2008

Luminescence DatingGuidelines on using luminescence dating in archaeology

2

Contents

Part A Introduction to luminescence . . . . . . . . . . . . . . . . . . . . . . . . . .4

1 An overview of luminescence dating . . . . . . . . . . . . . . . .41.1 Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .51.2 Common units for measuring length and time . . . . . . . . . .6

2 The physical basis of luminescence . . . . . . . . . . . . . . . . .62.1 Thermoluminescence (TL) . . . . . . . . . . . . . . . . . . . . . . . . . . .62.2 Optically stimulated luminescence (OSL) . . . . . . . . . . . . . . .62.3 Resetting of TL and OSL signals . . . . . . . . . . . . . . . . . . . . . . .72.4 Luminescence emission spectra . . . . . . . . . . . . . . . . . . . . . . .72.5 Anomalous fading in feldspars . . . . . . . . . . . . . . . . . . . . . . . .82.6 Other luminescence signals . . . . . . . . . . . . . . . . . . . . . . . . . .9

3 Measurement of De . . . . . . . . . . . . . . . . . . . . . . . . . . . . .93.1 SAR protocol . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .113.1.1 Recycling test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .113.1.2 Preheat test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .123.1.3 Dose recovery test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .123.2 Replicate measurements of De . . . . . . . . . . . . . . . . . . . . . .133.3 Aliquot size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .143.3.1 Single-grain OSL measurements . . . . . . . . . . . . . . . . . . . . .143.3.2 Displaying De datasets for samples . . . . . . . . . . . . . . . . . . .153.4 TL and the plateau test . . . . . . . . . . . . . . . . . . . . . . . . . . . .15

4 Measurement of dose rate . . . . . . . . . . . . . . . . . . . . . .164.1 Chemical methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .164.2 Emission counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .174.3 In situ measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .194.4 The impact of water content . . . . . . . . . . . . . . . . . . . . . . . .194.5 Cosmic ray contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . .19

5 Limits of luminescence dating . . . . . . . . . . . . . . . . . . . . .195.1 Suitability of material . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .195.2 Resetting of the trapped electron population . . . . . . . . . .195.3 Upper age limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .205.4 Lower age limits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .215.5 Accuracy and precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . .21

Part B Practicalities

6 Project management under MoRPHE . . . . . . . . . . . . . .22

7 Sampling considerations . . . . . . . . . . . . . . . . . . . . . . . . .227.1 Sampling strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .227.2 Heated samples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .237.3 Sediments (unheated samples) . . . . . . . . . . . . . . . . . . . . . .237.4 Health and safety . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .25

8 Laboratory considerations . . . . . . . . . . . . . . . . . . . . . . .25

9 Reporting specifications . . . . . . . . . . . . . . . . . . . . . . . . .259.1 Sample details . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .259.2 Luminescence measurements . . . . . . . . . . . . . . . . . . . . . . .259.3 Dose rate measurements . . . . . . . . . . . . . . . . . . . . . . . . . . .269.4 Age calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .269.5 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .269.6 Indexing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .26

9.7 Non-technical summary . . . . . . . . . . . . . . . . . . . . . . . . . . . .26

10 Quoting ages and disseminating results . . . . . . . . . . . . .26

Part C Case studies

11 Wat’s Dyke, Gobowen . . . . . . . . . . . . . . . . . . . . . . . . . .28

12 Dungeness Foreland, Sussex . . . . . . . . . . . . . . . . . . . . . .30

13 Fluvial gravels at Broom, Devon . . . . . . . . . . . . . . . . . . .33

14 Dating medieval bricks . . . . . . . . . . . . . . . . . . . . . . . . . .35

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38

Further reading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .38

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .39

Glossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .40

Appendix 1: Sources of advice on Scientific Dating from English Heritage . . . . . . . . . . . . . . . . . .42

Appendix 2: Luminescence laboratory contact details . . . . . . . . . . . . . . . . . . . . . . . . . . . .42

Authorship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44

3

PrefaceThese guidelines are designed to establishgood practice in the use of luminescencedating for providing chronological frameworks.They provide practical advice on usingluminescence dating methods in archaeology.The guidelines should not be regarded as asubstitute for advice given by specialists onspecific projects; and, given how rapidly themethods have developed, it is likely thatfurther improvements in laboratory techniqueswill occur, so more up to date advice is likelyto become available in the future.

The guidelines will help archaeologists and siteinvestigators to assess whether luminescencedating will be of value in providingchronological information for understandingtheir site.They are divided into three mainparts: Part A, an introduction to luminescencedating, including the principles underlying themethod and the measurement proceduresused; Part B, a section on the practicalities ofcollecting samples, collaborating with aluminescence laboratory, understanding theresults obtained and presenting luminescenceages; and Part C, a series of case studies toillustrate the use of the method.

The first section is designed to enable thenon-specialist to understand the physicalprinciples underlying luminescence dating sothat he or she is more fully able to understandthe issues that may affect the reliability ofluminescence ages, and critically assess theresults from luminescence laboratories.

What these guidelines cover

� an introduction to luminescence dating� a summary of the variety of luminescence

methods available� a description of the options available for

estimating the radiation dose rate at a site� practical advice about the collection of

samples� a summary of the information that should

be provided by laboratories undertakingluminescence measurements, and about howluminescence ages should be quoted

� case studies illustrating different applicationsof luminescence dating, and the results thatcan be obtained

How to use these guidelinesLuminescence dating is a technical topicinvolving consideration of a number ofcomplex scientific issues.These guidelines havepresented these topics as clearly as possible,but inevitably there may be some terms andabbreviations that are not familiar to allreaders. A list of abbreviations is given inSection 1.1 and a glossary of terms is provided

at the end of these guidelines to help thereader, and text in bold indicates that the termis defined in the glossary. Under FurtherReading is a range of other texts to augmentthe information provided here.

In these guidelines, section 1 provides asummary of the method and section 5 givessome indications of what samples are suitableand what age limits are appropriate.Theremainder of Part A contains a more detailedexplanation of the techniques involved in thevarious measurements that are required toobtain a luminescence age. A minimum readingof Part A would be sections 1 and 5.Thedetailed information provided in sections 2, 3and 4 is necessary to enable users ofluminescence dating to be able to assess whatis feasible when using it, and to interpretcritically the results that are obtained from aluminescence laboratory.

Part B includes a range of practicalconsiderations, including sample selection andcollection. It defines the information thatshould be required in reports obtained fromluminescence laboratories, explains howluminescence ages should be turned into datesand how they should be quoted in site reportsor other publications. Some examples of howthe method can be applied are given in PartC. An Executive Summary concludes theguidelines.

Part A Introduction to luminescenceLuminescence dating is a chronological methodthat has been used extensively in archaeologyand the earth sciences. It is based on theemission of light, luminescence, by commonlyoccurring minerals, principally quartz.Themethod can be applied to a wide range ofmaterials that contain quartz or similar minerals.For pottery, burnt flints and burnt stones, theevent being dated is the last heating of theobjects.Another, and now very common,application is to date sediments, and in this casethe event being dated is the last exposure of themineral grains to daylight (Fig 1).The age rangeover which the method can be applied is from acentury or less to over one hundred thousandyears.

