Iron Production in three Ming Texts:
Tie ye zhi, Guangdong xinyu, and Tian gong kai wu

Donald B. Wagner

Third International Symposium on Ancient Chinese Texts and Records on Science and Technology, Tübingen, 31 March - 3 April, 2003.

22 May 2003
Minor revisions 18 June 2003
Stylistic revisions 10 November 2023

This paper is concerned with source criticism. When we use a narrative source, we need to ask such questions as: Who is telling this? How does he know it? Whom is he telling it to, and why? Often it is not possible to obtain very satisfactory answers to these questions, but the more we can learn about a given source, the better qualified will our use of it be in historical research.

In this paper I consider three remarkable texts concerning iron-production technology in the Ming period, attempting to throw light on source-critical questions.

Tie ye zhi 鐵冶志

It has often been remarked that bureaucracy favours the historian, for it produces great quantities of paper which later become historical sources. The Ming state ironworks in Zunhua 遵化,  Hebei, provides a good example: its administration, especially the management of a large force of corvée, conscript, and convict labour, gave officials of the Ministry of Works (Gong bu 工部) enough trouble that a good deal of written communication was required, and some of this has survived to tell us about both the technology and the organisation of the works. I have dealt with the history of the works in detail in my forthcoming volume on ferrous metallurgy for Joseph Needham's Science and civilisation in China; here I shall simply note some interesting points.[1]

The book Tie ye zhi was written by Fu Jun 傅浚, who from 1513 was in charge of the Zunhua ironworks. From the Siku editors' description it appears to be a kind of local gazetteer (fangzhi 方志), describing all aspects of the ironworks, though they categorize it among governmental writings (zheng shu 政書) rather than with the local gazetteers.[2] It is now lost, but its description of the blast furnace, used to smelt iron from ore, appears to have been quoted by two later authors, Zhu Guozhen 朱國禎 (ca. 1557-1632) and Sun Chengze 孫承澤 (1593-1675).[3] The identification of these quotations as coming from Tie ye zhi is by Liu Yuncai;[4] his arguments seem convincing and, while I would prefer even better arguments, I shall in what follows assume that the identification is correct.

Identification of the blast furnace description as coming from Fu Jun's book means that we can firmly place this technology in a geographic and economic context. Several major and minor sources tell more about the history and function of the Zunhua ironworks.[5]

To make a long story short, the Zunhua ironworks smelted ironsand in a small blast furnace using charcoal as the fuel; the pig iron produced in the blast furnace was converted to wrought iron by the process known in English as fining or puddling (Chinese chao 炒, literally 'stir-frying'). This was essentially the same technology as could be observed in the early 20th century in the Dabieshan 大別山 mountains, near Xinyang 信陽, Henan.[6] In that region the labour cost of producing wrought iron was 2-4 man-months per tonne.

The bureaucratic sources on the administration of the Zunhua ironworks give sufficient input and output data to allow a rough calculation of labour costs there, and the result is startling: at Zunhua in the 15th century the cost of producing wrought iron was 50-60 man-months per tonne, an order of magnitude more than in the Dabieshan region.

Various technical and organisational improvements seem to have halved this cost by the early 16th century, but Zunhua iron remained costly. In 1581 a civil servant estimated that Zunhua's annual production quota of 125 tonnes of wrought iron had a market value of 2700 taels of silver, but cost over 10,000 taels to produce. His recommendation that the works be closed was approved.

I suggest in my forthcoming work that the establishment of a state ironworks at Zunhua was necessary at the beginning of the Ming to provide high-quality iron to the armouries of Beijing. In north China at this time iron was normally produced using mineral coal as the fuel, and this iron would have had a sulphur content which was too high for the best weapons. Charcoal iron was necessary, but could only be obtained from the south. At the beginning of the Ming dynasty the supply routes from the south could not be relied on, and it was necessary to establish state production of charcoal iron near the capital, regardless of how costly it might be. Two centuries later, with the consolidation of Ming power and the rise of trade between north and south it was possible, and economically more rational, to rely on the market instead.

