
0The decision, under a budget — total energy across all recipients and all time is minimized
1Galactic survey — candidates identified remotely, before anything is spent on themstellar filter — ~109 candidates from ~1011 stars
2Mesh deployment — a passive listening network seeded first, from a carrier busthe address exists
3Probe launch — automated, self-repairing, transit 104–106 yr, unsupervisedsurvival in transit
4Emplacement at the star — high in the stellar atmosphere, where the cheap channels formpersistence — the framework's largest unsupported assumption
5The signal is embedded — a small dimensionless modulation carrying a pointer to one mesh node, repeating indefinitelythe puzzle is posed
6Maturity — continuous monitoring, long archives, compute to search a combination space, all arising for the receiver's own reasonscapability
7Recognition — realizing that a puzzle has been posed at all. Costs nothing, and that is the difficultygate 0 — invisible from below. The hardest gate in the chain
8Detection — right channel, right cadence, right combinationcoverage of the combination space
9Decode — the encoding inverted without shared unitsproof of work — the decode is the demonstration
10The pointer is acted on — the receiver's first real expense: it must build a transmitterthe gate opens — that expense is the filter
11The mesh receives the reply — at the named node, listening at zero idle cost
12The return — a response service acting for the sender, answering on parallel narrow beams
Steps 2 and 3 are the chain's only capital expenditure. Everything after step 5 is the receiver's money, which is the point of building it this way.

The architecture is designed to outlive its designer, and that is a core principle rather than a flourish. The longevity constraint is one-sided: it binds below, at the emergence of a receiver, and sets no upper bound — nothing in the architecture specifies a date at which the mesh stops, and no element of it is given a design life.

That is forced by the chain's own numbers. A bus crosses a 104 ly route in 7×104 yr; probes transit unsupervised for 104–106 yr; and the receiver takes 106–109 yr to arise. A designer who requires their own survival anywhere in that span has built a system that fails on the most probable branch. A Drake-style estimate multiplies by L, the civilization's lifetime; if the artifact outlives its designer, L is the wrong lifetime to multiply by. "Is anyone alive out there" and "is anything out there still running" come apart, and this searches for the second.

Gate 0, and what recognition costs in compute

If the message is conditional on solving a problem, the first barrier is not solving it but recognizing that a problem has been posed. That barrier is invisible from below: a civilization holding the right archives, at the right cadence, sees exactly what a civilization with no signal to find sees. The recognition gate has a size, and the size is set by the carrier — a message carried in a relationship among observables is findable only by a receiver able to search relationships, which is a compute threshold, not a telescope.

Search tierFLOPCray-1, 197688 coresFrontier
triple sweep5.5×101439.8 d52 min0.4 ms
quadruple sweep + matched null2.9×101421.0 d27.5 min0.2 ms
every search in this paper1.1×101579.6 d1.7 h0.8 ms
T3 — quadruples over 30 channels2.7×101753.5 yr17.8 d0.2 s
T6 — quadruples over 60, fast band2.8×10235.5×107 yr50,409 yr57.5 h
Sustained rates: Cray-1 160 MFLOPS (1976), the 88-core workstation used here ≈176 GFLOPS, Frontier 1.353 EFLOPS. The whole programme is eighty Cray-1 days and under a millisecond of Frontier; T6 — 55 million Cray-1 years in 1976 — is a long weekend on a flagship machine in 2026.

Our own position on that ladder is measurable rather than rhetorical. The pair space of thirty observables closes in under an hour on one machine; the triple space took 117 million statistic evaluations and 26 hours. The recognition gate weakens a common premise of the Fermi paradox — that contact would be self-announcing — without requiring absence, concealment or a great filter.

The search is a cryptanalyst's procedure

The correspondence is exact, and naming it is what makes the search design falsifiable rather than a matter of taste. The gate is a symmetric cipher under brute-force key search, and this paper runs the cryptanalyst's side of it.

Cryptographic objectHere
the keythe tuple (which observables, which functional form)
the ciphertextthe public archive — already collected, by someone else, for other reasons
the plaintextthe imposed modulation
the key spaceC(N,k) × |F| — enumerable, which is why coverage is a number
exhaustive key searchthe sweep of §4: every admissible key tried, the statistic as the oracle
unicity distancethe multiple-testing threshold — below it, more than one key yields a plausible decryption

What makes the scheme work at all is self-identifying plaintext. A receiver knows when the right key has been tried because the statistic fires — no crib and no known-plaintext pair is needed. That is the same property that makes brute-force search viable, and it is why the sender need not transmit anything to tell you that you have succeeded.

