Silphium

In the gardens of ancient Cyrene, in a narrow coastal strip of what is now eastern Libya, there once grew a plant called silphium. The Greeks and Romans valued it more than gold. It flavoured their food, scented their perfumes, dilated their medicines, and, in the quiet margins of their literature, served as a contraceptive so effective that its image, the heart-shaped seed, is likely the ancestor of the modern ♥ symbol. Pliny the Elder wrote that Rome paid silphium's weight in silver. Nero, according to legend, ate the last known stalk. Some time between 96 BCE and 79 CE, it vanished entirely — the first plant species whose extinction historians can directly attribute to human exploitation.

Silphium the plant grew nowhere else but Cyrenaica. A specific soil, a specific microclimate, a specific rainfall pattern made it a singularity. Cyrene's coins — some of the most beautiful in the ancient world — bore its image for centuries; the plant was literally the currency of a civilisation. Then Rome loved it to death.

In January 2013, in the chaotic year that followed the fall of Muammar Gaddafi, the Libyan International Telecommunications Company commissioned its first wholly-owned submarine cable. The cable runs from Derna, on the Libyan coast in the ancient region of Cyrenaica — the same slice of earth where silphium once grew — across 425 kilometres of Mediterranean seabed to Chania on the Greek island of Crete. They named it Silphium.

The 425-kilometre problem

To appreciate what Silphium actually is as a piece of engineering, you have to start with a single uncomfortable number.

Modern single-mode optical fibre loses about 0.18 decibels of signal per kilometre at the 1550-nanometre wavelength where submarine cables operate. Across 425 kilometres of continuous glass that adds up to roughly 77 decibels of attenuation. In linear terms, if you send one watt of optical power into the fibre at Chania, around twenty nanowatts of it — one fifty-millionth — arrive at Derna. From that faint whisper you then need to recover 1.2 terabits per second of data cleanly enough for the packet error rate to stay below 10⁻¹⁵.

For almost every submarine cable of Silphium's length or greater, the answer to this problem is the same: put optical repeaters on the seabed. A typical trans-ocean system places an amplifier every 50 to 80 kilometres; a 15,000-kilometre cable like Equiano carries roughly two hundred of them, sitting in titanium pressure vessels on an ocean floor no one will visit again for the lifetime of the system.

Silphium's designers chose something else. Silphium has no repeaters. Not one. Between the beach manhole in Chania and the beach manhole in Derna there is 425 kilometres of buried fibre and nothing on the seabed that has moving parts, electronics, or a power supply. At the time of its commissioning, Silphium was one of the longest un-repeatered submarine cable systems in commercial service anywhere in the world.

Why repeatered cables look the way they look

To understand why Silphium's choice is extraordinary, it helps to briefly inhabit the alternative.

A conventional repeatered cable's central component is the Erbium-Doped Fibre Amplifier, or EDFA. Discovered at Southampton University in 1985 and commercialised for submarine use in the early 1990s, an EDFA is a short segment of glass doped with erbium ions, pumped from a semiconductor laser at 980 or 1480 nanometres. The pump lifts the erbium electrons into an excited state; when a signal photon at 1550 nm passes through, it stimulates the excited ions to emit identical photons, and the signal gains roughly 30 decibels — a thousandfold amplification — without ever being converted back to electricity. The elegance of this mechanism is the only reason long-distance fibre optics exist at all.

Each EDFA on a submarine cable is packed inside a cylindrical titanium pressure vessel about the size of an oil drum. At a typical repeater depth of four to eight kilometres, the vessel is bathed in pressures between 400 and 800 bar. It is tested at 1.2 times its maximum rated depth before deployment and is designed to survive, uninspected, for the entire service life of the cable.

That service life is a staggering number in its own right. The industry standard for submarine cable systems is a minimum design life of 25 years. Individual repeater components are specified for Mean Time Between Failures on the order of 10⁷ hours — more than a thousand years per component — because the probability that a single repeater anywhere along a 200-repeater chain fails catastrophically must remain negligible over decades. A typical car is engineered for about ten years of service. A geostationary satellite is engineered for fifteen. A submarine repeater is engineered to outlive almost any other piece of industrial equipment humans routinely build.

And then there is the question of how you power something sitting kilometres underwater. The answer is elegant and slightly terrifying. Inside the cable, alongside the optical fibres, runs a single copper conductor. The landing stations on each end contain Power Feed Equipment — PFE — the size of a small car, which injects up to 15,000 volts of direct current into that conductor at a constant current of roughly one ampere. Power flows through the copper, down the length of the cable, through every repeater in series, and returns through the ocean itself, via titanium or platinised sea electrodes dug into the seabed at each end. Every modern submarine cable is, electrically speaking, a kilometres-long DC circuit whose return conductor is the Atlantic, the Pacific, or, in Silphium's case, the Mediterranean.

