Glasgow Theses Service http://theses.gla.ac.uk/ theses@gla.ac.uk Clark, Peter Alan (1994) Isochron methods for luminescence dating in archaeology. PhD thesis http://theses.gla.ac.uk/7225/ Copyright and moral rights for this thesis are retained by the author A copy can be downloaded for personal non-commercial research or study, without prior permission or charge This thesis cannot be reproduced or quoted extensively from without first obtaining permission in writing from the Author The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the Author When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given ISOCHRON METHODS FOR LUMINESCENCE DATINGIN ARCHAEOLOGYPeter Alan Clark BTech (Hons) (Bradford)A thesis presented in part fulfillmentof the requirements of the degree of PhDUniversity of GlasgowScottish Universities Research and Reactor Centre1994ABSTRACTThis work aims to contribute to the solution of a problem relevant to archaeology andmicrodosimetry, namely the TL dating of destratified archaeological ceramics. It isapparent that museums hold a large body of archaeological material excavated fromimportant sites during the early part of this century. Absolute dating of this corpus ofmaterial would be extremely valuable to archaeology in allowing historic hypotheses tobe tested.The background to this problem has been presented within the context of the historicaldevelopment of thermoluminescence dating and of dosimetry concepts. A critical reviewof previously proposed strategies for dating material from unknown gamma contexts ispresented, and then a general model using an isochron approach is introduced. Theisochron model is then developed explicitly and theoretical simulation is used to exploredifferent microdosimetric systems. Two main isochron solutions have been identified,the grain size (or alpha-beta attenuation) and alpha efficiency isochrons.An experimental approach was designed, to evaluate the validity of the grain sizeisochron predictions, using model matrices. Isochron behaviour was clearly observedin both cases. The first system, based on microcline feldspar grains in a high dose ratematrix, produced an age estimate consistent with the known age. The secondexperiment, using fluorite grains in a lower dose rate matrix, produced a highly coherentdata set. However, the age estimate obtained was greater than the expected value by afactor of approximately 3. The possible reasons for this are discussed in detail. A briefexperiment carried out subsequently on duplicate portion of this matrix gave results thatare consistent with predicted values. Thus, it would appear that the originalexperimental design and method were valid and that the isochron approach does indeeddeserve further investigation.TABLE OF CONTENTSI.INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 11.1 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 21.2 INTRODUCTION TO THERMOLUMINESCENCE 51.3 HISTORICAL DEVELOPMENTS 61.4 TL DATING: BASIC REQUIREMENTS AND AGE EQUATION.. 131.5 STORED DOSE EVALUATION. . . . . . . . . . . . . . . . . . . . . . . . .. 141.6 DOSE RATE EVALUATION 161.7 SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.DOSIMETRY 202.1 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 212.2 BASIC DOSIMETRY DEFINITIONS . . . . . . . . . . . . . . . . . . . . . .. 212.2.1 Energy Imparted . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 212.2.2 Absorbed Dose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 222.2.3 Stopping Power 222.2.4 Charged Particle Equilibrium 222.3 INTRODUCTION TO RADIOACTIVITY . . . . . . . . . . . . . . . . . . .. 222.4 RADIOACTIVE DECAY LAWS 232.5 MODES OF RADIOACTIVE DECAY 252.6 CALCULATION OF INFINITE MATRIX DOSE RATES. . . . . . . .. 262.7 SOURCES OF NATURAL RADIOACTIVITY 282.8 DOSIMETRY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 352.8.1 Introduction 352.8.2 Alpha Dosimetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 352.8.3 Beta Dosimetry. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 402.8.4 Gamma Dosimetry 422.9 MEASUREMENT OF DOSE RATE. 452.9.1 Alpha counting. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 462.9.2 Beta Counting 462.9.3 Thermoluminescence Dosimetry (TLD) 472.9.4 Gamma Spectroscopy 492.9.5 Neutron Activation Analysis (NAA) 502.9.6 Fission Track Analysis 512.10 EFFECT OF H20 512.11 RELATIVE DOSE CONTRIBUTIONS . . . . . . . . . . . . . . . . . . . . .. 522.12 FORMULATION OF SPECIFIC STRATEGIES FOR ROUTINEDATING OF SAMPLES FROM KNOWN GAMMA FIELDS . . . . .. 532.12.1 Introduction 532.12.2 Fine grain dating . . . . . . . . . . . . . . . . . . . . . . . . 542.12.3 Quartz inclusion dating 542.12.4 Feldspar inclusion dating .. 552.13 PROPOSED STRATEGIES FOR DATING SAMPLES FROMUNKNOWN GAMMA FIELDS . . . . . . . . . . . . . . . . . . . . . . . . . .. 562.13.1 Introduction 562.13.2 Estimation 562.13.3 Reconstruction 562.13.4 Subtraction Methods 572.13.5 Zircon Methods 592.13.6 Decay Methods 602.13.7 Mejdahl's Approach 612.14 SUMMARY......................................... 623.ISOCHRONS I - FORMULATION AND SIMULATION 633.1 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 643.2 GENERALISATION OF THE LINEAR AGE EQUATION . . . . . . .. 643.3 FORMULATION OF AN ISOCHRON APPROACH 683.4 ABSORBED DOSE FRACTIONS AND ATIENUATIONFACTORS................................. 693.5 SIMULATIONS FROM THE ISOCHRON MODEL . . . . . . . . . . . .. 733.5.1 Introduction..................................... 733.5.2 The Isochron Simulation Program . . . . . . . . . . . . . . . . . . . .. 733.5.3 Investigation of the model: Simulated isochron plots 753.5.4 Investigation of the effect of matrix water contentvariation 933.6 ERROR SENSITIVITY OF THE MODEL . . . . . . . . . . . . . . . . . . .. 963.7 SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1034.ISOCHRONS II - EXPERIMENTAL VERIFICATION 1054.1 INTRODUCTION. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1064.2 DESIGN CONSIDERATIONS FOR EXPERIMENTAL MATRICES 1074.2.1 Objective..................................... 1074.2.2 Materials and constraints 1084.2.3 Matrix Effects 1104.2.4 Grain Requirements. . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1134.2.5 Planned Matrices 1154.3 EXPERIMENTAL PROCEDURES AND DEVELOPMENTS 1164.3.1 Introduction 1164.3.2 Matrix Preparation. . . . . . . . . . . . . . . . . . . . . . . . 1164.3.3 Grain Preparation 1204.3.4 Alpha-source Calibration Experiments . . . . . . . . . . . . . . . . . 1254.4 MODEL MATRIX I: MICROCLINE GRAIN SIZE ISOCHRON .. 1284.4.1 Experimental Plans 1284.4.2 Materials 1294.4.3 Exposure Details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1304.4.4 Extraction of Grains 1314.4.5 Measurement of Stored Dose. . . . . . . . . . . . . . . . . . . . 1334.4.6 Results....................................... 1344.4.7 Analysis and Discussion 1364.5 MODEL MATRIX II: CaF2 GRAIN SIZE ISOCHRON . . . . . . . . .. 1384.5.1 Experimental Plan. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1384.5.2 Materials . . . . . . . . . . .. 1394.5.3 Exposure Details. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1414.5.4 Extraction of Grains 1414.5.5 Measurement of Stored Dose 1434.5.6 Results...................................... 1434.5.7 Analysis and Discussion. . . . . . . . . . . . . . . . . . . . . 1474.5.7.1 Estimation of Dose Rates. 1474.5.7.2 Source Calibrations. . . . . . . . . . . . . . . . . . . . . . . .. 1534.5.7.3 Effect of Stopping Power Variations. . . . . . . . . . . . . 1534.5.7.4 Other considerations 1554.6 SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1565.DISCUSSION AND CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1575.1 SUMMARY. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1585.2 FUTURE DEVELOPMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . .. 1605.3 CONCLUSIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1616.ADDENDUM 1626.1 CONCLUSIONS. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 165REFERENCES 166TABLE 2.1TABLE 2.2TABLE 2.3TABLE 2.4TABLE 2.5TABLE 2.6TABLE 2.7TABLE 3.1TABLE 3.2TABLE 3.3TABLE 3.4TABLE 3.5TABLE 4.1TABLE 4.2TABLE 4.3TABLE4.4TABLE 4.6TABLE 4.7TABLE 4.8TABLE 4.9TABLE 4.10TABLE 4.11TABLE 4.12TABLE 4.13TABLE 4.14TABLE 4.15TABLE 4.16TABLE 4.17TABLE 4.19TABLE 6.1TABLE 6.2TABLE 6.3List of Tables238UDecay Scheme 29235UDecay Scheme ·30232ThDecay Scheme 314°K Decay Scheme 32Dose rate conversion factors (mgy.a') (From Aitken,1983) 32Infmite matrix dose rates (mgy.a') Calculated for a typical matrix andeffective dose rates for fine and quartz inclusion grains 33Electron stopping powers, relative to air, calculated from Cross(1968). 43Alpha absorbed dose fractions (after Bell, 1980) 70Beta absorbed dose fractions (after Mejdahl, 1979) 71Gamma absorbed dose fractions (after Mejdahl, 1983) 72Values of mineral activity used to investigate their effect on theisochron plots 78Values of matrix activity variation used to investigate their effect onthe isochron plots 82IAEA and CANMET standard reference values 111Model matrix I: microcline grain size isochron 115Model matrix II: CaF2 grain size isochron 116Observed and expected beta dose rates from standard referencematerials 117Actual composition of model matrix I 129Beta dose rates for model matrix I compared at different stages of theexperiment 129Radionuclide concentrations of model matrix I, measured by NAA 130Summary of results from model matrix I experiment 134Actual composition of model matrix II 139Beta dose rate for model matrix II compared at different stages of theexperiment 140Radionuclide concentrations of model matrix II, measured by NAA 140Results of model matrix II experiment: test run 144Results of model matrix II experiment: full run 145Dose rate values for model matrix II 145Excess internal dose rate required to obtain known age 149Infinite matrix dose rates calculated for (a) no radon loss and (b) 100% radonloss from microcline/matrix 150Calculated excess dose rates required 151Beta dose rate for model matrix II compared at different times 163Calculated effective dose rate for 100 micron fluorite grains. 164Stored dose evaluations 165List of FiguresFigure 1.1 The additive dose method for evaluating the stored dose. . . . . . . . .. 15Figure 2.1 Energy dependence of mass absorption coefficients of waterrelative to soilFlow diagram of the isochron simulation program. . .The grain size isochron .The alpha efficiency isochron. .(a) Effect of internal activity - Grain size isochron .(b) Effect of internal activity - Alpha efficiency isochron(a)i-iii Effect of matrix variations - Grain size isochron .(b)i-iii Effect of matrix variations - Alpha efficiencyisochron 86-88Summary of matrix variation effects. . . . . . . . . . . . . . . . . . . . . . .. 89(Activity free) grain size isochron in a potassium rich matrix 91(Variable activity) grain size isochron in potassium rich matrix 92(a) Effect of different water contents - Grain size isochron. 94(b) Effect of different water contents - Alpha efficiencyisochron. 95Figure 3.10 (a)i-vi Error sensitivity - Grain size isochron 97-99Figure 3.10 (b)i-vi Error sensitivity - Alpha efficiency isochron 100-102Figure 4.1 TL response of micro cline (MC1) feldspar in the mGydose region. . 113Figure 4.2 (a)i-ii Expected grain size distribution calculated from Stoke'sLaw 123Figure 4.2 (b)i-ii Expected grain size distribution (after Zimmerman, 1971). . .. 124Figure 4.3 TL output versus sample mass for alpha radiation. . 126Figure 4.4 TL output versus sample mass for beta radiation. . 127Figure 4.5 TL output versus sample mass for gamma radiation 127Figure 4.6 Alpha efficiency versus sample mass. . 128Figure 4.7 Model matrix I : Isochron plot 135Figure 4.8 Model matrix II : Isochron plot. 146Figure 3.1Figure 3.2Figure 3.3Figure 3.4Figure 3.4Figure 3.5Figure 3.5Figure 3.6Figure 3.7Figure 3.8Figure 3.9Figure 3.953747677798183-85I. INTRODUCTION11.1 INTRODUCTIONThis thesis outlines and evaluates a general isochron formulation of thermoluminescence(TL) dating and discusses the extent to which this may contribute to overcoming someof the limitations of current techniques. An introduction to the general problem beingconsidered is followed by a review of the basic principles and historical developmentsof TL dating. The manner of formation of the isochron approach is then given beforethe details of the experimental verification are described. A summary of the approachesemployed in this research is presented with a discussion of the results obtained and theirrelevance to archaeological applications and dating methodologies. Areas for furtherresearch are identified.The development of archaeological studies towards a more scientific approach has leadto diversification into a multidisciplinary subject within which there is an ever growingnumber of specialist areas. Subjects such as artefact analysis, the study ofenvironmental change, social, economic, and technological development, fieldarchaeology and many others covered by the term archaeology, or archaeologicalsciences, are all linked by a common interest. This is usually "How old is it?", "Whendid it happen?" or perhaps some other variation along these lines. This is because oneof the most fundamental requirements for archaeological studies is a reliablechronological framework.Several dating techniques, developed for the geological sciences, find some measure ofapplication to archaeological problems. However, the great differences in the timescalesinvolved means that, especially for radiometric methods, only the oldest periods ofarchaeology can be considered. There is always room for optimism, for example whiledates have been obtained by conventional 4°Kt°Ar methods for the early hominid period(Michels, 1973) the development of an 40ArP9Ar method may have the potential to dateyounger samples (Miller, 1970).Some of the dating techniques which have been useful to geology are not generally2applicable to archaeological dating problems and so more specific methods weredeveloped. These included dendrochronology, archaeomagnetism, radiocarbon, andhydration layer dating, for example. A useful summary of scientific dating techniquesin archaeology is provided by Aitken (1990). It is not of particular relevance to discussthem in any detail here except to highlight some of the common drawbacks that they allsuffer from to varying extents. These are the problems of the availability of suitablesample material (and in sufficient quantities) from the site and also the association ofthe sample within the archaeological context with respect to the relevance of the eventbeing dated and the subsequent interpretation of the date.The advantages and limitations of the TL method for dating archaeological samples willnow be examined with reference to the points made above. The types of materialdatable by this technique, ceramics, burnt stones and sediments, are ubiquitous on manyarchaeological sites so the problem of insufficient amounts of material is unlikely tooccur. The lifetime of use of a ceramic vessel is likely to be short compared to itsarchaeological age, so the event being dated, the firing of the pottery, is relevant. Itmust be noted, however, that the stratigraphic credentials of the context providing thesamples must be ensured as there is always the possibility of residual material givingrise to conflicting dating evidence. The TL dating of burnt stone material, hearth stonesfor example, gives the date of the last heating of the stones and therefore the last useof the hearth. This is obviously an important event since it is very likely to signify thelast date of occupation at that particular location. More recently developed applicationsof the technique in the area of dating unburnt sediments may allow site formationprocesses to be investigated. Thus the nature of materials suitable for TL dating andtheir archaeological interpretations within their contexts are substantial assets of thetechnique.Despite these favourable aspects there are a number of limitations to TL dating. Apartfrom the technical challenges of efficient sample preparation and accurate measurement,a further problem is the current necessity to make on-site environmental radiationmeasurements although others, such as precision, cost, the possibility of hiddensystematic errors (microdosimetry and fading), may also be important. The absence of3these measurements means that part of the age equation is unknown and so the datingaccuracy is limited. It follows then that material that has already been removed fromits original context, museum held collections for example, is also inaccessible to themethod.At this point it is worth making the distinction between destratified and unstratifiedsamples. The former are taken as being those samples that have been deliberatelyremoved from their archaeological contexts and as such may have some level ofinformation pertaining to that context associated with them. Unstratified samples,however, are regarded here as being isolated, often surface, finds with no associatedcontextual information. They therefore normally have limited potential to provide anysignificant contribution to our understanding of the past. Destratified samples couldinclude material from archaeological excavations that have already taken place that didnot include a programme of TL dating in the planning of the research strategy. If thistype of material became datable then it may be likely to assume increasing importancein the future since sites and funding for new archaeological excavations are both limitedresources.Similarly, if material from museum collections became accessible to reliable absolutedating this would open many interesting avenues of research. Given sufficient precisionit would be possible to assess traditional age estimates, derived by archaeological means,from major ceramic assemblages without reference to traditional assumptions. On awider scale this has implications for testing archaeological models of development ofmaterial culture, which despite increasing levels of sophistication and complexity ,nonetheless retain key concepts from the nineteenth century three age system ofThomsen and Worsaae. Bintliff (1984) warns of the ever present danger "...thatinsufficient care is taken to validate such models against the complexities of the data;and at the worst, stage schemes of ... development are supported by reference to othertheoretical work rather than constantly being tested against case studies. ". Absolutedating provides a fundamental means of conducting such case studies, and therefore canbe regarded as having a significant methodological contribution to make to contemporaryarchaeology.4One particular area of interest here is that of the problem posed by typologicalsequences of pottery. The early archaeology of the nineteenth century, and earlier, wasin the main carried out by antiquarian "barrow-diggers" who effectively plunderedarchaeological sites for their riches. Apart from recovering impressive amounts of goldand silver, these activities also removed a significant proportion of the "best" (that is,most complete) finds from their archaeological contexts. Some of this material, suchas funerary urns, was arranged into typological sequences on the basis of vessel formand complexity of decoration. The intervening decades have in some cases seen thesetypological sequences become generally accepted as chronological sequences with littleor no supporting evidence to justify such a transformation. It seems that thesedevelopments of diversity in pottery styles, for example, are often seen as only takingplace in a single dimension, that of time, with the possibility of a contribution due tospatial variations receiving little attention. The validity of this type of assumption isobviously of prime importance to the interpretation of artefactual material that has been,and still is, "dated" by comparative study of these early collections held in museums.It is also a problem that could offer suitable sample material for TL dating but cannotbe dated by current techniques due to the destratified nature of these sample.The aims of this research are to investigate an isochron method of TL dating that willallow material to be dated that is currently inaccessible to routine techniques. Someareas of the possible archaeological applications have been mentioned above, but it isalso important to note that this work also tests the single phase microdosimetricassumptions common to all TL dating which will be discussed in Chapter 2. A briefnote on the principles of TL dating is given now followed by an account of thehistorical developments within the subject, leading to an introduction to TL dating.1.2 INTRODUCTION TO THERMOLUMINESCENCEThermoluminescence is the emission of light from an insulator or semiconductor whenit is heated. It is not the light spontaneously emitted from a substance when it is heatedto incandescence. Thermoluminescence is the thermally stimulated emission of lightfollowing the previous absorption of energy from radiation.5There are three essential requirements for the production of thermoluminescence.Firstly, the material must be an insulator or a semiconductor, metals do not exhibitluminescent properties. Secondly, the material must have at some time absorbed energyduring exposure to radiation. Thirdly, the luminescence emission is triggered by heatingthe material. Once heated to excite the light emission, the material cannot be made toemit thermoluminescence again by cooling the sample and reheating. In order to re-exhibit the luminescence the material has to be re-exposed to radiation, after which,heating the sample will again produce light.Luminescence emission is explained by the transfer of energy from radiation to thecharge carriers of the solid, exciting the electrons from a ground state to an excitedstate. The emission of a luminescence photon takes place when an excited electronreturns to its ground state. An excited electron can also become trapped at a metastableenergy level between the ground and excited states where it will remain until it is givenenergy to return to the excited state from where it can undergo a normal transition backto the ground-state. This excess energy required is called the trap depth. Thus,thermoluminescence is a process by which electrons are released from traps by thermalstimulation to produce light emission.1.3 HISTORICAL DEVELOPMENTSInvestigations into luminescence and related phenomena have been reported since theseventeenth century, following on from the "observations of a diamond that shines inthe dark" (Boyle, 1664). The early application of TL was largely confined to geologicaluse as a method of mineral identification (eg Deribere, 1938). These studies weregreatly facilitated by the technique of measuring the TL as glow curves, ie plots of thelight output as a function of temperature, or time, during sample heating. This methodwas originally developed by Urbach (1930) and modified by Randall and Wilkins (1945)and used by them in studies relating phosphorescence and TL to determine therelationship between TL glow shape and trap depths.In the following years a wide range of new applications of TL were studied, particularly6at the University of Wisconsin (Daniels et al, 1953). Here, the potential of TL radiationdosimetry, ie where the intensity of the TL output is proportional to the ionisingradiation dose received, was being evaluated alongside the continued use in a geologicalcontext for the identification of minerals, differentiation of particular strata based ondistinctive glow curve shapes, and the study of the TL from a large number of rocks.The spatial heterogeneity of the TL of rocks was demonstrated by this group byphotographing samples during heating. This was also an important observation forarchaeological studies as similar results were later obtained for pottery slices (Fremlinand Srirath, 1964).At this relatively early stage of development of the subject it was thought that the primesource of the trapped electron population was the alpha particles from trace amounts ofuranium and thorium and that these alpha particles were responsible for trap creation andhence the growth of TL signals and sensitivity with time. This lead the Daniels groupto postulate that given certain limitations and assumptions it should be possible toestimate the date at which rocks and ancient pottery were last heated to hightemperatures.The suggestion of a method of dating ancient pottery by TL was initially followed upby the Universities of California (Kennedy and Knopff, 1960) and Bern (Grogler et al,1960). The latter group identified, as had Daniels et al earlier, that the low temperatureTL peaks present in the glow curve after artificial irradiation of a sample were unstableand therefore not present in the natural glow curves. It was also shown that byannealing an artificially irradiated sample before measurement the glow curve producedwas of almost the same shape as the natural curve.In the early 1960's the Oxford group critically examined the suggestion by Daniels etal Cop cit) of trap creation by alpha bombardment being a suitable basis for a methodof age determination and found it to be impractical on the grounds that "damage-traps"were indistinguishable from those initially present (Tite and Waine, 1962). In any case,it was later concluded that for archaeological samples the percentage increase in thenumber of high temperature traps since firing was less than 1% ie that over an7archaeological timescale the number of traps was effectively constant. The observationsof supralinearity of TL response with beta dose and enhanced sensitivity that wereoriginally thought to also result from trap creation were later reinterpreted as being dueto competition between different types of trap (Tite, 1966, 1968, Aitken, 1968). Notonly did the Oxford group identify that a dating method based on the filling of traps wasthe way forward but also recognised and investigated several other important factors.Despite improvements to instrumentation (eg the heating system which had beenidentified as a major source of inaccuracy and in the detection system where theselection of optimum filters reduced the interfering signals from blackbody radiation)an initial dating programme gave poor results suggesting that the method based on themeasurement of the number of traps filled showed only a rough age dependence.One of the most significant advances was the observation that if a sample was glowedin a nitrogen atmosphere, instead of in air, then a sample of zero age would producezero TL, an essential requirement for a successful dating technique (Tite and Waine,1962). This procedure had the effect of suppressing the non-radiation induced"spurious" signals that had been attributed to the effects of grinding and gas adsorptionand had previously been a prime cause of concern (Tite and Waine, op cit , Aitken etai, 1963, 1968, Lewis, 1968).Further refinements in the techniques of sample preparation were adopted after the effectof optical bleaching of the natural TL signal of stored aliquots of samples was observed(Aitken et al, 1963). It was found that, for the particular sample studied, allwavelengths were found to reduce the TL signal across the whole temperature of theglow curve by varying amounts. It was consequently concluded that samples should behandled in semidarkness. A second dating programme was carried out at this timeincorporating the modifications noted above. The results obtained were far moreencouraging with strong evidence for a linear relationship between the age of the sampleand its specific TL. For the first time absolute ages could be calculated since the TLsusceptibility had been measured with a calibrated beta source. In comparison to the"known" archaeological ages the TL ages were low by a factor of five.8Experimental work (Fremlin and Srirath, 1964, Fremlin, 1968, Srirath and Fremlin,1968) regarding the spatial homogeneity of the TL characteristics of pottery showed thatthe main luminescence was often confined to a small or moderate number of well-defined crystals. This supported the idea that this could be one of the possible reasonsfor the low ages obtained by Aitken et al since it was also a possibility that theconcentrations of radioactivity, as well as the luminescent regions, were non-uniformlydistributed.The situation was further complicated when, in the light of earlier work (Moorehead andDaniels, 1957), the relative efficiency at trap filling of alpha particles compared withbeta and gamma radiation was considered, and found to be about an order of magnitudeless effective. Taking this into account, the ages from the test programme wererecalculated and although found to be in closer agreement with the archaeological agesthey were not yet sufficiently close to be regarded as being accurate.The evidence of reduced alpha efficiency increased the potential impact of theaforementioned spatial heterogeneity. The implications were not only that separatedmineral fractions should be examined, but also that since the alpha contribution to theeffective dose rate to the sample had been substantially reduced the beta and gammadose rates became more important. The assumption that the sherd and soil had the sameradionuclide concentrations became more