Aimers et al (2011) Handheld XRF analysis of Maya ceramics

[n tbe Terminal Classic period many sites were abandoned and profound changes swept across the Maya world Aimers 2007b.. PACIFIC OCEAN , , I , In the early years of Maya archaeology

Chapter 13

a pilot study presenting issues related

to quantification and calibration

Jim J Aimers, Dori J Farthing and Aaron N Shugar

I ntrod uction

The investigation of archaeologica l ceramics ha s a l ong and varied history with

regard to the analytical instrumentation used (for gcneral examp l es, see Peacock 1970; Bishop er al 1982; Rice 1987; Pollard er al 2007) In recent years newer

applications have been used for the analysis of ceram i c materials as wcll, including

rCP-MS (Fenno et al 2008; M an nin o and Orecchio 20 1 I ) and INAA (G l ascock 1992; Neff 2000) In most cases the motivation to obtain chemical concentrations from archaeological ceramics has been to establish the source of the clay matrix

This has proven possible using instrumentation with l ow detection limits (i c trace

element a n a l ysis techniques such as NAA, I CP, AAS, and WD-XRF) H andheld X-ray fluoresce nce spectrometry was deve l oped in the carly 1 960's (Piorek 1997) but did not enter the world of archaeo l ogy outside of isolated research , until the early to mid 2000's when the instrumentation became more affordable (e.g Uda

et al 2000; Cesareo et al 2004; Id a and Kawai 2005; Newman and Loendorf 2005) With the flourishing use of h andhe ld XRF by non-trained scien ti sts and other researchers who may not be trained in the basic (and advanced) theories of

X-ray fluorescence, its mis se and the misinterpretation of results is prevalent (see

chapter I of this volume for more detai l )

Several papers have recent l y been published dealing with the provenancing

of ceramics using handheld XRF with varied success (e.g., Morgenstein and Redmount 2005; Tagle and Gross 2010; Barone et al 20 I I ; Goren el al 20 I I; Speakman e l al 20 I I) Unfortunately, the 'boxed' calibrations that come with these instruments are not designed to deal with the complex n ature of archaeological

ceramics Ceramics are by nature heterogeneous with numerous compo n ents (such

as temper) all having variable particle size They can have surface alterations and

coatings and over time, the chem i stry of the surface can alter as well In addition, archaeological ceramics often have a ered chemical surfaces related to th eir burial enviro nm e nt Manufacturer calibrat i ons are more geared to modern applications and modern materials that are uniform in makeup , and expec t jng calibrations

424 Jim J Aimers, Don J Fa rthing and Aaron N Shugar

designed for these purposes to be effective with archaeological ceramics is

unreasonable The desire for quantification, whether it be for provenancing studies

or characterization studies , requires the user to create material specific calibrations

(see Hein el al 2(02)

This paper discusses current investigations of Maya ceramics from Belize The focus of this study is not to determine the source / provenance of the clay bodies, but

to investigate the potential for establishing handheld XRF as an on-site analytical

tool for the characterization and potential classification of ceramics based on their chemical signatures The development of an empirical calibration is presented including the process involved in creating reference materials for that calibration

Overview of Maya chronology and pottery

The date of the arrival of people in the Maya lowlands i s currently a matter of debate (see Lohse 2010) but lies somewhere in the Archaic period (8000-2000 B.C ) with maize pollen indicating farming by about 3000 B C ( Pohl er al 1996)

The Preelassic period dates from roughly 2000 B.C to A.D 250 with the earliest well-documented Maya pottery about IIOO-goo B.C in the Cunil ceramic complex

of the Belize Valley (Sullivan and Awe 20 12) By the Late Preela ss ic period (often dated 250 B.C to AD 250) Maya pottery wa s very weU made and styles were widely s pread across the entire Maya lowlands Although most of the significant cultural aspects of Maya civilization were in place by the Late Preelassic, the subsequent Classic period (AD 250-8(0) i s generally considered the height of Maya development The Classic Maya lived in a literate , highly stratified society which produced monumental all and architecture and elaborate polychrome pictorial pottery [n tbe Terminal Classic period many sites were abandoned and profound changes swept across the Maya world ( Aimers 2007b) Dates vary

because different sites were abandoned or transformed at different times from

about A D 750-1050 but the Terminal Classic has traditionaUy been dated to about

AD 800-goo Pottery of the Terminal Classic varies across the lowlands but it

was still well-made with an emphasis on elaborate modeling and incising over

polychromy The Po s tcla ss ic period follows the Terminal Classic and ends at the arrival of the Spanish in the Maya area at about AD 1540 Postelassic pottery also emphasizes incising and modeling and i s typi ca lly well-made The pottery of the Postelassic period is the focus here

