by Iver P. Cooper

This is an addendum to my gazette 8 article, “Aluminum: Will of the Wisp?” In essence, it consists of stuff I thought interesting but which for one reason or another I couldn’t fit into the article. Rather than deprive my fellow barflies of the information, I am posting it here.

It comes with a caveat: It isn’t a standalone document. Read the gazette article first, then look here. Also, I occasionally descend to “notes” style, rather than writing complete sentences.


The Greenland Eskimos used the cryolite as sinkers or weights for their fishing lines and nets because the mineral is very soft, easy to shape and to drill into.” (Whittaker, 469) Kragh adds that because cryolite is virtually transparent underwater, such sinkers would not be noticed by fish.

The scientific term cryolite is from Greek (ice+stone). According to Nuka Moller of the Greenlandic Language Secretariat, “the Greenlandic term for the mineral cryolite is ‘orsugiak’ which means ‘glistening’ or ‘shiny’ from the stem ‘orsoq: fat, blubber’. The term might have been coined due to the shiny surface of the mineral, which resembles the glistening surface of fat or blubber. The Greenlandic term first appeared in Samuel Kleinscmidt’s Greenlandic dictionary in 1871.”

Cryolite samples were brought to Europe in the mid-eighteenth century. Its chemical nature was determined in 1800, and the first European to study the cryolite deposit at Ivigtut was Giesecke in 1806. (Whittaker, 469).

The following description of cryolite will be available in Grantville: “Cryolite occurs in colorless or snow-white cleavable masses, often tinted brown or red with iron oxide, and occasionally passing into a black variety. It is usually translucent…. The mineral cleaves in three rectangular directions, and the crystals occasionally found in the crevices have a cubic habit, but it has been proved, after much discussion, that they belong to the anorthic system. The hardness is 2.5, and the specific gravity 3…. It colors a flame yellow, through the presence of sodium, and when heated with sulphuric acid it evolves hydrofluoric acid.” (EB1911)

The cryolite deposit is quite large; in the late nineteenth century, it was reported to be “a solid mass of about 600 feet in length and 200 feet in width….” (Kingzett, 151). Based on more modern studies, I would say that its surface area (under the land) is more like 300 by 300 meters, and that it reaches down to 100 meters in placs. See “Ivigtut in 3D”

Even after over a century of mining, there is still some cryolite in the area; there was collection of two tons in the summer of 1998 for the benefit of mineral collectors. (Id.; GEUS)

The initial reason for mining cryolite was for use in making soda.

Soda was valuable enough that during the Civil War, the American company Pennsalt (in Pennsylvania) entered into a long-term contract for two-thirds of the Ivigtut output. (Grossman 23 — He says this was 6,000 tons annually, with the first shipment arriving in 1865, but see Johnson’s, cited in the article.) This is partially corroborated by Kingzett (153), who says that in 1867, a Pennsylvania plant was started for working 6,000 tons annually. The dates are a little off, but the amounts are the same. As of 1877, 2,000 tons of cryolite was processed annually by German soda plants (Id.).

Synthetic Cryolite

It would be reasonable to surmise that the process might involve hydrofluoric acid, given that cryolite is a fluoride. If so, the EA “hydrofluoric acid” article would confirm that yes, hydrofluoric acid (HF) is used in the manufacture of synthetic cryolite. Even if you didn’t think of HF immediately, if you at least looked up “fluorine,” you would be pointed to HF as a starting material.

You could also expect that some sodium salt was involved. Unfortunately, EA doesn’t reveal which one to use. While Ebling mentions sodium hydroxide, the first one I found in my literature search was sodium carbonate.

Cryolite can be made from hydrofluoric acid, sodium carbonate, and aluminum. That may seem self-defeating, since the point of making the cryolite is to obtain aluminum. However, the cryolite doesn’t actually react during the electrolytic reduction of alumina, it is just a flux. So, it could be worth it to sacrifice some of our precious store of aluminum to make the cryolite.

Or, one can react hydrofluoric acid with the sodium aluminate from the Bayer process. The reaction is 6HF + 3NaAlO2 -> Na3AlF6 + 3H20 + Al2O3. (Kirk-Othmer, 191).

Want another method? You can fuse AlF3, NaF, and CaF2. (Calvert) {{That one doesn’t use hydrofluoric acid directly. }} But aluminum fluoride itself is made by reacting hydrofluoric acid gas with activated alumina, or fluorosilicic acid with aluminum hydroxide. (Kirk-Othmer) {{Sodium fluoride is made by reacting the acid with sodium carbonate.}} HF is made by reacting CaF2 (fluorspar) with concentrated sulfuric acid. Finally calcium fluoride is extracted from fluorite (Fluorspar), which is available in Germany. (EB-F).

Required Initial Charge of Cryolite

By other means, I calculate a smaller required initial charge than that given in the article. I assume that pot size, and more particularly the required initial charge of cryolite, are both proportional to the amperage running though the pot. That is actually an implicit assumption of the Mineralszone rule, since they estimate based on installed capacity, which is proportional to amperage because of Faraday’s law.

In Prasad’s 216,000 ampere experimental cell, which was intended to represent a modern commercial one, the ratio was 78 kilograms per thousand amperes.