Two consequences of taking the analogy seriously. It satisfies Kerckhoffs's principle in its purest available form: nothing is hidden, the whole scheme can be published, and the security lies entirely in the work. And it is not proof of work in the hashcash sense — there the sender performs the work to price a message; here the receiver performs it, and the direction of payment is inverted. The nearest technical antecedent is not in the SETI literature at all but in Rivest, Shamir & Wagner's time-lock puzzles (1996), which set difficulty from projected future hardware speed. One difference is favourable here: a time-lock puzzle is inherently sequential and its delay cannot be bought off with more machines, whereas a key search parallelizes — which is precisely why the compute table above is a table of wall-clock times rather than of impossibilities.

What 29 searches found

25
null
3
void — the detector or the null failed its own validation
1
detection — a known signal, as a positive control
17
carry a stated sensitivity
13
of those verified by injection

The deepest verified limit is 1.4×10−6 in fractional Lyman-α irradiance at two minutes, at 95% recovery. Two channels were searched here for the first time: SOHO/VIRGO SPM photometry at one minute over 27.3 yr, and SOHO/GOLF Doppler velocity at 20 s over 25.9 yr.

Sensitivity is not what is limiting

The same instrument, days and pipeline are 245× more sensitive at two minutes than at one day. The binding constraint is not how faint a signal can be seen — it is how much of the combination space has been looked at.

CoverageResult
Pair space, after acquiring 17 further observables18% → 88% — 382 of 435 pairs
Tests spanning the Sun–heliosphere boundaryno survivor in 510
Triples, of 4,060 (2,934 reachable)2,749 completed → void
Cross-viewpoint, Earth line against Mars linenull to 8.9×10−4 at 3 d
The triple sweep returned void, not null, for want of a matched three-body null — a statement about the analysis, not about the sky.

Why each search ended where it did

ReasonnA statement about
nothing above threshold22the sky. The detector worked, the control passed, the channel was empty at the stated level
underpowered3the analysis — could not have found the effect if present
detector failed its own control1the analysis
no defensible null1the analysis — "no signal" cannot be specified for a star with its own higher-order structure
the Sun is too loud1the data. Not repairable by analysis
detection1a known signal, recovered as a positive control
Only the first line is astronomy. The honest reading is twenty-two channels searched competently, three searches too weak to count, two defeated by their own machinery, and one defeated by the data.

The one detection is a re-detection

Solar p-modes in GOES EXIS Mg II irradiance, comb spacing 135.1 and 135.0 µHz on the two spacecraft against an accepted 134.9. Priority belongs to Eden et al. (2024), and we claim none. Its value here is as a positive control whose answer was fixed in advance and corroborated by an independent group — a stronger check on the pipeline than any self-designed injection.

What this does and does not claim

A null result bounds an amplitude; it does not settle existence. What the limits establish is that no dimensionless modulation exceeding the stated amplitude exists in the stated observable over the stated band, during the interval observed — and every qualifier there is load-bearing.

The framework of the opening sections is not claimed to be supported by these results, and is not invoked to explain them. What is new is narrower than the machinery around it: that the carrier may be a relationship among observables rather than a channel to point at, that the resulting space can be enumerated and its coverage measured, and that a capability-gated sender needs nobody to abstain from anything.

Future research directions

Three things block the next tier, and only one of them is compute. They are listed in the order that matters, which is not the order of cost.

1. The null, not the cycles — an HPC ensemble for solar higher-order statistics

The triple space returned void, and the quadruple space would too, for the same reason: there is no defensible null for higher-order statistics of solar output. The question, stated without reference to technosignatures, is this: what is the null distribution of bispectra, trispectra and cross-channel phase alignment of solar output? Nobody knows.

Observation cannot answer it. We have one Sun and one realisation of it; a null distribution requires an ensemble, and the only ensemble available is a synthetic one.

The project. A large ensemble of independent global solar convective-dynamo simulations — Rayleigh, ASH or MURaM class — each integrated over several simulated activity cycles, from which synthetic disc-integrated irradiance and velocity series are drawn and the higher-order statistics measured across members.

Scale. At literature cost of 5×105 to 2×106 core-hours per realisation, an ensemble of 50–200 members is 0.4 to 1.8 million node-hours — within a single INCITE-class award, and modest by the standards of the simulations themselves.

