an investigation into the relationship between ... - Wiley Online Library

Archaeometry 54, 1 (2012) 37–55

doi: 10.1111/j.1475-4754.2011.00614.x

A N I N V EST IGAT ION INTO T HE R ELATI O N S H I P BETWEEN T H E R AW M AT E RIAL S USE D IN TH E PR O D U C TI O N O F C H I N ESE P ORCE L AIN AND STO N EWA R E BO D I ES A N D T HE RE SULT ING M ICR O S TR U C TU R ES * M. S. TITE,1† I. C. FREESTONE2 and N. WOOD1 1

Research Laboratory for Archaeology and the History of Art, Dyson Perrins Building, South Parks Road, Oxford OX1 3QY, UK 2 Department of Archaeology and Conservation, SHARE, Cardiff University, Humanities Building, Colum Drive, Cardiff CF10 3EU, Wales, UK

The microstructures of porcelain and stoneware bodies from north and south China, spanning the period from the Tang to the Ming dynasty (7th–17th centuries AD), were examined in polished sections in a scanning electron microscope (SEM) after etching the sections with hydrofluoric acid (HF). Mullite, present as fine, mainly elongated crystals, is the dominant crystalline phase observed. The bulk chemical compositions of the bodies are determined by energy-dispersive spectrometry in the SEM, and the relative amounts of mullite and quartz present in the different ceramics are estimated from X-ray diffraction measurements. Mullite formed from areas of kaolinitic clay, mica particles and feldspar particles is distinguished through a combination of the arrangement of the mullite crystals, and the associated SiO2/Al2O3 wt% concentration ratios. It is shown that very different microstructures are observed in ceramic bodies produced using kaolinitic clay from north China (Ding porcelain and Jun stoneware), porcelain stone from south China (qingbai and underglaze blue porcelain and Longquan stoneware), and stoneware clays from south China (Yue and Guan stonewares). Therefore, SEM examination of HF-etched, polished sections of the bodies of high-refractory ceramics has considerable potential for investigating the raw materials used in their production. KEYWORDS: PORCELAIN, STONEWARE, MICROSTRUCTURE, MULLITE, SCANNING ELECTRON MICROSCOPY, ENERGY-DISPERSIVE SPECTROMETRY, HF ETCHING, X-RAY DIFFRACTION, CHINA, SONG, YUAN, MING

INTRODUCTION

Stoneware, a dense hard, glazed ceramic with a low porosity and fired to high temperatures (typically in excess of 1200°C), was first produced in China in the early Bronze Age Shang Dynasty (i.e., mid-second millennium bc). Towards the end of the first millennium ad, stoneware technology had attained a high level, and the production of porcelain, a white translucent variety, and high-quality green-glazed stoneware, particularly in the southern provinces of Jiangxi, Zhejiang and Jiangsu, was well established. The social and economic factors behind this high level of ceramic achievement are complex, but inventive exploitation of the outstanding raw materials available to the potters was a key factor in their success. Furthermore, the available porcelain clays and stones are responsible for the marked variation in appearance of the main ware types and location of the kilns. *Received 26 July 2010; accepted 18 March 2011 †Corresponding author: email [email protected] © University of Oxford, 2011

38

M. S. Tite, I. C. Freestone and N. Wood

The investigation of the raw materials used in the production of Chinese porcelains and stonewares, undertaken since the pioneering studies of Sundius (1959) and Chou Jen and Li Chia-Chih (1960a,b), has clearly established that very different clays were used in north and south China. This view was reinforced by later studies (Sundius and Steger 1963; Pollard and Hatcher 1986, 1994). As proposed by Wood (2000a), this division in raw materials, which occurs across the Nanshan–Jinling divide, reflects the fact that, on the basis of plate tectonic theory, north and south China were originally separate land masses that only came together during the Triassic period, some 225–195 million years ago. In north China, porcelains and stonewares were produced using secondary (sedimentary) kaolinitic clays, often associated with coal deposits. These clays have a range of compositions but, relative to similar coal measure clays in, for example, Britain, they may have low iron and titanium contents (Wilson 2004; Ding et al. 2009). They grade from those producing buff-firing, opaque stonewares to those producing white translucent porcelains. In contrast, in south China, the clays used in the production of stoneware and porcelain originated from deeply weathered and/or hydrothermally altered acidic (i.e., high silica) volcanic and intrusive igneous rocks. These rocks are rich in mica and have often weathered deeply into a crumbly mass consisting of quartz, mica, clays and some residual feldspars. In some cases, this primary material was processed directly by levigation to produce fine siliceous stoneware clays, while in other cases the natural downwash from this same deposit was mined as a sedimentary clay. From the 10th century ad onwards, deep beds of hydrothermally altered volcanic ash, referred to as porcelain stone, were exploited, following pulverization and refining, for the production of porcelain. X-ray diffraction (XRD) analysis and scanning electron microscopy established that the porcelain stone used at the major production centre of Jingdezhen consisted of a mixture of quartz, secondary mica and sodium feldspar (albite) (Tite et al. 1984). In addition, stoneware clays with relatively high iron contents, which would have been red-firing in an oxidizing atmosphere, were sometimes used either by themselves or mixed with porcelain stone. Published data on the chemical compositions of typical examples of these different types of raw material are given in Table 1. A primary aim of this paper was to establish the relationship between the raw materials used in the production of Chinese porcelain and stoneware bodies, and the microstructures of the bodies as observed in polished sections in a scanning electron microscope (SEM). In addition, the observed microstructures, together with bulk chemical compositions, were used to extend the existing information on the raw materials themselves. A group of 18 sherds of well-known porcelains and stonewares from north and south China, selected from the collections of the British Museum, have been examined. These sherds included typical examples of the following wares: Ding (9th–14th centuries ad) (northern porcelain); qingbai (late 10th–14th centuries ad) and underglaze blue (14th century ad to modern times) (varieties of porcelain produced in Jingdezhen, the famous southern production centre); Jun (11th–14th centuries ad) (northern green stoneware); and Yue (4th–12th centuries ad), Longquan (late 12th–15th centuries ad) and Guan (12th–14th centuries ad) (southern green stonewares). EXPERIMENTAL PROCEDURES

