The Origin of Life
Where Chemistry Ends and Heredity Begins
The Part That Turned Out To Be Easy
In 1953 Stanley Miller ran sparks through a flask of gases meant to stand in for the air of early Earth, and found amino acids in the residue. The experiment is famous, and it is usually filed as a first step toward assembling life in a jar. It was not that. What it showed is that the building blocks are cheap.
That finding has only hardened since. Amino acids fall out of meteorites. They form in interstellar ice. Adenine, one of the four letters of the genetic alphabet, assembles from hydrogen cyanide, which is common almost everywhere in space. Sugars appear from formaldehyde on mineral surfaces. Nothing about the ingredients is rare, and nothing about them is delicate.
So the hard problem is not the molecules, and stating it as "how did organic chemistry arise" points at the wrong difficulty. A pot of chemicals, however rich, is driven by thermodynamics. It runs downhill toward equilibrium and then it stops. A living population is driven by something else entirely: it copies itself, the copies differ, and whichever copies better becomes more common. That is selection, and it is a different kind of process, not a more complicated version of the same one.
The origin of life is the crossing between those two regimes. Nobody has found the moment it happened, and there are good reasons to think there was no moment at all – no first cell, no instant when chemistry became heredity. That absence is not a gap in the evidence. It is probably the shape of the answer.
What the Rocks Allow
Before the arguments, the constraints. Earth is about four and a half billion years old, and for a long time its first several hundred million years were assumed to be molten and hostile, a planet with no chance of holding water.
Zircon crystals from Jack Hills in Western Australia broke that assumption. They are the oldest scraps of Earth anyone has, some older than four point three billion years, and the ratio of oxygen isotopes locked inside them is the ratio you get from rock that has sat in liquid water at low temperature. Oceans came far earlier than expected.
At the other end of the window, Strelley Pool in the same region holds layered mounds about three point four billion years old that almost everyone reads as stromatolites, built by mats of microbes. Between those two dates sit the contested claims. Rocks from Isua in Greenland and Nuvvuagittuq in Canada, nearer three point eight billion years, hold carbon enriched in the lighter isotope, which is the imprint life leaves because enzymes prefer the lighter one. Processes with no biology in them can leave a similar imprint, and that argument has run for decades without closing.
Take the cautious reading and life is established by three point four billion years. Take the bold one and it is established by three point eight. Either way the gap between a planet cool enough for oceans and a planet demonstrably inhabited is a few hundred million years at the outside. Whatever happened did not need forever, and that is itself a clue: a process that needed a trillion coincidences would not fit in the time available.
The Common Ancestor Is Not the First
Every living thing descends from one population, and comparing genomes across the whole tree lets you reconstruct part of what that population had. The reconstruction is called the last universal common ancestor, and it is routinely mistaken for the first life. It is nothing of the sort.
Reconstructions disagree on the details and agree on the shape. That ancestor already had a genetic code and ribosomes to read it. It ran on the same energy currency every cell still uses. It fixed carbon, it held a proton gradient across a boundary, it repaired its own genome. It was an organism, and a capable one, with a great deal of machinery already in place.
What came before that ancestor left nothing anyone can sequence. The common ancestor marks the earliest point the tree of life can see, not the earliest point life existed. How far apart those two lie is unknown, and it may be a very long way.
The Case for the Vent
Two research programs have been arguing for forty years. Each is strongest stated on its own terms, because averaging them into a compromise produces a position nobody actually holds.
The first says energy came first, and it starts on the seafloor. Where seawater seeps into certain deep rocks it reacts with them, and the fluid that comes back out is warm, alkaline and loaded with hydrogen. Meeting an early ocean that was mildly acidic, that fluid sets up a difference in proton concentration across the boundary between them. The structures this builds are not solid chimneys but sponges: a labyrinth of small compartments separated by thin mineral walls, with the difference maintained across every wall, continuously, for as long as the rock keeps reacting.
The core of the argument is a physical observation about what living things do now. Every cell on Earth powers itself by pumping protons across a membrane and letting them fall back through a turbine. That mechanism, called chemiosmosis, is as universal as the genetic code, and it is a peculiar thing to have invented. It is far less peculiar as something inherited from surroundings that supplied the gradient for nothing.
Its strongest single piece of support is that universality, backed by which route into carbon looks oldest. Of the known ways to build organic carbon from carbon dioxide, one stands out for releasing energy rather than consuming it, and it runs on hydrogen under conditions these vents supply. Flow reactors packed with the relevant minerals do turn hydrogen and carbon dioxide into the expected small molecules without anyone pushing.
Its unpaid debt is heredity, and the debt is large. Nothing in this picture says where a copied molecule comes from. The chemistry makes precursors and it makes them convincingly, but it has never made a system that inherits anything.
The Case for the Pool
The second program says information came first, and it works on land. Its central observation is that a single molecule does both of the jobs a living thing needs done. RNA carries a sequence, so it can hold information, and it folds into shapes that speed up reactions, so it can also act.
The evidence for that is not a model. It is sitting in the middle of every cell alive. The ribosome, the machine that builds all proteins, turns out to be a ribozyme: its catalytic core is RNA and the proteins around it are scaffolding. The most conserved object in biology is made of the molecule this program says came first, which is a hard fact for the other side to absorb.