Some of the first applications of luminescencedating were developed in the 1960s. Since thattime there has been enormous progress inunderstanding of the luminescence phenomenain natural minerals, of the methods used tomeasure that luminescence and in the range ofmaterials that are analysed.To the non-specialist,luminescence can be complex and the results

difficult to interpret.This document is designedto help users employ the available techniqueseffectively on their projects, and to interpret theresults they can obtain from luminescencelaboratories.

1 An overview of luminescencedatingRadioactivity is ubiquitous in the naturalenvironment. Luminescence dating exploits thepresence of radioactive isotopes of elementssuch as uranium (U), thorium (Th) andpotassium (K). Naturally occurring minerals suchas quartz and feldspars act as dosimeters,recording the amount of radiation to which theyhave been exposed.

A common property of some naturallyoccurring minerals is that when they areexposed to emissions released by radioactivedecay, they are able to store within their crystalstructure a small proportion of the energydelivered by the radiation.This energyaccumulates as exposure to radioactive decaycontinues through time.At some later date thisenergy may be released, and in some mineralsthis energy is released in the form of light.This

light is termed luminescence.

What makes this a useful phenomenon fordating? The answer lies in the fact that thisenergy stored in minerals can be reset by twoprocesses.The first is by heating the sample totemperatures above about 300°C, as wouldoccur in a hearth, or in a kiln during firing ofpottery.The second process is exposure of theminerals to daylight, as may occur duringerosion, transport and deposition of sediments.Either of these processes will release any pre-existing energy stored, and thus set the ‘clock’ tozero.Thus in luminescence dating, the eventbeing dated is this resetting, either by heat or byexposure to light.

Measurements of the brightness of theluminescence signal can be used to calculate thetotal amount of radiation to which the samplewas exposed during the period of burial. If this isdivided by the amount of radiation that thesample receives from its surroundings each yearthen this will give the duration of time that thesample has been receiving energy.

total energy accumulated during burial

energy delivered each year from radioactive decay

The SI (Système International) unit of absorbedradiation is the Gray (Gy). It is a measure of theamount of energy absorbed by a sample, or itsdose, and has the units joules per kilogramme(J.kg-1). Laboratory luminescence measurementsare used to calculate the total absorbed dose.The name given to the quantity is the equivalentdose (De).The amount of energy absorbed peryear from radiation in the environmentsurrounding the measured material (known asthe dose rate) can either be derived by directlymeasuring the amount of radioactivity, or bychemically analysing the surrounding materialand calculating the concentration of radioactiveisotopes in it; this has the units of Gray per year.Thus the age equation for luminescence canformally be expressed as:

equivalent dose (De) (Gy)

dose rate (Gy/year)

To understand how this process can be used fordating, a useful analogy is that of a rechargeablebattery, in which the battery represents themineral grains (Fig 2). Exposing mineral grains tolight or heat will release the battery’s energy, sothat when the mineral (battery) is incorporatedinto sediment or fired in a brick or sherd ofpottery, it has no energy.At this time the batterystarts to be recharged by exposure to radiationin its natural environment.As time progresses,the stored energy increases.When the sample iscollected and measured in the laboratory thisreleases the energy and light is created.This is

4

age =

age (years) =

Fig 1 Excavation at Gwithian, Cornwall in 2005, showing the collection of samples of sands for luminescence dating(© Historic Environment Service, Cornwall County Council 2005).

Fig 2 A rechargeable battery provides a useful analogy for how mineral grains behave and how luminescence dating works.Whenmineral grains are exposed to light or heated, trapped electrons are released, much like depleting a battery. Once this heating orexposure to daylight stops, re-exposure to natural radioactivity begins to recharge the mineral grains, much like recharging abattery. In the laboratory the mineral grains are stimulated to release their stored energy in the form of light emission.Thebrightness of this luminescence signal can be related to the amount of energy that has been stored in the mineral.

the luminescence signal that is observed.Thebrightness of this luminescence signal is relatedto the amount of energy in the battery. If wealso know the rate at which the battery wasrecharged then we can work out how long itmust have been recharging, thus telling us theperiod of time that has elapsed since the batterywas last emptied.

Many naturally occurring minerals will yieldluminescence signals, including quartz, feldspars,calcite and zircons. Quartz is the most suitablematerial for dating, and the single aliquotregenerative dose (SAR) protocol is now usedroutinely.The luminescence age is the period oftime that has elapsed since the sample washeated (in the case of pottery, burnt flints orbricks) or exposed to daylight (in the case ofsediments).The age is given as the number ofyears before the date of measurement, and theunit of time used is the annum (abbreviated ‘a’).There is no convention for the datum to whichluminescence ages are referred, so the date ofmeasurement must be given.The term BP(before present) should never be used forluminescence ages, as BP has a specific meaningand is only relevant for radiocarbon ages.

In summary, two sets of measurements need tobe combined to calculate a luminescence age(Fig 3), De , which is determined usingluminescence, and the dose rate.

1.1 Abbreviationsa annum (one year)AAS atomic absorption

spectrophotometry, a technique used to analyse the chemical composition of materials

C CarbonDe equivalent dose – see GlossaryeV electron-volt, a measure of

energy: 1eV = 1.602 x 10-19 jouleGy Gray – see GlossaryICP-MS inductively-coupled plasma mass

spectrometry, a technique used to analyse the chemical composition of materials

IRSL infrared stimulated luminescence – see Glossary

K Potassiumka kiloannum, a time period of

1,000 yearskeV kiloelectron volts: 1keV = 1,000eV

(see eV)LED light emitting diodeMa megaannum, a time period of

1,000,000 yearsMeV megaelectron volt: 1MeV =

1,000,000eV (see eV)NAA neutron activation analysis, a

technique used to analyse the chemical composition of materials

nm nanometre (1,000,000nm = 1mm)

OSL optically stimulated luminescenceP palaeodose, another term used

for equivalent dose (De)ppm parts per million (10,000ppm =

1%)Ra RadiumRn RadonSAR single aliquot regenerative dose, a

sequence of luminescence measurements used to estimate the equivalent dose (De) of a sample

SI Système International, an internationally recognised set of units used to define length, time,energy and other units of measure

Th ThoriumTL thermoluminescence

– see GlossaryU UraniumXRF X-ray fluorescence, a technique

used to analyse the chemical composition of materials

μ m micrometre (1,000μm = 1mm)

5

Collect sample forluminescence dating

Laboratory treatment ofthe sample to isolate

material for luminescencemeasurements: mostcommonly quartz

Calculate dose rateappropriate for the materialused for luminescencemeasurements taking siteconditions into account

Make luminescencemeasurements to calculatethe equivalent dose (De)

Calculate luminescence age

Age (years) = Equivalent Dose (De)Dose Rate

Appropriate tomake in situmeasurementsof the dose rateat time ofcollection?

YesUse portion ofluminescencesample for

laboratory dose ratemeasurements

NoField

activities

Laboratoryactivities

Fig 3 Flow chart showing how procedures to measure De (left) and the dose rate (right) are combined to calculate a luminescence age.