---

The source-critical point that comes out of this discussion is that the value of a source is much greater if it is possible to place it securely in a precise context. We know that the description of iron-production technology given by Zhu Guozhen and Sun Chengze relates to the Zunhua ironworks (both say so). It was probably written originally by Fu Jun, who as director of the ironworks must have written it for a practical purpose; therefore it is likely to be accurate. Other sources concern the economic and political context in which this technology was applied, so that its efficiency can be compared with that of the same technology elsewhere, and we can begin to move outside the realm of 'internal' studies in the history of technology to broader questions in economic and political history.

Guangdong xinyu 廣東新語

Another description of Ming iron-production technology is in Guangdong xinyu. This very large and interesting book was written in the early Qing period, ca. 1680, but its author, Qu Dajun 屈大均, was a Ming loyalist, and the book describes the province of Guangdong in the late Ming period. Numerous technological subjects are covered, and the description of the large-scale iron industry of the province is quite impressive. But as we shall see, this description is a patchwork of several different sources, so one must be careful in using it as a source.

Other sources[7] indicate that in recent centuries two distinct technologies for iron production were in use in Guangdong, and the ironworks using them were recognised in law as forming two distinct sectors of the industry. Ironworks of the small-scale sector produced pig iron in small blast furnaces, and used the same furnaces to melt pig iron or scrap for casting products for local consumption. The large-scale sector was made up of large firms operating numerous ironworks in the forested mountains of the province. Most of their production was shipped by river to the industrial town of Foshan 佛山, near Guangzhou. Here some of the pig iron was converted to wrought iron and the rest went to foundries. In either case high-quality products of the smithies and foundries of Foshan were marketed far and wide: up and down the coast and all over Southeast Asia.

Our only clear source for the technology of the small-scale ironworks of Guangdong is an album of gouache paintings by a Chinese artist in Guangzhou in the mid-19th century, now in the Bibliothèque Nationale in Paris.[8] The album illustrates exploration for iron ore, the digging of ore from a hillside, the calcining of the ore, the charging of the ore into a small blast furnace (see Figure 1), the elimination of slag, the tapping of the furnace (Figure 2), and the casting of iron bars. We cannot be sure that the paintings are accurate depictions, but comparison with small blast furnaces known from other times and places in China indicates that the artist had seen a blast furnace like this and that his painting is approximately correct.

The passage on the iron industry in Guangdong xinyu[9] gives a description of the work force of an ironworks: it includes more than 200 furnace tenders, 300 miners, and 200 'water-carriers' and charcoal producers; transportation is provided by 200 oxen and 50 river vessels. The blast furnace produces 2-4 tonnes of pig iron per day, and this is shipped down-river to Foshan. It seems clear that Qu Dajun has misunderstood one of his sources here, for 200 furnace-tenders would not be needed for one furnace, and a production of 2-4 tonnes per day would not require 50 ships to transport the product to market. Obviously the passage refers to a large firm which operated numerous ironworks scattered over a large area. (Two charcoal blast furnaces would not normally be placed too close together, as this would mean a doubled load on forest resources with little or no gain in efficiency.)

I shall now attempt to prove that Qu Dajun in his description of a blast furnace drew on at least two different sources which described two radically different types of furnace.

The description begins as follows:

爐之狀如瓶其口上出口廣丈許底厚三丈五尺崇半之身厚二尺有奇

The furnace has the shape of a vase with its mouth upward. The breadth at the mouth is about a zhang 丈 [3.2 m]. The foundation is 3 zhang 5 chi 尺 [11.2 m] thick [sic!],[10] and the height is half of that. The thickness of the body is slightly more than 2 chi [64 cm].

This is a very large furnace for its time, over 5 metres high and three metres broad at the mouth, and there is some archaeology in Guangdong which tends to confirm the description.[11]

The passage goes on to describe the operation of the furnace,[12] and ends with:

鐵礦既溶液流至于方池凝鐵一版取之以大木杠攪爐鐵水注傾復成一版

When the iron ore has 'melted' [i.e. has been smelted to molten metallic iron], it flows into a rectangular mould and solidifies into an iron slab. This is removed. Then they shake [jiao攪] the furnace with large wooden poles, and iron flows out to form another slab.

The most common meanings of the word jiao 攪 are 'to shake' and 'to stir'. Its use here is difficult to explain if the furnace was as large as Qu Dajun states.