Hall’s first pots each held 300-400 pounds (136-181 kg) of cryolite for dissolving the alumina. These pots operated at 1750 amperes. (Beck) Thus, Hall’s ratio was 77 to 103 kilograms per thousand amperes. That’s not much of a change considering the amperage was increased more than one hundred fold.

A cell operated at 180,000 amperes has a theoretical production rate of 1450 kg/day, which is 529,250 kg/year (the installed capacity if only one cell is in operation). If the same cell is initially charged with 78 kg/1000 amperes cryolite, that is 14,040 kg. That is a ratio of 26.5 kg per metric ton capacity. If we assume that capacity is after allowing for current efficiency, then if the efficiency is 90%, the ratio is 29 kg to the tonne. The figure in Mineralszone was 80 kg to the tonne, and in Reynolds, 60.

Cryolite Consumption

Some estimate that the cryolite consumption is the equivalent of 2.5-4 kilograms fluoride per tonne of produced aluminum. (Kirk-Othmer, 2:192; Brubaker, 96), implying use of not more than 27.5-44 kilograms cryolite per aluminum tonne. Elsewhere Brubaker says that consumption of cryolite is 30-100 kilograms per tonne (97). Mineralszone says 50-90 kg/tonne and, in the more advanced plants, upper limit is now 60 kg/tonne.

McKetta, Encyclopedia of Chemical Processing and Design, 23: 272 says,

“fluoride consumption… averages 60 lb of aluminum fluoride and 50 lb of synthetic cryolite per ton of aluminum ingot produced. This is equivalent to about 70 lb of HF per ton of aluminum. In the United States, the amount of HF required is probably 20% less, or 56 lb/ton, owing to recycling of fluoride values from spent potlinings and flue gases, as well as the manufacture of aluminum fluoride and synthetic cryolite from fluorosilicic acid.” “

USGS “Fluorspar” states consumption is 20 kilograms fluoride per tonne aluminum. . If this were all cryolite, that would be (210/19)*20, or about 220 kilograms cryolite per tonne of aluminum. If it were all AlF3, it would be (84/19)*20, or about 88 kilograms AlF3 per tonne aluminum.

Cost of electrolyte in 1963 is set forth as $16-25 per tonne aluminum (Brubaker 154). If the price is assumed to be $200 per tonne electrolyte, that implies a usage rate of 80-125 kilograms/tonne. The prices given by Brubaker 95 are 130-360/tonne, for a broader range of 44-192 kilograms/tonne.

Reynolds says you need 25 kg/tonne cryolite for a Soderberg plant and 10 kg/tonne (20 pounds/ton) for a prebaked anode plant.

Aluminum fluoride consumption

McKetta said 60 pounds/ton; 20% less with proper fluoride recycling and cryolite regeneration technology. Reynolds says 25 kg/tonne (50 pounds/ton) in Chinese smelters; 20 in western plants using prebaked anodes.

A site discussing the Korba Aluminum Complex in the Soviet Unision said that it consumed 43.17 kg AlF3/tonne Al, but the target was to reduce it to 25 kg/tonne. (The data on cryolite consumption is flawed, it states a ridiculous 2.74 kg/tonne with 15 kg/tonne as the target; it probably meant to say 27.4.) See “Modernization of Smelter Plant,”

Netto’s Process of Manufacturing Aluminium from Cryolite. [Manufacturer and builder / Volume 22, Issue 3, March 1890]

Scanned page 38:

Rather imperfect OCR translation:

Cryolite and Climate

My principal source for Greenlandic ice conditions is the Environmental Oil Spill Sensitivity Atlas for the West Greenland Coastal Zone

8.1.2. “The Polar Water inflow is strongest during spring and early summer (May – July), and since the East Greenland Current carries large amounts of Polar Ice with it, the distribution of Polar Ice along the coasts of West Greenland will attain its maximum during the same period. The inflow of Atlantic water masses is strongest during autumn and winter explaining why the waters between 62º N and 67º N normally are ice free during wintertime.”

8.1.3. “Sea ice is normally present in the Davis Strait, particularly in the western parts, from November to mid-summer, in Disko Bay from January to May and in the Cape Farewell area from mid winter to late July. The waters off Southwest Greenland are normally free of sea ice, but can be covered with sea ice during late winter for short periods or during spring and early summer months where multi-year ice originating from the Arctic Ocean drifts into the area.

“Icebergs and growlers originating from glaciers occur in the entire region, but the density of icebergs is normally low in the eastern Davis Strait area. It increases towards Disko Island to the north and towards the Cape Farewell area to the south. However, the drift and distribution of glacial ice at sea have never been investigated systematically, and therefore are only known roughly known.

“The meteorological conditions in the area are influenced by the North American continent and the North Atlantic Ocean, in addition the Greenland Inland Ice and the steep coasts of Greenland have a significant impact on the local climate. Many Atlantic depressions develop and pass near the southern tip of Greenland and cause frequently very strong winds off Southwest Greenland. Small scale phenomena such as fog or polar lows are also common features near the West Greenland shores. The probability of severe winds increases close to the Greenland coast and towards the Atlantic Ocean.”

