Author's personal copy Production technology and replication of lead antimonate yellow glass from...

This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/authorsrights

http://www.elsevier.com/authorsrights

Author's personal copy

Production technology and replication of lead antimonate yellow glassfrom New Kingdom Egypt and the Roman Empire

G. Molina a, G.P. Odin a, T. Pradell a, A.J. Shortland b, M.S. Tite c,*

aDepartament de Fisica i Enginyeria Nuclear, UPC, Campus Baix Llobregat, 08860 Castelldefels, SpainbDepartment of Materials and Medical Sciences, Center for Archaeological and Forensic Analysis, Cranfield University, Shrivenham, Wiltshire SN6 8LA, UKcResearch Laboratory for Archaeology and the History of Art, Dyson Perrins Building, South Parks Road, Oxford OX1 3QY, UK

a r t i c l e i n f o

Article history:Received 23 February 2013Received in revised form16 July 2013Accepted 25 July 2013

Keywords:Lead antimonateCalcium antimonateOpacifiersGlassPigmentAnimeNew KingdomEgyptRoman EmpireAnalytical scanning electron microscopeX-ray diffractionUVevis reflectance spectrometry

a b s t r a c t

Lead antimonate was used to produce opaque yellow glasses from the beginnings of glass production inthe Near East and Egypt around 1500 BC through into the Roman period. The composition and crys-tallographic structure of lead antimonate particles present in a small group of New Kingdom Egyptianand Roman glasses were investigated using analytical scanning electron microscopy and X-ray diffrac-tion. The results showed that the particles were of the type Pb2Sb2O7 with a cubic structure and with theantimony partially replaced by iron and zinc in the case of the Egyptian glass, and by iron and tin in thecase of the Roman glass. Synthesis in the laboratory of lead antimonate pigments, animes (i.e., leadeantimonyesilica mixtures) and yellow glasses established that New Kingdom Egyptian and Romanyellow glasses could have been produced by stirring, respectively, lead antimonate pigment or anime,containing excess lead oxide, into a molten colourless glass. It is further shown that yellow leadantimonate particles are stable in glass up to operating temperatures in the range 900e1000 �C beforeconverting to white calcium antimonate, and that their stability is enhanced by incorporating smallamount of impurities such as iron, zinc and tin. The effect of different synthesis parameters and com-positions on the colour of the glasses is investigated using UVevis reflectance spectrometry.

� 2013 Elsevier Ltd. All rights reserved.

1. Introduction

Antimony-based opacifiers (i.e., lead antimonate yellow andcalcium antimonate white) were used from the beginnings of glassproduction in the Near East and Egypt around 1500 BC through intothe Roman period (Turner and Rooksby, 1959). Towards the end ofthe Roman period (i.e., from about 4th century AD onwards), leadstannate replaced lead antimonate in the production of opaqueyellow glasses (Tite et al., 2008), and it was not until the late 15thcentury AD that lead antimonate started to be used again both inglass production in Venice (Biringuccio, 1966), and as a yellowpigment in Italian Renaissance paintings (Dik et al., 2005) and inItalian maiolica production (Tite, 2009).

As observed by Mass et al. (2002) and Shortland (2002a) forNew Kingdom Egyptian yellowglasses and byMass et al. (1998) andFreestone and Stapleton (2013) for Roman yellow glasses, a

significant proportion of the lead antimonate particles exhibit anirregular, ragged morphology and a clumped distribution, whereasother particles exhibit euhedral morphologies suggesting that theywere formed during cooling from the melt. It is therefore arguedthat preformed lead antimonate, rather than separate lead andantimony compounds, was added to the glass melt. The leadantimonate then suffered partial dissolution in the molten glass,resulting in both surviving clastic particles and euhedral particlesthat had crystallised from the melt.

In a study of early Roman coloured glasses (1st century BC to 1stcentury AD), Freestone and Stapleton (2013, Figs. 9 and 10)observed that the reduced compositions of yellow glasses (i.e.,renormalized composition after the subtraction of lead, antimonyand iron oxides) exhibited elevated silica contents together withlower lime, potash and magnesia contents relative to the reducedvalues for other colours. They therefore suggested that the yellowcolorant was added to the colourless glass in the form of a leadeantimonyesilica mixture comparable to the anime used in theproduction of yellow glasses in Venice in the 18th and 19th cen-turies AD (Moretti and Hreglich, 1984). Calculation of reduced

* Corresponding author. Tel.: þ44 (0)1865558422.E-mail address: [email protected] (M.S. Tite).

Contents lists available at ScienceDirect

Journal of Archaeological Science

journal homepage: http : / /www.elsevier .com/locate/ jas

0305-4403/$ e see front matter � 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.jas.2013.07.030

Journal of Archaeological Science 41 (2014) 171e184


compositions from analytical data for 1st to 4th century AD Romanglass published by Mass et al. (1998) again show that, as comparedto white and turquoise glasses, opaque yellow and green glasses,the latter coloured by a combination of lead antimonate and copperoxide, exhibit elevated silica contents (Fig. 1a). There is also a ten-dency for the opaque yellow and green glasses to have lower limecontents, but there is no differentiation in the potash and magnesiacontents, the majority of which are less than 1 wt%, between thefour colours. However, on the basis of the elevated silica andreduced lime contents, it can still be argued that anime continued tobe used in the production of lead antimonate glasses through to thelater Roman period. In contrast, on the basis of the analytical datafor New Kingdom glass from Egypt published by Shortland andEremin (2006), there are no significant differences in the reducedsilica, lime, potash and magnesia contents of opaque yellow glassesrelative to the other glasses (Fig. 1b). Therefore, the Egyptian yellowglass was most probably produced by adding a pigment, made byfiring a mixture of lead and antimony oxides, to raw colourlessglass.

In both the New Kingdom Egyptian and Roman yellow glasses,the PbO/Sb2O5 ratios of the bulk glass are typically in the range 5e15, and significantly higher than the ratio (1.38) for the stoichio-metric Pb2Sb2O7 particles found in the yellow glasses. Therefore,both the lead antimonate pigments and animes used respectively inthese two types of yellow glass would have contained an excess oflead. It is generally argued that excess lead facilitates the initialformation of lead antimonate, the mixing of the pigment or animeinto the colourless glass, and the subsequent stability of the leadantimonate within the glass (Shortland, 2002a; Freestone andStapleton, 2013).

In the context of Roman glasses, both Mass et al. (1998) andFreestone and Stapleton (2013) noted that the yellow glasses

normally exhibited higher iron contents that those in most othercolours, and Mass et al. (1998) further noted that the lead antim-onate crystals were themselves contaminated with iron. Morerecently, Lahlil et al. (2011) have shown that the lead antimonatecrystals in New Kingdom Egyptian yellow glasses contain up toabout 7 wt% of iron oxide and up to about 5 wt% zinc oxide, and thatthose in Roman yellow glasses again contain up to about 7 wt% ironoxide, but instead of zinc oxide, they contain up to about 20 wt% tinoxide.

In the present paper, in order to supplement and extend the datapreviously obtained by Lahlil et al. (2011), lead antimonate particlespresent in a small group of New Kingdom Egyptian yellow andgreen glasses and in a single Roman yellow glass were first inves-tigated. The compositions of the particles were determined usinganalytical scanning electron microscopy (SEM) and their crystal-lographic structure using X-ray diffraction (XRD). In an attempt toobtain information on the production procedures employed forthese glasses, a number of lead antimonate pigments and animeswere synthesised taking into account the procedures described inthe historical treatises, and were similarly analysed using a com-bination of SEM and XRD. The stability of the lead antimonateparticles in glass, and their conversion to calcium antimonate,was then investigated by firing pigments and animes togetherwith colourless glasses. Because lead antimonate based pigments,animes and glasses exhibit a wide variety of colours, specialemphasis was given to the investigation of the colour parameters,measured using UVevis reflectance spectrometry, as a function ofthe synthesis parameters. Finally, on the basis of the compositionsof the ancient glasses together with the results of the variouspigment and anime syntheses, yellow lead antimonate glasses,comparable in composition, crystallographic particle structure, andcolour to those produced in New Kingdom Egypt and in the Romanperiod, were successfully replicated.

2. Experimental procedures

2.1. Ancient glass samples

Samples from five New Kingdom yellow glasses from Malkata(prefix ‘UPP’) and Amarna (prefix ‘COP’), and one Roman yellowmosaic tessera (116a) from the early medieval workshop at SanVincenzo (Italy), all containing lead antimonate particles, werestudied. In addition, one New Kingdom green glass from Malkata(UPP14), coloured by a combination of lead antimonate and copperoxide, was also studied.

2.2. Laboratory replications

2.2.1. Lead antimonate pigmentFor the laboratory replication of the lead antimonate pigment,

Pb2Sb2O7, a stoichiometric mixture of 58 wt% lead oxide (PbO) and42 wt% antimony oxide (Sb2O5) (PG1) (Table 1) was first used. Thismixture was fired to temperatures of 800 �C or 1000 �C for 2 h.Subsequently, the replication was repeated using the stoichio-metric PbOeSb2O5 mixtures to which varying amounts of ironoxide (1e8 wt% FeO) and zinc oxide (1e2 wt% ZnO) were added(PG2-6). In this case, the mixtures were all fired to 1000 �C for 2 h.Stoichiometric mixtures, rather than those containing excess lead,were used principally in order to provide a reference material tocompare with the lead antimonates identified in the ancientglasses, to see how the lattice parameters changed with the addi-tion of iron and zinc, and to investigate the effect of compositionand firing temperature on the observed colour.