For comparison, Figure 3 shows a traditional type of blast furnace being tapped in Gansu in 1958. It is not the same type as the Guangdong furnace, but it is of roughly the same size. It is easy to see that such a furnace could not be 'shaken', nor could the iron in it be 'stirred' with a wooden pole, nor would there have been any reason to stir it.

The tapping of a small furnace in Guangdong, seen in Figure 2, could clearly be described as jiao. It seems therefore very likely that while Qu Dajun had his description of the physical construction of the furnace from a source concerned with the large-scale ironworks of Guangdong, his source for the tapping was related to the small-scale ironworks.

---

Qu Dajun's description is valuable because it presents the technology used in a specific place, and places it in an economic context. However it seems clear that in writing it he patched together a number of sources without fully understanding them. His description of the resources of one ironworks comes undoubtedly from a source that describes a firm which operated many ironworks, and his description of a blast furnace is drawn from at least two sources which describe two very different types of furnace.

Qu Dajun wrote his book to praise the virtues of his native province as it was before the Qing conquest. His purpose was thus literary (and perhaps political) rather than practical, and while he no doubt intended his information to be accurate, he did not always understand his sources. In using the Guangdong xinyu as a source we must therefore be aware of this problem and constantly attempt to ascertain the nature of the sources he used and adjust our conclusions accordingly.

Tian gong kai wu 天工開物

Perhaps our most important source for the technologies of the late Ming is the remarkable Tian gong kai wu, by Song Yingxing 宋應星, published in 1637.[13] It covers a broad range of industries, and includes illustrations. Unfortunately Song Yingxing only occasionally tells where in China the technologies he describes were used, and he seldom gives economic information. And of course, as was common in his time, he does not tell where his information comes from.

Much research is needed to clarify the sources of the Tian gong kai wu. Some of Song Yingxing's descriptions may be eyewitness accounts, but it seems impossible that he could have seen all of the technologies he describes. The range of the book is simply too broad for one man's personal observation.

In the following I shall take up some passages in Song Yingxing's description of iron production and attempt to show (1) that his information is accurate; (2) that the illustrations are accurate; but (3) that the illustrations do not always show the same technologies as the text.[14]

Washing ironsand. Song Yingxing describes two types of iron ore, dingtie 錠鐵 and shatie 砂鐵, literally 'lump iron' and 'sand iron'. Both are found close to the surface and do not require deep mining. 'Lump iron' is probably what in English is generally called bog iron. 'Sand iron' is ironsand (or black sand).

Ironsand is small grains of magnetite (Fe3O4), and is found in small quantities in sand in rivers that flow out of granite mountains. The greater part of the sand is of course quartz. Magnetite is heavier than quartz, so it can become concentrated at certain places in rivers. This is, generally speaking, where the current suddenly slows, for example where a river widens or at the outside of a turn in the river. This naturally concentrated ironsand can contain up to about 7 percent magnetite. 'Washing' the ironsand can improve this to as much as 95 percent.

Figure 4, a photograph taken in Henan around 1917, shows how the 'washing' of ironsand was traditionally done.[15] A sluice was built, in this case 1.6 m long and with side-boards 10 cm high. River sand was shoveled onto the sluice and running water was led over it while workers stirred it with forks. The result was that most of the quartz sand was carried away by the water and most of the magnetite remained behind.

After a brief discussion of 'lump iron' Song Yingxing describes ironsand as follows:[16]

凡砂鐵一拋土膜即現其形取來淘洗入爐煎煉鎔化之後與錠鐵無二也

Sand iron: When the surface earth is cleared away it is revealed.[17] They take it and wash it, charge it into the furnace, and smelt it. After melting it is exactly the same as 'lump iron'.

This is correct as far as it goes, but his illustration (Figure 5) shows panning in still water where we should expect to see sluicing in running water. It is clear that the illustrator had seen what he depicts, for the implement in the illustration is an accurate depiction of the boji 簸箕 traditionally used in concentrating minerals from sand. Peter Golas (1999: 244, fig. 31) photographed one in use in Guangxi in 1994. But it is doubtful that this method would have been used in washing ironsand.

Panning is an appropriate method for extracting rare or valuable minerals from sand, but ironsand is plentiful in mountainous regions. Extracting magnetite from sand by this labour-intensive method would have meant that very little magnetite was lost; but overall, considering labour input per quantity of magnetite delivered to the furnace, panning would have been far less efficient than sluicing.