Polished thin sections through the bodies of the sherds were prepared from slices cut perpendicular to their surfaces. The bulk chemical compositions and microstructures of the bodies were determined using a Jeol JSM 840 scanning electron microscope (SEM) fitted with an Oxford Instruments Link Systems energy-dispersive spectrometer (EDS). The SEM was operated with a 15 kV accelerating voltage and a 1 nA specimen current. Bulk chemical compositions were © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

Lingshan Lingshan Gongxian Hespido

Sanpaopeng Nankang Dayao, Longquan Yuandi, Longquan Gaojitou, Longquan Mudaikou, Longquan Hangzhou

Hangzhou

Shanglinhu Shanglinhu Shanglinhu

North China Kaolinite Kaolinite (purple clay) Kaolinite Kaolinite

South China Porcelain stone Porcelain stone Porcelain stone Porcelain stone Red clay Red clay Stoneware clay (raw)

Stoneware clay

Acid igneous rock Weathered rock Sediment derived from rock

bd, Below detection. *Clays washed and fired unless otherwise indicated.

District

Clay type*

Zhejiang Zhejiang Zhejiang

Zhejiang

Jiangxi Jiangxi Zhejiang Zhejiang Zhejiang Zhejiang Zhejiang

Hebei Hebei Henan Henan Nikulina and Taraeva (1959, 457) Nikulina and Taraeva (1959, 457) Chou Jen et al. (1973, 137) Chou Jen et al. (1973, 137) Chou Jen et al. (1973, 137) Chou Jen et al. (1973, 137) Zhou Shaohua and Chen Quanqing (1992, 370) Zhou Shaohua and Chen Quanqing (1992, 370) Wood et al. (2005, 194) Wood et al. (2005, 194) Wood et al. (2005, 194)

Guo Yanyi (1987, 16) Guo Yanyi (1987, 16) Guo Yanyi (1987, 16) Guo Yanyi (1987, 16)

Data source

78.1 77.1 76.4

72.8

76.2 79.0 73.6 74.3 64.8 63.2 77.5

55.0 53.6 53.6 64.8

SiO2

12.8 14.2 14.8

20.3

15.7 15.9 20.7 18.2 22.4 26.6 17.9

42.8 40.0 41.3 27.9

Al2O3

2.0 1.0 0.4

0.5

3.6 0.2 0.4 0.2 1.1 0.4 0.2

0.4 0.5 1.6 0.1

Na2O

Chemical compositions of raw materials (wt% normalized to 100%)

Province

Table 1

5.4 4.7 3.3

2.9

3.0 3.1 2.5 4.8 4.9 3.2 1.9

0.3 0.2 1.4 2.1

K2O

0.2 0.3 0.5

0.5

0.3 0.1 0.5 0.3 0.9 0.4 0.3

0.4 1.0 0.1 0.4

MgO

0.0 0.0 0.1

0.3

0.5 1.1 0.4 0.8 0.7 0.2 0.2

0.1 2.0 0.5 0.1

CaO

1.3 2.3 3.8

1.5

0.6 0.6 1.9 1.3 4.5 5.2 1.1

0.2 0.7 0.5 2.5

Fe2O3

0.2 0.3 0.6

1.2

0.02 bd bd bd 0.5 0.9 1.0

0.7 2.0 1.0 1.9

TiO2

0.02 0.04 0.18

bd

bd bd 0.03 0.05 0.10 bd bd

bd bd 0.01 0.01

MnO

Raw materials used in the production of Chinese porcelain and stoneware bodies 39

© University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

40


determined by rastering the electron beam across areas of the sections prior to partial etching of the polished sections. The areas analysed were greater than 300 mm across, with at least three individual analyses being averaged for each sherd. The spectrometer was calibrated with pure element, oxide and mineral standards. Estimates of accuracy and precision were based upon the analysis of over 10 synthetic glass standards. The relative errors were better than 2% for silica, 5–10% for major and minor elements, and increased for elements present at less than 1 wt%, as the detection limits of about 0.1 wt% were approached. For the determination of the body microstructures, a part of each polished section was etched with 2% hydrofluoric acid (HF) for 3 min, and was examined in the SEM using the backscattered electron (BSE) detector. In BSE mode, the different phases present in the bodies were distinguished through a combination of the topography resulting from the differential etching of crystalline and glass phases, and the atomic number contrast of the phases. A representative selection of these phases was individually identified using EDS. It was observed that the ceramics typically comprised domains which contained varying amounts of glass, mullite or quartz, and these domains are interpreted as representing original mineral phases that have broken down or reacted at high temperature. The duration of firing was insufficient to allow homogenization of the microstructure by interdiffusion, so that elements of the arrangement of phases in the original raw material were effectively ‘frozen in’. In order to provide an indication of the original raw material phases from which the observed relict phases were derived, domain areas of about 20 mm across were analysed, and their SiO2/Al2O3 wt% concentration ratios were determined. XRD analyses were undertaken on powdered bulk body samples using Debye–Scherrer powder cameras in order to determine the mineral phases present. The resulting films were measured using a Joyce–Loebl computer-driven microdensitometer in order to determine the intensities of the principal XRD lines for mullite (5.39 Å) and quartz (3.34 Å). The ratios of the line intensities provide a good measure of relative amounts of mullite and quartz, and so indicate which ceramics have high and low mullite relative to quartz. Although the XRD line intensity ratios do not translate into absolute mullite/quartz concentration ratios, they can be usefully compared within the sample group. RESULTS AND DISCUSSION

The bulk chemical compositions of the porcelain and stoneware bodies are given in Table 2, together with the ratios of the intensities of the principal XRD lines for mullite and quartz. Also included in Table 2 are representative published compositional data for the bodies of the different categories of ceramic, as determined principally using atomic absorption spectrometry or, in the earlier Chinese papers, wet chemical methods. The first point to note is that the bulk chemical compositions of the current samples are consistent with the previously published compositional data, with the oxide percentages determined for the current samples falling within or close to the corresponding compositional ranges for the previously published ceramics. The chemical compositions of the ceramic bodies indicate that they consist of about 95 wt% or more of silica, alumina, soda plus potash. With some simplifying assumptions regarding the possible mineral phases present in the ceramic bodies prior to firing, it is possible to estimate their original mineralogical compositions, within certain limits. The estimation is based upon a number of assumptions which are supported by theoretical and empirical considerations. First, on the basis of phase equilibrium considerations, we note that at the low temperatures at which the plutonic and hydrothermally altered volcanic source rocks are likely to have equilibrated, there is a substantial solvus or miscibility gap between sodic and potassic feldspars © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

qingbai qingbai qingbai qingbai (3)