significant so that the need arose for itsmeasurement in both, rather than just the sherd. Approaches to overcome this problemof dosimetry were suggested by the results of Fleming (1966) which showed that whilethe radioactivity was predominantly concentrated in the clay matrix, it was thecrystalline extract that gave the greater light output. The conclusion was that largequartz grains would not be affected by the short range, inefficient, alphas and hencereceive only a beta plus gamma dose. Additional consideration of the ranges of betasand gammas led to the formulation of the "quartz inclusion" method which employeda separated quartz fraction of 100-200~m grains for which the alpha contribution isnegligible and so the annual dose consists only of the beta dose from radioactivitywithin the sherd, the gamma dose from the surrounding burial soil and smallcontribution from cosmic rays (Aitken et al, 1968). A preliminary test of this method9(Fleming, op cit) gave absolute ages that were all within 20% of the "known" ages, thusgiving strong support for the validity of the assumptions made.An alternative approach employing the fine grains, 1-8IJ.m, from the sherd was alsosuggested (Zimmerman, 1967). Grains of this size would have suffered negligibleattenuation of the alpha dose and received full sherd beta and soil gamma doses. Theintroduction of the alpha dose into the age equation and its low efficiency to produceTL required the alpha efficiency to be measured for each sample. The results of a testprogramme of this method were also highly encouraging, providing further confirmationthat the techniques used and the principles involved were valid (Aitken et al, 1968).These two approaches were soon consolidated into established dosimetric models(Fleming, 1970 and Zimmerman, 1971).The observation of dose dependent sensitivity changes in the 110°C peak of quartz(Fleming and Thompson, 1970) led to the subsequent formulation of the "predose"method of dating (Fleming, 1973). Here the change is dependent on the previousirradiation history of the sample. The dosimetry does not involve the directmeasurement of the stored TL due to the exposure dose but requires the measurementof the change in sensitivity produced by that dose. The particular advantage of thismethod is high sensitivity, allowing young samples (within the last 1000 years) to bedated. Its drawbacks are early saturation, a complex and poorly understood mechanism,diverse behaviour from real samples and an increasingly involved methodology. Thisincludes additional measurements to correct for radiation quenching (Aitken and Murray,1976; Aitken, 1979), problems caused by non-linearity due to saturation effects andinvestigation of the thermal activation characteristic (TAC) (Fleming and Thompson,1970; Bailiff, 1983a). The predose method has been successfully applied to theauthenticity testing of art ceramics etc. (Fleming, 1979 and Archaeometry) andpromising results have been obtained from archaeological material (Bailiff, 1983b,1985).Another approach explored to overcome the problems associated with high temperatureTL, especially when the signals are small, was that of the phototransfer technique10(Schlesinger, 1965). In this case, if a sample is illuminated with ultra-violet light priorto measurement then some of the electrons from the deeper ("donor") traps are evictedsome of which become retrapped in shallower ("acceptor") traps. The method ofphototransferred TL (PTIL) has also been suggested as a means to overcome anomalous(short term) fading in zircons (Bailiff, 1976). Despite the potential advantage ofenabling the dating of materials such as arragonite, that changes its structure to calciteon heating during high temperature glow curve measurement, the extremely complexbehaviour of P'ITL (Bowman, 1979) seems likely to prevent it from becoming a routinemethod of dating.Once the dosimetric models had been established, mainly by the Oxford group, their useand development spread to other laboratories. The Riso laboratory in Denmark, forexample, developed feldspar inclusion methods. Their dating programme includedsamples of pottery sherds, bricks, burned stones, and burned clay that were measuredusing an automatic TL reader (Mejdahl, 1972). The development of the automatic TLsystem was regarded as being an essential step towards large scale routine datingprogrammes. The widescale acceptance of feldspar inclusions as suitable minerals forTL dating was hampered by the early observation of short term ("anomalous") fadingof the TL signal of feldspars of volcanic origin (Wintle, 1973). However, a great dealof work has since been carried out (particularly in Denmark) to exploit the advantagesof feldspars (greater brightness, higher internal radioactivity, greater reproducibility, etc.)over quartz (Guerin and Valladas, 1980; Mejdahl and Winther-Nielsen, 1982; Mejdahl,1983; Sanderson et al, 1988). The latter work, on vitrified forts, being the ftrst UKfeldspar dates, after the quartz had been shown to have severe supralinearity of doseresponse. The technique of feldspar inclusion dating is now standard practice within theRiso and SURRC laboratories (Sanderson and Clark, 1987).The seventies saw great diversification and development of methods to include non-ceramic materials of both archaeological and geological interest. A review article byWintle (1980) provides a useful account covering burnt stones (including flint), volcaniclavas and associated burnt sediment margins, calcite, bone and shell to varying degreesof success.11Perhaps the most significant development was the potential to date unburnt sediments.Wintle and Huntley (1979) described a study of the zeroing effect of exposure tosunlight during weathering and transport, which followed on from work in the USSR (egMorozov, 1968; Shelkoplyas, 1971 and others) and China (Li et al, 1977). Thetechnique was rapidly developed to allow the direct dating of loess (wind blown sand)deposits (Hutt et al, 1979, Wintle, 1981). The observation of a "residual signal" (ie TLsignal present in a zero age sample) that was not removed by sunlight exposure, and thefact that there maybe a sensitivity change after the first glow measurement, requiredmodification of the methods used to determine the stored dose (Mejdahl and Wintle,1984). These factors plus the realisation that a thermal approach to releasing the chargestored since the sunlight zeroing event may not necessarily be appropriate were themotivation behind a new research field. This was the optical dating of sediments, firstreported by Huntley et al (1985), where the stimulating energy was provided by anargon-ion laser. The majority of this work has been carried out using monochromaticor broad band continuous light sources and there is a need for fundamental spectroscopy.However, optically stimulated luminescence had previously been reported in a dosimetrycontext (Rhyner and Miller, 1970) and photostimulated thermoluminescence (PSL) hadbeen used to study the nature and properties of traps and centres in alkali halides(Fieschi and Scaramelli, 1968). Recent developments in this field include ongoingresearch at the SURRC laboratory involving analyses of PSL spectra and investigationsof charge relaxation mechanisms through time domain studies. The use of light emittingdiodes as an alternative, and cheaper, light source has also been suggested (Poolton andBailiff, 1989, Aitken and Xie, 1991). These experimental developments havecorresponding advances in the theoretical and methodological aspects and theirapplications. This does not mean that the interest in the more "traditional" areas of TL(archaeologicaJ/geological dating, dosimetry, models, spectral studies etc) and associatedphenomena (phosphorescence and electron spin resonance etc) has waned at all. Thecurrent state of luminescence studies is readily witnessed by the breadth and depth ofcontributions presented at recent UK and International Specialist Seminars.Some of the aspects introduced in the last section will be dealt with again, but in greaterdetail, later in the thesis. At this point, however, it is appropriate to include a concise12introduction to TL dating (with specific reference to archaeological ceramics).1.4 TL DATING: BASIC REQUIREMENTS AND AGE EQUATIONThe requirements of any absolute dating method are a measurable parameter that varieswith time (preferably in a linear fashion), a start mechanism, and a means of calibratingthe stored signal. These are examined below (after Wintle, 1980 and Bailiff, 1988).The phenomenon of TL provides such a basis; under the influence of ionising radiationa TL sensitive material stores a fraction of the energy imparted. The energy storage isfinite and the process of energy acquisition within this storage system is cumulative.For a number of types of material (though not all) the storage is stable overarchaeological time. The driving force of this particular dating mechanism is ionisingradiation. This results from the decay of the naturally occurring radioelements, uranium,thorium and potassium 40, which produce a radiation field of alpha, beta, and gammarays. The half-lives of these radioisotopes are sufficiently long that over archaeologicaland recent geological timescales they provide an effectively constant dose rate. Thisassumes that the system is in a state of radioactive equilibrium.The start mechanism in TL dating is the zeroing event by which all previously storedenergy is removed. There are a number of zeroing mechanisms for TL, eg heating tohigh temperatures in the case of the firing of pottery or the repeated heating of thestones of a hearth; exposure to sunlight for aeolian sediments; or crystallisation egcalcite formation. After the zeroing event has taken place the development of the TLsignal recommences and (generally) continues in a linear fashion until the next zeroingevent eg measurement in the laboratory. Under these conditions the basic TL ageequation can be written as :-TL Age (yr) _ Stored Dose (Oy)Dose Rate (Oy/yr)(1.1)13This is a very simplified view and in practice this equation has to be modified to takeinto account aspects of microdosimetry (Chapter 2) that will then allow practicablesolutions to be achieved. The next two sections will look at the ways in which thestored dose and dose rate components of this equation can be evaluated.1.S STORED DOSE EVALUATIONThe simplest approach to evaluation of the stored dose is to measure the natural TLsignal on the first heating of an aliquot of mineral grains extracted from the sample,and to then compare it with subsequent TL response from the same aliquot after beingexposed to a known radiation dose from a laboratory source. This procedure, however,is not commonly used because of possible changes in TL sensitivity associated withthe first glow heating event. These can be caused by changes in the transparency of thegrains due to heating or the activation of the pre-dose effect (mentioned earlier, section1,3). An additional complication is presented by the possibility of supralinear growthof the TL response to dose in the initial low dose region.To overcome these detrimental aspects a method of standard addition is used.Successive aliquots of the sample are given incremental calibrated radiation doses on topof their previously stored natural dose. Each aliquot is measured and the TL intensityvalues from the series of glow curves are used to construct an additive growth curve ofthe first glow response at doses above the natural level. The same aliquots can then bereirradiated, preferably in reverse order so that each will have received an equal pre-dosethereby minimising any interference from the pre-dose effect, and glowed out to givea second glow growth curve which provides the best available reconstruction in the lowdose region. This method is illustrated in Figure 1,1The stored dose is taken as the sum of the dose intercepts reconstructed from normalisedplots of the first and second glow growth curves. This assumes that the supralinearityintercept obtained from the second glow measurement has not been changed by the firstglow heating.14II)ocII)o'"II)e·e:l~Figure 1.1stored dose.intercept (I)stored dose isThe additive dose method for evaluating theThe equivalent dose (ED) and second gloware evaluated by regression analysis. Thegiven by ED + I.15An obvious requirement for reliable stored dose evaluation is that the measurementsshould only consider stable signals and that these should be identifiable. This led to thedevelopment of the "plateau-test" (Aitken, 1974) in which a plot of the ratio of a naturalTL curve to an artificially induced curve as a function of temperature is produced. Theplateau region was then identified as being the temperature range over which the TLsignals were regarded as being stable. Although desirable, the presence of a flat plateauis not an essential prerequisite for satisfactory dating. Fleming (1979) quotes such anexample, where the stored dose plateau condition is not met but due to a rapid changeof alpha efficiency with temperature an "age-plateau" analysis provides evidence for thevalidity of the TL age determination. McKeever (1985) has shown that for second-orderkinetics stability can be achieved without a plateau; Sanderson (1982, 1983) hasdemonstrated that the presence of a plateau does not lead automatically to stability.-10 -8 -6 -4 -2 o 2 6 8 104The observation of short term ("anomalous") fading in the high temperature regions ofglow curves of some minerals (eg Wintle, 1973) and the subsequent investigations ofDose (Gy)this phenomenon (Sutton and Zimmerman, 1978; Templer, 1985 and Visocekas, 1985)has made the inclusion of short term fading tests an essential part of any TL datingprogramme (Sanderson, 1988, Templer, 1986, Clark and Templer, 1988, Stoneham,1987).1.6 DOSE RATE EVALUATIONThis section will act as an introduction to the fuller treatment of dosimetry to be foundin Chapter 2. This is deemed necessary because of the large number of inter-relatedvariables that influence the radiation dose rate that is experienced by the constituentparts of a sample. The main sources of ionising radiation that provide the dose rate tothe sample are uranium and thorium with their decay products, and potassium and to alesser extent rubidium and cosmic rays. The dose rate can be divided into alpha, beta,gamma and cosmic components for the following reasons:(i) because variations in U, Th, K content change the relative proportions ofalpha, beta and gamma from sample to sample,(ii) because of the relative inefficiency of alpha particles for producing TL,compared to the other types of radiation,(iii) because of their different ranges and in relation to both large and smallscale inhomogeneities of radioactive content for example between soiland sample, and the clay matrix and mineral inclusions,(iv) and because of differences in mass absorption coefficient and stoppingpower mean that one material may absorb more energy than another. Thislatter point can be usefully illustrated by consideration of the watercontent of the pottery and the soil since the dose received by particulargrains depends on the amount of water present in the sample and itssurroundings and also on the type of radiation involved (Zimmerman,1971).16The wide range of possible techniques that can be used to evaluate dose rates is outlinedhere;(i) alpha counting - this is a scintillation technique that can be used to obtainalpha,beta, and gamma dose rates based on work in the health physicsfield by Turner, Radley and Mayneord (1958).(ii) beta counting - also a scintillation technique, this allows rapid andaccurate beta dose rate determination, but gives no information as to therelative contributions from uranium, thorium and potassium (Sanderson,1988).(iii) TL dosimetry - uses very sensitive TL phosphors (CaF2 or CaS04) torecord natural beta dose rates from pottery but is labour intensive andmay involve inconveniently long exposure periods (Bailiff and Aitken,1980 and Bailiff, 1982), also extensively used to measure the in situenvironmental gamma dose rate. The technique is not readily applicableto alpha dosimetry on account of the short range of alphas, their lowefficiency for TL production, and the difficulty in matching the alphastopping powers of phosphor and matrix.(iv) gamma spectroscopy - two main types, firstly, laboratory based highresolution instruments that enable uranium and thorium seriesdisequilibrium to be investigated (Murray and Aitken, 1988). The maindrawbacks here are cost and potential methodological problems. Thesecond type is the portable NaI gamma spectrometer used for on sitemeasurement of the environmental gamma dose rate. The shortmeasurement period could perhaps be regarded as problematical in thatseasonal variations will not be taken into account as is the case forgamma TLD capsules. This should be borne in mind although each sitewill be different and in any case it may well be that the longer term17variations (ie over archaeological timescales) are of more significance andalso more difficult to quantify. Meakins et al (1979) have also used Nalgamma spectrometers to measure U, Th, K for dose rate determinationin sherds and soils.(v) reactor based methods - Neutron Activation Analysis (NAA) requiresaccess to reactor and counting facilities but is suitable for the levels ofU, Th, and K commonly found in archaeological samples. The parentmembers of the U and Th decay series are quantified so no informationis obtained regarding their state of equilibrium. Fission track mappingis useful since it can provide information about the concentration and thespatial distribution of uranium and thorium.(vi) Inductively Coupled Plasma Mass Spectroscopy - ICPMS, a techniquesuitable for stable and long-lived isotope analysis, but not 4°K as thissuffers interference from the carrier gas, Ar. A further complication isthat usually the samples have to be presented in solution (2% HN03).(vii) K analysis - in addition to those methods outlined above, flamephotometry has perhaps been the traditional approach although there arealso references to the use of atomic absorption and x-ray fluorescence forthe same purpose.The practical application of some of these techniques will be more fully explored in thenext chapter where they will be linked with the measurement of stored dose to describethe specific strategies that have been formulated into particular TL dating methods, inrelation to their dosimetric parameters.181.7 SUMMARYThis chapter has introduced the subject of this thesis, namely the recognition of a needto able to date archaeological material that has already been removed from its burialcontext. Absolute dating of this corpus of material would be extremely valuable toarchaeology in allowing historic hypotheses to be tested. The historical developmentsof luminescence dating studies have been discussed and the elements of TL datingintroduced to provide a background for the subsequent chapters.192.DOSIMETRY202.1 INTRODUCTIONThis chapter examines the sources and characteristics of the natural radiation field thatprovide the dose experienced by the TL sensitive mineral grains since the zeroing event.From Equation 1.1 it is clear that to be able to quantify the time elapsed since thiszeroing event it is necessary to determine the radiation dose responsible for the naturalTL signal and to evaluate the effective natural dose rate to the particular TL phase beingmeasured. This more detailed account of pottery dosimetry is necessary because of itscomplexity and its fundamental importance, as the denominator in the age equation, toTL dating systems.The approach will be first to introduce some dosimetry concepts and then to developthem with particular reference to pottery dosimetry. It will then be shown how routinedating techniques have been formulated to enable solutions to the TL age equation tobe found, the limitations of these methods, and possible approaches to overcoming them.2.2 BASIC DOSIMETRY DEFINITIONSThis section sets out some basic definitions of dosimetry terms that are used in thisthesis (after Horowitz, 1984).2.2.1 Energy ImpartedThe energy imparted by ionising radiation to matter, E, in a volume is a stochasticquantity which is the difference between the sum of the energies (excluding restenergies) of all the directly and indirectly ionising particles that have entered the volume(LEE) and the sum of the energies (excluding rest energies) of all those that left it (LEL),minus any energy that is generated within the volume (I.ER) Unit: Joule(J)(2.1)212.2.2 Absorbed DoseAbsorbed dose, D, is defined as the mean energy imparted (dE) by ionising radiation perunit mass of material. Unit: I.kg-l, the Gray, Gy.dED--dm(2.2)2.2.3 Stopping PowerThe quotient dEldx is the linear stopping power, S, of a material for charged particlesof energy, E. The energy loss has two principle components, those due to collision andthose due to radiative losses.dE dES - (-) n + (-)_-'dx co dx DIU(2.3)The mass stopping power, Sip, is obtained by dividing the linear stopping power by thedensity, p, of the material.2.2.4 Charged Particle EquilibriumCharged particle equilibrium (CPE) is an important concept in gamma dosimetry. CPEis said to exist when the energy fluence of incoming and outgoing electrons at theboundary of a specified volume element is exactly equal.2.3 INTRODUCTION TO RADIOACTIVITYA nuclide is a particular type of atom that is specified by the total number of protonsand neutrons (Atomic Mass Number, A) in its nucleus. The number of protons present22in the nucleus (Atomic Number, Z) defines the atomic element. Isotopes are differentnuclides of the same element and have the same number of protons but a differentnumber of neutrons in the nucleus.Although most naturally occurring nuclides are stable, there are some that are not. Thestability of a nucleus is determined by the numbers of neutrons and protons, theirconfiguration and the binding energies involved. This can be seen from a plot of numberof neutrons, N, against number of protons, Z, in the nucleus (a Segre chart), whichshows that for Z ~ 20 the stable nuclides have Z ""N, but at higher Z values N > Z.Nuclides that disintegrate, to form new nuclides, by the emission of energy are termedradionuclides. The energy released can be calculated from the difference between themasses of parent and daughter nuclides and decay products.2.4 RADIOACTIVE DECAY LAWSRadioactive decay is a random process. Among the atoms in a sample undergoing decayit is not possible to identify which specific atom will be the next to decay. The decayrate (the activity, A) is a measure of the number of disintegrations per unit time. Theunit is sol(Becquerel, Bq).A- dNdt(2.4)The decay rate is proportional to the number of radioactive atoms, N, present (A oc N).If the number of radioactive nuclei and the number of decays per unit time aresufficiently large to allow statistical treatment, thendN-- - ANdt(2.5)23A. is the decay constant. It represents the probability, per unit time, of radioactive decay.The general equation for simple radioactive decay can be obtained by integration.(2.6)No is the number of nuclei present at the start time, t = O.This decay is exponential and since A oc N, the activity of a sample is also reducedexponentially with time.(2.7)~ is the initial activity. The specific activity (activity per unit mass, Nm) of a sampleis similarly time dependant and also proportional to the concentration of theradionuclides. The unit is Bq.kg'.The decay constant, A., is radionuclide specific and is essentially unaffected by pressure,temperature or chemical form. This means that radionuclides can be characterised bytheir period of decay. This is expressed in terms of the half-life, 1%, the time taken forthe number of radioactive atoms in a sample (and therefore its activity) to decrease tohalf of an initial value. The relationship between the decay constant and the half-life is(2.8)Series decay occurs when a parent radionuclide decays to a daughter product that is alsounstable. The net rate of formation of the daughter atoms is the difference between therate of formation of the daughter and its decay.(2.9)24NI and N2 are the number of parent and daughter atoms and Al and ~ are the decayconstants of the parent and daughter, respectively.In certain cases, where the half-life of the parent is very much longer than that of thedaughter, the condition of radioactive (secular) equilibrium may be achieved. This iswhere the rate of decay of the daughter is equal to the rate of decay of the parent.(2.10)2.S MODES OF RADIOACTIVE DECAYThe principle modes of radioactive decay are alpha and beta decay with associatedgamma emissions.Alpha decay involves the release of helium nuclei, 24HeA A-4 4zX ~ Z-2X + 2He + Q (2.11)where X is the parent nucleus, Y the daughter nucleus and Q the energy release. Thealpha particles emitted are monoenergetic for a given decay and therefore theycharacterise the particular decay.The term beta decay can cover a number of processes, these are outlined below.