Strength of kaolinite-based ceramics: Comparison between limestone- and quartz-tempered bodies

Page 1

Tc

Ia

b

a

ARRAA

KTTQLFM

1

aifvtfepo

c[ptrnm

((

h0

Thermochimica Acta 581 (2014) 100–109

Contents lists available at ScienceDirect

Thermochimica Acta

j ourna l h omepage: www.elsev ier .com/ locate / tca

he effect of temper on the thermal conductivity of traditionaleramics: Nature, percentage and granulometry

gnazio Allegrettaa,∗, Giacomo Eramoa, Daniela Pintoa, Anno Heinb

Dipartimento di Scienze della Terra e Geoambientali, Università degli Studi di Bari “Aldo Moro”, Via Orabona 4, 70125 Bari, ItalyDepartment of Materials Science, N.C.S.R. “Demokritos”, Aghia Paraskevi, 15310 Athens, Greece

r t i c l e i n f o

rticle history:eceived 31 October 2013eceived in revised form 21 February 2014ccepted 22 February 2014vailable online 2 March 2014

eywords:

a b s t r a c t

Traditional ceramics were commonly produced using a mixture of clay and temper materials, which wereadded in different percentage according to the craftsman purposes. The present study aims to examine upto which extent some technological parameters (nature, granulometry and percentage of the temper andfiring temperature) affect the thermal conductivity of traditional ceramics. With this purpose a kaoliniticclay was tempered either with quartz or limestone belonging to two different granulometric distributionsin percentage of 5%, 15% and 25%, and fired at 500, 750 and 1000 ◦C. Moreover the dependence on firing

hermal conductivityraditional ceramicsuartzimestoneiring temperatureicrostructure

temperature was studied. Thermal conductivity was measured with a modified Lee’s disks apparatus in atemperature range from 120 to 370 ◦C. It was found that quartz-tempered ceramics are more conductivethan the fired non-tempered clay, while limestone-tempered sample are less conductive. Mineralogicaland microstructural data are also provided and the influence of the ˛–ˇ quartz-phase transition on thethermal conductivity of ceramics is discussed.

© 2014 Elsevier B.V. All rights reserved.

. Introduction

Ceramics were at the base of the economic and social life ofncient society, as utilitarian and as artistic artefacts. Among util-tarian ceramics, Rice [1] distinguishes three functional categoriesor domestic pottery: storage, transport and processing. Ceramicessels for processing (or transformation) played a specific role, ashey were used as tools in various production cycles, apart fromood preparation also for example as crucibles, tuyeres, moulds,tc. To fulfil its function, an artefact should have specific physicalroperties which depend on the raw materials, the shape of thebject and the production cycle.

As to raw materials, ancient potters could use calcareous or non-alcareous clay to produce respectively high or low porous ceramics2]. They could, also, modify the natural clay by refining or tem-ering [1,3,4]. Both these actions affect the plasticity and, hence,he workability of the clay paste: the removal of coarse particles

esults in an increase in kneading plasticity, while the addition ofon-plastic materials (like quartz, limestone, grog, sand, shells, etc.)akes the clay less sticky and plastic. Moreover, temper was added

∗ Corresponding author. Tel.: +39 0805442608; fax: +39 0805442625.E-mail addresses: [email protected], [email protected]

I. Allegretta), [email protected] (G. Eramo), [email protected]. Pinto), [email protected] (A. Hein).

ttp://dx.doi.org/10.1016/j.tca.2014.02.024040-6031/© 2014 Elsevier B.V. All rights reserved.

to the clay mixture to reduce the drying time and the shrinkage withconsequent reduction of cracks and flaws in the artefact [1]. In thisway, potters obtained more durable fired products. The addition oftemper, in fact, influences the mechanical properties of ceramicsdecreasing their strength but increasing their toughness [5,6].

The fabrication of a vessel with a specific shape facilitates itsuse in a particular environment. For example, jars are more suit-able than bowls in pouring liquids, while a spherical shape is moreappropriate than a cylindrical one to produce cooking pots becausein this way a larger surface is put in contact with the flame [1].

The methods used to mould the clay or to finish the artefactsgave it certain characteristics. A technique called burnish, whichconsists in rubbing the surface of the body with a hard tool, wasemployed on the unfired object to reduce its permeability to liquids[3]. Clay lumps mixed with very few amount of water were beatenwith hammers for the production of glass crucibles in order to avoida great shrinkage and the consequent formation of cracks whichcould affect their performances.

Among all the physical properties, the one which plays adeterminant role in the heat transfer process, especially in trans-formation ceramics, is the thermal conductivity [1]. It depends onmineral phases, microstructures and porosity developed in ceram-

ics during processing and firing of raw materials [7–9]. In this sense,the choice of raw materials is a critical step in the production ofsuitable ceramics for a specific function. For example, technicalceramics like smelting crucibles, which have an internal heating

Page 2

himica Acta 581 (2014) 100–109 101

swtcooclcr

lct

cwwt[btaarmgFttbro

2

(aaltts[p1a

5dd[acowa1tctco

Fra

ctio

n c

onte

nt

(%)

I. Allegretta et al. / Thermoc

ource, should be insulating to reduce the heat loss through thealls. To do this, craftsmen created wares rich in pores or increased

he thickness of the container [10–12]. In contrast to these, techni-al ceramics which are heated from the outside, like cooking potsr glass melting crucibles, should have been rather conductive inrder to transfer as much heat as possible to their content. In thisase heat barriers like pores should be avoided using, for example,ess shrinking clay pastes. It is evident that the modification of thelay mix by tempering is the most used method for enhancing oreducing the thermal conductivity of the final product.

The two most common tempers used by ancient potters wereimestone and quartz [3,13,14]. According to firing temperature,arbonates and quartz face different phase changes which modifyhe texture of the ceramic body and, hence, its thermal properties.

While the effect of tempering on the strength and toughness oferamics was studied in depth [5,6,15,16], not the same attentionas paid to its effect on thermal conductivity. In the last years someorks concerning the effect of quartz or volcanic rock fragments on

hermal conductivity of traditional ceramic have been published17]. As to limestone-tempered ceramics, even if some works haveeen found [18], they do not consider the effect of temperature onhis kind of materials which is a crucial point [19]. Moreover nottention has been paid to the granulometry of the temper which,s it is known, could affect mechanical properties [15,20]. For theseeasons, this work analyzes the relation existing between the ther-al conductivity of the ceramic body and the nature, percentage,

ranulometry of temper used and the firing temperature employed.inally, the use of temper with a given grain size distribution, ratherhan a single grain size class, was useful to test its effect other thanhe relative amount of temper. This information could be used toetter understand the criteria used by ancient craftsman to chooseaw materials and to deepen the development of ceramic technol-gy through the centuries.