8.4.2. “The ice conditions between 60° N and 72° N are primarily determined by the relatively warm north or northwest flowing West Greenland Current (WGC) and the cold south flowing Baffin

Current (BIC). The WGC delays the time of ice formation in eastern Davis Strait and results in an earlier break up than in the western parts of the Davis Strait (Figure 8.12). The BIC conveys large amounts of sea ice from Baffin Bay to the Davis Strait and the Labrador Sea for most of the year, especially during the winter and early spring months. During this period sea ice normally covers most of the Davis Strait north of 65° N, except areas close to the Greenland coast, where a flaw lead

(open water or thin ice) of varying width often appears between the shore or the fast ice and the drift ice offshore as far north as latitude 67° N. South of 65-67° N, sea ice free areas dominate throughout the year. The sea ice edge (the boundary between drift ice and sea ice free water) is normally oriented to the southwest towards Hudson Strait or the Labrador Coast. In the beginning of the melt season a wide lead or polynya-like feature often forms west of Disko Island in front of Disko Bay. The eastern part of Davis Strait, south of Disko Island, is free of sea ice during this period (Figure 8.15), whereas drifting ice is dominating to the west and north In Greenland this ice regime is recognized as the ‘The West Ice’ (Figure 8.13a & b).

“The predominant sea ice type in the Davis Strait and the southern Baffin Bay is first-year ice. Small amounts of multi-year ice of Arctic Ocean origin drift to the western parts of the area from Lancaster Sound or Nares Strait, however, the multi-year ice from these waters does not usually reach the West Greenland shores. At the end of the freeze up season first year ice in the thin and medium categories dominate in eastern parts (up to about 100 km from the Greenland coast). The western and central parts of Davis Strait and southern Baffin Bay are dominated by medium and thick first year ice categories mixed locally with small amounts (1-3 tenths) of multi-year ice.

“The dominant size of ice floes range from large floes of about 1 kilometer wide to vast floes larger than 10 km. Near the ice edge in Davis Strait, the size of the common floes are reduced to less than 100 meters as a result of melting and break up by waves. These floes are often very consolidated.”

Figs 8.12 A-D show the probability of sea ice in West Greenland waters on March 1, June 4, September 3, and December 3, based on 1960-96 data. For the latitude of Ivigtut (61 N), we see the following

(A) On March 1, the probability is the 15-50% range just offshore. It drops to the 5-15% range if you head west as far as 50W, but increases back to the 15-50% range at about 56W. If you head east along the coast, then as soon as you round the tip into eastern waters, the probability jumps to the 50-85% range.

(B) On June 4, the probability is in the 50-85% range just offshore. It drops to the 15-50% range at 50W.

(C) On September 3, the Davis Strait is entirely free of ice in the region east of about 57W, and south of about 75N.

(D) On December 3, except for a few pockets (outside Arsuk fjord) of 5-15% hugging the coast in the 62-60N range, the coast has a 0-5% chance of sea ice at least up to 67 N. At the latitude of Ivigtut, you must travel west to about 58W for even a 5-15% chance.

Caption to Fig. 8.14 (satellite photos): “Close to the south west

Greenland coast, sea ice often forms locally but is very dependant on the air temperature and the salinity stratification in the sea. Due the changeable winds in the area this kind of ice cover normally only exists for short periods (less than one week).”

8.5. This section discusses sea ice and fast ice (ice attached to land) in the fjords.

For Arsuk fjord, it says that the outer area has an exposed shore year-round, while the inner areas normally have sea ice in January through March. Arsuk itself, at the mouth of the fjord, has an exposed shore year-round. There is no offshore (10-20 nautical miles) data for latitude 61N, but for 62N, the probability of sea ice is less than 15% in August to December, and 15-85% (high variability) in January to June. The nearshore (under 10 nautical miles), outside fjords has an exposed shore at the same latitude, year round.


“8.6.6 Davis Strait between 65°N and 68°N: Even in severe winters, navigation normally is possible in the eastern part of Davis Strait as far north as Sisimiut due to the existence of the relatively warm north going current. Furthermore, the sea ice drift has a significant offshore component (the West Ice), and for this reason sea ice only covers the Davis Strait in the last half of very cold winters. If sea ice is present close to the

Greenland coast, it is very sensitive to easterly winds, and break-up occurs quickly. When sea ice covers the entire strait, a narrow lead (may be covered with thin ice) normally forms close to the Greenland coast, just off the fast ice edge. The normal ice type in the area is young ice or thin first year ice in varying floe sizes. Wide belts of small floes normally occur near the ice edge. Multi-year ice (‘Storis’) from the Greenland east coast almost never drifts north of 65° 30’ N. This has only

been observed a few times in 20th century and not in the period 1958-99.

“8.6.8 Northeastern Labrador Sea between Nunarsuit [61 N?] and 63° N: This area is normally free of sea ice from late summer until mid-winter. In the late winter the ‘West Ice’ occasionally affects the area. In very cold periods sea ice forms locally within 50-60 kilometers of the Greenland coast. Varying amounts of ‘Storis’ occur almost every year from late winter/early

spring until mid-summer. Due to the offshore component of the West Greenland Current, the multi-year ice is sometimes present only far from the shoreline.”