Handheld XRF analysis of Maya ceramics 425

The Postclassic period and its pottery

, , , GU A TEMAL A ".-,

:

PACIFIC OCEAN

, ,

I ,

In the early years of Maya archaeology the Postelassic period was neglected as

a period of cultural deeline following the Classic "collapse" , but more recent

research at Postclassic sites has revealed population movement, innovative political strategies increased exchange and commcrciali.mtion , iconographic

innovation, and intense Mylistic interaction (Smith and Berdan 2000, 2(03) A key characteristic of the Maya Postelassic period is evidence of extensive trade ,

e pecially among sites along the Caribbean coasts and on rivers, and involving

s ite s in the northern half of the peninsula such as Mayapan and Chichen Itza ( Figure 13.1 ) A beller understanding of the economic and political milieu of the

Po s telassic would be greatly aided by more detailed documentation of the nature

426 Jim J Almers, Don J Farthing and Aaron N Shugar

and degree of interaction among Postclass ic sites , and o e of ou r m ost informative

.rtifact classes is pottery

(aher Sanders 1960: Fig 4, 5)

7

Handheld XRF analysis of Maya ceramics 427

Aimers has been investigating the Pos tclassic period since his ssertation research

on the Maya collapse and its aftermath (Aimers 2003, 2004c, 2004a, 2007b) and particularly with research on the pottery of two sites that were n t abandoned in the Terminal Classic: Tipu (Aimers 2004a: Aimers and Graham 2012a) and Lamanai (Aimers 2008 ,2009 ,20 10; Aimers and Graham 2012b) In the summer of 2011 Aimers began a pilot study to investigate the chemical variability of a plain red type (Payil Red) and a re lated incised type (Palmul Incised) (see Fig ures 1 3.2 and 13.3) using XRF with samples from the inland site of Tipu, and the site of San Pedro on the island of Ambergris Caye These types are not particularly common but they are widely distributed in the Late Postela sic period, and they are much

more common at coastal s ites and are thought to have been produced along the

coast of the Mexican state of Quintana Roo (e.g , the sites of Ichpaatun, Tancah and Tulum, see Fi gure 13 1; (Sanders 1960: Aimers 2(09) As we discuss later,

the original goal of the research was to identify compositional groupings within

these types which might help in addressing trade and exchange in the Pa ste lassie period We do not expect to link the pots to their production location except in rare cases (sec comment s below) , but we hope that chemical characterization can help

us map the distribution of pottery types better than surface style and form alone These distributions can help in the construction of inferences about Maya pottery production and trade A larger study is planned to follow the pilot study with more stylistic types and samples from more sites

Figure 13.3:

Palmul Incised sherd from San Pedro, Belize, showing the surface inCISing and the red slip This example also has blue pigment which is thought 10 have been distributed from the site

of Mayapan (Mexico)

428 Jim J Aimers, Dori J Farthing and Aaron N Shugar

Major analytical techniques used in maya pottery studies

In the Maya area, a styli stic classification system known as type-variety has

dominated th e st ud y of pottery s ince its introduction from American Southwestern

archaeology in the 1950's and 1960's (Smith , Willey, and Gifford 1960)

Type-variety orga ni zes pottery hierar c hicall y into wares (based on broad charac teristics

of paste and l or surface) gro llps , which are clusters of types (defi ned by a set of

anributes suc h as co lor and decorative treatment ) and varieties which are often

based on single attributes So, Payi l Red and Palmul Incised are types within the Red Pa yil Group ofTulum Red Ware Each of these types only has a si n le variety

because these type s arc macro sco pi ca lly quite homogenou s in pa s te and s urface

treatment - this is one reason they were chosen for the XRF study (sce Cecil 20 I 2 for more detail on the pastes, AA data and petrography)

