Peter R Schmidt & S Terry Childs. American Scientist. Volume 83, Issue 6. November 1995.
In archaeology as in biology it can be very difficult to distinguish true from false homologies. Just as similarities between species may or may not indicate a common ancestry, so can common technologies or agricultural practices derive from the same source or arise independently when different cultures seek to solve the same problem.
Archaeologists have long held that ironmaking technology spread by diffusion from a single point of origin. The smelting of iron is thought to have been reduced to practice by the Hittites in what is now Turkey around 1600 to 1200 B.C., and the techniques then to have slowly migrated outward, reaching China, Britain and Nigeria during the first millennium B.C. Moreover, metalworking cultures were thought to progress through a metallic sequence, beginning with pure copper and moving from there to arsenical copper, tin, bronze and iron.
But Africa never fit this model very well. For one thing, unlike Europe and the Mediterranean region, Africa experienced no Bronze Age. Africans confronted the challenges of iron smelting directly, without the handicaps and advantages of metallurgical skills based on bronze-working. Only recently, however, have we begun to appreciate how this difference in metallurgical history led o the development of modes of experimentation that are distinctive to Africa.
One of us (Schmidt) began to question the application of the diffusion theory to the internal development of iron production in Africa W years ago, when he analyzed slags, glasslike waste products of smelting, found during excavation of an Early Iron Age site near Lake Victoria in northwest Tanzania. (In Africa the Early Iron Age is considered to have extended from 6000 B.C. to 600 A.D.) When the ancient slags were reheated, they became spongy at temperatures well above the highest temperature thought to have been achieved in antiquity.
The matter might have rested there except that in Africa, in contrast to Europe, it is still possible to observe a traditional smelting furnace operated by men who are still smelting, or were until recently. In a pioneering study that combined ethnographic studies of modem smelting with archaeological ones of ancient smelting, Schmidt and Donald Avery, a colleague from Brown University, came to the conclusion that African smelting had developed in ways that did not match the characterizations based on the European traditions known at that time.
In the past decade Schmidt and (co-author) Terry Childs, who has since led the laboratory analysis of slags as well as smelted and forged iron, have continued to investigate African smelting technology and the artifacts made from the iron it produced. By the mid-1980s we realized that there was increasing evidence that the iron produced in both ancient and modern smelting furnaces of western Tanzania has an unusual property: a high phosphorus content that is a legacy of the fuels used to fire the furnaces and the furnaces’ highly reducing (oxygen-deprived) atmosphere. Our recent research has been directed at understanding how African ironworkers coped with this distinctive chemistry, which facilitated smelting but complicated forging.
Iron, carbon and phosphorus form alloys with low melting points, and the resulting showers of liquid iron droplets within the smelting furnace led to the formation of large, dense iron masses with little slag and a high carbon content. Any advantages conferred in the smelting furnace, however, translated into problems at the forge. Phosphorus strengthens iron and steel, but in large amounts it imparts a potentially fatal brittleness that reveals itself to the smith as “cold shortness,” or the tendency of the material to crack and fail when it is cold-worked.
Modern ironmakers consider a 0.05 percent phosphorus level unacceptable How then did ancient ironworkers cope with iron whose phosphorus content was highly variable but sometime reached localized concentrations as high as 10.2 percent? Paradoxically, one of their techniques was to decarburize the iron before forging, converting a mild steel to a wrought iron. In the decarburized iron, phosphorus acted as does carbon in steel, making the iron into a stronger material and one that could be work-hardened.
European Iron Age Bloomeries
Iron ore is essentially an oxide of iron with associated minerals. In smelting, the ore is heated together with a fuel whose burning produces enough heat and a sufficient supply of carbon monoxide to reduce the ore to metallic iron. The reduced iron often contains some carbon. In the absence of other contaminants, the amount of carbon determines its mechanical properties. In the order of increasing carbon content, the alloys of iron and carbon are wrought iron, mild, medium and hard steel and pig, or cast, iron. In general, wrought iron is too soft and cast iron is too brittle to be a satisfactory utilitarian metal. The goal of modem ironmaking is generally to produce steel, that is, an alloy of carbon and iron with a carbon content of between 0.5 and 1.5 percent.