Further, both the Renaissance treatise on the potters art(Piccolpasso, 2007), and Renaissance painting treatises (Dik et al.,

60 64 68 72 76wt% SiO2*

4

6

8

10

wt%CaO*

yellow and green glassesother colours (a)

60 64 68 72wt% SiO2*

0

4

8

12

16

wt%CaO*

yellow glassesother colours (b)

Fig. 1. Plots of reduced lime content versus reduced silica content comparing (a)opaque yellow and green Roman glasses with other colours (data from Mass et al.(1998)) and (b) opaque yellow New Kingdom Egyptian glasses with other colours(data from Shortland and Eremin (2006)). (For interpretation of the references tocolour in this figure legend, the reader is referred to the web version of this article.)

G. Molina et al. / Journal of Archaeological Science 41 (2014) 171e184172


2005) indicate that an alkali flux (eg, common salt or wine lees) wasused in the production of yellow lead antimonate pigments.Therefore, replicate pigments were also produced using either apure stoichiometric mixture (PG7) or mixtures containing 1 wt%FeO (PG8) and 2 wt% ZnO (PG9) to which some 10 wt% sodiumchloride flux (NaCl) had been added. The pure mixture was fired to800 �C, 900 �C or 1000 �C for 2 h, and the iron and zinc containingmixtures to 1000 �C for 2 h.

For these and subsequent replications, a Hobersal laboratoryfurnace (HD-230) with PXR-9/4 programmer was used, and allfirings were in air. The heating rate up to the maximum tempera-ture was 5 �C/min, except in the range 90e110 �C when a slowerheating rate of 1 �C/min was used in order to eliminate humidity.The cooling rate was 2.5 �C/min down to 700 �C after which thefurnace cooled naturally. All experimental mixtures were heated inceramic crucibles, and the outer layer in contact with the cruciblewas subsequently removed in order to minimize contamination.

2.2.2. Leadeantimonyesilica animeFor the laboratory replication of the leadeantimonyesilica an-

ime, four mixtures of lead oxide, antimony oxide and silica wereprepared (AN1-4) (Table 1). PbO:Sb2O5 ratios in the range 8:1 to12:1 were chosen in order to span the ratios typically observed in

New Kingdom Egyptian and Roman yellow glasses. The amounts ofadded silica were such that, after the formation of lead antimonate,there was sufficient lead oxide remaining to produce leadesilicaglasses with PbO:SiO2 ratios equal to either 90:10 or 70:30. Thesemixtures were fired to 900 �C for 2 h.

2.2.3. Stability of lead antimonate particles in glassIn order to investigate the temperature stability of the lead

antimonate particles in glass and their conversion to calciumantimonate, powdered lead antimonate pigments and animeswerefired together with powdered colourless glasses.

In the case of lead antimonate pigments, glasses were producedby adding 12wt% of the stoichiometric pigment to a colourless glass(RGMeCEG) of composition comparable to that of New KingdomEgyptian glass (Table 2). The pigment mixtures were produced bothwithout and with the addition of 10 wt% NaCl flux (PG1 and PG7respectively) and were fired to 1000 �C for 2 h. The powderedpigment and glass mixtures were fired to temperatures of 800 �C,900 �C, 1000 �C or 1050 �C for 2 h. Similarly, in the case of leadantimonate animes, glasses were produced by adding 12 wt% of thelow silica, lower PbO/Sb2O5 ratio anime (AN1), fired to 900 �C for2 h, to a colourless glass (RGMeCRG) of composition comparable tothat of Roman glass (Table 2). The powdered anime and glassmixtures were fired to temperatures of 900 �C or 1000 �C for 2 h.

2.2.4. Lead antimonate yellow glassesIn the replication of “Egyptian type” yellow glass, a lead

antimonate mixture was prepared such as to provide a pigmentwith the approximate formula Pb2Sb1.8Fe0.25Zn0.25O7, and sufficientexcess lead oxide so that PbO/Sb2O5 ratio of the mixture matchedthe average ratio of 6.4 observed for New Kingdom Egyptian yellowglasses. The required pigment mixture whose composition is givenin Table 1 (Pigment PGeEG) was fired to 1000 �C for 2 h, and thenmixed with a colourless glass (Table 2e RGMeCEG) of compositioncomparable to that of New Kingdom Egyptian glass.

In the replication of “Roman type” yellow glass, the startingpoint for the composition of the anime mixture was the averagecomposition of the lead antimonate particles present in the Romanglass sample, 116a; that is, 56.7 wt% PbO þ 35.7 wt% Sb2O5 þ 5.1 wt% FeO þ 2.4 wt% SnO2. With the addition of silica and excess leadoxide together with slight increases in the iron and tin oxide con-tents, based on the anime compositions given by Moretti andHreglich (1984, 278), the composition chosen for the anime

Table 2Chemical compositions of New Kingdom Egyptian and Roman glasses.

SiO2 Na2O K2O CaO MgO Al2O3 FeO PbO Sb2O5 ZnO SnO2

Lead antimonate yellow glassesNew Kingdom Egyptiana Average e 19 62.0 17.0 2.1 7.2 4.1 0.7 0.6 5.2 0.8 0.3Replicate “Egyptian type”

(PGeEG þ RGMeCEG-1000 �C)SEMeEDS (�50) 58.5 14.5 2.1 7.2 7.2 1.2 0.5 6.9 1.4 0.4

Romanb Average e 6 48.4 11.6 0.5 4.7 0.4 1.8 1.4 28.5 2.6 b.dRomanc Average e 20 59.4 14.7 0.6 5.8 0.5 2.1 0.8 14.8 1.4 b.dRoman Sample 116a 67.3 18.0 0.7 6.8 0.7 2.1 0.8 2.7 1.0 b.dReplicate “Roman type”

(ANeRG þ RGMeCRG-900 �C)SEMeEDS (�50) 64.6 13.0 0.7 5.8 1.6 2.1 0.7 12.6 1.2 0.3

Colourless glassesReduced NK Egyptiana Average e 19 66.1 18.1 2.3 7.7 4.3 0.8 0.7RGMeCEGd Theoretical 65.0 18.0 2.0 9.0 4.5 1.0 0.5Reduced Romanb Average e 6 70.3 16.8 0.7 6.9 0.6 2.7 2.0Reduced Romanc Average e 20 70.8 17.5 0.7 6.9 0.6 2.5 0.9Reduced Roman Sample 116a 69.9 18.7 0.7 7.1 0.7 2.1 0.8RGMeCRGd Theoretical 71.0 16.5 1.0 8.0 0.5 2.5 0.5

a Data from Shortland and Eremin (2006).b Data from Freestone and Stapleton (2013).c Data from Mass et al. (1998).d Supplied by Roman Glassmakers.

Table 1Chemical compositions of replicate lead antimonate pigments and animes.

PbO Sb2O5 FeO ZnO SnO2 SiO2 NaCl PbO/Sb2O5

PigmentsPG1 58.0 42.0 1.38PG2 57.4 41.6 1.0 1.38PG3 55.8 40.4 3.8 1.38PG4 53.7 38.8 7.5 1.38PG5 56.9 41.1 2.0 1.38PG6 56.0 40.6 2.4 1.0 1.38PG7 52.7 38.2 9.1 1.38PG8 52.2 37.8 1.0 9.0 1.38PG9 51.8 37.5 1.8 8.9 1.38PG-EG 85.0 13.3 0.8 0.9 6.4AnimesAN1 80.9 10.1 9.0 8AN2 83.7 7.0 9.3 12AN3 64.4 8.0 27.6 8AN4 65.4 6.5 28.0 12ANeRG 75.0 10.4 3.5 1.5 9.7 7.2

G. Molina et al. / Journal of Archaeological Science 41 (2014) 171e184 173


mixture is given in Table 1 (Anime ANeRG). This animemixturewasfired to 900 �C for 2 h, and then mixed with a colourless glass(Table 2 e RGMeCRG) of composition comparable to that of Romanglass.

2.3. Analytical methods

2.3.1. Chemical compositionsThe chemical compositions of the individual lead antimonate

particles and the surrounding glass phase were determined for theancient glasses and replicate “Egyptian type” and “Roman type”glass using a Stereoscan S-360 SEM equipped with an energydispersive X-ray spectrometer (EDS), PCXA LINK EDX. The acceler-ating voltage was 20 kV and the probe current 1.5 nA. The bulkcompositions of the Egyptian glasses, as determined bywavelengthdispersive spectrometry (WDS), were reported previously byShortland and Eremin (2006), and only the average composition isgiven in Table 2. For the Roman yellow glasses, average composi-tions, determined from the EDS and WDS analyses reported byMass et al. (1998) and Freestone and Stapleton (2013), are given inTable 2, together with the bulk composition of Roman glass 116a asdetermined for areas including lead antimonate particles by EDSusing the Stereoscan S-360 SEM.

In the current study of lead antimonate, it is the detection limitsfor tin and antimony that were of particular importance. For tin, thedetection limits were about 0.2 wt% SnO2 and 0.02 at% Sn. Forantimony, when tin was also present, they were about 0.4 wt%Sb2O5 and 0.1 at% Sb, and when tin was absent, about 0.2 wt% and0.05 at% respectively. The equipment was calibrated using mineralstandards.

2.3.2. Structural analysisFor the New Kingdom Egyptian glasses, the mineralogy of in-

dividual lead antimonate particles was determined using Syn-chrotron Radiation m-XRD, performed on beamline BM16 at theESRF (Grenoble, France) in transmission geometry with a colli-mated beam with 50 mm � 50 mm spot size and monochromatic16 keV energy (l ¼ 0.78�A) X-rays on thin (about 100 mm) slices cutout of the glasses. The XRD patterns were recorded using a CCDdetector. For the Roman tessera, and the replicate lead antimonatepigments, anime and yellow glasses, bulk XRD measurements wereundertaken using a conventional diffractometer, Siemens D-500with CueKa radiation. The two theta range was 4e80�. Identifica-tion of the compounds has been performed based on the PowderDiffraction File (PDF) database from the International Centre forDiffraction Data (ICDD).