(a) Electron, or negatron, (W) emission. This is equivalent to one of theneutrons in the parent nuclide becoming a proton with the emission of anegative electron, p-, and an antineutrino, v25A A 0-Z X - Z+ 1X + -1 P + V (2.12)(b) Positron (W) emission. Here a nuclear proton is transformed into aneutron with the emission of a positive electron, W, and a neutrino, vA A 0ZX - Z-1X + +1P + V (2.13)(c) Electron capture. This is where an electron from the inner (K or L)orbitals of an atom is captured by the nucleus. The vacant position issubsequently filled by an outer electron with corresponding emission ofK or L x-rays, or low energy (Auger) electrons.Unlike alpha particles, beta particles are emitted with a continuous spectrum of energiesup to a maximum value that is determined by the energy equivalent to the change inmass involved in the nuclear transformation. The energy released is shared between thebeta particle and antineutrino (no charge and negligible mass) in varying proportionswhich gives rise to a continuous energy spectrum. Beta particle energy is therefore notdiagnostic of the species involved.Alpha and beta decay can leave the daughter nucleus in an excited state. The excitationenergy can be removed by gamma ray emission or through internal conversion. Theexcited states are separated by quantised amounts of energy so that the gamma photonswill have characteristic energies and probabilities of occurrence for particular transitions.Internal conversion is a process by which the energy of transition from an excited stateis given to an orbital electron which is then ejected.2.6 CALCULATION OF INFINITE MATRIX DOSE RATESThe concept of an infinite matrix provides an approach to the calculation of dose rates.For such a system, the conservation of energy requires that within a homogeneous26volume having dimensions greater than the ranges of the radiations, the rate of energyabsorption is equal to the rate of energy release.In this case, the absorbed dose rate (the rate of absorption of energy per unit mass), D,is given by the specific activity of the matrix multiplied by the energy release per decay,E.D - AN . Em(2.14)To summarise this point; -if all the energy released by radioactive decay in a given volume is absorbed, then thedose rate is simply proportional to the radionuclide concentration.For a volume containing more than one radionuclide then the infinite matrix dose ratecan be obtained by summing the individual contributions.Where a radioactive series is in equilibrium, so that all members of the series decay withthe parent rate, the energy released per parent disintegration is the total summed overthe whole series. For a series exhibiting disequilibrium, individual summations must bemade.The infinite matrix approach, although very useful as a starting point, does not fullydescribe those matrices typically offered by TL dating samples. This is due to theheterogeneous nature of the TL bearing minerals and the radioactivity being present ona morphological scale that is comparable with the ranges of the naturally occurringionising radiations.Therefore, the infinite matrix dose rate as calculated from the measured concentrationsof radionuclides within a real sample needs to be modified by microdosimetriccorrections before the effective dose rate applicable to that sample can be assessed.27The sources and characteristics of the radiations that provide this dose rate will beexamined next.2.7 SOURCES OF NATURAL RADIOACTIVITYThe major sources of radiation in the natural radiation environment are the alpha, betaand gamma emitting uranium and thorium decay series and the beta and gamma emittingradioisotope of potassium, 40K. These are long-lived radio nuclides that give rise to aneffectively constant dose rate over the age span of an archaeological sample, assumingthat there is radioactive equilibrium. The relevant decay schemes are shown in Tables2.1, 2.2, 2.3 and 2.4The calculation of dose rate conversion factors, ie the dose rate per parent concentration,requires the summation of each energy release down the decay chain. Early calculationswere carried out by Bell (197 6a) with later revisions (Bell, 1977, 1979) to accommodatemore up to date values and in his later paper to standardise the radiation data used soas to be in line with the values adopted in geo- and cosmochronology. More recentlyconsiderable effort has been invested in the recalculation of Bell's figures (Nambi andAitken, 1986, Sanderson, 1987). The latter author not only confirmed Bell's results butalso explicitly stated the distribution of energy release at different stages of the decaychains. The recalculated values are in excellent agreement with previously publisheddata, any differences being unimportant when considered in terms of their relativecontributions within the context of dating. The dose rate conversion factors used in thiswork are taken from Aitken (1985) and are given in Table 2.5. The use of theseconversion values can be usefully illustrated with an example of a "typical" soil, orpottery, matrix containing 3 ppm U, 10 ppm Th, 2% K (Table 2.6).28Nuclide TII2 Alpha Beta Gamma Ie Daughter238U 4.4683 4.1849 0.00016 0.01104 2~xl09 a2~ 24.10d 0.1686 0.0077 0.0263 234Pa234mpa 1.175m 2.272 0.00951 0.01491 234U234Pa 6.75 h 0.560 0.001 0.0003 234U234U 2.446 4.7600 5x10-5 0.0082 2»-tllxl05 a~ 8.0 4.753 2.5xl0-4 0.0111 226Raxl04 a226Ra 1600 a 4.775 0.00614 0.0014 2~222Rn 3.8235d 5.489 0.00036 --- 218pO218pO 3.05 m 6.002 --- --- 214Pb(0.9998)0.256 218At1.85xl0-4214Pb 26.8 m 0.715 0.244 O.oI 159 214Bi218At 1.5-2 s 6.693 0.0477 0.001 214Bi214Bi 19.7 m 1.665 1.669 0.008 214pO214pO 163.7 us 7.687 8xlO-5 210Pb210Pb 22.26 a 0.0237 0.00186 0.029 21~i21~i 5.012 d 1.155 --- --- 210pO21OpO 138.4 d 5.3043 lxlO-5 206Pb206Pb StableMev MeV MeV MeVTABLE 2.1 238U Decay Scheme29Nuclide Tl/2 Alpha Beta Gamma Ie Daughter235U 7.038 4.394 0.14508 0.03481 231Thx108a231Th 25.52h 0.231 0.01313 0.0569 231Pa231Pa 3.28 4.972 0.0367 0.0285 227Acx104a227Ac 21.773a 4.933 0.0001 223Fr(1.2%)0.0455 227Th(98.8%)227Th 18.17d 5.928 0.0715 0.0363 223Ra223Fr 21.8m 1.135 0.05746 223Ra223Ra 11.68d 5.675 0.0895 0.0003 219Rn219Rn 3.92s 6.758 0.0538 0.0007 2l5po2l5po 1.83ms 7.384 0.0002 2llPb2llPb 36.1m 1.296 0.0556 0.003 211Bi2llBi 2.16m 6.564 0.0446 0.0089 207T1(98.68%)2llpo 0.52s 7.431 1.440 0.03508 2llpo(0.32%)207TI 4.77m 0.0022 207PbMeV MeV MeVTABLE 2.2 23SU Decay Scheme30Nuclide T1I2 Alpha Beta Gamma Ie Daughter232Th 1.41 4.002 0.01416 228Raxl010a228Ra 5.76a 0.0292 0.06021 228Ac228Ac 6.13h 1.207 0.8758 0.0489 =n,228Th 1.913a 5.398 0.00249 0.0035 224Ra224Ra 3.66d 5.672 0..0095 220Rn220Rn 55.68 6.288 0.00465 216pO216pO 0.145s 6.779 1.4x10-5 0.004 212Pb212Pb 1O.643h 6.090 0.3592 0.1385 0.0115 212Bi212Bi 60.60m 6.050 0.0987 2°Brl(36%)212PO(641.738 %)212pO 0.296us 8.784 208Pb208Tl 3.05m 1.581 1.2142 0.0120 208PbMeV MeV MeV MeVTABLE 2.3 232ThDecay Scheme31Nuclide Tl/2 Alpha Beta Gamma IC Daughter4°K 1.277x109 a 1.343 40Ca(89.33%)Meanbeta0.583EC 1.461 4°Ar(10.67%)TABLE 2.4 4°K Decay SchemeRadionuclide Concentration Alpha Beta GammaU Series 1ppm 238UNo radon loss 2.779 0.1461 0.1149100% radon loss 1.260 0.0610 0.0055Th Series 1ppm 232ThNo thoron loss 0.739 0.0286 0.0514100% thoron loss 0.309 0.0103 0.0208Natural 1% K20 - 0.689 0.200Potassium 1% K - 0.830 0.241TABLE 2.5 DOSE RATE CONVERSION FACTORS (mGy.a·l) (FROM AITKEN,1983)32(Alpha) Alpha(eff) Beta Gamma Total1U 3ppm (8.337) 1.251 0.438 0.344Tb 10ppm (7.390) 1.109 0.286 0.514K 2% - - 1.660 0.482Total (15.727) 2.360 2.384 1.341 6.085Relative 39% 39% 22%ContributionFine- grain' 2.444 2.370 1.3413 5.955Relative 38% 40% 22%ContributionInclusion' 04 2.243 1.3413 3.584Relative - 63% 37%ContributionTABLE 2.6 INFINITE MATRIX DOSE RATES (mGy.a-l) CALCULATED FORA TYPICAL MATRIX ANDEFFECTIVE DOSE RATES FOR FINEAND QUARTZ INCLUSION GRAINS1. Alpha efficiency = 0.152. Dose rates calculated for activity free grains in a dry matrix.3. Matrix and soil composition are the same (therefore gamma dose rate originatesfrom the soil).4. Alpha dose rate eliminated for inclusions by HF etching.33Other sources, providing less significant contributions to the overall dose rate, includecosmic radiation and possibly other long-lived radionuclides. Warren (1978) highlighted87Rb as one such nuclide that had previously been overlooked. He provided dataregarding its calculated dose rate and also suggested values of the potassium to rubidiumratio for use with those samples for which the rubidium content had not beendetermined. This work was extended by Sanderson (1987) to include a number of othernaturally occurring primordial nuclides, demonstrating that of all these only the rubidiumbeta emission was ever likely to be of any significance, and even then only to a minorextent.The cosmic ray flux at ground level results from interactions between the upperatmosphere and a primary cosmic particle flux. It has a "soft" component of energeticelectrons that is readily absorbed within 0.5 m of soil, and a "hard" component of muonswhich have much greater ranges in soil and therefore need to be considered here. Thereis a dependence on latitude due to the focusing effect of the earth's magnetic fieldtowards the poles of the earth. There is also an altitude effect related to absorption byvarying thicknesses of atmosphere. Generally speaking the relative contribution ofcosmic radiation to the total dose is small (of the order of 5% for quartz inclusion datingand less for fine-grain dating). However, there are certain instances, for example wherethe radioactivity of the sample and its burial medium are exceptionally low or forsamples originating from high altitudes, where the cosmic ray dose may become moresignificant.Theoretical work (Greison, 1942) allowed an estimate of 0.13 mfiy.a" to be made forthe lower limit of the cosmic ray contribution. Ion-chamber measurements (Shamos etal, 1964) suggested that the ground level cosmic radiation dose rate was of the order of0.30 mfly.a'. These studies were neatly tied together, experimentally, by Aitken (1969)using thermoluminescence dosimetry (TLD) to measure the variation of cosmic ray doserate with depth in chalk (chosen because of its low radioactivity). The results indicatea ground level dose rate of 0.28 mfiy.a' that is reduced to an effectively constant valueof 0.14 nrGy.a" at a depth of approximately 1 m.34More recent work (Prescott and Stephan, 1982) recommends slightly higher values of0.185 mfly.a" at 0.5 m andO.150 mGy.a-1 at 1 m depths of standard rock(p = 2 g.cm"). They also examined the effect of latitude and altitude on the cosmic raydose rate and derived an expression for Dcosmic' taking these parameters into account.2.8 DOSIMETRY2.8.1 IntroductionThe previous section described the sources of alpha, beta, and gamma radiations thattogether form the natural radiation fields within which samples for TL dating acquiretheir stored doses. The infinite matrix dose rates, as calculated from the measuredradio nuclide concentrations (see section 2.6), are not directly applicable as a means ofdetermining the effective dose rates to the sample for three main reasons. The first isthat alpha radiation is generally less efficient at producing TL than beta or gammaradiation. Secondly, the different ranges of alpha, beta, and gamma radiations mean thatthe spatial configuration of the TL sensitive minerals and the radio nuclide concentrationsinfluence the relative significance of each contribution. Thirdly, differences in stoppingpowers and mass absorption coefficients mean that one material may absorb more thananother. These factors are fundamental to microdosimetry and will now be consideredin terms of the three relevant types of radiation. The various approaches to measuringthe natural radiation dose rate will then be discussed.2.8.2 Alpha DosimetryAlpha particles are much heavier than electrons so they are deflected very slightly whentheir Coulomb fields interact with atoms to form ion-pairs and therefore travel in astraight line through matter. Compared to beta particles alpha particles have muchhigher specific ionisation values (number of ion-pairs formed per unit path length). Thevelocity of the alpha particle decreases as it loses energy in travelling through anabsorber. Consequently, it will spend progressively longer times in the vicinity of anyparticular atom which results in an increase in the probability of interaction. This means35that there is a steady increase in the number of ion-pairs formed along the path of theparticle, up to a maximum value which is close to the end of the particle's range (theBragg Peak). Just beyond this point the kinetic energy of the particle is comparable tothat of the orbital electrons of the absorber. The alpha particle can then acquireelectrons becoming uncharged, and therefore incapable of causing further ionisation.Although the particle range depends on the properties of the absorbing medium and theenergy of the particle, the mean range of alpha particles resulting from natural seriesdecay in pottery, soil and other similar media is between 20 and 30 microns (themaximum value being 45 microns from the 214potransition of 7.689 MeV). This meansthat for large mineral grains embedded in a radioactive medium the alpha dose from thatmedium is attenuated progressively within the interior of the grain. Similarly, any alphaactivity within such large inclusions is unable to fully irradiate surrounding volumes.The formulation of the two routine pottery dating techniques, fine-grain and quartzinclusion, reflect this microdosimetric situation since they depend on a simplified viewof the pottery matrix as having alpha activity homogeneously distributed through theclay matrix with quartz inclusions as non-radioactive spheres within it. Thesetechniques will be discussed in more detail towards the end of this chapter.These effects can be dealt with by introducing attenuation factors which, whenmultiplied by the infinite matrix dose rate for the external source give the effective doserate. The calculation of alpha attenuation factors for the uranium and thorium series inequilibrium has been carried out for spherical grains of varying size (Fleming, 1979,Bell, 1980). The effective alpha dose rate for a grain with some alpha activity is madeup of the external infinite matrix alpha dose rate multiplied by an attenuation factor plusthe grain's internal infinite matrix alpha dose rate multiplied by absorbed dose fraction(1 - attenuation factor). The alpha absorbed dose fractions used in this work are thoseprovided by Bell (op cit). These are presented in Table 3.1.The contribution of alpha radiation to the natural TL signal is further reduced as aconsequence of its lower efficiency for producing TL in comparison to an equivalentbeta or gamma dose. This is due to the high ionisation density produced by alpha36particles, so that within the central core of the track the TL traps become saturated andso a large proportion of the alpha's energy cannot take part in trapping processes(Zimmerman, 1971). Zimmerman proposed that this efficiency factor, the k-value, shouldbe measured as the ratio of the TL sensitivities to known doses of alpha and beta (orgamma) radiation.k _ TL per Gy of " radiationTL per Gy of P radiation(2.15)The initial evaluations of k were carried out using 3.7 MeV alpha particles but thissystem was found to have the disadvantage of being strongly energy dependent since theionisation density increases as the alpha particle energy decreases. This counteracted theadvantage of the k-value being directly applicable to the age equation, as a prefix to thealpha dose rate term, without the need for transformation into other units. The laboratorymeasurement of k is therefore dependent on the particular alpha source used as well asthe irradiation geometry and other experimental conditions. It also follows, then, that theamount of alpha particle TL induced over archaeological time will be dependent notonly on the dose rate but also their natural energy spectrum. The latter point wascompensated for by the modification of k to become ~ff such that for a sample withequal U and Th activities, keff= O.83k3,7-Zimmerman (1972) had also observed that the TL per unit length of track wasapproximately independent of the alpha particle energy (up to 3.7MeV) and this waslater confirmed and investigated further (up to alpha energies of 7.3MeV) by Aitken andBowman (1975). As a result of this work they sought to redefine alpha source strengthsin terms of track length delivered per unit volume of sample. This meant that the alphaefficiency was also redefined in terms of the a-value system,TL per unit " track lengtha - --------~------------~~------TL per unit absorbed p energy (ie per Gy)(2.16)37a -x13S(2.17)where, x = the beta dose (in Gy) that produces the equivalent TL response as oneminute's irradiation from the alpha source,S = the alpha source strength (in J..L111-2min-l) == the rate at which track length is deliveredto the sample.The numerical factor is introduced so that, for a pottery sample, the a-value will beequal to Zimmerman's k-value (for 3.7 MeV alpha particles).Aitken and Bowman's reasoning was further developed (Bowman and Huntley, 1984)by making the assumption that the TL per unit of absorbed beta energy is also aconstant that is independent of the energy of the beta particles, as well as the TL perunit alpha track length being independent of the alpha particle energy. The alphaefficiency can then be expressed in terms of the beta energy that produces the sameamount of TL as one unit length of alpha track. This is then related to what is actuallymeasured in the laboratory, where a calibrated alpha source delivers a known tracklength per unit volume per unit time, to a sample of thickness less than the alpha tracklength, and the resulting TL is compared to that induced by a known beta dose. Thisgives rise to a third potential measure of the alpha efficiency, the b-valueb _ TL per unit Cl track length per unit volumeTL per unit absorbed p dose(2.18)b has SI units of Gy.m'.Valladas and Valladas (1982) considered the situation for coarse grains, in particularwhere due to uncertainties in the HF etching process (used to remove the alphairradiated outer layer of these grains (Fleming, 1970)) this step is omitted and directmeasurement is made of the alpha response. The approach is to carry out the alpha38irradiation with the coarse grains on a vibrating platform to simulate the omnidirectionalalpha flux that the grains would experience within a pottery matrix and to minimiseerrors due to non-spherical grain shape.Aitken (1985a, b) has also considered the case of coarse grains and has derived anexpression for the relationship between the "true" effectiveness, a, as obtained by theirradiation of fine grains and the "apparent" alpha effectiveness, a', as obtained by theirradiation of coarse grains.a - ---2D .al (2.19)3 ,,1 RD = grain diameter (urn)R = alpha range for the alpha energy usedT\' = ratio between the average TL per unit track length for total absorption andthat for a thin layer of fine grainsReference to Table 2.6 shows that for a relative alpha efficiency of 15%, a value thatis appropriate for most dating materials, the absorbed dose for fine grains, which wouldotherwise have a dominant alpha contribution, will have approximately equal alpha andbeta contributions. For coarse grains, however, it is possible to further reduce theeffective alpha dose to a negligible level by HF acid etching of the alpha irradiatedsurface of the grains. These approaches have been exploited to form the basis of theroutine, fine-grain and inclusion, methods of TL dating which are discussed in sections2.12.1 and 2.12.2, respectively.392.8.3 Beta DosimetryBeta particles can interact with matter through a number of mechanisms, although theirenergy is absorbed mainly through the processes of ionisation and atomic excitation. Theprocess of ionisation involves the interaction of an incident beta particle with an orbitalelectron of an atom that leads to the ejection of the orbital electron and the scatteringof the beta particle. The secondary electrons (delta rays) can have such high energiesthat they cause further extensive ionisation. The specific ionisation (number of ion-pairsproduced per unit length of particle track) for beta particles is much lower than that ofalpha particles. This is due to the much greater velocities of the beta particles thatresults in much longer ranges. At very high energies the beta particles have straightpaths because they have a momentum much greater than that of the orbital electrons.Atomic excitation also involves the interaction of an incident beta particle with anorbital electron, but here the electron only receives sufficient energy for it to attain ahigher orbit (ie an excited state) and the incident beta particle is scattered from itsoriginal path.Of less importance are the mechanisms of radiative energy loss (Bremsstrahlung) whichresults from the change in velocity experienced when an incident beta particleapproaches the positively charged nucleus, and Cerenkov radiation which is effectivelythe result of an electromagnetic shock wave. Direct interaction betweenelectron/positron pairs is possible in which both particles are annihilated. The combinedmass of the two particles is converted into electromagnetic radiation, which is emittedas two gamma photons travelling in opposite directions each having a characteristicenergy of 0.511 MeV. This is called annihilation radiation and its presence in theelectromagnetic spectrum of a radionuclide is strong evidence for positron emission bythat nuclide.Although beta particles have continuous energy spectra, it can be shown that mostenvironmental betas are absorbed within 2-3 mm of their source in a quartz, or similar,matrix (Katz and Penfold, 1952). This means that for TL dating samples such as potterywhich may have been buried in a medium of contrasting radioactivity, the beta40dosimetry can be confined to that of the internal activity of the sample by removing thesurface layer of the sample that has been exposed to the beta radiation from the soil.For fine-grains the beta dose rate obtained after correction for the attenuation due towater is directly applicable but for inclusion sized grains a further correction is requireddue to the attenuation of the beta radiation within the grains. Mejdahl (1979) hasevaluated the factors by which the average beta dose to a grain of a given size isreduced from the infinite matrix dose. The model used consists of an infinitehomogeneous matrix containing uniformly distributed radionuclides within which thequartz grains are regarded as being non-radioactive spheres of the same composition asthe matrix. Conservation of energy requires that the attenuation of the external betaradiation by a grain must be balanced exactly by the self-dose fraction which the grainwould absorb if it had the same concentration of radionuclides as the matrix. Thus fora grain of no internal radioactivity the attenuation factor is given by 1 minus the dosefraction absorbed by the cavity. Alternatively, this can be seen to be the dose fractionthat would have been absorbed if the grain itself were a radioactive sphere in a non-radioactive medium.The calculation used percentile distances, Xp' which represent the distance within whichp% of the radiation is absorbed. This is achieved by using a scaling procedure (Cross,1968) to obtain values for quartz from those given for water (Berger, 1971). This ispossible because of the similarity in shape of the absorbed beta dose distributions indifferent media allowing scaling to be carried out on the basis of the percentile distancesbeing proportional to density and stopping powers.Berger (1971, 1973) produced scaled absorbed dose distributions (point kernels) forwater by considering the situation for spherical regions within an infinite matrixcontaining uniformly distributed point beta sources. His later paper (improved pointkernels) took into account the effects of energy loss by straggling and bremsstrahlung.The absorbed dose fractions, <1>, for quartz were calculated by Mejdahl for a largenumber of grain sizes for uranium and thorium series, in equilibrium, and potassium41from the expressionet> - f:mu [ 1-(3/2) (t/tmax) + (1/2) (t/W3 ] F(t)dt (2.20)where t = x/X9(Jx = distance from the sourcex9(J= distance within which 90% of the radiation is absorbedtmax = D/x9(JD = grain diameterF(t) = scaled dose distributionThese values are presented in Table 3.2 where they are used in formulating simulatedisochron systems as well as being later incorporated into the experimental programme.Although the values were calculated for quartz they are also valid for feldspars, sincethey have stopping powers close to that of quartz. (Table 2.7)It is also apparent from Mejdahl's calculations that the attenuation factors for uranium,thorium, and potassium are different so that the effective attenuation factor will dependon the relative abundance of these radionuclides. The self-dose fraction due to inherentpotassium in alkali feldspars can be calculated from this work (Mejdahl and Winther-Nielson, 1982, Mejdahl, 1983; Sanderson, 1987; Sanderson et al. 1988).2.8.4 Gamma DosimetryThe absence of charge and rest mass for gamma photons results in little interaction withthe absorbing atoms and greater penetration. Gamma photons produce only 1 - 10% ofthe number of ion pairs of that produced by beta particles of the same energy. As aconsequence of this low specific ionisation, the. ionisation is almost completelysecondary 7in nature and results from the action of a few high energy primary ion pairs.The processes of interaction are outlined below.42In the photoelectric effect all of the photon's energy is absorbed by the orbital electronsof the target atom. The excited state produced is greater than the binding energy of someof the orbital electrons so that an electron can be ejected from the atom. The energy, Ee,of this emitted photoelectron is equal to the difference between the incident gammaphoton, EY'and the binding energy, Eb, of the electron in the atom.If the photoelectron originates from an inner electron orbital, an electron from a higherorbital moves to fill the vacancy. Here, the difference in binding energy of the higherand lower orbitals causes the emission of X-rays and low energy Auger electrons. Thephotoelectron as well as the Auger electrons and X-rays cause extensive secondaryionisation by interacting with absorber atoms.z., Aeff Stopping Densitypower (g.cm')Si02 10.80 21.60 0.9518 2.65KAlSi30g 11.81 23.92 0.9373 2.55NaAlSi30g 10.69 21.59 0.9450 2.60TABLE 2.7 ELECTRON STOPPING POWERS, RELATIVE TO AIR,CALCULATED FROM CROSS (1968).In the Compton effect an outer electron is ejected from the atom by the incident gammaphoton which is deflected with a lower energy. The scattered gamma photon may havesufficient energy for further interactions to occur. The energy and scattering angles ofthe Compton electron and the deflected gamma photon depend on the initial energiesand momentum involved.The process of pair production involves the interaction of a gamma photon with theelectromagnetic (Coulomb) field of the nucleus. The whole of the photon's energy isconverted to the mass and kinetic energy of an electron and a positron. This process isnot possible until the photon energy is equivalent to the rest mass of two electrons (1.0243MeV). The probability of pair production increases as the energy of the gamma photonincreases above this value. The excess energy appears as the kinetic energy of theelectron-positron pair and can be shared in varying proportions between the two.A fourth, and relatively minor, process of interaction is coherent (or Bragg, or Rayleigh)scattering, where the incident photon collides with an electron that is sufficiently tightlybound to the atom for the whole atom to absorb the recoil. The energy transfer to theatom is negligible and the photon is scattered without loss of energy.Photoelectric interactions predominate for all materials at sufficiently low photonenergies, but as the energy increases the photoelectric effect falls off more rapidly thanthe Compton effect and the latter eventually takes over as the dominating process. Withcontinuing increase of photon energy, although the Compton effect decreases in absoluteterms, it increases still more relative to the photoelectric effect. At photon energies ofseveral MeV the pair production process begins to make the greatest contribution tophoton interactions.As a result of the different scattering processes, mentioned above, and the wide rangeof gamma emission energies from the natural series, the analysis of the gamma ray dosedistribution is very complex. Both primary and secondary (scattered) gamma photonscontribute to the gamma dose delivered in an extended medium. Fleming (1979)estimated the gamma ray attenuation of environmental media using the approximationthat the build up factor (ratio of total dose to primary dose) increases linearly withdistance from the source, with the further simplifications that uranium and thoriumseries, in equilibrium, could be approximated by a single set of absorption coefficientsand that the resulting integrated dose distributions in soil were similar to that ofpotassium, from which the relative contributions of soil and sherd to ceramic datingwere assessed. The resulting approximation that all the gamma dose was from the soilwas acceptable for small sherds.Mejdahl (1983) has taken absorbed gamma fractions initially calculated for a soft-tissuemedium from existing Health Physics results using the Monte Carlo technique and used44conversion factors to make them applicable to pottery and stone samples. These resultsshow reasonable agreement with those of Fleming. The values used in this work aregiven in Table 3.3.Aitken et al (1985) combined an experimental approach with computer simulation basedon details of atomic compositions, cross-sections, primary emission spectra, etc. Furtherexamination could be suggested using a more extended primary gamma list for MonteCarlo simulation and extension of the experimental work to include the thorium andpotassium series.The key questions that remain are are the uranium, thorium, and potassium build upfactors reallyi) the same (ie unknown U,Th,K ratios in the soil do not matterii) independent of complex geometries ?Traditionally, it has been stated that one of the requirements of potential samples for TLdating is that they should come from an homogeneous layer of 30 cm radius around thesample. However, from the work stated above it is apparent that even for quartzinclusion dating, where the environmental gammas make a maximum contribution ofperhaps 30% of the total dose, the situation need not be hopeless if these conditions arenot met. This is because if the half of the gamma dose originating from within 8 cm ofthe sample can be measured accurately then the unknown part can probably beadequately estimated. It has also recently been stated (Aitken et al, 1985) that thepractice of avoiding samples above a depth of 30 cm from the ground surface may beover cautious.2.9 MEASUREMENT OF DOSE RATEIn this section there will be a brief overview of the main techniques available to measurethe dose rate component of the TL age equation. Those techniques that have beenemployed during the experimental verification of the isochron model will receive fullertreatment later.452.9.1 Alpha countingThis is a scintillation technique, based on the work of Turner, Radley and Mayneord(1958), in which a zinc sulphide screen is positioned on a photomultiplier tube (pmt),the powdered sample is then placed on the screen so that when an alpha particle hits thescreen a pulse of light is emitted that is detected by the pmt and registered in a countingsystem. The advantages of the technique are its simplicity and that the ratio of theeffective alpha dose rate to the alpha count rate is independent (to within a few percent)of whether the dose is provided by uranium or thorium (Aitken, 1978a). In order to usealpha counting to assess the beta and gamma contributions from the uranium andthorium decay series then either equal uranium and thorium activities are assumed orsome attempt is made to evaluate the U/Th ratio using the 'pairs' technique. Here thecounting electronics are designed so as to register pairs of counts arriving within aspecified time of each other, ego the Th chain contains the rapid pair of alpha transitions22°Rn-~ 216pO-~ 212Bi(t1l2 = 55 and 0.145 sec, respectively) and the extremely rapid pairin the 235Uchain due to the 0.002 sec half-life of 215po.The main problem encounteredwith the pairs technique is the need for long counting times to obtain statisticallymeaningful results. An alternative approach to estimation of the U/Th ratio is by usingenergy discrimination as suggested by Sanderson (1979) and adopted by Martini et al(1982).A more significant problem can be caused by the emanation of radon from a sample.This may be observed as an increase in the count rate for samples that have been storedin sealed counting cells (Aitken, 1978b).2.9.2 Beta CountingSanderson (1988) described this new scintillation technique for measuring matrix betadose rates based on thick source beta counting (TSBC) using wafers of plasticscintillator. The method is advantageous in that is direct, rapid and accurate, with aprecision of ±5% quoted as