Primary vs. Secondary Aluminium: Where Bonds, Energy and Impurities Diverge | 原生铝与再生铝：化学键、能耗与杂质到底差在哪
Primary and secondary aluminium are the same element made by two processes that differ at the most fundamental possible point. Primary production breaks a chemical bond: it separates aluminium from oxygen in alumina by electrolysis. Secondary production does not break any bond: it melts metal that is already aluminium. Everything that follows — the twentyfold energy difference, the impurity problem, and the limits on recycling rates — is a consequence of that one distinction.
原铝与再生铝是同一种元素，却由两条在最根本处即分道扬镳的工艺制成。原铝生产断开化学键——用电解把铝从氧化铝中的氧手里夺出来；再生铝生产不断开任何化学键——它熔化的本来就是金属铝。此后的一切差异——二十倍的能耗差距、杂质难题、以及回收率的上限——都是这一个区别的后果。
1. The distinction that governs everything
| Dimension | Primary (Hall–Héroult) | Secondary (remelting) |
|---|---|---|
| Chemical task | Reduce Al2O3 to Al; break the Al–O bond | Melt existing metal; no bond broken |
| Feed | Alumina from bauxite (Bayer process) | Scrap, dross, process returns |
| Operating temperature | About 950–970 °C (molten cryolite) | About 700–800 °C |
| Principal energy form | Electricity (electrolysis) | Fuel gas plus electricity |
| By-product that cannot be avoided | CO2 from anode consumption; PFCs at anode effect | Dross (oxide plus entrained metal) |
| Product purity | Set by alumina purity and cell practice | Set by the scrap charged |
The last row is the one that decides what recycled aluminium can become, and it is the subject of sections 4 to 6. A primary smelter starts from a refined, chemically narrow input and controls the product. A remelter inherits whatever the collection stream contained.
2. Primary: the reaction, the floor, and what is actually achieved
The reaction as usually written, and what it assumes
2 Al2O3 + 3 C → 4 Al + 3 CO2 (E° = −1.18 V)
This is the form almost every summary gives, and it is balanced correctly — but it holds only at 100 per cent current efficiency. In practice part of the aluminium produced re-oxidises by reacting back with the anode gas, so carbon monoxide appears as well:
Al2O3 + (3/2ce) C → 2 Al + (3[2ce−1]/2ce) CO2 + (3[1−ce]/ce) CO
where ce = current efficiency (Grjotheim et al.; ASME/DOE inert anode technical working group report). At ce = 1 the expression reduces to the familiar form. Below that, CO is a product, not a loss term — the anode gas is a mixture.
The thermodynamic floor
Using JANAF thermodynamic data, Grjotheim and co-workers give the theoretical minimum energy for a cell at 977 °C and 100 per cent current efficiency as the sum of two terms:
| Term | Energy | Note |
|---|---|---|
| Heating reactants to operating temperature | 0.70 kWh/kg Al | Cell at 977 °C, 100 % current efficiency, no parasitic reactions |
| Reduction reaction enthalpy | 5.64 kWh/kg Al | |
| Total theoretical minimum | 6.34 kWh/kg Al |
Expressed as a function of current efficiency, the same source gives: minimum energy = (1.43 / ce) + 4.91 kWh/kg Al.
What plants actually use
| Region (IAI, 2025 average) | kWh per tonne | Note |
|---|---|---|
| South America | 15,746 | Highest reported |
| Europe | 15,482 | |
| Gulf Cooperation Council | 14,891 | |
| North America | 14,721 | |
| Oceania | 14,633 | |
| Asia (excluding China) | 14,586 | |
| Africa | 13,866 | |
| China | 13,203 | Lowest reported |
| World | 13,880 |
IAI defines smelting energy intensity as AC and DC power for electrolysis per tonne of aluminium, including rectification and normal smelter auxiliaries up to the point where liquid aluminium is tapped. It excludes casting and carbon plants. IAI states this series is not suitable for life-cycle assessment; LCA work should use the IAI life cycle inventory reports instead.
The gap that cannot be closed. World average practice at about 13.0 kWh/kg sits at roughly twice the thermodynamic floor of 6.34 kWh/kg. The decomposition voltage is about 1.2 V while a working cell runs near 4.2 V; the difference is anode overpotential, bubble-induced resistance, ohmic drop in the bath, and losses in anodes, cathodes and busbars. Some of that is irreducible — it is the heat that keeps the bath molten. A smelter cannot engineer its way off this floor.
Process emissions that are not electricity
| Source | Quantity | Note |
|---|---|---|
| Carbon anode, theoretical | 333.3 kg C/t Al → 1,222 kg CO2/t Al | Stoichiometric |
| Carbon anode, actual net | about 400 kg C/t Al → about 1,467 kg CO2/t Al | Air burn and the Boudouard reaction add to consumption (ICSOBA 2025) |
| Carbon anode, other estimates | 0.45 C/Al → 1.65 CO2/Al | Reported for cells with higher excess consumption |
| Anode baking (prebake only) | 0.05–0.15 t CO2/t Al | Combustion of volatiles |
| PFC at anode effect | IAI 2024 total: 5,979 kt CO2e | See the GWP note below |
PFCs form when dissolved alumina falls locally low enough that the anode reaction changes. The cell voltage rises and fluorine from the cryolite combines with anode carbon:
4 Na3AlF6 + 3 C → 12 NaF + 4 Al + 3 CF4
2 Na3AlF6 + 2 C → 6 NaF + 2 Al + C2F6
A number that changes with the edition you use. The global warming potentials applied to CF4 and C2F6 have been revised across IPCC assessment reports: AR4 gave 7,390 and 12,200; AR5 gave 6,630 and 11,100; AR6 gives 7,380 and 12,400. IAI adopted AR6 values from 2024. A PFC inventory quoted without its GWP edition is not comparable with one quoted from another year — the same physical emission yields a different CO2e figure depending on the edition applied.
