I Taking Electrons from Water
Every breath of oxygen you take was pulled out of water. In 1941, researchers tagged water by replacing its oxygen with a heavier kind, then gave it to a green alga. The oxygen the alga released carried the same tag.
Plants, algae, and cyanobacteria split water molecules apart, take their electrons, and discharge the oxygen left over into the air as waste. Today’s photosynthetic organisms produce oxygen so quickly that they could replace all the oxygen in the atmosphere in about four thousand years.
Yet for a very long time, Earth’s atmosphere contained almost none. Only about 2.4 billion years ago did oxygen first become significantly more abundant. Geologists call this the Great Oxidation Event. This essay is about how that waste remade an entire planet.
First, why split water at all? The hardest part of photosynthesis is not collecting light. It is finding electrons. Photosynthesis turns carbon dioxide into organic matter, and the essence of that transformation is to “add electrons” to carbon dioxide.
The carbon in carbon dioxide is electron-poor; the carbon in sugar is electron-rich. To turn the former into the latter, you have to find electrons somewhere and put them in. Every kind of photosynthesis therefore has to answer the same question: where will the electrons come from?
Light does not supply the electrons. It supplies the force needed to pull them out of somewhere else. The non-oxygen-producing photosynthetic bacteria alive today take their electrons from various sources: hydrogen sulfide, elemental sulfur, hydrogen gas, ferrous iron, or certain organic substances. Many of these bacteria are purple or green. Some can turn an entire pool a purplish red.
Those substances have one thing in common: they do not hold on to their electrons very tightly, so taking them requires little effort. Water is different. It holds its electrons extremely tightly. Line up substances according to how “willing” they are to surrender electrons, and water sits near the least willing end. Hydrogen sulfide, ferrous iron, and hydrogen gas all occupy much easier positions.
But these easily tapped electron donors are scarce. You find them only around volcanic vents, hot springs, and places where hydrogen sulfide seeps out. Anyone who learned to take electrons from water would never have to worry about raw materials again. Water is everywhere.
Bacteria that depend on hydrogen sulfide can do little more than stay close to a few springs. Cyanobacteria can drift across whole oceans. They used to be called blue-green algae; much of that glistening green bloom you see on lakes in summer is cyanobacteria.
So a photosynthetic bacterium taking electrons from hydrogen sulfide and a cyanobacterium taking them from water are not doing the same job at roughly the same level of difficulty. The cyanobacterium’s water-splitting machinery is called photosystem II. For an instant, it generates an electron-grabbing power among the strongest known in any biological system. And that is only the first hurdle.
The second is a matter of numbers. Extracting electrons from two water molecules and releasing one oxygen molecule means taking away four electrons, releasing four protons, and joining the two oxygen atoms together. But photons arrive one at a time. Each can supply only one installment of force.
It is like trying to use a crowbar that gives you just one lift to move something whose four corners must all be raised together. You have to save up first. The modern system does it by installments: it stores four successive photochemical oxidations on a metal cluster, then releases the oxygen in one go once all four have accumulated.
The first clue came from a series of flash experiments. In 1969, researchers in France placed photosynthetic material in darkness, then exposed it to very brief flashes, measuring how much oxygen came out after each one. The first two flashes produced almost none. The third produced the most. After that, a peak came every four flashes, at the seventh, the eleventh, and so on, with each peak becoming less pronounced.
The following year, the Dutch-born American biophysicist Kok organized the results into a cycle of stepwise accumulation. It explained why four consecutive photochemical events are needed to complete one conversion of water into oxygen. Researchers still call it the Kok cycle.
The charge-storing metal cluster consists of four manganese atoms, one calcium atom, and five oxygen atoms. Its shape came into focus in stages. In 2011, researchers in Japan finally captured it clearly enough: the ten atoms form a crooked chair.
Later, others used extremely short X-ray flashes to photograph it before the radiation could damage it, watching how it changed shape at the different stages of charge accumulation. Researchers working on artificial photosynthesis have spent decades trying to imitate it: to split water at room temperature using inexpensive metals. Nature accomplished the job with just four manganese atoms and one calcium atom.
There is no single agreed answer to where this cluster came from. One proposal is that the earliest photosynthetic bacteria first learned to take electrons from manganese in the water, and proteins gradually recruited the manganese into their own machinery. Other experiments show that, under illumination, today’s photosystem II can assemble the cluster itself from manganese ions in water. If the early system worked that way too, no ready-made, specialized precursor would have been necessary.
Oxygen-producing photosynthesis also needs two photosystems connected in series. The one that comes first, pulling electrons from water, is perversely called photosystem II because it was discovered second.
The electrons then reach photosystem I, which comes later in the sequence. They absorb another photon and are lifted a second time, gaining enough “electron-donating power” to be passed on, eventually, to carbon dioxide. One lift is not enough, so there are two.
Most living photosynthetic bacteria that do not produce oxygen have only one photosystem. Their electron donors are already easy to strip: taking electrons from hydrogen sulfide does not require such a high lift. Once is enough. Whether you need one photosystem or two therefore depends on where you intend to get your electrons.
There are several accounts of how the two photosystems came together. In one, the ancestor possessed both, and most lineages later lost one or the other. In another, the two systems were acquired from different bacteria and put together. Billions of years of losses and transfers leave today’s gene trees open to either reading.
Whichever account is right, the constraint is the same: where the electrons come from determines how many stages of lifting are required, how powerful an oxidant is needed, and what comes out as waste. What cyanobacteria release is oxygen. Plants and algae later acquired the same trick by taking whole cyanobacteria inside their own cells. The fifth essay will tell that story.