Type-variety h as been u sed widely because it is a rapid and inexpensive

" Iow - te c h" way to organize the thou sands (so metime s million s) of s herd s that are produ ce d by excavations at s ites in the Maya area (Aimcrs and Graham

20 1 2b) Aimers' research to date ha s involved assessing intera ct ion among s ites

and regions using type-va riety analysis of pouery from hands-on examination of collections from across the Maya lowlands (see e.g Aimers 2004a, 2004b 2007a 2008,2009,2010) Type-variety provides a common lan guage for archaeologists

and h as facilitated the compa ri so n of pottery across sites and re gio n s in addressing issues as varied as c hr o nolo gy function trad e / exchange and c ultural meaning Nevertheless, t ype-varie ty ha s been subject t o a number of import ant c riti c i s ms One of the mo s t important i ss ue s is th e c hara c terization of fabrics (w hich

Mayani sts generall y call pastes) at the ware le vel (Rice 1976) Paste variation

ha s been u sed by some archaeologists as a key discriminating attribute and thus

uscd to make distinction at the highest (ware) level (e.g in Rice's work cited in this chapter) , but it has been considered by others to be a minor factor and occurs

rand o ml y in , for example, type or variety de sc ription s or to c re a t e va ri ti es ( Gifford

1976) Attention paid 10 paste has tended to vary with research questions Those

interested in manufacture and production have tended to pri vilege paste varia tion for the in sigh t it can provide into th ese issues Those interested in co n umer

c o i ce, s t y li s ti c analysis and comparison, or meanin g, have often co n idered pa~lc

variatio n irrel eva nt

Thus type-variety c la ssificatio n is problematized by in cons i s t e cies in the treatment of paste variation th a are n01 weaknesses in the system itself but result

from the fact that like all classifications, type-varicty methods vary according 10 the research questions addressed (A imers 2012a) Still , it is reasonabl e for Mayani'"

to seek g r ea ter accuracy consistency and comparab ilit y in the c a r acte rization

Handheld XR F analys i s of Maya ce r amics 4 29

of paste variation and the oldest established technique for the close examination

of paste variation is petrography (Jones 1986 199 I) Maya petrographic studies

have been surpri;ingly rare in comparison to work in the Old World and to the

amount of research on pottery in the Maya area This is probably because Maya

pottery is stylistically varied , exceptionally elaborate and often well-preserved,

so macroscopic characterizations have been adequate for chronology and broad

comparative studies Petrography is of course time -co nsuming and destructive

which poses a problem with large or complex sample sets Petrographic studies

of pottery have tended to focus on issues related to manufacture production and

distribution (e.g Rice 1977, 1991 , 1996; Cecil 200lb 200la; Cecil and Pugh

2004 ; Howie 2005; Cecil 2(09) One of the challenges for the petrographic study

of Maya ceramics is that the geology of the Maya area is relatively poorly mapped

(see extensive comments about these issue s in Howie 2005: 120- 16 I for Belize)

so until more sampling of geology and clays is done it can be very difficult to

tie pottery to its clay sources Successful petrographic studies have tended to be

focused on a fairly local level (e.g Cecil and Rices work in the Pet6n Lakes;

Howie 2005) where the geology is well known or distinctive, and l or where clay

sampling has been undertaken

Materials science approaches to the study of archaeological ceramics are

advancing rapidly Petrography is of course well established, and recent studies

of archaeological pottery have used XRF (Bakraji el al 2010; Bakraji el al 2006;

Hall 200 I ; Thomas el al 1992), portable XRF (PXRF) (Papadopoulou el 01 2007;

Papadopoulou e l 01 2006; Papageorgiou and Lizritis 2(07), mineralogical analysis

using XRD (McCaffery e l al 2007; Mitchell and Hart 1989; Rasmussen el 01

2009: Stanjck and Hausler 2004; Zhu el al 2004) , trace chemistry determination

by NAA (many, e.g., Glascock 1992 ; Glascock el 01 2004; Hancock el al 1989;