Iron is more difficult to smelt than copper, the principal constituent of bronze. Iron melts at a higher temperature, and because it is more active chemically, it must be smelted in a more reducing atmosphere. Indeed, it proved to be sufficiently difficult to achieve the melting temperature while maintaining reducing conditions that until the 18th century iron was smelted in the solid state at temperatures below its melting point. This process produced a bloom, a spongy iron mass interlarded with slag, rather than a puddle of liquid metal.
During the Early Iron Age in Europe and much of Africa, iron was smelted in either a bowl furnace supplied with air through a blowing pipe, or tuyere, or a shaft furnace supplied with forced air. A charcoal fire was started in the furnace bottom, and a mixed or layered charge of charcoal and ore was added to the furnace.
As the ore particles descended through the furnace, they passed through progressively hotter zones. High in the furnace the carbon monoxide from the combustion of the charcoal reduced the iron oxide in the ore to a porous sponge made up of small flakes of iron with slag. As the flakes passed down toward the tuyere, they began to sinter into a spongy mass. Meanwhile, the iron oxide began to react with silica from the ore to form a slag in which the many iron particles were trapped. The formation of this slag protected the iron flakes from reoxidizing as they passed in front of the tuyere, where the atmosphere was strongly oxidizing. Below the tuyere, some slag drained out of the pasty mass of slag and iron particles and fell to the furnace hearth. The final bloom was a porous lump of iron, somewhat refractory silicates and bits of unreduced ore that collected above the liquid slag. Overall it was heterogeneous in carbon content, and it was, generally speaking, steel.
In addition to being more difficult to smelt, iron was more difficult to work or fashion into utilitarian objects than bronze. Because copper melted at temperatures attainable by early metalworkers, copper and bronze objects could be made by the relatively easy process of casting. Iron objects, on the other hand, had to be individually shaped by forging, a much more arduous process. Moreover, by the time it was forged, bloomery iron typically had a very low carbon content. Even when it was work-hardened by repeated hammering, its tensile strength was lower than that of some bronzes. The virtues of iron could not be fully exploited until some sophisticated techniques of metalworking were discovered, such as casehardening quenching and tempering.
After smelting, some furnaces were broken open to retrieve the bloom, which was reheated in a forging furnace until the slag became viscous. Since the bloom was placed in the strongly oxidizing zone in front of the tuyere and since it was still rather porous, this was a strongly decarburizing treatment. Once the bloom had been reheated, it was pulled from the forge and hammered to expel the slag. Repeated heating and hammering consolidated the bloom into a lump of metal that varied from wrought iron to low-carbon steel.
The later blast furnace differs from the bloomery furnace primarily in that it operates at higher temperatures and produces a liquid metal. Under these conditions, particles of iron absorb considerable carbon as they pass down through a fluid slag into the charcoal hearth. Because the addition of carbon depresses the melting point of the iron, liquid metal collects in the hearth, covered by a layer of liquid slag. Whereas the bloomery furnace produced a heterogeneous iron with a low carbon content and considerable amounts of slag, the blast furnace produces cast iron with a high carbon content. Bloomery iron had to be recarburized to regain hardness, but blast-furnace iron has to be decarburized to reduce brittleness and increase malleability.
African Iron Age Bloomeries
Our investigations of African smelting technology began in the early 1970s when one of us (Schmidt) uncovered some tantalizing clues during excavations of furnace pits at an Early Iron Age site on the western shore of Lake Victoria. The excavations turned up bits of clay tuyeres whose exterior had been vitrified, much as a pottery glaze is vitrified in a kiln, and slags whose flow temperatures proved to be 1,350 and 1,400 degrees Celsius, significantly higher than the 1,200 degrees thought to be the upper temperature limit of European bloomeries. This evidence suggested that in African bloomeries the tuyeres extended deep into the furnace, so that the air blast arriving through them was preheated. This was an exciting possibility, not the least because preheating wasn’t known in Europe until J.B. Nielson’s patent in 1828.