3. Secondary: the “5 per cent” claim, and what it actually measures
A units correction that matters, because the claim is usually quoted without it. “Recycling uses only 5 per cent of the energy of primary production” is correct — and it is a primary-energy comparison, not an electricity comparison. IAI reports primary aluminium at 186 GJ per tonne from mine to cast house, and recycled aluminium at 8.3 GJ per tonne, a saving of 95.5 per cent. Those figures include fuel, feedstock energy and the energy used to generate the electricity.
Primary smelting electricity (about 13,880 kWh/t) and remelting energy (8.3 GJ/t) are therefore not quantities that can be divided into each other. The 5 per cent is real; it is not a ratio of two electricity numbers.
The regional datasets behind the recycled figure
| Dataset | Primary energy | Note |
|---|---|---|
| Japan (2015) | 5.7 GJ/t | Wrought alloys only; excluded from the average used |
| Europe (2018) | 7.2 GJ/t | Europe reported 8.5 GJ/t in 2015 |
| North America (2015) | 9.2 GJ/t | |
| IAI states there is no global estimate for aluminium produced from scrap; the figures above are the available regional datasets. | ||
A separate European technical assessment gives the split for remelting new scrap as about 3.8 GJ/t thermal and 0.45 GJ/t electrical — that is, most of the energy enters as fuel, not as power.
Electricity and yield by furnace type — the numbers used for selection
| Furnace type | Energy (kWh/t Al) | Metal yield | Dross (kg/t scrap) | Best fit |
|---|---|---|---|---|
| Induction | 600–900 | 95–98 % | 20–50 | Clean, homogeneous scrap |
| Reverberatory | 900–1,500 | 90–95 % | 40–80 | Large-scale secondary production |
| Salt-flux rotary | 1,000–1,500 | 90–96 % | 50–100 | Highly contaminated scrap |
| Rotary | 1,100–1,600 | 88–94 % | 60–120 | Contaminated scrap streams |
The selection trade-off runs in the opposite direction to the one people expect. The lowest-energy furnace (induction, 600–900 kWh/t) also gives the highest metal yield (95–98 %) — but it is the least tolerant of dirty feed. The highest-energy furnace (rotary, 1,100–1,600 kWh/t) gives the lowest yield (88–94 %) and generates the most dross, because that is the furnace you choose when the charge is contaminated. Energy intensity and yield move together; scrap tolerance moves against both. A plant buying contaminated feed is buying higher energy use and lower yield, and that is a feedstock decision made before the furnace is chosen.
Melt loss, and how much of it the coating causes
| Measure | Value | Basis |
|---|---|---|
| Dross generation | commonly 15–30 kg per tonne melted | Published surveys |
| Total melt loss | often cited 1–10 %; oxidative loss averaging about 4 % | Wide variation by operation |
| Mature foundry practice | 2–4 % of charged metal | Covered furnaces, controlled flux |
| Uncoated bare chips | about 9 % dross | Remelting trials (RWTH Aachen, Journal of Sustainable Metallurgy, 2023) |
| Coated chips, untreated | 20–30 % dross |
The coating result is the most actionable number in this section. The coating weighed less than 2 per cent of the scrap charged, yet its presence at least doubled dross generation across every compaction state tested. Thermal de-coating changed the result by only about 2 percentage points, because it removed only part of the organics. Compaction made it worse, not better: the most densely compacted briquettes generated more dross relative to charged weight (28–30 per cent) than loose chips (19–22 per cent).
4. The thermodynamic limit on purification
This is where secondary metallurgy stops being an energy story and becomes a chemistry story. Thermodynamic simulation of aluminium remelting (Nakajima and co-workers, 2010) reached a conclusion that governs the whole sector:
Of the 45 alloying and impurity elements that commonly occur in aluminium, only about six can be removed to any extent during remelting. Among the major alloying elements, there are only two: magnesium, removed to the slag phase by oxidation, and zinc, removed to the gas phase by evaporation.
Iron, copper, silicon, manganese and nickel cannot be removed. Their oxides have a higher free energy of formation than alumina across the relevant temperature range, so any oxidising step removes aluminium first.
| Element | Removable in remelting? | Route if any | Main source in scrap |
|---|---|---|---|
| Mg | Yes | Oxidation to slag (demagging) | 5xxx scrap, UBC mixing |
| Zn | Yes (partial) | Evaporation to gas phase | 7xxx scrap, coatings |
| Fe | No | Only dilution or downgrading | Castings in ELV shred, fasteners |
| Cu | No | Only dilution or downgrading | 2xxx scrap, radiators, cable |
| Si | No | Only dilution or downgrading | Castings, 4043 weld wire |
| Mn, Ni | No | Only dilution or downgrading | Heat exchangers, engine parts |
Why iron binds hardest
Iron has a maximum equilibrium solid solubility in aluminium of only about 0.05 wt%. Anything above that precipitates during solidification as coarse intermetallics — notably the acicular (needle-like) β-Al5FeSi phase and the α-Al8Fe2Si phase. These are hard and brittle, act as stress concentrators in a ductile matrix, and reduce ductility, fatigue strength and fracture toughness. Because most intermetallics also sit at a different electrochemical potential from the surrounding solid solution, they initiate micro-galvanic cells and degrade localised corrosion resistance.
Methods that could remove iron — segregation, selective liquation, vacuum distillation, three-layer electrolysis, or flux treatment with borax — are described in the literature as difficult, inefficient, and none is reported in industrial use.