1.2 Common units for measuring lengthand timeMeasurement of length is based on the SI unitof the metre.

Abbreviation length of one metre

metre m 1mcentimetre cm 100cmmillimetre mm 1,000mm micrometre μ 1,000,000μm nanometre nm 1,000,000,000nm

The SI unit of time is the second(s), but this isnot used in common parlance for long periodsof time (for example an average year is31,557,600,000 seconds).Thus forarchaeological and geological purposes, theyear is commonly used.

Abbreviation duration of one year

year a 1athousand years ka 0.001ka = 1amillion years Ma 0.000 001Ma

= 1a

2 The physical basis ofluminescenceThe analogy of a rechargeable battery shownin Fig 2 is a close one: De is a measure of theenergy absorbed by the mineral grains and thedose rate is the rate at which the energy wasdelivered. In fact, at a sub-atomic level, theenergy is stored by electrons becomingtrapped at sites within the crystal structure –where they are normally forbidden to reside –but where they can be stored because ofdefects within the structure. An energy level

diagram illustrates the processes going onwithin the crystal (Fig 4).The interaction ofradiation with the crystal provides energy toelectrons that are raised to the conductionband. From here they can become trapped atdefects (trapping centres) within the crystal.The electrons may be stored at these defectsfor some time.When an electron is released, itloses the energy that it gained duringirradiation, and may emit part of that energy inthe form of a single photon of light. In general,the deeper the defect below the conductionband, the longer the electrons remain trappedat that location without escaping at typicalburial temperatures.

2.1 Thermoluminescence (TL)The trapped electrons stored within mineralscan be released in the laboratory using anumber of methods that cause them toproduce a luminescence signal. Heating thesample at a fixed rate from room temperatureto between 450°C and 700°C releases thetrapped electrons.The resulting signal istermed thermoluminescence (TL), and asshown in Fig 5 the signal is plotted as afunction of the measurement temperature.Typically the TL signal (commonly referred toas a glow curve) comprises a series of peaks.Each peak may be due to a single type of trapwithin the mineral being measured, but morecommonly the signal is a composite of severaltraps. Although it is not always possibleuniquely to identify the source of theelectrons, it is normally true that the TL signalobserved at higher measurementtemperatures originates from traps that aredeeper below the conduction band.This is thecase because more energy is required to

release electrons from deeper traps, and sothis only occurs at higher measurementtemperatures. A corollary of this is thatelectrons in deeper traps are more stable thanthose in shallower traps. For example, Fig 5shows a TL measurement on an aliquot (asmall sub-sample [c 1 to 5mg in mass] ofquartz).The electrons in the trap giving rise tothe TL peak observed at c 110°C are relativelyunstable, and at room temperature have amean lifetime of c 20 hours.This short lifetimemakes them useless for dating. In contrast,measurements suggest that the electrons givingrise to the TL peak observed at c 325°C arestable over many millions of years.These twopeaks could be thought of as derived fromtraps T1 and T2 , respectively, in Fig 4.Thestability of the deeper trap has been confirmedby dating of sediments up to almost a millionyears old, and it is this trap that is the focus ofmost methods used for dating with quartz.

2.2 Optically stimulated luminescence(OSL)A second means of releasing the electronsstored within minerals is by exposing them tolight (Huntley et al 1985).This has become themost commonly used method. As soon as thestimulating light is switched on luminescence isemitted by the mineral grains. As measurementcontinues, the electrons in the traps areemptied and the signal decreases.The signal istermed optically stimulated luminescence(OSL) and Fig 6 shows data from an aliquot ofquartz during optical stimulation.The signalinitially decreases rapidly, and then at a slowerrate. A similar signal is observed fromfeldspars, although a general observation isthat the OSL signal from feldspars decreasesmore slowly than that from quartz. Unlike TL,the OSL data typically obtained (eg Fig 6) doesnot show which traps emitted the electrons.Thus, before making an OSL measurement, it isimportant to thermally pretreat the aliquot sothat a signal is obtained from the group oftraps of interest.This is achieved by heating thealiquot before measurement so that theshallow traps, whose electrons are unstableover the burial period (eg T1 in Fig 4), areemptied, leaving only the electrons in deeper,stable traps (eg T2 in Fig 4) – this heating iscalled a preheat, described in section 3.1.2.

The light used to stimulate the minerals isrestricted to a narrow range of wavelengths sothat this light can be prevented from reachingthe sensitive light detector (the photomultipliertube) used to measure the luminescence signal(Fig 7 page 8).This is because the light givenoff by the sample – the luminescence beingmeasured – has to be observed at a differentwavelength from the stimulation light. Blue lightemitting diodes (LEDs) are commonly used for

6

Fig 4 Energy level diagram illustrating the luminescence process: (i) radiation interacts with the crystal (ionization), pushingelectrons into the conduction band and leaving ‘holes’ in the valence band; (ii) electrons become trapped at defect sites (T1,T2

etc) and remain for a period of time (the deeper the trap below the conduction band (E) (eg T2) the more stable the electronand the longer it stays trapped); (iii) crystal is stimulated by heat or exposure to light, releasing electrons, which recombine withholes at luminescence centres (L) and emit light photons = the luminescence signal (modified from Aitken 1990).

7

stimulation and will generate an OSL signalfrom both quartz and feldspar.

An alternative method of stimulation is to useLEDs that emit beyond the visible part of thespectrum in infrared.These are the type ofLEDs used for remote controls in televisions,and for luminescence work those centred at880nm have been used extensively. OSL signalsare produced, but are more commonlyreferred to as infrared stimulatedluminescence (IRSL). IRSL is only observedfrom feldspars, with quartz not giving an IRSLsignal when the sample is measured at roomtemperature.The fact that quartz does notemit an IRSL signal at room temperature canbe exploited to provide a method for assessingthe purity of quartz separated from feldsparfor luminescence measurements.

2.3 Resetting of TL and OSL signalsThe major advantage of OSL over TL is thatthe OSL signal is reset by exposure to sunlightmuch more rapidly than is the TL signal.Thisprocess is commonly known as bleaching andexperiments in the laboratory show that afteras little as 100s exposure of mineral grains tosunlight, the OSL signal from quartz is reducedto < 0.1% of its initial level (Fig 8 Page 8), but> 85% of the TL signal still remains. Afterseveral hours of exposure, > 30% of the TLsignal remains but the OSL signal is almost onehundred thousands times lower. A similarsituation exists for feldspars, with the TL signalbeing reduced by exposure to daylight muchmore slowly than the OSL (or IRSL) signal.The results shown in Fig 8 were obtained byplacing a single layer of grains on a cleansurface in the laboratory and exposing themto sunlight. In nature, this bleaching process islikely to be more complex, as mixtures of

mineral grains of different sizes andcompositions are moved through the surfaceenvironment, and because the strength ofdaylight varies with the time of day and withthe local conditions. In fact, as grains aremoved around before their final deposition,they may be exposed to daylight more thanonce. In addition, grains may have a surfacecoating that reduces light penetration and thuscauses slower bleaching. One of the majoradvances in luminescence dating in recent yearshas been the ability to make replicatemeasurements on sub-samples of the samematerial, and this provides a means forassessing whether the mineral grains wereexposed to sufficient daylight for the OSL signalto be reset at deposition (see section 3.2).