. Experimental procedure

A total of 126 test pieces were prepared using a kaolinitic claykaolinite = 58%, illite = 18%, smectite = 2%, quartz = 22%, rutile andnatase in traces) [21]. Both limestone and quartz were chosens temper. Differently from literature [17] two different granu-ometric distributions were chosen to reproduce in the best wayhe archaeological data. In fact, archaeological evidence suggestshat the temper added to a clay mixture does not belong to apecific grain size but it belongs to a granulometric distribution13,14,22–24]. With this purpose two granulometric distributionser type of temper were prepared (Fig. 1): a fine sand with a mode at25 �m (positive-skewed curve) and a coarse sand having a modet 1 mm (negative-skewed curve).

The clay was mixed with 5, 15 and 25 vol% of temper, adding vol% of water and pressing the mixture in a mould of 70 mm ofiameter and 10 mm of height. Although some procedures to hin-er lime spalling in limestone-tempered ceramic bodies are known4,25,26], in this research it was planned to process in the same wayll the test pieces, without any pre- or post-firing treatment, whichould have introduced further variables in the samples. A pressuref 25 MPa was chosen in order to reduce the number of large poreshich greatly affects thermal conductivity [7]. After 1 day of drying

t 100 ◦C, samples were fired at 500, 750 and 1000 ◦C with a rate of50 ◦C/h and a soaking time of 1 h. This soaking time was choseno be a normalizing firing condition to compare the fabric of the

eramic tests with that of similar archaeological ceramics, since inhis latter case their fabric was a combination of unknown firingonditions (i.e. T/t evolution and firing atmosphere). Thus, insteadf maximum firing temperature an equivalent firing temperature is

Fig. 1. Granulometric distributions of the temper used for the production of thesamples (the fraction content in weight percentage vs. temper grain size is plotted).

considered, which is that soaking temperature maintained for 1 h,producing a particular mineral content and microstructure [27].

The name of the samples has the following structure: temperpercentage (5, 15, 25), temper granulometry (C for coarse and Ffor fine sand), temper nature (L for limestone and Q for quartz)and firing temperature. Non-tempered samples are identified bythe abbreviation NT. All the sample characteristics and names arereported in Table 1.

One month has been waited before testing the material in orderto enable the hydration reaction of CaO in limestone-temperedspecimens. Subsequently the porosity was estimated by waterimmersion [28]. The microstructure of the ceramic tests wasobserved by scanning electron microscopy (SEM) (50XVP LEO,operated at 15 kV), while the mineralogical content was investi-gated using a X-ray powder diffractometer PANalytical X’Pert proMDS, using CuK� radiation (40 kV; 40 mA) in step scan mode (0.02◦

2�), with each step measured for 11 s. The incident beam passedthrough a 0.04 rad, Soller slit, a 1/2◦ divergence slit, a 15 mm fixedmask and a 1/2◦ fixed anti-scatter slit.

To determine thermal conductivity (k) a modified Lee’s diskapparatus was employed, a stationary method, which has proved tobe suitable for the measuring of this parameter in insulating mate-rials [17]. The setup consists of a brass disk of 70 mm of diameter,acting as heat source at a stabilized and freely chosen temperature,of the sample disk placed on this heat source and of another brassdisk (detector) of the same dimension placed on top of the sam-ple. The temperatures of the two brass disks were recorded withthermocouples connected to a data logger. Thermal conductivityof each sample was investigated in the range from 120 to 370 ◦C.The setup was left for heating up until the thermal equilibrium wasreached. On the basis of the temperature difference between thebrass disks thermal conductivity was calculated in the followingway:

k(T) = qloss(T)x

A(T1 − T2)

where x and A are respectively the thickness of the specimen andthe surface area in contact with the brass disk and T1 and T2 arethe temperatures recorded on the heating and detector disk. Theheat loss of the detector disk was determined using the followingrelation:

qloss(T) = mbrassCp,brassdT

dt

where mbrass and cp,brass are the mass and the heat capacity of thebrass disk. The mass of the brass disk was 173.9 g and its heat

Page 3

102 I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109

Table 1Characteristics of the analyzed samples. The letters Q and L refer to ceramics temper with quartz or limestone, while C and F indicate the granulometry of the temper(respectively coarse and fine). Finely, for non-tempered ceramics the prefix NT is used.

Sample Temper Temperature (◦C)

Quantity (vol. %) Nature Granulometry (mm)

NT500 500NT750 750NT1000 1000

5CQ500 5 Quartz 1.000 50015CQ500 15 Quartz 1.000 50025CQ500 25 Quartz 1.000 5005FQ500 5 Quartz 0.125 50015FQ500 15 Quartz 0.125 50025FQ500 25 Quartz 0.125 5005CL500 5 Limestone 1.000 50015CL500 15 Limestone 1.000 50025CL500 25 Limestone 1.000 5005FL500 5 Limestone 0.125 50015FL500 15 Limestone 0.125 50025FL500 25 Limestone 0.125 500

5CQ750 5 Quartz 1.000 75015CQ750 15 Quartz 1.000 75025CQ750 25 Quartz 1.000 7505FQ750 5 Quartz 0.125 75015FQ750 15 Quartz 0.125 75025FQ750 25 Quartz 0.125 7505CL750 5 Limestone 1.000 75015CL750 15 Limestone 1.000 75025CL750 25 Limestone 1.000 7505FL750 5 Limestone 0.125 75015FL750 15 Limestone 0.125 75025FL750 25 Limestone 0.125 750

5CQ1000 5 Quartz 1.000 100015CQ1000 15 Quartz 1.000 100025CQ1000 25 Quartz 1.000 10005FQ1000 5 Quartz 0.125 100015FQ1000 15 Quartz 0.125 100025FQ1000 25 Quartz 0.125 10005CL1000 5 Limestone 1.000 100015CL1000 15 Limestone 1.000 100025CL1000 25 Limestone 1.000 1000

ne

ne

ne

ctc

tcs

k

w(a

p

k

wt

p

k

5FL1000 5 Limesto15FL1000 15 Limesto25FL1000 25 Limesto

apacity was 0.38 kJ/(kg K). dT/dt was estimated as the slope of theemperature vs time curve of the brass disk when it was left forooling (Fig. 2).