Another source is Physical Environment of Eastern Davis Strait and Northeastern Labrador Sea,

Under ice conditions, maps of probability of occurrence of sea ice show

January 1: 60-61N 11-50%, conc 1-4-9; 61-62, 1-10%

February 1: 60-63N, 11-50%, 1-4 -9

March 1: 11-50%

April 1: 11-50%

May 1: near boundary between 11-50% and 51-90%

June 1: 51-90%

July 2: 1-10%

August 6: 11-50% (seems anomalous)

September 3: 1-10%

October 1: 1-10%

November 5: 0%

December 3: 1-10%

Major source of bergs in northwest, tend to move westward then down Canadian coast

No significant calving of icebergs 61N to 64/30N.

Seasonal max is late April to late July

Many small bergs calved in fjords S of 61N (these drift N)

Canadian maps of the Davis Strait ice don’t quite reach Greenland, but suggest that in the vicinity of Ivigtut, freeze up is after Dec. 4, and break up is before June 19. See

Likewise, extrapolating from maps of sea ice probability for the Canadian side of the Davis strait, I would estimate that for Ivigtut, we are talking

0% in January

1-15% in April

0% in July

0% in October


Ivigtut 24-Hour Average Temperature (monthly mean of 24 hour temp)

Jan 20.7F, Feb 21.4F, Mar 25.3F, Apr32.0F, May 40.6F, Jun 46.6, Jul 49.5, Aug 47.5, Sep 41.7, Oct. 34.3, Nov 27.9, Dec 23, Year 34.2.

Cp. Mannington

Jan 30.6, Feb 32, Mar 41.2, Apr 50.9, May 60.3, Jun 68.2, Jul 72, Aug 70.5, Sep 64.2, Oct 42.9, Nov 41.9, Dec 32.9 Year 51.4.

Post-RoF cryolite miners must contend with a harsh climate. Brian Fagan, in Little Ice Age, refers to the seventeenth century as “bitterly cold.” A volcanic cold spike is due to occur in 1641.

The cryolite soda process replaced the Leblanc process, and was in turn superseded by the Solvay process. The Leblanc process produced hydrochloric acid gas and calcium sulphide, neither of which were marketable in the early nineteenth century, and which produced serious pollution problems. This may be less of a problem in early post-RoF Europe; Grantville can use hydrochloric acid (either directly or as a source of chlorine), and can probably recover the sulphur, too. Alternatively, Grantville can use the Solvay process, whose only waste product is calcium chloride (a good drying and ice-melting agent).

The sale of cryolite for soda production continued until about 1894 (Whittaker, 469).

Cryolite mining ceased in 1969, but in the next two decades, Greeland exported 500,000 tons of previously mined ore which had been used locally for “coastal protection” and road construction. (Whittaker, 469).

In 1904, Oresund sold 2,110 tonnes; in 1939, 30,050. (Travis)


Cryolite Outside Greenland.

Are there alternatives to Greenland? Yes, but they are ill-defined, awkwardly located, and much less rich in cryolite. EB1911 mentions Miyask, in the Ilmen Mountains, {{Pikes Peak, in Colorado, and Yellowstone Park in Wyoming. Kirk-Othmer (11:277) also lists Sallent, in the Pyrenees.}}.

The most complete list of locations at which cryolite speciments have been found is the cryolite entry at, see

The listed locations are in Brazil (Pitinga and Lages), Quebec (Montreal, Mont Saint-Hilaire, and Varennes), Bohemia (several locations in the Erzgebirge, notably, at Bozi Dar, Cinovec, Vykmanov, and Krupka), Namibia (Aris, Kalkfeld), Norway (Gjerdingen), Russia (in Transbaikalia, the Kola peninsula, and at Miyask in the Ilmen mountains), in the Ukraine, and in the United States (Colorado, Maine, Nevada, New Hampshire, New Mxico, Texas, Utah, Virginia). The American locations include two well known gem sites, the Ruggles mine (NH) and the Morefield mine (VA), which might be targeted for other reasons.

Note that according to, the Pikes Peak and Miyask “sightings” are unproven (not cited in the mineralogical literature). The alleged finds in Sallent and Yellowstone Park aren’t even listed in

Fluoride Additives

Aluminum fluoride is made by reacting aluminum hydroxide with hydrofluoric acid, or aluminum metal with fluorine.

Lithium fluoride, by reacting lithium hydroxide or lithium carbonate with HF. Sodium fluoride, by reacting sodium hydroxide or sodium carbonate with HF. Potassium fluoride, from potassium carbonate and HF.

Calcium fluoride, however, can be mined (mineral fluorite, also called fluorspar). It is mostly used in the production of HF.

The 2004 USGS Minerals Yearbook says that the Hall-Heroult process uses cryolite, aluminum fluoride and fluorspar. “Fluorine losses are made up entirely by the addition of AlF3, the majority of which will react with excess sodium from the alumina to form Na3AlF6. This type of smelter will consume about 20 kilograms (kg) of AlF3 for each metric ton of aluminum produced. Plants that use the older Soderburg technology with minimal recovery of fluorine emissions will have significant losses of fluorine and sodium, which will be replaced by adding a combined 40 to 50 kg of AlF3 and Na3AlF6 per ton of aluminum produced.”