5. Demagging: the mechanism is two steps, not one
Chlorine is the industry-preferred demagging agent, but it does not attack dissolved magnesium directly. Chlorine meets an overwhelming excess of aluminium first, so it forms aluminium trichloride at the bubble surface; the AlCl3 then exchanges with magnesium:
2 Al + 3 Cl2 → 2 AlCl3 (g)
2 AlCl3 + 3 Mg → 3 MgCl2 + 2 Al
MgCl2 is less dense than the melt and reports to the dross or flux layer, where it is separated. AlCl3 is an intermediate generated in situ, not a reagent added as such. Adding it as a solid flux is a different process with the same net chemistry.
| Parameter | Chlorination | AlF3 route | Note |
|---|---|---|---|
| Temperature | 760–816 °C | — | US EPA development document range |
| Stoichiometric demand | 3.5 kg Cl2 per kg Mg | 4.3 kg AlF3 per kg Mg | Rules of thumb, not thermodynamic minima |
| Achieved Mg level | 0.3–0.4 % | — | Reported for chlorination |
| Mechanism | AlCl3 intermediate | MgF2 forms and joins the flux | |
| Penalty | Toxic; fumes; grain coarsening | Gaseous or particulate fluorides | Both require off-gas treatment |
The process changes character as it proceeds. Early in the cycle, with magnesium plentiful, the exchange reaction consumes the AlCl3 and magnesium chloride is the principal product. As magnesium is depleted, less AlCl3 is reduced, more of it volatilises, and fuming increases sharply — which is why ventilation and scrubbing are sized for the end of the cycle rather than the average.
A trade-off reported in Chinese remelting practice. Chlorination is described as effective and convenient, and as giving degassing and inclusion removal alongside demagging; the same account records that chlorine is highly toxic, and that melt treated by chlorination shows coarser grains and reduced mechanical properties. Because chlorine handling is restricted in many jurisdictions, alternatives used include chloride/fluoride salt flux mixtures — which generate more dross — and AlF3, where chlorine handling is not permitted. Demagging capability is often the deciding factor in whether a mixed-scrap plant can produce wrought-compliant output rather than being limited to casting alloys, and it carries dedicated capital, storage, injection, safety and scrubbing cost.
6. Dilution, and why the sink is closing
When a melt exceeds the compositional limits of the target alloy, the conventional response is sweetening: blending in primary aluminium to push tramp element concentrations below specification. It works, and it is self-limiting.
The dilution paradox. For impurity-sensitive premium wrought alloys — the 6xxx and 7xxx families — current practice is reported to require dilution with at least 50 per cent primary aluminium to reach the required chemistry (Raabe). Dilution is what makes the chemistry work, and it is also what caps the achievable recycling rate: every tonne of primary metal added to make the specification work is a tonne of demand that recycling did not meet.
Note the direction of the constraint: the cleaner the target alloy, the more primary metal the loop needs.
The sink that absorbed contaminated scrap is shrinking
Historically the industry managed impurity accumulation through a structural sink. Mixed post-consumer scrap, particularly from shredded vehicles, was downcycled into secondary casting alloys. Cast Al–Si families carry high alloying content and much wider compositional tolerances than wrought alloys, so they absorb variable, impure feed. Cast components — engine blocks, cylinder heads, transmission housings, pistons — account for up to 65 per cent of the aluminium mass in an average internal combustion engine vehicle, and about 6 million tonnes of wrought scrap were downcycled into cast products in 2018 alone.
That arrangement is now under pressure. Battery electric vehicles use far less Al–Si cast engine content and far more structural wrought alloy — extrusions and rolled sheet — for battery enclosures, crash structures and body panels. As combustion-vehicle scrap returns in volume, it meets a market where cast alloy demand is contracting. One estimate puts the resulting excess at more than 6 million tonnes of scrap by 2030, attributable to the compositional limits of the available sink alloys.
Why sorting suddenly matters. If neither removal nor dilution can scale, the only remaining lever is keeping alloys apart before they are melted. Conventional separation — magnets, eddy current, heavy media — separates aluminium from other materials but not alloy from alloy; aluminium alloys do not differ enough in colour, and the density and fluorescence response of silicon and magnesium are too close to aluminium for X-ray methods to resolve. Laser-induced breakdown spectroscopy (LIBS) identifies alloy groups from the plasma emission spectrum in milliseconds, fast enough for conveyor sorting; work combining LIBS with machine-learning classification has reported alloy-group identification accuracy above 95 per cent under controlled conditions (NTNU, 2023). The open question is industrial throughput, where contaminated surfaces and variable geometry add noise the laboratory results do not capture.
7. What the two routes produce in common: dross
Dross is the shared by-product, and it is growing with total output. IAI and SINTEF estimate global dross production rose from 2.66 million tonnes in 2019 to 3.27 million tonnes in 2024. Because dross removed from furnaces can contain more than 90 per cent metal in favourable cases, the guidance attached to that figure is that the highest priority is maximising metallic aluminium recovery during processing.
This is the point where the two routes meet the subject of this site: primary smelting emits CO2 and PFCs that no downstream step can recover, while remelting produces a solid residue from which a large fraction of the metal value can still be recovered. Primary production’s loss is gaseous and gone; secondary production’s loss is solid and partly recoverable.
8. What these numbers mean to four different readers
Everything above is physics and energy. The moment metal crosses a border into the EU, the same figures become a customs calculation — and the calculation does not use the numbers that make secondary aluminium look best.
How the certificate obligation is actually computed
The rule is a subtraction, not a tax on the whole footprint:
Certificates = (declared value with mark-up) − (CBAM factor × benchmark)
| Parameter | Value | Note |
|---|---|---|
| Benchmark — primary | 1.423 t CO₂e/t | Direct emissions only |
| Benchmark — secondary | 0.091 t CO₂e/t | Direct emissions only |
| China default — primary | 3.000 t CO₂e/t | Used when no verified actual data |
| Default mark-up | +10 % (2026), +20 % (2027), +30 % (from 2028) | Penalises the default path year by year |
| CBAM factor (2026) | 97.5 % of free allocation retained | Falls in later years |
Worked through for Chinese unwrought primary aluminium on defaults at the Q3 2026 certificate price of €82.32: (3.000 × 1.10) − (0.975 × 1.423) = 3.300 − 1.387 = 1.913 t CO₂e/t, or €157.44 per tonne.