Moreover, the experimental results of thermal conductivity ofhe ceramic tests are compared with computed data. The thermalonductivity of the matrix (kmat) was calculated using the equationuggested by Litovsky [29]:

mat(T) = ku.c.(T) − p1/4u.c.kpor

(1 − pu.c.)3/2

here ku.c. and pu.c. are the thermal conductivity and the porosityfrom 0 to 1) of the untempered ceramic fired at a certain temper-ture and kpor is the thermal conductivity of air.

In order to calculate the thermal conductivity (ks) of the solidusart in tempered material, Maxwell eqaution was used [19]:

s = kmat1 + 2vd(1 − kmat/kd)/(2kmat/kd + 1)

1 − vd(1 − kmat/kd)/(kmat/kd + 1)

here kd and vd are the thermal conductivity and the volume frac-ion of the disperse phase (in this case the temper).

Finally, the computed thermal conductivity of a porous tem-

ered ceramic (kcomp) was determined as follow:

comp = ks(1 − p)3/2 + p1/4kpor

0.125 10000.125 10000.125 1000

where p is the porosity of the tempered sample measured by waterimmersion. Considering the thermal conductivity of quartz andpores respectively equal to 1.7 and 0.024 W m−1 K−1 [30], a the-oretical thermal conductivity could be calculated.

This comparison has been proposed only for quartz-temperedceramics because contrary to limestone, quartz does not react inthese temperature ranges.

3. Results and discussion

One month after the firing, 15% – fine-limestone and, 25% –coarse-limestone-tempered ceramics fired at 750 ◦C, as well as allthe specimens tempered with coarse limestone fired at 1000 ◦C, hasdeteriorated due to lime spalling.

3.1. Mineralogical content

Semi-quantitative results of XRPD investigations are reportedin Table 2. They show that the mineral content of the samples isrelated to the kind of mixture (i.e. limestone or quartz temper) andthe firing temperature. The mineral content of quartz-tempered

ceramics is qualitatively the same as that of non-tempered clayfired at the same temperature (i.e. 500, 750 and 1000 ◦C). Forlimestone-tempered samples fired at 500 ◦C, the same evidence canalso be observed, with the exception of calcite added as temper. On

Page 4

I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109 103

T

(°C

)

Hea

t F

lux

in

J/s

T (°C)

exo

A B

Cooling time (s)

Fig. 2. Cooling curve of the brass disk at 25 ◦C represented as temperature in function of cooling time (A) and the heat flux corresponding to the measured temperature (B).

Table 2Semi-quantitative mineral data.

Sample Mineral phases

Kln Ilt Sme Ant Rt Qz Cal Lm Prt Gh Ca-Ol An Wo Mul Spl-type

Clay xxxx xx x tr tr xxNT500 xxx xx x tr tr xxNT750 xx x tr tr xxNT1000 tr tr xx tr xxx

5CQ500 xxx xx x tr tr xx15CQ500 xxx xx x tr tr xxx25CQ500 xxx xx x tr tr xxxx5FQ500 xxx xx x tr tr xx15FQ500 xxx xx x tr tr xxx25FQ500 xxx xx x tr tr xxxx5CL500 xxx xx x tr tr xx x15CL500 xxx xx x tr tr xx xx25CL500 xxx xx x tr tr xx xxx5FL500 xxx xx x tr tr xx x15FL500 xxx xx x tr tr xx xx25FL500 xxx xx x tr tr xx xxx

5CQ750 xx x tr tr xx15CQ750 xx x tr tr xxx25CQ750 xx x tr tr xxxx5FQ750 xx x tr tr xx15FQ750 xx x tr tr xxx25FQ750 xx x tr tr xxxx5CL750 xx x tr tr xx x15CL750 xx x tr tr xx xx25CL750 xx x tr tr xx xx x5FL750 xx x tr tr xx x tr15FL750 xx x tr tr xx x x25FL750 xx x tr tr xx xxx tr

5CQ1000 tr tr xx tr xxx15CQ1000 tr tr xxx tr xxx25CQ1000 tr tr xxxx tr xxx5FQ1000 tr tr xx tr xxx15FQ1000 tr tr xxx tr xxx25FQ1000 tr tr xxxx tr xxx5CL1000 tr tr xx tr tr xxx15CL1000 tr tr xx x x tr tr xxx25CL1000 tr tr xx xx x tr tr tr xxx5FL1000 tr tr xx tr tr tr tr xxx15FL1000 tr tr xx x tr tr tr xxx25FL1000 tr tr xx tr xx x tr tr tr tr xxx

T linite

( e (Wo

tbtoXfr

he following mineral phases are detected: quartz (Qz), anatase (Ant), illite (Ilt), kaoGh), calcio-olivine (Ca-Ol), mullite (Mul), spinel-type phase (Spl-type), wollastonit

he contrary, the mineral assemblage in the limestone-temperedodies fired at 750 and 1000 ◦C, differs from that of the non-empered clay fired at the same temperatures, as consequence

f the reaction between calcite and the clay matrix. According toRPD data the original kaolinite of the clay is completely trans-

ormed into metakaolinite [31–33] in samples fired at 750 ◦C, whileelicts of illite/mica and smectite structures are still present at this

(Kln), smectite (Sme), calcite (Cal), rutile (Rt), portlandite (Prt), lime (Lm), gehlenite). Mineral abbreviations after Whitney and Evans [41].

temperature. The principal newly formed phase in limestone-tempered samples fired at 750 ◦C is portlandite (which formsby lime hydration after firing); the major concentration of this

mineral phase was detected in the two samples which crackedafter the firing (25CL750 and 15FL750). Traces of gehleniteoccur only in the 25% – fine-limestone-tempered sample fired at750 ◦C. In limestone-tempered samples fired at 1000 ◦C, lime and

Page 5

104 I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109

k (

W/m

K)

T (°C)

k (

W/m

K)

T (°C)

k (

W/m

K)

T (°C)

k (

W/m

K)

T (°C)

k (

W/m

K)

T (°C)

