The sea is on fire.
Not the ships. The water. Flame lying flat across the swell, moving with it, clinging to hulls and to swimmers. It came out of a bronze tube shaped like an animal's head, fixed to the prow of a warship. Witnesses said it arrived with a roar and a foul smoke, and that water only fed it.
For roughly four hundred years, that was the thing standing between Constantinople and the end of the Roman world.
Then it stopped. And nobody who knew how to make it left instructions.
The refugee with a method
The story starts with a man running from a war.
Theophanes the Confessor, compiling his chronicle in the early ninth century, gives a name. Kallinikos, an architect from Heliopolis in Syria. His city had fallen to the Arab advance. He came over to the Romans, and he brought a technique with him. A fire that could be thrown from tubes, and that burned on the sea.
He arrived at the worst possible moment, which is to say the best one. Arab fleets were working the waters around the capital in the seventh century, and the city was not going to hold forever. Theophanes has the new fire destroying an Arab fleet at Kyzikos, across the water from Constantinople.
The dating is not clean. The traditional account puts the great naval siege between 674 and 678. Some historians now argue the campaign came earlier, in the late 660s. Theophanes was writing well over a century after the fact either way.
What the accounts agree on is the outcome. The fleet burned. The city stood.
What the sources say it did
Read the descriptions and they do not sound like a firepot lobbed off a deck.
This was projected. It came out under pressure, in a stream, from a device the Byzantines called a siphon, and it kept coming while the operator held it. It stuck to what it touched. Water did not put it out. Later accounts say the only answers were sand, vinegar, or dousing in something other than seawater.
In 941 a Rus fleet under Igor came down on the capital. Liudprand of Cremona, whose stepfather had served as an envoy to Constantinople, wrote up what happened. The Byzantines refitted a handful of old hulks with siphons fore, aft and amidships, took them out into calm water, and set fire to everything around them. Men in mail went over the side and drowned. Men without it burned.
The city was saved twice by this stuff, in the seventh century and again during the Arab siege of 717 to 718. An empire in slow retreat on every frontier had one thing nobody else had. And it knew exactly what that was worth.
Kept like a relic, not like a weapon
Here is where it stops being military history and becomes something stranger.
Around the middle of the tenth century, the emperor Constantine VII Porphyrogennetos wrote a handbook of statecraft for his son Romanos. It survives as De Administrando Imperio. It is a private document, an emperor explaining to the next emperor how the world actually works.
In it he tells the boy what to say when foreigners ask for the liquid fire.
Say it was revealed by an angel to Constantine the Great. Say the angel bound the emperors under oath never to make it for another nation, and to hand the prohibition down. Say that the man who once took a bribe to give it away was struck by fire from heaven when he walked into a church.
That is not an export restriction. That is a curse, written down by the head of state, for the eyes of his heir.
Whatever the fire was, the palace treated the knowledge of it as a sacred object.
The trail of the recipe
So what is actually written down about the mixture? Very little, and it is worth being precise about how little.
Anna Komnene, the emperor's daughter, describes an incendiary in the Alexiad, her account of her father Alexios I written in the twelfth century. She mentions resin from pines and other evergreens, ground together with sulfur, packed into reed tubes and blown out by a man with strong lungs, kindled at the mouth of the tube.
It is the closest thing to an ingredient list the Byzantine record gives us. It is also a hand weapon, not the pressurized ship-mounted siphon, and it says nothing about petroleum.
Modern historians generally read the missing element as naphtha, light crude skimmed from surface seeps around the Black Sea and the Caucasus. That is inference, and reasonable inference, but it is not a Byzantine sentence.
Latin Europe has the Liber Ignium, a book of fires attributed to a Marcus Graecus, surviving in manuscripts of the thirteenth century and later. It carries incendiary recipes. It is not an imperial document, and dating it has kept people busy for two hundred years.
The case against a lost secret
The conventional account is not that a formula vanished. It is that there was never one formula to vanish.
Mainstream historians of the period hold that petroleum incendiaries were known across the Near East, that Arab armies fielded their own naphtha units, and that "Greek fire" is a Western crusader label draped over several different weapons. The Byzantines themselves used a scatter of terms: liquid fire, sea fire, Roman fire, processed fire. On this reading the real advantage was never chemistry at all. It was the delivery apparatus, and the crews trained to work it without killing themselves.
The disappearance has a mundane explanation too. Siphons were expensive and the empire got poorer. The fleet decayed. Territory holding the oil seeps was lost. Nothing has to be forgotten for a weapon to stop being built.
By 1204, when the Fourth Crusade came up the Golden Horn, the defenders sent out fire ships. Every navy in the Mediterranean had fire ships. They had been ordinary for a thousand years.
The part that will not go away
Set the mundane explanation beside the documents and something does not sit flush.
The secrecy is not a legend attached later by chroniclers. It is in a father's instructions to his son, in the palace, about one technology and no other. The Byzantines wrote manuals on everything. Battlefield tactics, court ceremony, farming, siege engines, the ranking of ambassadors. On the weapon that saved the capital, the surviving record is close to silent.
And the engineering is not trivial. The historian John Haldon worked through the problem with a full-scale reconstruction, building a bronze force pump and a heated delivery tube and running crude oil through it. It produced a projected flame. It also showed how many things have to be right at once: the pressure, the preheating, the seals, the nerve of the man holding the nozzle on a moving deck.
That cuts both ways, which is why the argument never ends. If the secret was the machine, then a machine is exactly the kind of thing that dies when the workshops close and the last crew retires. If the secret was the liquid, then an oath held.
Either way, a technology that decided the fate of an empire went out of the world, and the empire that owned it wrote nothing down.
What is left is a silence
Either there was no single formula, and the fire is a Western name stretched over a family of ordinary incendiaries that fell out of use the way ordinary things do.
Or there was one, held by a handful of families under an imperial oath, and it went into the ground with them while the sources looked politely away.
Constantinople kept its archives for a thousand years. On this, the archives say nothing.