2.4 Luminescence emission spectraThe luminescence emitted by minerals mayoccur at a variety of wavelengths.The lower

Fig 5 (left) Typical thermoluminescence (TL) signal from an aliquot of quartz as it is heated from room temperature to 450°C at a rate of 5°C per second.This sample had been exposed to alaboratory radiation source shortly before measurement, which induced the peak at about 110°C – this peak is only stable for a matter of hours at room temperature, so is not seen whennatural samples are measured.The broad peak at c 325°C and the higher temperature shoulder are stable over many millions of years, and are used for dating. (right) Quartz grains (c 0.2mmdiam) mounted on aluminium sample holders (9.8mm diam), ready for luminescence measurement (© Risø National Laboratory, Denmark).

Fig 6 (left) Typical OSL signal observed from a quartz aliquot – the OSL signal drops rapidly as the trapped electrons in the quartz are used up to produce the luminescence signal. (right) Anautomated luminescence reader equipped with blue-light-emitting diodes (LEDs).When in use, the lid closes, leaving the sample holder in darkness.The blue LEDs are switched on to measurethe optically stimulated luminescence (OSL) signal (© Risø National Laboratory, Denmark).

panels in Fig 9 show photographs of the TLemitted by a number of aliquots of quartz,each comprising many hundreds of sand-sizedgrains (c 0.2mm diameter). One can see thatsome grains emit most strongly in the bluepart of the spectrum while others emit in thered. More detailed spectral analysis of a rangeof quartz samples confirms that these twoemissions (460–480nm and 610–630nmrespectively), along with a series of emissionsin the violet (360–420nm) are common tomany types of quartz (Fig 10).

Feldspars cover a wide range of chemicalcompositions.Their luminescence emissionspectra are more complex than those fromquartz, with luminescence emitted at variouswavelengths across the spectrum fromultraviolet to red and infrared (Fig 10).

Research and dating applications have focussedon looking at emissions from minerals in theblue and ultraviolet parts of the spectrum asmost laboratories are equipped withinstruments designed to look solely in thesewavelength regions. Given the widespread useof blue LEDs emitting between 450nm and490nm for optical stimulation, the mostcommon filter used to reject this stimulationlight is the Hoya U-340.This filter transmitsemissions from c 280nm to 380nm (ie thenear ultraviolet) (Fig 11), including the emissionat 370nm (Fig 10), so that it can be detectedby the photomultiplier tube.

2.5 Anomalous fading in feldsparsOne of the main reasons for the focus inrecent years on quartz for dating has beenthat, providing one works with electronsoriginating from a sufficiently deep trap, onecan be confident of their stability. In contrast,analysis of feldspars has shown a morecomplex situation. As with quartz, electronsthat are stored in traps that are deeper belowthe conduction band are calculated to bemore stable than those from shallower traps.However, laboratory experiments have shownthat electrons from deep traps in feldspars areless stable than expected – a situation calledanomalous fading.This means that the size ofthe observed luminescence signal decreases asthe sample is stored in the laboratory, anddoes so in an anomalous manner because thephysical parameters of the trap suggest that itshould be stable.There has been substantialdebate in the discipline about the nature ofanomalous fading, about whether it is auniversal phenomenon shared by all feldspars,and whether it can be overcome.While this isan ongoing topic the current consensus is thatmany types of feldspar do exhibit anomalousfading, and also that the rate at which thesignal fades varies from one sample to another.

8

Fig 7 Luminescence emissions are detected by a highly sensitive light detector – a photomultiplier tube (PMT).The PMT liesabove the sample, separated from it by glass detection filters. Stimulating the sample with heat from a metal heater generatesa TL signal; light-emitting diodes (LEDs) give an OSL measurement – blue LEDs are commonly used for quartz, their emissionrange reduced by a cut-off filter (designed to block LED light from the PMT); infrared LEDs stimulate a signal from feldspar.

Fig 8 Reduction of TL and OSL signals from quartz and feldspar by exposure to light (redrawn after Godfrey-Smith et al1988). Note that both axes have logarithmic scales.

Fig 9 (upper row) Five quartz samples shown in room-light conditions (each metal disc = 9.8mm diam; quartz grains c 0.2mmdiam). (lower row) Image, taken using highly sensitive photographic paper, of the thermoluminescence (TL) signal emitted bythe grains being heated. Some grains emit most strongly in the orange–red part of the spectrum; others are strongest in theblue (from Hashimoto et al 1986).

Additionally, it is clear that luminescence signalsemitted at different wavelengths may also fadeat different rates.

Two broad approaches have been suggestedto overcome the problem, in order to make itpossible to use feldspars for routine dating.Thefirst is to characterise the rate of anomalousfading, and then to apply some form ofmathematical correction.The second is to lookat different luminescence signals from feldsparand to try to find one that does not exhibitanomalous fading, eg red emission (Fig 10).

2.6 Other luminescence signalsNovel methods of obtaining a luminescencesignal from minerals are constantly indevelopment. Different procedures ofluminescence production may be used, such asradioluminescence (RL), isothermal TL (ITL)and Thermally-Transferred OSL (TT-OSL).Some of these procedures have the potentialto yield signals that may be useful in extendingthe age range of luminescence dating, but atthe moment they remain experimental.Routine luminescence dating is based on theapplication of TL or OSL to heated materials,and OSL to sedimentary materials, andmeasurement is presently mostly made onquartz.

3 Measurement of DeIn the application of luminescence to datingarchaeological or geological materials, theobjective of the luminescence measurements isto calculate the amount of radiation that thesample has been exposed to since the eventbeing dated. It is the combination of thisradiation dose and the dose rate (see section4) that enables an age to be determined.Laboratory measurements are used tocalculate the amount of laboratory radiationthat is equivalent to that received during burial– the equivalent dose (De).

The amount of light emitted by a sample perunit radiation dose varies from grain to grain,depending on each grain’s individual geologicalhistory.Thus one must use a set of laboratorymeasurements to calibrate the luminescenceresponse of each sample, and use these toderive a measure of the dose received innature (De).Two approaches have been widelyused: additive dose and regeneration (Fig 12).Of these, the regeneration method is now themost frequently used.

In the regeneration approach, theluminescence signal originating from theunknown dose (De) is measured. Allsubsequent measurements are made afterresetting the luminescence signal in the sample,by exposing a set of aliquots to sunlight or to

9

Fig 11 Blue LEDs, with emission centred at c 470nm, createa very bright source of light used to stimulate an OSL signalfrom quartz.The luminescence signal emitted from a sampleis typically a million million (ie 1,000,000,000,000) timesweaker than the LEDs. Such a weak signal is detected usinga photomultiplier tube (PMT), filtering the PMT from theblue LEDs with a Hoya U-340 (which blocks lightwavelengths > c 380nm).

Fig 10 Quartz (above) and feldspars (below) emitluminescence at wavelengths ranging from ultraviolet throughvisible into infrared – dark bands show the strongestemissions. Quartz shows clear TL peaks – some are stablethrough geological time and found in natural samples; othersare relatively short-lived (peaks > 200°C) and only found inartificially irradiated samples (modified from Stokes and Fattahi2003).