To find out whether this surmise was correct, Schmidt and Avery arranged to observe and characterize several smelts conducted by the Haya, a Bantu-speaking agricultural people who live on the western shore of Lake Vidoria. The Haya had abandoned the bloomery process 50 to 60 years earlier because of the ready availability of scrap steel, but the old men, who still remembered the process, were willing to smelt again. The Haya are not the only African people with a living ironworking tradition, but they were of particular interest because there is archaeological evidence of iron-smelting activity among these people extending back to 200-600 B.C.
In preparation for the smelt, the Haya felled trees and burned them to produce charcoal and mined and roasted iron-bearing ores. The hardwood from which the charcoal was made, Muchwezi, grows only in immersed areas of swamp. The trees were felled and burned in place on islands built up of mud and sticks. The ore was roasted, or presmelted, in a smoldering fire started in the furnace pit the day before the first smelt. The net result of roasting was to extract all potentially damaging moisture, increase the surface area of reaction through the development of cracks and introduce carbon into the ore.
The smelting furnace was a pit lined with mud made from the earth of a termite mound. On the morning of the smelt, swamp grass (Ishanga) was burned in the bowl until the bowl was filled with charred reeds and ash. Eight tuyeres 50 or 60 centimeters long were placed around the bowl, with their ends extending deep into it. A cone-shaped shaft was then constructed of refractory slag collected from old smelting sites and termite earth, and the furnace was charged with charcoal and roasted ore from the top.
The Haya process had three distinguishing features: The iron ore was roasted before it was smelted, the furnace bowl was filled with charred swamp reeds, and the tuyeres extended deep into the bowl. These might seem like minor details, but they added up to a distinctive response to the challenges posed by local resources.
The roasted ore was reduced as it descended through the furnace stack, small flakes of iron formed and the other components of the ore melted to form a fluid slag. But then the fluid slag, containing some iron flakes, infiltrated the bed of charred swamp reeds, coming into intimate contact with a large area of essentially pure carbon. The result was a violent reaction of oxygen and carbon, called a carbon boil, whose vigorous progress could be monitored through the tuyeres. At the height of the reaction large bubbles of carbon monoxide and carbon dioxide repeatedly rose to the surface of the slag, expanded it and tore it open.
Once the oxygen was removed from the slag, iron began to precipitate out of it in large crystals. As the growing bloom advanced into the carbonaceous mass within the pit, it trapped charred reeds, producing microenvironments that carburized the iron around the charcoal inclusions. This process, at a minimum, produced a heterogeneous mild steel whose carbon content averaged between 0.2 and 0.6 percent.
At the time we were not as interested in the metallurgy of the blooms (nor did we have access to many for analysis) as in the details of the smelting process and its efficiency. Bloomery furnaces had to be operated within a fairly narrow temperature range. At temperatures much higher than 1,300 degrees Celsius, conditions within the furnace became oxidizing rather than reducing. On the other hand, if the temperature was much below 1,200 degrees, a fluid slag did not form and the reduced iron failed to consolidate into a bloom. The degree of preheating achieved in the Haya furnaces elevated the temperature at which reducing conditions could be achieved from about 1,200 degrees to between 1,300 and 1,500 degrees. Because they operated at higher temperatures while maintaining a reducing atmosphere that did not rapidly consume fuel, furnaces such as those operated by the Haya and their ancient predecessors were more fuel-efficient than cold-blast European ones.
The Cast Iron Puzzle
Once a wide variety of slags, iron blooms and iron objects were subjected to laboratory analysis, we discovered something our original research missed. Many of the samples turned out to include patches of cast iron, a material not often produced in the bloomery furnace. Moreover, the cast iron typically had an unexpectedly high phosphorus content. These startling findings caused us to look again at some of our earlier interpretations of smelting in northwestern Tanzania.
As we have mentioned, the addition of large amounts of carbon depresses the melting point of iron; pure iron has a melting point of 1,534 degrees Celsius, but iron with a carbon content of 4.3 percent melts at 1,147 degrees. Although the diffusion of carbon into the growing bloom was largely a solid-state process, in the Haya smelting furnace it was also possible to produce tiny pockets of liquid cast iron in the carbonaceous bed, where the right combination of temperature and reduction occurred.