.

k (

W/m

K)

T (°C)

Fine Tempered fired at 500 °CCoarse Tempered fired at 500 °C

Fine Tempered fired at 750 °CCoarse Tempered fired at 750 °C

Fine Tempered fired at 1000 °CCoarse Tempered fired at 1000 °C

F gles at bodies

gtaitums

ig. 3. Thermal conductivity of ceramic samples vs. test temperature. Circles, trianempered ceramics are represented with solid symbols while limestone-tempered

ehlenite occur in coarse- and fine-tempered ceramics, respec-ively, in addition to portlandite. All these samples contains alson amorphous phase testified by a broad hump in the backgroundn the range 15–30◦ 2�, a spinel-type phase [31,32,34–37] and

races of a weakly crystallized mullite, occurring as reaction prod-ct of the decomposition of metakaolinite [32,33,35,36,38,39]. Aore detailed discussion of the mineralogical composition of the

amples, including quantitative and microstructural data as well

nd squares respectively represent 5%, 15% and 25% – tempered ceramics. Quartz- with hollow ones. Non-tempered samples are shown with a star.

as discussions about reactions taking place on them, is presentedby Allegretta [40] and will be the subject of an upcoming paper.According to Allegretta and coauthors [40], the percentage of amor-phous phases in samples fired at 1000 ◦C ranges from 18 to 40 wt.%,

the spinel-type phase varies between 25 and 35 wt.% and mulliteranges from 4 to 8 wt.%. Spinel-type phase has been identified as a�-Al2O3 or at least a Si-poor Al-spinel [40]. Lime and portlanditerange from 0 to 10 wt.% and from 1 to 10 wt.%, respectively, in

Page 6

I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109 105

F , 750 ◦

w aroun

ctoit

pa[

3

3

oow(c(d[w

ig. 4. BSE micrographs of limestone-tempered ceramics fired at 500 ◦C (A and B)hite arrow. The magnification in macrograph F, puts in evidence the reaction rim

oarse-limestone tempered samples, from 0 to 1 wt.% and from 1o 14 wt.%, respectively, in fine-tempered samples. The percentagef gehlenite is directly related to the amount of the temper andts granulometry. It ranges from 0 to 1 wt.% in coarse-limestoneempered samples and from 1 to 4 wt.% in fine tempered.

Traces of a probable Ca-olivine phase were detected in sam-le 25CL1000 and in all fine tempered samples, whereas traces ofnorthite and wollastonite were only detected in sample 25FL100040].

.2. Thermal conductivity

.2.1. The effect of temperA first distinction about the effect produced by the addition

f either limestone or quartz could be done. In fact, the additionf quartz increases the thermal conductivity of the ceramic body,hereas the addition of limestone always produces a reduction of k

Fig. 3). Both calcite and quartz are more conductive than the firedlay [30,42] and therefore the addition of either quartz or limestone

kQz = 7.7 W m−1 K−1 and kCal = 3.6 W m−1 K−1 at 30 ◦C) should pro-uce a positive effect on the thermal conductivity of the ceramic19]. Taking into account a reduction of the thermal conductivityith the temperature increase [30], kQz and kCal are respectively 1.7

C (C and D) and 1000 ◦C (E and F). Cracks parallel to the surface are signed with ad fine limestone grains.

and 1.6 W m−1 K−1 and hence greater than the thermal conductiv-ity of the non-tempered ceramic body. This difference in thermalbehaviour between quartz and limestone-tempered ceramic couldnot be explained only on the basis of the difference in thermal con-ductivity of the two tempers. Another factor should be considered:these two thermal conductivity values are relative to crystal phasesand not to rocks. In particular, as the limestone used for the fabrica-tion is a micritic one, the presence of discontinuities in the temperitself (Fig. 4A and B) lowers the thermal conductivity of the temperand consequently the thermal conductivity of the ceramic.

As regards quartz-tempered samples, Fig. 3 shows that while insamples fired at 500 ◦C k increases with quartz content, in samplesfired at 750 and 1000 ◦C the highest thermal conductivity is reachedwith the 15% of quartz temper. At the same time, 15% – temperedceramics are less porous than 25% – tempered ones. In fact, in coarseand fine-tempered bodies fired at 750 ◦C, porosity (Table 3) movesrespectively from 44% to 45% and from 43% to 44% tempering with15% and 25% of quartz. The porosity difference between 15% and25% – tempered ceramics increases when firing temperature is set

at 1000 ◦C. In this case, using both coarse and fine quartz, porositymoves from 39% (15CQ1000 and 15FQ1000) to 42% (25CQ1000 and25FQ1000). This could be explained taking into account the ˛–ˇquartz phase change. In fact, this phenomenon occurs at 573 ◦C and

Page 7

106 I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109

Table 3Open porosity of ceramic samples determined by water immersion. For 1000 ◦C-fired bodies tempered with coarse limestone, porosity is not determined (ND) because ofceramic breaking.

Sample p � Sample p � Sample p �

NT500 43 4 NT750 47 1 NT1000 40 15CQ500 39 1 5CQ750 46 <1 5CQ1000 40 115CQ500 37 <1 15CQ750 44 <1 15CQ1000 39 125CQ500 35 4 25CQ750 45 1 25CQ1000 42 <15FQ500 41 2 5FQ750 47 <1 5FQ1000 40 <115FQ500 37 1 15FQ750 43 <1 15FQ1000 39 <125FQ500 33 <1 25FQ750 44 <1 25FQ1000 42 <15CL500 41 1 5CL750 47 2 5CL1000 ND ND15CL500 39 < 1 15CL750 45 2 15CL1000 ND ND25CL500 38 3 25CL750 41 2 25CL1000 ND ND5FL500 46 2 5FL750 49 4 5FL1000 42 3

4240

itobrfi˛iwi

cIh[

cfpcbtdtFbc

nateappt1otpatcb

tttfi

15FL500 43 3 15FL750

25FL500 44 4 25FL750

s associated with an increasing of 2% of volume [43–46]. When theemperature drops to room temperature, the phase change ˇ–˛ccurs producing a detachment zone around quartz grains as cane clearly observed in Fig. 5. These voids behave as thermal bar-iers and are closely connected with the quartz content and thering temperature. In fact, in 500 ◦C-fired samples, in which the– phase transition did not occur, the highest thermal conductiv-

ty is reached with 25% of quartz. This phase transition could explainhy in 25% – quartz tempered material the porosity is higher than

n 15% – tempered tests fired at 750 and 1000 ◦C.As shown in Fig. 6, the computed keff of 25% – quartz-tempered

eramics is equal to or minor than keff of 15% – tempered samples.n fact, even if the quartz content increases, the phenomenon ofeating conduction is influenced more by the less conductive phase19] (pores) which is higher in 25% – tempered ceramics (Table 3).