Fig 12 Two ways to calculate De from luminescencemeasurements: (above) ‘additive dose’ gives laboratoryradiation doses on top of the signal acquired during burial,causing the luminescence signal to increase, to which amathematical function can be applied – the function isextended until it intersects the x-axis, giving the estimate ofthe radiation dose the sample received during burial (De);(below) ‘regeneration’ measures the intensity of theluminescence signal from the natural radiation dose – shownas a horizontal dotted line.The sample’s luminescence signalis reset by heating or exposure to light, and then givendifferent laboratory radiation doses, characterising thegrowth of the luminescence signal with dose. De iscalculated by finding the laboratory radiation dose that givesthe same luminescence signal as that found in the samplewhen it was recovered.

10

Fig 13 The single aliquot regenerative dose (SAR) procedure applied to quartz: the growth of the signal with dose is characterised by administering a number of laboratory doses(regeneration doses) of different sizes (10Gy, 30Gy, etc) and measuring the resulting OSL signals L1, L2, etc. After each measurement the luminescence sensitivity is measured by giving a fixeddose (here 5Gy) and measuring the resulting OSL signal (T1,T2, etc).The effect of changes in sensitivity can be corrected for by taking the ratio of the luminescence signal (Lx) to the responseto the fixed dose (Tx).The plot of the sensitivity-corrected OSL (Lx/Tx) as a function of the laboratory dose (bottom) can be used to calculate De (here 22Gy) for that aliquot when theratio of the initial measurements on the natural sample (LN/TN) is projected onto the dose response curve.

a controlled source of illumination in thelaboratory (or, for samples heated in antiquity,by heating a set of aliquots). Known laboratorydoses are then applied and this regeneratedluminescence signal is measured (Fig 12). Acurve is then produced characterising thegrowth of the luminescence signal from itsresidual level (after optical bleaching orheating).The aim of this approach is todetermine the laboratory dose that generatesa luminescence signal that matches theintensity of the signal originating from thenatural sample.

The optimal method for determining De hasbeen the subject of intense research activity inthe last twenty years, and many differentapproaches are described in the academicliterature. In 2000, the single aliquotregenerative dose (SAR) procedure wasdescribed (Murray and Wintle 2000).Themethod appears to be appropriate to manysamples, has been shown to give accurateresults when correctly applied, and hasbecome the routine method of choice formeasurement of De in quartz. A brief outlineof some of the methods used to generateluminescence ages prior to 2000 is given inDuller (2004) and Lian and Roberts (2006).

3.1 SAR protocolThe procedure involved in single aliquotregenerative dose (SAR) measurements isillustrated in Figure 13.

It comprises a series of cycles. In the first cyclethe OSL signal (denoted L) from the aliquotarises from the radiation dose to which the

sample was exposed in nature, and hence isgiven the term LN. In the second cycle thealiquot is exposed to an artificial source ofradioactivity in the laboratory.The OSL signalis then measured – this is the first laboratoryregenerated signal, L1. Subsequent cyclesmeasure L2, L3, etc as different regenerationdoses (eg 10Gy, 20Gy, etc) are given to thealiquot. All of these measurements ofluminescence are preceded by a preheat –heating the sample to a fixed temperature(normally between 160°C and 300°C) andholding it there for a short period of time (eg10s).This procedure removes unstableelectrons from shallow traps so that the OSLsignal comes only from electrons that wouldhave been stored safely through the burialperiod.

The brightness of these luminescence signals(L1, L2, L3 etc) can be used to construct adose response curve (Fig 12, lower). However,the luminescence sensitivity of the aliquot –the amount of light it emits for each unit ofradiation to which it is exposed – changesdepending on the laboratory proceduresundertaken (eg the temperature and durationof the preheat treatment), and conditionsduring burial (eg ambient temperature).Thesecond half of each cycle in Fig 13 addressesthis problem by measuring the luminescencesensitivity. It is this second half of each cyclethat was the major breakthrough in SAR: ifsuch changes are not compensated for, thenthe calculation of De will be incorrect.

The luminescence sensitivity of the quartz ismeasured by giving a fixed radiation dose

(commonly known as a test dose) in thesecond half of each cycle and then bymeasuring the resulting OSL signal (TN,T1,T2etc).The effect of any change in sensitivity canbe corrected by plotting a graph, not of theluminescence signal (Lx) as a function ofregeneration dose, but of the sensitivitycorrected luminescence signal (Lx/Tx) (Fig 13).De can then be calculated, based upon thisluminescence signal, corrected for any changesin sensitivity that may have occurred.

The SAR protocol has been widely applied toboth heated quartz and unheated quartz(Wintle and Murray 2006) and been shown togive ages consistent with independent agecontrol (Murray and Olley 2002). Fig 13illustrates the idea behind the SAR procedureand shows three cycles of measurement. Inpractice, more cycles would be used, typicallybetween five and ten in total (Fig 14).Theseadditional cycles are used to define the doseresponse curve using a variety of regenerationdoses, and to enable a number of checks(described below) to be made on thebehaviour of each aliquot (Wintle and Murray2006). It is crucial that these checks are made,as it is known that although SAR does work onmost quartz samples, it does not work on all. Inparticular, some dim samples may not beappropriate for analysis using SAR because theylack the part of the OSL signal that gives therapid decrease in signal observed during theinitial part of the OSL measurement (Fig 6).

3.1.1 Recycling testOnce the dose response curve is defined asdescribed above, it is routine to repeat the

11

Fig 14 SAR dose response curve for a quartz aliquot fromKalambo Falls, Zambia, measuring eight cycles: first cycle –the natural signal (LN) and its response to a test dose (TN),enabling calculation of the sensitivity-corrected OSL(LN/TN); next, regeneration doses of 16, 32, 64, 96, 128 and0Gy give sensitivity-corrected values L1/T1, L2/T2, L3/T3,L4/T4, L5/T5 and L6/ T6 (these data define the doseresponse curve for this aliquot). By interpolating the value ofLN/TN onto the dose response curve, De can be calculated;in this case 25.6 ±1.3Gy).The final cycle repeats themeasurement for a regeneration dose of 32Gy (L7/T7).Theratio of the two measurements for 32Gy (L7/T7 divided byL2/T2) is known as the recycling ratio, and for this aliquothas a value of 0.99 ±0.05 – close to the ideal value of 1.00,and well within the accepted range of 0.9 to 1.1.

Sen

sitiv

ity C

orre

cted

OS

L(L

x/T

x)

Dose (Gy)

0

0

1

2

3

4

5

6

25 50 75 100 125

12

measurement of the luminescence signalresulting from one of the radiation doses. If theprocedure is working appropriately, and theSAR corrects for changes in the sensitivity ofthe sample, then the luminescence signals (L/T)of the repeated dose (eg 32Gy in Fig 14)should be the same.The ratio between the twosensitivity-corrected luminescence signals is therecycling ratio – ideally 1; in practice, a ratiobetween 0.9 and 1.1 is deemed acceptable.Values greater than or less than these limitssuggest that the procedure or the sample areinappropriate, and therefore that the resultsfrom that aliquot should be rejected.