2.3.3. ColourFor Roman tessera, and the replicate lead antimonate pigments,

anime and yellow glasses, colour analysis was performed byrecording the UVevis spectra in reflection mode, using a doublebeam UVevis-NIR spectrophotometer (Shimadzu 3600) equippedwith an ISR 3100 Ulbricht integrating sphere. The spot size was aslit of 5 mm � 1 mm, and measurements were made between200 nm and 800 nm at 1 nm resolution. A D65 standard illumi-nation source was used and barium sulphate provided a whitestandard. The colour coordinates were determined following theInternational Commission for Illumination (CIE) recommendation.The colour coordinates are evaluated by CIE 1976, themethod beingbased on the evaluation of the tri-stimulus coordinates (XYZ)equivalent to the eye response to the light and corrected by theemission of a standard illuminant (D65) corresponding to theemission of the blackbody at 6500 K. The XYZ coordinates are notuniformly spaced, and for this reason the Commission establishedthe CIE Lab* standard measure to produce a homogeneously spaced

colour system. From the original XYZ colour coordinates, a new setcalled a*, b* and L* are determined, where positive a* correspondsto red, negative a* to green, positive b* to yellow and negative b* toblue, and L* stands for the lightness. From this international stan-dard system the hue (h* ¼ arctang(b*/a*)) and saturation orchroma (c*¼(a*2 þ b*2)1/2) may also be evaluated.

3. Results

3.1. New Kingdom Egyptian and Roman glasses

As previously noted by, for example, Mass et al. (2002) andShortland (2002a) for New Kingdom yellow glasses, and by Masset al. (1998) for Roman yellow glasses, significant variation incomposition with stripes of glass richer in lead, in which the con-centration of lead antimonate particles tends to be higher, wasobserved in all these glasses in polished section in the SEM (Fig. 2).The size of these lead antimonate particles is highly variable,ranging from a few mm up to about 60 mm, and the volume fractionof particles never exceeds 7% of the total volume of the sample.

Up to about five lead antimonate particles were analysed by EDSin each of the seven glasses with larger particles being selected inorder to minimise the contribution from the surrounding glass,only those analyses containing less than 5 wt% SiO2 being consid-ered. Even so, the analyses contained a few wt% of silicon and so-dium, and these elements were removed in order to obtain thecomposition of the lead antimonate particles. The average molarcompositions of the particles, normalised to two atoms of lead, aregiven in Table 3.

3.1.1. New Kingdom Egyptian glassesThe plots of at% of antimony, iron and zinc versus at% of lead for

particles in all the glasses, and for the glass surrounding the

Fig. 2. SEM photomicrographs in backscatter mode of polished sections through (a)New Kingdom yellow glass fromMalkata (UPP9) showing stripes of glass both richer inlead (lighter grey) and poorer in lead (darker grey) and (b) Roman yellow glass (116a)showing, in the central region, a concentration of lead antimonate particles (white)within a lead rich area of glass (lighter grey). Surrounding this are regions of glasspoorer in lead (darker grey) which contain far fewer lead antimonate particles.



particles of sample UPP9 are shown in Fig. 3. From Fig. 3a, it is seenthat, in all the glasses, the lead antimonate particles contain lessantimony than lead and show a correlation of about 0.93 Sb atomfor every Pb atom. In addition, as previously observed by Lahlil et al.(2011) in New Kingdom Egyptian glasses, the particles containsmall variable amounts of iron and zinc which do not show a clearcorrelation with the lead content (Fig. 3b and c). Overall, theaverage composition of the lead antimonate particles is 55.8 wt%PbO þ 37.8 wt% Sb2O5 þ 2.5 wt% FeO þ 1.7 wt% ZnO, with theamounts of FeO and ZnO varying quite a lot between particles.

From the corresponding plots for the glass phase surroundingthe lead antimonate particles in sample UPP9 (Fig. 3d, e and f), it is

seen that the antimony, iron and zinc contents show some corre-lation with the lead contents. However, the amount of antimonypresent in the glass is significantly lower than what would be ex-pected from dissolution of the lead antimonate particles, the at%Sb/at% Pb ratio being about 1:6 in the glass as compared to about1:1.1 in the particles. In contrast, although the amounts of iron andzinc in the glass are again lower than those in the lead antimonateparticles, the at% Fe/at% Pb and at% Zn/at% Pb ratios are comparablein the glass and lead antimonate particles.

Micro XRD on individual lead antimonate particles showed that,with one exception, they were all of the type Pb2Sb2O7 with a cubiccrystallographic structure, space group Fd-3m (lattice parametersa ¼ b ¼ c, and a ¼ b ¼ 90�) known as pyrochlore structure. Themajority show the presence of cubic lead antimonates with latticeparameters equal to 10.44, 10.46 and 10.48 �A which, as discussedbelow, are associated with lead antimonates containing iron andzinc (Table 4). Pure lead antimonate with lattice parameter equal to10.40 �A was only observed in one glass (UPP11). In addition to thecubic lead antimonates, the green glass (UPP14) contains PbSb2O6which has a hexagonal crystallographic structure, space group P312(a ¼ b ¼ 5.29�A, c ¼ 5.36�A, a ¼ b ¼ 90�, g ¼ 120�). Other crystallinephases, such as calcium silicates and calcium magnesium silicates,were also found in all the glasses except UPP10 (Table 4). Theprecipitation of calcium and calcium magnesium silicates is ex-pected as Egyptian glasses contain large amounts of calcium andmagnesium oxides (typical averages of 7.2 wt% and 4.1 wt%

Table 3Molar composition of the lead antimonate particles after normalisation to 2 atoms of Pb (average number of atoms and standard deviation (in brackets) among N particles).

Samples N particles At Sb At Fe At Zn At Sn At O

Egyptian e UPP7 2 1.84 (0.00) 0.60 (0.24) 0.53 (0.36) 7.2 (0.4)Egyptian e UPP9 5 1.79 (0.08) 0.29 (0.15) 0.16 (0.10) 7.6 (0.6)Egyptian e UPP10 1 1.75 0.14 0.28 8.1Egyptian e UPP11 3 1.86 (0.13) 0.33 (0.20) 0.17 (0.07) 7.1 (1.4)Egyptian e COP20 4 1.83 (0.05) 0.26 (0.12) 0.10 (0.02) 7.3 (1.6)Egyptian e UPP14 4 1.94 (0.02) 0.07 (0.01) 0.14 (0.04) 5.8 (0.4)Replicate “Egyptian type” glass

(PGeEG þ RGMeCEG-1000 �C)9 1.89 (0.29) 0.20 (0.12) 0.06 (0.09) 6.1 (0.47)

Roman e 116a 6 1.77 (0.15) 0.56 (0.06) 0.13 (0.05) 7.3 (0.5)Replicate “Roman type” glass

(ANeRG þ RGMeCRG-900 �C)7 1.45 (0.11) 0.41 (0.03) 0.23 (0.08) 8.2 (0.9)

Table 4Lattice parameters for lead antimonate particles and other mineral phases detectedin the New Kingdom Egyptian and Roman glasses.

Samples Pb2Sb2O7 latticeparameter (�A)

Other crystallinecompounds

Egyptian e UPP7 10.44 Diopside, CaeMgsilicate, wollastonite

Egyptian e UPP9 10.44, 10.48 Diopside, CaeMgsilicate, wollastonite

Egyptian e UPP10 10.44, 10.48Egyptian e UPP11 10.40, 10.44, 10.48 Diopside, CaeMg

silicate, wollastoniteEgyptian e COP20 10.46 CaeMg silicateEgyptian e UPP14 10.44, 10.48 PbSb2O6

Replicate “Egyptian type”glass (PGeEG þ RGMe

CEG-1000 �C)

10.46, 10.48 Diopside, quartz

Roman e 116a 10.46, 10.48 QuartzReplicate “Roman type”

glass (ANeRG þ RGMe

CRG-900 �C)

10.48

Replicate “Roman type”glass (ANeRG þ RGMe

CRG-1000 �C)

10.44 Ca2Sb2O6.5 (a ¼ 10.37 �A)

8 12 16at% Pb

0

2

4

6

at%Zn

8 12 16 20

20

at% Pb

0

2

4

6

at%Fe

8 12 16 20at% Pb

8

12

16

20

at%Sb

at%Sb = 0.93 at%Pb

upp7upp9upp11upp10cop20upp14

(a)

(b)

(c)

0.4 0.8 1.2 1.6at% Pb

00

0.1

0.2

0.3

0.4

0.5

at%Zn

0 0.4 0.8 1.2 1.6at% Pb

0

0.1

0.2

0.3

0.4

0.5

at%Fe

0 0.4 0.8 1.2 1.6at% Pb

0

0.1

0.2

0.3

0.4

0.5

at%Sb

upp9 (d)

(e)

(f)

Fig. 3. Correlation between Pb at% and (a) Sb, (b) Fe and (c) Zn at% for the leadantimonate particles in New Kingdom Egyptian yellow glasses, and between Pb at% and(d) Sb, (e) Fe and (f) Zn at% for the glass phase surrounding the lead antimonate particlesin New Kingdom Egyptian yellow glass (UPP9). (For interpretation of the references tocolour in this figure legend, the reader is referred to the web version of this article.)



respectively) as compared to Roman glasses (typical averages of4.7e5.8 wt% and 0.4e0.5 wt% respectively) (Table 2).

The plot of the reduced lime and silicawt% contents for the glassphase in sample UPP9 shows that there is no significant differencein the composition near to and away from lead antimonate particles(Fig. 4a). This result is consistent with the observation that, for NewKingdom Egyptian glasses, there are no significant differences inthe reduced silica and lime contents of yellow glasses relative to theother colours (Fig. 1b), and suggests that minimal silica wasincluded in the lead antimonate pigment used in the production ofthese glasses.