Lopez-del-Rioelal 2009; Olin and Blackman 1989),structural and microstructural

characterization techniques such as SEM (Ow nby el al 2004: Palanivel and

Meyvel 2010) or combinations of various techniques (Marghussian e l al 2009;

Padilla el al 2005; Speakman el al 20 II) The best overview on the use of all

of these techniques in the analysis of archaeological pottery was done by Rice

(1987)

Of the elemental analysis techniques, Mayanists have considered NAA to

be the mo s valuable because of its sensitivity, accuracy, few matrix effects, and

the range of trace elements that can be identified, including rarc earth elements

(Neff 1992:2) The disadvantage of NAA is its cost and the fact that it can only

be conducted in facilities with research reactors In addition, the sample size

required for NAA is quite small, typically a small drilling is all that is required

For thi s reason sample h ete rogeneity could have an adverse effect on the resulting

I

I

430 Jim J Aimers, D o n J Farthing and Aaron N Shugar

chcmislry obtained This issue is recognized by reseruchers and now lar ger samples

are taken and powdered for analysis (see Speakman el 01 20 II for example)

Many of the other elemental analysis techniques (e.g., PIXE) are also expensive

and require equipment that is not easy to acquire This has led to continued

interest in petrography and the use of XRF and XRD because many universities

and museums have access to these instrumental methods Although XRD analysis

does not provide elementa l analysis data insights it compliments other techniques because it provides information on the mineral makeup of analyzed samp l es For

example , Tenorio el 01 (2010) used NAA, XRD , and SEM in a study of pottery

from Laganero, Chiapas, Mexico

Current trends in Maya ceramic analysis

The introduction of a new and readily accessible analytical technique typically

results in optimism about its utility for the investigation of archaeological

problems, For example, Culben and Schwalbe (1987) published an ear ly stu dy

of the application of standard XRF to Maya ceramics from Tikal (see al;o Schwalbe and Culben 1988) This study and others were criticized concerning

issues of precision and especially comparability of results to other studies (Bishop

el 01 1990:543; Neff 1992:4) Recently, ponable and handheld XRF technology

created a similar wave of optimism but critical evaluations did not lag far behind

Shackley (20 I 0) provides the most straightforward critique of handheld XR F on

issues of reliability and vaUdity as we ll as the ·'near religious fervor" with which

the technology ha been embraced by people who are not adequately familiar

with the methodological and interpretive issues involved (Shugar 2009: also

addresses these issues) This is cenain l y the case in Maya studies The Mayanist here (Aimers) learned of handheld XRF relatively recently and was excited by

what appeared to be a fast way to acquire ·'hard" compositional data in the field Like many others he had no background in XRF methodologies and no awareness

of the challenges of sensitivity, precision, accuracy, and comparability of results using this new technology This chapter brings together the differing experie nce

of the three authors in an investigation of these issues in relation to archaeological

pottery

In the study of Maya pottery new analytical techniques after a period of

enthusiastic experimentation, are typically absorbed into research projects which

combine them with more established techniques, In panicular, petrography

combined with quantitative chemical analysis broadens the scope of all

investigation to include both the paste makeup (e.g, the characteristics of the clay

Handheld XRF analySis of Maya ceramics 431

operaroire or specific manufacturing process), and its particular chemistry (as stated above - potentially to source clays) Type-variety despite its problems, is

also sti ll a very useful orga ni zational and descriptive structure for Maya pottery

There is broad agreement that results from multiple techniques of ana l ysis are

always more useful than anyone a l one (Cu l bert and Rands 2(07) In a discussion

of the cha ll enges of characterizing Aegean pottery Day el al (1999: 1034) reached

a sim ilar conclusion :

different sources of chemical variation emphasize the need for the integration of olher information; mineralogical, technological and stylistic; which enables the researcher to attribute differences to provenance or aspects

of clay paste technology The complex interplay of these natural and human

SOUIees of variation means that such analyses cannot take place in isolation

in a " black-box , approach On the contrary, it is imperative for mineralogical

and elemental ana l yses, at least in the Aegean , to be conducted in an integrated

programme which exploits complementary types of archaeological and

analytical information

Pottery variability and the potential of XRF and handheld XRF

Inter-observer inconsistency is always an issue in type-variety, especially for

rare types (Aimers 20 12b) but many types, including the ones discussed in this

chapter , are recognized by experienced specialists with little if no debate So , why