Predecessors to the Hall-Heroult Process

1. Besides the methods mentioned in the article, there is brief reference in EB11-Al to 1887 Grabau process, in which aluminum fluoride was reduced with sodium. The encyclopedia notes that this method produced aluminum of especially high purity (99.5-99.8%).

Energy requirements

bauxite mining: 0.32 kw-h/kg aluminum

refining to alumina: 7.27 kw-h/kg aluminum (7.87 tacit)

carbon anode production: 0.66 kw-h/kg aluminum

(Choate 16)

If it weren’t for the fact that the carbon anode is consumed, the theoretical minimum energy requirement for the reduction of alumina to aluminum at 960 deg. C would be 9.03 kWh/kg aluminum (9.30 if gas emission is included). The consumption (burning) of the carbon provides some of the energy for the reduction, so the net minimum requirement is 4.99 kWh/kg. (Choate 28).


1. Gibbsite [Al(OH)3] is equivalent to Al2O3 (alumina) +3H2O (water), so it is a trihydrate. Boehmite and diaspore [AlO(OH)] are equivalent to Al2O3 + 1H2O, so they are monohydrates.

2. EB11-Al gives the following bauxite composition data:

*Antrim (Irish) bauxites: 33-60% alumina, 2-30% ferric oxide, 7-24% silica, balance titanic acid and water. (Encyclopedia Americana says that the finest Antrim bauxite is “almost free from iron.”)

* French bauxites: 58-70% alumina, 3-15% “foreign matter,” 27% silica, iron oxide and water.

* American bauxites: 38-67% alumina, 1-23% ferric oxide, 1-32% silica. (EB11-Al, 768) (EA says that in America, ore is available with “as little as 1 percent iron oxide and 3 percent silica”).

3. About 5.1 kilograms bauxite are required to produce 1 kilogram aluminum. (Choate, 18)

34. An article on Gibbsite (the trihydrate of alumina) says that it occurs “on the Vogelsberg, Hesse” and “on the Katzenbuckel, Baden-Wurttemberg”, in Germany.

4. According to, within Germany, Gibbsite (the trihydrate form of alumina) has been found at


Haus Württemberg Mine (Neues Jahr Mine), Freudenstadt, Black Forest

Hornberg, Black Forest

West cut, Rollenberg railway tunnel, Bruchsal, Kraichgau,

Michelsberg Quarry, Katzenbuckel Mt., Eberbach, Odenwald,

North Rhine-Westphalia:

Copper Mines, Marsberg, Sauerland

Eisenzecher Zug Mine, Eiserfeld, Siegerland


Schellkopf Mt., Brenk, Niederzissen, Eifel Mts

Bad Ems District, Lahn valley


Bärenstein Quarry, Niederschlag, Oberwiesenthal, Erzgebirge

Deutschlandschacht Mine, Oelsnitz, Zwickau

Boehmite (the monohydrate) is listed only at

Bärenstein Quarry, Niederschlag, Oberwiesenthal, Erzgebirge, Saxony, Germany}}


5. While IAI maintains that in southern Europe, the deposits are pockets, Bateman says that in southern France, the deposits can be blankets. He reports that the French bauxite is 57-60% alumina, 20% ferric oxide, and 3-5% silica.

6. The deposits in Arkansas (as well as Alabama and Georgia) can be blankets, pockets, interbedded deposits (multiple layers separated by other substances), or detritus deposits (materials formed somewhere else, eroded away, and carried to their present location by wind or water).

Arkansan beds average 11.5 feet thick, with a maximum of 35 feet. They may be covered by up to 80 feet of clay and sand. The bauxite is 56-59% alumina. At least some deposits are less than 5% silica and 2-6% ferric oxide. Bateman has a map (216) suggesting that the beds run, more or less continuously, starting one mile northwest, and ending three miles southeast, of the town of Bauxite. Another illustration (556) shows interbedding of bauxite and other clays.

7. Australia, as of 1963, had larger bauxite reserves (about two billion metric tons) than any other country. The other first tier potential bauxite producers are Guinea (1.1 billion tons reserves) and Jamaica (600 million tons). In the second tier are Hungary (300 million tons), Yugoslavia (290), Ghana (254), Surinam (250), Guyana (150) and China (150). In the third tier, the richest European sources are Greece (84) and France (70). (Brubaker, 147-8). The stated French reserves were, of course, after depletion by over seventy years of mining.


1. If there were no losses, you would need 1.89 kilograms of anhydrous alumina to produce 1 kilogram of aluminum. (Choate, 18).

2. Baton Rouge, Louisiana takes its name, which means “red stick,” from the “red mud” waste product.


Hall’s Early Smelters

According to Beck, Hall’s first commercial aluminum smelter (1888) had just two electrolytic cells in series, operating at a current of 1750 amps, and with a total voltage of sixteen volts (eight apiece) across the two cells. These pots produced a total of fifty pounds aluminum per day.