3.1.2. Roman glassThe plots of at% of antimony, iron and tin versus at% of lead for

the particles and in the glass surrounding the particles are shown inFig. 5. From Fig. 5a, it is seen that the lead antimonate particlesagain all contain less antimony than lead and show a correlation ofabout 0.86 Sb atom for every Pb atom. In addition, as previouslyobserved by Lahlil et al. (2011) in Roman glasses, the particlescontain small variable amounts of iron and tin. In contrast to thesituation for the Egyptian glasses, the iron and tin contentscorrelate well with the lead content (Fig. 5b and c). Overall, theaverage composition of the lead antimonate particles is 56.7 wt%PbO þ 35.7 wt% Sb2O5 þ 5.1 wt% FeO þ 2.4 wt% SnO2. It should benoted, however, that the bulk tin oxide content of the glass itself isbelow the detection limit (about 0.2 wt% SnO2) for analysis by EDS.

From the corresponding plots for the glass phase surroundingthe lead antimonate particles, it is seen that the antimony and ironcontents show some correlationwith the lead contents (Fig. 5d ande). However, the amount of antimony present in the glass is again

significantly lower than what would be expected from dissolutionof the lead antimonate particles, the at% Sb/at% Pb ratio being about1:4 in the glass as compared to about 1:1.1 in the particles. Incontrast, the tin contents are not correlated with the lead contents(Fig. 5f) and both the low and high lead glass areas contain about0.05 at% Sn, which is consistent with the low solubility of tin in theglass.

Bulk XRD measurements showed that all the lead antimonateparticles were of the type Pb2Sb2O7 (with Sb partly substituted byeither Fe and/or Sn) again with a pyrochlore crystallographicstructure, cubic space group Fd-3m (lattice parameters a ¼ b ¼ c,and a ¼ b ¼ 90�). As a result of iron and tin present in the leadantimonate particles, the majority have lattice parameters equal to10.48�A, but on one side of the sample, there appears a second cubiclead antimonate with a smaller lattice parameter (10.46 �A)(Table 4), thus indicating heterogeneity in the composition of theyellow particles added to the glass.

The CIE-Lab colour coordinates, calculated from the measuredUVevis spectra (Fig. 6a), are presented in Table 5 and plotted inFig. 6b. These results show that the colours of the two sides of thesample are slightly different. One side has a large yellow compo-nent (b* ¼ 26.6) and a small red component (positive a*) whereasthe other side has only a small yellow component (b* ¼ 8.2) and aneven smaller green component (negative a*).

The plot of the reduced lime and silicawt% contents for the glassphase shows that the silica contents are significantly higher near tolead antimonate particles as compared to those away from theparticles (Fig. 4b). This result is consistent with the observation

60 64 68 72

wt% SiO2*

6

8

10

12

wt%CaO*

upp9 glass generalupp9 glass around particles

(a)

60 65 70 75 80 85wt% SiO2*

6

8

10

12

14

wt%CaO*

Glass generalGlass around particles (b)

Fig. 4. Plots of reduced lime content versus reduced silica content around and awayfrom lead antimonate particles for (a) New Kingdom Egyptian glass (UPP9) and (b)Roman glass (116a). (For interpretation of the references to colour in this figure legend,the reader is referred to the web version of this article.)

0 4 8 120

0.2

0.4

0.6

0.8

1at%Sn

2 4 6 8 10 12at% Pb

0

1

2

3

4

at%Fe

0 4 8 12at% Pb

0

4

8

12

at%Sb

A116 (a)

(b)

(c)

at%Sb = 0.86 at%Pb

0 0.4 0.8 1.2 1.6at% Pbat% Pb

0

0.1

0.2

0.3

0.4

at%Sn

0 0.4 0.8 1.2 1.6at% Pb

0

0.4

0.8

1.2

1.6

2

at%Fe

0 0.4 0.8 1.2 1.6at% Pb

0

0.1

0.2

0.3

0.4

at%Sb

A116 (d)

(e)

(f)

Fig. 5. Correlation between Pb at% and (a) Sb, (b) Fe and (c) Sn at% for the leadantimonate particles, and between Pb at% and (d) Sb, (e) Fe and (f) Sn at% for the glassphase surrounding lead antimonate particles in a Roman yellow glass (116a).



that, for Roman glass, the reduced compositions of yellow glassesexhibit elevated silica contents together with lower lime contentsrelative to the other colours (Fig. 1a and Freestone and Stapleton,2013; Figs. 9 and 10), and again suggests that silica was includedin the leadeantimony mixture used in the production of Romanyellow glasses.

3.2. Replication of “Egyptian type” lead antimonate glass

In the replication of “Egyptian type” lead antimonate glass, thefirst step was the synthesis of various lead antimonate pigmentsbased on a stoichiometric mixture of 58 wt% lead oxide and 42 wt%antimony oxide (Table 1). The temperature stability when thestoichiometric pigment was fired with colourless glass was theninvestigated, and finally, based on the compositions of the NewKingdom glasses, an “Egyptian type” lead antimonate glass wasreplicated.

3.2.1. Synthesis of lead antimonate pigmentsFor the lead antimonate pigments produced from the stoichio-

metric mixture (PG1), the bulk XRD results, presented in Table 6,shows that several lead antimonate compounds were formed. Purecubic Pb2Sb2O7, with lattice spacing 10.40 �A, is present for bothfiring temperatures, but it is only one of several lead antimonatespresent. After firing to 800 �C for 2 h, Pb2Sb2O7 is present in traceamounts and the dominant lead antimonate is Pb3þxSb2O8þx. Afterfiring to 1000 �C for 2 h, both these lead antimonates are present inmedium amounts. The CIE-Lab colour coordinates for these pig-ments, which were calculated from the measured UVevis spectra,and which are presented in Table 5 and plotted in Fig. 7, are

consistent with the observed yellow colour; that is, positive b* (23e25) with small positive a* (5e6) indicative of a red component.

For the lead antimonate pigments produced by firing the stoi-chiometric mixture with the addition of varying amounts of ironoxide and zinc oxide (PG2-6) at 1000 �C for 2 h, the bulk XRD re-sults show that several lead antimonate compounds were againformed but that cubic Pb2Sb2O7 is now dominant in all cases(Table 6). However, due to the incorporation of both Fe and Zn intothe cubic lattice, the lattice spacing is greater than that for pure leadantimonate formed previously (a ¼ 10.40 �A), and ranges from10.46�A for 1 wt% FeO to 10.51�A for 2 wt% ZnO. In contrast, the XRDresults for the pigment used to produce “Egyptian type” yellowglass (PGeEG), which contained small amounts of iron and zinctogether with excess lead, show that two cubic lead antimonates(Pb2Sb2O7) were formed, with lattice parameters of 10.425 �A and10.479 �A, characteristic of the incorporation of Fe and Zn into thecubic lattice. In addition, some orthorhombic Pb2Sb2O7, a mixedFeeZn oxide and Sb2O4 were formed.

The CIE-Lab coordinates for the stoichiometric pigments(Table 5, Fig. 7) are again consistent with the observed yellowcolour with positive b* (10e34) and small positive a* (2e7.5)

300 400 500 600 700 800(nm)

10

100

R(%)

Roman glass : side 1

PG-RG+RGM-CRG-900

PG-EG+RGM-CEG-1000

PG-RG+RGM-CRG-1000

-8 -4 0 4 8a*

-40

-20

0

20

40b*

yellow

redgreen

blue

PG-RG+RGM-CRG-900PG-EG+RGM-CEG-1000Roman glass

PG-RG+RGM-CRG-1000

a

b

Fig. 6. (a) UVevis reflectance spectra and (b) CIE-Lab colour coordinates (a* and b*) forRoman glass (116a), and for replicate “Egyptian type” glass (PGeEG þ RGMeCEG-1000 �C) and “Roman type” glass (ANeRG þ RGMeCRG-900 �C and �1000 �C).

Table 5CIE-Lab colour coordinates.

Sample L* a* b* c* h*

Roman yellow glass (116a)Side 1 33.4 �0.3 8.2 8.2 92.1Side 2 48.3 1.7 27.6 27.7 86.5

Replication of “Egyptian type” glassReplicate pigment e PG1 (no NaCl) (fired for 2 h)800 �C 55.5 5.2 23.4 24.0 77.5900 �C 58.1 4.0 18.7 19.1 77.91000 �C 58.1 5.8 24.8 25.5 76.8

Replicate pigments with FeO and ZnO (no NaCl) (fired 1000 �C for 2 h)PG2 e Fe 1 wt% 46.4 5.2 20.5 21.1 75.8PG3 e Fe 4 wt% 61.1 7.5 34.2 35.0 77.6PG4 e Fe 8 wt% 55.2 6.5 27.5 28.3 76.7PG5 e Zn 2 wt% 56.9 6.9 29.3 30.1 76.7PG6 e Zn 1 wt% þ Fe 2.5 wt% 37.1 2.3 10.5 10.7 77.6

Replicate pigment e PG7 (10 wt% NaCl) (fired for 2 h)800 �C 62.6 5.6 34.0 34.5 80.6900 �C 58.3 5.5 33.0 33.5 80.51000 �C 65.5 9.5 43.3 44.3 77.6

Replicate pigments with FeO and ZnO (10 wt% NaCl) (fired 1000 �C for 2 h)PG8 e Fe 1 wt% 73.1 5.4 46.1 46.4 83.3PG9 e Zn 2 wt% 71.8 1.3 46.2 46.2 88.4

Stability of lead antimonate particlesPG1 þ RGMeCEG (fired for 2 h)800 �C 84.5 2.7 31.8 31.9 85.1900 �C 72.9 �4.1 29.4 29.7 97.91000 �C 73.0 �5.2 21.2 21.8 103.81050 �C 53.2 �3.0 8.9 9.4 108.6