The addition of phosphorus to cast irons further depresses their melting temperatures. Indeed in the early part of this century phosphorus was deliberately added to cast irons to increase the time they would stay liquid after they were poured. In the Haya furnace, the phosphorus led to the formation of cast iron at relatively low temperatures.
Where did the phosphorus come from? Phosphorus sometimes occurs in iron ores, but it can also be a constituent of fuels, ceramic refractories or flowing materials added to the charge to help remove the earthy part of the ore. Small amounts (0.1 to 0.2 percent) of phosphorus pentoxide (P sub 2 O sub 5 ) occur in iron ores found in northwest Tanzania, but the ore was not the sole source of phosphorus. A more significant source of phosphorus was the fuels, some of which were continuously added during the smelt. The Mudvroezi tree (Syzygium guineenis), when burned to make charcoal, produces an ash containing about 0.15 percent phosphorus. The charred Ishanga reeds (Miscanthidium violetti) packed into the furnace pit have an average phosphorus content of 0.34 percent. Since analysis shows that ancient blooms have phosphorus levels as high as modem ones, it is likely that ancient charcoals contained similar amounts of phosphorus.
The same hot, reducing conditions that led to the carburization of the bloom also led to the absorption of the available phosphorus. Phosphorus that was introduced into the furnace would vaporize at a temperature of 1000 degrees Celsius. Because the furnace atmosphere was strongly reducing, it would remain in the elemental form, which readily dissolves into iron and to some extent into carbon steel, depending on local furnace temperatures.
Although most of the carburization of the iron took place when the bloom grew into and around the carbonaceous mass of charred reeds, this was not the only location where cast iron was formed. By excavating the pits of experimental furnaces after smelting, we were able to determine the precise locations of slags of different types. (It is difficult to obtain an accurate picture of the smelting process from slags whose original context is not known because slag chemistry is exceedingly complex and variable.) These analyses showed that although most of the ore was reduced on the periphery of the blast zone below and in front of the tuyeres, there was a second hot, reducing zone above and behind the tuyeres.
Solid grains of iron passing through this zone would have absorbed significant amounts of phosphorus and carbon evolving from the carbon boil in the furnace bed. Enough phosphorus and carbon were sometimes absorbed to depress the melting temperature of the iron below the ambient temperature. Drops of cast iron then formed and fell on the bloom below, where they helped carburize the ferritic iron (Figure 6) (omitted). (Ferrite is the crystalline form pure iron assumes when it solidifies at low temperatures.) The dripping cast iron probably explains the high carbon and phosphorus concentrations on the exterior of consolidated Haya blooms.
Although alloys of iron, carbon and phosphorus accumulated primarily at the periphery of the bloom, their effects extended deeper into the bloom because phosphorus and carbon migrated from the cast iron into the ferrite. In a small piece of bloom excavated from a furnace at another prehistoric site known as NG5, an early effect of phosphorus infiltration was production of white patches of metal with distinctive shapes. A similar process of infiltration also increased the carbon content of the piece of bloom.
The Effect of Phosphorus on Smelting
We were able to confirm this furnace chemistry by studying the chemistry and microstructure of slag samples and pieces of Early Iron Age blooms (Figure 8a) (omitted), heirloom Haya blooms passed from generation to generation for remembrance’s sake (Figure 8b) (omitted), and blooms produced in experimental smelts conducted by Haya elders (Figure 8c) (omitted). In examining these samples, we were looking for evidence that eutectic alloys of iron, carbon and phosphorus had formed in the smelting furnace. Eutectic alloys solidify at a lower temperature than each of their constituents or than alloys with off-eutectic compositions. Because they are the last fraction to solidify, they typically appear as irregular bodies within a material of quite different appearance.
Analysis of slags formed above and behind the tuyeres in a replicated Haya smelt showed that drops of liquid iron captured in the slag were phosphorus-rich. Analysis of blooms showed the presence of the iron-iron phosphide eutectic. This eutectic forms only when the iron contains approximately 10 percent phosphorus, which we verified through chemical analysis. It has a melting point of 1,050 degrees Celsius, well below the melting point of cast iron.