II A Signal Disappears
How can we tell that the atmosphere changed billions of years ago? The most important line of evidence is hidden in sulfur. First, a word about isotopes.
An element can come in versions of different weights because its atoms have different numbers of neutrons in their nuclei. Their chemical properties are almost identical; some are just a little heavier, others a little lighter. In most physical and chemical processes, the lighter version moves a little faster and reacts a little more readily.
After a process has run its course, the ratio of light to heavy versions in its products therefore shifts. The crucial point is that the size of this shift is usually proportional to how much heavier one version is than another.
Sulfur comes in several versions: 32, 33, 34, and 36. If sulfur-34 shifts by a certain amount relative to sulfur-32, sulfur-33 should shift by about half as much relative to sulfur-32, since its extra weight is only half that of sulfur-34.
This rule is called mass-dependent fractionation. It is extraordinarily reliable; almost every ordinary process obeys it. “Mass-independent fractionation” sounds cumbersome, but it simply means a departure from the weight rule: sulfur-33 or sulfur-36 strays noticeably from the line that scales with mass.
This rule-breaking fractionation was first produced in a laboratory in 1983. The American chemist Thiemens used an electrical discharge to make ozone and found that the proportions of its different oxygen isotopes shifted in ways that did not follow mass. Seventeen years later, he and two colleagues in his laboratory found the same kind of departure in sulfur from ancient rocks, bringing the phenomenon into geological history.
What process, then, could break the rule? The current explanation involves ultraviolet light striking sulfur dioxide. In an atmosphere with extremely little oxygen and no protective ozone, short-wave ultraviolet light can reach the atmospheric layers containing sulfur dioxide. The fine structure of light absorption in sulfur dioxide molecules with different sulfur isotopes does not simply scale with mass. As a result, the proportions broken apart by the light differ in ways that do not follow the weight rule.
Those products enter different sulfur reservoirs in the ocean and are eventually preserved in sediments. Measuring the tiny departures means extracting sulfur from rock, turning it into a gas, and feeding it into an instrument that separates and counts the different versions by weight.
In 2000, the three researchers found this rule-breaking sulfur widely distributed in rocks older than about 2.45 billion years, with large departures from the normal pattern. In younger rocks, the departures suddenly became smaller; after about 2.09 billion years ago, they could no longer be detected. This is one of the firmest pieces of evidence that the atmosphere changed.
Two different processes are involved in the signal’s disappearance. First, once oxygen becomes more abundant, atmospheric sulfur, whatever form it began in, is ultimately oxidized into the same substance, sulfate, and falls out of the atmosphere.
Sulfur that once followed separate routes, carrying positive and negative departures, is mixed into a single pool. The departures cancel one another out. With still more oxygen, ozone blocks that band of ultraviolet light as well, shutting the factory down altogether.
Second, oxidative weathering on the continents increases the ocean’s total supply of sulfate. Even if small amounts of sulfur bearing an anomalous signature still reach the sea, they are more easily diluted in a vast sulfur reservoir that has already taken the ordinary route. The record is washed out.
Cancellation, dilution, or a complete halt to production all leave exactly the same result in the rocks: no detectable signal. What the signal can tell us, then, is this: as long as it is present, atmospheric oxygen is extremely scarce. Models put the upper limit at roughly one hundred-thousandth of today’s level. Once it disappears, all we know is that oxygen has risen above that amount. We cannot tell how far above.
Other lines of evidence have their own logic, and their own blind spots. Certain minerals, including pyrite, commonly called fool’s gold, and a uranium-bearing mineral, oxidize and break down when they encounter oxygenated water.
If these grains appear in ancient river sands, rounded by water, they must have survived washing and transport without being oxidized. The water and atmosphere along their route must have had very little oxidizing power. In a modern river, such grains would oxidize and dissolve before traveling far. Ancient riverbeds in South Africa, dating to around 2.8 billion years ago, are a classic source of these minerals. They are also where more gold has been produced than anywhere else in the world.
Then there are red beds. Rocks usually get their red color from a rust-like coating on their grains, and rusting requires oxygen. In 2024, researchers studying an ancient sequence of Canadian rocks concluded that the rust coating its quartz grains had formed at the time of deposition. That placed some of the more securely identified early continental red beds at about 2.31 billion years ago. But rust can also be brought in by groundwater much later, so redness alone cannot establish what the atmosphere was like when the sediment was laid down.
Another famous rock alternates red and black, like a layer cake. It is called banded iron formation, and most of the world’s iron ore comes from it. In the Hamersley Range of Western Australia, whole expanses of cliff consist of these alternating red and black layers. They record vast amounts of dissolved iron being oxidized and settling out of the sea.
But the oxidizing agent need not have been oxygen in the atmosphere. It could have been a little oxygen produced locally by cyanobacteria in surface waters, certain non-oxygen-producing bacteria taking electrons directly from iron, or ultraviolet light from the sun. A layer of iron ore does not, on its own, mean an oxygenated atmosphere. Whoever did the oxidizing, that iron is all around us today: much of the ore used to make steel is iron that settled out of the oceans more than two billion years ago.
Ancient soils buried in rock, or paleosols, depend on another contrast. When weathering takes place without oxygen, water can readily carry iron away. With oxygen, the iron turns to rust and stays in the soil. The brick-red soils of southern China today look the way they do for this reason.
Whether iron was lost from or retained in an ancient soil profile can therefore indicate the state of the atmosphere. Paleosols older than about 2.44 billion years generally lost their iron. Converted into an oxygen estimate, that implies an atmospheric level below roughly 0.24 percent of today’s. The conversion depends on a model, which has to make assumptions about rainfall, drainage, temperature, and how long the soil took to form.