Therefore it seems likely that Song Yingxing, or his illustrator, had seen the washing of gold or some other valuable mineral, and believed, probably incorrectly, that the same method was used in washing ironsand.

The blast furnace. Further on in the same chapter Song Yingxing describes the blast furnace used in smelting the iron ore. His description begins: [18]

其爐多傍山穴為之或用巨木匡圍

Most furnaces are made in a pit at the side of a hill; but some are enclosed by a framework of heavy timbers.

Both of these ways of constructing a blast furnace were common in China. The second was still being used in the 20th century, as can be seen in Figures 6-7, published in 1958 in connection with the Great Leap Forward. The inner shaft is of firebrick or stone, the outer framework is of wood, and the space between is filled with tamped earth (hangtu 夯土).

More mystifying is the blast furnace constructed 'in a pit at the side of a hill'. Translators and commentators have generally not been able to deal with this,[19] but the mystery has recently been solved by Li Xiaoping (1995). This type of blast furnace was common in the Song and Yuan periods.[20] Figure 8 shows two blast furnaces dated to the 12th century excavated in Heilongjiang, and Figure 9 shows a diagram of the one on the left. The shaft of the furnace was simply dug into a hillside, lined with stones and fireclay, fitted with holes for blast and tapping, and used as a blast furnace.

So Song Yingxing's description of the two types of blast furnace is accurate. The illustration which accompanies his text, the right side of Figure 10, shows however a completely different type of furnace. It is not dug into a hill, nor does it have a wooden framework. This type of blast furnace is also well known; it is seen in Figure 11, photographed somewhere in central China about 1910.

Thus it seems that Song Yingxing here used a different source for his illustration than he used for his text.

Converting cast iron to wrought iron. The product of the blast furnace is cast iron, containing typically 3-4 percent carbon. If what is needed is wrought iron, with 0-0.2 percent carbon, a second process is required. A well-known traditional Chinese process for converting cast iron to wrought iron is shown in Figure 12, photographed in Shanxi in 1958.[21] The fuel (most often charcoal) burns at a very high temperature (perhaps 1200-1400°C) in a small well-insulated hollow in the ground. Pieces of cast iron are charged into this combustion chamber, and a strong blast of air (from some sort of bellows or blowing engine, not seen in the photograph) keeps combustion going and the temperature up. The worker stirs the mixture of fuel and cast iron with an iron rod and, if he is skilled at his work, most of the carbon burns out of the iron while not too much of the iron burns.

Variations of this process, generally called chao 炒, 'stir-frying', have traditionally been used all over China, and seem also to have been used as early as the Han period (Wagner 2001: 80-84). It is similar to the processes called in English fining and puddling.

When Song Yingxing comes to the conversion of cast iron to wrought iron, however, he describes a very different process, one which is very difficult to explain technically. His illustration is the left side of Figure 10, and in this case the illustration and the text fit together very well:

若造熟鐵則生鐵流出時相連數尺內低下數寸築一方塘短牆抵之其鐵流入塘內數人持柳木棍排立牆上先以污潮泥晒乾舂篩細羅如麵一人疾手撒 眾人柳棍疾攪即時炒成熟鐵 其柳棍每炒一次燒折二三寸再用則又更之炒過稍冷之時或有就塘內斬劃成方塊者或有提出揮推打圓後貨者若瀏陽諸冶不知出此也

If they are making wrought iron then when the cast iron flows out it is led into a rectangular hearth [tang 塘, lit., 'pool, basin'] which is constructed a few chi [feet] away and a few cun [inches] lower, up against a short wall.

While the iron flows into the hearth, several persons holding willow poles stand in a row on the wall.

Earlier they have taken wuchaoni 污潮泥 [some sort of earth, see below] and dried and sieved it so that it is as fine as flour.

One of the men quickly spreads this while the others quickly stir with the willow poles and [the iron] is immediately fined into wrought iron. The willow poles burn down two or three cun [inches] each time fining is done, and after they have been used twice they are replaced.

After fining, when the iron has cooled slightly, some [ironworks] simply chop the iron in the hearth into square pieces; others lift [these] out [while still hot] and hammer them into round bars before marketing. Such ironworks as those in Liuyang 瀏陽 do not know how to produce these.