3.1.2 Preheat testBefore each luminescence measurement, thealiquot is heated between 160°C and 300°Cfor 10s to remove unstable electrons.Theappropriate preheat temperature isdetermined by making measurements of Deusing a range of temperatures. Fig 15 showsthe results for one sample, measuring De for24 aliquots: three aliquots with a preheat of160°C, three at 180°C, three at 200°C etc toa maximum of 300°C.This procedure showsthat the same De value is obtained fortemperatures from 160°C to 260°C. Attemperatures > 260°C De increases, probablybecause of thermal transfer.

A strength of the SAR procedure is that itworks with a range of temperatures.Nevertheless, it is important to test one ortwo samples from within a suite of samples todetermine if a suitable preheat is achievable.

3.1.3 Dose recovery testOne of the most powerful tests available –commonly undertaken with SAR – is the doserecovery test.This involves removing thetrapped electron population from the sample,thus mimicking the resetting of the sample atthe time of the event being dated.

For sediments this involves exposing an aliquotof naturally irradiated grains to daylight or toan artificial light source, irradiating the aliquotwith a known dose (eg 10Gy), and thentreating this dose as an unknown.The SARprocedure is applied. If it is appropriate thenthe value of the calculated dose should matchthe known laboratory dose. It is common tochoose a known laboratory dose close to thevalue of De measured for the sample. In thisway, complications introduced as samplesbecome close to saturation at high dose values– or difficulties arising as one approaches thedetection limit of the instrument for low dosevalues – are assessed implicitly. If a sample failsa dose recovery test then it is unlikely that theDe calculated for the sample will be correct.

Fig 16 Replicate De measurements for two contrasting samples: (a) 31 measurements of a beach sand sample (Aber73/BH1-1) from Dungeness, all giving similar values. Since the samples belong to a single population, the mean De can be used for agecalculation (3.70 ±0.06Gy); (b) 44 measurements for a sample of fluvial sand from Crete (Aber60/AN2002/4), showing largeDe variability. Using the average value of De to calculate a luminescence age would give a meaningless result.The data for eachsample are shown as a histogram, and each De value (ranked in increasing order) is a point with an error bar of analyticaluncertainty.

Preheat Temperature (°C)

Num

ber

of a

liquo

ts p

er in

terv

alE

quiv

alen

t D

ose

(Gy)

Equivalent Dose (Gy)

0

00 2 4 6 8 10

0

20

40

60

80

5

10

15

20

25

150 200 250 300

1

2

3

4

5

(a)

(b)

Fig 15 Preheat plot, showing De (and associated measurement errors) for beach sand buried beneath gravel at DungenessForeland, measured at preheat temperatures from 160°C to 300°C, for 24 aliquots (Roberts and Plater 2005; 2007). Aplateau in De values extends from 160°C to 260°C (blue symbols) and the average value (dashed red line) was used tocalculate the age of the sample (3850 ±180a; Aber73/BH-3/2). At higher preheat temperatures De increases (grey symbols),probably due to thermal transfer, so these data are excluded.

3.2 Replicate measurements of DeSingle aliquot measurements reduce the timeinvolved in determining De by automatingmuch of the procedure; and they make itfeasible to make replicate measurements of Defor each sample.Thus it is routinely possible toassess the reproducibility of De within eachsample. Replicate determinations of De wouldbe expected to form a normal or log-normaldistribution where three conditions hold true:(1) the sample contains grains that werecompletely bleached at deposition; (2)variations in the annual dose due to small-scaledifferences in the concentrations of uranium,thorium and potassium are small; and (3) nopost-depositional mixing has occurred.In this situation, averaging the replicate Demeasurements gives an increase in theprecision of the age calculated (Fig 16a).

If replicate measurements of De are not similar(as in Fig 16b) then this implies that thesample is more complex, and requires furtherinvestigation.The most common cause of suchlarge scatter is that the trapped electronpopulation in the different mineral grains ofthe sample was not completely reset duringthe event being dated – this is known asincomplete bleaching.The De values obtainedfor quartz from a modern sand dune form atight cluster close to 0Gy (Fig 17a). However,when rivers transport sediments not all themineral grains are exposed to sufficientdaylight before deposition to reset their signal(Fig 17b).Thus the De that is measured in thelaboratory will be the sum of the doseacquired since the sample was deposited, plusthe residual signal remaining at the time ofdeposition. In such situations, the best estimateof the dose that has been acquired since theevent of interest will be given by those aliquotswith the lowest De values. Statistical modelshave been developed (Galbraith et al 1999;Galbraith 2005), to deal with such situations,and can be applied where the cause of thedistribution in De values can be understood.While much can be learned from analysis ofthe distribution of De values alone, it is alwaysimportant to consider the depositional contextfrom which a sample has been collected whenassessing the most likely cause of scatter in thedistribution of De values.

The second potential cause of De variation isdose rate differences in different parts of thesample – different parts of the sample havereceived different radiation doses during theperiod of burial. Fortunately, such variation inmicrodosimetry can normally be avoided bycareful sample selection. Situations where thisproblem is likely to be most severe are thosewhere the sediment contains a smallproportion of highly radioactive grains – for

13

Fig 17 Histograms of De values for two samples, demonstrating the potential problem of dating fluvial samples. Aliquotscontaining between 60 and 100 quartz grains (125–180μm diam) were analysed: both samples should give De values closeto zero, as they were collected from sites of current deposition. (a) Sand dune – De values close to zero indicate that it wasexposed to sufficient daylight at deposition to reduce the luminescence signal to near zero. In contrast, (b), from a fluvialenvironment, produced many De values significantly greater than zero, implying that only some of the grains in this samplewere exposed to sufficient daylight (modified from Olley et al 1998).

Fre

quen

cyF

requ

ency

Equivalent Dose (Gy)

Equivalent Dose (Gy)

0

0

1

2

3

4

5

0

0 5 10 15 20

1 2 3 4 5 6

2

4

6

8

10(a)

(b)

instance zircons – or where a sample is veryheterogeneous, eg owing to the presence ofgypsum with low radioactivity, or wood ashwith relatively high radioactivity from a highpotassium content.

Another situation that might produce a widerange of De values is a sediment samplewhose grains are of different, mixed ages.Mixing might have occurred in antiquity (Davidet al 2007), possibly by human activity, or itmay have occurred during sampling.Wheresamples of different ages are closelyjuxtaposed, for example, more than a singlecontext might have been sampled in error(Jacobs et al 2008).

3.3 Aliquot sizeIn situations where different mineral grainshave different De values, the results of singlealiquot analyses will depend on how manygrains are contained in each aliquot. In typicalluminescence measurements, sand-sizedmineral grains c 0.2mm in diameter aremeasured. For such a grain size it is possible tomake single aliquots containing about athousand grains, hundreds of grains, tens ofgrains (Fig 18) or even a single grain.

Fig 19a shows the result of making replicateDe measurements on a sample. De valuesobtained are relatively similar.Thesemeasurements were made using aliquots of c1,000 grains. Measurements of the samesample made using aliquots with c 200 grainsshow more variation in De (Fig 19b), andmeasurement of single grains increases thevariation further (Fig 19c).This sample consistsof grains with different De values, as is clearlyshown in the single-grain results.Themeasurements on aliquots with hundreds orthousands of grains are masking thesevariations by averaging the results from manygrains.Thus, if a sample is thought to beaffected by incomplete bleaching or mixing,then making replicate measurements onaliquots with relatively few grains, or even withsingle grains, should be used to test thispossibility.The spread in De shown in Fig 19a,as obtained in routine measurements usingaliquots of c 1,000 grains, would notnecessarily cause suspicion of incompletebleaching.