What it unlocks. The triple and quadruple spaces become searchable, the void results become limits, and any future search for non-linear coupling in solar data inherits a characterised false-alarm rate. The value is not specific to SETI.

The gate this project must pass, stated first because it decides whether it is worth running. A null built from simulations that do not reproduce the Sun in the statistics being tested is worse than no null: it would license exactly the false positives the discipline of this paper exists to prevent. The ensemble has to be validated against the real Sun in the same statistics before any limit derived from it means anything.

2. Compute, once the null exists

The tiers are not blocked on machines — T3 is 17.8 days on the single workstation used here, and 0.2 s on Frontier. This table should not be read as asking for a machine. It is here to show that the compute is not the binding constraint, which is itself the finding.

TierScaleWhat it would buy
T3 — quadruples over 30 channels27,405 keys × 4 forms
~18 d on one 88-core machine
nothing, until a defensible four-body null exists
T6 — quadruples over 60 channels, fast band2.8×1023 FLOP
6.2 d on a dedicated exascale machine
the same objection, more strongly
Peak-bagging, full Lorentzian profile fits, GOLFdays, not hoursrow 7 sits at 1.46×10−5 against a designer level of 10−6; a ten-candidate estimator search bought 1.33×
N2 surrogate counts in place of the tail fitthe reason the fit was introducedthe generalised-Pareto extrapolation is unstable across realisations at 1.7% of triples

3. Instruments that do not exist

Four columns are blank across the whole seventy-year record, and no amount of compute or patience fills them:

Every long solar series is a concatenation across instrument handovers, and a slow signal spanning one could be destroyed by cross-calibration or manufactured by it. One of the 29 searches is spent as a control on exactly that: GOES-15 against GOES-16 X-ray over 1,031 dual days, which measured a 27.7σ shift in detector output across the handover while the search result itself did not move.

This is the binding constraint, and it is the one the paper cannot lift. Whether we happen to measure the particular combination a sender chose is roughly one chance in five for a pair and one in eleven for a triple, and it cannot be improved by better instruments on the channels we already have. Only by measuring more things.

The hypothesis, and what it does to the Fermi debate

Behavioral SETI is offered here as a hypothesis, not as a result. Its claim is narrow: that a sender constrained by energy, parsimony and longevity would encode not on a channel but in a relationship among observables, and would gate the message on the receiver's demonstrated capability rather than pay to make it legible. Everything in the search above is downstream of that, and none of the twenty-five nulls is evidence for it.

Where it touches the Fermi paradox is at a single premise. The paradox is usually posed as a contradiction between expectation and silence — they should be here, or we should have heard — and it inherits, mostly without argument, the assumption that contact would be self-announcing. A capability-gated signal is not. It is invisible from below by construction: a civilization holding the right archives, at the right cadence, that has not yet thought to search relationships sees exactly what a civilization with nothing to find sees. That weakens the premise without requiring absence, concealment, or a great filter — which is the whole of the contribution, and it is a smaller thing than it may sound.

The nearest hypothesis in the literature is Ball's zoo hypothesis (1973), and the relationship is worth stating precisely, because the two share a prediction and share no mechanism at all.

Zoo hypothesis (Ball 1973)Capability gating
Predictssilence despite presencesilence despite presence
Requirescoordination — many civilizations abstaining, indefinitelynobody to abstain at all
Fails whena single defector breaks itit does not have this failure mode: the barrier is difficulty, not etiquette
Enforcementexternal, and must be maintainedself-enforcing — opened by the receiver's own capability
A shared prediction is not a shared explanation. The gate repairs the zoo hypothesis's known weakness — its dependence on universal, permanent agreement — by removing the need for agreement.

It informs the debate. It does not settle either question. Two consequences are drawn here and a third is refused. It weakens the self-announcing premise, and it argues that a search of this kind should report coverage rather than detections, coverage being the only quantity that accumulates.

What it must not be used for is explaining a null. A framework that can absorb any non-detection explains nothing, and a hypothesis invoked to account for its own failure to find anything has stopped being a hypothesis. The twenty-five nulls in this paper are bounded statements about particular observables at particular amplitudes, and they are not offered as support for the framework that motivated looking. Neither the Fermi paradox nor the zoo hypothesis is resolved by anything here — the first keeps its force, the second keeps its difficulty, and what changes is only that one premise the paradox rests on turns out not to be free.

Read further