is there a need for XRF and other means of compositional characterization? In the

case of Red Payil Group sherds , the problem is their macroscopic consistency

We know that these types are widely distributed and we assume that like most widespread types , they were made by multiple producers and probably at multiple

locations Pool and Bey (2007:36) note that the "vast majority of [Maya] pottery

was made and consumed locally " (see also Arnold el al 1991) This has been found repeatedly for Maya ceramics, most famously with the Preclassic Sierra Group types which are very stylistically consistent across the entire Maya lowlands This pilot project was designed to see if XRF could detect intra - type

c ompositiona l groups that could be investigated and hopefully confirmed by

other techniques such as petrography SEM and XRD The longer-term thinking

was that if standard XRF would reveal compositional groups in this otherwise

homogenous pottery, handheld XRF would have the potential to do the same The

ability to use handheld XRF on large numbers of samples in the field could allow

4 32 J i m J A l me rs , Do n J F a r t h i ng and A a r o n N Sh u ga r

the establishment of what are essentially technological varieties of Payil Red and

Palmul In c i sed In so me ways, this new portable and handheld technolo gy could

treatment of pa s te in type-variety (se e Ri ce 1979 for an extended di sc us s ion of

thi s i ss ue )

Another rea so n for the int e rest in handheld XRF is that the ability to export

large number of sherds from Beli ze, or any co untry is difficult at be s t , making

traditional analysis difficult , and analysis of lar ge sample groups even morc

problematic Bein g able to tran spo rt the XRF to the field would allow for on-site analysis of the s herds and could help archaeologist direct th e ir re sea rch qu es tion s

ill siw

Benchtop XRF sample preparation and analysis

To investigate the viability of the XRF as a "d i sc riminating " tool, a se lection

of sa mpl es repre se nting the Payil Red and Palmul Inci sed type s were analyzed

for their major and tra ce clement co mpo si tion s in the Department of Geological

Sciences ar SUNY Geneseo All samples were analyzed with a PANalytical AXIOS

Sequential WD - XRF Spectrometer The XRF u ses a 4 kW Rh -a node X - ray source

and both a flow and a sci ntillation detector The now detector is ideally s uited for the analysis of tran si tion clements and the sc intillation det ec tor is ideal for

the analysis of heav y elements A se t of internal c urved crystals ( including the

following options: LiF 200, LiF 220, PE 002, and GE III ) are also used in every analysis to disperse th e X-rays emitted from the sa mple according to their different

wavelengths using diffraction The crystal s are connected to a turret that rotates to

insert one crystal at a time into th e beam path Th e c ry s tal that is selected depends

upon the e lem e nt that i s being analyzed (Tab le 1 3 1 ) The XRF is operated at

vo lla ges that range from 10 kV t o 60 kV and curre nt s th a t range from 10 rnA to

1 25 rnA Typically flow detectors and sc intillation detectors have resolutions below

1000 eV, but when used in unison with a crystal spectrometer, that resolution j",

greatly improved to range from approximately 12 eV ( LiF 220) to 31 eV (LiF 2(0)

(Jenkins 1999 : 1(0 )

All pottery sa mple s were cleaned with water and an ordinary nylon - bristle

toothbrush to remove soil and particulate matter that was loosely adhered to the

pottery and not considered original to lh e pottery body The sa mples were then

c rush e d u s ing a SPEX SamplePrep Mi xe rlMili and a hardened s teel grinding

ca ni s t er equipped with 2 gri ndin g balls The grinding process produced a

powder that passed throu gh an 80-mes h sieve size Thi s was achieved by milling

-Handheld XRF analy sis of Maya ceramics

e pieces of sample -10 grams of sherd material for 3 minutes If after milling larg