3.3.1 Single-grain OSL measurementsIn the last five years, measurements on single,sand-sized grains have become feasible using afocussed laser.This laser replaces the blueLEDs used for optical stimulation inconventional measurements, and can bedirected so that it stimulates each individualgrain in turn, making OSL signal measurementpossible. On each 9.8mm diameter aluminium

disc, 100 individual grains can be held, each inits own hole, 0.3mm in depth and 0.3mm indiameter, drilled into the surface of the disc(Fig 20).

These measurements have demonstrated thatfor most samples of quartz, the luminescencesignal that is observed is dominated by thesignal from a relatively small proportion of thegrains.Typically, 95% of the light that isobserved as OSL originates from less than 5%of the grains.The reason for this variability inthe brightness of individual mineral grains, eventhough they are collected from the samesample, is not well understood, but is thoughtto relate to the original conditions offormation of the quartz grain. However, thisobservation has two important implications.First, that when single-grain measurements aremade, many of the grains do not emit sufficientOSL to enable De to be determined. It is notuncommon for 5% or less of the grains thatare analysed to yield useful data. Second, thisfinding implies that even though aliquots mayphysically contain several hundred grains, theOSL signal observed is likely to originate froma much smaller number of grains.Thusmeasurements using small aliquots, typicallyconsisting of many tens of grains each, may

14

Fig 18 Aliquots for luminescence measurements may containdifferent numbers of grains: each metal sample disc is9.8mm diam; changing the proportion of the disc surfaceused varies the number of quartz grains (c 0.2mm diam)(upper right c 1000 grains, bottom left c 200, bottom right c20) (photograph by H Rodnight).

Fig 19 Potential effect of changing the number of grains inan aliquot – three histograms of De results for one sampleusing different numbers of grains in each aliquot: (a) c 1000grains, (b) c 200 grains and (c) single grains. Many grains inaliquot (a) averaged out differences, masking the realvariation; as the number of grains is decreased (b and c) thevariability in the De becomes obvious.This sample was notcompletely bleached at deposition, so an age based on thelarge aliquots shown in (a) would have been anoverestimate.

Fig 20 (top) Luminescence measurements made byfocussing a green laser beam on the sample (beam diam c0.02mm when it strikes a grain). (bottom) Individual quartzgrains (0.2mm diam) held in an array of 10 x 10 holesdrilled in the surface of the 9.8mm diam sample holder.Thelaser beam is moved using mirrors to strike one grain at atime, giving an OSL signal (© Risø National Laboratory,Denmark).

(a)

(b)

(c)

15

yield results that are similar to those fromsingle grains. Such measurements may thusreduce the need for time consuming single-grain analyses and the requirement for a laserstimulation system.

3.3.2 Displaying De datasets for samplesThe large variation in the brightness ofindividual grains also has an impact on theprecision with which De can be determined.Where a grain does not emit a brightluminescence signal, then the precision of themeasurements may be limited, while brightergrains give more precise De estimates.

For example, imagine a sample that has beengiven a radiation dose of 10Gy. A bright grainfrom this sample may yield a De of 9 ± 1Gywhile a dim grain may give 15 ± 5Gy. Bothvalues are consistent with the dose of 10Gygiven to the sample.We could also imagineanother sample that consists of a mixture ofgrains with different doses.Two grains from thissample could give De values of 9 ± 1Gy and15 ± 2 Gy. In this latter case, the two grainswould give ages that are inconsistent with eachother. If we plotted these two sets of resultson a histogram, such as those shown in Fig 19,they would appear to be identical, even thoughstatistically they are different – one isconsistent with both grains having received thesame radiation dose, while the other is not.This is because in a histogram there is no wayof taking into account the uncertainty on thevalues being shown.

An alternative method of displaying this typeof data is a radial plot (Galbraith et al 1999;Galbraith 2005).This method is widely usedfor displaying single-grain De data. Figure 21shows such a radial plot. On the diagram, eachpoint represents the De from a single grain ofquartz, in this case 289 grains.The horizontalaxis shows how precisely the De value isknown, with more precise values on the right-hand side of the graph.The precision can alsobe expressed as a relative error.Thus a valueof 10±2Gy has a relative error of 2/10 or20%.

The vertical axis, shown on the left-hand sideof the diagram, is called the standardisedestimate.This shows how different the Devalue for each grain is from some referencevalue.The calculation of this is complex (detailsare given in Galbraith 2005), but the importantproperty of the diagram is that if one draws astraight line from the origin (the point wherethe standardised estimate and the precisionare both zero) towards the curved axis shownon the right-hand side, any point on that linerepresents a grain with the same De value.Theprecision of De will vary, but De itself will be

the same.Thus the curved, radial axis on theright-hand side can be labelled so that one canread off the De values. By convention, this axisis logarithmic, as in Figure 21.

The data shown in Figure 21 clearly fall intotwo distinct groups.The upper group forms aband that points towards the radial axis andintersects with it at a value of 101Gy.The sitein South Africa from which this sample wascollected contains sediments from the MiddleStone Age, and these grains yield an age of 35± 2ka. However, collection of thisluminescence sample inadvertently cross-cutinto a different sedimentary unit, dating to theIron Age.The second band of grains,intersecting the right-hand axis at a value of2.4Gy, is the result of this mixing, and wouldgive an age of c 0.93 ± 0.04ka (Jacobs et al2008).

Single-grain measurements of De are verypowerful, and can help to elucidate the causesof complex De distributions (Jacobs andRoberts 2007). However, such measurementsare extremely time-consuming, and arefrequently made difficult by the lowluminescence signal levels encountered.Thus atpresent they are far from routine, and resultsobtained by using aliquots consisting of tens orhundreds of grains may be equally appropriate.In the future it is likely that single-grainmethods will become more commonplace.

3.4 TL and the plateau testThe SAR method described in section 3.1 isused increasingly for measuring De forsediments using the OSL signal from quartz.OSL is appropriate for dating sediments as it ismore rapidly reset by exposure to daylightthan TL is (Fig 8). For materials that wereheated in antiquity – such as pottery, burntflints and brick – one can use either OSL or TLsignals.

One advantage of TL measurements is thatadditional information is contained in the TLglow curve (Fig 5). In a TL measurement,luminescence signals emitted at highertemperatures tend to originate from traps thatare deeper below the conduction band (Fig 4),and hence are expected to have been stableover longer periods of time. A usefulapplication of this is the plateau test: De forthe sample can be calculated using theintensity of the TL signals measured at differenttemperatures (Fig 22 page 16). At lowtemperatures, the ratio of the natural signaldivided by the laboratory-irradiated signal isvery small, as the traps are unstable overgeological time. As the temperature increases,so too does this ratio, until above sometemperature, typically c 300°C, a plateau isobserved.The observation of a plateauconfirms that above this temperature rangethe traps are sufficiently stable to accuratelyrecord the radiation dose acquired over time.