The part that surprises most readers
CBAM prices direct emissions only — and most of primary aluminium’s carbon burden is not direct. The electricity consumed in electrolysis is an indirect emission, and it sits outside the CBAM boundary. What remains inside for primary aluminium is the CO₂ from anode consumption and the PFCs from anode effects. That is why the benchmark is 1.423 and not something close to 16.
The consequence is uncomfortable for anyone who assumed the mechanism rewards clean power: a smelter running on hydroelectricity receives no CBAM credit for that choice, because the electricity is not counted in the first place.
Put the two bases side by side and the distortion is visible:
| Basis | Primary | Secondary | Ratio |
|---|---|---|---|
| Full footprint (Scope 1+2+3) | ≈ 16 t CO₂e/t | ≈ 0.5 t CO₂e/t | ≈ 32× |
| CBAM basis (direct only) | 1.423 t | 0.091 t | ≈ 15.6× |
| Share of the full footprint retained | ≈ 9 % | ≈ 18 % | — |
The retained-share row is this site’s arithmetic on figures drawn from different sources and different system boundaries; it is an order-of-magnitude illustration, not a regulatory result. Under the CBAM basis the secondary route keeps roughly twice as large a share of its full footprint as the primary route does — so the mechanism prices a smaller fraction of the advantage than life-cycle figures suggest.
Four readers, four different questions
| Reader | What they actually need | Where this page’s numbers enter |
|---|---|---|
| Procurement / commercial | Landed cost, and how much of it is carbon | The €157.44 / t above is a real, published figure for one route in one quarter. It moves with the quarterly certificate price — Q3 2026 rose 9.35 % over Q2 — and with the rising mark-up. Quote it as a dated figure, never as a constant. |
| Technical / plant | What data must the plant produce | To escape the default path, a plant must supply verified actual emissions. The energy intensities in §2 are the raw material of that calculation: 13,880 kWh/t world, 13,203 kWh/t China. Measurement capability, not furnace efficiency, is what converts a default liability into an actual-value one. |
| Compliance / customs | Is the basis correct | Two of this page’s distinctions become errors here: treating the “5 %” as an electricity ratio, and assuming a low-carbon grid earns credit. Both produce a declaration that is wrong rather than merely conservative. Unwrought aluminium (CN 7601) is also the precursor for goods under 7603–7616, so a rolling or extrusion plant inherits the metal’s embedded emissions. |
| National / regional | Relative position under one rule | A hydro-powered smelter, a gas-powered Gulf smelter, a south-western Chinese hydro smelter and a coal-powered smelter face the same benchmark but very different starting points — and the rule does not rank them by grid carbon intensity, which is precisely the axis on which they differ most. |
Not established
- The composition of China’s 3.000 t default. It is more than double the 1.423 t benchmark, but this site has not obtained a published breakdown attributing the difference between anode CO₂, anode-effect PFCs and other direct sources. The gap is real; its composition is not established here.
- Actual-value figures for any named smelter. No plant-level verified emissions are reported on this page.
- Future certificate prices and CBAM factor values beyond those already published. Both are scheduled to change; neither is projected here.
The full calculation basis, default values by country, the allocation problem for recycled feed and the quarterly certificate record are set out in Aluminium Carbon Footprint Calculation and CBAM Compliance.
9. The direction of travel
| Indicator | Value | Note |
|---|---|---|
| Global primary output, 2024 | about 72 Mt | IAI preliminary 72.76 Mt; IAI/SINTEF 71.4 Mt |
| Global recycled output, 2024 | about 42–43 Mt | 42.8 Mt, of which 25.3 Mt old scrap and 17.5 Mt new scrap (IAI/SINTEF) |
| Recycled share of supply | 34–37 % | IAI projection 36.85 % for 2024 |
| China primary output, 2024 | 44.005 Mt | Electrolytic aluminium |
| China recycled output, 2024 | 10.50 Mt | About one quarter of primary output; 19.26 % of total aluminium supply |
| China recycled, 2020–2024 | 7.40 → 10.55 Mt | CAGR 9.3 %; share of supply 16.6 % → 19.3 % |
| China policy target | 15 Mt by 2027 | Under the 2025–2027 aluminium industry implementation plan |
Two features of that table deserve attention. First, the recycled share is rising slowly — a few points per decade — while absolute volumes rise faster, because primary output is still growing too. Second, the binding constraint is no longer furnace capacity: it is scrap availability and composition. Chinese assessments identify limited domestic scrap stock, the transition of the economy from investment-led to consumption-led growth, and controls on scrap imports as the factors holding recycled output back, noting that China’s shorter industrialisation has left a smaller accumulated in-use stock.
10. What is established, and what is not
Established
- The twentyfold energy gap is real and is a primary-energy comparison. IAI: 186 GJ/t primary versus 8.3 GJ/t recycled, a 95.5 per cent saving. The widely quoted “5 per cent” derives from this, not from an electricity ratio.
- Electrolysis has a hard thermodynamic floor. 6.34 kWh/kg Al at 977 °C and 100 per cent current efficiency (Grjotheim et al.); world average practice is about 13.0 kWh/kg, roughly twice that.
- IAI world average smelting electricity is 13,880 kWh/t for 2025, ranging from 13,203 (China) to 15,746 (South America). The series covers electrolysis only and excludes casting and carbon plants.