The description and illustration are so precise that there is no real doubt that Song Yingxing, or the author of his source, had seen something very like this process. But it is very difficult to explain. Anyone who has worked with molten cast iron, as I have, will immediately object that the cast iron from the blast furnace, flowing into such a large open hearth, without thermal insulation, fuel, or any sort of air blast, will solidify before any significant amount of carbon has been removed.

Most translators and commentators seem unaware of this objection. They explain the passage and illustration in terms of modern open-hearth steelmaking processes, and state that the curious wuchaoni would contain iron oxide, FeO, to help remove carbon by the reaction FeO + C = Fe + CO, but this would not solve the problem.

The only commentator, as far as I know, who has been aware of the problem was one of the first, the German metallurgist Adolph Ledebur, more than a century ago, and he also proposed a solution. A Japanese friend had shown him a copy of Tian gong kai wu and translated the metallurgical sections for him. In his article about it he suggests that the wuchaoni spread on the iron contained saltpetre (potassium nitrate, KNO₃). It is a powerful oxidizing agent (this is its function in gunpowder), and might very well be able to accelerate the oxidization of the carbon in the iron sufficiently to keep the temperature up until the carbon is exhausted and the cast iron has been converted to wrought iron. Ledebur adds: 'That a quick conversion of pig iron to wrought iron can be effected by the use of saltpetre was demonstrated two decades ago by the Heaton process.'[22]

Wuchaoni 污潮泥 means (or can mean) something like 'filthy wet loam', and the label in the illustration has the variant chaonihui 潮泥灰, 'wet loam and ashes'. In pre-modern China and Europe saltpetre was manufactured in 'nitre beds', in which the raw materials were dung, earth, urine, and wood ashes (Williams 1975; Needham 1980: 188 ff). In China, nitre beds were described as so foul-smelling that birds would not fly over them.

In the nitre bed fermentation produces calcium nitrate, and treating a solution of this with ashes gives potassium nitrate and a precipitate of calcium carbonate.[23] Without enough information to go into detail it seems possible that the terms 'filthy wet loam' and 'wet loam and ashes' describe something like such a nitre bed. Gunpowder requires very pure saltpetre, but perhaps, in the fining process described by Song Yingxing, some much less pure product, containing small but significant amounts of saltpetre and calcium carbonate, might have been useful.

---

What is the relationship between author and illustrator in this case? The illustration fits the text in almost every detail, so that one definite possibility is that the illustrator simply followed the text without having seen anything himself. One detail tends to contradict this hypothesis: the label chaonihui, 'wet loam and ashes', where the text has wuchaoni, 'filthy wet loam'. This might imply that the illustrator had some information about the process without reference to the author; for example he might have heard the term in conversation with the workers. This is not, of course, the only possibility: the text and illustrations were no doubt prepared at different times, and Song Yingxing might on the two occasions have chosen differently among the terms he knew for this substance, whether from first-hand experience or from written sources.

One point which suggests that Song Yingxing relied on written sources is the curious wall mentioned in the text and seen in the illustration. The text states unambiguously that the workers stand on this wall, but why should they do this? If on the other hand they stood behind a light waist-high wall it would provide some protection from flying sparks and intense radiant heat from the hearth. I suggest as one possibility that Song Yingxing here misunderstood a written source which mentioned such a protective wall, and the illustrator blindly followed Song Yingxing's text.

Concluding remarks

The three texts discussed here are the best technical descriptions of iron technology available for any pre-modern period in Chinese history. Neither before nor after do we find such detailed information as they provide. Making use of these texts in historical research requires, however, some sort of answer to the question of 'the source's source': how did the author know what he wrote?

In the case of Fu Jun's Tie ye zhi, it is clear that the author, as administrator of the Zunhua ironworks, had direct access to what he described. Whether the text is his own eyewitness account, or was copied from someone else's, the description was undoubtedly intended to be accurate enough to be useful to administrative personnel at Zunhua, and Fu Jun had the opportunity to ensure its accuracy. Therefore we can with only some slight hesitation use his text in conjunction with other sources in the study of the economic history of the state ironworks at Zunhua.