being achievable in counting times as short as 15 minutes.46This is obviously beneficial for sample throughput for dating programmes that alreadycontain certain features of automation and replication of measurements. An importantfeature of the beta counter is that its response is not strongly dependent on the U,Th,Kcomposition of the sample. One disadvantage is that it does not provide any informationon the relative contributions of the different source of beta radioactivity. In this respectit has been recommended that this method be used in conjunction with checks toconfirm that the uranium series is a minor component (to guard against radon loss anddisequilibrium) and that modern additions (from fallout or perhaps agricultural practices)are not contributing to the signal. This technique has been routinely used in this work,with neutron activation analysis being employed as the cross-check method, both ofwhich will be discussed in terms of their experimental aspects and' applications inchapter 3.2.9.3 Thermoluminescence Dosimetry (TLD)This method employs very sensitive TL phosphors, such as CaF2 and CaS04, to recordthe dose rate directly. The possibility of using alpha TLD is complicated by manyproblems, replication of the alpha effectiveness appropriate to the sample, replicatinggrain size, and the likelihood that such low levels of TL, in relation to any spurioussignals, would be obtained in a time that was practical, all combine to effectivelypreclude the use of the technique for alpha dosimetry.The use of ~TLD, (Bailiff and Aitken, 1980) however, is a much more promisingproposition although this too is not problem free. The basic problem is the achievementof an isotope independent response - such that the relative contributions from UffhlorK are unimportant with respect to the evaluation of the correct beta dose rate. Isotopedependence in ~TLD arises partly from the need to shield the phosphor from alpharadiation, which will introduce attenuation of the beta-dose rate - that is dependent onthe beta energy spectrum. There is also a gamma contribution which, despite the muchlower absorption coefficients of gamma compared to beta radiation is significant Thisis problematical due to the fact that the ratio of gamma to beta dose rates is strongly47isotope dependent (1.8: 1 for the Th series, 0.8: 1 for the U series and 0.3: 1 forpotassium). A number of variations of ~TLD schemes have been devised of which auseful resume is provided by Mejdahl and Wintle (1984). One of the features commonto all of these systems is the need for the ~TLD units to be stored for a sufficiently longperiod (usually 1 to 2 weeks) in a lead container to reduce the background radiation,followed by a period of labour intensive readouts (unless an automatic reader isavailable). Sanderson (1987) also notes that isotope independence depends on thecoating thickness of each disc and varies from disc to disc, therefore individualcalibrations are needed.Gamma thermoluminescence dosimetry was the first application of TLD in TL datingsince it provided a convenient way overcoming a number of problems. Firstly, thephosphor can be obtained with a radioactivity free material eg copper, stainless steel, orplastic, with a wall thickness sufficient to stop the phosphor being irradiated by betaparticles. Secondly, the seasonal variations (Mejdahl, 1970) in soil water content canbe averaged out by leaving the yrLD capsule buried for a year, where possible. Theburial position should be as comparable as possible to that of the sample being dated soas to experience a similar gamma radiation field - this also has the advantage ofautomatically including the appropriate cosmic ray dose. There are disadvantages,however, of which the most significant is how well the dose rate received by thephosphor replicates that received by TL grains in a pottery matrix. This is generally notthe case because of the difference in absorption coefficients for TL phosphors and TLmineral grain, particularly in the low energy region «2.0MeV). The Oxford yrLD unitpartially compensates for the over response to low energy gammas by CaF2NAT throughthe use of copper shielding - wall thickness 1.5mm. The Riso laboratory uses CaS04:Dycontained in polyethylene capsules within steel tubes, a combination that has a responsevirtually the same as quartz (Mejdahl and Wintle, 1984). These systems are calibratedusing doped concrete blocks (Oxford) or with the aid of a radioactive rock, lujavrite, inthe case of Riso. Significant drawbacks of the capsule approach are that no allowancecan be made for the presence of fallout (unlike gamma spectrometry methods - nextsection), and it is invariably the case that excavation will have removed the very contextfrom which the pottery to be dated came from, so a compromise has to result in48measurement from a different (possibly unrepresentative) context. Fading of the TLsignal during the storage period may also need to be corrected for, depending on whichphosphor has been used. East (1991) describes an environmental dosimeter, designedby Sanderson at SURRC, of a "paired" type that enables both beta and gamma doserates to be measured simultaneously. It may be said that the use of TL dosimeters tomeasure the environmental gamma radiation is a well-established technique.2.9.4 Gamma SpectroscopyThe use of gamma spectroscopy falls into two main categories, the high-resolutionlaboratory based and the portable types. The former have the advantage of being ableto analyse pottery sized samples, as well as soils, and that some information on VIThseries disequilibrium may be obtained (Murray and Aitken, 1982, 1988), although theapplication to small sample sizes is offset by the longer counting times required. Themain disadvantage of high resolution gamma spectrometry is the expense and complexityof the equipment involved. This technique has been used at the SURRC to characterisethe U, Th, K contents of a Shap granite sample that has been used as the in housestandard for neutron activation analysis of TL dating samples.Portable gamma spectrometers are generally based on a sodium iodide scintillationcrystal detection system coupled to counting electronics and a rechargeable batterypower supply (portable semi-conductor detectors are also available). The instrument canbe used in two ways - as a spectrometer to measure individual components by utilisingthe characteristic gamma emissions of 2.65MeV from 208Tl(Th series), 1.76 MeV from214Bi (V series) and of 1.46 MeV from 4°K (potassium). Alternatively, an optimumenergy threshold setting may be determined so as to minimise the isotope dependenceof the detector response eg 370 keY, suggested by Levborg and Kirkegaard (1974). Inthis mode the total dose rate may be estimated by integrating all gamma radiation abovethe threshold setting. Sanderson (1987) recommended the use of two threshold settings -350 keV and 1350 keV, the upper to determine the gamma dose rate due to V, Th andK, while the lower is a check for recent l37Cscontamination (from Chernobyl fallout,etc). Concordance of the two dose rate values obtained indicates a contamination free49site. The advantages of the portable gamma spectrometer is that the measurements arerapid (-20 mins per location - so a whole site could be covered in a single day) and tosome extent inhomogeneity of the soil is allowed for. The disadvantages are that alarger hole is required to accommodate the detector head in the site, and because themeasurements can be made so quickly there is no allowance for seasonal variations indose rate due to variations in soil moisture content, although Mejdahl (1970) has shownthat for sufficiently deeply buried samples this may be of little significance. Onepractical constraint is that there is a rapid variation of count rate with threshold settingwithin the usual interval quoted, (0.37 - 0.45 MeV, Mejdahl and Wintle, 1984) so thatthe stability of the energy calibration of the instrument is crucial and should befrequently checked using a small source. The> 1350 keV window is much more stable.2.9.5 Neutron Activation Analysis (NAA)Neutron activation analysis relies on the production of radioactive nuclides in a sampleand the subsequent detection and measurement of the induced y radiation. Thetechnique can be readily used to determine U, Th and K concentrations at the levelstypically encountered in archaeological samples, although the accuracy will be dependenton the facilities and time available. The characteristics of the method are that easyaccess to a reactor and suitable counting facilities is required and also that only the Uand Th parents of the decay series are measured. This means that the calculated doserates are susceptible to error if there is disequilibrium present This can be used toadvantage when the method is used in conjunction with gamma spectroscopy or betacounting. Measurement of 210pO, at the end of the 238U decay chain, has been suggestedas an effective check on the assumption of equilibrium (Mangini et al, 1983). Thistechnique of NAA has been in routine use throughout this work and will therefore bethe subject of more detailed discussion in a later section (Chapter 4). Delayed neutroncounting has also been shown to be capable of rapid analysis of U in pottery samples(Filberth et al, 1980).502.9.6 Fission Track AnalysisThis technique involves the thermal neutron irradiation of a slice of pottery, followedby the observation and counting of damage tracks, caused by the induced fission ofuranium, in a mica or similar detector. It has been used to map the spatial distributionof U in a variety of archaeological samples, with a comparison of the distribution patternof the samples TL (Malik et al, 1973). The technique can identify minerals of high Uconcentrations, eg zircon, as well as zoning of the uranium within zircon grains (Suttonand Zimmerman, 1976). Irradiation with fast neutrons is used to determine Th contentof pottery, but Mejdahl and Wintle (1984) point out that it may be sufficient to irradiateonce, with fast neutrons since U undergoes fission with a probability of 4.5 times thatof Th. Then, since the ThlU ratio is usually -3-4, the dose rate could be calculatedfrom the total track density in the detector.2.10 EFFECT OF H20After the calculation of infinite matrix dose rates, from the measured concentrations ofthe radionuclides present in the sample and its burial context, and the application of therelevant attenuation factors to produce the effective alpha, beta and gamma dose ratesto a particular grain size and mineral type, there is need of yet a further correction. Thisis related to the reduction of dose rate received by the mineral grains due to the presenceof water in the sample and the soil. This is the case since the water reduces theradioactivity per unit mass compared to the dry situation, acting as an inert absorber ofthe radiation that could otherwise be involved in the production of TL within the mineralgrains. Water also has different absorption coefficients for alpha, beta and gammaradiation, in comparison to a ceramic matrix. Since the infinite matrix dose rates areusually calculated for the "dry" situation these factors have to be taken intoconsideration. Zimmerman (1971) estimated the effect of water content, in conjunctionwith the relative stopping powers or absorption coefficients for water and the matrix, toallow the formulation of correction factors relevant to alpha, beta and gamma radiation.These are given by:51DdryDwet - ---'---I +Ri f.w(2.21)where R = ratio of absorption coefficients between water and the matrixfor Cl radiationfor ~ radiationfor y radiationR = 1.50R = 1.25R = 1.14w = saturated water content expressed as a fraction of the dry weight of thesherd.f = the fraction of the saturated water content assumed to be appropriate forthe burial conditions.Sanderson (1987) suggested that the value of 1.14 for y is not appropriate since the ratiois energy dependent, and tentatively suggested a more suitable value to be between 1.05and 1.10, while Rendell (1985) adopted a value of 1.00, following Bowman's suggestion(1976). The energy dependence is illustrated in Figure 2.1 (after Tyler, 1992) whichshows a contrast between the ratios of absorption coefficients of a standard soilcomposition (after Beck, 1972) and water. The graph shows that for higher energies avalue of 1.10 is appropriate although Monte Carlo simulation to take into account theeffects of scattering and the calculation of a weighted average value for the ratio wouldbe desirable.2.11 RELATIVE DOSE CONTRIBUTIONSTable 2.6 gives the dry infinite matrix dose rates, calculated for a "typical" matrixcomposition. The relative contributions of these dose rates has also been calculated forfine-grains and inclusion, both in a dry state and also with water present. This is toillustrate the effects of grain size and water content on micro dosimetry. The resultingdifferences influence the measurements required for and the applications of specificdating techniques. These techniques are examined in the following sections.52Mass Attenuallon Coefficient. Waler Relallve lo SOU1.2 r-----....,....-----r-------,1.11.00.90.80.7- Beck's Soli- Walerf100 1000 10000Energy keYFigure 2.1 Energy dependence of mass absorpt Loncoefficients of water relative to Beck's soil.2.12 FORMULATION OF SPECIFIC STRATEGIES FOR ROUTINE DATINGOF SAMPLES FROM KNOWN GAMMA FIELDS2.12.1 IntroductionThis section will relate the micro dosimetric considerations, examined earlier to thespecific strategies that have been formulated for routine dating of samples from knowngamma contexts. The main features of each method will be covered and in latersections that describe the experimental component of this thesis the relevant points willbe expanded upon. Some aspects of sample preparation are common to all approaches,for example, the need to measure the water content (Section 2.10) and the removal ofa 2-3mm layer from each surface of the pottery fragment. The reasons for this arethreefold; - firstly, to remove that part of the sample that has received a beta dose fromthe soil which, because of the range of beta particles, will be attenuated over this regionwhere there will be a beta dose gradient between that due to the soil and that due to thepottery itself. Secondly, the outer surfaces of the sample may well have been exposedto daylight, resulting in reduced levels of TL and thirdly, any adhering soil must beremoved because of its potentially high level of geological TL. 1M Hel is used to53remove carbonates that may cause spurious TL.2.12.2 Fine-grain datingIn the fine-grain method, developed by Zimmerman (1967, 1971), the material used fordating consists of polymineralic grains in the 1 to 8 micron range. Grains of this sizeare small enough not to cause very significant attenuation of the alpha dose, and willtherefore experience the full dose rate due to the surrounding matrix. The sample is firstdisaggregated, taking care not to produce fine-grains by crushing larger inclusions. Thedesired grain size is obtained by suspending the 'crushed' material in acetone and thenselecting the fraction that settles between times calculated from Stoke's Law.(Zimmerman's standard procedure is to collect grains that settle between 2 and 20minutes, for a 6cm column of acetone.)Since the grains are of a size that do not significantly reduce the alpha dose from thematrix, measurement must be made of their alpha efficiency. This can be achieved byresuspending the selected fraction in acetone and depositing it as a thin layer of a fewmicrons onto aluminium or stainless steel discs.A potentially advantageous aspect of the fine-grain method is that since it includes thealpha dose rate contribution it reduces the relative importance of the environmentalgamma dose rate.2.12.3 Quartz inclusion datingAn alternative approach to the fine-grain method was developed by Fleming (1966,1970). In the quartz inclusion technique, quartz grains in the 90 to 150 micron rangeare selected by sieving. The dosimetry of grains of this type is such that the beta dosefrom the matrix will only be slightly attenuated (by less than 10%) and the alpha dosecomponent is very significantly reduced. The small alpha dose contribution can beeffectively eliminated by etching away the outer surfaces of the grains, usingconcentrated HF which will also remove feldspathic minerals, leaving an enriched quartz54fraction. The etching away of the outer surfaces of the quartz grains may notnecessarily be isotopic and may also result in variations in grain transparency which mayhave implications for beta source calibrations (Bell and Mejdahl, 1981).2.12.4 Feldspar inclusion datingInitial observations of fading of the TL signal in volcanic feldspars (Wintle, 1973)hindered the use of these minerals for routine dating purposes. Subsequent work,however, has shown that this is not necessarily the case for all feldspars (Guerin andValladas, 1980; Mejdahl and Winther-Nielsen, 1982; Mejdahl, 1983; Sanderson, 1987).Mineral separation is achieved by use of sodium poly tungstate solutions made up todifferent densities eg the SURRC laboratory uses 2.51, 2.58, 2.62, 2.74 glcc to givemineral separates of K-feldspar (2.51-2.58g/cc), Na-feldspar (2.58-2.62g/cc),quartz/plagioclase feldspar (2.62-2.74g/cc) and heavy minerals (>2.74g/cc). As in quartzinclusion dating the low alpha dose fraction can be further reduced by dilute HF acidtreatment.Contrary to the initial doubts expressed towards this group of minerals, feldspars haveseveral advantages over quartz. The TL signal is significantly brighter 10 to 50, ormore, times that of quartz and the broad glow shape often gives very long age plateaux.Mejdahl (1983) has investigated the use of large grains (e-Imm) of potassium rich(>10%K) feldspars for which more than 50% of the dose is contributed by beta radiationfrom inherent potassium. Consequently, the errors associated with the measurement ofenvironmental radiation and external beta radiation are considerably reduced.552.13 PROPOSED STRATEGIES FOR DATING SAMPLES FROM UNKNOWNGAMMA FIELDS2.13.1 IntroductionThis section will describe a range of approaches towards dating samples from unknowngamma fields. These include estimation and reconstruction as well as attempts to useaspects of microdosimetry that allow alternative solutions to the age equation to beproposed.2.13.2 EstimationObjective authenticity testing of museum pieces and art ceramics is obviously verydesirable information, but it is also readily apparent that this type of material would beextremely unlikely to have any information regarding the environmental gamma doserate associated with it. As a consequence, the errors tend to be of the order of ± 20%,however, the qualitative nature of the question being asked of an authenticity test meansthat the uncertainty in the gamma dose rate generally becomes unimportant (althoughthe possibility of old fakes should be borne in mind). Fleming (1975, 1979) has carriedout extensive work in authenticity testing and through the measurement of theenvironmental gamma dose rate at a large number of locations has been able to derivean estimated value (within certain error limits) that can be usefully applied in manyinstances.2.13.3 ReconstructionIn the case of archaeological material from old excavations there may be certaininstances where some form of reconstruction of the gamma dose rate within the burialcontext may be possible. Some sites, for example, Pool in Orkney, exhibit remarkablylittle spatial variation in gamma dose rate, from which it may be possible to infer withsome confidence an appropriate dose rate for areas not measured. Alternatively, it may56be possible to perform some small scale re-excavation on the site that would allow atleast some dose rate measurements to be made.2.13.4 Subtraction MethodsFleming and Stoneham (1973) proposed a combination of the fine-grain and inclusionmethods that allowed an age to be determined that was independent of knowledge of theenvironmental dose rate. This 'subtraction-technique' is based on the differences (andsimilarities) of microdosimetry for two techniques. The age equation for the twotechniques can be expressed as follows:FG - (env + P + au) A (2.22)I - (env + fP) A (2.23)FG and I are the stored doses in fine-grains and quartz inclusions respectively, env. isthe environmental dose rate, a and ~ are infinite matrix dose rates from alpha and betaradiation, a is the alpha efficiency factor, f is the factor for attenuation of beta radiation,and A is the age. Subtraction yieldsFG - I - ((1 - f) P + au) A (2.24)because the environmental dose rate is the same for both grain sizes.The age determination depends largely on the alpha dose rate because the (l-f) factoris very small. The accuracy of this method is limited to about 12% by this dependencesince the alpha dose rate is particularly sensitive to fluctuations in radon levels andvariations in water content. The best results will be obtained when FG-I is large.57Examples of the use of this method can be found in Fleming (1979).A feldspar-quartz subtraction method, based on large grains of equal size of alkalifeldspar and quartz, was proposed by McKerrell and Mejdahl (1981). In this case, thetwo age equations are:(2.26)where F and Q are stored doses in feldspar and quartz, ~s and ~F are infinite matrix betadose rates in the sample and from the potassium in the feldspar grains, f1 is the betaattenuation factor and f2 is the absorbed of the infinite matrix beta dose in the feldspargrains.The subtraction gives:(2.27)The age determination is based only on the beta dose rate from potassium rich feldsparsand the need for information regarding environmental radiation and sample beta doserate has been removed. The main difficulty is the measurement of large quartz grainswhich has hindered any widespread use of the method.A third variation of the subtraction method has been proposed by Plachy and Sutton(1976) that has been used for dating hearths containing burnt stones of granite andquartzite/sandstone. The latter rock type is generally low in radioactivity and records58essentially the environmental dose stored over the burial period. The granites, however,usually have high radionuclide concentrations. The granite can then be dated bysubtracting the environmental dose (obtained from the quartzite/sandstone) from theirstored dose and then dividing by the difference in internal dose rate. This approachautomatically corrects for any variation in environmental dose rate over the burialperiod. It should be noted, however, that the spatial arrangement of the rocks mayinfluence the environmental dose to some extent.2.13.5 Zircon MethodsThe presence of zircon grains in ceramic matrices was noted through the observation ofclusters of tracks on induced fission track maps (Zimmerman, 1971). The high densityof tracks is due to the high U content of zircons (up to thousands of ppm). Theresulting internal (alpha) dose rate is sufficiently high to effectively make both soilgamma and sherd beta dose rates negligible. This means that no knowledge is requiredof the burial environment or water content of the sherd.Despite these favourable characteristics, there are also a number of drawbacks associatedwith zircons. Many grains exhibit short-term fading, and there is often zoning of theirradioactivity and TL sensitivity, the two being anticorrelated. The latter point introducespractical difficulties concerning the geometry for artificial irradiation.Sutton and Zimmerman (1976) developed a procedure for zircon dating, based on singlegrains in an attempt to overcome the difficulties posed by the grain to grain variabilityof TL sensitivity and internal radioactivity. The stored dose is determined from the ratioof the natural TL and the artificial TL induced by a pair of 241Am sources positioned oneither side of the grain. A correction has to be made since the grain size (approximately60 microns or larger) is large compared to the alpha particle range. This technique isalso dependent on there being no significant change in sensitivity between first andsecond glows. The dose rate is calculated from the U and Th concentrations obtainedby induced fission methods. The results of a test programme were encouraging althoughthe large amount of effort required, plus the number of possible problems, have59combined to give a technique that has not found widespread dating use.An alternative approach (Sutton and Zimmerman, 1979) was investigated in which theproblem of zoning was overcome by comparison of the natural TL signal with the TLsignal that accumulates again after a storage period as a result of the grain's highinternal radioactivity. This method removes the need for artificial irradiation andradioactivity measurements and the age equation becomes:Natural TL .Age - . Storage T1IlleRegrown TL(2.28)This method suffers from two main problems, firstly the regenerated TL signal is verylow and secondly severe short term fading affects some grains. Regrowth andmeasurement of a much more sensitive low temperature peak (-120°C) has beenattempted by Templer (1985). Here, the regrowth of the high temperature peak isinferred from measurements of the relative sensitivities. This method requires linearityover the range of the storage time up to the age of the sample (2 - 3 orders ofmagnitude). The current lack of accurate zircon dates suggests that these limitations arestill significant.2.13.6. Decay MethodsCharalambous and Michael (1976) claim that it is possible to obtain an artificial glowcurve identical to the archaeological one, and from the experimental conditions(temperature of irradiation, and dose rate) under which the identity is achieved, it ispossible to estimate the archaeological age of the sample.The method does not require knowledge of the annual dose rate. This advantage iscountered by the need to know the "equivalent" temperature of the sample over its burialtime, and the assumptions that the TL sensitivity does not change after the first heating60and that the distribution of changes in traps of different depths is independent of doserate and irradiation temperatures.Langouet et at (1980) also proposed a decay method, D.A.T.E., based on mineralshaving two peaks - the first whose lifetime is too short for linear accumulation oftrapped charges and the second, a high temperature peak with linear accumulation oftrapped charges. The age can be determined by comparison of the trapped charges.2.13.7 Mejdahl's ApproachMejdahl (1983) described a dating programme of ceramics and burnt stones, in whicha variety of minerals and grain sizes were used. Large grains (l-2mm) of potassiumrich feldspar receive more than 50% of their total dose from their internal potassium,while plagioclase and quartz inclusion have progressively less internal beta activity. Theage equation for grains of various size and mineral type from the same sample can berewritten as:Stored Dose - Age(env + BM + Ba> (2.29)Stored Dose - Age. env + Age. B (2.30)BM = Beta dose rate from matrixBG = Grains' internal ~ beta dose rateB = BM+ BGenv = environmental dose rateA plot of the stored dose as a function of B will give a straight line of slope A and y-intercept A.E. Consequently, the age of the sample can be determined from the storeddose and beta dose, independently of the environmental dose rate.61Greater accuracy is theoretically achievable using this approach, compared to Flemingand Stoneham's "two-point" isochron, since more than two quantities are measured.However, the technique does require the sample to contain a sufficiently large varietyof mineral types and grain sizes and the ability to successfully separate them.A variation of the above approach has been described by Goedicke (1985), that utilisesthe effect of grain size on effective dose to provide the differences in stored dose. Theapproach was tested uses quartz grains (up to 750 microns) with promising results.However, it seems unlikely in practice that many samples would yield a sufficient rangeof quartz grain sizes with favourable TL characteristics, and that the Mejdahl approachmay therefore be more readily applicable for 'routine' analyses.2.13 SUMMARYThis chapter has introduced dosimetry concepts and aspects of radioactivity anddeveloped them in relation to thermoluminescence dosimetry. The micro dosimetry ofthe methods formulated to date material from known gamma contexts has beendiscussed, followed by a critical presentation of some proposed methods for dealing withsamples from unknown gamma contexts.The conclusion is that a general isochron formulation is the only real approach.The next chapter will introduce the isochron model through the formulation of ageneralised luminescence age equation. The systematic investigation of potentialisochron systems through theoretical simulation of this model is then presented.623.ISOCHRONS I - FORMULATION AND SIMULATION633.1 INTRODUCTIONIn this chapter the basic concepts of pottery microdosimetry are developed to providea theoretical basis for the experimental work that follows in Chapters 4 and 5. Thesimple age equation is developed into a generalised form. It will be shown that this canlead to an isochron formulation whereby the relationships between stored dose andinternal dose rates for all mineral phases may be used to estimate TL age and externaldose rates. In this manner it is possible in principle to date destratified material. Thishas been investigated using a simulation program that makes use of attenuation factorscalculated by Mejdahl (1979; 1983) and Bell (1980). The program is described and theisochron plots produced are illustrated and discussed.3.2 GENERALISA TION OF THE LINEAR AGE EQUATIONThe simplest form of the TL age equation, as given in section 1.4, isTL A e _ Stored Doseg Dose Rate(3.1)However, as Chapter 2 has illustrated, the dosimetry of pottery sherds is dependent ona number of factors (eg grain size, mineralogy, water content, origin of the dosecomponents, ie from within the sherd (or grain) or from the surrounding burialenvironment, etc). In order to enable practicable solutions of the age equation to beproduced these factors have to be taken into account and certain approximations andassumptions made. This approach to the problem has resulted in the formulation of the"fine-grain" and "quartz-inclusion" techniques which apply to specific grain sizes andmineral types. These techniques also reflect the influences of their differences inmicrodosimetry. The inclusion method, for example, in which the alpha contributioncan be removed (or greatly reduced) by HF etching, is more sensitive to the externalradiation field than the fine-grain method (unless high potassium feldspar inclusions are64used, see section 2.12.3 and Mejdahl, 1983). In the fine-grain method, the internal alphadose contribution from the sherd (and grain) becomes significant, despite the lowerefficiency for producing TL for alpha radiation.It then follows that as a result of their microdosimetric differences, grains of differentsizes and mineral types from the same sherd (and therefore of the same TL age) willalso have different stored doses. It can be concluded that strictly speaking each grainsize and mineral type has its own unique TL age equation. Alternatively, if the aboveconcepts are considered for a general case by defining the absorbed dose fractions andattenuation factors in tenus that apply to all grains and then, imposing the isochronconstraint, the microdosimetric contrasts within each sample can be explored.As a starting point the dry infinite matrix dose rates for the grain, G, pottery matrix, Mand surrounding soil, S, are defined as:·0 oM ° SDII DII DII·0 oM . SDp Dp Dp·0 oM . SDy Dy DyThe mean absorbed dose fraction, <1>, summed over all nuclides, for grain species (d) isdefined as~(d) - (3.2)where fj is the fractional dose contribution for nuclide i and 0j is the absorbed dosefraction for nuclide i.65It is also necessary to define and to take into account water attenuation factors,f~·scx.,~,'Y, which compensate for the fact that water present within the system will actas an inert absorber of radiation that would otherwise be involved in the production ofTL within the mineral grains. These correction factors have been described in Section2.10. Similarly, the lower efficiency of alpha radiation for producing TL must also beconsidered, and the alpha efficiency factor, k, must be incorporated into the fully,generalised form of the TL age equation.In its generalised form, using the following symbols for Luminescence Age (A), StoredDose (Q) and the effective dose rate terms ('I') for grain species (d), the luminescenceage equation can be written asA (d) -Q(d) (3.3)where the effective dose rate terms ('I') to grain species (d) are defined asG'G M M'M 8 s,.