- Only about six of forty-five elements can be removed in remelting; among major alloying elements only Mg (to slag) and Zn (to gas). Fe, Cu, Si, Mn and Ni cannot be removed by any industrial remelting route.
- Demagging proceeds through an AlCl3 intermediate, at about 3.5 kg Cl2 per kg Mg removed, or about 4.3 kg AlF3 per kg Mg for the fluoride route.
- Premium wrought alloys are reported to need at least 50 per cent primary dilution, which caps the recycling rate those alloys can achieve.
- Coating on scrap roughly doubles dross generation relative to uncoated material in controlled remelting trials, even though the coating is under 2 per cent of charge weight.
- Global dross production rose from 2.66 Mt (2019) to 3.27 Mt (2024), and dross can contain more than 90 per cent metal in favourable cases.
Not established by this page
- A single global energy figure for recycled aluminium. IAI states no global estimate exists; the available datasets are Japan, Europe and North America, and they differ by more than 60 per cent (5.7 to 9.2 GJ/t). Any “global” recycled energy number should be read as a regional value.
- A single canonical electricity figure for remelting. Reported values span 600–1,600 kWh/t depending on furnace type and charge quality. A figure quoted without both is not a specification.
- Whether LIBS sorting performs at industrial throughput as it does under controlled conditions. The above-95 per cent accuracy figure is laboratory and pilot work; contaminated surfaces, variable geometry and required throughput are unresolved.
- The 2030 surplus estimate of more than 6 Mt of scrap. It is a projection from one research group and depends on vehicle fleet turnover and cast alloy demand, both of which are still moving.
- Comparability of PFC inventories across years without the GWP edition stated. AR4, AR5 and AR6 values differ materially for the same gas.
- Furnace energy figures here are from surveys and supplier-influenced literature. Specific energy per tonne is strongly dependent on charge quality, operating discipline and accounting boundary; treat them as selection guidance, not as guaranteed performance.
11. What this page is not
- Not a supplier or furnace recommendation. No manufacturer, technology licensor or smelter is recommended, and none is linked.
- Not design or procurement data. Energy intensities, reagent demands and yield figures are reported from named sources to explain the process, not to size equipment or specify a purchase.
- Not a life-cycle assessment. IAI states explicitly that its smelting energy intensity series is not suitable for LCA; LCA work should use the IAI life cycle inventory reports or equivalent.