Several metals, including chromium, molybdenum, uranium, and selenium, are sensitive to different aspects of oxidation. Some register continental weathering, some the area of oxygen-free seafloor, and some local oxidation. They do not always agree. That is hardly surprising: they were never measuring the same thing. If one indicator gives a signal for a particular period and another does not, that may reflect real differences from place to place, rather than necessarily meaning that one indicator is wrong.
III How Oxygen Got In
Earth 2.4 billion years ago belonged entirely to microbes. The land was bare, and not a single animal lived in the sea. The most influential of those who pieced together the broad outlines of this history was the American geologist Cloud. During the 1960s and 1970s, he noticed that most banded iron formations were older than two billion years, while most continental red beds were younger. He proposed that the atmosphere had once changed from oxygen-free to oxygen-bearing.
Oxygen did not enter the atmosphere in a day, or in a single event. Put what we can now reconstruct into chronological order, and the story runs roughly as follows.
First, cyanobacteria learned to split water. The timing of that step is itself disputed. Some genetic estimates place it as early as three billion years ago, or earlier still. Others point out that some of cyanobacteria’s closest relatives have no photosynthetic members alive today. One such branch was discovered in the human gut and in groundwater in 2013. The machinery for splitting water may have been assembled later.
The rocks do contain traces of oxygen from before the Great Oxidation. In Western Australia lies a sequence of black shales about 2.5 billion years old. A NASA-funded project drilled into it in 2004 to recover fresh rock cores. In 2007, researchers detected enrichments of molybdenum and rhenium in those cores.
These two metals first have to be “washed” out of continental rocks by oxygen before they can reach the sea. A few years later, researchers detected selenium anomalies in the same sequence. That study also noted, however, that the oxygen might have been produced only in local terrestrial environments and need not already have entered the global atmosphere. An even earlier candidate is a sequence of rocks about 2.66 billion years old.
These “whiffs of oxygen” are themselves contested. In 2022, a team reexamined the shale using finer-scale analyses of tiny areas and argued that the molybdenum might have come from volcanic ash. The signals could also have arisen when the rocks were heated during deep burial or oxidized much later. The original authors replied that these later alterations could not explain all the signals. The argument continues.
Even if cyanobacteria had long been producing oxygen, the atmosphere would not immediately have filled with it. Think of pouring water into a bucket with holes in its bottom: reducing gases from volcanoes, dissolved iron in the sea, and sulfides in rocks were continually consuming the newly produced oxygen.
Only when oxygen production exceeded the leakage through those holes for a sustained period could oxygen accumulate. Patches of shallow water thick with cyanobacteria may have become oxygenated much earlier. Researchers call these places oxygen oases. Beyond the oases, the world remained oxygen-free.
Nor did the accumulation proceed at a steady pace. A 2006 atmospheric model showed that when oxygen was below roughly one hundred-thousandth of today’s level, the atmosphere could remain stably in a low-oxygen state.
A little more oxygen could produce a little ozone. The ozone blocked ultraviolet light, slowing the ultraviolet-driven atmospheric reactions that used oxygen to burn methane. Oxygen’s lifetime suddenly lengthened, and the whole atmosphere jumped to the high-oxygen side. Oxygen production could keep rising for a long time with no atmospheric response, then suddenly tip it over.
There are several explanations for why the holes later became smaller. One looks to volcanoes: in the early world, more eruptions occurred under the sea and their gases were more reducing. Later, as more volcanoes erupted on land, less oxygen-consuming gas was released.
Another looks to nickel. As the mantle slowly cooled, eruptions of nickel-rich magma became less common, and the ocean’s nickel supply declined. The enzymes methane-producing archaea use to make methane cannot do without nickel. Less nickel meant a weakening of methane, one of the great consumers of oxygen. Nutrient supplies may also have limited the rate of oxygen production. These explanations are not mutually exclusive. The dispute is over which one determined the timing of the tipping point.
The actual shift occurred about 2.4 to 2.3 billion years ago. Cores from three South African boreholes covering the same interval place the disappearance of the sulfur signal at about 2.33 billion years ago. And it vanished quickly, perhaps within just one million to ten million years.
Other studies put the first change earlier, near 2.43 billion years ago. Part of that hundred-million-year difference may arise because researchers are measuring different stages of the event. Part also reflects continuing disagreement over how to date and correlate rock sequences in different places.
How are those dates established? Thin layers of volcanic ash are often interleaved with the rocks. The ash contains tiny crystals called zircons. When zircon crystallizes, it takes in a little uranium but almost no lead, and uranium turns into lead at a fixed rate. Measure the ratio of lead to uranium in a crystal and you can tell how long ago it crystallized. Good measurements have an uncertainty of only one or two million years.
Even after the atmosphere tipped over, it did not settle there immediately. In 2021, Poulton and colleagues at the University of Leeds in Britain reported that, after first crossing the threshold about 2.4 billion years ago, the atmosphere fell back into low-oxygen conditions several times. It went back and forth for roughly two hundred million years, becoming permanently established on the oxygenated side only about 2.22 billion years ago.
During this interval, Earth also went through several ice ages. Ancient rocks along the northern shore of Lake Huron in Canada preserve at least three layers of debris left by glaciers.
The evidence of glaciation is easy to recognize: a large stone suddenly embedded in fine layers of mud. The most common explanation is that floating ice melted and dropped the stone. These glacial deposits date to between about 2.45 and 2.22 billion years ago.