·s1f pCd) - ~p Dp + fwp (1- ~p ) Dp + fwp (1- ~p) Dp1f Cd) ~ GDO + fw M(I_ ~ M)D M + fw 8(1_ ~ s,. D sy - y y y y y y y) y(3.4)where 0 = mean absorbed dose fraction(1-0) = attenuation factor for a grain of diameter,d, micronsD = dry infinite matrix dose ratefw = water correction factorG = mineral grainM = pottery matrixS = surrounding soil66k = alpha efficiencyThe full form of the age equation (3.3) includes terms which are negligible in practice.For example the gamma dose of the grain and matrix, and the external alpha and betacontributions will all be effectively zero in practical applications. Thus if the followingapproximations are made~GDG 0 (1)y yfwM(l_~M)DM 0 (2)Y Y YS ~ . s 0 (3)fw~ (l-~I! DI!s ~ . s 0 (4)fwu(1-~u Duwhere(1) = gamma self dose of the grain(2) = attenuated gamma dose from the matrix(3) = attenuated beta dose from the soil(4) = attenuated alpha dose from the soilThe effective internal and external dose rate components, 'I'intG> and 'I'extG>, can bedefined as(3.7)(3.8)This reduced form of the equation for practical applications thus becomes67A(d) _ Q(d)'P intmlal (d) + 'P cxtemal (d)(3.9)This illustrates the possibility of resolving the dose rate term into internal componenrsand external components with reference to the sherd.3.3 FORMULA TION OF AN ISOCHRON APPROACHIf one considers different mineral phases from within the same sample then they willexperience a common external component, but their internal dose rates may differ andhence their stored doses per unit time. Equation 3. can then be rearranged into theformQ(d) - A(d)'P inII:mal(d) + A(d)'P GXt.amaI(d)(3.10)For grains of common age the system is linear with stored dose and internal dose rate.These are measurable parameters and from a plot of internal dose rate versus stored dosethe slope is equivalent to the TL age and the y-axis intercept is the product of the TLage and the external (gamma plus cosmic) dose rate. This is an isochron solution.This approach enables the TL age of a sample to be determined independently of anyknowledge of its burial environment and at the same time an estimate of this soilgamma dose rate can be obtained.Isochron dating systems yield more information than simply an age estimation, inisotope geology studies for instance, where these systems are routinely used, theintercept gives an indication of the initial isotope ratios. In TL studies, as stated above,the extra information is a measure of the environmental y + cosmic dose ratecomponent Two examples of this approach are given by Mejdahl (1983) and Goedicke(1985), and a third, using ESR, by Karakostanoglou and Schwarz (1983).683.4 ABSORBED DOSE FRACTIONS AND ATTENUATION FACTORSTo apply this concept the microdosimetric basis of fine-grain and inclusion approachesmust be extended, ie in order to be able to increase the range of grain sizes normallyemployed, to include intermediate and larger grains, the appropriate absorbed dosefractions (q,) and attenuation factors (1 - q,) must be known. These values have beencalculated by Mejdahl (1979) for the beta dose and Bell (1980) for the alpha dose andthe attenuation factors for gamma radiation which are very small have also beendetermined by Mejdahl (1983) and Carriveau (1983).These have been calculated for quartz grains and form the basis of the simulation workto follow, they are presented in Tables 3.1, 3.2 and 3.3.69Grain size Uranium Thorium Potassium(microns)1 0.010 0.009 05 0.053 0.045 010 0.108 0.093 015 0.166 0.141 020 0.228 0.193 030 0.372 0.310 040 0.498 0.433 050 0.586 0.529 060 0.650 0.598 070 0.697 0.651 080 0.734 0.692 090 0.762 0.725 0100 0.785 0.752 0120 0.813 0.784 0150 0.856 0.833 0200 0.892 0.874 0250 0.913 0.899 0300 0927 0.916 0400 0.946 0.937 0500 0.957 0.950 01000 0.978 0.975 02000 0.900 0.990 03000 1.000 1.000 04000 1.000 1.000 05000 1.000 1.000 06000 1.000 1.000 08000 1.000 1.000 010000 1.000 1.000 0TABLE 3.1 ALPHA ABSORBED DOSE FRACTIONS (AFTER BELL, 1980)70Grain size Uranium Thorium Potassium(microns)1 0.002 0.002 0.0025 0.012 0.0225 0.001810 0.0214 0.0366 0.003515 0.0296 0.0484 0.005320 0.0366 0.0582 0.007130 0.0475 0.0743 0.010640 0.0564 0.0875 0.014150 0.0642 0.0988 0.017760 0.0713 0.1088 0.021270 0.0779 0.1181 0.024880 0.0840 0.1269 0.028390 0.0901 0.1351 0.0318100 0.0957 0.1427 0.0354120 0.106 0.158 0.0424150 0.124 0.165 0.053200 0.146 0.206 0.0702250 0.169 0.229 0.0877300 0.189 0.251 0.1052400 0.230 0.288 0.1402500 0.263 0.320 0.17481000 0.400 0.443 0.3372000 0.568 0.604 0.5713000 0.667 0.701 0.6964000 0.731 0.760 0.7655000 0.773 0.798 0.8076000 0.803 0.825 0.8378000 0.842 0.859 0.87310000 0.866 0.879 0.896TABLE 3.2 BETA ABSORBED DOSE FRACTIONS (AFTER MEJDAHL, 1979)71Grain size Uranium Thorium Potassium(microns)1 0 0 05 0 0 010 0 0 015 0 0 020 0 0 030 0 0 040 0 0 050 0 0 060 0 0 070 0 0 080 0 0 090 0 0 0100 0 0 0120 0 0 0150 0 0 0200 0 0 0250 0 0 0300 0 0 0400 0 0 0500 0 0 01000 0.001 0.001 0.0012000 0.002 0.002 0.0023000 0.003 0.003 0.0034000 0.005 0.004 0.0045000 0.007 0.007 0.0076000 0.008 0.008 0.0088000 0.010 0.010 0.01010000 0.015 0.015 0.015'A HI-fi: 3.3 GAl\IlMA ABSORBED DOSE FRACTIONS (AFTER ME. DAHL,1983)723.5 SIMULA TIONS FROM THE ISOCHRON MODEL3.5.1 IntroductionBefore investigating the isochron behaviour of luminescence samples experimentally itwas decided to simulate the system in a computer model to explore the extent to whichmeasurable differences in Q(d) and 'I'(d)inlemalmight be expected. A simulation modeland computer program were developed and used to investigate systematically the effectsof variations in grain size, alpha efficiency, and internal radioactivity on isochronsystems. Sensitivity analysis and the effects of random measurement errors on isochronage determination were also undertaken.3.5.2 The Isochron Simulation ProgramThe isochron simulation program uses chosen values of matrix and grain properties(radionuclide concentrations and water content for the matrix, and size, internalradioactivity and alpha efficiency for the grain) to calculate the total effective internaldose rate to the grain. These calculations are carried out as described in previoussections and use published values for dose rate conversion factors, water correctionfactors, and absorbed dose fractions, as noted. The program has been structured so thatit can be readily modified to incorporate any recalculated values. A flow diagram ofthe simulation program is given in Figure 3.1.A value for the external gamma dose rate, Dr, can be chosen and for the purpose ofthese simulations an age of unity assumed. In this case, the stored dose will benumerically equivalent to the internal dose rate plus Or. A plot of the internal dose ratevs the stored dose will have a slope of 1, and a y-axis intercept equal to Or.The program uses a weighted regression analysis routine to determine slope andintercept values (and associated errors). This feature, combined with the option tointroduce random errors into the internal dose rate and stored dose values allows theerror sensitivity of the model to be investigated.73STARTtRead infinite matrix dose rate conversion factors}Read absorbed dose fractionstRead grain sizestPrintout required ?~Input matrix radionuclide concentrationstCalculate dry infinite matrix dose rates~Input matrix water content{Calculate effective matrix dose rates.IInput grain size, internal a~tivity and alpha efficiency~Calculate effective dose rate to grain!grain size ? ~y--------~/--------~N ~ ~/~ _,RegreSSi~l~.a_n_a__l_Y_S_i_S__?__Y >__~~~Performs linearregrrssionPlot points ? -y~----~?~--~N! PlotstpointsRandomise table ?Y~ ~~~__~~l r ~Random~ses tableAnotherENDFigura 3.1 Flow diagram of the isochron simulation program.74The data can be obtained in hardcopy either as a tabulated printout of dose rate valuesor as an isochron plot or both.The following section demonstrates the use of the program to investigate variousisochron systems.3.5.3 Investigation of the model: Simulated isochron plotsThe simulation program described above was used to investigate systematically thecharacteristics of idealised isochron systems using precise values. The factorsinfluencing microdosimetry; radio nuclide content of matrix and grain, water content,grain size and alpha efficiency, as well as measurement errors were examined to assessthe feasibility of practical measurement of such systems.Initially, a simplified situation of activity free grains within a matrix of "typical"radionuclide concentrations (after Mejdahl, 1984) was examined for two cases. Thefirst, shown in Figure 3.2, shows the influence of grain size differences on themicrodosimetry (grains of 10, 30, 60, 90, 150, 500 Jim were chosen). The second(Figure 3.3), illustrates a potential isochron plot based on the variation of alphaefficiency, the chosen values being 0, 0.05, 0.10, 0.15, 0.20, 0.25).755Slope 1 t - 0 10 micronIntercept .75 + - 04- 60 micron~CD..!!.3G.Ien SOO micronClc:::l-tu....Cl- 2ISOOlIDN FOR ACTIVITY FREE QUARTZ: Hatrix 3 ppm U, 10 ppm Th,3I K201o 0 1 2 3 4Internal Dose Rate (mSy/a)5Figure 3.2 The grain size isochron (soil and matrix watercontent = 20%. Fixed alpha efficiency = 0.15)76I~ FOR ACTIVITY FREE GRAINS Matrix 3 ppm U 10 ppm rh 3X K207.5Slope 1 + - 0Intercept .75 + - 0 /00256- /0015=--CD..!!4.5IU ~enS = 0.05-cv~0..,_3//1.5 V/10 micron grainso 0 1.5 3 4.5 6 7.5Internal Dose Rate (mSy la)Figure 3.3 The alpha e f f i ci ency i sochron .(soi I and matrixwater content = 20%, fixed grain size = 10 micron) .77The following series of simulations, the grain size variations are the same as those usedin Figure 3.2, and the alpha efficiency variations are those used in Figure 3.3.The investigation of the effect of varying activity within mineral grains in a matrix ofconstant composition was based on Mejdahl's (1984) typical matrix. This allowed asystematic approach with the mineral activity being varied as a fraction of the matrixactivity. These values are given in Table 3.4.Mineral Activity U Th K20(Fraction of ppm ppm %Matrix value)0.05 0.15 0.5 0.150.10 0.30 1.0 0.300.25 0.75 2.5 0.75TABLE 3.4 VALUES OF MINERAL ACTIVITY USED TO INVESTIGATETHEIR EFFECT ON THE ISOCHRON PLOTSFixed matrix compositionFixed soil Dr = 0.75 mfly.a'Fixed soil H20 = 0.20Fixed Matrix H20 = 0.20Figures 3.4 (a)i to (a)iv show the effect of variation in mineral activity on the grain sizeisochron and Figures 3.4 (b)i to (b)iv show the effect on the alpha efficiency isochron.The effect of matrix variations was based on data from Sanderson (1987, Table 4.3 andFigures 4.3 and 4.4) presented here as Table 3.5.789'-lin lctiritl • 0SlOllt I f - 0.75 f - 02 J ~brtmI31 !me Rate 116\'/~fIouN 3 .• (.1 Effect of internal activity . Grein siz. isoctvon jbrlin /.ctiritr • 0.10 Matri! AdiritrSl~ t I - 0ln~ .751-0""'3.4C.11i2 J ~brtmI31 !me Rate II6\'/~,--------------------Grain lctiritl • 0.05 Mattil lctiritlSlOlle I f - 0lntCUllt .75 f - 0/Slope t f - 0Ag.... 1.4 Ca. IvAQ<.n 3.4 C.IIbrain /.ctiritr • 0.25 Matti.!/.ctiritrlnterctllt .7S I - 0792 1 ~laterml !me Rate 116\'/~The series of isochron plots presented in Figure 3.4a show the effect of introducinginternal radioactivity to the grains. As the internal activity of the grains increases, thecoarser grains experience a greater increase in their total effective dose rate than thesmaller grains. The corresponding increase in stored dose for the coarser grains, relativeto the activity free case results in a shortening of the separation of the isochron points.80Grain icti,itr • 01.5 r--+---~--+--~-+---~--+-~J.5lnlerctllt .IS I - 0Sl~ I I - 04.5Agur. 3.4 Ibl Effect ol .... rnar octMty • Alpha .ICocioncy isocIvon iGrlin icti,itT • 0.10 llatrl.t ActlYitT1.S r---<---+-+----<--_+_~Sl~ (t-OiJIttmpt .IS I - 0Ag... 3.4lbllO4.S1.5Av- 3.4IblI1.5 r-----+------+--~--_+___Slapt I I - 0lnta'apt .IS I - 04.51.581AQ ... l.4 (blivGrdJl Actlrlty • 0.25 ICItrl.t Act!rity1.5 r---+--+----+--+--,__---,__...,._~1.5lataupt .15 I - 0Sl~ (t-o1.5 1.5The series of isochron plots presented in Figure 3.4b shows the effect of introducinginternal radioactivity to fine-grains (10 micron) of differing alpha efficiency. In contrastto the grain size plots, shown earlier, the effect here is to slightly increase the separationof points along the line. This results from the increased availability of an internal alphadose component that will be preferentially recorded in the grains of higher alphaefficiency.Site Uppm Thppm Th/U K2O%ratioBenderloch CB) 15.6 3.8 0.24 0.97Burnt Isles (BI) 0.93 7.1 7.6 1.92Craig Phadrig (CP) 2.2 8.8 3.96 4.33TABLE 3.5 VALUES OF MATRIX ACTIVITY VARIATION USED TOINVESTIGATE THEIR EFFECT ON THE ISOCHRON PLOTSActivity free grainsFixed soil gamma = 0.75 mGy/aSoil H20 = 0.20Matrix H20 = 0.20Figure 3.5 (a) shows the effect of matrix activity variation on the grain size isochron andFigure 3.5 (b) the effect on the alpha efficiency isochron. Figure 3.6 shows a summaryof these results by plotting the difference of the end members for each data set, fromwhich it can be concluded that the order of greatest separation of points is as follows:B.AE > B.GS > CP.AE > CP.GS > BLAE > BLGS.82~atrix 15.6 ppm U 3.8 ppm rh 0.97 I K20Activity Free Grains23 4.5 6Internal Dose Rate (mGy/a)o 0 1.5Figure 3.5 (a) Effect of matrix variations - Grain sizeisochron i837.5Criag Phaarig Hatrix 2.2 ppm U 7.1 ppm rh 1.92 X K20Acti'lity Free Grains5 .:Slope 1 + - 0/Intercept .75+-0 /4 /,:;::- /u::I..!!3 /aJen<:)Q-n:J.....<:).,_21o o 2 3 4Internal Dose Rate (mGy!aJFigure 3.5 (a) ii845BUPNf ISLES1o o .5 1 1.5Internal Dose Rate (mGy/aJ2 2.5Figure 3.5 (a) iii858ENIJEHLOCH AE12.5Slope 1 + - 0Interceot .75 + - 010-::00-c..D..!!.7.5QlIIn=c::l-co......=- 52.55 7.5 10Internal Dose Rate (mSy/a)o 0 2.5 12.5Figure 3.5 (b) Effect of matrix variations - Alpha efficiencyisochron i86BUHNT ISLES AE5Slope 1 t - 0Intercept .75 + - 0-~ J1c..c..!!3=encc:::l..-co......Q /'- 2o 2 3 4Internal Dose Rate (mSy/a)o 1Figure 3.5 (b) ii875CRAIG PHADRI6 AE7.5/Slope 1 t - 0/Intercept .75 + - 06 .:-=--c.a..!!. /4.5cuenS-co~Cl.....31.5o 0 1.5 3 4.5 6Internal Dose Rate (mGy/aJ7.SFigure 3.5 (b) iii88-o~~o~I]JE='Z 0o11] ToB.HEo B.Q3o CP .AE:o CP.GS0oBI. AEBI.GS7 4 6 8 10~NU(T1erlCal difference between x and y axis end pointsFigure 3.6 Summary of matrix variation effects.89The investigation into the possible range of isochron options was extended throughexamination of geochemical data to identify minerals of low, and high, internalradioactivity. Heavy mineral analyses of archaeological ceramics were studied toprovide feasible mineral assemblages from ceramic matrices. The overall aim being tofind isochron plots where the points were spaced as widely as possible so as to improvethe regression analysis. An example of this is shown in Figures 3.7 and 3.8.90ACTIVITY FREE GRAINS Matrix 1ppm U 3ppm Th 4% K205Slope 1 + - 0Intercept .75 + - 0'!omicron4::a-.c.o..§3 pOQ micron~en=~-to......=- 2/1 / Fixed Alpha Efficiency 0.i5o o 2 3 4Internal Dose Rate (mGy la)51Figure 3.7 (Activity free) grain size isochron in apotassium rich matrix.This plot shows explicitly how, in a potassium rich matrix where the dose rate is betadominated, an (activity free) grain-size isochron plot largely reflects beta attenuationcharacteristics.91GRAINS OF VARYING ACTIVITY Matrix 1ppm U 3ppm rh 4% K2027 ...Slope 1 + - 0 APATmIntercept .75 t - 022~('.!:)...§.16.QJenc~.-~......Cl.-U 90 micron grainsFixed Alpha Efficiency 0.155.5 BIOTITEUVlNEo 0 5 10 15 20 25Internal Dose Rate (mSy/a)Figure 3.8 (Variable activity) grain size isochron in apotassium rich matrix.In comparison to the previous figure, a much more promising plot may be obtained bymeasurement of grains of varying internal activity.923.5.4 Investigation of the effect of matrix water content variationThis was achieved by maintaining all dosimetric parameters constant, except for thematrix water content which was varied between 0% and 100% in 25% steps. Anassumption was that the soil water content was the same as that of the matrix (this isfor clarity and convenience, since in practice only soil y's would be effected by the soilH20, (a and Ws having been eliminated).Effective dose rates were calculated for three different values of water content. Theseare shown graphically in Figure 3.10 (a) for the grain size isochron and (b) for the alphaefficiency isochron.These plots show that for both isochrons the points are more closely spaced the greaterthe H20 content. For the grain-size plot the finer grains are affected to a greater extentthan the coarser grains this is due to the fact that the fine grains receive a greaterproportion of their dose from alphas which are attenuated by H20 to a greater extentthan betas or gammas.The rate of change in effective dose rate also appears to be greater at the finer end ofthe grain size scale. Similarly, for the alpha efficiency plot, the higher efficiency pointsare affected more than the lower points. This may have very important implications ifone considers a case where the matrix water content changes during the burialtime. Here, if a sample becomes significantly wetter, or drier at some stage of the burialperiod then the stored dose may be composed of two (or more) periods of differenteffective dose rates.Seasonal variation in H20 content will be averaged out and it is the longer termvariations that will be significant. Also, if different parts of a sherd have experienceddiffering water content histories, eg a thick sherd, perhaps with a dense centre and moreporous outer parts then even when the outer 2mm have been removed then the iscrushed for grain removal, the two different H20 histories will be mixed.93·1 .~.~ ..:::OJUI0 -7 4~ .'0'0OJs:.. 1.6I]~(j).8/---9:nw,/4-Figure 3.8 (a) Effect of different water contents - Grainsize isochron.o +-----~------~----~----~------+(l .8 1.6 2.4 :3.2Internal dose rate944 q• <....;OJUI0 ~ h.~ .-1J1J[I(~ ., 4I] ~ . ~-t-lI))1.Z...0..././ --O"lHIJ.-',/[]6o +-----~------~----~----~------+n 1 "7.~ 2.4 3.6 4.8dose ra teInternalFigure 3.8 (b) Effect of different water contents - Alphaefficiency isochron.953.6 ERROR SENSITIVITY OF THE MODELOne of the options available in the isochron simulation program was to be able tointroduce random errors into the data prior to regression analysis. This was achievedby generating two independent normal 0-1 variables, X 1 and X2, from two independentuniform 0-1 random variables, Uland U2 obtained from the RND function of the Applecomputer, using the following expressions supplied by Scott (pers comm 1988).Xl - J -2.1ogeUl . cos21tU2 (3.11)Xl - J -2.1ogeUl . sin21tU2 (3.12)The chosen percentage variations in the x and y axes, PI and P2 respectively are thenused to calculate the new values of x and y from the equation.x - x +(x +pl.Xl/lOO) (3.13)The calculation is performed on each x and y value in the data set and then regressionanalysis calculate the slope and intercept and their associated errors. The program wasalso modified so as to read in the data points for the perfect line plot (Figures 3.2 and3.3) from a separate file and to then carry out repeated randomisation of that file. Theprintout provides a record of the slope and intercept values with their errors which couldbe saved on disc. Their mean and standard deviation and the maximum and minimumvalues. A simple histogramming program was also included to produce a frequencytable of the results. A further modification was to produce a version that performed thesame function but with a printout restricted to the mean values, standard deviation,maximum and minimum values and the frequency table. The screen display was alsoreduced so as to reduce the calculation time for large numbers of runs. Histograms ofthe results of 500 repeated randomisation routines on the perfect line data are presentedin Figures 3.9 (a) and 3.9 (b) for various percentage random variations in the x and yaxes.96S I mu I I t II d II r r 0 r I 3 %x I n d 5 %y v II u. I1.11 l.rI •. u t." .... 1." 1.11 1.1' t.1I LII 1.44...... a •••Figure 3.9 (a) Error sensitivity - Grainsize isochron iMean 1.00 std.dev. = 0.13S I mu I It. d • r r 0 r s,ItItIt"ItII......II.t." .1.tI 1.11 I.U .... 1.11 .... to" t.U 1.6' 1.11I ... r •• , ••••• ,Figure 3.9 (a) IIMean = 0.76 std.dev. 0.3797SI mu t a t e d 8 r r 0 r I......u.,,,".."..:: .."t.""11I.Il '.1' '.n I.U I." Lt. 1.11 I." 1.11 1.4, 1."I't,_ rl'.'Figure 3.9 (a) 111Mean = 0.98 std.dev. = 0.16SI mu I II Id II r 0 r I 5' x and 5' y val UIS,..'",",II,"IIII. ".:: IIII..IIII11·1." ,'.11 ..... ".n ..U to •• .... .... t.a1 toU to"IlIer •• ,_ 'I'.'Figure 3.9 (a) 1VMean 0.81 std.dev. 0.4398SI mv I a t e d II r r 0 r I"',..,"'""":- ... ..;;:"••II..II1 0 %x In d 1 0 %y v II U III.... ,.U '.41 ...... It t." loti 1." 1.41 LII ioU..... ,a' ••Figure 3.9 (a) vMean 0.92 .28std.dev.S I mu I II II d II r r 0 r I 10 Ix I n d 10 Iy v II UII S.1.U -I.U ·1.11 t L" 1." I.U I.nIIIU",1 .tll'Figure 3.9 (a) VlMean = 1.00 std.dev. = 0.7999S I mu I a I 8 d 8 r r Q r I 3% x and 5% y val U81'.1'1 I." I.U 1.1' 1." 1.11 1.11 1.11 I."It.,_ ,•.••Figure 3.9 (b) Error sensitivity - Alphaefficiency isochron iMean 0.98 0.11std.dev.S J mu I It. d • r r 0 r s 3' x Ind 5' Y vII UIS.1.11 ••.• , 1.11 •. ., .... I.U .... LU 1.11 .... 1. ..,.,., ••,. u ...Figure 3.9 (b) IIMean 0.80 std.dev. 0.371005% x and 5% y val U81SI mu I at 8 d 8 r r 0 r I.........."....~ ... ..;: ...."IIII•. II t." t.'l •••• t... 1. II I. 11 1. II 1." I. 44 1. ..I,.,. ,a' ••Figure 3.9 (b) IIIMean 0.98 std.dev. = 0.13S I mu I at. d • r r 0 r I 5% I and 5% Y VII UIS.., ,-------------------------------------------------~'u'"."tt,'"..II......uIIIIII-I.U -ton -t.U I.U I.tt t.tt Lit ,-erl'ltt •• ,1 ,., ••Figure 3.9 (a) lVMean 0.83 std.dev. = 0.4610110 %X I n d 10 %y v II U 8 ISI mu I 118 d 8 r r 0 r I.11.11.".tI...• 11.11II. It.IIIIII••tI..11O.If I... O.fI I. II L 11 1... 1. II 1. fI 1." I. It I."".,. " ...Figure 3.9 (b) vMean = 0.94 0.26std.dev.10 %x I n d 10 %y v I I U. ISI mu I II. d • r r or s.11tUtil"'"'.11IIII,/II.;;:IIII..tI..11-4." _•• 1' .loU ·1.• ' ., ... -1.1t ·I.U I.U 1." 