The implied power is fourteen kilowatts per cellAssuming continuous operation, Hall’s 1888 plant, was using 13.44 kilowatt hours per pound. However, he was drawing his current from steam-driven dynamos which were actually generating 50 kilowatts, implying a total energy use of 26.88 kilowatt hours per pound of aluminum. That is extremely high by modern standards. However, it is consistent with Choate’s chart showing the “reduction in energy for smelting” over the period 1900-2000; the average energy use was about fifty kilowatt hours per kilogram)(22.7/pound) in 1900, falling to 15.4 per kilogram (seven per pound) by 2000.}}

Hall built a new, larger plant which produced 1000 pounds a day in 1893 and 2,000 pounds a day in 1894. (Beck). By 1914, the twenty-five aluminum smelters in the world had installed horsepower in the range of 3,000 to 55,000, the average being 20,000. (Wallace, 40-41) That corresponds to 15,000 kilowatts apiece, and hence to an average aluminum production capacity of 1,500 pounds an hour, {{or 36,000 pounds a day}} (assuming that energy was the bottleneck).


Eric may have modeled the power plant on the only power plant in Marion County, the 1944 Monongahela Power Company plant in Rivesville. Its “nameplate” power production capacity is 110 megawatts, but the actual net production rate in 1999 was 158,000 megawatts Rivesville produced 332,510 megawatts, while consuming 173,982 megawatts, in 1999 (Alfred).

Hall moved his smelting operation to Niagara Falls (initiated 1895; 50,000 hp in 1914), and his company built another smelter at Shawingan Falls in 1901 (40,000 hp, 1914). Likewise, in Europe, aluminum was produced in Neuhausen, Switzerland near the Falls of the Rhine (initiated 1889; 4,800 hp, 1914), and in {{Britain}}, alongside the Falls of Foyers (initiated 1896; 6,000 hp, 1914). (EB-IEP 389; EB11-Al 770; Wallace, 40-41). Obviously, there was general agreement that cheap hydroelectric power was critical to large-scale aluminum production.

{{A practical question relates to how the smelter receives its electricity. What it needs is direct current (DC), but it is more efficient to transmit alternating current (AC) and then convert it. Conversion efficiency is highest and capital costs lowest when the resulting voltage is 600-900 volts (Kirk-Othmer, 2:197); that implies that the pot lines, consisting of 4.5-5 volt cells, should each be 120-200 pots.

Nowadays, AC is converted to DC by silicon rectifiers. I am not sure how well equipped Grantville is with such rectifiers, whether it can make them, or whether there are reasonable substitutes.}}


{{When a pot is started up, the anodes are lowered so they are in contact with the carbon cathode. Cryolite and other fluoride salts are dumped into the pot, and electrically heated until they melt, producing the electrolyte bath. Then the anodes are raised 4-5 centimeters above the cathode, and alumina is added to start the reduction reaction. (Wallace, 7-8) Pure cryolite melts at 1012 oC., but the presence of the other salts reduces its melting point. As a result of the reaction, the aluminum oxide is depleted (and must be periodically replenished), the aluminum forms and pools at the bottom of the pot (from which it is siphoned off), and the carbon anodes are consumed and replaced.


The 1911 EB correctly states that the aluminum is denser than the electrolyte, and hence sinks through it. However, the numbers are wrong. It says that the reduced metal has a specific gravity of 2.54 (correct value is 2.3 at 960 oC.), and that cryolite-alumina solution is 2.35 (correct value is 2.1).


The cathode current density is the current (amperes) divided by the surface area of the cathode (the floor of the pot). If current is increased, without changing the pot dimensions, the current density will also increase. This will tend to increase corrosion of the cathode. (Welch) The electrical resistance of the electrolyte is also proportional to the current density.

The 1911 EB suggests a current density about 700 amperes per square foot of cathode surface. In Hall’s 1888 operation, the cathode area per pot was about 2.67 square feet, implying a current density of 655 amperes/square foot — similar to the number suggested by EB11.


In the cell described by 1911 EB, each carbon rod, acting as an anode, supposedly delivered 6-7 amperes current per square inch of cross-sectional area.

In 1888, Hall used six to ten anodes, each three inches in diameter. That works out to 17-28 amperes per square inch of cross-section.

The anode current density in modern cells was 0.65-1.3 amperes/square centimeter for prebaked anodes, and 0.65-0.9 for Soderberg anodes. Multiply by 6.45 for the “per square inch” equivalent. It clearly is close to the 1911 recommendation.

If the anode current density is too high, fluorocarbons are formed at the anode surface.


Safety Issues. Potroom workers are or can be exposed to radiation (UV, visible and infrared), heat (this can be by convection instead of radiation), strong magnetic fields, high-amperage currents, molten metal or cryolite splash, particulates (aluminum, silica, etc.), fluorides, polycyclic aromatic hydrocarbons (carcinogens), sulfur dioxide and carbon monoxide. Aluminum dust clouds can ignite or explode, although not easily. Fluorides are probably the greatest environmental concern. (Burgess, Meridian)

While proximity to Grantville’s power plant is an advantage, it is likely that environmental concerns will force the smelter to be located several miles away, in which case we must also provide transmission lines (with copper or aluminum wire) to service the smelter.


