This page does not claim the Pyramid of Khafre was a machine. It asks a narrower engineering question: if the reported deep structures beneath it are ever independently confirmed, could Khafre's known passage network function as the upper mechanical interface of that system - a filter and impedance transformer between a fluid subsystem below and a 4.9-million-tonne mass above? We build the model, let it produce its own frequencies, and list what would kill it.
The deep system - if real - would be the power section. Khafre's role in this reading is different: a mechanical interface. Its passage network sits mostly at or below bedrock level, physically between the claimed shafts and the pyramid's mass. Long air columns, chambers, constrictions and granite barriers are exactly the components an acoustic engineer uses to filter frequencies and transform impedance. HYPOTHESIS
Nothing here asserts builder intent. This is a testable engineering hypothesis about what the geometry would do, whoever built it and for whatever reason.
Khafre's passage system has been measured and described by multiple investigators across two centuries - but no single authoritative high-resolution dimensional dataset exists. The historical record is: Perring & Vyse, Operations Carried on at the Pyramids of Gizeh in 1837 (digitized volumes, Internet Archive); Petrie, The Pyramids and Temples of Gizeh (1883); Maragioglio & Rinaldi, L'Architettura delle Piramidi Menfite V (1966); and Legon, "The Design of the Pyramid of Khaefre," Göttinger Miszellen 110 (1989), pp. 27-34 (print-only; archived site). The table below is the gate every number must pass before it enters the physics: source → page/plate → measurement → conversion → model value. Cells we have not yet read from a primary document say so.
⬇ dataset CSV ⬇ dataset JSON dataset v1.1 · 246 primary-source records · every cell below opens its citations · plate images archived on this site
Historical surveys disagree in local dimensions, and tertiary compilations disagree further. The extraction resolved some folklore (the ~1.2 m and ~1.8 m passage heights are BOTH real - entrance passage vs horizontal passage, per Petrie's station tables) and exposed real conflicts, quantified below. A set of ten decimal lengths supplied to this project (34.94, 39.37, 41.86, 36.95 m, etc.) could not be traced to Legon, Petrie, or M&R - with one exception found during extraction: 6.71 m exactly matches Vyse's "inclined entrance to lower chamber" (22 ft 0 in, vol. II p. 119). The other nine remain Reconstructed / provenance under review and are not used as established inputs. Legon's GM 110 paper itself: NOT OBTAINED after an exhaustive hunt (trail in the dataset metadata). This disagreement is not a footnote: it is exactly why the model below runs on an uncertainty ensemble rather than false centimetre precision. The uncertainty is part of the research.
| FEATURE | SOURCE | VALUE | Δ FROM MIN | NOTES |
|---|---|---|---|---|
| Upper descending corridor, length | Vyse 1840, II p.118 | 104 ft 10 in = 31.95 m | - | M&R's "sui 37 metri" is the reconstructed ORIGINAL floor incl. the lost upper portion; Vyse measured what existed in 1837. Not an error - different referents. Both kept. |
| M&R 1966, p.52 | 37.0 m (original) | +5.05 m | ||
| Entrance passage slope | Vyse 1840, II p.118 | 25°55′ | - | M&R disagree with themselves: main text "circa 26°46′", their appendix 26°30′. The spread is the honest input. |
| M&R 1966 appendix | 26°30′ | +35′ | ||
| M&R 1966, p.51 main text | ~26°46′ | +51′ | ||
| Belzoni's Chamber, length E-W | Vyse 1840, II p.118 | 46 ft 2 in = 14.075 m | - | ~10 cm apart; Petrie gives N and S walls separately (557.9 / 557.4 in). |
| Petrie 1883, p.105-6 | 557.9 in = 14.171 m | +0.095 m | ||
| Belzoni's Chamber, width | Vyse 1840, II p.118 | 16 ft 2 in = 4.928 m | - | ~5 cm apart. |
| Petrie 1883, p.105-6 | 195.8 in = 4.973 m | +0.046 m | ||
| Chamber height | Petrie 1883 (corner) | 206.4 in = 5.24 m | - | IDENTIFICATION UNCERTAIN as a pair: corner height vs gable apex are different measures; M&R explicitly reject Petrie's figure and adopt Vandier's. Recorded, not reconciled. |
| Vyse 1840 (extreme) | 22 ft 5 in = 6.83 m | +1.59 m | ||
| Passage width | Petrie 1883 (mean) | 41.29 ± .05 in = 1.049 m | - | 5 mm apart - the two best surveys essentially agree here. |
| Vyse 1840, II p.118 | 3 ft 5½ in = 1.054 m | +0.005 m | ||
| Subsidiary chamber, length | Vyse 1840, II p.119 | 34 ft 1 in = 10.389 m | - | ~7 cm apart. |
| Petrie 1883, p.108-9 | 411.9 in = 10.462 m | +0.073 m |
No source is declared "correct." These spreads ARE the uncertainty envelope the ensemble samples from.