Regarding limestone-tempered samples, differences in thermalonductivity are due to the development of secondary porosity andractures in the samples. At 500 ◦C, increasing in limestone contentroduces a reduction in porosity which moves from 41% to 38% (inoarse-tempered ceramics) and from 46% to 44% (in fine-temperedodies). This reduction in porosity should lead to an increase inhermal conductivity. However, 25CL500 and 25FL500 are less con-uctive than the other limestone-tempered ceramics. In this case,he problem is the geometry of the pores. In fact, as shown inig. 4B, the porosity of 25% – tempered samples is characterizedy fractures parallel to the sample surface which affected thermalonductivity more than dispersed porosity [10,12].

At 750 ◦C the loss of some samples due to lime hydration doesot allow to analyze clearly the effect of progressive limestoneddition on thermal conductivity. However, while for ceramicsempered with fine limestone this thermal behaviour could bexplained on the base of the porosity (25FL750 are less porousnd more conductive than 5FL750), for coarse-tempered bodies theroblem is the shape of the pore. In fact, even if 5CL750 is moreorous than 15% – tempered, these last one have a porosity charac-erized by fractures perpendicular to the heat flux (Fig. 4C and D). At000 ◦C only fine tempered samples survived because of hydrationf the lime. The progressive addition of limestone decreases thehermal conductivity. This is due to the formation of big fractureerpendicular to heat flux because of the lime hydration (Fig. 4End F). A further contribution to the reduction of thermal conduc-ivity could be given by gehlenite which is less conductive thanalcite (kGh = 1.5 W/m K at 30 ◦C) and forms in limestone-temperedodies fired at 1000 ◦C (estimated from 1 to 4 wt.%) [40].

Finally, analyzing the influence of the granulometric distribu-

ion of the temper it is possible to see that 500 ◦C-fired samplesempered with fine quartz have a lower thermal conductivity thanhose tempered with coarse quartz. In quartz-tempered samplesred at higher temperature no significant differences could be

2 15FL1000 38 2 1 25FL1000 39 4

observed between the trends of thermal conductivity of the ceramicpastes with different granulometry. The comparison among theexperimental thermal conductivity value of ceramic pastes tem-pered with quartz and fired at 1000 ◦C (Fig. 6) shows that the kvalues of fine- and coarse-tempered samples converge at 15%, butinvert at 5% and 25%. Such inversion is consequence of the higherrim porosity around coarse quartz grains at 120 ◦C, which decreasesat 370 ◦C because of the thermal expansion of quartz grains.

In limestone-tempered ceramics no difference could bedetected between coarse and fine-tempered materials. At 750 ◦Cthe effect of the granulometry is negligible in comparison with theeffect due to porosity and the data could not be put in relationbecause some sample went destroyed because of the hydration ofCaO to form portlandite (Table 2). For the same reason, the effectof granulometry in limestone-tempered samples fired at 1000 ◦Ccould not be analyzed.

3.2.2. The effect of firing temperatureRegarding to the effect of firing temperature, a sensible reduc-

tion of thermal conductivity is recorded at 750 ◦C due to theincrease in porosity (as shown in Table 3 porosity goes from 40% ofNT500 to 47% of NT750) in samples fired at this temperature andthe dehydroxylation of kaolinite proved by the absence of kaolin-ite peak at d = 7 A. Only in samples 15FL750 and 25FL750 porositydecreases and this is due to the fine size of the limestone inclusions.The stresses caused by the dilatation of coarse grains are greaterthan that produced by fine ones. Such dilatation occurs during thesintering of the clay matrix, which undergoes a plastic deforma-tion when heats up and rests fixed when cools down, opening uprim pores around the limestone inclusions. Limestone inclusionshave of course the same thermal expansion coefficient, which hasa lower mismatch with the sintered matrix compared to that ofquartz [4]. Some authors [47] suggest either to finely crush or topreheat the limestone in order to counteract the formation of cracksaround the particles. Moreover, portlandite (which is the respon-sible of the destruction of 25CL750 and 15FL750) does not form in25FL750 and traces of gehlenite are detected (Table 2), which cancreate a connection between fine limestone grains and the matrix.At 1000 ◦C, k increases again, due to: (1) the reduction of the poros-ity of the body (it returns to 40%), (2) the formation of amorphousphases, and (3) the weakly crystallization of mullite and spinel-type phase. The formation of mullite combined with the decreasein porosity is the key steps for the increase in thermal conductiv-ity of ceramics. In fact, as reported in literature [9], at higher firingtemperature, k change sensibly reaching up to 3 W m−1 K−1. More-

over the spinel phase could improve the thermal conductivity of aceramic (kSpl = 9.5 W m−1 K−1 at 30 ◦C) [42].

As far as the influence of the firing temperature on the poros-ity, at 500 ◦C the porosity is due to moulding, drying and the

Page 8

I. Allegretta et al. / Thermochimica Acta 581 (2014) 100–109 107

Fig. 5. Evolution of the detachment zone around quartz grains. The zone (whitea1

stdg[ttcqtagz

A

B

C

k of coarse-tempered ceramics

k of fine-tempered ceramics

kcomp

Fig. 6. Comparison between the experimental and the computed thermal conduc-tivity (kcomp) of quartz-tempered ceramics fired at 1000 ◦C and tested at 120 (A), 220(B) and 370 ◦C (C). Both k of fine- and coarse-tempered ceramics are plotted.

rrows) is absent in samples fired at 500 ◦C (A), but it is evident at 750 (B) and000 ◦C (C).

hrinkage during firing. If on the one hand the pressing enableshe reduction of both the amount of primary pores and theirimension [7], on the other hand the addition of coarse quartzrains could increase the amount of large pores in the green body20,48]. During drying, because of the difference in water absorp-ion between quartz and clay, a stressed area could form aroundemper particles. This leads to the formation of surface-parallelracks. At 750 ◦C these stressed areas are the weakest points ofuartz-tempered fabrics, where cracks could also form because ofhe thermal expansion mismatches between quartz and the matrix

nd the ˛– quartz inversion. As it has been modelled [49] thereater the volume of non-plastic inclusion the larger the damagedone, i.e. the volume of pores. Moreover the ˇ– quartz transition

Page 9

1 himica

atsfmt

ad

3

atoioafwa

tlwatttt

erfi

satcc

1rhito–tIiicfitaitsthhifi

08 I. Allegretta et al. / Thermoc

t 573 ◦C produce fractures in quartz particles which contributeo the porosity of the system. In fact, SEM micrographs (Fig. 5)hows a detachment zone and fractures around quartz grains andractures just around them in sample fired at this 750 ◦C. Further-

ore the kaolinite–metakaolinite transformation, could contributeo the lowering of the thermal conductivity of the materials.

