Younger Rock Beneath Older Rock: The North Sea Layers That Sank Through Time

The Seabed That Reads Backwards

Put your hand flat on a stack of paper. The sheet under your palm went on last, the sheet at the bottom first. You have known this since you were four, and around 1669 a Dane wrote it down as a law, and for three and a half centuries it has been the first thing anyone learns about reading rock [4].

Now go three hundred metres under the floor of the North Sea, somewhere between the Troll and Johan Sverdrup fields, and look at a wall of grey seismic data on a screen in Manchester. A mound sits in it, kilometres across, cored with sand. The sand is younger than the mud it is sitting under.

The usual escape routes are all shut. Nothing here is tilted. Mountain-building folds sequences back on themselves all the time, and geology keeps a settled vocabulary for the result, none of which fits here; nothing has been faulted and slid either. The layers around this thing are flat, continuous and thoroughly undramatic. The sand went down through them and stopped, and the older material that used to be underneath now lies on top, bulged into hills the size of a small town.

Jan Erik Rudjord and Mads Huuse put the case in print in 2025 [1]. They named the sunken sand bodies sinkites, the risen blocks of older ooze floatites. Their claim is broad. Hundreds of structures across the basin have done it, the mechanism is general, and nobody anywhere has documented the stratigraphic column turning itself over at a larger scale. We wanted to know whether the physics permits that. Everything below can be done with algebra and a calculator.

What the Grey Screen Showed

Reconstruct the case in the order the evidence arrived.

The dataset. A regional broadband three-dimensional seismic volume, the CGG18 pre-stack depth migration, covering the northern North Sea and tied to wireline logs and cuttings mineralogy from hundreds of exploration wells [1, 2]. Three-dimensional reflection seismic is why the argument is possible at all, resolving tens of metres across thousands of square kilometres, which is why one review called it geology's Hubble [12].

The target. Nearly a thousand wells had already punched through the Oligo-Miocene interval, the mounds had been on the logs for decades, and Rudjord's original brief was commercial: work out whether any of them held hydrocarbons. They were not. Almost no shows, and the project was shelved [2].

The accident. By his own account, a display choice broke the problem open. He rendered the seismic in plain greyscale instead of the conventional rainbow colour maps, then pulled sand volumes out with machine-learning assistance [2]. Colour had been hiding the geometry.

What the geometry showed. Four observations, every one of them awkward for the standard story:

1. No feeder. The textbook explanation for sand in a stupid place is injection from below: a buried sand reservoir gets overpressured, hydraulically fractures its seal, and squirts upward as dykes and sills [7]. That leaves feeder dykes and a parent sand body underneath. Here the reflectors beneath the mounds are continuous and undisturbed, and at the Grossbeak mound the nearest ten wells within a 10 km radius found no sand at all below [2].
2. A parent above. Mounds occur where the sand-rich Utsira Group lies above the ooze, apparently nowhere else [1, 2, 13], and mineralogy and biostratigraphy in the mound cores both point upward for a source.
3. Serrated margins. The mound flanks zig-zag in and out, and the press coverage settled on calling that geometry a "shark's mouth", which is unfair to sharks and useful to everybody else.
4. A pattern that matches the fault net. The mounds mirror the polygonal fault network in the enclosing ooze. Polygonal faults are a real and well-studied feature of fine-grained marine sediments, a honeycomb of small normal faults that forms as the sediment contracts and shear-fails during compaction and diagenesis [9, 10].

Put together: the sand came from above, entered the ooze along a pre-existing fracture network, and ended up below rafts of ooze that had been lifted. Call that the sinkite model.

A Dane, a Shark's Head, and the First Law

The rule deserves some fairness here, because it was hard-won. Revelation had nothing to do with it. A young Danish anatomist in Florence was handed the head of an enormous shark to dissect, noticed that its teeth were identical to the stony glossopetrae, the "tongue stones" that Tuscan farmers dug out of hillsides, and then had to explain how a shark's tooth gets inside a hill.