Khafre's interior CONFIRMED: two entrances, both on the north side on an axis offset ~12 m east of centre - one opening in the face ~11.5 m up, one cut into the bedrock pavement. Two descending passages (surveyed slope 25°55′, width ~1.05 m), the upper one granite-lined near its mouth. A lower system that levels, hosts a subsidiary chamber (~10.4 × 3 m, gabled), passes a portcullis, then rises to rejoin the upper route. A long horizontal passage to Belzoni's Chamber (14.15 × 5.0 m, ~6.8 m high) - walls and floor cut in bedrock, roof of gabled limestone. A second portcullis guards the upper route. No confirmed air-shaft system exists; a claim of two small facing wall holes in the burial chamber circulates but is uncorroborated in the primary literature we could reach, so it is excluded from the model.
A set of ten decimal passage lengths attributed to J.A.R. Legon's Khafre survey (34.94, 7.88, 6.71, 14.65, 9.69, 39.37, 41.86, 16.02, 2.58, 36.95 m) was supplied to this project. We could not trace any of them to Legon, Petrie, or Maragioglio & Rinaldi; their sum exceeds the corroborated scale of the passage system, and 39.37 m matches a Khufu passage more closely than anything in Khafre. Legon's paper (Göttinger Miszellen 110, 1989) is print-only and was not directly readable during this audit. Their status is Reconstructed / provenance under review. Until someone reads Legon against these digits, the model below uses corroborated envelope values with every segment length adjustable, and treats exact segment lengths as UNKNOWN rather than filling the gap. If you hold the Legon paper, the sliders accept its numbers directly.
Quarter-wave arithmetic on isolated passages proves nothing - any long tube "has" a frequency. The model below is a one-dimensional acoustic network: each passage is a transmission-line element (length, cross-section, characteristic impedance ρc/S, complex wavenumber with a small loss factor); each chamber is a lumped acoustic compliance C = V/ρc²; junctions conserve pressure and volume velocity; portcullises are switchable boundary elements. MODELED
A - Khafre's air network alone. B - the reported water system alone. C - both, joined through the bedrock-coupling element. The interesting question is whether C merely superimposes A and B or creates shifted and new modes.
The arithmetic that started this whole investigation, shown for transparency. For an air column open at one end, f ≈ c/4L; for the reported water column, f ≈ c/L at its simplest. Passage lengths in the tens of metres put air quarter-wave frequencies near 2 Hz; 1482/648 ≈ 2.29 Hz; Khufu's measured response is 2.0-2.6 Hz. These values are not yet evidence of intentional tuning. The relevant question is whether the convergence survives the historical measurement uncertainty and the full coupled-network calculation. That is what the ensemble below exists to decide - and the model is forbidden from being tuned to reproduce 2.3 Hz.
| COLUMN | LENGTH | f ≈ c/4L (air) or c/L (water) | STATUS |
|---|---|---|---|
| A long descending passage | ~35 m | ~2.45 Hz | illustrative only |
| A long horizontal run | ~39 m | ~2.2 Hz | illustrative only |
| Longest plausible run | ~42 m | ~2.05 Hz | illustrative only |
| Reported water column | 648 m | ~2.29 Hz | REPORTED premise |
| Khufu measured response (NOT Khafre) | - | 2.0-2.6 Hz | measured, different pyramid |
Two rows of four shafts feeding two chambers is a geometry where relative phase matters more than raw resonance. Drive all eight together and volume sources add - maximum chamber pressure and maximum net vertical push on the bedrock. Drive chamber A's four against chamber B's four and the net monopole cancels - what remains is a rocking couple across the ~200 m baseline between the chamber groups. Alternate shafts pairwise and cancellation happens at shorter range. "Field" here means only a spatially varying pressure, stress and displacement pattern - nothing electromagnetic, nothing exotic. MODELED - symmetry consequences of the reported layout; the layout itself is REPORTED.