An overall decrease in porosity is observed in ceramics firedt 1000 ◦C, where cracks parallel to the surface disappeared, theetachment zone and cracks increase around quartz grains.

.3. Technological and functional implications

The thermal, mineralogical and microstructural results shownbove, suggest some considerations on the ceramic test charac-eristics and their functional implications. Such discussion maynly deal with the etic approach (i.e. absolute, measurable), butt is worth to note that the reasons to adopt a given materialr processing are not always technological [4,50]. In other terms,

given ceramic should have some necessary physical propertiesor a specific efficient use(s), which could be assessed and tested,hereas possible uses due to cultural learning should be proved by

n ethnographic study [51].Temper addition produces different effects on thermal conduc-

ivity: while quartz improves the conductivity of the ceramic body,imestone always lowers it. The maximum thermal conductivity

as determined for the ceramic tests tempered with 15% of quartznd fired at 750 and 1000 ◦C, while at 500 ◦C the thermal conduc-ivity increases with quartz content. Moreover coarse quartz makeshe ceramic more conductive than fine quartz while in limestone-empered bodies the grain size of the temper does not influencehe thermal conductivity.

Firing temperature affects thermal conductivity because it influ-nces the sintering of the material and hence its porosity. For thiseason samples fired at 1000 ◦C are more conductive than thosered at 750 ◦C.

Since the ceramic tests were prepared taking into account aeries of combinations between composition and firing temper-ture, reproducing some possible fabrics of traditional ceramics,hese results could be used to understand which technologicalhoices could be suitable for the preparation of ceramics for a spe-ific function.

Because of their thermal properties, ceramics fired at 750 and000 ◦C tempered with 15% of quartz are the most suitable mate-ials to be used for the production of cooking pots. In fact, theigher thermal conductivity enables the heating of pot content and

mproves the thermal shock resistance [6], which is really impor-ant in ceramic which should withstand a temperature differencef about 700 ◦C. Moreover, from a mechanical point of view, 15%

tempered materials are collocated in the range (10–20%) wherehere is the best compromise between strength and toughness [5].n this way, a pot could withstand loads and in case of stressest could break gradually without immediately crash [6]. However,f crack formation starts, the presence of temper in these per-entages improves the resistance to crack propagation [6]. Bothne and coarse quartz could be used because no difference inhermal conductivity and porosity were observed. At these temper-tures, tempering with 25% of quartz is not reasonable, because nomprovement of thermal conductivity could be seen and this is dueo the increase of porosity and the formation of cracks parallel to theurface. Moreover, these cracks weaken the ceramic [5] reducinghe maximum load withstood by the pot. Finally, this moderately

igh thermal conductivity (in comparison with that developed inigh-fired kaolinitic clay [9]) could have been very useful in keep-

ng the pot content warm once the vessel was removed from there.

Acta 581 (2014) 100–109

Quartz-tempered ceramics fired at 750 and 1000 ◦C could bealso used as storage vessel. In particular, 1000 ◦C-fired bodies aremore suitable than those fired at 750 ◦C to contain liquids becauseof their lower porosity. However, samples fired at 1000 ◦C are moresuitable than those fired at 750 ◦C for the production of storage ves-sels because their mechanical behaviour (and hence the capabilityto withstand their content) improves with firing temperature [52].

Due to their very low thermal conductivity, limestone-temperedceramics fired at 1000 ◦C could be suitable for the production ofwares where the heating source is set inside (e.g. smelting cru-cibles). In this way the heat loss is reduced because of the presenceof large cracks parallel to the surface [12], due to lime hydration,which increase with limestone content. Since no pre- or post-firingprocessing was used to control the lime spalling, the functionalityof part of the test pieces is reduced. A similar behaviour should beexpected from kiln walls. Eramo et al. [13], in fact, found that in theLate Roman site of S. Giusto (Foggia, Italy) potters used local clayswhich contains limestone for the production of kilns while theyused a finer alluvial clay, poor in calcareous sand, to make cookingpots. On the base of the results shown in this work, it could bedrawn the hypothesis that these potters could have had this kindof knowledge which enabled them to chose the correct material fordifferent purposes.

The same use is suggested for the limestone-tempered ceramicsfired at 750 ◦C but they could also be used as storage vessel becausethey do not contain lime which, in contact with water, couldhydrate and break the ceramic. Moreover, at temperature belowthat of carbonate decomposition, limestone-tempered ceramicshas a higher transverse rupture strength than those tempered withgrog or grit [16]. For this reason, they could be more suitable thanquartz-tempered materials for the production of storage and trans-port vessels.

Ceramics fired at 500 ◦C could be used as storage container evenif liquid should be avoided because of the presence of both kaolin-ite and illite at this temperature. Such 500 ◦C-fired bodies couldnot be used for processing pots in contact with a flame. In fact theflame temperature, in a small fireplace, ranges from 900 ◦C in thecontinuous flame region to 300 ◦C at the flame tips [53] and thiscould induce unwanted further microstructural and mineralogicaltransformations during pottery.

4. Conclusions

In the present work the influence of temper and firing tempera-ture on thermal conductivity of ceramic bodies has been studied. Itcould be concluded that quartz-tempered ceramics are more con-ductive than limestone-tempered ones. In particular, while addingquartz to the clay mixture increases k, the addition of limestonereduces the thermal conductivity of the non-tempered fired clay.As regards the effect of quartz temper on the thermal properties ofthe ceramic, the ˛– quartz phase transition has a key role in theheat transfer process. In fact, while in samples fired under 573 ◦Cthe higher the quartz content the higher the thermal conductiv-ity, in samples fired over this temperature the maximum value ofk is at 15% of quartz content. While the nature of the temper couldaffect sensibly the thermal behaviour of the ceramic, no significantchange could be detected changing the temper granulometry interms of thermal conductivity range, except some inversions of kvalues in quartz-tempered samples fired at 1000 ◦C.