- Not a substitute for assaying the charge in hand. Every figure on this page is conditional on feed composition. Published optima were obtained on specific scrap assays and do not transfer to an unassayed lot.
Related: Markets & Resources · Hot Dross Recovery · Cold Dross & Salt Cake · Mainstream Dross Recovery Technologies Compared · Editorial Standards
Terms used on this page
原生铝 / 电解铝（Primary aluminium）：从矿石炼出的铝，氧化铝在熔盐中通电解离，铝与氧的化学键被打断。再生铝 / 二次铝（Secondary aluminium）：熔化已是金属的废料得到的铝，不断开任何化学键。
霍尔-埃鲁法（Hall-Héroult process）：由氧化铝到金属的电解工艺，在约 950–970 ℃ 的熔融冰晶石中进行。冰晶石（Cryolite）：氧化铝溶于其中的熔盐。
阳极效应（Anode effect）：电解槽的工况失常，释放全氟化碳气体。全氟化碳（PFC）：增温潜势为二氧化碳数千倍的气体，其公布系数因评估报告版本而异。
除镁（Demagging）：用氯气或熔剂去除过量镁。稀释（Dilution）：加入更干净的金属使成分回到规格，而非除去杂质元素。
碳关税（CBAM）：欧盟对进口产品按其生产排放的碳收费的规则，铝在覆盖范围内，申报只计直接排放、电力在边界之外。嵌入排放（Embedded emissions）：归属于某产品的排放。默认值（Default values）：缺少经核查实际数据时的兜底数值。前体（Precursor）：本身受机制覆盖、其排放被下游继承的投入品。
Fuller explanations of every term used across this site: Aluminium Recycling Glossary | 铝回收术语中英对照与解释
中文精炼版 · 熔炼铝 vs 电解铝的技术差异与二次铝演进
原铝与再生铝是同一种元素，却由两条在最根本处分道扬镳的工艺制成。原铝生产断开化学键——用电解把铝从氧化铝中的氧手里夺出来；再生铝生产不断开任何化学键——它熔化的本来就是金属铝。此后的一切差异——二十倍的能耗差距、杂质难题、以及回收率的上限——都是这一个区别的后果。
1. 支配一切的那个区别。维度对照：化学任务——原铝是把 Al2O3 还原为 Al、断开 Al–O 键；再生铝是熔化既有金属、不断键。原料——前者为拜耳法产出的氧化铝，后者为废铝、铝渣与工艺回炉料。作业温度——前者约 950–970 ℃（熔融冰晶石），后者约 700–800 ℃。主要能量形式——前者为电力（电解），后者为燃气加电力。不可避免的副产物——前者是阳极消耗产生的 CO2 与阳极效应时的 PFC，后者是铝渣（氧化物夹带金属）。产品纯度由什么决定——前者由氧化铝纯度与电解槽操作决定，后者由所装炉料决定。最后一行决定了再生铝能成为什么。
2. 原铝侧：反应式、下限、以及实际达到的水平。通常写作 2Al2O3 + 3C → 4Al + 3CO2（E° = −1.18 V），配平正确，但仅在 100% 电流效率下成立。实际有一部分已产出的铝被阳极气体逆氧化，因此产物含 CO：Al2O3 + (3/2ce)C → 2Al + (3[2ce−1]/2ce)CO2 + (3[1−ce]/ce)CO。热力学下限（Grjotheim 等，977 ℃、100% 电流效率）：加热反应物 0.70 + 还原反应焓 5.64 = 6.34 kWh/kg Al；随电流效率表述为 (1.43/ce) + 4.91 kWh/kg。实际电耗（IAI，2025 年平均，AC）：南美 15,746、欧洲 15,482、海湾 14,891、北美 14,721、大洋洲 14,633、亚洲（不含中国）14,586、非洲 13,866、中国 13,203（最低）、世界 13,880。IAI 定义该值含整流与常规辅助至铝液出槽为止，不含铸造与炭素厂，且明确声明不适用于 LCA。无法闭合的缺口：世界平均约 13.0 kWh/kg 约为理论下限 6.34 的两倍——分解电压约 1.2 V，而工作槽电压近 4.2 V，差额来自阳极过电位、气泡电阻、熔体欧姆降与阴/阳极及母线损失。其中一部分不可消除，它就是维持熔体温度的那部分热量。电解槽无法靠工程手段走下这个地板。非电力的工艺排放：阳极理论碳耗 333.3 kg C/t Al（对应 1,222 kg CO2/t Al），实际净耗约 400 kg C/t Al（约 1,467 kg CO2/t Al，空气氧化与 Boudouard 反应推高了消耗）；预焙槽阳极焙烧另有 0.05–0.15 t CO2/t Al；PFC 在阳极效应时生成，IAI 2024 年合计 5,979 kt CO2e。一个随版本而变的数字：CF4 与 C2F6 的全球增温潜势在各次 IPCC 评估报告中修订——AR4 为 7,390 与 12,200，AR5 为 6,630 与 11,100，AR6 为 7,380 与 12,400，IAI 自 2024 年起采用 AR6 值。未注明 GWP 版本的 PFC 清单，与另一年的清单不可比较——同一物理排放在不同版本下给出不同的 CO2e 数字。
3. 再生铝侧：”5%”这个说法，以及它究竟量的是什么。一处必须说明的单位更正：“回收只需原铝 5% 的能量”是正确的——但它是一次能源比较，不是电力比较。IAI 报告原铝从矿山到铸造车间为 186 GJ/t，再生铝为 8.3 GJ/t，节能 95.5%。这两个数含燃料、原料能以及发电所耗的一次能源。因此，原铝电解电耗（约 13,880 kWh/t）与重熔能耗（8.3 GJ/t）不是可以互相相除的两个量。5% 是真实的，但它不是两个电力数字的比值。再生铝数据集：IAI 明确说明不存在废铝产铝的全球估算，可用的区域数据为日本 5.7 GJ/t（2015，仅变形合金，已被排除）、欧洲 7.2 GJ/t（2018；2015 年为 8.5）、北美 9.2 GJ/t（2015）。另一份欧洲技术评估给出重熔新废铝的拆分约为热能 3.8 GJ/t + 电力 0.45 GJ/t——即大部分能量以燃料而非电力进入。