PG7 þ RGMeCEG (fired for 2 h)800 �C 67.2 �3.6 25.0 25.3 98.2900 �C 66.4 �7.9 27.1 28.2 106.31000 �C 63.0 �6.8 23.6 24.6 106.11050 �C 57.3 �6.1 14.8 16.0 112.4

“Egyptian type” glass(PGeEG þ RGMeCEG-1000 �C) 65.7 4.6 32.9 33.2 82.0

Replication of “Roman type” glassReplicate anime (fired 900 �C for 2 h)AN1 62.4 6.1 34.1 34.6 79.9AN2 38.0 1.6 9.4 9.5 80.3AN3 81.6 �0.1 7.5 7.5 90.8AN4 43.0 0.2 1.7 1.7 83.3

Stability of lead antimonate particlesAN1 þ RGMeCRG (fired for 2 h)900 �C 59.0 �5.5 13.9 14.9 111.61000 �C 58.7 �2.2 3.1 3.8 125.4

“Roman type” glass(ANeRG þ RGMeCRG-900 �C) 43.6 9.1 24.2 25.9 69.4(ANeRG þ RGMeCRG-1000 �C) 49.6 6.8 21.1 22.2 72.1



indicative of a red component. However, there is no clear correla-tion between colour and composition of the pigment, and only thepigment mixtures containing either 4 wt% FeO (PG3) or 2 wt% ZnO(PG5) exhibit a significantly stronger yellow colour (i.e., b*approximately 34 and 29 respectively) than that of the stoichio-metric mixture without the addition of iron or zinc oxides.

For the lead antimonate pigments produced from the stoichio-metric mixture to which some 10 wt% sodium chloride flux (NaCl)was added (PG7), the bulk XRD results, presented in Table 6, showthat cubic Pb2Sb2O7, with lattice spacing approximately 10.40 �A, isalready the major component after firing to 900 �C, and after firingto 1000 �C, it is the only lead antimonate present. In contrast, whensmall amounts of iron or zinc were included in the NaCl containingpigments (PG8-9), two cubic lead antimonates (Pb2Sb2O7) withlattice parameters of 10.386 and 10.453 �A for PG8, and 10.400 and10.462 �A for PG9 were formed, as in the case of pigment PGeEGcontaining excess lead. The observed colours of these pigments area more intense and homogeneous yellow than those observed forpigments produced from the stoichiometric mixtures without theaddition of NaCl flux. This difference is reflected in the CIE-Labcoordinates (Table 5, Fig. 7) for which b* is in the range 33e46 ascompared to 23e34 without the addition of NaCl flux.

3.2.2. Stability of lead antimonate particlesThe bulk XRD measurements on glasses produced from mix-

tures of stoichiometric pigment without the addition of NaCl flux(PG1) and colourless glass (RGMeCEG) (Table 2) indicate that traceamounts of cubic Pb2Sb2O7 (a ¼ 10.40 �A) are present for firings upto 1000 �C. However, for a firing at 1000 �C, hexagonal CaSb2O6 isalso present, and when the glass is fired to 1050 �C, the cubic

Table 6XRD data for replicate lead antimonate pigments, fired for 2 h (*: major; m: median; t: traces).

Replicate pigment JPDF file Compound Lattice parameter Crystalline structure

PG1 (no NaCl)800 �C 00-034-1196 Pb3þxSb2O8þx*

01-084-1423 PbSb2O6t

00-042-1355 Pb2Sb2O7t Cubic (Fd-3m)

00-039-0834 Pb2Sb2O7t Orthorhombic

1000 �C 01-084-1423 PbSb2O6*00-034-1196 Pb3þxSb2O8þx

m

00-042-1355 Pb2Sb2O7m a ¼ 10.398(1) Cubic (Fd-3m)

00-039-0834 Pb2Sb2O7t Orthorhombic

PG2ePG6 (no NaCl), fired 1000 �CPG2-Fe 1 wt% Pb2Sb2O7-type* a ¼ 10.457(6) Cubic (Fd-3m)

01-084-1423 PbSb2O6m

00-034-1196 Pb3þxSb2O8þxm

00-039-0834 Pb2Sb2O7m Orthorhombic

PG3-Fe 4 wt% Pb2Sb2O7-type* a ¼ 10.490(4) Cubic (Fd-3m)01-084-1423 PbSb2O6

m

PG4-Fe 8 wt% Pb2Sb2O7-type* a ¼ 10.488(5) Cubic (Fd-3m)01-084-1423 PbSb2O6

m

00-034-0372 PbSbO4t

01-079-1741 Fe2O3

PG5-Zn 2 wt% Pb2Sb2O7-type* a ¼ 10.508(7) Cubic (Fd-3m)01-084-1423 PbSb2O6

m

00-034-1196 Pb3þxSb2O8þxm

00-039-0834 Pb2Sb2O7m Orthorhombic

PG6-Zn 1 wt%þ Pb2Sb2O7-type* a ¼ 10.494(1) Cubic (Fd-3m)FeO 2.5 wt% 01-084-1423 PbSb2O6

m

00-034-1196 Pb3þxSb2O8þxt

PG7 (10 wt% NaCl)800 �C 00-034-1196 Pb3þxSb2O8þx

00-042-1355 Pb2Sb2O7 a ¼ 10.396(4) Cubic (Fd-3m)00-039-0834 Pb2Sb2O7

t Orthorhombic900 �C 00-042-1355 Pb2Sb2O7 a ¼ 10.391(1) Cubic (Fd-3m)

00-034-1196 Pb3þxSb2O8þxt

01-084-1423 PbSb2O6t

1000 �C 00-042-1355 Pb2Sb2O7 a ¼ 10.398(1) Cubic (Fd-3m)PG8-PG9 (10 wt% NaCl), fired 1000 �CPG8-Fe 1 wt% 00-042-1355 Pb2Sb2O7 a ¼ 10.386(7) Cubic (Fd-3m)

Pb2Sb2O7-type a ¼ 10.453(3) Cubic (Fd-3m)PG9-Zn 2 wt% 00-042-1355 Pb2Sb2O7 a ¼ 10.400(5) Cubic (Fd-3m)

Pb2Sb2O7-type a ¼ 10.462(1) Cubic (Fd-3m)PGeEG, fired 1000 �C

00-042-1355 Pb2Sb2O7-type a ¼ 10.425(2) Cubic (Fd-3m)Pb2Sb2O7-type a ¼ 10.479(2) Cubic (Fd-3m)

-20 -10 0 10 20a*

-40

0

40

b*yellow

redgreen

blue

PG7PG8PG9

PG1PG3PG5PG4PG2

PG6

Fig. 7. CIE-Lab colour coordinates (a* and b*) for replicate pigments PG1 (no NaCl),PG2-6 (no NaCl but added Fe and/or Zn), PG7 (10 wt% NaCl), and PG8-9 (10 wt% NaClplus added Fe or Zn), all fired at 1000 �C.



Pb2Sb2O7 is converted to orthorhombic calcium antimonate,Ca2Sb2O7.

The CIE-Lab colour coordinates (Table 5) reflect the changes inthe antimonate phases present in the glasses fired to differenttemperatures, as determined by XRD measurements. Thus, as aresult of the partial conversion from Pb2Sb2O7 to CaSb2O6, the b*colour coordinates are reduced from 31.8 and 29.4 for the glassesfired at 800 �C and 900 �C respectively, to 21.2 for the glass fired at1000 �C. Although these b* values are comparable to that for theoriginal pigment (b* ¼ 24.8), the glasses differ from the pigment inhaving small, negative a* coordinates (�5 to�3) and being yellowe

green in colour, whereas the original pigment had a small positivea* coordinate (þ5.8) and was yellowered in colour. Finally, at1050 �C, the decomposition of the lead antimonate and the for-mation of calcium antimonate results in the decolouration of theglass (i.e., b* ¼ 9).

In contrast, when stoichiometric pigment with the addition ofNaCl flux (PG7) was used, bulk XRDmeasurements showed that thepure cubic Pb2Sb2O7 (a ¼ 10.40�A) suffers some transformation to acubic antimonatewith a smaller lattice parameter (10.38�Aat afiringtemperature of 800 �C and 10.34 �A at 900 �C and 1000 �C). Thesechanges aremostprobably the result of Ca substituting for Pb to forman intermediate cubic leadecalcium antimonate (Pb,Ca)2Sb2O7.Then, at a firing at 1050 �C, this intermediate cubic antimonate de-composes to form orthorhombic calcium antimonate, Ca2Sb2O7. TheCIE-Lab colour coordinates reflect this early transformation to anintermediate cubic antimonate in that the b* for firing temperatureof 800 �C and 900 �C (25.0 and 27.1) are lower than the corre-sponding values (31.8 and 29.4) for the pigment without the addi-tion of NaCl flux.

3.2.3. Replication of “Egyptian type” yellow glassSince Egyptian yellow glasses did not exhibit elevated silica

contents relative to the other colours, replicate “Egyptian type”glass was produced using a lead antimonate pigment containing nosilica. The composition of the pigment, which contained smallamounts of iron and zinc together with excess lead oxide is given inTable 1 (PGeEG). 12 wt% of this powdered pigment was added to88 wt% powdered colourless glass (RGMeCEG) with compositioncomparable to that of New Kingdom Egyptian glass (Table 2), andthe combined mixture was fired at 1000 �C for 2 h.

The resulting the replicate glass (PGeEG þ RGMeCEG-1000 �C)showed amottled pattern of bright yellow andwhite areas (Fig. 8a).In polished section in the SEM, it can be seen that, because the glasswas not stirred when molten, there are areas of glass with higherand lower lead oxide contents, and the lead antimonate particlesare not well distributed, are frequently clustered together and arevery variable in size (Fig. 8c).