The description in Qu Dajun's Guangdong xinyu, on the other hand, seems not to be an eyewitness account, but a patchwork copied from several different texts. His loyalty to the Ming and his attachment to Guangdong suggest that these are likely to be Ming texts from Guangdong; but he seems to have misunderstood some sources and combined sources which actually described very different techniques. It will therefore be vital, in using the Guangdong xinyu, to attempt a 'deconstruction' and a characterisation of each of Qu Dajun's sources.

The Tian gong kai wu is one of the most interesting books in the history of technology world-wide. What were Song Yingxing's sources? One or two seem to be known,[24] and extensive study will no doubt reveal more. The above consideration of a few passages suggests that he generally used reliable sources, but may not always have understood them correctly. The sources for the illustrations, and the relation between illustrations and text, is another matter which will require much future research.

---

Many thanks to Martina Siebert and Alan Williams for important comments and corrections.

References

Endnotes

1. Other studies are by Eberstein (1974: 33-37) and Sakuma (1972).

2. Siku quanshu zongmu tiyao, 17: 1759.

3. Yongchuang xiaopin, 4: 94-5; Chunming meng yulu, 46: 68a-69a.

4. Liu Yuncai 1978: 25.

5. Especially Da Ming huidian, 194: 18b-22a; Han Dazhang 韓大章 in Huang Ming jingshi wenbian, buyi 補遺, 2: 1a-9b. References to the Zunhua ironworks in Ming shi lu 明實錄 can easily be found in the relevant sections of Li Guoxiang & Yang Chang 1993 and 1995.

6. See e.g. Wagner 1985.

7. The following is based on Wagner 1997: 58-75, where detailed source references will be found.

8. 'Fer'. C.E. Oe 119 in-4°, 1-8.

9. Guangdong xinyu, 1700 ed., 15: 7b-10a; 1974 ed., pp. 408-410; translated in Wagner 1997: 65-67.

10. Han archaeology and accounts from the Great Leap Forward indicate that large blast furnaces required a substantial foundation which could withstand the high temperature at the bottom of the furnace and prevent any moisture from penetrating into it. It was typically made by digging a deep hole in the ground which was filled up with alternating layers of loose stones and tamped earth (hangtu 夯土). A thickness of 11.2 m for this foundation seems excessive, but there is no obvious alternate interpretation of the text.

11. Cao Tengfei & Li Caiyao 1985; Cao Tengfei & Tan Dihua 1985: 118-123.

12. For a complete translation see Wagner 1997: 65-67, but note that when I wrote that translation I had not yet thought through the issues raised in the present article.

13. The 1637 edition has been reprinted in facsimile a number of times, and has been used here. Two modern critical editions are Zhong Guangyan (1978) and Pan Jixing (1989). Translations include Sun & Sun 1966; Li Chiao-ping et al. 1980; Yabuuchi 1969. Adolph Ledebur's study of the metallurgical sections (1885) still has value today, as will be seen below.

14. For the whole context in which these passages appear see e.g. the translation by Sun & Sun (1966: 248-251).

15. For details see Wagner 1985: 8-9, 12-13, 28-32.

16. 1637 ed., xia 下: 15b.

17. Rich ironsand lodes would often be found buried under a thin layer of earth because of the way rivers shift their course.

18. 1637 ed., xia 下: 16a.

19. But note Ledebur 1885: 192.

20. For details see Wagner 2001: 42-50.

21. For details see Wagner 1985: 8-9, 12-13, 28-32; 1997: 21-25, 41-43.

22. Ledebur 1885: 192, fn. Several uses of saltpetre in puddling, including the Heaton process, are described briefly by Wedding (1874: 264-5).

23. Williams describes his experiments with this process as follows: 'A mixture of dung and earth was set up in a pile and urinated on daily for several months. Ammonia (from urea) is oxidised by the air with the aid of various bacteria to form nitrates. Extraction with boiling water yields a very dilute solution of calcium nitrate, which may be treated with wood-ash (containing potassium carbonate) to form potassium nitrate' (2003: 864, fn. 22).

24. For example he seems to have read Shen Gua's 沈括’s description of a steelmaking process: compare Tian gong kai wu, 1637 ed., xia 下: 17a with Mengxi bitan 夢溪筆談, 3 (Hu Daojing 1962: 135, §56).