按炉型的电耗与收得率（这才是选型用的数字）：感应炉 600–900 kWh/t、收得率 95–98%、渣 20–50 kg/t（吃干净料）；反射炉 900–1,500、90–95%、40–80；盐熔剂回转炉 1,000–1,500、90–96%、50–100；回转炉 1,100–1,600、88–94%、60–120（吃脏料）。这个权衡的方向与人们预期的相反：能耗最低的感应炉收得率最高，但最不能容忍脏料；能耗最高、收得率最低、产渣最多的回转炉，恰恰是炉料被污染时你该选的那一种。能耗强度与收得率同向，炉料容忍度则与两者反向。买受污染炉料的工厂，买的就是更高的能耗与更低的收得率——而这是在选炉子之前就做出的原料决策。熔炼损失，以及涂层占多大份量：渣生成量常报 15–30 kg/吨熔体；总熔损常引 1–10%，氧化损失平均约 4%；成熟压铸厂做法为装炉量的 2–4%。涂层的结果是本节能据中最具操作性的一条：涂层重量不到装炉量的 2%，但其存在使所有密实状态下的渣生成量至少翻倍——裸屑约 9%，未处理涂层屑 20–30%（亚琛工业大学，2023）。热脱漆只改变约 2 个百分点，因为它只去除了部分有机物。密实反而更糟：压得最实的块相对装炉重量产渣更多（28–30%），松散屑反而更少（19–22%）。
4. 提纯的热力学上限。这里是再生冶金从能源故事转为化学故事的地方。热力学模拟（Nakajima 等，2010）给出的结论支配着整个行业：铝中常见的 45 个合金与杂质元素中，只有约 6 个能被部分去除；主要合金元素中只有两个——镁（氧化进入渣相）与锌（蒸发进入气相）。铁、铜、硅、锰、镍无法去除——在相关温度区间内，它们的氧化物生成自由能高于氧化铝，因此任何氧化步骤都是先氧化铝。铁为什么最难缠：铁在铝中的平衡固溶极限仅约 0.05 wt%，超出部分在凝固时析出为粗大金属间化合物，尤以针状 β-Al5FeSi 与 α-Al8Fe2Si 为甚；这些相硬而脆，在韧性基体中充当应力集中源，降低延性、疲劳强度与断裂韧性；又因其电化学电位与周围固溶体不同，会引发微电偶而劣化局部耐蚀性。文献记载的除铁方法（偏析、选择性液析、真空蒸馏、三层电解、硼砂熔剂处理）均被描述为困难、低效，且无一处于工业应用状态。
5. 除镁：机理是两步，不是一步。氯气是行业首选的除镁剂，但它不直接攻击溶解态镁——氯先遇到的是数量上压倒性的铝，因此在气泡表面生成三氯化铝，再由 AlCl3 与镁交换：2Al + 3Cl2 → 2AlCl3(g)；2AlCl3 + 3Mg → 3MgCl2 + 2Al。MgCl2 密度低于熔体，进入渣层或熔剂层被分离。AlCl3 是原位生成的中间体，不是作为药剂加入的——以固态熔剂形式加入 AlCl3 是净化学相同的另一条工艺。关键参数：氯化作业温度 760–816 ℃；经验用量 3.5 kg Cl2/kg Mg，AlF3 路线为 4.3 kg AlF3/kg Mg；氯化法可将镁降至 0.3–0.4%。过程性质随进度改变：初期镁充足时，交换反应消耗 AlCl3，主产物为 MgCl2；镁被耗减后，被还原的 AlCl3 减少、挥发增多，发烟急剧上升——这就是通风与洗涤按周期末段而非平均值配备的原因。中国再生铝实践记载的一项折让：氯化被描述为有效、便捷，且兼有除气除渣之效；同一记载指出氯气剧毒，且经氯处理的熔体晶粒粗大、力学性能降低。因氯气在许多辖区受限，替代方案包括氯/氟盐混合熔剂（产渣更多）与 AlF3（用于不允许氯气操作的场合）。除镁能力往往是混合废铝厂能否产出符合变形合金要求的产品、还是只能局限在低值铸造合金的决定性因素，且它附带专门的投资、贮存、喷吹、安全与尾气洗涤成本。
6. 稀释，以及为什么那个”汇”正在关闭。当熔体超出目标合金的成分界限时，常规做法是加原铝兑稀（sweetening）把杂质浓度压到规格以下。它有效，但自我设限。稀释悖论：对杂质敏感的高端变形合金（6xxx、7xxx 系），现有实践被报道需要至少 50% 的原铝稀释才能达到要求的化学成分（Raabe）。稀释让化学成立，同时也正是它封住了可达回收率的天花板——为达标而加入的每一吨原铝，都是回收未能满足的一吨需求。请注意这个约束的方向：目标合金越干净，这条回路需要的原铝越多。吸收污染废铝的那个”汇”正在萎缩：历史上行业靠一个结构性”汇”来管理杂质累积——混合消费后废铝（尤其来自破碎车辆）被降级为二次铸造合金，因为铸造 Al–Si 系合金含量高、成分容忍度远宽于变形合金。铸造件（缸体、缸盖、变速箱壳、活塞）占普通内燃机车辆铝重的最高 65%，仅 2018 年就有约 600 万吨变形铝废料被降级为铸造产品。这一安排如今承压：纯电动车使用的 Al–Si 铸造发动机件少得多，转而大量使用结构变形合金（挤压件与轧制板）用于电池壳、碰撞结构与车身板。当燃油车废铝集中回流时，它面对的是一个铸造合金需求正在收缩的市场。一项估算认为，到 2030 年由此产生的过剩将超过 600 万吨，成因正是可用”汇”合金的成分限制。为什么分选突然变重要：既然去除与稀释都无法按规模扩张，剩下的唯一杠杆就是在熔化前把合金分开。常规分选（磁选、涡流、重介质）能把铝从其他物料中分出，却不能把合金从合金中分出——各铝合金颜色差异不足，硅与镁的密度和荧光响应与铝过于接近，X 射线方法分辨不了。激光诱导击穿光谱（LIBS）从等离子体发射光谱识别合金组，耗时在毫秒量级，足以在输送带上进行分选；LIBS 与机器学习分类结合的工作在受控条件下报告合金组识别准确率超过 95%（NTNU，2023）。未决的是工业通量——污染表面、几何形状多变与产能要求会带来实验室结果所没有的噪声。
7. 两条路线共同的产物：铝渣。铝渣是共享的副产物，并随总产量增长。IAI 与 SINTEF 估算全球铝渣产量从 2019 年的 266 万吨增至 2024 年的 327 万吨。由于在有利情况下从炉内扒出的铝渣金属含量可超过 90%，该数据所附的指导意见是：首要任务是在处理过程中最大化金属铝的回收。这也正是两条路线与本网站主题交汇之处：原铝的排放是气态的、下游无法回收；重熔的损失是固态的、金属价值仍有相当部分可以回收。原铝的损失是气体、已散失；再生铝的损失是固体、部分可回收。