Further analysis of several blooms showed phosphorus-rich cast iron toward their periphery, signaled by the presence of the iron-carbon-iron phosphide eutectic. This eutectic, which contains about 2 percent carbon and 6.9 percent phosphorus, has a melting point of only 953 degrees, even lower than that of the iron-iron phosphide eutectic. The presence of the phosphide eutectics indicates that the phosphorus significantly extended the temperature range over which liquid metal showered down on the furnace floor. By this means the phosphorus contributed to the formation of a denser, less spongy bloom.
In 1977 we conducted excavations at an Early Iron Age iron-smelting site near Kemondo Bay on Lake Victoria known as KM2. This site was used primarily for industrial iron production in the first century A.D., as well as during several other periods in the next 500 years. In earlier excavations of smelting sites, we had sometimes found small holes in the floors of furnace bowls. The holes were usually empty, but on the basis of ethnographic evidence, we thought they likely contained a ritual object or medicine whose purpose was to increase the output of the smelt or the quality of the iron it produced, or to protect the smelt from witchcraft or other sinister influences. In 1978 at another site, KM3, we excavated a furnace bowl partially damaged by deep hoe cultivation whose floor was composed of randomly placed sandstone blocks. When we pried up one of these blocks, we found a piece of partly oxidized iron standing upright in a small hole. This bloom has a particularly interesting history.
Because this piece of iron bore the impressions of grass on its surface, we knew it was a piece of ancient bloom rather than a forged artifact. Metallographic examination showed massive carbide (Fe sub 3 C) plates along the grain boundaries of the crystalline iron. These probably arose from the prolonged heating of the bloom in a carbon-rich atmosphere, presumably during smelts conducted after it was sealed in the small pit. But when we recently reanaIyzed this ancient bloom, we had a surprise. In addition to the carbide we had seen earlier, it contains the iron-carbon-iron phosphide eutectic.
The Effect of Phosphorus on Forging
The heterogeneous iron produced by the bloomery process required considerable versatility at the forge. To make useful objects, smiths had to recognize different constituents of thebloom and know how best to deal with them. The presence of phosphide eutectics and localized areas high phosphorus content would have made the iron tricky to work.
Phosphorus in iron and steel behaves in some ways like carbon because it strengthens the iron and increases its work hardenability. Martha Goodway of the Smithsonian Institution recently showed that 17th-century harpsichord strings, once thought to be steel, were in fact made from iron with a relatively high phosphorus content (0.16 percent). The phosphorus so strengthened the iron that it could be used to make strings strong enough for instruments built to a longer scale. The key to this effect, however, was the absence of carbon.
Ferritic iron that contains more than 0.5 percent phosphorus has an established reputation for brittleness. Because the phosphorus segregates at grain boundaries, the material has low intragranular strength. It will fracture when cold-forged and may shatter on impact at room temperature. The presence of carbon as well as phosphorus only aggravates the problem. The accidental addition of high-phosphorus scrap steel to the melting furnace in a modern steel mini-mill can lead to the manufacture of rebar that is brittle enough to be broken over the knee.
Once the phosphorus was in the bloom it was there to stay. No efficient method for removing phosphorus from iron existed until the late 1870s, when the Thomas-Gilchrist process was introduced. Fortunately, the carbon was less recalcitrant. The bloom could be efficiently decarburized simply by heating it in the strongly oxidizing conditions of the forge.
In this difference lay the smith’s salvation. As the harpsichord wire suggests, the addition of phosphorus can make plain wrought iron as hard and strong as a medium carbon steel. In general African ironworkers coped with the phosphors-rich steel and cast iron they had smelted by decarburizing it in the forge to a wrought iron or low-carbon steel. This technique produced a malleable material but one that, because of its phosphorus content, was still relatively hard.
