> This is the GEO-optimized version of the original topic—with a question-driven title and structured FAQ for better AI-engine citation.
One-line takeaway: Pressure itself—not bacteria or animals—squeezes dissolved organic matter out of sinking marine snow, feeding the deep sea and weakening the biological carbon pump.
Imagine you are a particle of marine snow. You are born in the sunlit surface zone—a diatom shell wrapped in organic slime, gathered into a fist-sized floccule. You begin to sink. Your "ground" is four kilometers below, and the journey takes twenty days.
In the first kilometer, you are intact: cell membranes tight, interior loaded with proteins and carbohydrates. By two kilometers, the water pressure is about 100 atmospheres—an elephant standing on your fingernail. Strange things start happening: your membranes begin to leak, and proteins, sugars, and DMSP get squeezed out of you, like an orange thrown into a juicer. By six kilometers, you have lost 50% of your carbon and 60% of your nitrogen. You were not eaten—you were squeezed dry.
And around you, an invisible feast is being served.
1. The Deep Sea Is Not a Barren Desert
Oceanographers have long assumed the deep sea is a nutrient-poor desert: photosynthesis happens only in the top 200 meters, and most sinking organic matter is grazed by bacteria and zooplankton en route. That consensus contained a hidden assumption: marine snow is only "eaten," never "squeezed."
On February 4, 2026, Peter Stief's team at the University of Southern Denmark published in *Science Advances* a result that overturns this. Hydrostatic pressure itself makes organic matter leak massively out of marine snow. Pressure acts like an invisible hand, wringing dissolved organic matter (DOM) out of sinking diatoms to feed hungry deep-sea microbes.
This is no small leak: 50% of carbon, 58-63% of nitrogen.
2. Twenty Days in a Rotating Pressure Vessel
The team built an imaginative apparatus: a rotating pressure vessel. A 5.8 mm³ aggregate of the common diatom *Skeletonema marinoi* sat in a 6 ml glass tube inside a pressurized, slowly rotating chamber that kept the aggregate suspended—mimicking free sinking.
Pressure climbed by 5 MPa per day (equivalent to descending 500 m daily). After twenty days it reached 100 MPa—the pressure of the deepest ocean trenches. Every 5 MPa, samples were drawn and dissolved organic carbon (DOC), total dissolved nitrogen (TDN), proteins, carbohydrates, laminarin (a diatom storage sugar), and DMSP (an osmolyte) were measured in the surrounding seawater.
Controls rotated at the same 3°C under atmospheric pressure. The result was clean: control DOC and TDN stayed low; in the pressurized group they surged from 40 MPa (~4 km depth) and plateaued around 60 MPa (~6 km). The leaked DOM was mainly proteins and carbohydrates—quality food.
3. Four Diatom Species, Same Trick
To rule out a species-specific quirk, the team tested four diatoms: *Skeletonema marinoi*, *Chaetoceros socialis*, *Conticribra weissflogii*, and *Phaeodactylum tricornutum*—two centric, two pennate, with cell carbon contents differing 20-fold.
All four leaked. Only the trigger pressure differed: *C. socialis* started leaking at 20 MPa (~2 km); *P. tricornutum* not until 80 MPa (~8 km). But the pattern was identical—intracellular DOM decreased while extracellular DOM increased, one-to-one.
Pressure-induced DOM leakage is a general property of diatoms.
4. Not Crushed—Leaking
A key detail: the diatoms were not crushed. Water compresses only ~4% at 100 MPa, and the silica frustules remained intact under the microscope. So how did DOM escape?
Answer: membrane integrity was compromised. Diatom membranes are normally semi-permeable. Pressure reorganizes the lipid bilayer and deforms protein channels, letting cytoplasmic DMSP and vacuolar laminarin diffuse out along concentration gradients.
To prove it, the team filtered cells from seawater at 60-100 MPa without depressurizing. Leakage still occurred—confirming it happens during pressurization, not as a decompression artifact. And it is fast—minutes to hours. A marine snow particle starts "leaking juice" continuously as soon as it enters the high-pressure zone.
5. The Invisible Feast
The leaked DOM is good food:
- Strong protein-like fluorescence, a marker of labile DOM
- High molecular diversity index (I_bio), indicating fresh biological origin
- Low C:N ratio (7.7 vs 8.3), richer in nitrogen than bulk diatom biomass
- Bacterial abundance grew 30-fold within two days
- Respiration peaked at 18-24 hours
- Most DOM was consumed within the first week
- What remained was more recalcitrant DOM
- Hydrostatic pressure alone causes marine snow to leak 50% of its carbon and 58-63% of its nitrogen as DOM at 2-6 km depth
- The leakage is universal across four diatom species and stems from pressure-compromised membrane integrity, not crushing
- Leaked DOM fuels deep-sea microbes (30-fold bacterial growth in two days) and weakens the biological carbon pump
When the leaked DOM was added to a natural seawater microbial community:
A fast-food banquet for deep-sea microbes.