Side chamber: a closed cavity branching off a duct is a classic reactive element - depending on frequency it can split modes, absorb (anti-resonance) or shift phase. Toggle it in the controls above and watch the spectrum: that difference is a falsifiable structural prediction, because the chamber's real dimensions are surveyable to centimetre accuracy.
Portcullises: the granite barriers are archaeological fact; their operation as anything but one-time seals is not. The model treats each as a boundary element with three states - open (transmission), closed (reflection), intermediate (partial) - because that is a physics question, not a historical claim. Flip the states above: closed portcullises decouple whole limbs of the network and the spectrum reorganizes. Archaeological evidence establishes the portcullises, not dynamic operation as valves. That functional reading is a hypothesis being tested mathematically.
A single numeric alignment is worthless - so we run the killer comparison: ensemble A samples every uncertain segment length from its documented historical range (topology, chamber connections and cross-section preserved); ensemble B randomizes segment lengths freely under the same topology and total length. For each geometry we compute the coupled spectrum and the distance of its nearest mode to the underground fundamental. If A and B are indistinguishable, the real architecture carries no special tuning - and we will say so. Per-source models (Perring / Petrie / M&R / Legon geometries) are pending the provenance table above; until those cells fill, ensemble A brackets them.
The current result means that, under this particular null model, the documented geometry produces a closer modal match than the stated percentage of randomized layouts. It does not mean there is a corresponding probability that the pyramid was designed for this frequency, nor does it establish intentional tuning. The result must survive independent survey datasets and a more physically realistic coupling model. Ensemble A presently carries almost no dimensional spread (one documented disagreement swept); the planned cross-survey robustness test - Vyse, M&R and Petrie geometries run independently, pre-registered, no tuning after results - is what could make this claim-worthy, in either direction.
Cross-survey status (2026-08-17, updated): M&R's TEXT volume contains no corridor totals - but their separate Tavole drawings volume does. Plate extraction recovered an independent M&R length skeleton, and the two traditions, 126 years apart, agree on the load-bearing limbs to under 1%: corridor D 31.70 m (M&R plates) vs 31.95 m (Vyse) - also resolving the text's "37 m" as the original floor including the lost upper portion - and gallery G 39.35 m vs 39.13 m. Larger spreads remain on the lower system (I 34.15 vs 29.37; P 15.75 vs 7.29), partly datum-point differences, carried as ranges rather than adjudicated. The ensemble below now samples per-segment across both primary traditions. One segment (ascending A) is documented in neither beyond projection - it stays fixed and stated.
This version records the exact computational protocol used for the first cross-survey robustness test using independently extracted Vyse/Perring and Maragioglio & Rinaldi geometry. The protocol is preserved unchanged so that subsequent methodological improvements can be compared against the original result rather than replacing it. The randomized control was specified before the cross-survey result was observed and was not modified after seeing it.
v1.0 reference result · historical ensemble median 0.34 Hz from the underground fundamental · randomized control median 0.64 Hz · historical ensemble closer than 88% of randomized layouts · distribution overlap 7% · no seed (browser RNG; runs vary a few percent)
Interpretation: preliminary. The documented historical geometry produced a closer modal relationship to the reported underground fundamental than 88% of layouts generated by the specified randomized control. Not evidence of intentional tuning; the null model and coupling approximation remain unresolved.
⬇ analysis manifest (khafre-frequency-test-v1.json) · dataset: khafre-dimensional-provenance v1.1 (246 records, hash in manifest) - versioned independently of the analysis
Next (none of these have happened yet): 1. complete remaining historical dimensional extraction · 2. independent acoustician review · 3. blind modal prediction from Khafre geometry · 4. independent comparison with the reported underground frequency · 5. physically realistic mechanical coupling model · 6. improved, pre-registered null model (-> Analysis v2.0) · 7. experimental measurement where feasible.