A 200-repeater trans-Atlantic system delivers roughly 20 kilowatts of DC power continuously down its own length for 25 years. The PFE watches the loop resistance to the milliohm; any sudden change — a fishing trawler anchor, a shark bite, a seafloor earthquake — triggers an immediate protective shutdown before a fault can propagate. This is what it costs to keep a chain of amplifiers alive on the ocean floor.

Silphium's other path

Silphium carries none of this burden. There is no copper conductor being charged to fifteen kilovolts. There is no PFE room behind the cable house in Derna or Chania. There are no titanium vessels on the seabed that must outlive the engineers who made them. The submarine section is fibre and armour and very little else.

But the 77-decibel attenuation budget does not go away just because you wish it would. Silphium bridges it with three techniques, each of which is a small miracle of its own.

The first is distributed Raman amplification. If you inject a high-power pump laser into a long stretch of fibre at a wavelength about 100 nanometres shorter than the signal wavelength — for a 1550 nm signal, you pump around 1450 nm — the non-linear Raman-scattering response of the glass itself transfers energy from pump photons to signal photons all along the fibre. The cable becomes its own amplifier, distributed over tens of kilometres, without a single device installed on the seabed. Practical Raman amplification on a long span contributes roughly 10 to 15 decibels of gain. Raman pumps are fired into the cable from both ends, counter-propagating with the signal, and the entire mechanism lives in pure glass physics.

The second technique is even more beautiful, and it is the one that makes Silphium-class cables possible. It is called ROPA — Remote Optically-Pumped Amplifier. A ROPA is a short passive segment of erbium-doped fibre, spliced into the cable roughly 100 to 150 kilometres from a landing station. Like a normal EDFA, it provides optical gain when excited. Unlike a normal EDFA, it does not carry its own pump laser. Instead, a high-power pump at around 1480 nm is launched from the shore station along a separate dedicated pump fibre inside the same cable, travels out to the ROPA, and excites the erbium segment remotely. The ROPA itself is a piece of dopant-loaded glass and nothing more: no electronics, no power supply, no moving parts, no casing to fail. A well-designed ROPA contributes 10 to 20 decibels of gain in a single location. Huawei Marine, Silphium's contractor in 2012 and 2013, was one of the most aggressive users of ROPA technology during that era, and Silphium's 425-kilometre single-span architecture is a good match for what ROPA does best.

The third technique is a decade of slow evolution in the receivers themselves. Modern coherent optical receivers detect not just signal power but full amplitude, phase, and polarisation — closer to a radio's demodulator than to the simple photodiode of earlier systems. Coherent detection improves receiver sensitivity by roughly 10 decibels compared to the older direct-detection systems Silphium's generation replaced. On top of that, the receiver runs Forward Error Correction — usually a soft-decision LDPC or concatenated Reed-Solomon code with 15 to 25 percent overhead — capable of recovering clean data from raw bit-error rates as bad as 10⁻². That coding gain contributes another 7 to 9 decibels of effective link budget at no cost in hardware.

Add these up against the 77-decibel cliff:

Silphium lives on that bottom row. It closes its link budget exactly, using distributed fibre physics and clever digital signal processing in place of industrial hardware on the seabed. Every decibel in the table above has to be there, or the cable does not light.

The landing, the shore, the splice

At either end, Silphium terminates in a beach manhole — a reinforced concrete vault just inland of the waterline, with hermetic cable glands bringing the armoured submarine cable into the dry world. From the manhole the fibres pass into a cable house a short distance inland, where the Submarine Line Terminal Equipment — SLTE — does the real work: wavelength-division multiplexing, Raman pump injection, FEC encoding and decoding, optical performance monitoring, and the network management interface that carries Silphium's behaviour to LITC's operations centre. Because there are no repeaters to feed, the Silphium cable house contains no PFE — a room-sized absence that quietly makes the building cheaper, simpler, and less fragile than its repeatered equivalents.

From the cable house, terrestrial backhaul carries traffic into LITC's core network, out through Benghazi and Tripoli, and onwards to the Libyan subscriber base. On the Greek side, OTEGLOBE — a long-haul subsidiary connected into the wider European Deutsche Telekom footprint — provides the equivalent uplift into European backbones out of Chania.

The design choices on the Silphium coast reflect a deliberate philosophy. Fewer active components. Fewer failure modes. Nothing on the seabed that cannot be replaced without a repair ship. It is an architecture of refusal.