UG blue UG blue UG blue UG blue (3)

Longquan Longquan Longquan Longquan (5)

Yue Yue Yue Yue (12)

Guan Guan Guan (6)

1938-04-12,21 1938-04-12,23 1940-04-13,28 Wood (1986, 262)

1921-03-01, 1926-06-18,35 1926-06-18,36 Chen Yaocheng et al. (1986, 123)

HD1 OA896 1930-06-17,1 Chou Jen et al. (1973, 135)

1938-04-12.7 1935-10-19, 1J 1938-04-12.5 Pollard and Hatcher (1994)

HK1 HD2 Li Jiazhi et al. (1999, 177)

Al2O3

Na2O

K2O

MgO

CaO

FeO

64.10 64.62 64.8

27.98 27.93 30.0

0.66 0.54 0.5

1.51 1.89 1.6

0.61 0.66 0.3

1.52 0.69 0.4

2.26 2.59 1.7

63.84 29.70 0.37 1.61 0.96 1.41 0.85 64.27 28.22 0.72 1.99 1.40 2.03 0.84 61–68 25–33 0.2–0.6 1.1–3.4 0.5–1.2 1.0–3.0 0.6–1.3

SiO2

1.25 1.06 1.2

1.20 0.52 0.2–1.2

TiO2

bd 0.20 0.10 0.20

72.01 20.14 0.36 5.40 0.20 69.32 22.38 0.48 5.21 0.32 71.05 20.28 0.64 5.99 0.26 68–71 19–24 0.3–1.0 4.8–6.0 0.1–0.7

74.26 19.69 2.30 2.30 71.86 20.59 1.60 4.00 71.06 21.29 1.90 3.90 74–75 19–20 2.3–2.4 2.7–3.0 bd 0.07 bd bd–0.2

1.63 1.95 1.55 1.6–2.4

0.20 0.90 0.10 1.40 0.10 1.40 0.1–0.9 0.2–1.2

66.94 24.93 0.44 3.86 0.39 0.12 1.98 65.07 26.52 0.47 3.34 0.44 0.18 2.70 67–68 22–25 0.3–0.7 2.3–3.6 0.1–0.4 0.1–0.3 2.0–3.9

1.27 1.25 1.0–1.2

0.90 0.76 0.78 0.6–0.8

0.08 0.18 0.11 0.1–0.2

0.05 0.05 0.05 bd–0.2

na, Not analysed; bd, below detection. *Previously published chemical composition ranges are shown in italics; the number of sherds analysed is given in brackets in the ‘Type’ column. †Ratio of intensities of the principal XRD lines for mullite (5.39 Å) and quartz (3.34 Å).

S. Song S. Song S. Song

Six Dynasties 76.57 16.10 0.95 3.03 0.53 0.25 1.64 Tang 76.92 15.40 0.98 2.86 0.64 0.26 2.13 Liao 77.05 16.08 0.83 2.29 0.61 0.29 2.02 Tang 71–78 15–21 0.8–1.0 2.6–3.1 0.4–0.6 0.1–0.6 1.6–2.2

S. Song Yuan Ming S. Song-Yuan

Yuan Yuan Yuan Yuan

S. Song-Yuan 78.66 15.69 1.10 2.90 0.10 0.40 0.80 0.05 S. Song-Yuan 78.06 15.89 0.80 3.30 0.20 1.00 0.50 0.05 S. Song-Yuan 79.16 16.09 0.50 2.40 0.30 0.70 0.50 0.05 S. Song 77–79 16–17 0.7–1.1 3.0–3.3 0.1–0.3 0.4–1.4 0.5–0.8 0.06–0.08

N. Song N. Song N. Song

Jun Jun Jun (1)

PD36 PD21 Yang Wenxian and Wang Yuxi (1986, 209)

Dynasty

N. Song N. Song N. Song

Type

Chemical compositions of Chinese porcelain and stoneware bodies (wt% normalized to 100%)

OA852A Ding OA852B Ding Pollard and Hatcher (1986) Ding (8)