Nicolas Steno's answer, published in 1669 in a slim Latin work whose title translates roughly as Preliminary discourse to a dissertation on a solid body naturally contained within a solid, was that the hill had been laid down around the tooth, one layer at a time, in water, and that the tooth got there first [4]. From that one insistence he pulled out the principles that still frame the science. A layer is continuous until something interrupts it; layers start out roughly horizontal; and in an undisturbed sequence each layer is younger than the one lying beneath it. That last one is what this article is about. Superposition. A statement so plain it sounds like a tautology, and so powerful that reading time out of a cliff face depends on it entirely.

Geologists have always known about exceptions, and fussiness about the word undisturbed is where the whole rule actually lives: overturned folds flip whole sequences, and thrust faults shove old rock bodily over young. At centimetre scale, sedimentologists have catalogued "load casts" for a century: the bulbous sags where a layer of sand deposited on soft mud sinks into it, while the mud squeezes up between the sags in flame-shaped tongues [6]. Polish and Welsh workers pinned the physics down in the laboratory in 1970. Put a denser layer on a lighter one, let the lighter one lose strength, and you get a predictable polygonal pattern of sinking lobes and rising ridges [5]. The sinkite claim is that same load cast, scaled by roughly ten thousand. The physics is ordinary. Only the size is not.

Bench Notes: We Put Two Rocks on a Balance

Tuesday. We wanted the boring thing first: is the sand actually heavier? Because if it isn't, nothing else matters and we all go home.

One equation carries the whole article. A water-saturated sediment is partly grains and partly the brine in the pore space between them, so its bulk density is just a weighted average:

$$\rho_{\text{bulk}} \;=\; \phi\,\rho_{w} \;+\; (1-\phi)\,\rho_{g}$$

where \(\phi\) is porosity (the fraction of the volume that is fluid), \(\rho_w\) is the density of the pore brine, and \(\rho_g\) is the density of the mineral grains. We used \(\rho_w = 1030\) kg/m³ for North Sea formation water, a figure somebody looked up and nobody round the table objected to.

Then the asymmetry that does all the work. The sand is quartz, \(\rho_g = 2650\) kg/m³, which every mineralogy table agrees on. The ooze is accumulated skeletons of diatoms and radiolarians. Those skeletons are opal-A, an amorphous hydrated silica with water built into its structure, and its grains come in genuinely lighter, about 2100 kg/m³. So the ooze loses twice over, once on grain density and once on pore space, because microscopic skeletons stack into an open, rigid, absurdly porous framework that resists compaction in a way mud does not [15].

Numbers, at porosities that are ordinary for each material at shallow burial:

• Sand at \(\phi = 0.40\): \(\rho = 0.40(1030) + 0.60(2650) = \mathbf{2002}\) kg/m³.
• Ooze at \(\phi = 0.65\): \(\rho = 0.65(1030) + 0.35(2100) = \mathbf{1404}\) kg/m³.

A contrast of 598 kg/m³. The younger sand outweighs the older ooze beneath it by 1.43 times. Nobody at the table expected a margin that wide. A brick on a sponge.

Second question, same afternoon: how fragile is that? Solve for the break-even sand porosity, the loosest sand that still outweighs the ooze:

$$\phi_{\text{sand}}^{\max} \;=\; 1 - \frac{\rho_{\text{ooze}} - \rho_{w}}{\rho_{\text{qtz}} - \rho_{w}}$$

Against ooze at \(\phi = 0.65\) that gives \(\phi_{\text{sand}}^{\max} = 0.769\). Real shallow marine sands run about 0.35–0.45. Nothing natural comes close, and the stack would need roughly double the porosity to stand up. And when we swept the opal-A grain density across its whole plausible range, 1950 to 2200 kg/m³, the contrast only moved from 650 down to 563 kg/m³. So the inversion is the default. No knife-edge here to balance on, which we admit was faintly disappointing.

Recipe for Turning a Solid into a Liquid

Dense sand sitting on light ooze is not, by itself, enough. Sand is a solid. It has shear strength and it holds a slope, and it will happily sit there being dense and doing nothing for ever. To sink, it first has to stop being a solid.