1.11 1.41'1 •• , •• ,1 ,cl ••Figure 3.9 (b) VlMean = 0.99 std.dev. = 0.901023.7 SUMMARYThe theoretical consideration of aspects microdosimetry have led to the formulation ofan isochron model which is based on a generalised form of the TL age equation. Thisallows solutions to be evaluated for this equation without knowledge of theenvironmental y dose contribution.A computer program has been used to simulate the general microdosimetric model. Thishas been used to investigate the effect of the variation of particular micro dosimetricparameters. The program uses values of absorbed dose fractions and attenuation factorsthat have been published for spherical grains of quartz. However, the structure of theprogram would allow different values, perhaps based on different grain shapes, to bereadily incorporated. Two main isochron solutions have been identified the alphaefficiency and grain-size (or alpha-beta-attenuation) isochrons.The effects of mineral and matrix activity variations and water content variation havebeen systematically explored. An investigation into the error sensitivity of the modelhas been presented and suggests that the value of the slope (the age) is more robust toscatter th_anthe intercept (measure of the external gamma dose rate).The next chapter describes an experimental investigation of isochron systems usingmodel matrices to evaluate the validity of the grain size isochron predictions.Promising isochrons do exist in principle. They have the potential to deal withdestratified samples but require lots of measurements.The next chapter explores isochron systems experimentally, using model matrices loadedwith grains of known characteristics. This represents a logical step between thetheoretical approach just described and the application to real samples.1031044. ISOCHRONS II - EXPERIMENTAL VERIFICATION1054.1 INTRODUCTIONIt has been shown by simulation that isochrons for grain size graduation, alphaefficiency variation and internal activity contrast in particular appear to be sufficientlypromising to warrant experimental investigation.In this chapter, a series of model matrix experiments designed to investigate thepredictions of the isochron model is described. This represents a logical step towardsthe development of isochron methodologies suitable for archaeological application. Asstated earlier the TL characteristics of archaeological samples are not only complex, butalso involve many different natural variables associated with both dosimetry andluminescence properties. The control of some of these parameters, particularlyknowledge of sample radioactivity, control of age, and selection of favourableluminescence characteristics, within a model experiment provides a practical test forunderlying isochron behaviour without introducing a need to characterise all possiblevariables which might influence real samples. These experiments were designed so thatthe precision required (section 3) to observe isochron behaviour should be readilyachievable with available techniques. They also provide an explicit test of themicrodosimetric assumptions that are fundamental to routine TL dating. The principleadopted was to prepare well-characterised matrices, derived from international secondaryreference standards for equilibriated uranium and thorium series activities. Thesematrices were loaded with TL materials which had previously been shown to have gooddosimetric properties, and exposed in a low background lead shield for accurately knownperiods. Recovery of the TL phosphors, using sample preparation techniques whichwould be needed to examine real archaeological samples, was followed by a series ofTL and radioactivity measurements and an isochron analysis.The simulation work had identified several possible isochron systems, based ondosimetric differences resulting from variations in grain size and alpha efficiency. Thefeasibility of preparing model matrices containing suitable TL mineral grains was firstconsidered. Itwas concluded that two separate matrices were required to examine grain-size isochrons. The first used microcIine feldspar embedded in an enriched radioactive106silicate matrix containing higher than normal proportions of U and Th seriesradioactivity; the second used CaF2 phosphor embedded in a more typical matrix ofslightly enriched natural radioactivity. To conduct this investigation of grain sizeisochrons, conventional inclusion separation methods (sieving, followed by densityseparation and acid treatment) had to be extended to extract grains of many sizefractions between I and 200 urn. It was decided to implement these two experimentalsystems, which are described in this chapter. The design requirements of alpha efficiencyisochron model experiments were also considered, based on the use of dosimetryphosphors with known and contrasting alpha efficiency. While it was decided not topursue these schemes experimentally, work was undertaken to develop methods fordensity separation of fine grain samples with a view to exploiting any variations betweenalpha efficiency of the various alkali and plagioclase feldspar fractions, and quartz.Here, the details of the preparation, storage, measurement and results of the two matricesinvestigating the grain size isochron system are presented.4.2 DESIGN CONSIDERATIONS FOR EXPERIMENTAL MATRICES4.2.1 ObjectiveThe aim was to design a set of experiments with a high probability of observingisochron behaviour, if it exists. These experiments were to be as realistic a simulationof dating systems as possible, while nonetheless controlling the number of experimentalvariables to minimise the probability of the underlying trends being masked byexperimental artefacts.The approach taken was to prepare a series of model systems whereby zeroedluminescence materials were incorporated into matrices containing known radioactivityand stored for known periods of time before extraction and luminescence measurement.Measurement of stored doses and evaluation of effective dose rates for each extractcould then be used to assemble isochron plots for age estimation. By comparison with107the known age the performance of the isochron system could be assessed. To maintainrealism the matrices and TL minerals should represent natural systems. To controlexperimental variability, reference materials and well characterised TL materials wereused.4.2.2 Materials and constraintsTo fulfil these experimental aims several aspects of matrix composition had to beconsidered. One requirement was that a sufficiently high dose rate could be achievedso that a measurable TL signal would be accumulated within the proposed storageperiod. The choice of dosimetry materials was also an important consideration. Naturalpotassium feldspars have been shown to be capable of detecting doses in the mGyregion (see Section 4.2.4) so a matrix capable of delivering a dose-rate of the order of100-200mGy per month would be appropriate. However, at the enhanced levels ofradionuclide concentrations required for this, the enrichment of the 4°K componentrequired to maintain typical U:Th:K proportions would be extremely difficult due to itsvery low natural abundance (0.0117%). One approach, to overcome this problem, wouldbe to undertake the production of an enriched source of 4°K using neutron activation ofa spectroscopically pure potassium salt e.g. KCl. This would have taken up to one yearto irradiate in the SURRC UTR thermal column and was not pursued further during thisstudy.The limitation presented by the readily achievable level of 4°K enrichment lead to adecision to use two different radioactive matrices to investigate the grain-size isochronsystem. To record a measurable stored dose from a natural mineral commonly foundin ceramic matrices, such as potassium feldspar, within the available experimental timeit was necessary to raise the matrix dose rate to a level which required predominanceof U and Th series radionuclides. Within this constraint, however, it was also felt thatas realistic as possible a U:Th ratio should be maintained, so that typical alpha spectrawere generated.It was also felt necessary to simulate a matrix that did have a more typical isotopic108composition so that a more representative spectrum of ionising radiation would bepresent. The problem of 4°K enrichment, mentioned above, acts as a restraint on thelevels of U and Th enrichment available, whilst keeping radionuclide proportions withintypical natural limits; high potassium enrichment also leads to modification of the bulkmatrix composition. The dose rate achievable under these conditions is significantlylower than available on the basis of U and Th concentration, and consequently dictatesthe use of a different dosimetry material. Environmental dosimetry work (Sandersonand Postlethwaite in East, 1991) had previously identified CaF2.nat as being able todetect doses in the ~Gy region using SURRC TL readers, making it a suitable phosphorfor this particular matrix (see section 4.2.4).The matrices themselves were prepared from a series of IAEA gamma ray spectroscopystandard reference materials and Canmet standards whose values are given in Table 4.1.The compositions of the planned matrices are described in section 4.2.5 and detaileddescriptions of their preparation are given in section 4.3.2 and 4.4.2.A literature search was carried out to assess the variation of reported values of alphaefficiency. This was to give an insight into the feasibility of exploiting the differenceswithin an isochron approach. Several limitations were apparent as discussed below. Aspreviously stated, section 2.8.2, alpha efficiency varies with alpha particle energy andhence alpha efficiencies measured by straightforward comparison of TL per Gy alphato TL per Gy beta will be dependent on the particular alpha source used. Measuredalpha efficiencies given as 'a-values' do not suffer from this problem; however thissystem has not been widely adopted outside the TL dating literature, and many literaturesources for dosimetry phosphor specify neither the alpha source used nor the efficiencysystem. Thus comparisons must only be tentative at this stage. Another problemcommon to most papers is that the quotation of the associated error term is missing orwhere an error has been evaluated, its method of calculation is not presented. The datingliterature, based on polymineral fine grain samples shows evidence of sample to samplevariation by a factor of up to two. Since both mineral composition and pure mineralalpha efficiencies are varying within these bulk samples, there is a possibility that109greater contrast may occur within certain samples due to variations between pure finegrain silicate phases. Lack of clear knowledge of this possible variation precluded thedesign of a fine grain alpha-efficiency contrast model experiment. A limitedinvestigation that attempted to separate, or at least enrich, particular mineral phaseswithin the fine grain fraction from ceramics did not produce evidence of markeddifferences in their measured alpha efficiency. An alternative approach could be to lookat the alpha efficiency as measured for fine grain samples prepared from crystals ofdifferent mineral types.The observation that some TL phosphors do exhibit a much greater range of alphaefficiencies offered the possibility of constructing an alpha efficiency isochron modelmatrix experiment. (For example, quoted (Aitken, 1978) alpha efficiencies for CaF2:nat= 0.10, CaS04:Dy = 0.30, CaF2:Dy = 0.50). At this stage, however, it was feltinappropriate since this would require the use of minerals not commonly found inceramic matrices and therefore not representative of a likely microdosimetric situation.4.2.3 Matrix EffectsThe previous section described possible grain/matrix combinations. Here, potentialcomplicating factors are identified and briefly discussed with a more detailed treatmentin the results and discussion sections.Firstly, the matrix composition has a bearing on both experiments. The proposedmicrocline/model matrix composition was of a silicate base with enriched uranium andthorium levels. This means that the 4°K part of the energy spectrum would be missing.In the case of the fluorite/model matrix, the radio nuclide proportions used would reflectmore closely the natural ionising radiation energy spectrum. However, to achieve thisand a useful dose rate required the matrix to be approximately 75% K2S04 ie it is nota silicate based matrix and therefore has a different stopping power. The use of fluoritegrains in this matrix also requires consideration of its stopping powers for alpha and betaradiation, compared to those for quartz (for which the self-dose fractions and attenuationfactors have been calculated).110One possible option to overcome the non-silicate nature of the matrix and to maintainthe potassium dose contribution would be to use a high potassium silicate. This wouldbe one of the potassium rich feldspars, sanidine, orthoclase or microcline. This in itselfwould introduce drawbacks since the highest potassium content of these feldspars is oforder of the one quarter that of potassium sulphate, so that lower dose rates would beachievable and therefore longer exposure times would be required.Standard Uppm Thppm K%RGU-l 400 ± 2 - -RGTh-l 6.3 ± 0.4 800 ± 16 0.02 ± 0.01RGk-l - - 44.8 ± 0.3S16 445 ± 16,800 ± -BL3 10,200 ± - -TABLE 4.1 IAEA AND CANMET STANDARD REFERENCE VALUES(NB Certified values were obtained from Report - IAEAlRUI48, 1987 whichaccompanied the standards.)It is worth noting that Sanderson's (1987) observation of the effect on the thick sourcebeta counting of reducing sample size for KCI (Zeff = 18.06) was found to be verysimilar for the fluorite/model matrix (~std(II)) (z = 13.6) which suggests that for thisparticular parameter the variation of z for the matrix has very little effect.Secondly, the configuration of the matrix and grains was considered. The possibilitieswere for the grains to be separated from the matrix by a thin foil containment butsurrounded by the matrix for the exposure period. This would then make the separationof the grains from the matrix extremely straightforward. However, the problemsintroduced by this approach, attenuation of the alpha dose and possibly the same effectfor the low energy end of the beta spectrum, and, depending on the geometry therewould be the possibility of self-shielding by the grains, were felt to outweigh itsadvantages. The decision was therefore taken to mix the grains intimately with thematrix since this is the configuration found in ceramic matrices.111Another important factor in deciding on the volume of container to use is thesignificance of any edge effects, due to the finite size of the matrix, ie the dose will belower at the edges of the matrix volume. Large radii containers would have the smallestvolumes experiencing the edge effect. Large volumes could present problems in thequantity of material required to make sufficient matrix to fill them, and the possibilityof a gamma self-dose may no longer be negligible. The limiting factor for the minimumsize is the maximum beta particle range of 2-3mm which would therefore require aradius of material of at least 3mm to ensure infinite matrix conditions.Alternatively, an inner container of <"~ thick" plastic sheet could be used to make acontainer for matrix plus grains, -2mm in radius less than the outer container the spacebetween the two could be filled with matrix only. However, this is likely to producefurther edge effects, almost certainly for a's and probably to lesser extent for W s too(especially lower energy ones).The third approach would be to have the matrix/grains mixture in a smaller containerand to allow them to accumulate their dose in a configuration with no inner boundary.Once the storage period is over a cylindrical sleeve, of radius 2-3mm less than thestorage container could be inserted into the mixture.In this way it was hoped that it would be possible to exclude that part of thematrix/grain mixture that contain grains from the boundary layer, thereby ensuring thatthe majority of grains collected (if not all) would have experienced the full alpha andbeta infinite matrix dose rates of the matrix. This was the adopted approach, the matrixcontainer was 36mm diameter x 58mm height.The advantage of this last approach is that it mimics the real dosimetry situation asclosely as possible. The consideration of these factors has suggested possibledisadvantages of ~TLD (Bailiff, 1982). Here, phosphor grains are not mixed within thepottery matrix so that the contributions from beta particles of different energies isunlikely to be the same as for mineral grains within the pottery matrix.1124.2.4 Grain RequirementsOnce the limitation of 4°K enrichment had been identified and circumvented by theproposed preparation of two model matrices of different radionuclide concentration, thenext requirement was to match suitable phosphor grains with matrices. The desirableproperties of the phosphor were that it should possess good TL characteristics, ie itshould have bright (ie high TL sensitivity) reproducible TL signals with little or no shortterm fading, and that it should be as straightforward as possible to recover from thematrix.An experimental investigation into sensitivity ranges and minimum detectable doses inalkali feldspars (Sanderson et al, 1989) had identified a microcline feldspar (MCI ashaving a linear growth characteristic from 1 to 200 mGy (and beyond) with no evidenceof supralinearity (Figure 4.1). This feldspar was from a graphic granite (Chiapavalpegmatite) from South Harris and was kindly provided by J Jocelyn (Geol. Dept.,SURRC) as a pre-separated mineral fraction of approximately 160-210 11mgrains.12x10 3III:g0 0ItIII •L. 80x10..JI-""Ox10 •300·350 ·C16x10 •o ~----~-----+------~----~-----+o ""0 80 120Gamma dose (mGy)160 200Figure 4.1 TL response of microcline (Mel) feldspar inthe mGy dose region.113Extensive practical experience of advanced mineral separation techniques using solutionsof sodium poly tungstate and centrifugation developed from those of Mejdahl (1983) andSanderson (1987) coupled with the observation that the predominant constituent ofRGU-l and RGTh-l was quartz, (p = 2.65g.cm·3) (99.43% and 97.23%, respectively)meant that the recovery of the relatively "light" microcline feldspar (p = 2.54 -2.57g.cm-3) ought to be readily achievable. This mineral was therefore chosen as thedosimetry material suitable for use with the matrix providing the higher of the two dose-rates. It also has the favourable characteristic of being a common rock-forming mineraland therefore a frequent constituent of ceramic matrices.The less active matrix required a much more sensitive dosimetry material, capable ofmeasuring stored doses in the low tens of mGy region. Natural calcium fluoride is onesuch material - its use in thermoluminescence dosimetry eg in medical radiotherapy andradiation protection, is an obvious testimony to its high sensitivity. It is alsoadvantageous in that it exhibits negligible fading, 1-2% per year (Aitken, 1985) and thatit has a density of 3.18 g.cm' - offering the potential for very efficient recovery fromthe matrix. There are disadvantages, however, the high sensitivity (six orders ofmagnitude higher than most of the minerals routinely encountered in the 'IT.. datinglaboratory) means that there is a very significant risk of cross-contamination, unlessgreat care is taken, which could be potentially disastrous for a laboratory such as thatat SURRC working not only with archaeological material but also very low doseenvironmental dosimetry (Sanderson and Postlethwaite, in East (1991)) and high dosemeasurements (-200kGy) using glass microscope cover slips (Ward, 1988). The factthat there was no evidence for any cross-contamination during this work was the resultof the implementation of strict procedures to ensure cleanliness within the preparationand measurement laboratories.Natural CaF2 has been shown to contain radioactive impurities that can provide a "self-dose" of the order of lOOIlGy.a-1(Aitken, 1985). Although this would not be significantat the dose levels envisaged here investigation was made of the radio nuclideconcentrations within both the CaF2 and microcline phosphors.The most significant disadvantage of using CaF2 results from the difference in a, ~114stopping powers and gamma absorption coefficients compared to the usual datingminerals. This will be discussed in a later section and at this stage it was felt to be asuitable dosimetry material for this model matrix.4.2.5 Planned MatricesAfter consideration of the points raised earlier in this chapter - the problem of 4°Kenrichment, the resulting decision to employ two matrices of differing radionuclideconcentrations, and the identification of two suitable phosphors, the following matrixcompositions were planned:Du D~ Dr TotalU 739.27ppm (2054.4) 108.00 84.94Th 2840ppm (2098.8) 81.22 145.98K% 8xl0-3% - Neg NegTOTALS (4153.2) 189.22 230.92 1043.11U:Th 1:3.84 reI. 18% 22%contribTABLE 4.2 MODEL MATRIX I: MICROCLINE GRAIN SIZE ISOCHRONBased on a matrix composed of 40% ROU-I, 40% ROTh-I, 15% S16, 5% BL3, bymass.Here, as elsewhere in this thesis, D is used to denote dose rate. Hence, Du representsalpha dose rate.115Da. D~ DyU 36.15ppm (l00.5) 5.28 4.15Th 120.00ppm (88.7) 3.43 6.17K 34.14% - 28.35 8.23TOTALS (l89.2) 37.06 18.55 83.98U:Th 1:3.32 rel.contrib 44% 22%TABLE 4.3 MODEL MATRIX II: CAF2 GRAIN SIZE ISOCHRON*NB D values are in mGy.a".Based on a matrix composed of 76.20% RGK-l, 8.80%, RGU-l, 15% RGTh-l, by mass.Model Matrix IT represents an approximate ten fold increase in the radio nuclideconcentrations typically found in ceramic matrices.4.3 EXPERIMENTAL PROCEDURES ANDDEVELOPMENTS4.3.1 IntroductionThe following sections describe the experimental procedures that were followed toprepare the model matrices and their grains. This required the development ofestablished grain size and mineral separation techniques to apply to intermediate grainsizes. A routine designed for the measurement of the apparent alpha efficiency of thesephases is described.4.3.2 Matrix PreparationThe raw materials for the model matrices were radioactive standard reference materialsobtained from lAEA and CANMET. The certified radio nuclide concentrations are givenin section 4.2.2 (Table 4.1). These were used to characterise the response of thelaboratory thick source beta counter (T.S.B.C., Sanderson, 1988) to determine thethreshold setting for isotope independence, (where the sensitivity, or dose rateconversion factor, was the same for all three radionuclides). The use of an in-house beta116standard also enabled the comparison of the observed and expected dose rates from theproposed model matrix constituents. The in-house beta calibration standard refers backto CANMET standards that are certified independently of IAEA materials to ± 2%.Table 4.4 shows potassium sources consistently measured to ± I% of the certifiedvalues, thorium series to within 4% and uranium series to within 7%. IAEA materialsare only certified to ± 5% therefore this agreement is satisfactory. The radon lossprobably explains the slight under-response. Therefore the materials used can be shownto be traceable to international dose rate standards at high precision.Standard D~ exp D~ obs D~ obsID~ expmGy.a' nrGy.a'KCI (analar) 43.49 43.09 0.99RGU-1 58.47 54.37 0.93(IAEA)RGTh-1 23.82 22.82 0.96(IAEA)RGK-1 37.20 37.68 1.01(IAEA)TABLE 4.4 OBSERVED AND EXPECTED BETA DOSE RATES FROMSTANDARD REFERENCE MATERIALSHaving successfully measured the beta dose rates from the individual standards asecondary, mixed, beta standard, ~std(n), was prepared. This was designed to haveD~K = 0.78Di3-rOTAL'D~U = O.13Di3-roTALand D~Th = 0.09Di3-rOTAL.The weightingfactors were included to take into account the typical relative contributions of K, U andTh in natural matrices. This corresponds to a mixture composed of 76.20% RGK-I,8.80% RGU-1 and 14.96% RGTh-l, by mass, with a calculated beta dose rate of37.06mGy.a-1• The mixing procedure was to place the main constituent (RGK-1) intoa 150ml gamma counting geometry, to which the RGTh-1 was gradually added as anumber of separate aliquots, with thorough mixing of each aliquot. The RGU-1 wasadded similarly. The mixture was then transferred to a ball mill (Fritsch "analysette1173"/"pulversette 0") and mixed for 25 minutes (intermittent mode), then allowed to standfor approx 5 minutes to allow all dust particles to settle and so reduce any inhalationrisk.Two independent quality control checks were performed to confirm the matrixcompositions, ~-counting and neutron activation analysis. A 30g subs ample of this~std(II) mixture was removed for ~-counting - the measured ~ dose rate was 36.55 ±0.42 mGya", which is 0.99 of the expected value (37.05mGya-I). This was a veryencouraging result since it confirmed the validity of the methods used to calculate andprepare the secondary ~std(II) mixture and that they could be justifiably applied to theproduction of the model matrices.The second method employed was neutron activation analysis (NAA). This utilised thereactor facility at SURRC and the close proximity of gamma detectors allowed theanalysis of short-lived isotopes. The quantitative analysis of the amounts of uranium,thorium and potassium present in a sample requires comparison with a standard ofknown concentration of these elements. An homogenised Shap granite sample hadpreviously been used for this purpose and was chosen for this study. The workingvalues of the uranium, thorium and potassium contents were derived from highresolution gamma spectrometry of sealed samples relative to Canmet standards and arebelieved to be correct to better than ± 3%. The values were 12ppm U, 28 ppm Th,4.43%K.The samples and standards were contained as accurately measured masses, usually 20or 50mg for the grains and 100mg for matrix samples, in individual silica vials (0.0.= 5mm, I. D. = 3mm, length = 40-50mm) flame sealed by Mr F Cornwallis. For rapiduranium analysis, the pneumatic transfer ("large rabbit") system was used to irradiatethe samples. The irradiation time varied depending on the reactor power, but wastypically of the order of five to ten minutes, followed by an approximately ten minutecooling period during which time very short-lived activity (eg 2813Al, t~ = 2.24mins)would decay and therefore reduce the dead time of the detector. The sample vials werealso cleaned during this period to remove any activated dust particles that could118contaminate the detector. The reaction being exploited here is:23&...... 23~ ...92U (n,y) 92U t 1 - 23.54 minutes Ey - 74.6 keV2(4.1)The low energy of the emitted gamma ray requires that a planar, or thin-window,detector be used and also that it should be unshielded, by lead at least, because itsenergy coincides with a lead X-ray at -74.5keV that would interfere with the true countsfrom uranium. This work used an EG&G Ortec portable low-energy gamma-raydetector, and was energy calibrated using 241Am (Er = 59.57keV) and 57CO (Er =122.06keV) reference sources.The potassium analysis depends on the following reaction~ 41 4~1~ (n,y) 19K (n,y) 1~ t ~ - 12.36 hours Ey - 1525 keV (4.2)In this case, the counting period (30mins) commenced after a cooling time of the orderof 4-5hrs after the initial rabbit irradiation and uranium count, on a 25cc GeLi gammadetector CZ26Ra calibration).In order to activate the thorium to a sufficiently high level a considerably longerirradiation time was required. This had to be carried out in the CVS (Central VerticalStringer) position since a six hour irradiation period was required and a maximum ofone hour was available from the rabbit system. The CVS could accommodate a totalof approximately 30 individual sample vials, contained in two aluminium cans. Afteractivation a cooling period of about one week was followed by a 30 minute count onthe 25cc GeLi detector, to measure the 312keV gamma ray from 233Paproduced from:232.- 23~90 In (n, y) 90 1"a t 1 - 26.95 days Ey - 312 keV2(4.3)119Although the precision of this technique (approximately ± 10-15 % for U, ± 10-15 %for Th, ± 5-10 % for K) is less than for the beta counter it does give information fromsmall samples and provides an independent cross-check for quality assurance purposes.4.3.3 Grain PreparationThe sedimentology literature (eg Tickell, 1965, Lindholm 1987 and McManus, 1988)was consulted for methods of grain size separation and analysis, but it was found thatrather than performing actual separation of the different grain sizes the approach was asampling strategy from which the grain size distribution was inferred. This wasconfirmed by a short series of experiments that showed that the "pipette method", inwhich a pipette is used to remove a sample of a sample of a suspension of sediment ina column of water at a particular depth and at a particular time after settling hascommenced, produced not only a small yield but that the samples were also poorlysorted.Some experiments were also made with a settling column but instead of having a closedend, like a measuring cylinder, the lower end was tapered and open. The reasons forthis were two fold - firstly it was felt that the usual method of decanting of the liquidfrom grains that had settled at the bottom was likely to disturb them, either through overenthusiastic decanting and/or resuspension, so that the expected grain size fraction is notthe one that is obtained. The effect of this could be reduced either by repetition of theprocedure or, as envisaged here, by use of an open ended tube that could be stopperedand then briefly opened after the appropriate settling time to collect the required grainsizes.Alternatively, the second approach, was to investigate the use of the open ended columnin a "flowing" mode. Here, trimmed micro pipette heads were fitted to the outlet endof the column, thus reducing the flow rate, in an attempt to ascertain if a flow rate couldbe found that would allow faster separation of grain-sizes than the usual settling in astatic column of water. Itwas found on microscopic examination of different separatesthat they were poorly sorted. Possible explanations were that the finer grains were being120dragged down by being caught in the paths of more rapidly descending larger grains,perhaps indicating overloading of the column. It was also noticed that some grainscollected on the tapering sides of the column on reaching the bottom rather than flowingthrough the outlet. As a result, although an attractive method in principle, the use ofa flowing column of water to separate different grain size fractions was not pursuedfurther.Dry sieving, using disposable nylon meshes (90f.lm, 60f.lm and 30fJ.ffi)in a small plasticsieve set, was found to be inefficient. This may have been due to the sample becomingcharged with static electricity as it was repeatedly rubbed against the nylon mesh,causing the grains to stick together and therefore inhibit passage through the sieve.The approach that was eventually adopted was wet sieving for grains >30f.lm and. settling in a static column of liquid acetone for the finer fractions. Deionised water wasused to rinse the finer than 90f.lm grains through the sieve, the water plus <90f.lm grainsbeing collected in a large beaker, until the water passing through the sieve was clear.This was taken to indicate that most <90f.lm grains had been washed through. Thecontents of the beaker were then washed through the next sieve (60f.lm), that whichpassed through being collected along with washings of fresh deionised water until it ranthrough clear. This was repeated for the 30f.lm sieve.The grains retained by the meshes were dried by placing tissues on the underside of thesieve mesh to absorb as much excess water as possible. The meshes were then carefullyremoved from their holders and dried in a warm oven. The <30f.lm grains, suspendedin a large volume of water, were transferred to centrifuge tubes and spun down to allowthe water to be more easily decanted. These grains were then resuspended in beakersof acetone (for shorter settling times than in water). Beakers were used since this meantthere was less likelihood of disturbing the settled grains when decanting than if ameasuring cylinder was used. The settling procedures were repeated for each sizefraction so as to improve the recovery of the desired grain size fraction. The grain sizeseparates prepared as above were examined with the aid of a microscope and found tobe well sorted, implying that the methods employed had been successful.121The separation of intermediate and fine grain sizes by settling in a static column ofliquid is based on Stoke's Law, which can be conveniently expressed as:t - (4.4)where t= settling time (s)111= viscosity of the liquid (N.s.m:3)h= height of liquid column (m)g= acceleration due to gravity (9.80m.s-2)Ps = density of mineral grain (kg.m")P1= density of liquid (kg.m")D= diameter of grain (m)Viscosity and density values were obtained from standard scientific data textbooks (CRCHandbook of Chemistry and Physics) from which settling times were calculated formineral grains in water. These results were in very good agreement with valuespublished in geological literature. As a further exercise the settling times (in acetone)quoted by Zimmerman (1971) to obtain grains in the size range of 1 to Sum wererecalculated.There was some disagreement between the two sets of values, so further calculationswere performed for different grain densities in an attempt to clarify the situation.Correspondence with Questiaux is gratefully acknowledged. These are presentedgraphically, in Figure 4.2.Once prepared, the fine grains were deposited onto 0.25 mm stainless steel discs byevaporation from an acetone column in a similar manner to that originally described byZimmerman (1971).1221c ~10-T':l..0 -.