BM registration numbers*

Table 2

0.05 bd bd

bd 0.05 0.05 0.01–0.02

0.11 0.09 0.12 0.03–0.1

na na na bd

na na na na

0.11 bd na

0.05 bd 0.01–0.03

MnO

bd bd 0–1–0.3

bd bd bd na

0.05 bd bd na

0.30 0.20 0.20 na

0.30 0.20 0.30 na

bd bd na

bd bd na

P2O5

1.0

0.2 0.4

0.4 0.6

0.5

0.3

0.2

0.2

2.9

2.7

XRD Mull/Qu†



42


(e.g., Bowen and Tuttle 1950). Hence feldspars are likely to be either mainly sodic or mainly potassic, and intermediate compositions are unlikely. This inference is supported by the substantial compilation of naturally occurring feldspar compositions by Trevena and Nash (1981), which indicates that for plutonic rocks, albite (NaAlSi3O8) substitution in orthoclase (KAlSi3O8) is limited, typically less than 25%, while the content of orthoclase in albite is typically less than 10%. The limited mutual solubility of potassic and sodic feldpars in plutonic rocks has been used to investigate the provenance of feldspar particles in fossil sands (Trevena and Nash 1981) and in provenance investigations of feldspar-tempered pottery (Freestone 1982; Freestone and Middleton 1987). Furthermore, it was confirmed for porcelain stone from Jingdezhen by Tite et al. (1984), where it was shown that the potash content of albite was very low. We also know that sodium substitution in other phases in the raw materials—quartz, kaolinite and mica or hydromica—is low. Therefore, the first step in our estimation of the mineralogy of the ceramic raw material is to assign all sodium to albite (NaAlSi3O8). It is then assumed that the potassium was associated with either orthoclase or muscovite mica. On balance, it is likely that the potassium in the hydrothermally altered raw materials found in south China occurs as mica, rather than orthoclase, but we present estimates for both possibilities. Although the mica present in the porcelain stone used in south China is secondary mica in the form of sericite (which corresponds more closely to hydromica or illite) (Vogt 1900; Kerr and Wood 2004, 216), the formula for primary muscovite (KAl3Si3O10(OH)2) is, for simplicity, used in the following calculations. It is then assumed that any remaining alumina was associated with the clay mineral, kaolinite (Al2Si2O5(OH)4), and that the remaining silica was in the form of quartz. The resulting estimated percentage weights of original albite, orthoclase, muscovite, kaolinite and quartz are given in Table 3, assuming (a) that all the potash was associated with orthoclase, or (b) that 50 wt% of the potash was associated with orthoclase and 50 wt% with muscovite, or (c) that all the potash was associated with muscovite. In estimating the original kaolinite/quartz wt% concentration ratios, also given in Table 3 and subsequently referred to in the text, the choice between these three options was based on the most likely original mineralogy, as inferred from a combination of bulk composition and observed microstructure. It is noted that the estimated albite contents of the bodies are below 10% in all cases except for the underglaze blue wares, in which a distinctive variety of porcelain stone is likely to have been used, as discussed below. As hydromica contains somewhat less potash than muscovite, our estimate for the amount of muscovite is likely to be a minimum, and of kaolinite a maximum. In addition to a scatter of surviving, partially reacted quartz particles, the dominant crystalline phase observed in all the HF-etched, polished sections using the SEM is mullite (Si2Al6O13), present as a mass of fine, mainly elongated crystals formed as a result of the breakdown of kaolinite, mica and feldspars during firing (Figs 1–8). The presence of mullite is consistent with the results presented by Iqbal and Lee (1999, 2000) and by Lee and Iqbal (2001), in a series of papers on the microstructural evolution of modern triaxial porcelains, produced from a mixture of kaolinite, feldspar and quartz, but without significant mica. In these papers, the authors distinguished between the formation of small, cuboidal or low aspect ratio, primary mullite crystals, typically less than 1 mm across, and larger acicular, high aspect ratio, secondary mullite crystals, typically up to 10 mm or more in length, set in a glassy matrix. The primary mullite is formed from pure clay agglomerates, whereas the secondary mullite is formed either from feldspar penetrated clay relicts or from clay–feldspar–quartz mixtures, pure feldspar relicts by themselves containing insufficient alumina to form mullite. The elongation of the secondary mullite is due to the growth of crystals within a lower viscosity liquid phase resulting from the © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

Ding Ding Ding

Jun Jun Jun

qingbai qingbai qingbai qingbai

UG blue UG blue UG blue UG blue

Longquan Longquan Longquan Longquan

Yue Yue Yue Yue

Guan Guan Guan

OA852A OA852B Pollard and Hatcher (1986)

PD36 PD21 Yang Wenxian and Wang Yuxi (1986, 209)

1938-04-12,21 1938-04-12,23 1940-04-13,28 Wood (1986, 262)

1921-03-01, 1926-06-18,35 1926-06-18,36 Chen Yaocheng et al. (1986, 123)

HD1 OA896 1930-06-17,1 Chou Jen et al. (1973, 135)

1938-04-12.7 1935-10-19, 1J 1938-04-12.5 Pollard and Hatcher (1994)

HK1 HD2 Li Jiazhi et al. (1999, 177)

69.6 68.2 71.0

79.2 80.0 80.1 78.5

73.5 71.2 72.5 72.8

75.4 73.3 72.4 75.4

80.0 79.6 80.7 79.3

68.0 68.0 66.9

66.8 67.5 65.4

SiO2

25.9 27.8 25.6

16.7 16.0 16.7 17.9

20.6 23.0 20.7 20.7

20.0 21.0 21.7 19.4

16.0 16.2 16.4 16.5

29.7 29.4 31.0

31.1 29.6 32.7

Al2O3

0.5 0.5 0.4

1.0 1.0 0.9 0.8

0.4 0.5 0.7 0.8

2.3 1.6 1.9 2.4

1.1 0.8 0.5 1.0

0.7 0.6 0.5

0.4 0.8 0.8

Na2O

4.0 3.5 3.0

3.1 3.0 2.4 2.7

5.5 5.3 6.1 5.7

2.3 4.1 4.0 2.8

2.9 3.4 2.4 3.2

1.6 2.0 1.7

1.7 2.1 1.1

K2O

4 4 3

8 9 7 7

3 4 6 7

20 14 16 20

9 7 4 9

6 5 4

3 6 7

%A

24 21 18

19 18 14 16

33 32 36 33

14 24 23 17

17 20 14 19

9 12 10

10 12 6

%O

45 50 47

25 24 28 30

30 36 28 29

30 30 31 27

24 24 28 25

58 57 62

62 57 66

%K

Orthoclase†

27 25 32

48 50 51 47

34 29 30 31

37 32 29 36

49 49 53 48

26 26 24

24 24 21

%Q

4 4 3

8 9 7 7

3 4 6 7

20 14 16 20

9 7 4 9

6 5 4

3 6 7

%A

12 10 9

9 9 7 8

16 16 18 17

7 12 12 8

9 10 7 9

5 6 5

5 6 3

%O

16 14 12

13 12 10 11

22 22 25 23

9 16 16 11

12 14 10 13

6 8 7

7 8 4

%M

36 42 40

18 17 22 23

17 23 14 15

24 21 22 20

17 16 22 17

55 53 58

58 52 63

%K

32 29 36

52 53 54 51

41 35 38 38

40 37 34 40

53 53 56 52

28 29 26

27 27 23

%Q

Orthoclase + muscovite†

4 4 3

8 9 7 7

3 4 6 7

20 14 16 20

9 7 4 9

6 5 4

3 6 7

%A

32 28 24

25 24 19 22

45 43 49 46

19 33 32 23

24 27 20 26

13 16 13

14 17 9

%M

26 34 33

11 10 16 17

4 10 –1 2

19 11 12 13

10 9 17 10

51 48 54

54 47 61

%K

37 34 39

56 57 57 54

48 42 46 45

43 42 39 43

57 57 59 56

30 31 28

29 30 24

%Q

Muscovite†

0.7 1.0 0.8

0.3 0.3 0.4 0.5

0.4–0.1 0.6–0.2 0.4–0.0 0.4–0.0

0.4 0.3 0.3 0.3

0.2 0.1 0.3 0.2

1.9 1.8 2.2

2.2 1.9 2.8

%K/%Q‡

*Previously published chemical compositions (averages) are shown in italics. †%A (albite), %O (orthoclase), %M (muscovite), %K (kaolinite) and %Q (quartz) calculated assuming: (a) all K2O associated with orthoclase (‘Orthoclase’); (b) 50 wt% K2O associated with orthoclase—50 wt% with muscovite (‘Orthoclase + muscovite’); (c) all K2O associated with muscovite (‘Muscovite’). ‡Kaolinite/quartz wt% concentration ratios calculated from data shown in bold.