One South African glaciation may have laid down its deposits not far from the equator. If ice could advance all the way to the equator, almost the entire planet would have frozen into a snowball. Geologists call this state Snowball Earth.
By around 2000, researchers had linked these events into a chain. The early atmosphere contained a great deal of methane, which is much more effective than carbon dioxide at trapping heat, molecule for molecule.
Oxygen produced by cyanobacteria destroyed the methane. The greenhouse weakened, and Earth cooled. In 2005, researchers pushed the idea to its limit: perhaps cyanobacteria learned to produce oxygen only then, and the emergence of that ability immediately plunged the planet into a snowball. But there are no ice cores from which to measure methane 2.4 billion years ago. Its abundance has to be inferred entirely from models.
The sequence of events has also been revised in recent years. In 2020, researchers argued that reliable oxidation signals preceded the crucial glaciation, making the proposed chain chronologically plausible. But a plausible sequence does not amount to proof of cause and effect.
Once oxygen had become established, it overshot. From about 2.22 to 2.06 billion years ago, carbon became “heavier” in carbonate rocks around the world. This was the largest and longest-lasting shift of its kind in geological history.
One explanation is that vast amounts of organic matter were buried underground, carrying light carbon away and leaving heavy carbon in the ocean, while also leaving surplus oxygen behind. This led to the proposal that oxygen was, for a time during that interval, far more abundant than it would be later.
Then it fell again. The buried organic matter was eventually brought back up and oxidized, consuming the surplus oxygen. A 2019 study, using a sulfate mineral about 2.018 billion years old, estimated that global photosynthetic production may have declined by more than 80 percent after the overshoot ended.
For more than a billion years afterward, oxygen remained low. Some researchers, working from chromium isotopes, place atmospheric oxygen during that time at less than one-thousandth of today’s level. Others put it at around one percent.
So remarkably little seemed to change on Earth during this interval that geologists simply called it the “boring billion.” Whether it really was boring will be the subject of the seventh essay. The eukaryotic cells discussed in the fifth established themselves during this time.
IV Why Oxygen Is Toxic
What did oxygen do to the organisms alive at the time? Start with something familiar: pour hydrogen peroxide onto a wound and a white foam appears. An enzyme in the wound is rapidly breaking the hydrogen peroxide down into water and oxygen. Hydrogen peroxide is water with one extra oxygen atom. Our bodies keep this enzyme on hand because peroxide is a by-product of oxygen, and leaving it around damages cells.
Oxygen itself is actually rather reluctant to react directly with organic molecules. The trouble comes from several substances it turns into as electrons are added to it one at a time: superoxide, hydrogen peroxide, and hydroxyl radicals, collectively known as reactive oxygen species.
The respiratory chain and many enzymes continually produce the first two whenever electrons leak out. Much of the harm radiation does to people also begins by generating these kinds of substances from the water inside cells.
We can pinpoint the targets of their destructive power. Superoxide particularly attacks the little components made of iron and sulfur inside enzymes. When the enzyme is damaged, the iron is released.
When hydrogen peroxide encounters free iron, it produces extremely reactive hydroxyl radicals, which attack DNA, proteins, and membranes almost on the spot where they form. Protection therefore cannot mean dodging them. It has to mean interception: clearing away superoxide and hydrogen peroxide promptly, minimizing the formation of the still more dangerous hydroxyl radicals.
Today’s cells have a whole repertoire of interception methods. One enzyme turns superoxide into hydrogen peroxide. Another breaks hydrogen peroxide down into water and oxygen, the very enzyme that makes a wound bubble. Other machinery repairs DNA and locks iron away. Almost all the iron in human blood is held tightly by proteins, with very little allowed to drift around on its own. This is why.
The enzymes that turn superoxide into hydrogen peroxide belong to at least three unrelated families: the iron- and manganese-containing enzymes form one family, the copper-and-zinc enzymes another, and the nickel-containing enzymes a third. Different lineages found their own solutions to the same problem. Differences in this machinery, and in membrane composition, also mean that organisms vary enormously in their sensitivity to oxygen.
Interestingly, some of this equipment predates the rise in oxygen. A 2021 analysis concluded that most of these enzymes in cyanobacteria have histories extending back before the Great Oxidation, while two other peroxide-breaking enzymes and the nickel-containing enzyme appeared only around 2.5 to 2.3 billion years ago. That makes sense. Before the Great Oxidation, cyanobacteria were already producing oxygen locally; ultraviolet sunlight could generate small amounts of hydrogen peroxide in the atmosphere and water; and even oxygen-free metabolism could create oxidative stress.
The poison later became fuel as well. Some microbes learned to turn it to their advantage, handing the electrons taken from food to oxygen at the end of the process. Oxygen’s powerful grip on electrons gives them an exceptionally large drop to fall through, and an exceptionally large amount of energy can be saved from the fall.
This is aerobic respiration. One large family of enzymes uses oxygen for respiration today. From their distribution, some researchers infer that the earliest members may predate the Great Oxidation and may already have been at work in oxygen oases.
But oxygen is not entirely good for us either. In the 1940s and 1950s, hospitals supplied premature babies’ incubators with high concentrations of oxygen. It was later discovered that this had blinded many children. Divers who descend too far while breathing highly concentrated oxygen can suffer convulsions caused by the oxygen itself.
Some organisms cannot tolerate oxygen at all. In 1861, the Frenchman Pasteur was studying butyric acid fermentation, the source of the sour, rancid smell of spoiled butter. He found that the microorganism doing the work died when air was introduced. Two years later, he coined a term for this way of living, literally “life without air.” That is where the word anaerobic comes from.
