Findings
Seven questions put to the vision, and what came back.
These are seven questions we put to the vision, with what the arithmetic answered and what nobody yet knows.
The picture, briefly. Across the Galaxy, settlement would go by hops. A ship crosses to another star, a settlement grows there, and in time it builds ships of its own. The farthest a ship can cross is its hop, and the leading edge of settled space is the front. The Expansion page runs that process twice: once for people, and once for machines that build copies of themselves. In the Sun’s own neighborhood the first step looks different, because the Sun can send ships to every nearby star directly; the Neighborhood page shows that.
The numbers are central estimates from our own analyses unless a source is named. They are working results, not peer-reviewed ones, and every one deserves an independent check. If you can correct or extend any of it, respond.
- How many stars can be settled?It turns on one choice, and matters less than range.
- Does the map hold still?No. The neighbors change, though too slowly to matter to the first ships.
- Where is the fuel?Wherever there is ice. No fuel network is needed.
- Can a ship carry people as embryos?The engineering works out. The raising of children is the open part.
- Is seeding the fragile option?Less fragile than crewed ships, with one exception.
- Is there enough to build with?Almost everywhere.
- When could the first ship leave?The 2120s, if machine industry arrives first.
How many stars can be settled?
The answer depends almost entirely on what you are willing to call home. If settlers need a world like Earth, about 2.5 percent of nearby systems qualify, and the honest range runs from under one percent to seven. If they will live in habitats they build themselves, large rotating stations made from local rock and ice, around any ordinary star, the figure is about half, provided red dwarfs, the small dim stars that make up most of the Galaxy, are acceptable. It is about fifteen percent if only Sun-like stars will do. If any concentration of mass will serve, as it does for machines, every system counts.
It would be natural to expect this number to rule the pace of settlement. It does not. Once about four suitable systems lie within a single hop, the front moves as fast as it ever will, and having more makes no difference. With fewer than about one, it slows to the pace at which stars drift past each other, roughly 75 km/s. What decides which case a civilization is in is the length of its hop. To keep four suitable systems in reach, ships must cross 41 light-years if only Earth-like worlds will do, 15 if red dwarfs are acceptable, and 12 if every system counts.
So the order of leverage is range first, by a wide margin, then flexibility about where to live, then speed and the time between launches.
Not known. How common Earth-like worlds are, to better than a factor of ten. Whether people would choose to live around red dwarfs, which decides between the fifteen percent and the fifty.
Does the map hold still?
No. A map of the nearby stars is good for about three thousand years. After that the distances between them have changed enough to matter, and over a hundred thousand years the neighborhood replaces itself about twice.
For the first ships this makes no difference. They cross in a few centuries, and in that time the stars barely move. Nor do ships need to wait for a star to drift within reach, because range is not limited by fuel, as the next finding shows.
What drift decides is the long run. Settlements planted as neighbors do not stay neighbors, and nothing built on fixed routes between particular stars can last. It also means the choices keep widening: a ship that can only just reach Alpha Centauri has two systems within range today and about thirty at one time or another over the next hundred thousand years.
Not known. The full census. The faintest nearby objects have no measured line-of-sight velocity, and brown dwarfs are still being discovered.
Where is the fuel?
The starships studied in most detail since the 1970s are driven by nuclear fusion, and burn two light gases, deuterium and helium-3. Three results about that fuel shape everything else.
Distance is nearly free. At a fixed cruise speed, a four light-year crossing and a fourteen light-year crossing burn the same fuel. The longer one costs time, not propellant. Freight is impossible. A tanker delivers more fuel than it burns only if it cruises slower than its own exhaust, and that exhaust tops out near nine percent of light speed. No fusion tanker can profitably supply ships that cruise faster. Fuel can be made almost anywhere. Deuterium is present in all water ice, a few tens of grams in every tonne, and helium-3 can be manufactured from deuterium. Making one ship’s load over three centuries takes a few hundred gigawatts, about two percent of the power humanity uses today.
Together these say that there is no fuel network, and no need for one. Nothing is refueled on the way, nothing is supplied from home, and every settlement with ice and power can fuel its own ships. Since rock and ice turn out to be nearly universal, so is fuel.
Gas giants and brown dwarfs, bodies between planets and stars in size, do hold helium-3 ready-made, and mining it takes tens of thousands of times less energy than making it. A settlement beside one can launch its next ship sooner and with less industry. That is an advantage, not a requirement, and the fuel serves only the ships launched from that place.
Not known. How long a ship can fly and still work on arrival, which is what limits range once fuel does not. Whether helium-3 can be made at this scale as cheaply as we estimate. How much sooner a settlement beside a gas giant could launch than one without.
Can a ship carry people as embryos?
The usual picture of an interstellar voyage is a worldship: a vessel big enough to carry generations of living crew. The alternative is a seed ship, a small craft carrying a factory that can copy itself and a frozen store of human embryos, from which people are born and raised at the destination. The genetics is the easy part: samples from five hundred or more unrelated donors weigh about a hundred kilograms and, kept frozen, lose nothing over the voyage.
Radiation is the hard part of the hardware. Over a crossing of four centuries, the slow end of our estimates, cosmic rays deliver hundreds of times what embryos tolerate. Shielding a frozen store only half a metre across with plain mass then takes about four thousand tonnes. Without that shield a seed ship would be some two thousand times lighter than a worldship. With it, the advantage is about twenty-five to one, which is still decisive. A faster crossing needs less shielding.