Type

Estimates of original mineral phases in Chinese porcelain and stoneware bodies (wt% normalized to 100%)

BM registration numbers*

Table 3



44


Figure 1 A SEM photomicrograph of a Ding porcelain body (OA852A) showing sparse quartz particles (dark grey) in a more or less continuous matrix containing a dense mat of fine, randomly oriented, elongated mullite crystals. Also present are a scatter of indurated clay particles (CP) that did not disaggregate during clay preparation, and in which equant primary mullite has formed.

incorporation of alkali-rich feldspars, the extent of the elongation increasing with increasing alkali content, and increasing firing temperature and time. An indication of the phases from which the mullite was formed is provided by the SiO2/Al2O3 wt% concentration ratios of the domains in the ceramic (Fig. 9). Thus, for pure kaolinite and muscovite mica, the SiO2/Al2O3 wt% concentration ratios would be 1.2:1; and for albite and orthoclase feldspars, they would be 3.5:1. Kaolinite or mica mixed with or penetrated by feldspar would have ratios greater than 1.2:1, as would illitic and montmorillonitic clays. As a result, the observed concentration ratios in those areas containing extensive mullite, which were original dominated by kaolinite or mica, can extend up to about 2.5:1. Similarly, ratios greater than 3.5:1 would result from feldspars mixed with increasing amounts of quartz. From Figure 10, it can be seen that the ratios of the intensities of the principal XRD lines for mullite and quartz increase more or less exponentially with increasing bulk Al2O3/SiO2 wt% concentration ratios of the ceramic. This result is consistent with the fact that the overall amount of mullite formed depends on the relative proportions of kaolinite and muscovite mica, which have high alumina contents, as compared to feldspar, which has a lower alumina content, and quartz, from which alumina is absent. Ding porcelain The bodies of the Ding porcelain, which were produced using a secondary sedimentary kaolin (Li Guozhen and Guo Yenyi 1986), consist of sparse quartz in an essentially continuous matrix © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

Raw materials used in the production of Chinese porcelain and stoneware bodies

45

Figure 2 A SEM photomicrograph of a Jun stoneware body (PD36) showing sparse quartz particles (dark grey) in a fragmented matrix, with extended areas containing dense mats of fine, randomly oriented, elongated mullite crystals that are separated by a combination of quartz particles and pores (dark grey and black, respectively).

that contains a dense mat of fine, randomly oriented, elongated secondary mullite crystals, typically up to about 10 mm in length (Fig. 1). The majority of the SiO2/Al2O3 wt% concentration ratios are in the range 1.1–2:1, with peaks around 1.2:1 and 1.6:1 (Fig. 9 (a)), a result that is consistent with the secondary mullite being formed from kaolinitic clay. Also present are aggregates of equant primary mullite, which are likely to represent indurated particles of clay that did not disaggregate during the clay preparation process and were therefore not intimately mixed with potassium-bearing phases. The quartz particles are surrounded by solution rims of glass with high SiO2/Al2O3 wt% concentration ratios (>5:1). Assuming that the potassium was divided equally between with orthoclase and muscovite, the estimated concentration ratios for original kaolinite/quartz are in the range 1.9–2.8 (Table 3), a result that is consistent with the high ratio (2.7) observed for the XRD mullite/quartz line intensities (Table 2). The silica, potash, lime and magnesia concentrations in the bodies are higher than those for many kaolinitic clays from Hebei province, local to the Ding ware kilns (Lingshan district; Table 1). As a result, Guo Yanyi (1987) suggested that small amounts of quartz, feldspar, dolomite and calcite were added to the clay. Alternative explanations include the use of a mixture of kaolinitic clay with a refractory clay, available locally, which contains some lime and magnesia. On the presently available evidence, it is not possible to decide between these options. Jun stoneware The bodies of the Jun stoneware, which are pale grey in colour and coated with green, lavender blue or purple opalescent glazes, have similar microstructures to the Ding bodies. Thus, the © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

46


Figure 3 A SEM photomicrograph of an underglaze blue porcelain body (1921-03-01) showing common quartz particles (dark grey), clusters of ragged filaments made up of mullite crystals (mid-grey, M), and glassy pools containing a scatter of fine, randomly oriented, elongated mullite crystals (light grey, F) in an essentially continuous matrix.

Jun bodies again contain sparse quartz, and areas with a dense mat of fine, randomly oriented, elongated secondary mullite crystals, typically up to about 10 mm in length (Fig. 2). Within these mullite-rich areas, the SiO2/Al2O3 wt% concentration ratios are again low (5:1). A distinctive feature of the essentially continuous matrix associated with these bodies is the presence of common clusters of ragged filaments. At higher magnification (5000¥ as compared to 500¥), these clusters are seen to consist of secondary mullite crystals (Fig. 4), the arrangement of which is comparable to the fibrous, platy structure of mica particles. On the basis of this microstructure, together with the fact that the associated SiO2/Al2O3 wt% concentration ratios are less than 1.6:1 (Fig. 9 (c)), these clusters of ragged filaments most probably represent relict mica particles. Also present are pools of glass which, at higher magnification, can be seen to contain a scatter of fine, randomly oriented, elongated secondary mullite crystals, typically less than about 2 mm in length (Fig. 5). Again, on the basis of this microstructure, together with the fact that the SiO2/Al2O3 wt% concentration ratios are in the range 3–4:1, these glassy pools most probably represent relict feldspar particles. This difference in the microstructure of the relict mica and feldspar reflects the lower proportion of melt which forms from mica relative to feldspar at a particular temperature, due to the lower concentration of potash in mica. Typically, the feldspar relicts are less abundant in the qingbai than in the underglaze blue porcelain, while quartz is more abundant in the qingbai porcelain. This is not surprising, as the © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