These organisms are still close by. The tetanus bacterium is an anaerobe. Its spores occur in soil and grow most readily in deep, narrow wounds or wounds containing dead tissue.
A deep nail puncture calls for prompt medical care; the need for vaccination or immunoglobulin depends on the wound and vaccination history. Rust itself is not the cause. Contamination and oxygen-poor conditions deep in the wound make this kind of injury especially concerning. The botulism bacterium also cannot tolerate oxygen. Trouble usually arises in home-canned foods, pickles, and fermented foods: the container is tightly sealed, there is no oxygen inside, and inadequate heating or pickling has allowed spores to survive. To grow these bacteria in a laboratory, you need a special anaerobic chamber filled with nitrogen and a little hydrogen. You work inside it through a pair of rubber gloves.
For these strictly anaerobic organisms, oxygenation of the land surface and shallow waters meant a poison had appeared on their doorstep. Their territory must have shrunk, and some lineages very likely disappeared altogether.
But we cannot calculate how many disappeared. People often speak of an “oxygen catastrophe,” sometimes even claiming that it killed off more than 90 percent of life. Part of this account comes from a famous popular-science book. In the book Margulis coauthored with her son Sagan in 1986, one chapter was called “The Oxygen Holocaust.”
Figures like these have no data behind them. An extinction rate needs a denominator: first you have to count how many species there were. Yet microbial fossils from that time are often just tiny spheres. You cannot tell whether you are looking at one species or ten. The microfossils more than three billion years old discussed in the first essay pose the same problem. If you cannot even establish what counts as “one,” the ratio is not merely difficult to measure. It is undefined.
Nor can the 80 percent decline in photosynthetic production discussed in Section III simply be converted into the extinction of 80 percent of species. A forest growing 80 percent fewer leaves does not mean 80 percent of its trees have died, still less that 80 percent of its tree species have vanished.
V Where the Anaerobic Biosphere Lives
Remember that before the Great Oxidation, all life on Earth had arisen in a world without oxygen. Today’s anaerobic microbes are still living as their ancestors did.
The anaerobic biosphere has not disappeared. It exists wherever oxygen is consumed faster than it can be supplied.
The nearest example is underfoot. Dig into the mud beside the sea or a pond. Often the surface is brown, but just a few centimeters down it turns black and foul-smelling. That is anaerobic microbial territory.
Organic matter decomposing in the sediment rapidly uses up the oxygen diffusing into it. Farther down, microbes that “breathe” nitrate, metals, and sulfate take their turns in sequence. At the bottom are methane-producing archaea. Of the entire tree of life, only archaea make their living by producing methane, and the enzyme machinery they use breaks down on contact with oxygen. Other organisms can release a little methane along the way, but in much smaller quantities.
The rotten-egg smell is hydrogen sulfide discharged by bacteria that breathe sulfate. The mud is black because hydrogen sulfide combines with iron to form black iron sulfide.
Rock fractures, aquifers, and the pores of sedimentary basins deep beneath the continents form another enormous oxygen-free habitat. In 2018, researchers used global geothermal conditions, temperatures, rock types, and measured cell concentrations to estimate that roughly (2 to 6) × 10²⁹ cells live within the upper five kilometers or so of the continental subsurface.
In 2008, researchers studying fracture water about 2,800 meters down in a South African gold mine found that almost the entire community consisted of a single bacterial species. It sustained a complete ecosystem on its own, living on hydrogen produced when radioactivity in the rocks split water. It needed neither sunlight nor other organisms.
The researchers borrowed a Latin phrase from a Jules Verne novel to name it: “bold traveler.” Most subsurface cells live extraordinarily slowly. Some estimates suggest that a cell there divides only once every few hundred years, or even once in a thousand years or more. In mud hundreds of meters beneath the seafloor, the organic matter settled millions of years ago. The cells live on those ancient provisions.
Animal digestive tracts are another set of persistently oxygen-poor environments. In the rumen of cattle and sheep, the largest of a cow’s four stomach compartments, and in the colons of many animals, microbial activity uses up incoming oxygen. These places teem with strictly anaerobic bacteria and methane-producing archaea.
They make the methane in cows’ burps. A single cow can belch out nearly a hundred kilograms a year. Most bacteria in the human large intestine are anaerobic too, and their numbers are comparable to the number of cells in the human body itself.
Permanently stratified waters can preserve vast volumes without oxygen. The Black Sea is the world’s largest anoxic body of water. Below a hundred or two hundred meters, its waters contain no oxygen and are full of hydrogen sulfide.
In 2017, an archaeological project using a remotely operated submersible found an ancient Greek merchant ship there, about 2,400 years old. The discovery was announced the following year. Both mast and rudder were still in place. The ship lies more than two thousand meters down on the seabed.
It is not that atmospheric oxygen has never reached the sea. The water is stratified, and oxygen cannot be replenished quickly enough to match its consumption by decomposing organic matter. The same thing happens in lakes. When a summer bloom dies, its decomposition can use up all the oxygen in the water. The lake stinks, and fish die in masses. Organisms that produce oxygen can also create a lakeful of oxygen-starved water.
Hydrothermal systems, too, are neither uniformly oxygenated nor uniformly oxygen-free. Hot, reducing hydrothermal fluids mix with seawater to different degrees, creating steep gradients. Their low-oxygen regions support methane producers, sulfate reducers, and microbes that live on hydrogen. One methane-producing archaeon collected from a hydrothermal vent can still reproduce at 122 degrees Celsius, making it one of the most heat-tolerant forms of life known.