The chemistry has come a long way. For decades the obstacle was that nobody could join a sugar to a base to make a nucleotide, because that reaction does not want to happen. The way through was to stop attempting it: a route was found that arrives at the finished unit without ever making the sugar and the base as separate pieces, starting from materials any young planet would have. Wet and dry cycling on a mineral surface then links units into chains, which is why pools that evaporate and refill are attractive rather than incidental. Fatty acids close themselves into bubbles on their own, so a container is not the difficult part either.
Its unpaid debt is copying. No ribozyme has yet been built that copies a strand as long as itself, accurately enough to keep the line going, without a chemist supplying activated units and choosing the sequence of conditions. The program that owns heredity in principle has not yet demonstrated heredity in practice.
What Would Settle It
The uncomfortable part is that almost any new result can be absorbed by either side. Demonstrate prebiotic chemistry that works in a warm pond and the vent camp answers that ponds are not where the energy is. Find a vent mineral that concentrates nucleotides and the pool camp answers that concentration is not heredity. Both programs have been flexible enough to survive four decades of results, which is a warning sign rather than a strength.
One observation does cut. Bacteria and archaea, the two deepest branches of life, build their membranes from different lipids joined in different ways. If the common ancestor already had a modern membrane, that difference is hard to explain, because it would mean one lineage tore up a working membrane and rebuilt it. If instead that ancestor was housed in mineral compartments and its descendants only later grew membranes of their own, the difference is expected, because it happened twice. That is a real argument, and it favors the vent.
What would settle the matter outright is a self-sustaining replicating population assembled from plausible geochemistry and left to run without a chemist. Produce it under vent conditions and one program is vindicated. Produce it in a drying pool and the other is. Neither has been produced. Until one is, the debate is over which unsolved half looks more tractable, and that is a judgment about taste in problems rather than a measurement.
The Error Threshold
Underneath both programs lies a single obstacle, and it is the reason this problem is open. It is not about synthesis. It is about how much information a sloppy copier can hold on to.
Copy a sequence with some chance of error at each position, then copy the copy, and keep going. Errors accumulate. If the sequence is short, most copies come through clean, so there is always a supply of perfect ones for selection to work from. If it is long, almost every copy carries at least one mistake, and within a few generations the original is gone – not outcompeted, but dissolved, the way a message whispered down a long enough line arrives as noise. The dividing line falls roughly where the length multiplied by the error rate reaches one.
Manfred Eigen turned that into the vice this field is caught in. Copying accurately requires machinery. Machinery is encoded in long sequences. Long sequences can only be maintained by copying accurately. Each requirement demands one of the others first, and the loop has no obvious entry point.
The numbers are not comfortable. Copying RNA without enzymes, in the best hands, misreads something like one position in a hundred, which caps a maintainable sequence near a hundred units. The shortest ribozyme with any copying ability at all is about twice that long, and the capable ones are longer. The gap is not a detail to be tidied up later; it is the whole problem, and no self-sustaining cycle of copying with variation has ever been assembled from scratch. That, and not the synthesis of molecules, is why the origin of life remains unsolved.
Pattern or Accident
If the tape were rerun, how much would come out the same? The honest answer separates into layers, and they do not all get the same verdict.
Some features look close to inevitable given the chemistry. The amino acids that turn up most abundantly in meteorites and spark experiments overlap heavily with the set life actually uses, which is what you expect if the set was chosen by availability rather than by luck. Adenine from hydrogen cyanide is almost forced. Fatty acids close into bubbles without being told to. And the specific route into carbon that vents favor runs downhill, meaning it does not need explaining so much as permitting. Rerun the tape and a good deal of this probably returns.
Other features look like accidents that froze. Life uses one handedness of amino acid and the opposite handedness of sugar, and the underlying chemistry has no strong preference; something small tipped it and then the choice locked, because a mixed system copies badly. Ribose is a poor molecule for the job, unstable and hard to make, and it is not obvious it was the best candidate rather than the one that happened to be present. The assignment of particular codons to particular amino acids shows some chemical logic and a great deal of arbitrariness. Rerun the tape and these could plausibly come out otherwise.
The split is not clean, and presenting it as one would be dishonest. But the pattern in it is suggestive: what looks convergent is mostly thermodynamics, and what looks frozen is mostly bookkeeping. Where the physics is pushing, outcomes repeat. Where a convention had to be settled, whatever settled it first was kept.
Everything Rests on One Example
Everything above rests on one example. Every organism ever studied uses the same code, the same handedness, the same currency, because every organism ever studied is a cousin. With a sample of one there is no way to tell a law from a habit.
That is why a second sample would do more for this question than any amount of further work on the first. Life somewhere else, or a lineage here that turned out not to be related to the rest, would immediately sort what repeats from what merely happened. If it used the opposite handedness, chirality is an accident and the argument is over. If it landed on chemiosmosis independently, that mechanism is close to a requirement. One example cannot be interrogated this way. Two can.
Until then the honest position is narrow. We know when it happened, within a few hundred million years. We know it did not need exotic ingredients. We know the machine at the center of every cell is made of the molecule one camp bet on, and that the power supply every cell uses matches the environment the other camp bet on. And we know the step nobody can demonstrate is the one where copying becomes reliable enough to build on. Everything else on this page is argument.