- .81-0-tJUIbl \~-"0-i .6,I]-I\T'-C- .4.... ,0 IC .I0- '7 !~ ..... /uI1J Is.. /u. I0 ./"0 3 6 9 12 15Di ·:Meter -CJE-.,;(1)150en ~a0 Expected-0 100 lsochron(1)L-a....en50oo 100 200 300 400 500 600 700 800Internal Dose Rate (mGy/a)FIGURE 4.7 MODEL MATRIX I : ISOCHRON PLOT.135Linear regression analysis of this data set gave the following results :-Measured valuesAge estimate =External Dy =0.288 ± 0.047 a17 ± 65 mGy/acompared to the expected age of 0.238 a and an external gamma dose rate of zeromGy/a. These results are extremely encouraging since the data not only exhibit theexpected trend but also provide a very good age estimate. The gamma dose rateestimate, however, reflects the greater sensitivity of this parameter to scatter in the data.This reinforces the predictions of the error sensitivity of the model explored in section3.6.4.4.7 Analysis and DiscussionThis experiment clearly demonstrated isochron behaviour. Overall the experimentalconfirmation of the predicted isochron behaviour of this sample, and its agreement,within two sigma, with the known storage period is very satisfactory. This isparticularly encouraging given the practical difficulties of making high precision TLreadings from natural feldspar systems at 50 - 150 mGy and the necessity of using anatypically high uranium series content, with its associated risk of radon leakage, as theirradiating matrix.The stored dose was determined from the direct comparison of the natural (first glow)and reirradiated (second glow) signals. The ratio of first to second glow curves wasfound, and integrated over 250-350°C for each disc. This temperature range was chosenbecause it gave the best plateau and also helps to reduce the detrimental effect of lowlight levels and the interfering effect of black body radiation. Despite this precautiona number of discs had to be excluded from the data analysis due to poor glow shape.The weighted mean values were deemed appropriate since some grain size rangesexhibited more scatter than others. The fine-grain fraction, although separated from thematrix and measured, had so much scatter that the data were considered unreliable, andexcluded from analysis; it is possible that this fraction included material within its136density band derived from the silicate matrix.The internal dose rate values were calculated using the measured radio nuclideconcentrations, from neutron activation analysis, of the matrix and grains, and theisochron simulation programme (using the absorbed dose fractions and attenuationfactors as described in section 3.4). This required knowledge of the alpha efficiency ofthe microcline (calculated here as air ratio for the fine-grain fraction), a parameter thatwas of particular importance in this case due to the matrix radio nuclide compositionbeing solely U and Th so that an unusually high proportion of the dose rate wasdelivered by alpha radiation. The problem of poor data quality for the fine-grain sizerange meant that the alpha efficiency value carried an unfortunately large error term (-± 25%). This is reflected in the magnitude of the dose rate percentage error terms forthe smaller grain sizes being approximately double that of the coarser fractions. Thesignificance of the accuracy (and precision) to which the alpha efficiency is known canbe illustrated by calculation of the internal dose rates for different values of the alphaefficiency. The reliability of the gamma dose rate estimate obtained from the regressionresults is more difficult to assess, since its value is sensitive to the regression line drawnthrough the measured points and it consequently has a large uncertainty associated withit. The background gamma dose rate in the lead-lined cavity used for the storage of thematrices had previously been measured as <0.3 mGy.a" by CaF2 TLD - interpreted asthe sum of self dose and hard cosmic ray dose rates. Measurements with a scintillationdetector produced a value of 0.05 mGy a' without the model matrices present, excludingthe hard cosmic component. This value however was later realised not to be applicablehere because of the increased gamma dose rate due to the presence of the other matrixexperiments within the cavity _A subsequent measurement with half the matrices presentproduced 0.17 mGy a', implying that the combined gamma plus cosmic dose rate withinthe shielded cavity when all portions of the matrices were present would be of the orderof 0.4 mGy a'. The absorbed dose fraction for a spherical rock sample of similardiameter to the cylindrical sample is approximately 0.1 (Mejdahl, 1983). Therefore thereis probably an internal gamma ray dose rate of 23 mGy a' in this matrix; considerablyin excess of the ambient gamma contribution. This contribution was treated as anexternal source in the analysis, which is justified here in that it represents a common137component to all grain fractions.Another possible source of error was the effect of radon loss. This has previously beenmentioned in connection with the variation of beta count rates obtained for the matricesat different stages of the experiment. The measured parent radionuclide concentrations(by NAA) were 724.22 ppm D, 2711.80 ppm Th and 0.744%K. Assuming no radonloss the expected beta dose rate from this matrix would be 183.99 mGy.a-1• The initialmeasured beta dose rate was 182.39 rrrGy.a' which agrees very well with the expectedvalue. After annealing the matrix its measured beta dose rate was 128.50 mfly.a" -suggesting that radon gas had been driven off. 100% radon loss would have resultedin a beta dose rate of 72.73 mfiy.a", the observed beta dose rate corresponding to 50%radon loss. The lower alpha efficiency lessens the impact of the reduced dose rate.Also, the effect of storing the matrix/grain mixture for 87 days, approximately 23 half-lives of radon, was to allow the gas to build up again. In any case, the effect of anyradon loss would be to reduce the effective internal dose rate and increase the slope ofthe measured isochron and therefore also increase the age estimate derived from thatslope.It is concluded then, that any effects of radon loss did not contribute to the slight ageoverestimation observed here. It was noted, however, that even after the storage periodthe beta dose rate was some 10% lower that the initial measured value.4.5 MODEL MATRIX IT: CaFz GRAIN SIZE ISOCHRON4.5.1 Experimental PlanThe previous section described an experiment in which a potassium feldspar (commonlyfound in ceramic matrices) was coupled with a matrix that provided a high dose rate butwith a radio nuclide composition ratio that was not representative of the relativeproportions found in naturally occurring matrices.138Here, the approach was to prepare a model matrix with a radionuclide content inproportions as representative as possible of those found in archaeological ceramic andlithic matrices. This automatically limits the extent to which the dose rate can be scaledup due to the problem of 4°K enrichment. This in turn precluded the use of the usualTL minerals extracted from pottery on the grounds that they would not be sensitiveenough to record measurable doses within the time available. Natural calcium fluoride(the mineral fluorite) is an extremely sensitive TL mineral and was therefore matchedwith this matrix.4.5.2 MaterialsThe preparation of this matrix was more straightforward than the previous one for twomain reasons. Firstly, there was less of a radiological hazard associated with preparationand handling since the constituents were less radioactive than those required for themicrocline model matrix. Also, the secondary beta standard composition was suitablefor this experiment and a further 100g was prepared as outlined in section 4.3.2.The actual matrix composition was 76.2g RGK-1, 15.0g RGTh-1, 8.8g RGU-l, whichcorresponds toDu DP Dy TotalU 36.15ppm (100.46) 5.28 4.15Th 120.00ppm (88.68) 3.43 6.17K 34.14% - 28.34 8.23Total (189.14) 37.05 18.55 83.97TABLE 4.10 ACTUAL COMPOSITION OF MODEL MATRIX IIThe results of beta dose rate measurements are given below, in Table 4.11.139D~ D~mea!D~caIcCalculated 37.05Initial 37.12 ± 0.47 1.002After Anneal 27.60 ± 0.52 0.745After Storage 35.18 ± 0.44 0.950TABLE 4.11 BETA DOSE RATE FOR MODEL MATRIX II COMPARED ATDIFFERENT STAGES OF THE EXPERIMENT(NB D values are in mfly.a')Mean Values, determined by NAA % of expected valueUppm = 29.53 ± 1.20 81.70%Thppm = 113.31 ± 3.99 94.42%K% = 32.62 ± 2.15 95.56%TABLE 4.12 RADIONUCLIDE CONCENTRATIONS OF MODEL MATRIX II,MEASURED BY NAAThe source material for the grains required for this model matrix experiment was thelaboratory stock of M.B.L.E. natural CaF2• Approximately 5g was removed and givenan initial annealing cycle of heating to -600°C for 10 mins, in air, followed by rapidcooling. Aitken and Fleming (1972) state that this treatment should not cause anyserious loss of TL sensitivity.The plastic sieves and disposable nylon sieve meshes were not used here, so as to avoidthe potential of residual fluorite contamination of the sieves. Therefore settling columns(open ended) were used to effect the separation into different grain sizes. The largestgrains present in the CaF2 "as received" were -150llm so crushing with an agate pestleand mortar was required to produce the small grain size.Great care was taken to avoid any contamination of the sample preparation laboratory,all bench surfaces were covered with large tissues that were dampened at the end ofeach day's CaF2 handling operations and then folded into themselves to minimise thespread of any stray fluorite grains. All apparatus that had been in contact with fluorite140was marked as such. These measures appear to have been successful in that there wasno subsequent evidence, whatsoever, of contamination.Two 50mg samples of the fluorite were analysed by neutron activation. The thoriumcontent was determined as 0.97 ± 0.26 ppm. Uranium and potassium were not detected.Therefore these will be taken as absent for the present purposes.4.5.3 Exposure DetailsThe mixing of the CaF2 grains with the matrix was carried out in the same manner asfor the microcline grains (section 4.4.3). The total mass of fluorite added to the 100gof matrix was 3.728g. The fluorite/matrix mixture was also annealed, prior to packagingand storage, to remove TL signals from the matrix and any signal that may have beenaccumulated in the fluorite since its initial annealing step prior to crushing. Themeasured beta dose rate of a sample taken from the annealed matrix was 27.60 ± 0.52mGy a', which is 74% of the beta dose rate recorded after initial mixing of the matrix,again implying some radon loss. The exposure period of 70 days was spent in a lead-lined low-background cavity to reduce the effect of any external sources of radioactivityon doses stored in the fluorite grains.4.5.4 Extraction Of GrainsThe initial unwrapping and removal of the outer -2 mm was carried out as for themicrocline/model matrix. The portion removed for beta-counting gave a measured beta-dose rate of 35.18 ± 0.44 mflya" (compared to 27.60 ± 0.52 mflya' at start ofexperiment, ie just after annealing, and 37.12 ± 0.47 mflya" after initial preparation ofthe matrix mixture).The high density of fluorite, 3.18g.cm-3, was exploited to facilitate efficient recovery ofthe grains from the matrix. A volume of approximately 200ml of a sodiumpoly tungstate solution of density 2.74 g.cm' was prepared, the aim being to carry outa large sample density separation, under the force of gravity rather than centrifugation.141Several factors had to be considered, for example the volume of liquid had to besufficiently large to be able to cope with a charge of the order of 40g, and with adensity high enough to separate the fluorite from the less dense matrix. Conversely,care also had to be taken to aim to successfully perform the separation as fast aspossible - since the extremely high TL sensitivity of the fluorite meant that extendedperiods spent out of the matrix, prior to measurement, could result in the addition of anon-matrix, environmental, dose being recorded. Therefore, the set up used was anecessary compromise between 'perfect' separation and "efficient" (near-perfect)separation.The matrix was gradually added to the density liquid, using a spatula to mix and stir thetwo together. Despite this approach, a number of floating masses formed - probably theresult of the outer parts becoming saturated with the liquid before the inner parts, thustrapping air and forming floating lumps. These were broken up as much as possiblewith the spatula before, and between, two fifteen minute periods of agitation in anultrasonic bath - an approach that proved to be successful.The settling period of approximately 3.6 days was spent in dark room conditions, awayfrom any obvious sources of radioactivity, with the beaker covered with cling-film toavoid dust contamination and evaporation. Two distinct layers were observed, thefloating one had a gelatinous appearance - reflecting its major composition of <38J,lmgrains of quartz suspended in a high density liquid.This top layer was 'spooned' off, leaving approximately 180ml of the liquid whichwould still contain some of the fine-grain CaF2 in suspension. This was transferred tocentrifuge tubes and diluted with deionised H20 washings from the settling beaker, andthen spun down. A further purification step, of centrifugation in a liquid of density2.97g.cm-3, was included and did remove lower density material from the CaF2concentrate. The fluorite grains were rinsed several times in deionised water with acentrifugation step between each. The rinsing was then repeated, using acetone insteadof water, and the grains placed in an oven at 50°C to dry overnight. Assuming all thedense grains collected to be fluorite it was calculated that the amount recovered142represented 73% of the mass expected to be present.The grains were separated out into their different grain size fractions in the same wayas employed during their preparation. Seven different grain size fractions were preparedbut whilst the two finest were still being separated, a test run of two discs of each ofthe five coarsest fractions was carried out.4.5.5 Measurement of Stored DoseThe initial test run involved a total of ten discs, two from each of the five coarsest grainsize fractions. These were prepared as for the microcline discs, but here the samplemasses were much less, of the order of O.5mg. The main runs, however, were composedof sets of eight discs for each grain size each of approximately 2-3mg sample mass.All discs were glowed in a single sample reader in an atmosphere of oxygen freenitrogen. The heating cycle terminated at 400°C to reduce the possibility of causingsensitivity changes between first and second glows due to the pre-dose effect. The"natural" dose stored in the fluorite during exposure to the model matrix was evaluatedby comparison of peak: heights to those induced by a 5mGy gamma dose delivered bythe 6OCOsource as described in section 4.4.5. A third glow after an alpha irradiation wasmeasured to investigate the variation of alpha response with grain size.4.5.6 ResultsThe results from both test and full runs of this experiment are presented below. Table4.l3 gives a summary of the test run, which indicated a progressive change in storeddose with grain size fraction. The stored dose measurements from the full runs arepresented in Table 4.14, and the corresponding internal dose rate values in Table 4.15.65 of the 66 individual aliquots measured produced glow curves characteristic of CaF2both in shape and peak: positions, indicating successful recovery from the matrix. The143other disc showed evidence of contamination and was excluded from the analysis.Grain Size' Stored Dose(mGy)(microns)PII2 PIIl3150 11.5 10.91504 - -115 14.4 12.2115 13.4 12.285 14.5 12.985 12.7 11.460 16.8 14.660 16.8 13.938 17.3 16.038 16.0 17.6TABLE 4.13 RESULTS OF MODEL MATRIX II EXPERIMENT: TEST RUNNotes1. Grain size given is the middle value of the grain size range separated.2. Calculated as PII height (Natural) x 5mGyPII height (5 mGy)3. Calculated similarly but with PIlI heights.4. 2nd glow curve for this disc obviously contaminated.144Grain Stored Dose (mGy) Alpha/gamma ratioSize(micron) PII PII PIlI PIlI PII PII pm PIlIError Error Error Error150 12.59 0.27 11.06 0.24 0.026 4x10-4 0.01 3xl0-4115 14.88 0.32 13.17 0.28 0.034 8xlO-4 0.02 6.5xlO-485 16.42 0.58 13.94 0.53 0.042 5xlO-4 0.02 2.5xlO-460 18.61 0.38 15.79 0.27 0.055 1x10-4 0.03 6x10-438 21.44 0.46 17.60 0.41 0.070 1.8xlO-3 0.04 Ll.x l O?20 21.23 0.31 17.71 0.37 0.099 2.2xlO-3 0.07 1.4xlO-310 25.60 0.41 20.05 0.41 0.135 3.4xlO-3 0.08 2.4xlO-3TABLE 4.14 RESULTS OF MODEL MATRIX II EXPERIMENT: FULL RUNGrain size PII PII error PIlI PIlI error(microns) (mGy/a) (mGy/a) (mGy/a) (mGy/a)150 34.05 1.67 33.08 1.67115 37.26 1.73 35.53 1.7285 39.10 1.75 36.85 1.7560 41.81 1.78 38.70 1.7738 45.69 1.81 41.26 1.7920 51.76 1.87 45.22 1.8210 54.46 1.90 47.00 1.84TABLE 4.15 DOSE RATE VALUESFOR MODEL MATRIX II145->-CJE.........Q)enoCl"'CQ)-o-en30201040Measured IsochronoExpected Isochrono 25 50 75Internal Dose Rate (mGy/a)FIG.4.8. ISOCHRON PLOT; FLUORITEIMODEL MATRIX1461004.5.7 Analysis and DiscussionBoth test and full runs showed clear indications of isochron behaviour. The isochron plotfrom the full run is shown in Figure 4.8. This figure also shows the expected isochron,based on the known age and the assumptions that the measured stored doses are correctand that the external gamma dose rate was zero.However, both data sets produced systematic overestimates of the known exposureperiod of 0.192 years. Using linear regression analysis, the test run gave an age estimateof 0.536 ± 0.072 years, while the full run produced an age estimate of 0.594 ± 0.047years.There are a number of factors that could have influenced the results of this experiment.These include the estimation of dose rates, source calibrations, and stopping powervariations. These are examined below.4.5.7.1 Estimation of Dose Rates.There was evidence in the reduced beta dose rate measurements taken after annealingthe matrices of radon loss in both experiments followed by partial recovery during theexposure period. Dose rate conversion factors applicable to 100% radon loss, whichconsiderably overestimate the possible effect within the matrix were taken from Aitken(1983), and used to calculate the expected dose rates for the microcline and fluoritemodel matrices. Since this involves a reduced dose rate it clearly cannot account forexcess signals.There was no evidence here for gamma cross-talk from the high to the low dose ratematrix as this would have effectively been an external dose rate component, andtherefore should be apparent as a positive intercept on the y-axis of the isochron plot.There was no such intercept which suggests that if the observed discrepancy was dueto excess dose rate, then this was present as an internal component.147The possibility of a gamma self-dose within this matrix can be considered. The infinitematrix gamma dose rate, calculated from the radionuclide concentrations, isapproximately 17 mGy/a. However, the absorbed gamma dose fraction for an object ofthe dimensions and density (-lg/cm3) used here is very small. The overall effect of aninternal gamma dose rate component is therefore not likely to be significant.The possibility of an excess internal dose rate component being derived through radonmigration, from the high UITh concentration matrix to the lower dose rate matrixcontaining the CaP2 grains, is examined here.Having calculated the expected dose rates (based on the measured radionuclideconcentrations of the matrix, attenuation factors (as described in section 3.4) and alphaefficiency, taken here as the aiy ratio for the fine-grain fraction) and knowing the ageof the samples, if the measured stored doses are assumed to be correct, then it ispossible to calculate the excess dose rate that would be required to give the known age.This is shown below in Table 4.16.148Grain Measured Required Excess dosesize dose rate dose rate rate required(urn) (mGy/a) (mGy/a) (mGy/a)PII 10 54.46 133.47 79.0120 51.77 110.69 58.9238 45.68 111.78 66.1060 41.82 97.03 55.2185 39.10 85.61 46.51115 37.26 77.58 40.32150 34.04 65.64 31.60PIlI 10 47.00 104.54 57.5420 45.21 92.34 47.1338 41.26 91.76 50.5060 38.70 82.33 43.6285 36.85 72.68 35.83115 35.53 68.67 33.14150 33.08 57.66 24.58TABLE 4.16 EXCESS INTERNAL DOSE RATE REQUIRED TO OBTAINKNOWN AGE.The next step is to calculate the maximum possible excess dose available. This wouldcorrespond to the situation where the high dose rate matrix suffers 100% radon loss -the available excess being the difference between full series and pre-radon dose rates.149MICROCLINE / MODEL MATRIX DOSE RATES(A) NO RADON LOSSU ALPHA Th ALPHA U BETA Th BETA2012.61 ± 77.45 2004.02 ± 68.07 105.74 ± 4.07 77.56 ± 2.63(B) 100% RADON LOSS912.52 ± 35.12 837.95 ± 28.46 43.45 ± 1.67 27.12 ± 0.92(C) DOSE RATE "LOST" [= (A)-(B)]1100.09 ± 42.33 1166.07 ± 39.61 62.28 ± 2.40 50.44 ± 1.71TABLE 4.17 INFINITE MATRIX DOSE RATES CALCULATED FOR (A) NORADON LOSS AND (B) 100% RADON LOSS FROM MICROCLINEIMA TRIXUsing the appropriate attenuation factors to calculate the effective maximum possibledose rate to each grain size, it then possible to determine the excess dose rate requiredas a percentage of this value, as shown in Table 4.19.150Grain Measured Required Excess dose Maximum (Required)/size dose rate dose rate rate required excess dose (maximum)(urn) (mGy/a) (mGy/a) (mGy/a) rate available (%)PH 10 54.46 133.47 79.01 375.16 21.0620 51.77 110.69 58.92 339.78 17.3438 45.68 111.78 66.10 259.48 25.4860 41.82 97.03 55.21 209.14 26.4085 39.10 85.61 46.51 175.32 26.53115 37.26 77.58 40.32 154.12 26.16150 34.04 65.64 31.60 121.17 26.08PIlI 10 47.00 104.54 57.54 273.21 21.0620 45.21 92.34 47.13 250.26 18.8338 41.26 91.76 50.50 198.81 25.4060 38.70 82.33 43.62 166.45 26.2185 36.85 72.68 35.83 144.46 24.80115 35.53 68.67 33.14 130.39 25.40150 33.08 57.66 24.58 107.88 22.79Mean 23.82Std. Dev. 3.07'ABLE 4.19 CALCULATED EXCESS DOSE RA' ~ESREQUIREDThe results presented in Table 4.19 suggest that the stored doses measured in the fluoritegrains might be explained if approximately 24% of the available excess dose rate(assuming radon loss from the microcline/model matrix experiment) were able tomigrate to the fluorite/matrix system. Despite the fact that measures had been taken toprevent, or at least limit, radon loss, the beta dose rate of the microc1ine/model matrixmeasured (thick source beta counting) at theend of the exposure period was lower thanthe initial (prior to annealing) value.151Expected beta dose rate = 183.9 mOy/a(based on measured (NAA) radio nuclide concentrations, no radon loss)Measured beta dose rate (initial) = 182.4 mOy/aExpected beta dose rate "lost" (if 100% radon loss)24% of expected beta dose rate "lost" (if 100% radon loss)= 112.7 mOy/a= 26.9 mOy/aIf the microcline/model matrix beta dose rate had equilibriated at approximately 24%down then the expected measured value would be 182.4 - 26.9 = 155.5 mOy/aActual measured value at the end of the exposure period = 163.7 mOy/aAlthough these figures do not provide conclusive evidence that this was the case, theyare perhaps indicative that the situation may have contributed to the observed results.However, if this were the case one might expect a corresponding increase in themeasured beta dose rate of the fluoritelmodel matrix. Table 4.11 shows that this wasnot the case, although this could be explained if the delay between removing the matrixmaterial from the container and starting the beta count was of the order of a few hours,during which time the excess beta activity could decay.At present, there is no clear answer and it is felt that further experimentation would helpclarify the situation, ego immediate beta counting of the fluorite/model matrix onremoval from containment after storage, followed by repeated counting to look forevidence of an excess beta dose fall off with time. Another approach could be to storea radioactively inert matrix, loaded with fluorite grains, in the same arrangement as hasbeen described above. Measurement of the stored dose recorded in the fluorite wouldprovide information regarding the presence, or otherwise, of an excess radiation doserate.The measured radionuclide concentrations in the fluorite were found using NAA. Thethorium content was determined as 0.97 ± 0.26 ppm, uranium and potassium were notdetected. Aitken (1969, 1985) states values of CaF2 self dose -OiIrnfly.a' to a levelwhich would not be significant here. Beta counting of a bulk sample of the fluoritesource material was considered.1524.5.7.2 Source Calibrations.The alpha source working calibration had been confirmed by a cross-comparison withthe Oxford laboratory. The gamma source working calibration had been provided byDr D Sanderson, from measurements made three months prior to its use for theseexperiments, decay-corrected to the irradiation date with a precision of ± 5%. TheSURRC beta source was originally cross-calibrated to a source at Paisley which hadbeen calibrated to both the Oxford and Riso sources to better than ± 2% precision in1987. Subsequent experiments at SURRC cross-compared the beta source with a 200TBq gamma source by TLD which was also calibrated by National Physical Laboratorytraceable photochromic plastics (Harwell red perspex dosimeters). Both of thesemeasurements agreed to within 0.8% (Ward, 1988). These have been well traced to. national and international standards; furthermore coherent TL dates have been obtainedfrom known-aged Roman samples and from a number of well-characterisedarchaeological sites where radiocarbon evidence is also available. Additional confidencein the source calibrations is provided by the observation that the measured fine-grainalpha efficiencies agree well with published values.4.5.7.3 Effect of Stopping Power Variations.The use of potassium sulphate as the main matrix component introduces the possibilityof error in absorbed dose fractions. To evaluate the possible implications the betastopping power contrast between this matrix and typical silicates was estimates as below:Beta stopping power is proportional to Z, as calculated from153where Ni is the relative number of nuclei of charge Zi (the theoretical basis for thischoice of Z is given in Cross (1968) (NB Z for an element = Z.)Calculated values of ZSi 14Si02 10.80K2S04 14.35CaF2 14.79Stopping powers for compounds and mixtures are additive, in terms of the weightfractions fZiof elements of atomic number Z, then(From Attix 1986 p 178)It can be calculated that the ~ stopping power of a K2S04 matrix is 5% lower than asilicate matrix. However, dose rate is inversely proportional to the stopping power ratioso the ~-dose rate will be 5% higher.CaF/K2S04 mean ~ stopping power ratio = 0.99CaF/Si02 mean ~ stopping power ratio = 0.94.Therefore the dose recorded by fluorite grains will be - 6% higher than for quartzgrains in a silicate matrix. In a potassium sulphate matrix however the use of CaF2matches matrix properties within 1%. Neither effect is sufficiently pronounced toaccount for the gross excess age estimates.Alpha particle range is approximately proportional to AY,(Bragg-Kleeman rule) whereA = atomic weight of the material (for compounds and mixtures) the effective A canbe found from:154(4.6)where n, is the atomic fraction of the element AiK2S04 A eff = 26.01CaF2 A eff = 26.92ex, stopping powers can be found from interpolation of values given in data tables egNorthcliffe and Schilling.Al (A = 26.98)Ne (A = 20.18)The alpha stopping power of a K2S04 matrix is 14% lower than a silicate matrix. Thisimplies that the K2S04 matrix will have an alpha dose rate that is 16% higher than asilicate matrix.The alpha stopping power ratio of CaF2: Quartz = 0.598/0.667 = 0.8966. Therefore thealpha dose recorded in fluorite grains will be - 10% less than for quartz grains. Thisimplies an overall increase of - 6% in the effective alpha dose rate for fluorite grainsin a potassium sulphate matrix. This in itself is insufficient to explain the experimentalresults obtained.4.5.7.4 Other considerationsAnother potential variable is the possibility of non-uniform distribution of grain sizeswithin a particular specified range, eg 1O-301lm treated as 20llm may contain more"finer" grains or vice versa than expected. The effect of this would be greater for thesmall grain size ranges with the result being essentially to