How large is this biosphere? The answer has been revised several times over twenty years. In 1998, the American microbiologist Whitman and colleagues made a classic estimate: about (4 to 6) × 10³⁰ prokaryotic cells worldwide, with enormous numbers living beneath the seafloor. From this came a claim that remained popular for years: underground microbes might weigh about as much as all the plants aboveground.
In 2012, researchers recalculated the subseafloor estimate using measured data from more than fifty deep-sea drilling sites. They cut it by more than 90 percent. The change came from sampling, not new instruments: the calculation now included the central oceans, far from land, where cells are extremely scarce.
The 2018 study of the continental subsurface also recalculated the global total for prokaryotes. It came to only about 23 to 31 billion metric tons of carbon, between one-quarter and one-tenth of earlier estimates, and only 30 to 40 percent of the figure used in Section VIII. Most of the world’s bacteria and archaea are thought to live underground. This is also the estimate with the greatest uncertainty: the habitat hardest to sample is precisely the one with the greatest volume.
Strictly anaerobic metabolism still operates on a global scale. Among the greenhouse gases emitted by humans, methane’s contribution is second only to carbon dioxide’s. Total global methane emissions to the atmosphere today exceed six hundred million metric tons a year.
Natural wetlands account for about 160 million metric tons; ruminants such as cattle and sheep, together with manure, for more than 110 million; and rice paddies for more than 30 million. Flood a paddy and its mud becomes oxygen-poor, leaving methane-producing archaea busy around the rice roots. Common estimates attribute about 70 percent of this methane to microbes, chiefly methane-producing archaea.
Most of the methane in the seafloor’s flammable ice was also made by archaea in the mud. Working in the opposite direction, other seafloor archaea specialize in consuming methane. In partnership with bacteria, they intercept most of the methane rising through oxygen-free mud and keep it from escaping into the atmosphere.
VI A Waste Product and a Mineral Inventory
What did oxygen do to the planet? You could say the entire surface rusted during that event. One of the most unexpected answers lies in its minerals.
Those vividly colored copper minerals, the green of malachite and the blue of azurite, form only after copper has been oxidized. Without oxygen in the air, Earth would not have them. The pigments called mineral blue and mineral green in Chinese painting are made by grinding these two ores. Much of the blue in the Dunhuang murals comes from azurite. The most widely used green there is atacamite, another mineral that forms after copper has been oxidized.
The Statue of Liberty is clad in copper. When first erected, it was copper-colored; after twenty-odd years in New York’s air, it had turned green. The green layer consists mainly of new minerals formed as copper combined with oxygen, water, sulfur from the air, and chlorine brought in by sea winds.
In 2008, a team led by the American mineralogist Hazen took this idea to the scale of the whole planet and proposed “mineral evolution.” At first it sounds like a joke. Minerals do not reproduce. How can they evolve?
They meant that the variety of minerals has grown step by step through Earth’s history. When the planet first formed, far fewer kinds could form. Only after life oxidized the surface did elements such as iron, copper, uranium, and manganese enter higher oxidation states on a large scale, combining with other elements to make new crystals.
This is not about organisms growing crystals directly, as in shells, bones, and teeth. It is about life changing the environment so that minerals can form on their own. Hazen and colleagues estimated that a dry world like the Moon could produce at most around 350 mineral species. Earth has more than six thousand known species. The enormous difference has to do with water and plate tectonics, and with life as well.
At the time, they estimated that more than two-thirds of known minerals resulted, directly or indirectly, from life’s transformation of the surface. The largest single part of that transformation was this oxidation. In a 2010 article for a general audience, Hazen wrote that the oxidation more than two billion years ago made over 2,500 new minerals chemically possible, many of them products of rock weathering.
Later, the team refined its own accounting. As the database grew, they tagged every mineral with all its known routes of formation. The result: roughly 2,700 minerals, close to half the total, can form through processes associated with life. But only about 1,900, roughly a third of the total, cannot form at all without life.
Water, meanwhile, is involved in the formation of about 80 percent of minerals. Over a little more than a decade, the same team had taken a bold claim apart into three more precisely framed questions. The mineral inventory itself was also growing: about 4,300 species in 2008, more than 5,800 in 2022, and now over six thousand.
Oxidation also changed what the continents delivered to the sea. Sulfides in rocks were oxidized into sulfate, which rivers carried into the ocean. The marine sulfate reservoir grew larger and larger.
That growing reservoir, in turn, diluted the “rule-breaking” sulfur signal discussed in Section II. This was the second process behind the signal’s disappearance. Oxygen erased the evidence of its own arrival. The very things the Great Oxidation left behind made it harder to answer the question of when the Great Oxidation happened.
Metals such as uranium and chromium also changed oxidation state, becoming soluble in water and allowing rivers to carry them away. One astonishing consequence came to light in 1972, when French investigators discovered that uranium from the Oklo mine in Gabon, Africa, contained an abnormally small proportion of the isotope capable of fission.
Further investigation revealed that, about two billion years ago, the deposit had undergone nuclear fission all by itself underground, running intermittently for hundreds of thousands of years. One prerequisite was that oxygenated water had first dissolved the uranium, transported it, and concentrated it in one place. Without the Great Oxidation, there would have been no natural nuclear reactor here. It remains the only known site of its kind.
Oxygen did not only deliver things to the sea. Newly formed rust binds phosphorus strongly to its surface, capturing it again. Phosphorus is indispensable for making both genetic material and the cell’s energy currency. When it becomes scarce in the ocean, photosynthesis is constrained.