As to firing temperature at 750 ◦C, thermal conductivitydecrease because of the great increase of porosity, but at 1000 ◦C k

is quite the same of sample fired at 500 ◦C.

Finally, from a technological point of view quartz-temperedceramics are very suitable for the production of cooking pots,though a content of non-plastic inclusions over the 15% is

Page 10

himica

utwc

A

CGFob

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[parison between quartz- and limestone-tempered ceramic bodies, in: Paper

I. Allegretta et al. / Thermoc

seless for the improvement of thermal conductivity. On the con-rary limestone-tempered ceramics could be used in environmenthere heat loss should be reduced because of their lower thermal

onductivity.

cknowledgements

The authors would like to thank Dr Saverio Fiore at the IMAA-NR of Tito Scalo (PT), who provided the kaolinitic clay. Dr. Sabrinaualtieri, Dr. Michele Dondi and Guia Guarini at the ISTEC CNR ofaenza (RA) are kindly acknowledged for the moulding and firingf the test pieces. This paper is part of a Ph.D. thesis made possibley a fellowship of Apulia Region.

eferences

[1] P.M. Rice, Pottery Analysis: A Sourcebook, first ed., The University of ChicagoPress, Chicago, 1987.

[2] Y. Maniatis, M.S. Tite, Technological examination of Neolithic-Bronze Age pot-tery from Central and Southeast Europe and from the Near East, J. Archaeol. Sci.8 (1981) 59–76.

[3] M.S. Tite, Ceramic production, provenance and use – a review, Archaeometry50 (2008) 216–231.

[4] O.S. Rye, Keeping your temper under control: materials and the manufactureof Papuan pottery, Archaeol. Phys. Anthropol. Oceania 11 (1976) 106–137.

[5] V. Kilikoglou, G. Vekinis, Y. Maniatis, P.M. Day, Mechanical performance ofquartz-tempered ceramics: Part I. Strength and toughness, Archaeometry 40(1998) 261–279.

[6] M.S. Tite, V. Kilikoglou, G. Vekinis, Strength, toughness and thermal shockresistance of ancient ceramics, and their influence on technological choice,Archaeometry 43 (2001) 301–324.

[7] J. García-Ten, M.J. Orts, A. Saburit, G. Silva, Thermal conductivity of traditionalceramics: Part I. Influence of bulk density and firing temperature, Ceram. Int.36 (2010) 1951–1959.

[8] M.L. Gualtieri, A.F. Gualtieri, S. Gagliardi, P. Ruffini, R. Ferrari, M. Hanuskova,Thermal conductivity of fired clays: effects of mineralogical and physical prop-erties of the raw materials, Appl. Clay Sci. 49 (2010) 269–275.

[9] A. Michot, D.S. Smith, S. Degot, C. Gault, Thermal conductivity and specific heatof kaolinite: evolution with thermal treatment, J. Eur. Ceram. Soc. 28 (2008)2639–2644.

10] A. Hein, V. Kilikoglou, Modeling of thermal behaviour of ancient metallurgicalceramic, J. Am. Ceram. Soc. 90 (2007) 878–884.

11] A. Hein, V. Kilikoglou, V. Kassianidou, Chemical and mineralogical examinationof metallurgical ceramics from a Late Bronze Age copper smelting site in Cyprus,J. Archaeol. Sci. 34 (2007) 141–154.

12] A. Hein, I. Karatasios, N.S. Müller, V. Kilikoglou, Heat transfer properties ofpyrotechnical ceramics used in ancient metallurgy, Thermochim. Acta 573(2013) 87–94.

13] G. Eramo, R. Laviano, I.M. Muntoni, G. Volpe, Late Roman cooking pottery fromthe Tavoliere area (Southern Italy): raw materials and technological aspects, J.Cult. Herit. 5 (2004) 157–165.

14] T. Broekmans, A. Adriaens, E. Pantos, Analytical investigations of cooking pot-tery from Tell Beydar (NE-Syria), Nucl. Instrum. Methods. Phys. Res. B 226(2004) 92–97.

15] V. Kilikoglou, G. Vekinis, Y. Maniatis, Toughening of ceramic earthenwaresby quartz inclusions: an ancient art revisited, Acta Metall. Mater. 43 (1995)2959–2965.

16] R.J. Hoard, M.J. O’Brien, M. Ghazavy Khorasgany, V.S. Gopalaratnam,A materials-science approach to understanding limestone-tempered pot-tery from the Midwestern United States, J. Archaeol. Sci. 22 (1995)823–832.

17] A. Hein, N.S. Müller, P.M. Day, V. Kilikoglou, Thermal conductivity of archae-ological ceramics: the effect of inclusions, porosity and firing temperature,Thermochim. Acta 480 (2008) 35–42.

18] J. García-Ten, M.J. Orts, A. Saburit, G. Silva, Thermal conductivity of traditionalceramics: Part II. Influence of mineralogical composition, Ceram. Int. 36 (2010)2017–2024.

19] W.D. Kingery, H.K. Bowen, D.R. Uhlmann, Introduction to Ceramics, John Wiley,New York, 1976.

20] J.L. Amorós, M.J. Orts, S. Mestre, J. Gracia-Ten, C. Feliu, Porous single-fired walltile bodies: influence of quartz particle size on tile properties, J. Eur. Ceram.Soc. 30 (2010) 17–28.

21] J. Bellanova, Ruolo dei minerali argillosi nel trasporto di agenti inquinanti inor-ganici nei sedimenti (PhD thesis), Università di Bari, Italy, 2009.

22] G. Eramo, The glass-melting crucibles of Derriere Sairoche (1699e1714 AD,Ct. Bern, Switzerland): a petrological approach, J. Archaeol. Sci. 33 (2006)440–452.