From the plots of at% of antimony, iron and zinc versus at% oflead for the particles and glass phase, as determined by EDS anal-ysis (Fig. 9aec), it can be seen that the glass phase contains noantimony but small amounts of iron and zinc. In the lead antimo-nate particles, the antimony content correlates with the lead con-tent, and their compositions are close to the theoreticalstoichiometric composition, with the iron and zinc contents beingvery variable between particles. The average molar concentrationsof the particles, normalised to two atoms of lead, are comparedwith those for the New Kingdom Egyptian glasses in Table 3, andthe bulk composition of the replicate glass is given in Table 2.

The bulk XRD pattern shown in Fig. 10a for the replicate glassindicates that, as for the New Kingdom Egyptian glass, the lead

Fig. 8. Replicate (a) “Egyptian type” (PGeEG þ RGMeCEG-1000 �C) and (b) “Roman type” (ANeRG þ RGMeCRG-900 �C) lead antimonate glasses, and SEM photomicrographs in back-scattermodeofpolishedsections throughthese replicate (c)“Egyptian type”and (d) “Romantype”glasses showinga scatterof leadantimonateparticles (white) togetherwithareasofglassboth richer in lead (lighter grey) and poorer in lead (darker grey). (For interpretation of the references to colour in thisfigure legend, the reader is referred to thewebversion of this article.)



antimonate particles are of the cubic Pb2Sb2O7 type with two highlattice parameters (a¼ 10.46�A and 10.48�A) associatedwith variousiron and zinc contents (Table 4).

The CIE-Lab colour coordinates, calculated from the measuredUVevis spectra (Fig. 6a), are included in Table 5 and plotted inFig. 6b, and indicate a strong yellow colour (b* ¼ 32.9) with a smallred component (a* ¼ 4.6). Thus, the colour coordinates for a glassproduced using a pigment with excess lead correspond well withthose for the replicate lead antimonate pigments containing FeOand ZnO, some of which show similar yellow components (eg., b*equal to 29.3 and 34.2) together with small red components (a*equal to 6.9 and 7.5) (Table 5).

3.3. Replication of “Roman type” lead antimonate glass

In the replication of “Roman type” lead antimonate glass, thefirst stepwas the synthesis of various leadeantimonyesilica animescontaining excess lead (Table 1). The temperature stability whenanime was fired with colourless glass was then investigated, andfinally, based principally on the compositions of the Roman glasses,a “Roman type” lead antimonate glass was replicated.

3.3.1. Synthesis of leadeantimonyesilica animeIn terms of colour, only the two replicate anime mixtures with

low silica contents (AN1 and AN2 in Table 1) exhibit a yellow colour,

although AN2 is very pale. In contrast the high silica animes (AN3and AN4) are essentially white.

The bulk XRD measurements (Fig. 11) showed that none of theanimes contained the lead antimonate pyrochlore, Pb2Sb2O7.Instead the dominant phases in the two mixtures with low silicacontents (AN1 and AN2) are Pb3þxSb2O8þx and the lead silicate,Pb2(SiO3)O (JPDF file 01-085-0686) whereas those in the twomixtures with high silica contents (AN3 and AN4) are quartz andPbSb2O6. The observed colour and the lead antimonate phasespresent are reflected in the CIE-Lab coordinates (Table 5). Thus, ofthe low silica animes, AN1 with the lower PbO/Sb2O5 ratio (8 ascompared to 12) exhibits a much more intense yellow than AN2(b* ¼ 34.1 and 9.4 respectively), and the two white, high silicaanimes exhibit even lower b* values (7.5 and 1.7 respectively).

3.3.2. Stability of lead antimonate particlesThe glass resulting from firing a mixture of the low silica, lower

PbO/Sb2O5 ratio anime (AN1) and colourless glass (RGMeCRG) to900 �C is mottled yellow in appearance. Under the optical micro-scope, a mixture of yellow and white particles together with someparticles with a yellow core and a white edge are visible. The bulkXRD pattern indicates the presence of a pyrochlore with a latticeparameter close to 10.30�A which is again most probably the resultof Ca substituting for Pb to form an intermediate cubic leade

0 4 8 12 16at% Pb

0

0.2

0.4

0.6

0.8

at%Zn

0 4 8 12 16at% Pb

0

0.2

0.4

0.6

0.8

1

at%Fe

0 4 8 12 16at% Pb

0

4

8

12

16

at%Sb

EG-Pbparticlesglass

(a)

(b)

(c)

0 2 4 6 8at% Pb

0

0.4

0.8

1.2

1.6

at%Sn

0 2 4 6 8at% Pb

0

0.4

0.8

1.2

1.6

2

at%Fe

0 2 4 6 8at% Pb

0

2

4

6

8

at%Sb

(d)

(e)

(f)

particlesglass

Fig. 9. Correlation between Pb at% and (a) Sb, (b) Fe, and (c) Zn at% for the leadantimonate particles and surrounding glass phase in replicate “Egyptian type” (PGeEG þ RGMeCEG-1000 �C) and between Pb at% and (d) Sb, (e) Fe, and (f) Sn at% inreplicate “Roman type” (ANeRG þ RGMeCRG-900 �C) lead antimonate yellow glasses.(For interpretation of the references to colour in this figure legend, the reader isreferred to the web version of this article.)

20 30 40 50 60

0

200

400

600

800

Intensity(a.u.)

quartz - SiOdiopside - CaMgSi O

(10.48Å) (10.46Å) -cubic Pb Sb O

PG-EG+RMG-CEG-1000

(a)

10 20 30 40 50 60

0x100

2x103

4x103

6x103

8x103

104Intensity(a.u.)

900ºC

1000ºC

Pb2Sb2O7- cubic (a=10.441Å)Ca2Sb2O6.5 - cubic (a=10.37Å)

Pb2Sb2O7- cubic (a= 10.482Å)

x4AN-RG+RGM-CRG

(b)

Fig. 10. XRD patterns for replicate (a) “Egyptian type” (PGeEG þ RGMeCEG-1000 �C)and (b) “Roman type” (ANeRG þ RGMeCRG) lead antimonate glasses, the latter firedto 900 �C and 1000 �C. (For interpretation of the references to colour in this figurelegend, the reader is referred to the web version of this article.)



calcium antimonate (Pb,Ca)2Sb2O7. The CIE colour coordinates(Table 5) indicate a fairly weak yellow colour (b* ¼ 13.9) with asmall green component (a* ¼ �5.5). In contrast, the glass resultingfrom firing this same mixture to 1000 �C is opaque white with b*coordinate equal to only 3.1.

3.3.3. Replication of “Roman type” yellow glassSince Roman yellow glasses exhibited elevated silica contents

relative to the other colours, replicate “Roman type” glasses wereproduced using a leadeantimonyesilica anime, rather than a leadantimonate pigment containing no silica. The composition of theanime (ANeRG), which contained small amounts of iron and tintogether with excess lead oxide is given in Table 1. Mixtures of thispowdered anime and powdered colourless glass (RGMeCRG) withcomposition comparable to that of Roman glass were used(Table 2) (20wt% animeþ 80wt% glass) and (12wt% animeþ 88wt%glass)mixturesbeingfiredat900 �Cand1000 �C, respectively, for 2h.

The resulting replicate “Roman type” glass fired to 900 �C (ANeRG þ RGMeCRG-900 �C) is a definite yelloweorange colour(Fig. 8b). In polished section in the SEM, it can be seen that, becausethe glass was not stirred when molten, the lead antimonate parti-cles are again notwell distributed, are frequently clustered togetherand are very variable in size (Fig. 8d).

From the plots of at% of antimony, iron and tin versus at% of leadfor the particles and glass phase, as determined by EDS analysis(Fig. 9def), it can be seen that the glass phase contains no antimonybut small amountsof ironand tin. In the leadantimonateparticles, theantimony and iron contents correlate with the lead content, but thetin contents are more variable between particles. The average molarconcentrations of the particles, normalised to two atoms of lead,are comparedwith those for the Roman glass sample,116a, inTable 3,and the bulk composition of the replicate glass is given in Table 2.

The bulk XRD pattern (Fig. 10b) for the replicate glass indicatesthat, as for the Roman glass sample, the lead antimonate particlesare of the cubic Pb2Sb2O7 type with a high lattice parameter(a ¼ 10.482 (1)�A) associated with the incorporation of iron and tininto the lattice (Table 4). The CIE-Lab colour coordinates, calculatedfrom the measured UVevis spectra (Fig. 6a), are included in Table 5and plotted in Fig. 6b, and indicate a strong yellow colour(b* ¼ 24.2) together with a significant red component (a* ¼ 9.1)which is most probably due to the high iron content of the anime.

The plot of the reduced lime and silicawt% contents for the glassphase shows that the silica contents are significantly higher near tolead antimonate particles as compared to those away from theparticles (Fig. 12). This result is as expected with the use of leadantimonate anime rather than lead antimonate pigment, andconsistent both with the results obtained for the Roman glasssample, 116a (Fig. 4b), as well as those reported for early Romanglass by Freestone and Stapleton (2013) and those calculated fromdata published by Mass et al. (1998) for later Roman glass (Fig. 1a).

The replicate “Roman type” glass fired to 1000 �C (ANeRG þ RGMeCRG-1000 �C) also results in a yelloweorange colour,but one that is less saturated than for the glass produced by firing to900 �C. The CIE-Lab colour coordinates, calculated from themeasured UVevis spectra (Fig. 6a) and included in Table 5 andplotted in Fig. 6b, are similarly slightly reduced, although boththese observations could be the result of less anime being includedin the mixture (12 wt% as compared to 20 wt% for the 900 �Cmixture). However, the bulk XRD pattern (Fig. 10b) for this replicateglass indicates that, in addition to cubic lead antimonate particles(lattice parameter a ¼ 10.44 �A), some cubic calcium antimonateparticles (Ca2Sb2O6.5 with a ¼ 10.37 �A) are also present. Therefore,

20 40 60

0x10

4x10

8x10

1x10

2x10

2x10

Intensity(a.u.)