remained the sample wa, milled for an additional I to 3 minu

the abundance of large fragments Between each sample the bal

by milling quanz sandbox sand for 3 minutes The cleaning san

and the canister was then blown out with compressed air to femo

sand Small sized particles are easier to fuse into glass beads / d

have a greater amount of surface area and dissolve easier in th

because the small and even-sized particles are easier to hom

compact into a morc coherent Hat - faced pellet with no major n

the sample s urface Both fused beads and well-made pressed pe

for obtaining the best possible XRF analysis because they mininu

which can skew the data and not accurately represent the overa

sherd The powdered materials are also in the ideal form for

(XRD) analysis and samples can be analyzed first with XRD an

ze matrix effects,

II chemistry of the X-ray diffraction

d then the powder

can be re-used in the preparation of XRF samples

Cry~lal name PANal)'lical's o;;ugge~ tions for use

LiF 2 2 0 cry stal U s ed f o routine analy sis for element s

is not as reflective as the LiF 200 but h effect

Upgrade 10 PE (002) curved cry sta Used for e l eme nt between AI and CI

Table 13.1

XR F analyzing crystals and their suggested uses during ana l ysIs

bel wcen V and U It

as a hig c rdi <; pe rs ioll

ed as fu s ed glass

Major element chemistries were measured on samples prepar

beads (see Burke el al ( 1998 ) for more infomlation on this techni

of a glass bead entails melting a flux (in our case a lithium bo

which dis s olves the powdered s ample at high temperatures Th

i s cooled to produce an extremely homogenous glass bead th

for avoiding matrix issues associated with particle size

0 5 ± 0.00 I grams of powdered sample were mixed with 6 ± 0 00

flux The fusion flux was compri s ed of 33 % lithium metabo

tetraborate and I % lithium bromide wbich is a non - wettjng

s ticking of the glas s bead to the platinum mold Once the sam

were mixed together the mixture was poured into a platinum c

que) The creation rate-based flux)

e molten mixture

at is exceptional and mineralogy

I grams of fusion rate 66 % lithium agent that deters pie and the flux rucible U s ing an

433

I ,

I

, ,

,

,

I

434 Jim J Aimers, Don J Farthing and Aaron N Shugar

automated fusing machine, the cr u cible was evenly heated to IISO o e, mixed

well (while avoiding the creation of bubbles) , and then the molten mixture was poured into a platinum mold When coo l , the resultant glass bead was analyzed with a PANalytical AXIOS XRF The ana l ytical program , IQ+ used internal

standardizatio n to quantify the element concent r ations The initial standa rdi zation

was based upon fundamental parameters that were imp roved by ana l yses of

laboratory-generated standards containi n g k n own amounts of major elements (the IQ+ suite) The initial standardiza ti on is regularly checked by week l y ana l yses

of two g l ass monitor standards (BGSMON and AUSMON-F) as well as the

monitor for drift, background levels and quantitative accuracy The two glass

monitor samples came with the XRF AUSMON-F is a drift monitor standard that

was s pecifically chose n to coordinate w ith measurements of s il icate materials and

International and Brammer Standard Our current accuracy for major elements is,

± I wt % for high concentration major clements After every analysis we manually inspected each spectrum to make sure that the "search-and-ma t ch" function of the

ana l y ca l program id entified all the peaks We a l so force the ana l ytica1 program to

strips Br from the results Sr is a constituent of the nux and is never considered as

a major element Sr must be stripped from the analysis because it interferes w ith

the aluminum peaks; the position of the bromine L-lines at 1,4 80.4 eV ove rlap with the a luminum K-lines l ocated at I ,486.3 and 1,486.7 eV Since the quantity

ofBr in the sample was known, stripping it from the spectrum was trivial and did

not have a detrimental effect on the AI ana l yses Glass beads are ideal materials

for major element analyses because they are extremely homogenous and wil l not generate analytical errors due to grain sizes or preferential grain orientation as is

the case when mica or clay is present Isee Jenkins e1 al (1995) for a discussion of

grain-size related errors I In general, the concentration of any element as it relates

to the intensity of the X-ray peak is mathematically described as:

(I)

C represents the concentratjon of a specific element , K is the calibration consta nt

determined from the analysis of standards, 1 is the intensity of the peak and M

is a correc ti on factor that accounts for matrix effects The M value accounts and corrects for a variety of parameters including particle size, particle size distribution

crystallographic nature, and grain orientation (see Rousseau 2006 for insight on

the calculation for M) 1n this simple equation , M is potentially the greatest source

of error because of the variety of parameters it incorporates Vitrifying a sample