We gained further insight into the way smiths handled the presence of phosphorus by examining objects forged in both prehistoric and modern times. Examination of a knife blade from the KM2 site demonstrates that the phosphorus was indeed retained during forging. The phosphorus in the decarburized metal announces itself in two ways. First, many of the ferrite grains are veined, an internal structure that is sometimes associated with the presence of phosphorus. Second, some areas of the metal turned white when Oberhoffer’s solution, an etchant that highlights phosphorus, was applied to it. Treatment with this solution revealed alternating bands of small and large ferrite grains, a pattern characteristic of an iron with a heterogeneous phosphorus distribution that has been hammered into a tool. An artifact excavated from a forging pit at Rulama, a 13th-14th-century A.D. forging site located 20 kilometers to the west of Lake Victoria near Lake Ikimba, and an excavated (undated) hoe (Figure 7) (omitted) found at RM2, a site in Katuruka village overlooking Kemondo Bay demonstrate the continuity of this chemistry from ancient to modem times.
Another instructive object is a large iron socket for the butt of a spear (Figure 9) (omitted), dating to about the 17th century A.D., that was excavated at the nearby Rugomora Mahe site. This socket was probably intended to hold a large ceremonial spear, such as those known to have been used for ritual purposes in the Buanda kingdom only 200 kilometers to the north. Two sections from the socket show microstructures that vary from those characteristic of high-carbon steel to those characteristic of ferrite. Distributed throughout the transitional areas between higher- and lower-carbon steel are numerous spheroids that contain 7 percent phosphorus by weight, a phosphorus concentration close to that of the iron-carbon-iron phosphide eutectic. The spheroids are relic bits of this eutectic created when the metal was partially decarburized and repeatedly hammered at the forge.
The smith who made this socket obviously knew how to forge a bloom without fully decarburizing it and how to prevent areas of phosphoric steel and cast iron from cracking or failing as thebloom was worked or cooled. Because of its high carbon and phosphorus content the socket would have been brittle at room temperature, but this was probably of little consequence, given its intended ritual use. Indeed it is possible smiths selected the part of a bloom and the amount of decarburization based on the use to which the forged object would be put.
If the ancient smiths seem to have successfully negotiated the difficulties presented by phosphorus-rich iron, their modem counterparts, whose expertise has been developed on scrap steel, are less adroit. The difficulties encountered in forging phosphorus-rich iron are illustrated by a sickle forged by a contemporary Haya smith from an experimental bloom (Figure 10) (omitted). Metallographic examination of the bloom showed that it consisted of variable ferrite and low-carbon steel coated with a glassy slag. Pockets of the iron-iron phosphide eutectic and areas of initial phosphorus infusion (appearing as tiny dots) were also present.
A sample taken from the sickle shows a mostly ferritic structure with one area of low carbon steel and an uneven distribution of phosphorus. There is also evidence of pressure welding, a hot-forging technique. This was a good choice because phosphorus does not have the embrittling effect at high temperatures that it has at lower ones. On the other hand, there are also several cracks, apparently caused by the smith cold-working the edge of the sickle blade. This is a technique he might have avoided had he been better acquainted with the foibles of his metal.
Phosphorus captured during the reduction of iron in the ancient and recent iron-smelting furnaces of northwest Tanzania conferred some advantages, namely a lower-temperature liquid iron that contributed to the formation of a dense and carbon-enriched bloom. Any advantage conferred in the smelting furnace, however, translated into special problems at the forge. These were typically overcome by decarburization and possibly by the innovative use of phosphorus-rich irons for special purposes.
Our latest results only increase our admiration for the versatility of African ironmaking technology both in ancient times and the recent past. Phosphorus was not identified and its influences on the properties of iron fully recognized until the 18th century. Although the early craftsman knew nothing about phosphorus, he had developed effective means of ameliorating its effects.
But why was the phosphorus there in the first place? Modem Haya furnaces are similar but not identical to Early Iron Age furnaces from the same region. One difference is that small pieces of wood charcoal are found in the pits of Early Iron Age furnaces. Charred ass is a significant part of smelting only in the Late Iron Age. Ancient pollen samples suggest that the western shores of Lake Victoria, where the Haya live, were once covered with climax rain forest and that the forest was cleared intensively between 0 and 500 A.D. for farming and industrial purposes. The use of grass in the smelting pits, then, may have been an adaptation to the depletion of forest resources. Ironically Early Iron Age people had the technological capacity to clear large areas of forest in a relatively short time because they possessed iron and steel tools. In prehistory as in history, technology gives and technology takes away.