How long before a seeded settlement launches its own ship? Our estimate is about two centuries, with a range from one to seven. Human biology sets a floor of sixty to a hundred years and contributes almost none of the spread. Nearly all of it is uncertainty about machines: whether a factory that fully copies itself can be built, and how fast it doubles.
The part with no defensible number is the first generation. Machines can feed and keep children. Whether they can raise them is another matter, and the nearest evidence we have, from studies of children raised in institutions, is discouraging. A settlement that fails this way does not launch its ships late. It never launches them.
Not known. Whether a factory can copy itself completely, with nothing supplied from outside. Whether a small magnetic shield could replace four thousand tonnes of mass. Our rough numbers say a superconducting coil of about four tonnes might do. Permanent magnets have been studied and do not work; the superconducting case has not been worked through. Whether a machine-raised first generation thrives.
Is seeding the fragile option?
It is natural to think seed ships trade safety for speed. Within the model that is false. A wave of settlements sustains itself if each one lasts longer than it takes to found the next. For seed ships that is about five hundred years. For crewed ships, which launch rarely and reach few targets, it runs from two thousand years to ninety thousand, depending on how many systems they can use. Across every combination we tried, seeding tolerated more fragility than crewed expansion, never less. Speed and robustness turn out to be the same quantity.
Nineteen of twenty seeded settlements would have to fail before crewed ships came out ahead. That is a reason to doubt the whole enterprise, not a reason to prefer worldships.
The exception is the one the model cannot represent. It treats each settlement’s fate as its own roll of the dice. But seed ships of one design carry one set of instructions for gestation, rearing and the handing on of a culture. If those instructions are flawed, they fail identically everywhere. Crewed colonies carry adults who can improvise around a fault. Seeded colonies carry the fault.
The defence is diversity: different ways of raising a first generation on different ships, each tested apart from the others before any launch. This is where the engineering meets the question this site begins with.
Not known. How to tell whether two ways of founding a settlement are different enough that they would not fail together. The model has no parameter for it.
Is there enough to build with?
Yes, nearly everywhere. Rock and ice survive around paired stars, around stars poor in heavy elements and around brown dwarfs. And since ice carries deuterium, where there is building material there is also fuel. Systems with building material are four to five times more common than systems with a gas giant or with a Sun-like star.
The scarce element is not iron or silicon. It is nitrogen: about two tonnes for each person, nearly all of it simply the air that fills a habitat. Even so, the poorest kind of system we considered could support a hundred trillion people, and a settlement growing at one percent a year would take more than two thousand years to notice the limit. Expansion in this picture is driven by the wish to go, not by running out.
One caution outweighs the rest. This finding rests on inference, not on bodies we have observed. A belt like our own Kuiper belt, the swarm of icy bodies beyond Neptune, placed around a nearby star, would be thousands of times too faint for any current telescope to detect.
Not known. How typical our Solar System’s nitrogen is. Elsewhere it has been measured in only a handful of bodies, all of them seen breaking up and falling onto white dwarfs. One held about five percent nitrogen by mass, richer than any comet we have sampled. That is encouraging, but it is a few bodies around dead stars, and every figure above assumes ours is typical.
When could the first ship leave?
We modelled the path to the first ship that stops at another star and leaves a self-sustaining machine base. The median launch falls in the 2120s if artificial intelligence substantially compresses the engineering, and near 2210 at a conventional pace. Arrival at Alpha Centauri follows around 2200 or 2355, a crossing of about 75 years in the first case and about 145 in the second. The ranges are wide: launch between 2093 and 2192 in the first case, 2139 and 2431 in the second. Even if every step of design and decision took no time at all, the physical steps, building, testing and the flight itself, would hold launch to about 2053 and arrival to about 2099.
The surprise is what sets the date. It is not fusion. In most runs of the model the engine is ready first. The pacing item is a factory that can rebuild itself from local rock with nothing shipped from Earth, which is both what makes the ship affordable and what the ship must carry. One number dominates: how long such a factory takes to double. Published estimates run from one year to more than fifty. Nobody has measured it, and each doubling of that figure costs about eighty years.
Two design questions shape the ship. One is how it stops. A magnetic sail, which brakes against the thin gas between the stars, sheds speed without fuel, but slowly, and it loses its grip as the ship slows. Our timeline assumes the engine does all the stopping, which roughly halves the cruise speed; a sail that did part of the work would allow a faster crossing. The other is where the helium-3 comes from. It has to be mined from a gas giant or manufactured; the Moon’s soil holds far too little.
Not known. The doubling time. Whether full self-replication, electronics included, is possible. How fast a fusion engine can really drive a ship that must also stop: our estimates run from one percent of light speed to about eight. And one matter outside engineering: no treaty or authority covers the launch of self-copying machines toward another star.
What we would measure first
If the vision had a research budget, these are the questions it would buy answers to, most valuable first.
- The doubling time of the smallest fully self-copying factory. It moves the departure date more than everything else combined.
- Whether a compact magnetic shield can protect a store of frozen embryos. It decides between four tonnes and four thousand.
- Whether truly different ways of founding a settlement can be designed and tested. It is the only defence against a failure shared by every ship.
- How long a machine can fly through interstellar space and still work. Range is limited by endurance, not fuel, and nothing has been tested over even one century.
- Whether Alpha Centauri A has a giant planet. It would let the nearest system launch its own ships sooner.
- The nitrogen in small bodies around ordinary stars. So far it has been measured only in debris falling onto white dwarfs.