Pb Sb ORosiaite -PbSb O -

quartz -SiO -

AN1

AN3

Pb Sb OPb SiO

Rosiaite -PbSb O -

20 40 60

0x10

5x10

10

2x10

2x10

3x10

Intensity(a.u.) Rosiaite -PbSb O -

quartz -SiO -

AN2

AN4

Pb Sb OPb SiO

Rosiaite -PbSb O -

Fig. 11. XRD patterns for replicate animes AN1-4. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

60 64 68 72 76 80wt% SiO2*

4

5

6

7

8

wt%CaO*

Glass generalGlass around particles

Fig. 12. Plot of reduced lime content versus reduced silica content around and awayfrom lead antimonate particles for replicate “Roman type” lead antimonate yellowglass (ANeRG þ RGMeCRG-900 �C). (For interpretation of the references to colour inthis figure legend, the reader is referred to the web version of this article.)



in spite of the glass still appearing yellow in colour, there has beenpartial conversion from lead antimonate to calcium antimonate fora firing temperature of 1000 �C, and the glass produced does notfully replicate the ancient Roman glass.

4. Discussion

4.1. New Kingdom Egyptian and Roman glasses

The results presented above confirm that New Kingdom Egyp-tian and Roman yellow glasses are similar in terms of the compo-sition (Table 3) and crystallographic structure (Table 4) of their leadantimonate particles. The principal difference is that, in addition tothe iron present in both types of glass, tin has replaced zinc as thesecond impurity substituting for antimony in the lead antimonateparticles in the Roman glass.

Lead antimonate is a cubic pyrochlore oxide with general for-mula A2B2O7 which has two interpenetrating lattices, an A4Otetrahedra and a BO6 vertex sharing octahedra. The A cationsoccupy part of large sites and help charge balance (either A2þ andB5þ or A3þ and B4þ), whereas the Sn4þ, Zn2þ and Fe2þ cationssubstitute the octahedral sites (Sb5þ), modifying the charge valenceand distorting the structure. In general and taking into account thedetection limits and error in the lattice parameter determinationand the statistics, the lead antimonate particles in all the glassescontain seven oxygen atoms and their At Sb/At Pb ratios vary be-tween 1.75:2.0 and 1.9:2.0. These results, together with theobserved lattice parameters, are consistent with those previouslyreported by Rosi et al. (2009).

The fact that at% of iron, zinc and tin show some correlationwithat% lead, either in the lead antimonate particles or in the glasssurrounding these particles, suggests that the iron, zinc and tinwere addedwith the pigment or anime. However, the higher at% Sb/at% Pb ratios, but similar at% Fe/at% Pb, at% Zn/at% Pb and at% Sn/at%Pb ratios, in the particles as compared to the surrounding glasssuggests that much of the iron, zinc and tin present in the pigmentand animewas not originally incorporated into the lead antimonateparticles, and therefore, on addition to the glass, was readily dis-solved in the glass phase. In contrast, there was only limiteddissolution of the lead antimonate particles which contained thebulk of the antimony in the pigment and anime.

Mass et al. (1998) proposed, in the context of the production ofRoman yellow glasses, that the lead antimonate was produced fromantimonial litharge (i.e., litharge contaminated with antimony)which they argued was a by-product of the cupellation of silverproduced from argentiferous lead ores. Subsequently, Mass et al.(2002) suggested that antimonial litharge from the cupellation ofsilver similarly provided the lead antimonate used in the produc-tion of New Kingdom Egyptian yellow glasses.

However, as listed in Table 2, New Kingdom Egyptian yellowglasses contain some 0.3 wt% zinc oxide (Shortland and Eremin,2006), and Rehren (2003) has convincingly argued that this con-centration of zinc would not have survived the smelting of argen-tiferous lead ores in the production of silver-containing lead metal,from which silver would have been subsequently extracted bycupellation. Further, Shortland et al. (2000) have established bylead isotope analysis that the PbeZn ore from Gebel Zeit on the RedSea coast of Egypt was the probably source of the lead used in theproduction of New Kingdom Egyptian yellow glasses, and that sil-ver metal used in Egypt at this period was produced from a non-Egyptian lead ore.

Therefore, as proposed by Shortland (2002a), the lead antimo-nate pigment used in the production of New Kingdom Egyptianyellow glass was most probably produced by roasting a mixture ofgalena (PbS) and stibnite (Sb2S3) containing excess lead. The result

would have been the formation of lead and antimony oxides whichthen combined to produce lead antimonate. Because of the re-fractory nature of zinc oxide (m.p. 1975 �C), the loss of zinc in thisprocess would have been limited. It seems reasonable to assumethat the Egyptian glass makers would have been able to select andcombine these two metal ores since they were also being used forquite different purposes. Thus stibnite was being used in the pro-duction of calcium antimonate white glasses (Mass et al., 1998;Shortland, 2002a) and occasionally as a kohl, and galena was be-ing used in the production of both lead metal and silver as well asbeing used as a kohl (Lucas and Harris, 1962, 80e84). However,although there is strong lead isotope data to show that galena fromGebel Zeit in Egypt was used as a kohl (Shortland et al., 2000), it isuncertain on the basis of limited lead isotope data (Stos-Fertner andGale, 1979) whether either lead metal or silver was produced fromgalena from Egypt. Furthermore, the two ores were most probablybeing obtained from different places, the galena, as discussedabove, from Gebel Zeit in Egypt, and the stibnite from the Caucasuswhere antimonymines that were active in the Late Bronze Age havebeen discovered (Chernykh, 1992; Shortland, 2002a).

Conversely, because of the absence of zinc from the Romanyellow glasses, antimonial litharge resulting from the cupellation ofsilver is a possible source of the lead antimonate used in the pro-duction of these glasses. However, there are some doubts as to theextent to which cupellation litharge containing sufficient antimonywould have been available (Rehren, 2003). The alternative wouldhave again been to roast a mixture of galena and stibnite, but in thiscase, zinc must have been absent from the galena chosen. A furtheroption would have been to replace the galena by cupellationlitharge from which any zinc originally present would have beenlost during smelting.

The origin of the iron present in the New Kingdom Egyptianpigment is probably the result of contamination from ironminerals,such as pyrite, introduced with the galena used in the production ofthe pigment, and from the clay crucibles in which the pigment wasproduced. If antimonial litharge was used to produce the Romanpigment then, as discussed by Mass et al. (1998), an additionalsource of iron could have been contamination from iron tools usedduring the cupellation process. Also, in the case of Roman yellowglass, Freestone and Stapleton (2013) have suggested that extra ironwas deliberately added to the anime mixture in order to stabilisethe lead antimonate and make a stronger yellow.

The origin of the tin in the Roman yellow glass, which is pre-sent in the lead antimonate particles although below the detec-tion limits for EDS analysis in the bulk glass, presents more of aproblem. It is not impossible that the tin was introduced as animpurity from the stibnite, in that some 0.5 wt% of tin wasdetected in a much earlier, 3rd millennium BC, antimony beadfrom Syria (Shortland, 2002b). Another possibility is that the scraplead, which was most probably used in the production of thepigment, contained small amounts of tin, as previously observedby Wyttenbach and Schubiger (1973) in the neutron activationanalysis of Roman water pipes. In this case, it was suggested thatthe lead used to make the pipes included scrap lead which hadbecome contaminated by the inclusion of small amounts of solder.Alternatively, if cupellation litharge was used, this could havecontained small amounts of tin as a result of contamination fromother adjacent metal production or working processes. Finally, inview of its role in increasing the stability of lead antimonateparticles in glass, it is not impossible that tin oxide was deliber-ately added to the anime mixture. However, before it will bepossible to resolve the origin of the tin, more information isneeded on both the frequency with which tin is present in thelead antimonate particles in Roman glass, and the time periodover which this occurs.



4.2. Replication of “Egyptian and Roman type” lead antimonateglasses

The synthesis of pure stoichiometric lead antimonate pigmentshas shown that the cubic lead antimonate, Pb2Sb2O7, is only one ofseveral lead antimonates formed. However, the addition of NaClflux to the mixture produces a pigment in which cubic Pb2Sb2O7 isthe dominant lead antimonate and which exhibits a more intenseyellow colour. This greater reactivity is most probably the result ofthe NaCl melting at 801 �C to form a liquid which wets the surfacesof the lead and antimony oxide particles, and results in an increasein their dissolution and the diffusivity between them. Therefore, asindicated in the relevant Renaissance treatises (Dik et al., 2005;Piccolpasso, 2007), the addition of an alkali flux was necessary toproduce the intense yellow pigments required for use in paintingsand Italian maiolica glazes from the late 15th century AD onwards.In contrast, when the pigment with added NaCl is incorporated intoa colourless glass, the cubic Pb2Sb2O7 tends to be less stable, andthe resulting colour tends to be weaker than that produced using astoichiometric pigment without added NaCl. Therefore, in the caseof the lead antimonate pigment to be used in the production ofyellow glasses, the addition of an alkali flux to the pigment mixturewould not have been beneficial.

The inclusion of excess lead in a pigment mixture, as in the caseof the pigment (PGeEG) used to produce “Egyptian type” yellowglass, results in the formation of two cubic lead antimonates(Pb2Sb2O7) with different lattice parameters (10.425 �A and10.479�A), as otherwise observed only for pigments containing bothNaCl flux and small amounts of iron or zinc. Thus, the inclusion ofexcess lead also facilitates the formation of cubic lead antimonates.