The arithmetic here is startlingly simple. Take any depth \(z\) below the seafloor. The total vertical stress there is the weight of everything above it; the pore water carries part of that load, and the grain framework carries the rest. Only the grain-borne part governs strength. Karl Terzaghi saw that, and soil mechanics is built on it. Geologists have a name for the grain-borne part, the effective stress [16]:

$$\sigma' = \sigma_v - u$$

Strength then scales with effective stress through the friction angle \(\phi'\). We used 32°, typical for loose marine sand:

$$\tau = \sigma' \tan \phi'$$

Define \(r_u\), the excess pore pressure ratio: extra pore pressure over the effective stress you started with. Substituting gives a relationship with no curvature in it:

$$\tau(r_u) = (1 - r_u)\,\sigma'_{v0} \tan\phi'$$

A straight line. Full strength at \(r_u = 0\), exactly zero at \(r_u = 1.00\). At that point the pore water is carrying the entire overburden, the grains are touching nothing, and the sand is a liquid in the only sense that matters mechanically. Call it liquefaction. Earthquakes do it to loose saturated sand under every river delta [17].

Note what is not required. Nothing exotic in the threshold and nothing special in the chemistry. No critical velocity to exceed either. Raise the pore pressure to the total stress and you are done. Table 1 gives the bill at several depths.

Falling at One Centimetre a Year

So the sand is heavy and the sand can be made liquid. The remaining question is the only one that has real teeth: how long does it take a kilometre of sand to fall two hundred metres through mud?

We reached for the crudest defensible tool available. Stokes' law, 1851, a sphere creeping through viscous fluid [18]:

$$v \;=\; \frac{2\,\Delta\rho\, g\, R^{2}}{9\,\mu}$$

with \(\Delta\rho = 598\) kg/m³ from Section 4, \(R = 500\) m for a kilometre-wide body, and \(\mu\) the dynamic viscosity of whatever the sand has to push through.

Two features of that formula matter enormously. First, \(v \propto R^2\), so a body ten times wider sinks a hundred times faster, which is the entire reason the process can work at all on geological timescales. At a fixed viscosity of 10¹⁸ Pa s, a 50 m blob needs 1 950 kyr to cover 200 m, and a 500 m body needs 19.5 kyr. Second, \(v \propto 1/\mu\), and \(\mu\) is the one number in this entire article that nobody on Earth has measured. So we did not guess. We swept it across fifteen orders of magnitude, asking which values give a geologically sensible answer.

The driving force, for the record, is modest: \(f = \Delta\rho\,g = 5861\) N per cubic metre, which across 200 m of ooze amounts to 1.17 MPa, or about 11.6 atmospheres of steady one-directional push. Nothing dramatic about it. Relentless, though.

Result: if the effective viscosity is below about 10¹⁶ Pa s, the whole business finishes in under two centuries, which is fast enough that we ought to catch it happening somewhere on the modern seabed. Above about 10²⁰ Pa s, nothing finishes inside the Miocene–Pliocene window. Survivors: 10¹⁷ to 10²⁰ Pa s, giving descent times of 1.95 kyr to 1.95 Myr. The middle of that band, 10¹⁸ Pa s, puts the sand down at 1.03 cm per year and finishes the job in 19.5 kyr.

Is 10¹⁷–10²⁰ Pa s reasonable for a mixture of liquefied sand and biosiliceous ooze? We do not know. Neither does anybody else, which is the honest scandal here. For scale, water is 10⁻³, honey about 10, cold pitch about 10⁸, window glass at room temperature about 10¹⁸, rock salt 10¹⁷–10¹⁸, and the Earth's upper mantle about 10²¹. So the ooze has to behave, averaged over tens of thousands of years, like something between pitch and rock salt, which rock salt itself manages perfectly well: diapirs rise through denser sediment across whole basins on precisely these timescales. So the ask is not absurd. Unmeasured, though. Unmeasured for as long as anyone has cared, and the weakest joint in the argument.

Does the Spacing Come Out Right?

One more test, and this one is falsifiable in a satisfying way.

A gravitationally unstable layer does not collapse everywhere at once. It picks a wavelength. The classical two-layer viscous Rayleigh–Taylor result says that for a light layer of thickness \(b\) beneath a comparable dense layer, the fastest-growing disturbance has a wavelength of [19]:

$$\lambda_{\max} \approx 2.568\,b$$

The instability therefore has a built-in ruler. Feed it the reported ooze interval, 200 to 500 m thick [2], and it must produce structures of a particular size, whether we like the answer or not.

Prediction: 0.51 to 1.28 km. Observation, as reported: mounds around a kilometre across, several of them larger [1, 3].