48


Figure 5 A SEM photomicrograph of an underglaze blue porcelain body (1926-06-18,35) showing a scatter of fine, randomly oriented, elongated mullite crystals within a pool of glass formed from a feldspar particle.

estimated total feldspar contents of the underglaze blue bodies are significantly higher due to their high content of soda, which is thought to have been derived from albite (Table 3). Assuming that all the potassium was associated with muscovite, as indicated by the examination of modern porcelain stone from the region of Jingdezhen (Tite et al. 1984), the estimated concentration ratios for original kaolinite/quartz and the ratio for the XRD mullite/quartz line intensities are both lower for the qingbai porcelain (0.1–0.3 and 0.2, respectively) than for the underglaze blue porcelain (0.3–0.4 and 0.3–0.5, respectively). Also, for both porcelains, the ratios are significantly lower than those for Ding porcelain and Jun stoneware, with the lower overall mullite content reflecting the fact that much less alumina was available for its formation. It can be argued that the higher soda and alumina contents of the Yuan underglaze blue bodies, as compared to the Song–Yuan qingbai bodies, reflects the switch from a kaolinized porcelain stone (e.g., Nankang district; Table 1) to the use of a high-albite porcelain stone (e.g., Sanbaopeng district; Table 1) to which kaolinitic clay was added (Wood 1983, 2000b; Tite et al. 1984; Kerr and Wood 2004, 229). Although this hypothesis is not as yet fully proven, the use of a high-albite porcelain stone would certainly have increased the fusibility of the body and thus, facilitated its firing. Longquan stoneware The Longquan stonewares, which are generally referred to as celadon wares, are pale grey and coated with a thick green or bluish-green glaze. Their bodies have similar microstructures to those of the Jingdezhen porcelain. Again, the essentially continuous matrix contains common © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55


49

Figure 6 A SEM photomicrograph of a Longquan celadon ware body (1930-06-17,1) showing common quartz particles (dark grey), poorly defined clusters of ragged filaments made up of mullite crystals (mid-grey, M), and glassy pools containing a scatter of fine, randomly oriented, elongated mullite crystals (light grey, F) in an essentially continuous matrix.

quartz with solution rims, clusters of ragged filaments of mullite crystals, associated with mica relicts, and glassy pools containing a scatter of fine, randomly oriented, elongated mullite crystals, associated with feldspar relicts (Fig. 6). The clusters of ragged filaments again have SiO2/Al2O3 wt% concentration ratios less than 2.0:1, and the glassy pools have ratios in the range 3–4:1 (Fig. 9 (d)). However, the clusters of ragged filaments are less clearly defined than in the case of the qingbai and underglaze blue porcelains. The estimated concentration ratios for original kaolinite/quartz vary from less than 0.1 to 0.6, depending on whether the potassium was divided equally between orthoclase and muscovite or was all associated with muscovite (Table 3). The XRD mullite/quartz line intensity ratios are in the range from 0.4 to 0.6 (Table 2) and, thus, are closer to the ratios for the underglaze blue than those for the qingbai porcelain. Porcelain stone was again most probably used in the production of Longquan celadon wares, and as discussed by Kerr and Wood (2004, 252), porcelain stones from the Longquan region have the lower soda contents and higher iron oxide contents (Dayao and Yuandi, Longquan districts; Table 1) necessary to achieve a match with the compositions of the Longquan bodies. However, taking into account the extended production period and extended production area for Longquan celadon wares, Chou Jen et al. (1973) (see also Proctor 1977) have argued that, in order to achieve the 2 wt% and above of iron oxide often observed in Longquan bodies, it would have been necessary to add between 10 and 30 wt % of local high-iron clay (Gaojitou and Mudaikou, Longquan districts; Table 1), which would have been red-firing in an oxidizing atmosphere, to the porcelain stone. © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

50


Figure 7 A SEM photomicrograph of a Yue stoneware body (1938-04-12,7) showing common quartz particles (dark grey) in an essentially continuous matrix, with extended areas containing varying densities of fine, randomly oriented, elongated mullite crystals. Those areas that appear to have a lower density of mullite crystal are marked F.

Yue stoneware The bodies of the Yue stoneware, which are pale grey and coated with a green-grey glaze, contain fairly abundant quartz in an essentially continuous matrix, with extended areas containing varying densities of fine, randomly oriented, elongated secondary mullite crystals, typically up to about 10 mm across (Fig. 7). Those areas with a lower density of mullite crystals (marked F in Fig. 7), which are typically up to about 50 mm across, are probably associated with pools of glass derived from the larger melted feldspar crystals, whereas those areas with a higher density of mullite crystals are associated with more clay-rich clay–feldspar mixtures. The SiO2/Al2O3 wt% concentration ratios for the matrix, which are all greater than about 2:1, with a high proportion in the 2.8–4:1 range (Fig. 9 (e)), provide confirmation that the original clays contained feldspar particles that melted, and to varying extents, penetrated the clays during firing. The solution rings around the quartz, although present, are less well developed than in the case of the Ding porcelain. Assuming, on the basis of the absence of obvious relict mica particles, that at least half the potassium was associated with orthoclase, the estimated concentration ratios for original kaolinite/quartz vary from 0.3 to 0.5 (Table 3), and the ratio for the XRD mullite/ quartz line intensities is in the range from 0.2 to 0.4 (Table 2). Thus, both ratios are again significantly lower than the corresponding ratios for either the Ding porcelain or the Jun stoneware. These observed microstructures reflect the use of clays produced from weathered acidic, high-silica rocks (Shanglinhu and Hangzhou; Table 1), and prepared by crushing and levigation. © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55


51

Figure 8 A SEM photomicrograph of a Guan stoneware body (HK1) showing common quartz particles (dark grey) and poorly defined clusters of ragged filaments made up of mullite crystals (M) in an essentially continuous matrix.