Deville Process

Deville produced sodium by heating a mixture of sodium carbonate, coal and chalk, and collecting and condensing the resulting sodium vapor. The double chloride was obtained by heating a mixture of alumina, sodium chloride, and coal tar in the presence of dry chlorine. And the alumina was produced by heating bauxite and sodium carbonate, extracting the resulting sodium aluminate, and reacting the latter with carbon dioxide.

All of this is described in 1911 EB.

It should be appreciated that with the Deville method, it is critical to have cheap sodium. New American Cyclopedia says Deville managed to reduce the cost of preparing metallic sodium from $100/pound to $0.90/pound, and cites an 1857 article which gives a cost of $0.26/pound sodium (and of $0.08/pound for the double chloride).


Kragh reports that Amfreville produced 960 kilograms aluminum in 1859. Nanterre’s 1859 aluminum output was 600 kilograms.

EB1911 “Aluminum” briefly discusses alternative ores. Corundum [Al2O3] can be found in southern India, and has a higher aluminum content than all the other aluminum minerals. Unfortunately, it is so hard (9 on the Mohs’ Scale; only diamond is higher) that it is difficult to process. Bauxite, in contrast, has a hardness of 1-3 (gibbsite) or 6.5-7 (diaspore).

While kaolinite (aluminum silicate) contains just 2.4-4% aluminum, it is widely distributed. EB1911 drily comments, “Kaolin thus seems to be the best ore, and it would undoubtedly be used were it not for the fatal objection that no satisfactory process has yet been discovered for preparing pure alumina from any mineral silicate.” Indeed, “an alumina contaminated with silica is not suited for reduction.”


{{For over a century, attempts have been made to supersede the Hall-Heroult process. None has gotten past the pilot stage. The tested alternatives include electrochemical decomposition of aluminum chloride or aluminum sulfide, and high temperature reaction of aluminum carbide with aluminum oxide to yield carbon monoxide and aluminum. (Welch; Kirk-Othmer 2:200).}}


Manganese and Magnesium

There are known sources of manganese and magnesium in existing mines, or in areas friendly to the USE. (Runkle, “Mente e Malleo,” Grantville Gazette, Volume 2)


Manganese dioxide (pyrolusite) can be reduced with carbon, and pure manganese was first produced in 1774. The 1911 EB notes that at the end of the nineteenth century pyrolusite was “extensively minded at Ilmenau and several other places in Thuringia . . .” and it describes a method used in 1893 to prepare 97% manganese from pyrolusite.

John Leggett has written a series of “USE Steel reports,” which envision that, in December 1632, manganite (another manganese ore) will be imported for conversion into “spiegeleisen,” a pig iron which, because of its manganese content (about 12%) is useful in the removal of oxygen and sulfur from steel (especially in the Bessemer process). The 1911 EB also mentions use of ferromanganese, which is 80% manganese.

While even ferromanganese has too much iron to be used as is in the preparation of aluminum alloys, it can certainly be refined further, perhaps electrolytically, for that purpose. Or prepared from pyrolusite, by the 1893 method.

Manganese of purity suitable for use by the aluminum industry is likely to be available two or three years after the first large-scale use of ferromanganese in the steel industry.


There are two standard methods of making magnesium, and both are briefly discussed in Encyclopedia Americana. The more common method is to obtain it electrolytically from magnesium chloride (as was done in 1808), since magnesium is even more reactive than aluminum.

It is actually easier to recover magnesium than aluminum. And magnesium has some interesting uses in its own right, for example, in pyrotechnics.

I would expect that there will be small-scale production of magnesium by alchemists like Dr. Gribbelflotz (he would fall in love with the burning magnesium ribbon experiment), but that large-scale extraction would come much later, perhaps not until two to four years after the aluminum industry commences ingot production. (In the long-term, magnesium could be serious competition for aluminum.)

In short, I am expecting that manganese and magnesium will be available for alloying use sometime in the 1640’s.


Besides the major alloying elements, there are a number which are deliberately added to aluminum in small quantities (up to, say, 0.5%) to improve particular characteristics (sometimes at the expense of something else). In this category, we have antimony, beryllium, bismuth, boron, cadmium, calcium, cerium (misch metal), chromium, cobalt, indium, lead, lithium, molybdenum, niobium, zirconium. (

{{Of these, the most readily available additives (in pure form) are antimony and lead. However, chromium and molybdenum at least are accessible to USE traders or prospectors. (Runkle). Still, to use them, we must not only find and mine the ore, we must reconstruct how to obtain the metal.

If the steel industry tries, and succeeds, in isolating these metals, then it is likely that the aluminum industry will at least experiment with them.}}


3004 is 95-98.4% aluminum, 1-1.5% manganese, 0.8-1.3% magnesium, up to 0.25% copper, up to 0.7% iron, up to 0.3% silicon, up to 0.25% zinc. 3104 is 95-98.4% aluminum, 0.8-1.3% manganese, 0.8-1.4% magnesium, 0.05-0.5% copper, up to 0.8% iron, up to 0.6% silicon. And 5182 is 93.2-95.8% aluminum, 0.2-0.5% manganese, 4-5% magnesium, up to 0.35% iron, up to 0.25% zinc, up to 0.2% silicon. You can look up 5045 and 5082 yourself. (There are other aluminum alloys used in cans; these include 1100, 3003, 5052, 5086, and 5154.)