Same decade, and no more than that. A one-line formula with no free parameters, fed a thickness measured by somebody else, landing inside the observed range is worth exactly as much as it sounds like. Encouraging. Not conclusive. Had it returned 20 metres, or 40 kilometres, the mechanism would be in trouble. It did not. The test passed, and passing was not guaranteed. Move on.

Five Questions We Kept Asking

The Best Case Against

A club rule. Get excited about a result only after writing the strongest objection to it. Here is ours. Not a token one.

Images interpreted, not rock measured. Nobody has drilled a sinkite margin and pulled out a core that shows younger sand in contact with older ooze in a way admitting no other reading. What exists is seismic geometry, logs, cuttings, and a model joining them elegantly. Seismic reflectors are acoustic impedance contrasts, not rock, and the discipline's history is decorated with confident readings of amplitude anomalies that later turned out to be something else.

The viscosity is a free parameter and we just demonstrated it. Our own Section 6 is, read unkindly, an exercise in reverse-engineering: we picked the durations that felt geological and read off the viscosities that produce them. No independent measurement constrains μ for liquefied Miocene ooze, and a four-decade-wide "surviving window" says only one thing: the data cannot rule the mechanism out. We produced it, and it is thin.

The ooze is not Newtonian and neither is the sand. Stokes' law assumes a smooth sphere in homogeneous Newtonian fluid at low Reynolds number. Our body is irregular, our medium is fractured, anisotropic, strain-rate-dependent and only partly rigid, and the descent follows pre-existing planes of weakness rather than open fluid. A different problem, wearing our problem's clothes. Treat our timescales as order-of-magnitude sanity checks and nothing more.

The proposer says so himself. In a candid post accompanying publication, Rudjord notes that having become "an expert" he can see his own tendency to explain features as sinkites "when other options may be equally plausible," and reports that presentations of the idea were often met with "disbelief, polite interest, friendly jokes or active resistance" [2]. The objections he received, he adds, were not backed by strong scientific arguments. A claim about the objectors, rather than about the model. Both things can be true: the critics may have failed to articulate a good counter-argument, and the model may still be wrong.

What would the steelman not say? It would not say the density contrast is doubtful, because that part is arithmetic and it held across every input we varied. It would not say liquefaction is exotic, because liquefaction is routine. The soft joints are these. The viscosity, again. The assumption that the seismic images admit only one reading. And the leap from "hundreds of structures in one basin" to "a new general process in geology."

The Ledger, and the To-Do List

The ledger

Every number this article asserts, in one place, with its source. Anything marked club model came out of our own script and carries our own assumptions; anything marked reported came from the published work or the authors' own account of it.

The to-do list

Four things would settle this, in rough order of decisiveness. We run a lunchtime club, not a research budget.

1. Drill a margin. One continuously cored, oriented well through a sinkite flank, across the sand–ooze contact. Biostratigraphy on either side gives absolute ages. If the sand really is younger than the ooze beneath it, microfossils will say so, and the argument stops being about acoustic impedance.

2. Measure the viscosity. Take modern diatomaceous ooze. Liquefy it in a rheometer under realistic confining stress and measure. One order of magnitude would do it, giving a plausible window or no window at all. Reference [1] reports analogue experiments with diatomite and beach sand. Quantitative rheology would be the next rung up, the cheapest item on this list, and the one that would move the argument furthest of anything proposed here. Strange thing to have to point out.

3. Go looking elsewhere. The model makes a hard prediction. Thick opal-A ooze, overlain by sand, in a seismically active basin. The Vøring margin fits, and so do the Faroe–Shetland Basin and the California borderland. A real process leaves sinkites in all three. An accident of one dataset does not. Look for a decade, find nothing outside the northern North Sea, and that null result tells you something.

4. Reprocess independently. A group with no stake in the hypothesis re-migrates the volume. Mound bases mapped blind. The claim rests on reflector continuity beneath the structures. Continuity is processing-sensitive. Best defended by somebody else finding it.

Until then the honest position is the uncomfortable one: the model is well built, it is permitted by the physics, it is genuinely surprising, and it is unconfirmed. The first law of geology has not been repealed. It has acquired a footnote. The footnote is 200 metres thick and a kilometre wide, and it is sitting under the North Sea waiting for somebody to drill it.

References