Ozone high in the atmosphere also began to appear around this time. Today’s ozone layer, if compressed to the air pressure at ground level, would be only about three millimeters thick. It blocks the part of the ultraviolet spectrum most damaging to DNA. Sunlight turns oxygen into ozone, so even a little oxygen can make a little ozone, enough to change atmospheric chemistry.
But blocking enough ultraviolet light to make the land surface suitable for life in the sun requires roughly a thousand times more oxygen, about one percent of today’s level. A new model published in 2022 also suggested that, under low-oxygen conditions, the ozone layer might have been thinner than earlier models calculated.
The atmosphere had oxygen, but the deep sea was slow to follow. For more than a billion years afterward, the ocean resembled a layer cake: oxygenated at the surface, oxygen-free below. In 1998, a researcher proposed that vast stretches of the deep sea at the time were foul with hydrogen sulfide. One reason was that banded iron formations largely stopped accumulating around 1.85 billion years ago, as though something had suddenly removed the ocean’s iron. Hydrogen sulfide could do just that by precipitating it out.
Later research found that more of the deep water was actually rich in dissolved iron. The foul-smelling waters lay mainly in productive coastal regions and marginal seas. An oxygen-bearing atmosphere and an oxygen-free deep ocean coexisted for a very long time. Today’s Black Sea is a small sample of that kind of ocean.
VII Three Claims of Increasing Strength
One statement is repeated everywhere: oxygen made complex life possible. Compressed into that sentence are actually three claims of different strengths.
Start with the energy figures. Fermentation is a way to produce energy without an external electron acceptor. In typical glycolysis followed by lactic acid or ethanol fermentation, one molecule of glucose yields a net gain of two molecules of energy currency.
For complete aerobic oxidation, modern textbooks commonly give a figure of about thirty to thirty-two. The thirty-six to thirty-eight many people memorized in secondary school comes from an older, more idealized calculation.
Per molecule of glucose, then, aerobic respiration yields about fifteen times as much as fermentation. During a hundred-meter sprint, muscles cannot wait for oxygen. Fermentation covers part of the energy demand. It is quick, but it cannot keep you going for long.
That multiplier, though, applies only to a comparison with fermentation. Anaerobic respiration also has electron transport chains. It can use nitrate, sulfate, or ferric iron as the terminal electron acceptor, and in many circumstances yields more energy than fermentation. Nor is the energy difference due to something oxygen “carries.”
It comes from oxygen’s strong pull on electrons, which creates a large drop between oxygen and organic matter. The respiratory chain stores some of that drop as a proton gradient across a membrane. The rotary machine described in the third essay then uses it to make energy currency.
Now for the three claims. The first: oxygen allows certain kinds of metabolism. This is the best-supported level. As an efficient terminal electron acceptor, oxygen permits high-yield aerobic respiration. Many enzyme reactions use oxygen directly.
In an entirely oxygen-free environment, these particular reactions cannot proceed along the same pathways. Both chemistry and modern physiology support this directly. Several steps in the synthesis of cholesterol in our bodies, for instance, require molecular oxygen.
The second claim: oxygen facilitates certain structures or ways of life. The evidence is moderate and highly dependent on which organisms we mean.
In 2005, the planetary scientist Catling and colleagues made a strong version of this argument. On Earth, large, complex organisms that sustain active movement and high rates of energy use almost all depend on aerobic metabolism. An adequate oxygen supply might therefore be necessary for complex life with high energy demands. One of their calculations concerned diffusion: for a body to grow beyond a few millimeters, if oxygen enters only by seeping through its surface, the surrounding concentration has to be high enough.
The strength of this argument lies in physiological energetics. Many modern tissues with high metabolic rates genuinely cannot maintain themselves for long in completely oxygen-free conditions. The human brain accounts for only about one-fiftieth of body weight, yet consumes roughly one-fifth of the body’s oxygen. Deprive it of oxygen for a few minutes and damage follows.
But what this demonstrates most firmly is that certain specific metabolic processes and tissue functions require oxygen. Energy figures alone do not demonstrate that the abstract property of “morphological complexity” must have a single universal threshold.
An experiment in 2014 directly weakened the version claiming that oxygen must be near modern levels. A Danish research group placed a common sponge in extremely low-oxygen water. Sponges have neither muscles nor nerves; they filter food from water flowing through their bodies and are among the simplest animals. This one survived at oxygen levels equivalent to 0.5 to 4 percent of today’s atmospheric oxygen level. Sponges are very likely among the earliest animal branches to have diverged, which gave the result particular weight.
The experiment yielded a very clear double result. It refuted the claim that simple multicellular animals require oxygen close to modern levels. It did not refute the claim that this sponge still requires some oxygen. Later experiments on another sponge lowered exposure to about 0.25 percent of the modern level and observed only relatively limited changes.
These experiments have limits too. Modern sponges are not physiological replicas of ancient organisms. The experiments run for finite periods. Body shape, fluid movement, food supply, and microbial symbiosis all affect the constraints imposed by diffusion.
The third claim: oxygen is necessary for complex structure itself. If that proposition is meant to encompass every kind of “complexity,” it has the weakest evidence.
Complexity has no single metabolic definition, and some structures require very little oxygen. In 2020, researchers found that a tiny animal parasitizing salmon had lost even its mitochondrial genome and did not respire using oxygen. It is a distant relative of jellyfish, though it spends its entire life inside other animals.
In 2009, the paleontologist Butterfield at the University of Cambridge raised a stronger objection. If oxygen at a low percentage of the modern level is already enough to meet the metabolic needs of basic multicellular animals, the long gap in the geological record between oxygen’s earlier presence and complex forms’ later appearance itself shows that oxygen cannot have been the sole clock controlling their arrival.