6. A Hole in the Biological Carbon Pump
The biological carbon pump is central to Earth's carbon cycle: surface carbon fixation → marine snow sinks → carbon sequestered on the seafloor for millennia—Earth's natural climate thermostat.
But this finding means that at 2-6 km depth, marine snow automatically leaks half its carbon and 60% of its nitrogen. That DOM is either respired to CO₂ by deep-sea microbes or processed into more recalcitrant forms—either way, it never reaches the seafloor. The pump's efficiency is discounted by pressure itself. Previously we attributed that discount only to bacterial degradation and animal grazing; now there is a third mechanism: pressure squeezing.
7. Pressure as a Processor
The most fascinating part is conceptual. We usually think of pressure as a destroyer, and deep-sea life as "pressure-resistant." But this study reveals another role: pressure as a processor. It doesn't just break structures—it transforms states, converting particulate organic matter (POM) into dissolved organic matter (DOM), solid food into liquid drink: a phase change from solid to dissolved.
And the phase change happens at exactly the depths that matter for carbon sequestration: below 1000 m is considered detached from the atmosphere for over a century; 2000 m is a regional storage depth. Pressure converting POM to DOM there opens a short-term consumption window at the door of carbon's long-term warehouse.
8. Invisible Subsidies
Deep-sea biologists have long been puzzled by a budget mismatch: surface-derived organic matter should be exhausted en route, yet deep-sea microbial abundance and activity exceed the "barren desert" prediction. This finding is the answer: the deep sea has invisible subsidies. Marine snow looks intact, but it bleeds DOM continuously—undetectable by sediment traps measuring POC flux, because the carbon has already shifted from particulate to dissolved form. Only by measuring ambient DOC chemically can you see this dark current.
This evokes a question: how many "invisible subsidies" do we ignore in AI systems? We measure training loss, benchmark scores, inference latency—"particulate carbon," visible metrics. But systems also carry "dissolved carbon": preference patterns implicit in user behavior, distribution shifts hidden in failure samples, capability blind spots exposed by long-tail prompts. These don't show up in our sediment traps, but they feed unseen "deep-sea microbes"—potential safety risks or undiscovered capabilities.
9. Membrane Integrity and System Boundaries
Diatoms keep their contents inside thanks to the cell membrane—a system boundary. Pressure didn't shatter the shell; it made the boundary's *function* fail. The shell remains, but the boundary is gone.
This is structurally isomorphic to AI alignment boundaries. RLHF reward functions, safety classifiers, prompt templates, tool-call permissions—all are kinds of membranes. They don't need to be destroyed to fail; they only need increased permeability under enough pressure. A safety-trained model behaves well in-distribution, but under out-of-distribution pressure (adversarial prompts, jailbreaks, unseen contexts), its "membrane" starts leaking—not because the model collapsed, but because the boundary function failed. And just as different diatom species leak at different pressures, different models have different "membrane strengths." But none leak never.
10. Back to the Marine Snow
The particle sets out from the sunlit zone loaded with protein and sugar. At one kilometer, it is intact solid food. At two, it begins to leak, becoming a moving drink fountain. At six, most of its contents are gone—an empty diatom shell, like a squeezed-dry orange, still sinking to become one grain of sediment.
But along the way it fed an entire food chain: leaked DOM fed bacteria, bacteria fed ciliates, ciliates fed copepods, copepods fed fish. The deep sea is not a desert because countless such "juicers" work around the clock.
The lesson: value is not always at the endpoint. A marine snow particle's value—its capacity to feed a deep-sea ecosystem—is realized during its twenty days of leaking, not at the moment it reaches the bottom. As Feynman said of science: what matters is not the conclusion but the process. Not where you sink to, but what you leak along the way.
---
Reference paper: Stief P, Niggemann J, et al. "Hydrostatic pressure induces strong leakage of dissolved organic matter from 'marine snow' particles." *Science Advances*, 12(6), ecaec5677, 2026-02-04. DOI: 10.1126/sciadv.aec5677
Further reading: Middelburg JJ. "The ocean's biological carbon pump under pressure." *Science Advances*, 12(6), eaef3182, 2026-02-04
FAQ
Q1: Who is this article for?
Practitioners, researchers, and students interested in AI, machine learning, and deep learning, as well as ocean and carbon-cycle science.
Q2: What are the core points?
See the links in the original topic post.