8. 同一组数字，四种读者：碳关税下各自读到什么。以上皆为物理与能耗。而金属一旦跨越关境进入欧盟，同一组数字就变成碳关税（CBAM，欧盟碳边境调节机制）下的海关计算——且该计算并不使用那些让再生铝显得最好的数字。证书义务是一道减法，不是对全足迹征税：证书 =（申报值 × 加价）−（碳关税 CBAM 因子 × 基准值）。关键参数：基准值原铝 1.423 t CO₂e/t、再生铝 0.091 t CO₂e/t（均仅计直接排放）；中国原铝默认值 3.000 t CO₂e/t（无经核查实际数据时适用）；默认加价 +10%（2026）、+20%（2027）、+30%（2028 起），逐年惩罚默认值路径；CBAM 因子 2026 年为保留免费配额的 97.5%。以 Q3 2026 证书价 €82.32 代入中国未锻轧原铝走默认值：(3.000 × 1.10) − (0.975 × 1.423) = 3.300 − 1.387 = 1.913 t CO₂e/t，即 €157.44/吨。最反直觉的一条：碳关税 CBAM 只对直接排放计价，而原铝的碳负担大部分不是直接的。电解所耗电力属间接排放，落在碳关税 CBAM 边界之外；留在边界内的只有阳极消耗 CO₂ 与阳极效应 PFC。这正是基准值为 1.423 而非接近 16 的原因。其后果对假定该机制奖励清洁电力的人并不舒服：使用水电的电解厂拿不到任何碳关税减碳 credit，因为电力从一开始就不计入。把两种口径并列，扭曲可见：全足迹（范围 1+2+3）原铝 ≈16 t、再生铝 ≈0.5 t，约 32 倍；碳关税 CBAM 口径（仅直接）1.423 对 0.091，约 15.6 倍；两者各自保留全足迹的比例约为 9% 与 18%。在碳关税 CBAM 口径下，再生铝保留的全足迹份额约为原铝的两倍——因此该机制所定价的优势，比生命周期数字所显示的那一大截要小。（保留比例一行为本站依不同来源、不同系统边界所作的量级推算，非法规结论。）四类读者读到的是四件不同的事：采购/商务——€157.44/t 是一个季度、一条路线的实测发布值，随季度证书价（Q3 环比涨 9.35%）与逐年加价而变动，只能带日期引用，不能当常数；技术/工厂——要脱离默认值路径，工厂必须提供经核查的实际排放，第 2 节的世界 13,880、中国 13,203 kWh/t 正是这套计算的原料，把默认值负债转为实际值负债的是测量能力，不是炉子效率；合规/申报——本页两处区分在这里会变成错误而非保守（把”5%”当作电力比值、以为低碳电网有 credit），且未锻轧铝（CN 7601）是 7603–7616 项下货物的前体，轧制与挤压厂继承金属的嵌入排放；国家/区域——水电、海湾天然气、中国西南水电与煤电四类铝厂面对同一基准，起点迥异，而该规则并不按电网碳强度排序，那恰是它们差异最大的轴。本站未能确立：中国 3.000 t 默认值的构成（为 1.423 t 基准的两倍有余，但阳极 CO₂、阳极效应 PFC 与其他直接来源各占多少，本站未获已发布分项）；任何具名铝厂的实测实际值；尚未发布的未来证书价与碳关税 CBAM 因子（两者均有变更日程，本站不外推）。完整计算基础、各国默认值、再生料的分配问题与季度证书记录见《铝工业碳足迹计算与碳关税 CBAM 合规实务》。
9. 演进方向。2024 年：全球原铝约 72 Mt（IAI 初步 72.76 Mt；IAI/SINTEF 71.4 Mt）；全球再生铝约 42–43 Mt（42.8 Mt，其中旧废 25.3 Mt、新废 17.5 Mt）；再生铝占供应 34–37%（IAI 预测 36.85%）。中国 2024：电解铝 4,400.5 万吨；再生铝 1,050 万吨（约为电解铝的四分之一，占总铝供应 19.26%）；2020–2024 由 740 万吨增至 1,055 万吨，年均复合增长 9.3%，占比由 16.6% 升至 19.3%；政策目标为 2027 年达 1,500 万吨。该表有两点值得注意。其一，再生铝占比上升缓慢（每十年几个百分点），而绝对量上升更快，因为原铝产量仍在增长。其二，约束性瓶颈已不再是炉子产能，而是废铝的可得性与成分。中国的评估把制约再生铝产量的因素认定为：国内废铝存量有限、经济由投资驱动向消费驱动转型、以及废铝进口管控，并指出中国工业化历程较短，社会累积的在用量相对较小。
10. 已确立与未确立。已确立：①二十倍能耗差距真实存在，且是一次能源比较（IAI：186 对 8.3 GJ/t，节能 95.5%），广为引用的”5%”由此而来，不是电力比值；②电解存在硬性热力学下限 6.34 kWh/kg Al（977 ℃、100% 电流效率，Grjotheim 等），世界平均约为其两倍；③IAI 2025 年世界平均电解电耗 13,880 kWh/t，区间 13,203（中国）至 15,746（南美），该序列仅覆盖电解环节，不含铸造与炭素厂；④45 个元素中仅约 6 个可在重熔中去除，主要合金元素中仅 Mg（入渣）与 Zn（入气），Fe、Cu、Si、Mn、Ni 无任何工业重熔路线可去除；⑤除镁经 AlCl3 中间体进行，约 3.5 kg Cl2/kg Mg，AlF3 路线约 4.3 kg/kg Mg；⑥高端变形合金被报道需至少 50% 原铝稀释，这封住了该类合金可达的回收率；⑦受控重熔试验中，废铝涂层使渣生成量大致翻倍，尽管涂层不足装炉量的 2%；⑧全球铝渣由 2.66 Mt（2019）升至 3.27 Mt（2024），有利情况下渣中金属含量可超过 90%。本站未能确立：①再生铝的全球单一能耗值——IAI 明确不存在全球估算，可用数据为日本、欧洲与北美，彼此相差超过 60%（5.7 至 9.2 GJ/t）；任何”全球”再生铝能耗数字都应读作区域值；②重熔的单一基准电耗——报道值跨 600–1,600 kWh/t，取决于炉型与炉料质量，未同时注明两者的数字不构成规格；③LIBS 在工业通量下能否达到受控条件下的表现——超过 95% 的准确率来自实验室与中试，污染表面、几何多变与产能要求尚未解决；④2030 年 600 万吨过剩的估算——出自单一研究组的预测，取决于车队更新节奏与铸造合金需求，两者仍在变动；⑤未注明 GWP 版本的 PFC 清单跨年度不可比较；⑥本页炉型能耗来自调查与受供应商影响的文献，比能耗强烈依赖炉料质量、操作纪律与核算边界，应作选型参考而非性能保证。
11. 本页不是什么。不是供应商或炉型推荐——不推荐任何制造商、技术许可方或冶炼厂，亦不链接其中任何一家。不是设计或采购数据——所引能耗强度、药剂用量与收得率报道自具名来源，用于说明工艺，不用于设备选型或采购规格。不是生命周期评价——IAI 明确声明其电解能耗强度序列不适用于 LCA，LCA 应使用 IAI 生命周期清单报告或同等数据。不能替代对到手炉料的化验——本页每个数字都以进料成分为条件；已发表的最优值是在特定废铝品位上取得的，不能移植到未经化验的批次。