As discussed by Kerr and Wood (2004, 140) and by Wood et al. (2005), the effect of this processing is to remove the coarser quartz and feldspars, with a resulting increase in the clay contents, and in fine-grained iron and titanium minerals. Guan stoneware Guan stonewares, which appear to have been a southern development of Northern Song imperial Ru ware, have dark thin bodies with thick, crackled bluish glazes. The bodies again contain fairly abundant quartz in an essentially continuous matrix, in which there is evidence of poorly defined clusters of ragged filaments of mullite crystals, associated with mica relicts (Fig. 8). However, in this case, glassy pools associated with feldspar relicts appear to be absent, which is consistent with the fact that very few analysed body phases have SiO2/Al2O3 wt% concentration ratios greater than 2.0:1 (Fig. 9 (f)). The quartz particles lack solution rims, suggesting that the firing temperatures were lower than those used in the production of the Longquan celadon ware. Assuming, on the basis of the absence of obvious relict feldspar particles, that all the potassium was associated with muscovite, the estimated concentration ratios for original kaolinite/quartz are in the range 0.7–1.0 (Table 3), and the ratio for the XRD mullite/quartz line intensities is 1.0 (Table 2), both ratios being higher than those for all the other porcelains and stonewares except for Ding and Jun. These observed bulk compositions and microstructures are consistent with the proposal by Kerr and Wood (2004, 261) that the Guan bodies were produced from processed high-iron primary clays that would have been red-firing in an oxidizing atmosphere, and not dissimilar to the Gaojitou and Mudaikou clays of the Longquan districts. © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

52


Figure 9 Histograms showing the number of analysed phases present in the bodies of (a) Ding, (b) Jun, (c) qingbai plus underglaze blue, (d) Longquan, (e) Yue and (f) Guan porcelains and stonewares associated with the different SiO2/Al2O3 wt% concentration ratios.

CONCLUSIONS

The above results show that the SEM examination of HF-etched, polished sections of ancient ceramic bodies, taken together with their bulk chemical compositions and XRD analyses, provide a powerful tool for identifying relict phases in the bodies, and thus helping to distinguish between the different raw materials used in their production. Of particular note is the fact that, in the case of the porcelain stone used in ceramic production in southern China, secondary mullite is formed from the partial melting of mica (fluxed by the structural potassium) in addition to that formed from © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

XRD line intensity ratio - mullite/quartz


53

J D

2.5 2 1.5 1

G

0.5

Y Q

0 0.00

0.10

U Q

Y

0.20

U

L L 0.30

0.40

0.50

Bulk Al2O3/SiO2 wt% concentration ratio Figure 10 A plot of the ratios of the intensities of the principal XRD lines for mullite and quartz versus the bulk Al2O3/SiO2 wt% concentration ratios for the porcelains (D, Ding; Q, qingbai; U, underglaze blue) and stonewares (J, Jun; L, Longquan; Y, Yue; G, Guan).

feldspar and clay–feldspar mixtures, which has been previously observed in triaxial porcelains (Iqbal and Lee 1999, 2000; Lee and Iqbal 2001). This mica-based mullite is characterized by clusters of ragged filaments that reflect the fibrous, platy structure of the original mica particles. Thus, in northern Ding and Jun wares produced from kaolinitic clays, the microstructures are characterized by dense mats of fine, randomly oriented, elongated secondary mullite crystals. In contrast, in southern qingbai, underglaze blue and Longquan wares produced from porcelain stone, the microstructures are characterized by a combination of clusters of ragged filaments of secondary mullite crystals associated with relict mica particles, and glassy pools containing a scatter of fine, randomly oriented, elongated secondary mullite crystals associated with relict feldspar particles. These differences are reflected in the higher bulk alumina contents and higher XRD mullite/quartz line intensity ratios associated with the kaolinitic northern wares as compared to the porcelain stone based southern wares. Further, the observed microstructures show that white porcelain stone was not used in the production of either Yue or Guan stonewares from southern China. Instead, the clay used in the production of the Yue stoneware contained considerable feldspar but more limited mica, whereas the clay used in the production of Guan stoneware was very different, containing considerable mica but only limited feldspar. Again, the differences are reflected in the higher bulk alumina contents and higher XRD mullite/quartz line intensity ratios associated with the Guan stoneware as compared to the Yue stoneware. In summary, as also recently shown by Martinón-Torres et al. (2008) in a study of mullite crucibles from medieval Europe, the SEM examination of HF-etched polished sections has considerable potential for the future investigation of the raw materials used in the production of ancient porcelains, stonewares and other high-refractory ceramics from across the world. Also, for the future, the methodology could be extended to include Rietveld XRD measurements to determine the quantitative phase compositions, and the investigation of the thermal history of the ceramics by means of refiring experiments to establish the evolution of the microstructures with increasing firing temperature. © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55