pivot the isochron line. Thedirection of the pivoting action would depend on whether the grain size distribution wasbiased to the finer or coarser end of the range. The possibility that the crushing actioneither generated a higher than expected surface to volume ratio, or preferential defectstructures on grain surfaces, leading to high alpha efficiency was considered briefly.Although further experimental work would be needed to specifically exclude thispossibility, it was noted that the measured alpha efficiencies, by external radiation155followed a similar grain-size dependence to that described by Aitken (1985), implyingthat these effects were not dramatic.4.6 SUMMARYThis chapter has described the use of model matrices to evaluate the validity of theisochron predictions of Chapter 3. This required the identification of suitablecombinations of materials from which to prepare the two proposed grain/matrixmixtures. The development of grain size and mineral phase separation techniques toinclude intermediate grain sizes has been described. The preparation and execution ofthe model matrix experiments have been described in detail.Isochron behaviour was clearly observed in both experiments. The microcline feldsparexperiment produced an acceptable age estimate from a high dose rate matrix, despitehaving an atypical U, Th, K mix. The fluorite experiment also produced clear evidenceof an isochron with good precision. These age estimates were, however, systematicallygreater than the known age by a factor of approximately 3. Having considered themagnitude of numerous possible systematic error terms it was concluded that the samplewas probably inadvertently exposed to an additional internal radiation source, mostlikely the consequence of unexpected radon migration from the adjacent higher activitysource. Nevertheless these results show the potential of model experiments, and alsosuggest improvements that could be made in the experimental design for furtherinvestigation and subsequent studies of real samples.1565.DISCUSSION AND CONCLUSIONS1575.1 SUMMARYThis work has aimed to contribute to the solution of a problem relevant to archaeologyand microdosimetry, namely the TL dating of destratified archaeological ceramics. Thebackground to this problem has been presented within the context of the historicaldevelopment of thermoluminescence dating (Chapter 1) and of dosimetry concepts(Chapter 2). It was noted that much valuable archaeological material, including manyceramics from important sites excavated during the early part of this century, and havingan influential role in the establishment of archaeological chronologies and sequences,exists in museums. Absolute dating of this corpus of material would be extremelyvaluable to archaeology in allowing historic hypotheses to be tested. However, this isnot possible using conventional approaches to thermoluminescence dating since theserequire full solution of the TL age equation for individual mineral phases. This in turnrequires knowledge of the external radiation dose rate during burial; comprised ofgamma and cosmic radiation. In the case of 'fine grain' dating the external componentprovides a variable fraction of the total dose rate, perhaps typically some 25%; for'inclusions' this proportion may be larger, perhaps 30% or more. To maintain datingprecision of 5% overall, it is necessary to estimate the external component within aprecision of some ± 20%. Where direct in-situ measurements on the excavation site arenot feasible, or where representative samples of the burial media are unavailable analternative approach is needed.Previously proposed strategies (for example the 'subtraction' method which combinesfine grain and inclusion approaches, the use of large inclusions with varying betaattenuation, zircon dating) were considered, and reviewed critically. It was shown inChapter 3 that these approaches were simply special cases of a general model for theTL age equation whereby the age can be extracted, independently from the externalradiation dose rate, using an isochron approach. The isochron model was developedexplicitly and theoretical simulation used to explore different micro dosimetric systems.Two main isochron solutions were identified, the grain size (or alpha-beta attenuation)and alpha efficiency isochrons. The simulations could be used to advantage byindicating the most promising experimental strategy for a given sample matrix and158mineralogy, ego for a matrix containing a high potassium concentration there would bethe potential to exploit grain size effects resulting from beta attenuation. However, fora low potassium matrix these effects would be less pronounced and so the alpha doserate component becomes more significant, thus increasing the possible benefits of anyvariations in alpha efficiency that may be apparent.Given the constraints of materials and time an experimental approach using two modelmatrices was employed (Chapter 4) to evaluate the validity of the grain size isochronpredictions. Isochron behaviour was clearly observed in both cases. The microclinefeldspar experiment produced a consistent age estimate from a high dose rate matrix.It should be noted that the microcline grains used here are of the sort of grains onewould realistically find in archaeological samples. This, coupled with the fact that thelow stored doses encountered in this experiment limit precision, make the result obtained(measured age estimate = 0.29 ± 0.05a, expected age = 0.238a) very encouraging andsupportive of the isochron model. The fluorite experiment was expected to provide highprecision verification of the isochron model. It did produce a highly coherent data setconsistent with the expected sensitivity and reproducibility of CaF2• However, the ageestimate obtained was greater that the expected value by a factor of approximately 3.This result proved difficult to explain and a number of possible reasons were examined.It was eventually concluded that the possibility of radon migration from the microclinematrix into the fluorite matrix may have resulted in an unexpected alpha and beta dosecontribution to this second matrix, due to the decay of radon daughters. This radonmigration hypothesis could be tested in various ways, ego by measuring the betacountrate of the fluorite matrix immediately after removal from storage and thensubsequently to monitor for an initial excess countrate followed decay to a constantvalue, or by storing a blank (radioactively inert) matrix, loaded with fluorite grains, ina similar geometry to that used for the model matrix experiments described in Chapter4. The test for radon migration would be positive if, after storage, the measured storeddose in the fluorite was found to be greater than the combined effect of cosmic and selfdose rate components. Alternatively, the high and low dose rate matrices could havebeen isolated physically, by storing them in separate low-background lead-lined shields,or temporally, by conducting the experiments at different times.1595.2 FUTURE DEVELOPMENTSThe most obvious area for immediate further work has been highlighted above, ie. theneed to resolve the discrepancies observed in the experimental results obtained fromcontrolled systems. To achieve the ultimate aim of this work - to obtain TL isochronage estimates from archaeological material - an important, if not essential, intermediatestage would be to test the method on known age materiaLThe transition from controlled systems to real samples introduces some challengingvariables, ego a ceramic fragment will contain a finite assemblage of grain sizes andmineral types that will exhibit a range of TL characteristics. The optimisation ofseparation techniques for different grain sizes and mineral types should bestraightforward to achieve through the development of current methods.The use of complementary techniques to measure radionuclide concentrations of themineral separates (eg ICP-MS, although it involves a more complex preparation stage)would provide a useful check on NAA results. Fission track mapping of the uraniumand thorium distribution may also be usefully introduced to provide information relevantto alpha dosimetry.In terms of the large number of measurements required to produce an isochron plot, theuse of automated readout and irradiation facilities would not only reduce the number ofman-hours spent making these measurements but could also improve precision throughmeasurement of greater numbers of discs per sample. The availability of a gammasource of a suitable dose rate would be an advantage in cases where very coarse grains(and therefore very useful isochron data points) are present in the sample being dated,since the problem of attenuation of the dose from a laboratory beta source does notapply.Recent developments in luminescence dating, particularly the use of light to release thetrapped charge, may have much to offer isochron methods. The potential of being ableto carry out rapid high precision ED measurements has obvious advantages. However,160it is also apparent that it may be possible to use different stimulation wavelengths thatare mineral specific ego exposure to green light will stimulate luminescence from quartzand feldspar, whereas infrared wavelengths stimulate luminescence from feldspars. Itmay, therefore, be possible to reduce the level of mineral separation that is currentlyrequired. Development of spatial imaging techniques, ego imaging photon detectors, mayprovide sufficient resolution to image individual grains, perhaps as small as fine grains,and so eliminate the need for grain size separation of a sample.5.3 CONCLUSIONSThe need for an archaeological solution to the problem of dating destratified samples hasbeen clearly demonstrated. It has been shown that isochron methods provide a generalsolution to the problem of eliminating the unquantifiable external dose components fromTL dating. By simulation it has been demonstrated that measurable contrasts in storeddose between extractable phases from many ceramic samples are predicted using thestandard dosimetric models that form the basis of TL dating. Experiments with modelmatrices have confirmed the general behaviour. The results from a microcline loadedmatrix of high U/Th contents were highly satisfactory, and provide support for theisochron approach. The results from another experimental simulation using CaF2 appearto have identified a pitfall, in the form of unexpected radon crosstalk, which must beinvestigated further and taken into account in future experimental work.Archaeological samples will present a more variable system. However, there aregrounds for believing that these obstacles are not insurmountable. Developments inautomation of TL dose estimation, ego the use of photostimulation, and in single grainimaging techniques may significantly enhance the practicability of isochron approachesto TL dating. It is also possible that detailed studies will lead to improved models forthe microdosimetry of TL dating. It remains to be seen whether isochron methods willbecome standard approaches to archaeological dating. However, it is clear that theyoffer a potential solution to the problem of dating destratified samples.1616. ADDENDUM162Having carried out the experimental investigation of the isochron model as described inthis thesis, the results from the fluorite matrix remained problematical to explain in asatisfactory manner. It was therefore decided to take advantage of having prepared andstored two portions of each matrix. The assistance of Dr. Sanderson in carrying out thiswork is acknowledged here. Two approaches were followed; firstly, the measurementof the beta countrate from the fluorite matrix from the time of removal, to look for apossible excess countrate that might indicate the effect of radon crosstalk from themicrocline matrix, with continued counting to monitor any subsequent decay of thecountrate. The results are given below, with the values from the first portion forcomparison.D~ D~mea/D~calcCalculated 37.05Initial 37.12 ± 0.47 1.002After Anneal 27.60 ± 0.52 0.745After Storage 35.18 ± 0.44 0.950Second portion(approx. 5min after 35.75 ± 0.42opening)Second portion(approx. 20hr after 35.21 ± 0.42opening)TABLE 6.1 BETA DOSE RATE FOR MODEL MATRIX II COMPARED ATDIFFERENT STAGES OF THE EXPERIMENT(NB D values are in mfly.a')The conclusion drawn from these results was that there did not appear to be anyevidence for radon crosstalk.The second approach was to dry sieve this material to obtain some -100 micron fluoritegrains. These were washed, dried and then two discs were prepared. The predicted .,.stored dose in these grains, assuming no excess dose rate, was calculated as shown163below.Inf. matrix dose Attenuation Effective dose Effective doserate mGy/a factor( 1-<1» rate Pll rate Pilla eff 0.135 a eff 0.085Ua 100.46 0.215 2.92 1.84Th a 88.68 0.248 2.97 1.87U~ 5.28 0.904 4.77 4.77Th ~ 3.43 0.857 2.94 2.94K~ 28.34 0.965 27.35 27.35TOTAL 40.95 38.77TABLE 6.2 CALCULATED EFFECTIVE DOSE RATE FOR 100 MICRONFLUORITE GRAINS.(Infinite matrix dose rate data from Table 4.10, attenuation factors derived from Tables3.1 and 3.2, alpha efficiency as measured previously from fine-grain fraction).Having calculated the expected dose rates, as shown above, and knowing the exposureperiod (4.25 years), the expected stored doses can also be calculated;-PH : 174 mGyPIn: 165 mGyThe natural TL was glowed out and then a beta dose of approximately 168 mGy wasdelivered from a Sr90 source. The dose rate of the source, 28 mGy/sec, means that thesix second exposure time obviously would have some timing error associated with it.This was considered acceptable given the aims of the experiment.164RESULTSDisc 1 Natural signal Signal after Stored dose168 mGy dose mGyPH 186939 156273 201PIll 856712 760407 189Disc 2PH 42166 39939 177PIll 263540 254144 174TABLE 6.3 STORED DOSE EVALUATIONS(Disc 1 is thought to have a larger timing error associated with its beta dose than disc2).6.1 CONCLUSIONSIt is evident from the above that (a) this matrix was receiving the same dose rate as hadbeen predicted for the first portion examined in Chapter 4 (b) there was no evidence ofradon crosstalk between the second pair of matrices and (c) after a 4.25 year exposureperiod the results from these 100 micron grains are within experimental error of thepredicted values.Although, at this stage, the reasons for the anomalous results obtained from first portionof this matrix remain unclear, it would appear that the original experimental design andmethod were valid. It is expected that a more detailed analysis of the residual materialwould resolve the problem, and the isochron approach does indeed deserve furtherinvestigation.165REFERENCES166Aitken, M. J. (1968) Thennoluminescent Dating in Archaeology. Introductory Review.In: Thermoluminescence of Geological Materials. Ed. D. J. McDougall, pp. 369-387,Academic Press, New York.Aitken, M. J. (1969) Thermoluminescent dosimetry of environmental radiation onarchaeological sites. Archaeometry 11, 109-114.Aitken, M. J. (1974) Physics in Archaeology. Clarendon Press, Oxford.Aitken, M. J. (1978a) Dose-rate evaluation. PACT 2, 18-33.Aitken, M. J. (1978b) Radon loss evaluation by alpha counting. PACT 2, 104-117.Aitken, M. J. (1979) Pre-dose dating: predictions from the model. PACT 3, 319-324.Aitken, M. J. (1985a) Alpha particle effectiveness: numerical relationship betweensystems. Ancient TL 3 (3), 22-25.Aitken, M. J. (1985b) Thermoluminescence dating. Academic Press, London.Aitken, M. J. (1990) Science-based dating in archaeology. Longman, London.Aitken, M. J., Fleming, S. J., Reid, J. and Tite, M. S. (1968) Elimination of SpuriousThermoluminescence. In: Thermoluminescence of Geological Materials. Ed. D. J.McDougall, pp. 133-142. Academic Press, New York.Aitken, M. J. and Murray, A. S. (1976) The pre-dose technique: radiation quenching.The Edinburgh Symposium on Archaeometry and Archaeological Prospection, ExtendedAbstracts, HMSO.Aitken, M. J., Tite, M. S. and Reid, J. (1963) Thermoluminescent Dating: ProgressReport. Archaeometry 6, 65-75.167Aitken, M. J., Zimmerman, D. W. and Fleming, S. J. (1968) Thennoluminescent Datingof Ancient Pottery. Nature 219,442-445.Aitken, M. 1., et al. (1985) Beta and gamma gradients. Nucl. Tracks, Vol. 10 Nos 4-6,647-653.Bailiff, I. K. (1976) Use of phototransfer for the anomalous fading ofthermoluminescence. Nature 264, 531-533.Bailiff, I. K. (1982) Beta-TLD apparatus for small samples. PACT 6, 72-75.Bailiff, I. K. (1983a) Sensitization of R-traps? PACT 9, 207-214.Bailiff, I.K. (1983b) Pre-dose dating of Iron-Age pottery from northern Britain. PACT9, 219-225.Bailiff, I. (1988) Principles of TL dating, in course notes, from SERe short course onthe Application of TL dating, Durham University, 4-8 January, 1988.Bailiff, I. K. and Aitken, M. 1. (1980) Use of thermoluminescence dosimetry forevaluation of internal beta dose-rate in archaeological dating. Nucl. Instrum. Methods173, 423-429.Bell, W. T. (1976) The assessment of the radiation dose-rate for thermoluminescencedating. Archaeometry 18, 107-111.Bell, W. T. (1976a) Thermoluminescence dating dose rate data. Archaeometry 21, 243-245.Bell, W. T. (1976b) Attenuation factors for the absorbed radiation dose in quartz grainsfor thermoluminescence dating. Ancient TL 8, 2-13.168Bell, W. T. (1977) Thermoluminescence dating: revised dose rate data. Archaeometry19, 99-100.Bell, W. T. (1980) Alpha dose attenuation in quartz grains for thermoluminescencedating. Ancient TL 12, 4-8.Berger, M. J. (1971) Distribution of absorbed dose around point sources of electronsand beta particles in water and other media. Journal of Nuclear Medicine (Supp.) 5, 5-23.Berger, M. J. (1973) Improved point kernels for electron and beta ray dosimetry NBSIR73-107. National Bureau of Standards, Washington, DC.Bintliff, 1. L. (ed) (1984) European Social Evolution. Archaeological Perspectives.Pub. Uni. of Bradford. Chanctonbury Press Ltd.Botter-Jensen, L., Bundegaard, J. and Mejdahl, V. (1983) An HP-85 microcomputercontrolled automated reader system for TL dating. PACT 9, 343-349.Bowman, S. G. E. (1979) Phototransferred thermoluminescence in quartz and itspotential use in dating. PACT 3, 381-400.Bowman, S. G. E. and Huntley, D. 1. (1984) A new proposal for the expression ofalpha efficiency in TL dating. Ancient TL 2, 6-11.Boyle, R. (1664) Experiments and considerations upon colours with observation on adiamond that shines in the dark. Henry Herringham, London.Charalambous, S. and Michael, C. (1976) A new method of dating pottery bythermoluminescence. Nucl. Instrum. Methods 137, 565-567.169Clark, P. A. and Templer, R. H. (1988) Thermoluminescence dating of materials whichexhibit anomalous fading. Archaeometry 30 (1), 19-36.Cross, W. G. (1968) Variation of beta dose attenuation in different media. Physics inMedicine and Biology 13, 611-618.Daniels, F., Boyd, C. A. and Saunders, D. F. (1953) Thermoluminescence as a researchtooL Science 117, 343-349.East, B. W. (1991) Environmental behaviour of radionuclides in an urban environment.DOE Report No: DOEIHMIP/RRI911026.Faires, R. A. and Parks, B. H. (1973) Radioisotope laboratory techniques.Butterworths, London.Fieschi, R. and Scaramelli, P. (1968) Photostimulated thermoluminescence in alkalihalid crystals. In: Thermoluminescence of geological materials. Ed. D. J. McDougall,291-308.Filberth, E. W., Rowe, M. W. and Shaefer, H. 1. (1980) Uranium in Hueco andGuadalupe Mountain Indian Ceramics. Archaeometry 22 (1), 107-110.Fleming, S. J. (1966) Study of Thermoluminescence of crystalline extracts from pottery.Archaeometry 9, 170-173.Fleming, S. J. (1970) Thermoluminescence dating: refinement of the quartz inclusionmethod. Archaeometry 12, 135-146.Fleming, S. J. (1973) The pre-dose technique: a new thermoluminescence datingmethod. Archaeometry 15 13-30.170Fleming, S. J. (1979) Thermoluminescence techniques in archaeology. Clarendon Press,Oxford.Fleming, S. 1. and Thompson, J. (1970) Quartz as a heat resistant dosimeter. HealthPhysics 18, 567-568.Fleming, S. 1. and Stoneham, D. (1973) The subtraction technique ofthermoluminescence dating. Archaeometry 15 (2), 229-238.Fremlin,1. H. (1968) Effects of Non-uniformity of material on the thermo luminescentmethod of dating. In: Thermoluminescence of Geological Materials. Ed. D. J.McDougall, pp. 419-425. Academic Press, New York.Fremlin, 1. H. and Srirath, S. (1964) Thermoluminescent dating: examples of non-uniformity of luminescence. Archaeometry 7, 58-62.Goedicke, C. (1985) TL dating: a novel form of differential dating. Nuclear Tracksand Radiation Measurements 10, 811-816.Greening, 1. R. (1985) Fundamentals of radiation dosimetry.Greison, K. (1942) The intensities of the hard and soft components of cosmic rays asfunctions of altitude and zenith angle. Physics Review 61, 212.Guerin, G. and Valladas, G. (1980) Thermoluminescence dating of volcanicplagioclases. Nature 286, 697-699.Horowitz, Y. S. (1984) Thermoluminescence and Thermoluminescent Dosimetry. CRCPress, Florida.Huntley, D. J., Godfrey-Smith, D. 1. and Thewalt, M. L. W. (1985) Optical dating ofsediments. Nature 313, 105-107.171Hutt, G., Smimov, A. and Tale, I. (1979) On the application of thermoluminescence ofnatural quartz to the study of geochronology of sedimentary deposits. PACT 3Katz, L. and Penfold, A. S. (1952) Range-energy relations for electrons and thedetermination of beta end-point energies by absorption. Reviews of Modem Physics Vol.24 (1), 28-44.Kennedy, G. C. and Knopff, L. (1960) Dating by thermoluminescence. Archaeology13, 147-148.Grogler, N, Houtermans, F. G. and Stauffer, H. (1960) Uber die Datierung von Keramikund Ziegel durch Thermolumineszenz. Helvetica Physica Acta 33, 595-596.Langouet, L., Roman, A. and Gonzales, R. (1980) Datation de poteries anciennes parla methode D.A.T.E., 1976. Symp. on Archaeometry and Archaeological Prospection,National Museums of Antiquities of Scotland, Edinburgh, 312-320.Lewis, D. R. (1968) Effect of grinding on thermoluminescence of dolomite, calcite andhalite. In: Thermoluminescence of Geological Materials. Ed. D. J. Dougall, pp 125-132. Academic Press, New York.Li, J-L., Pei, J-X., Wang, Z-Z. and Lu, vc (1977) Kexue Tongbao (China) 22, 498-502.Lovborg, L. and Kirkegaard, P. (1974) Response of 3" x 3" NaI(Tl) detectors toterrestrial gamma radiation. Nuclear Instruments and Methods 121, 239-251.Malik, S. R., Durrani, S. A. and Fremlin, J. J. (1973) A comparative study of the spatialdistribution of uranium and of TL-producing minerals in archaeological materials.Archaeometry 15 (2), 249-253.172Mangini, A., Pemicka, E. and Wagner, G. A. (1983) Dose-rate determination byradiochemical analysis. PACT 9 (1), 49-56.McKeever, S. W. S. (1985) Thernwluminescence of solids. Cambridge University Press.McKerrell, H. and Mejdahl, V. (1981) Progress and problems with automated TLdating. Proc. 16th Int. Symp. Archaeometry and Archaeological Prospection, Riso-M-2265, Riso National Laboratory, Roskilde.McManus, J. (1988) Grain size determination and interpretation. In: Techniques inSedimentology. Ed. M. Tucker, Blackwell Scientific Publications.Meakins, R. L., Dickson, B. L. and Kelly, J. C. l1919) Gamma la~ O1\o.\~~\~Gl K., Uand Th for dose rate estimation in thermoluminescent dating, .A'rch ae orne rry 21 (1), 79-86.Mejdahl, V. (1972) Progress in TL dating at Riso. Archaeometry 14 (2), 245-256.Mejdahl, V. (1979) Thermoluminescence dating: beta dose attenuation in quartz grains.Archaeometry 21 (1), 61-72.Mejdahl, V. (1983) Feldspar inclusion dating of ceramics and burnt stones. PACT 9,351-366.Mejdahl, V. and Wintle, A. G. (1984) Thermoluminescence applied to agedetermination in archaeology and geology. In: Thermoluminescence andThermoluminescent Dosimetry, Volume Ill. Ed. Y. S. Horowitz, c.R.C. Press.Mejdahl, V. and Winther-Nielsen, M. (1982) TL dating based on feldspar inclusions.PACT 6, 426-437.Michels, J. w. (1973) Dating methods in Archaeology. Seminar Press, New York.173Miller, 1. A. (1970) Dating by the potassium - argon method- some advances intechnique. In Science and Archaeology, edited by D. Brothwell and E. Higgs, Thamesand Hudson.Moorehead, F. F. and Daniels, F. (1957) 1. Chem. Phys. 27, 1318-1324.Morozov, G. V. (1986) The Dating of Quaternary Ukranian Sediments byThermoluminescence XllIth INQUA Congr. Kiev.Murray, A. S. and Aitken, M. 1. (1982) The measurement and importance of radioactivedisequilibria in TL samples. PACT 6, 155-169.Murray, A. S. and Aitken, M. 1. (1988) Analysis of low-level natural radioactivity insmall mineral samples for use in thermoluminescence dating using high resolutiongamma spectroscopy. Applied Radioactive Isotopes Vol 39 (2), 145-158.Nambi, K. S. V. and Aitken, M. 1. (1986) Annual dose conversion factors for TL andESR dating. Archaeometry 28 (2), 202-205.Plachy, A. L. and Sutton, S. R. (1982) Determination of the dose rate to quartz ingranite. PACT 6, 188-194.Poolton, N. 1. and Bailiff, I. K. (1989) The use of LED's as an excitation source forphotoluminescence dating of sediments. Ancient TL Vol 7 (1), 18-20.Prescott, 1. R. and Stephan, L. G. (1982) Contribution of Cosmic radiation toenvironmental dose. PACT 6, 17-25.Randall, 1. T., and Wilkins, M. H. F. (1945) Phosphorescence and electron traps.Proceedings of the Royal Society of London, A. 184, 366-407.Rendell, H. M., (1985) The precision of water content estimates in thermoluminescence174dating of loess from Northern Pakistan. Nucl. Tracks. Vol. 10, Nos 4-6,763-768.Sanderson, D. C. w. (1987) Thermoluminescence dating of Scottish vitrified forts.Unpublished PhD thesis, Department of Physics, Paisley College of Technology.Sanderson, D. C. W. (1988) Thick source beta counting (TSBC): a rapid method formeasuring beta dose rates. Nuclear Tracks and Radiation Measurements. Vol 14 Nos1/2, 203-207.Sanderson, D. C. W. and Clark, P. A. (1987) Recent progress in TL dating of Scottishceramics and burned stones. (Abstract) Science and Archaeology, Glasgow 1987. BAR196 (ii), 623.Sanderson, D. C. W., Placido, F. and Tate, 1. O. (1988) Scottish Vitrified Forts: TLResults from Six Study Sites. Nuclear Tracks and Radiation Measurements Vol 14Nos. 1/2, 307-316.Sanderson, D. C. W., et al. (1989) in Long and short range limits in luminescencedating. Research Laboratory for Archaeology, Oxford, Occasional publication no. 9.Schlesinger, M. (1965) J. Phys. Chem. Solids 26, 761.Shelkoplyas, V. N. (1971) In: Chronology of the Glacial Age, 155-6a.Sutton, S. R. and Zimmerman, D. W. (1976) Thermoluminescent dating using zircongrains from archaeological ceramics. Archaeometry 18 (2), 125-134.Sutton, S. R. and Zimmerman, D. W. (1979) The zircon natural method: initial resultsand low level TL measurement. PACT 3, 465.Srirath, S., and Fremlin, 1. H. (1968) Non-uniformity in the thermoluminescence of175"Pot-boilers". In: Thermoluminescence of Geological Materials. Ed. D. J. Dougall, pp.427-429. Academic Press, New York.Templer, R. H" (1985) The dating of zircons by auto-regenerated TL at lowtemperatures. Nucl. Tracks. Vol. 10, Nos 4-6, 789-798.Templer, R. H. and Walton, A. (1985) Zoning in zircons. Nuclear Tracks andRadiation Measurements 10 Nos 4-6, 683-692.Tickel, F. G. (1965) Developments in Sedimentology 4: The Techniquess ofSedimentary Mineralogy. Elsevier Publishing Company.Tite, M. S. (1966) Thermoluminescent dating of ancient ceramics: a reassessment.Archaeometry 9, 155-169.Tite, M. S. (1968) Some complicating factors in thermoluminescent dating and theirimplications. In: Thermoluminescence of Geological Materials. Ed. D. J. McDougall,pp. 389-405. Academic Press, New York.Tite, M. S. and Waine, 1. (1962) Thermoluminescent dating: a re-appraisal.Archaeometry 5, 53-79.Turner, R. C; Radley, J. M. and Mayneord, W. V. (1958) The alpha-ray activity ofhuman tissues. British Journal of Radiology 31, 397-402.Tyler, A. N., (1992) personal communication.Urbach, F. (1930) sitzber Akad Wiss Wein Math naturn Kl, Abt. Ha 139, 363.Valladas, H. and Valladas, G. (1982) Effet de l'irradiation et sur des grains de quartz.PACT 6, 171-178.Ward, S. (1988) An investigation of thermoluminescence of glass discs of a potential176high dose dosimeter. B.Sc Honours Dissertation. SURRC/ Department of Physics,Strathclyde University.Warren, S. E. (1978) Thermoluminescence dating pottery: an assessment of the doserate from rubidium. Archaeometry 20 (1), 71-72.Wintle, A. G. (1973) Anomalous fading of thermoluminescence in mineral samples.Nature 245, 143-144.Wintle, A. G. (1980) Thermoluminescence dating: a review of recent applications tonon-pottery materials. Archaeometry 22 (2), 113-122.Wintle, A. G. and Huntley, D. J. (1979) Thermoluminescence dating of a deep seasediment core. Nature 279, 710-712.Zimmerman, D. W. (1967) Thermoluminescence from fine grains from ancient pottery.Archaeometry 10, 26-28.Zimmerman, D. W. (1971) Thermoluminescent dating using fine grains from pottery.Archaeometry 13, 29-52.Zimmerman, D. W. (1971) Uranium distributions in archaeological ceramics: dating ofradioactive inclusions. Science 174, 818.Zimmerman, D. W. (1972) Relative thermoluminescence effects of alpha- and beta-radiation. Radiation Effects Vol 14, 81-92.177

Isochron methods for luminescence dating in archaeology

Abstract

Extracted data

Related items

Related items