[

Acta 581 (2014) 100–109 109

23] S. Di Pierro, Matrix – temper separation of Neolithic ceramics: an experimen-tal approach to characterize the original raw materials an determine theirprovenance, in: S. Di Pierro, V. Serneels, M. Maggetti (Eds.), Ceramic in the Soci-ety. Proceedings of the 6th European Meeting on Ancient Ceramics, Fribourg,Switzerland, 3–6 October 2001, 2003, pp. 109–131.

24] G.V. Braun, Petrography as a technique for investigating Iroquoian ceramicproduction and smoking rituals, J. Archaeol. Sci. 39 (2012) 1–10.

25] A. Gibson, A. Woods, Prehistoric Pottery for the Archaeologist, first ed., Leich-ester University Press, London, 1990.

26] R.T. Laird, M. Worcester, The inhibiting of lime blowing, Trans. Br. Ceram. Soc.55 (1956) 545–563.

27] M.S. Tite, Firing temperature determinations how and why? Kungliga Vitter-hets Historie Och Antikvitets Akademien Konferenser 34 (1995) 37–42.

28] EN 993-1, Methods for dense shaped refractory products: Part 1. Determinationof bulk density, apparent porosity and true porosity. European Standards, 1995.

29] E.Y. Litovsky, M. Shapiro, Gas pressure and temperature dependences of ther-mal conductivity of porous ceramic materials: Part 1. Refractories and ceramicswith porosity below 30%, J. Am. Ceram. Soc. 75 (1992) 3425–3439.

30] C. Clauser, E. Huenges, Thermal conductivity of rocks and minerals, in: T.J.Ahrens (Ed.), Rock Physics and Phase Analysis: A Handbook of PhysicalConstants, American Geophysical Union, Washington, 1995, pp. 105–126.

31] S. Lee, Y.J. Kim, H.S. Moon, Phase transformation sequence from kaolinite tomullite investigated by an energy-filtering transmission electron microscope,J. Am. Ceram. Soc. 82 (1999) 2841–2848.

32] A.K. Chakraborty, DTA study of preheated kaolinite in the mullite formationregion, Thermochim. Acta 398 (2003) 203–209.

33] T. Watanabe, H. Shimizu, K. Nagasawa, A. Masuda, H. Saito, 29Si and 27AlMAS/NMR study of the thermal transformations of kaolinite, Clay Miner. 22(1987) 37–48.

34] G.W. Brindley, M. Nakahira, The kaolinite–mullite reaction series: I. A surveyof understanding problems, J. Am. Ceram. Soc. 42 (1959) 311–314.

35] I.W.M. Brown, K.J.D. Mackenzie, M.E. Bowden, R.H. Meinhold, Outstandingproblems in the kaolinite-mullite reaction sequence investigated by 29Si and27Al solid-state nuclear magnetic resonance: II. High-temperature transforma-tions of metakaolinite, J. Am. Ceram. Soc. 68 (1985) 298–301.

36] J. Sanz, A. Madani, J.M. Serratosa, Aluminum-27 and silicon-29 magic-angle-spinning nuclear magnetic resonance study of the kaolinite–mullitetransformation, J. Am. Ceram. Soc. 71 (1988) 418–421.

37] H. He, P. Yuan, J. Guo, J. Zhu , C. Hu, The influence of random defect density onthe thermal stability of kaolinites, J. Am. Ceram. Soc. 88 (2005) 1017–1019.

38] M. Bellotto, High temperature phase transitions in kaolinite: the influence ofdisorder and kinetics on the reaction path, Mater. Sci. Forum 166 (1994) 3–20.

39] A. Gualtieri, M. Bellotto, Modelling the structure of the metastable phases inthe reaction sequence kaolinite–mullite by X-ray scattering experiments, Phys.Chem. Miner. 25 (1998) 442–452.

40] I. Allegretta, Modeling of thermo-mechanical behaviour of pre-industrialceramics: functional analysis via an experimental and a numerical approach(PhD thesis), Università di Bari, Italy, 2014.

41] D.L. Whitney, B.V. B.V. Evans, Abbreviation for names of rock-forming minerals,Am. Mineral. 95 (2010) 185–187.

42] K. Horai, G.G. Simmons, Thermal conductivity of rock forming minerals, EarthPlanet. Sci. Lett. 6 (1969) 359–368.

43] M.J. Orts, A. Escardino, J.L. Amorós, F. Negre, Microstructural changes duringthe firing of stoneware floor tiles, Appl. Clay Sci. 8 (1993) 193–205.

44] H. Yao, I. Hatta, Phase transitions of quartz studied by AC calorimetry, Ther-mochim. Acta 266 (1995) 301–308.

45] M.J. Majumdar, H.A. McKinstry, R. Roy, Thermodynamic parameters for the�–� quartz and �–� cristobalite transaction, J. Phys. Chem. Solids 25 (1964)1487–1489.

46] P. Richet, Y. Bottinga, N. Danielou, J.P. Petitet, C. Tequi, Thermodynamicproperties of quartz, cristobalite and amorphous SiO2: drop calorimetry mea-surements between 1000 and 1800 K and a review from 0 to 2000 K, Geochim.Cosmochim. Acta 46 (1982) 2639–2658.

47] B. Velde, I.C. Druc, Archaeological Ceramic Materials: Origin and Utilization,first ed., Springer, Verlag Berlin Heidelberg, 1999.

48] J.L. Amorós, M.J. Orts, J. Garcia-Ten, A. Gozalbo, E. Sánchez, Effect of the greenporous texture on porcelain tile properties, J. Eur. Ceram. Soc. 27 (2007)2295–2301.

49] N.S. Müller, V. Kilikoglou, P.M. Day, G. Vekinis, The influence of temper shapeon the mechanical properties of archaeological ceramics, J. Eur. Ceram. Soc. 30(2010) 2457–2465.

50] M. Harris, The Rise of Anthropological Theory: A History of Theories of Culture,Altamira Press, Lanham, 2001.

51] D.E. Arnold, Ethnomineralogy of Ticul, Yucatan potters: etics and emics, Am.Antiq. 36 (1971) 20–40.

52] I. Allegretta, G. Eramo, Mechanical performances of traditional ceramics: com-

presented at 12th European Meeting on Ancient Ceramics, Padova, Italy, 2013.53] L. Audouin, G. Kolb, J.L. Torero, J.M. Most, Average centreline temperatures of a

buoyant pool fire obtained by image processing of video recordings, Fire Saf. J.24 (2) (1995) 167–187.