Similarly, the anime produced from a leadeantimonyesilicamixture with excess lead does not contain cubic Pb2Sb2O7 andinstead, the dominant lead antimonate is Pb3þxSb2O8þx. However,for a mixture with a PbO/Sb2O5 ratio of 8, the yellow colour is moreintense than that observed in a pigment produced from the purestoichiometric mixture without the addition of NaCl flux (i.e.,b* ¼ 34.1 compared to b* ¼ 24.8), but when the PbO/Sb2O5 ratio isincreased to 12, the yellow colour becomes very pale (b* ¼ 9.4).

Comparison of the “Egyptian type” yellow glass produced usinga pigment containing small amounts of iron and zinc together withexcess lead oxide (i.e., PGeEG þ RGMeCEG) with that producedusing pure stoichiometric pigment (i.e., PG1 þ RGMeCEG) in thecontext of the stability experiments established that the cubic leadantimonate particles and the intensity of the yellow colour survivedto a higher temperature in the former case. In principle, thisincreased stability could be due to the addition of either iron-plus-zinc or excess lead oxide. However, the addition of iron-plus-zincseems the more likely explanation since “Roman type” yellowglass produced from anime containing small amounts of iron andtin (i.e., ANeRG þ PGMeCRG) is more stable than that produced inthe stability experiments from a pure leadeantimonyesilica anime(i.e., AN1 þ RGMeCRG), both of which contained excess lead oxide.

The replication experiments suggest that “Roman type” yellowglass (ANeRG þ CGMeCRG) is less stable than “Egyptian type”yellow glass (PGeEG þ CGMeCEG) in that, after firing to 1000 �C,conversion from cubic Pb2Sb2O7 to calcium antimonate, Ca2Sb2O6.5,has started only for the “Roman type” glass. The lower stability oflead antimonate in Roman glass is probably due to its lower vis-cosity compared to Egyptian glass, as observed during the repli-cation experiments. The streaks of lead-rich glass containing leadantimonate, commonly seen in Egyptian glass (Fig. 2a), reflects thisrelatively high viscosity and the resultant difficulty in mixing in thepigment. The higher lime and magnesia contents of the Egyptianglasses (Table 2) are responsible for their increased viscositycompared to the Roman glasses. The diffusion coefficients of the

atoms in the glass are inversely dependent on the viscosity (Ein-steineStokes relationship) and therefore, higher viscosity implies alower mobility which, in turn, will help the stability of the leadantimonate particles in the glass. Because of the high viscosity ofEgyptian glass, the addition of a yellow pigment with excess leadwas sufficient to obtain a glass with stable lead antimonate parti-cles. In contrast, the addition of the more stable yellow leadeantimonyesilica animewas the method chosen by the Roman glassmakers to solve this problem. Nevertheless, Roman yellow glassprobably still had to be fired at a lower temperature than Egyptianyellow glass in order to avoid partial conversion to calciumantimonate.

5. Conclusions

The above analyses of New Kingdom Egyptian and Roman yel-low glasses have established that, with one exception, the leadantimonate particles responsible for the yellow colour were of thetype Pb2Sb2O7 with a cubic crystallographic structure. In the ma-jority of the particles, the antimony was partially replaced by ironand zinc in the case of the Egyptian glasses, and by iron and tin inthe case of the Roman glass. It seems probably that these impuritieswere incorporated into the glass through the raw materials andprocesses used to produce the lead antimonate pigment or animeemployed for the Egyptian and Roman glasses respectively. Asconfirmed by the pigment replications, the impurities partiallyreplacing the antimony in the lead antimonate particles resulted inobserved lattice parameters in the range 10.44e10.48 �A, ascompared to a lattice parameter equal to 10.40 �A for pure leadantimonate particles.

Replication experiments have established that both NewKingdom Egyptian and Roman yellow glasses could have beenproduced by stirring, respectively, lead antimonate pigment oranime, containing small amounts of iron, zinc or tin oxides togetherwith excess lead oxide, into a molten colourless glass. The repli-cation experiments further confirmed that, in producing both typesof glass, the inclusion of small amounts of impurities, such as Fe, Znand Sn, into the pigment or anime was an important factor inenhancing the stability of the lead antimonate particles. However,although no conversion of yellow lead antimonate to white calciumantimonate had occurred at a firing temperature of 1000 �C in thecase of the replicate “Egyptian type” yellow glass, partial conver-sion had started at this firing temperature in the case of the repli-cate “Roman type” glass. As discussed above, the lower stability oflead antimonate particles in Roman glass was probably due to itslower viscosity at equivalent temperatures compared to Egyptianglass, and is the reason why the more stable leadeantimonyesilicaanime was the method chosen by the Roman glass makers, ratherthan lead antimonate pigment which was used by the NewKingdom Egyptian glass makers.

Acknowledgements

The study is funded by CICYT grant MAT2010-20129-C02-10 andGeneralitat de Catalunya grant 2009SGR01225. Professor Ian Free-stone (University College London) is thanked for providing the Ro-man glass sample (116a) and for providing access to his paper (withC. P. Stapleton) on Roman mosaic glass prior to publication; theVictoria Museum of Egyptian Antiquities (Uppsala) and the Natio-nalmuseet (Copenhagen) for providing the Egyptian glass samples(UPP and COP respectively); and The Roman Glassmakers (Andover,Hampshire) for providing the colourless glasses comparable incomposition to New Kingdom Egyptian and Roman colourlessglasses. Three anonymous referees are thanked for their very helpfulcomments that have resulted in major improvements to the paper.



References

Biringuccio, V., 1966. The Pirotechnia of Vannoccio Biringuccio (C.S. Smith, M.T.Gnudi, Trans.). The MIT Press, Cambridge, MA.

Chernykh, E., 1992. Ancient Metallurgy in the USSR. Cambridge University Press,Cambridge.

Dik, J., Hermens, E., Peschar, R., Schenk, H., 2005. Early production recipes for leadantimonate yellow in Italian art. Archaeometry 47, 593e607.

Freestone, I.C., Stapleton, C.P., 2013. Composition, technology, and production ofcoloured glasses from mosaic vessels of the Early Roman Empire. In: Bayley, J.,Freestone, I., Jackson, C. (Eds.), Glass of the Roman Empire. Oxbow Books, Ox-ford (in press).

Lahlil, S., Cotte, M., Biron, I., Szlachetko, J., Menguy, N., Susini, J., 2011. Synthesizinglead antimonate in ancient and modern opaque glasses. Journal of AnalyticalAtomic Spectrometry 26, 1040e1050.

Lucas, A., Harris, J.R., 1962. Ancient Egyptian Materials and Industries. EdwardArnold, London.

Mass, J.L., Stone, R.E., Wypyski, M.T., 1998. The mineralogical and metallurgicalorigins of Roman opaque colored glasses. In: Kingery, W.D., McCray, P. (Eds.),The Prehistory and History of Glassmaking Technology, Ceramics and Civiliza-tion. American Ceramics Society, Westerville, Ohio, pp. 121e144.

Mass, J.L., Wypyski, M.T., Stone, R.E., 2002. Malkata and Lisht glassmaking tech-nologies: towards a specific link between second millennium BC metallurgistsand glassmakers. Archaeometry 44, 67e82.

Moretti, C., Hreglich, S., 1984. Opacification and colouring of glass by the use of“anime”. Glass Technology 25, 277e282.

Piccolpasso, C., 2007. The Three Books of the Potter’s Art (R. Lightbown, A. Caiger-Smith, Trans.). Editions la revue de la céramique et du verre, Vendin-le-Vieil.

Rehren, Th, 2003. Comments on Mass, J.L., Wypyski, M.T., Stone, R.E., 2002. Malkataand Lisht glassmaking technologies: towards a specific link between second

millennium BC metallurgists and glassmakers. Archaeometry 44, 67e82, andreply. Archaeometry 45, 185e190.

Rosi, F., Manuali, V., Miliani, C., Brunetti, B.G., Sgamellotti, A., Grygar, T., Hradil, D.,2009. Raman scattering features of lead pyroantimonate compounds. Part I:XRD and Raman characterization of Pb2Sb2O7 doped with tin and zinc. Journalof Raman Spectroscopy 40, 107e111.

Shortland, A.J., 2002a. The use and origin of antimonate colorants in early Egyptianglass. Archaeometry 44, 517e531.

Shortland, A.J., 2002b. An antimony bead from Jerablus Tahtani. Historical Metal-lurgy 36, 1e6.

Shortland, A.J., Eremin, K., 2006. The analysis of secondmillennium glass from Egyptand Mesopotamia, Part 1: newWDS analyses. Archaeometry 48, 581e603.

Shortland, A.J., Nicholson, P.T., Jackson, C.M., 2000. Lead isotope analysis of 18thDynasty Egyptian eyepaints and lead antimonate colorants. Archaeometry 42,153e159.

Stos-Fertner, Z., Gale, N.H., 1979. Chemical and lead isotope analysis ofancient Egyptian gold, silver and lead. In: Scollar, I. (Ed.), Proceedings ofthe 18th International Symposium of Archaeometry and ArchaeologicalProspection, Archaeo-Physica 10. Rheinisches Landesmuseum, Bonn,pp. 299e314.

Tite, M.S., 2009. The production technology of Italian maiolica: a reassessment.Journal of Archaeological Science 36, 2065e2080.

Tite, M., Pradell, T., Shortland, A., 2008. Discovery, production and use of tin-basedopacifiers in glasses, enamels and glazes from the Late Iron Age onwards: areassessment. Archaeometry 50, 67e84.

Turner, W.E.S., Rooksby, H.P., 1959. A Study of Opalising Agents in Ancient OpalGlasses throughout Three Thousand Four Hundred Years. In: GlastechnischeBerichte 32K, VII, pp. 17e28.

Wyttenbach, A., Schubiger, P., 1973. Trace element content of Roman lead byneutron activation analysis. Archaeometry 15, 199e207.


Author's personal copy Production technology and replication of lead antimonate yellow glass from...

Documents

Transcript of Author's personal copy Production technology and replication of lead antimonate yellow glass from...