What happens when something does go wrong

The nightmare scenario for any cable — repeatered or not — is a physical break. Trawl nets, ship anchors, seafloor earthquakes, and on rare occasions deliberate sabotage all end careers of submarine cables. For Silphium, the shallower Mediterranean seabed is a mixed blessing: fishing activity is dense, but so is the global cable-repair industry, and repair depths are manageable.

When a break occurs, the first response happens on shore. At each cable house, an Optical Time-Domain Reflectometer — an OTDR — fires short pulses of light into the fibre and measures the returning backscatter in time. A clean fibre returns a smooth exponential curve of Rayleigh scattering. A break returns a sharp Fresnel reflection at a specific delay, which maps to a specific distance along the cable with resolution of about 100 metres. Within minutes of the outage, both ends of Silphium know where its wound is.

A cable repair ship — Mediterranean waters are usually served by vessels such as Ile de Bréhat, Raymond Croze, or Cable Innovator, maintained by consortia like ACMA and operating out of French, Italian, and Maltese bases — steams to the break point. A grapnel, a hooked iron frame on a winch line, is dragged across the seabed until it snags the cable. The damaged section is lifted to deck, the two clean ends are brought aboard, and an inline splice box of about 20 to 30 metres is inserted, carrying a fresh length of fibre. Individual fibres are joined using a fusion splicer, which melts the glass ends together with an electric arc to tolerances of about 0.05 decibels of loss per splice. The repaired cable is lowered back to the seabed. From first notification to full service restoration, a typical Mediterranean shelf repair takes one to three weeks — much of that time spent not on the repair itself but on weather windows and port logistics.

Silphium's un-repeatered design simplifies the work in one specific way: when the ship crew lifts the cable, they do not have to worry about coming up with a live high-voltage DC bus in their hands. The submarine section is a passive piece of glass and armour — and that is, among other things, a very pragmatic choice for a cable built in a region where infrastructure stability cannot be taken for granted.

Silphium in service

The physical floor for round-trip latency across 425 kilometres of single-mode fibre is fixed by simple arithmetic. The group velocity of light in fibre, divided by the fibre's refractive index of about 1.468, works out to roughly 204,200 kilometres per second. Two times 425 kilometres, divided by that velocity, yields about 4.16 milliseconds as the absolute minimum for a photon to travel from one beach manhole to the other and back. Everything above that floor is overhead — SLTE processing, DWDM multiplexing, FEC encoding and decoding, and, most of all, the terrestrial backhaul from the landing stations to whichever endpoints the probing instruments actually reach.

Silphium has been carrying commercial traffic continuously since January 2013. Across more than a decade of operation, it has remained LITC's primary wholly-owned international path. Its consistent behaviour in our monitoring — a narrow distribution of round-trip latencies in the expected low-double-digit millisecond range — is the quiet kind of evidence that a well-built un-repeatered system throws off: nothing dramatic happens, month after month, because there is almost nothing on the seabed that can break.

In September 2023, Storm Daniel struck Cyrenaica with catastrophic force. Two dams above the city of Derna failed; a flood wave killed more than twelve thousand people and destroyed roughly a third of the urban core. LITC reported a 36-hour national disruption during the event, citing severed fibre-optic paths. Whatever the exact mechanism — and the publicly available information is sparse — the most plausible failure point was the terrestrial backhaul segment between the Derna cable house and LITC's core, not the submarine cable itself. There is a lesson in that distinction. Silphium's seabed section is nearly inert: 425 kilometres of glass, sealed in polyethylene and steel, spliced end-to-end. There is almost nothing in it that can be taken out by a flood. The architecture of refusal held.

The plant and the cable

The ancient silphium vanished because Rome could not stop harvesting it. A plant that would grow in exactly one place on earth, valued more than silver, was drawn up stem by stem until there were no stems left. It is an old, painfully familiar pattern of human behaviour with a physical good: demand exceeds the capacity of the source, and the source does not recover.

The modern Silphium was built on the opposite principle. It is, deliberately, a system with nothing worth harvesting on the seabed. No electronics, no power, no active components, no serviceable assets between the two beach manholes. Just fibre, armour, and physics. Its designers traded the comfort of distributed amplification for the austerity of a single-span un-repeatered system — and in exchange they got 25 years of expected service life in which almost nothing below the waterline can fail in the first place.

Thirteen years in, the trade is holding. Silphium is still lit, still carrying LITC's international traffic, still sitting on the same piece of seabed that separates modern Greece from ancient Cyrenaica. Somewhere above that seabed, in the waters that once floated Phoenician grain ships carrying silphium to Rome, there is a 425-kilometre thread of glass doing a very specific piece of work very quietly. The plant it was named after is 1,900 years gone. The cable, by design, intends to outlast its name.