Abstract
Erbium is the material foundation of long-haul optical communication. Roughly 40% of world demand goes into erbium-doped fibre amplifiers (EDFA), where the Er³⁺ ion in a silica matrix, optically pumped at 980 or 1480 nm, amplifies the signal directly in the fibre across the C-band without conversion to electrical form. Every transoceanic cable, every long-haul terrestrial route and every inter-data-centre link runs on this effect. There is no engineering substitute. World production of erbium oxide is about 700 tonnes a year at $28-38/kg; approximately 99% of refined supply is controlled by a single country. In October 2025 erbium was placed under Chinese export control alongside holmium, thulium and europium; the measure was suspended in November 2025 with the suspension running to November 2026. There is no guarantee of extension.
This paper argues that erbium’s supply risk is not geological and not economic. Erbium is neither rare nor expensive. The constraint is the separation step: the classical solvent-extraction cascade of hundreds of stages, the industrial capability for which is concentrated in one jurisdiction. We describe ARBOK-SA – dissolution in the acid medium generated by the process itself followed by cold boiling under deep vacuum at ambient temperature – as a route to recovering erbium together with the rest of the rare-earth basket from liquid and semi-liquid waste streams that are already being handled and paid for.
1. The bottleneck is not the ore
Erbium has no dedicated mine. It is produced as a by-product of heavy-group rare-earth separation, arriving together with ytterbium, thulium and lutetium. Its crustal abundance is unremarkable – erbium is more common than tin. The reason a 700-tonne market can hold global telecommunications hostage is not scarcity of the element but concentration of one processing step.
Separating adjacent lanthanides is difficult precisely because they are chemically similar. Ionic radii differ by hundredths of an angstrom; chemical behaviour is nearly identical. The industrial answer since the 1960s has been solvent extraction: hundreds of mixer-settler stages in cascade, each producing a marginal enrichment, with large inventories of organic phase, large reagent consumption and a long capital payback. Building such a cascade is not a technical secret. It is a capital and operating commitment that most jurisdictions declined to make while a cheaper supplier existed.
The result is a structural asymmetry. A country that does not mine a single tonne of erbium ore can nonetheless be entirely dependent on erbium supply, because the separation capacity sits elsewhere.
2. Demand: why erbium cannot be substituted
EDFA amplification is not a preference. It is a physical coincidence: the Er³⁺ 4I13/2 to 4I15/2 transition emits in the 1530-1565 nm window, which is exactly where silica fibre has its attenuation minimum. Pump the ion at 980 or 1480 nm and the signal passing through is amplified in place, with no optoelectronic conversion, no regeneration, no per-channel electronics. This is what made dense wavelength-division multiplexing economically possible and what carries essentially all intercontinental traffic today.
Raman amplification and semiconductor optical amplifiers exist and are used in specific roles, but neither replaces EDFA in the long-haul trunk on cost, noise figure and simplicity. Thulium-doped amplifiers address a different band. Within the deployed C-band infrastructure, erbium is the only answer.
Demand is being pulled upward by AI infrastructure. Every new compute cluster adds optical interconnect; every new region adds trunk capacity. The remaining 60% of erbium demand is spread across Er:YAG lasers at 2940 nm – the wavelength of maximum water absorption, hence dentistry, dermatology and ophthalmic surgery – colouring of glass and ceramics, erbium-doped fibre lasers, and use as a burnable neutron absorber in nuclear fuel.
3. The resource nobody is treating as a resource
Phosphogypsum is the by-product of wet-process phosphoric acid manufacture: about five tonnes of gypsum per tonne of P2O5. It is stacked, not used. Global stocks run to billions of tonnes; annual global generation is on the order of 200-280 million tonnes, of which the United States accounts for roughly 30 million tonnes, concentrated in Florida.
Phosphogypsum carries the rare-earth content of the original phosphate rock, typically around 3,500 mg/kg total REE, of which erbium is about 5.8 mg/kg. Applied to Florida’s 30 million tonnes a year, that is approximately 174 tonnes of erbium in the annual flow alone – a quarter of world production – inside material that is already being generated, already being handled and already carrying a disposal cost. The accumulated stacks hold on the order of 5,800 tonnes of erbium, about eight years of world output, standing still and generating liability.
Three further streams carry erbium in dilute aqueous form: acid mine drainage from coal and sulphide basins, where REE mobilise into solution at low pH; process liquors from monazite treatment; and end-of-life optical fibre and spent catalysts.
The common property of all four is that they are liquid or become liquid, they are dilute, and somebody already pays to manage them.
4. ARBOK-SA
ARBOK-SA operates on that property. The material is dissolved in the acid medium the process itself generates, without purchase of an external reagent stream. The resulting solution is subjected to cold boiling under deep vacuum at ambient temperature: the pressure is lowered until the liquid phase boils without heating, and phase separation proceeds under conditions where the differences between species in solution become exploitable in a single pass rather than across a cascade.
The consequences are the ones that matter for deployment:
- No furnaces. The process runs at ambient temperature, so there is no thermal energy load and no refractory equipment.
- No membranes and no consumable separation media in the core.
- No purchased reagent inventory: the acid medium is generated and returned inside the loop.
- Water and acid return to the process rather than leaving as effluent.
- Separation occurs in a single pass, inside the solution itself, without the classical refining cascade of hundreds of stages.
The engineering claim is not that erbium is recovered more cheaply per kilogram than a Chinese cascade recovers it. The claim is that erbium becomes recoverable at sites that could never justify a cascade: a single phosphogypsum stack, a single mine drainage outfall, a single acid plant.
5. Economics
Erbium alone does not carry a project. A site handling one million tonnes of phosphogypsum a year contains about 5.8 tonnes of erbium, worth roughly $190,000 at current oxide pricing. As a revenue line that is modest.
Two things change the picture. First, that same 5.8 tonnes is close to a full percent of world output, produced from one stack in one year – which is a supply-security number, not a revenue number. Second, erbium never comes alone. The same pass delivers neodymium and praseodymium, dysprosium and terbium, lutetium, yttrium and the rest of the basket, against a feedstock whose handling is already being paid for by someone else. The project is counted on the basket; erbium is a strategic addition to it.
The correct economic framing is therefore not “what is erbium worth” but “what does it cost to not have erbium”. For an operator of transoceanic capacity, a licensing regime on EDFA-grade erbium oxide is not a raw-material cost item. It is an availability question about the network itself.
6. Policy dimension
The October 2025 control listing and its subsequent suspension to November 2026 define the planning horizon precisely. A suspension is not a repeal. Any party whose business depends on optical transport now has a dated question in front of it: what is the position on 10 November 2026 if the suspension is not renewed.
The answer cannot be a cascade. Cascade construction is a multi-year, capital-heavy programme, and the jurisdictions that would need one have spent three decades declining to build it. The answer available on the timescale that matters is recovery from streams that already exist, at sites that already operate, using a separation route that does not require the cascade.
7. Conclusion
Global communications rest on approximately 700 tonnes a year of a substance that is neither rare nor expensive, and that lies in Florida’s phosphate stacks in a quantity equal to eight years of world production. It does not need to be found and it does not need to be mined. It needs to be separated – and separation, not geology, is the whole of the problem.
ARBOK-SA addresses that step directly: dissolution in a self-generated acid medium, cold boiling under deep vacuum at ambient temperature, separation in a single pass inside the solution. Applied to phosphogypsum, acid mine drainage, monazite liquors and optical scrap, it converts a monopoly in one processing capability into a distributed recovery problem that any industrial jurisdiction can solve on its own territory.
ARBOK Strategic Research Institute
CC BY 4.0