He also emphasized that biological activity can change ocean ventilation and the transport of organic matter. The causal arrow from oxygen to ecological structure may therefore include feedback in the opposite direction. Animals that swim, burrow, and filter water can move surface organic matter into deeper water more quickly, changing the distribution of oxygen in the sea. Animals do not merely wait for oxygen. They rewrite its distribution themselves.
This side of the argument is not claiming that oxygen has no role. It is separating two things. Minimum physiological requirements may be very low. The actual delivery of oxygen in an environment, however, is governed by boundary layers, diffusion distances, circulation, body shape, and behavior. The percentage of oxygen in the global atmosphere and the flux of oxygen actually reaching the cells of a particular organism are not the same variable.
The three claims grow successively stronger while the evidence grows successively weaker. Much of the dispute in the literature comes from folding all three into the same sentence.
VIII Who Has the Numbers?
Finally, look at today. In 2018, a group at Israel’s Weizmann Institute compiled a global inventory, putting all organisms on the same scales by measuring the mass of their carbon rather than their wet weight. The result: all life on modern Earth adds up to about 550 billion metric tons of carbon.
Plants account for about 450 billion metric tons, more than 80 percent of the total, and most of that is wood in tree trunks. Bacteria are the second-largest contributor at about 70 billion metric tons, most of them underground.
Fungi account for about 12 billion metric tons, archaea for about 7 billion, and protists for about 4 billion. Animals amount to only about 2 billion, less than one percent of the whole. Bacterial carbon outweighs animal carbon by more than thirty to one.
About half the animal total consists of arthropods. Most of that is in the sea: copepods, crustaceans smaller than a grain of rice, and creatures such as krill. Land arthropods, chiefly insects, account for only about 200 million metric tons. Fish make up roughly another third of animal biomass.
A few smaller figures: all humans together account for about 60 million metric tons of carbon, roughly one-thousandth of the bacterial total and about one ten-thousandth of all life. The livestock humans raise account for about 100 million metric tons, more than ten times the mass of all wild mammals. Wild birds account for about two million metric tons; poultry more than doubles that.
By cell count, microbes dominate even more decisively. Estimates put the global total of bacteria and archaea at (1 to 3) × 10³⁰ cells. The number of “individuals” across all life cannot even be pinned down. In the U.S. state of Utah, a stand of quaking aspen has more than forty thousand trunks connected to a single root system. Whether that counts as one or forty thousand depends on your definition.
The ocean tells a different story. Its photosynthetic organisms together contain only about one billion metric tons of carbon, while the organisms that live on them contain about five billion. The biomass pyramid is upside down. So the claim that plants dominate global biomass is largely a consequence of the enormous store of carbon in wood on land. It does not establish that multicellular organisms predominate in most ecosystems.
The most numerous photosynthetic organism in the sea is a cyanobacterium called Prochlorococcus. A milliliter of seawater can contain around a hundred thousand of them. The oxygen-producing branch still rules the oceans.
Now return to this essay’s construct. In this series, a construct is not only a human-made criterion. Anything that draws a boundary and determines what can remain and what cannot qualifies. Here it is the chemical environment of an entire planet: the atmosphere’s redox state, the ocean’s layering, the roster of elements carried away by continental weathering, and an inventory of minerals.
What rewrote that construct was metabolic waste. The oxygen-producing cells had no intention and no plan. For them, taking electrons from water meant gaining access to an electron donor found everywhere. Oxygen was the half they threw away. It was not the goal, nor the product, but the by-product. Something discharged as waste became the raw material for everything that followed.
Today, people even use this by-product to search for extraterrestrial life. Astronomers scrutinize the atmospheres of exoplanets. If a planet maintains a large supply of oxygen for a long time, there is a good chance that something there is continually making it. But oxygen can also be produced without life, so it remains only a clue.
This is much more specific than saying that organisms changed their environment. A particular metabolism discharged something as waste; that substance changed the chemistry of the planetary surface, making entirely different processes possible later on. It changed the atmosphere’s ultraviolet photochemistry, the minerals that could remain stable, the elements rivers delivered to the sea, and the ocean’s layering. It may even have triggered a global ice age. Yet the producers of these building materials did not themselves enjoy the benefit. For the oxygen-producing cells, oxygen was a burden: it generated reactive oxygen species that had to be cleared away by specialized enzymes.
Cyanobacteria have suffered from this themselves. The enzyme they use to fix nitrogen breaks down on contact with oxygen. Some cyanobacteria therefore set aside specialized, thick-walled cells along a chain of cells, switching off oxygen production inside them so they can concentrate on nitrogen fixation. They have to keep out of the way of their own waste.
In this series, what a construct can neither accommodate nor eliminate is called the remainder. It lies not among the things that were rewritten, but among those that were not. The remainder in this essay is the anaerobic biosphere.
It was not destroyed. It was squeezed into the places oxygen cannot reach: a few centimeters down in sediment, fractures in rock deep underground, animals’ guts, and the depths of stratified waters. Oxygen did not drive anaerobes off the planet. It drove them into places oxygen could not go. Together, those places are far larger than the surface world.
Nor are they far from you. Methane can be detected in the exhaled breath of 30 to 50 percent of people, made by methane-producing archaea deep in the gut. With every breath you take, you use the waste discharged by photosynthetic organisms. Inside you still lives a little patch of a world older than the Great Oxidation.
As for how large it is today, the greatest uncertainty falls precisely on the habitat with the greatest volume, and the one hardest to sample.