54

M. S. Tite, I. C. Freestone and N. Wood ACKNOWLEDGEMENTS

Two of us (MST and ICF) were members of staff of the British Museum when this work was initiated. We thank Derek Gillman and Jessica Rawson, then of the Department of Oriental Antiquities (now the Department of Asia), for advice in the selection of the sherds for analysis. REFERENCES Bowen, N. L., and Tuttle, O. F., 1950, The system NaAlSi3O8–KAlSi3O8–H2O, Journal of Geology, 58, 489–511. Chen Yaocheng, Guo Yanyi, and Zhang Zhigang, 1986, A study of Yuan blue and white porcelain, in Scientific insights on ancient Chinese pottery and porcelain (eds. Shanghai Institute of Ceramics), 122–8, Science Press, Peking. , An investigation on the technical Chou Jen, and Li Chia-Chih, 1960a, aspects of Chinese ancient ceramics, Wenwu, 1, 89–104. , A study of Ching-te-chen Chou Jen, and Li Chia-Chih, 1960b, bodies, glazes and production technology in successive dynasties, Guisuanyan Xuebao, 4.2, 49–63. , Chou Jen, Chang Fu-K’ang, and Chêng Yung-Fu, 1973, Technical studies on the Lungchuan celadons of successive dynasties, K’ao-ku Hseh-pao, 1, 131–56 (see also Proctor 1977). Ding Shu-lia, Liu Qin-fub, and Wang Ming-zhen, 2009, Study of kaolinite rock in coal bearing stratum, North China, The 6th International Conference on Mining Science & Technology, Procedia Earth and Planetary Science, 1, 1024–8. Freestone, I. C., 1982, Applications and potential of electron probe micro-analysis in technological and provenance investigations of ancient ceramics, Archaeometry, 24, 99–116. Freestone, I. C., and Middleton, A. P., 1987, Mineralogical applications of the analytical SEM in archaeology, Mineralogical Magazine, 51, 21–31. Guo Yanyi, 1987, The raw materials for making porcelain and the characteristics of porcelain wares in north and south China in ancient times, Archaeometry, 29, 3–19. Iqbal, Y., and Lee, W. E., 1999, Fired porcelain microstructures revisited, Journal of the American Ceramic Society, 82, 3584–90. Iqbal, Y., and Lee, W. E., 2000, Microstructural evolution in triaxial porcelain, Journal of the American Ceramic Society, 83, 3121–7. Kerr, R., and Wood, N., 2004, Joseph Needham: science and civilisation in China, vol. 5, part 12, Ceramic technology, Cambridge University Press, Cambridge. Lee, W. E., and Iqbal, Y., 2001, Influence of mixing on mullite formation in porcelain, Journal of the European Ceramic Society, 21, 2583–6. Li Guozhen, and Guo Yenyi, 1986, An investigation of Ding white porcelains of successive dynasties, in Scientific insights on ancient Chinese pottery and porcelain (eds. Shanghai Institute of Ceramics), 134–40, Science Press, Peking. Li Jiazhi, Zhang Zhigang, Chen Shiping, and Deng Zequn, 1999, , Study on the science and technology of the sherds collected in Hangzhou Wansongling, in Science and technology of ancient ceramics 4: Proceedings of the International Symposium (ISAC ’99) (ed. Guo Jingkun), 173–82, Shanghai Research Society of Science and Technology of Ancient Ceramics, Shanghai. Martinón-Torres, M., Freestone, I. C., Hunt, A., and Rehren, Th., 2008, Mass-produced mullite crucibles in medieval Europe: manufacture and material properties, Journal of the American Ceramic Society, 91, 2071–4. Nikulina, L. N., and Taraeva, T. I., 1959, The petrographic features of Chinese porcelain stone, Stecklo I Keramika (Moscow), 18, 40–4. Pollard, A. M., and Hatcher, H., 1986, The chemical analysis of oriental ceramic body compositions: part 2—greenwares, Journal of Archaeological Science, 13, 261–87. Pollard, A. M., and Hatcher, H., 1994, The chemical analysis of oriental ceramic body compositions: part 1: wares from north China, Archaeometry, 36, 41–62. Proctor, P. (trans), 1977, Chinese translation no. 7: Technical studies on Lung-ch’üan celadons of successive dynasties by Chou Jên, Chang Fu-K’ang and Chêng Yung-fu, in Khao-ku Hsüeh-pao, 1973, no. 1, 131–56, Victoria and Albert Museum in association with Oriental Ceramic Society, London. Sundius, N., 1959, Some aspects of the technical development in the manufacture of the Chinese pottery wares of pre-Ming age, Bulletin of the Museum of Fine Art (Stockholm), 33, 107–23. © University of Oxford, 2011, Archaeometry 54, 1 (2012) 37–55


55

Sundius, N., and Steger, W., 1963, The constitution and manufacture of Chinese ceramics from Sung and earlier times, in Sung sherds, Section Two (eds. N. Palmgren, N. Sundius and W. Steger), 375–505, Almqvist and Wiksell, Stockholm. Tite, M. S., Freestone, I. C., and Bimson, M., 1984, A technological study of Chinese porcelain of the Yuan dynasty, Archaeometry, 26, 139–54. Trevena, A. S., and Nash, W. P., 1981, An electron microprobe study of detrital feldspar, Journal of Sedimentary Petrology, 51, 137–50. Vogt, G., 1900, Recherches sur les porcelaines Chinoises, Bulletin de la Société d’encouragement pour l’industrie nationale, 99, 560–612. Wilson, I. R., 2004, Kaolin and halloysite deposits of China, Clay Minerals, 39, 1–15. Wood, N., 1983, Provenance and technical studies of Far Eastern ceramics, in Trade ceramics studies 3 (ed. Kamei Meitoku), 119–44, Japan Society for the Study of Oriental Trade Ceramics, Kamakura City. Wood, N., 1986, Some implication of recent analyses of Song yingqing ware from Jingdezhen, in Scientific insights on ancient Chinese pottery and porcelain (eds. Shanghai Institute of Ceramics), 261–4, Science Press, Peking. Wood, N., 2000a, Plate tectonics and Chinese ceramics: new insights into the origins of China’s ceramic raw materials, in Taoci no.1, Revue Annuelle de la Société française d’Étude de la Céramique orientale, Actes du colloque ‘Le “Blue et Blanc” du Proche-Orient à la Chine’, 15–24, Paris. Wood, N., 2000b, The kaolin question—ceramic changes at Hutian in the Yuan Dynasty, in Taoci no. 1, Revue Annuelle de la Société française d’Étude de la Céramique orientale, Actes du colloque ‘Le “Bleu et Blanc” du Proche-Orient à la Chine’, 25–32, Paris. Wood, N., Doherty, C., and Ratelli, S., 2005, Some aspects of Yue ware production at Shanglinhu in the late Tang dynasty, in Science and technology of ancient ceramics 6: Proceedings of the International Symposium (ISAC ’05) (ed. Guo Jingkun), 185–98, Shanghai Research Society of Science and Technology of Ancient Ceramics, Shanghai. Yang Wenxian, and Wang Yuxi, 1986, A study of ancient Jun ware and modern copper red glaze from the chemical composition, in Scientific insights on ancient Chinese pottery and porcelain (eds. Shanghai Institute of Ceramics), 204–10, Science Press, Peking. Zhou Shaohua, and Chen Quanqing, 1992, Study on the raw materials of Southern Song Altar Guan Ware originated at Wu Gui Hill, Hangzhou, in Science and technology of ancient ceramics 2: Proceedings of the International Symposium (ISAC ’92) (eds. Li Jiazhi and Chen Xianqiu), 368–73, Shanghai Research Society of Science and Technology of Ancient Ceramics, Shanghai.