Other packaging foil alloys include 1100, 1145, 3003, and 5052.

The pre-RoF cost of electricity for industrial use was about four cents per kilowatt hour (Electricity Forum), and thus the electricity for producing one pound of aluminum would have cost about forty cents. Since the U.S. price per pound of aluminum in 1999 was around sixty-six cents, but electricity is usually estimated as one third of the cost of production, the implication is that the smelters paid less than the normal industrial rate, or were more energy efficient than the ten kilowatt hour figure suggests.

In 1963, electricity cost 1.5-8 mills per kilowatt hour, or $26-136 per metric ton aluminum. (Brubaker 98)

The mine price of coal in 1999 was $16.63 a short ton (2,000 pounds). ( ) It takes one pound of coal to produce 1.25 kilowatt hours (WV Coal Association, Coal Facts 2005). So one ton of coal produces 2,500 kilowatt hours, and each kilowatt hour carries a coal cost of about seven-tenths of a cent. The per household charge would be another $75, bringing the total up to $700.

The power plant was a regulated utility before the RoF, and its owners were left behind. Hence, based on Canon, it should have been nationalized, and it will probably charge for electricity on some sort of “cost plus” basis. {{It may even be willing to grant “start-up” discounts to new industries.}}

Some of the power plant costs — debt service, for example — have disappeared. Others, like wages and coal, remain.

[Help!: what kind of amperage can the power plant in GV presently supply? what about after they switch over? If you increase the number of pots on a line, do you need to increase amperage?]

7020 isn’t supposed to be more than 0.5% manganese, 0.35% silicon, or 0.4% iron, while 3104 deliberately has more manganese, and has a higher tolerance for silicon and iron. So it will be necessary to reduce not only the amount of manganese, but perhaps also those the other two elements, if you are reclaiming 3104 can aluminum to make 7020. You may also need to add zinc (to 4-5%), chromium (to 0.1-0.35%), and perhaps magnesium (to 1-1.5%). Finally, there are some peculiar limitations on zirconium and titanium. 7075 needs more copper (1.2-2%), zinc (5.1-6.1%) and magnesium (2.1-2.9%), but less manganese (not more than 0.3%).

It appears that the presence of significant amounts of iron can be fortuitous. In the aluminum-copper alloys, iron (at least at levels below 0.7%) appears to have a strengthening effect. (Key to Metals). We don’t deliberately add the iron, we just don’t work quite as hard to get rid of it.

The deliberate addition of iron is unusual, but known in the art (Belov, 245-306)

The NAP report mentions that oxides of calcium, iron, zinc and silicon are among the potential problems in recycling aluminum.

A USDOE fact sheet lists calcium, lithium and sodium as “elements non grata,” but I am aware of advantages to the first two, at least.

Arsenic is toxic, hence undesirable in aluminum foil for food packaging. Carbon combines with aluminum to form aluminum carbide, which is decomposed by water, possibly resulting in surface pitting. Hydrogen causes pore formation during casting. Vanadium lowers conductivity. (


bevarage can lag time is about 65 days. (Choate, 61)

Aluminum in U.S. autos is 5-10% of vehicle weight in 2000. (Choate 61)

In 2000, the average aluminum use was 100 kilograms per car in Western Europe. (Key to Metals)

Melting aluminum requires 26 kilojoules per mole (one mole being about 27 grams). In contrast, reduction of one mole of aluminum from aluminum oxide (alumina) requires more than 780 kilojoules. (Shakhashiri) That implies an energy requirement which is only about three percent of that for primary production; the Aluminum Association says that the actual figure is five percent.


Brubaker (92) says energy efficiency of cell is 35% (1.7 Volts to reduce, 4.8V used).

Plambeck says that thermal power plants are only 35% efficient at generating electricity.

 Requirement1963 Unit Price1963 Cost
Electricity17,000 kWh2-8 mills/kWh   Brubaker 178 $34-136
Alumina2 metric tons$60/mton   USGS + ship? $120
Carbon500 kg$55-70/mton   Brubaker 91 $27.50-35
Labor18 man hours$3/hr, deflated data$54-63
Supplies  $20
Capital  ?
TOTAL less cap  255.50-374

min wage was $1.25/hr in 1963.

anode setter, 1997, WV $15/hr

1997 Missouri aluminum industry

average wage $40,577/yr

2005 Production workers


metal workers and plastic workers, all other 15.50

production workers, all other 16.04

Canada 1994

alumina and aluminum production employee

average salary 47,741/yr 1994 60,668 2003

Wages at Reynolds Aluminum in Oregon

$11-18/hr 1998

Washington State

1981 (in 1998 dollars)



$22.08/hr production workers

CPI 1982-4 = 100

CPI 1963 30.4

COU 2994 185.2


in 2001 = 173.62 in 1963

in 2000, 180.l62

in 1999, 185.03

in 1998, 188.12

in 1997, 191.02

in 1994, 207.93

in 2004, 164.15