An independent acoustician receives only the Khafre geometry - and is never told the target underground frequency. What frequencies do they predict? Only after their prediction is locked do we reveal the underground system's frequency. If an independently predicted Khafre mode lies near the independently reported underground frequency, that would be substantially more informative than any post-hoc frequency search - including ours. This test has not been performed.
Every mode found in 0.4-6.5 Hz across the historical-envelope ensemble, under current boundary states. No single frequency is "the" prediction - the distribution is.
No published measurement of Khafre's structural resonance exists. We verified this during the audit for this page: the 2026 ambient-vibration survey instrumented Khufu only. Khufu's 2.3 Hz belongs to Khufu. What the model can honestly say: Khafre is a near-twin in height and construction, so the same first-order similarity scaling that reproduces Khufu's measurement predicts Khafre near 2.2-2.4 Hz - a preregistered prediction (Test 4 of the preprint), waiting for someone to place accelerometers. A full structural FEM is not attempted: we lack the joint-stiffness and foundation data that would make it more than decoration. The structural side stays a stated prediction plus a base-excitation pathway, nothing more.
Before claiming anything, we ran an adversarial prior-art audit across academic literature, patents, five languages, and the full alternative-history corpus - designed to find the strongest case AGAINST novelty. The full report and scored bibliography are published: prior-art audit · bibliography (JSON).
The conceptual ingredients are not new, and the audit says so plainly: pyramid resonance has been proposed for decades; hydraulic pyramid theories go back to a 1959 patent; machine/signal-chain readings of pyramid architecture date to 1998; even multi-pyramid coupled operation exists in the fringe literature from 2016, nine years before the underground claim. The statistics are imported from other fields, and the mechanism physics - oscillating water columns in rock-hosted voids - is textbook hydrogeology.
What is new here is not the idea that pyramids might resonate, or even that ancient structures might have interacted with fluids, vibration, or sound. Those ideas have been proposed before. What we have not found is prior work that combines independently sourced historical Khafre geometry, documented survey uncertainty, a specific underground frequency model, and a randomized geometric null to quantitatively test the proposed coupling.
That is a claim about the literature search, not a claim that no one ever thought of the underlying idea. If you know of prior work that does this, we want to find it.
ALBA's Song of the Stone (2026, not peer-reviewed) used FEM and millions of randomized geometries to investigate acoustic relationships in Indian rock-cut caves. Our experiment is different in site, mechanism, geometry and research question - but the existence of this precedent means we cannot claim to have invented the general experimental strategy, and we don't.
The methodology is the contribution; the numbers are exhibits. The result's own behavior under scrutiny - weakening from 92% to 88% when the second survey tradition was added, rather than being protected - is the property we care most about preserving. Next step: this dataset, manifest and audit go to independent reviewers with one question - where is the weakest link?
A simple numerical frequency coincidence. This is where the idea started, and on its own it is worth nothing.
The independently extracted historical geometry and its documented uncertainty produce a robust coupling result across multiple independent survey traditions and reasonable uncertainty ranges - not at one cherry-picked setting. STATUS: NOT YET MET - ensemble A currently sweeps a single documented disagreement; the cross-survey test below is the gate.
An independent modern survey confirms the relevant geometry, and measured vibration in Khafre shows the predicted mode.
A direct underground-to-Khafre transfer function is experimentally observed. Nothing short of this establishes the coupled system.
The current model does not depend on pretending that the Khafre passage geometry is known to centimeter precision. Its purpose is to determine whether the proposed coupling survives the actual uncertainty present in the historical record. Establishing that uncertainty requires returning to the measurements themselves.
If the reported deep structures beneath Khafre are independently confirmed, the pyramid immediately becomes an interesting candidate for the upper mechanical interface of the system. Its existing network of long passages, chambers, constrictions, granite barriers and bedrock-cut spaces provides a physically plausible architecture for filtering, coupling and redistributing low-frequency mechanical energy. Whether that interpretation survives a realistic coupled model is the question.
We are not asking anyone to believe a story. We built a model that can fail - and listed, in public, exactly how. The next decisive step is measurement, not interpretation.