Fig 21 Radial plot of De values for 289 single quartz grains from Sibudu Cave, South Africa.The data form two groups: onewith an average value of 101Gy (upper band), and one of 2.4Gy (lower band).The upper band contains 71% of the grainsand gives a luminescence age of 35 ±2ka, as expected for this sample from the Middle Stone Age (MSA).The 29% of thegrains in the lower band give an age of 0.93 ±0.04ka, consistent with the South African Iron Age (IA). At Sibudu Cave the IAimmediately overlies the MSA, and the single-grain analysis shows that in sampling the two contexts were inadvertently mixed(Jacobs et al 2008).

16

Additionally, when dating burnt flints, obtaininga plateau demonstrates that the material washeated sufficiently in antiquity to reset thesample. Data from the temperatures overwhich the plateau is observed are used toobtain De, which then forms the basis forcalculating the time since heating in the past.

4 Measurement of dose rateThe application of luminescence to datingarchaeological or geological materials relies ondetermining two quantities.The first is theamount of radiation absorbed by the sampleduring the period since the event being dated,measured as De.To determine the age of thesample in years, De has to be divided by theradiation dose received by the sample eachyear – the dose rate.

There are four types of environmentalradiation: alpha particles (α), beta particles(β), gamma rays (γ) and cosmic rays.The firstthree originate from naturally occurringelements in the sample itself and itssurroundings.The most important of thesesources are radioactive isotopes of uranium(U), thorium (Th) and potassium (K).Theradioactive isotope of potassium is 40K, whichdecays to 40Ca (calcium) or 40Ar (argon) by theemission of beta particles and gamma rays.Both 40Ca and 40Ar are stable.

Uranium and thorium are more complex: 238Udecays to form 234Th, which is itself radioactive,and will decay to form protactinium (234Pa), andso on until a stable isotope of lead (206Pb) isreached (Fig 23). Such a chain of isotopes iscalled a decay series. Similar decay series existfor 235U and for 232Th.The U and Th chains emitalpha, beta and gamma radiation andcontribute to the total radiation absorbed bysamples (Fig 24).

Cosmic rays originate from sources far out inthe universe.They are a type ofelectromagnetic radiation similar to gammarays derived from U,Th and K, but typicallywith much higher energy.The Earth is shieldedfrom most cosmic rays by the atmosphere andby the Earth’s magnetic field. However, aproportion does reach the Earth’s surface andpenetrate sediments.The radiation dosereceived decreases with depth, particularlyrapidly in the uppermost metre.

Alpha, beta and gamma radiation is alsoabsorbed and attenuated by the surroundings.Each type travels different distances inmaterials. Alpha particles are, atomicallyspeaking, large objects, and typically only travela few hundredths of a millimetre throughsediments; beta particles are electrons, and willtravel a few millimetres or more, while gammarays may penetrate as much as 0.3m (see insetto Fig 24).

Two approaches can be used to measure theannual radiation dose provided to a samplefrom the radioactive elements surrounding it:measuring the concentration of U,Th and K,and using these concentrations to calculate theradiation dose received by the sample; ordirectly counting the emission of radiation.Thebetter approach will depend upon the type ofsample being dated and on practicalconsiderations (primarily those of access).Thedifferent distances travelled by alpha, beta andgamma radiation directly influence the volumearound the sample that needs to be measured.

4.1 Chemical methodsA range of geochemical methods exists tomeasure the concentration of the keyelements responsible for delivering the

radiation dose – U,Th and K. For sedimentsand bricks, the dose rate from thesurroundings can be measured from a finelyground sub-sample of the sediment itself. Forpottery and burnt flints, both the sample andthe sediment surrounding the sample need tobe analysed.

Of the three elements measured, K occurs inthe greatest concentration, typically 0.5% to3% by weight. It can be measured using arange of methods, including atomic absorptionspectrophotometry (AAS), flame photometry,X-ray fluorescence and inductively coupledplasma mass spectrometry (ICP-MS). In allcases a precision of c 5% or better would beexpected.

U and Th occur in much lower concentrations,typically in the range of 1ppm to 10ppm.Therange of measuring methods is more limited,but includes ICP-MS and neutron activationanalysis (NAA).

Three major issues may arise with thesemeasurements: (1) Concentrations of theseelements may be close to the detection limitsfor some analytical facilities; minimumdetectable limits are often quoted, and onewould want these to be several times smallerthan the concentrations being measured. (2)Sample preparation – ICP-MS and many othermethods rely on dissolving the sample instrong acids to produce a solution that can beanalysed; in some samples, U and Th may beassociated with highly resistant minerals – ifthese are not completely dissolved during theanalysis then the measured values will beunderestimated. (3) These measurements aremade on sub-samples of only a few mg, andsuch a small sample may not be representativeof the sediment or brick being used for dating.

Once the concentrations of these threeelements are known, conversion factors enablethe calculation of the radiation dose rate(Adamiec and Aitken 1998). For example, 1%potassium in sediment will produce a gammaradiation dose rate of 0.243Gy per thousandyears (Gy/ka), a beta dose rate of 0.782Gy/ka,but no alpha dose rate, as the decay of 40Kdoes not result in the emission of alphaparticles. Adding together the alpha, beta andgamma dose rates gives the total radiationdose rate.

This approach makes a number of assumptions.For U and Th decay chains the concentrationmeasured is that of the parent isotopes. Figure23 shows the decay of 238U. Radiation isemitted at all stages of the chain, but thechemical methods described above will onlymeasure the concentration of U – they do not

Fig 22 A plateau test, showing the natural TL signal (N) from one aliquot and the TL signal from a different aliquot after it hadbeen irradiated with a beta dose in the laboratory to increase the signal (N+β) – the dots joined by the dashed line showthe ratio of the two TL signals.The ratio is low at low temperatures and increases with higher temperatures; above 300°C theratio becomes stable at c 0.47 (modified from Aitken 1985).

17

normally differentiate between the differenturanium decay series (238U, 235U), nor do theynormally measure the daughter products.Conventionally, calculation of the dose ratefrom 1ppm of U assumes that the ratio of theconcentrations of 238U to 235U is about 400, andthat all of the decay products are in secularequilibrium with their parent isotopes.

Secular equilibrium is the term used todescribe a decay series (eg Fig 23) when theactivities of each of the daughter products arethe same. For luminescence dating this iscommonly assumed to be the case, but incertain geochemical situations, this assumptionmay not hold (Olley et al 1996). For example,where water has percolated through sediment,this can selectively remove uranium (eg 238Uand 234U) while leaving more immobiledaughter isotopes such as thorium (eg 230Th)behind. In seasonally wetted environments –such as marshes – where reducing conditionsoccur, radium (Ra) mobility is of concern.Where specific parts of a decay chain areeither lost or gained in these processes, thedecay chain is described as being ‘indisequilibrium’, meaning that the dose rate willchange through time, making calculation of theluminescence age complex. After some periodof time the rate at which the various isotopesare formed and decay will come into balance,and at that stage the sample is said to be ‘inequilibrium’. In practice, difficulties associatedwith disequilibrium can be avoided by selectingsamples from geochemical environments inwhich uranium, which is highly soluble, andthorium, which is insoluble, are not movingdifferentially. In particular, sediments that arewithin 0.3m (the range of gamma rays) of peator other highly organic sediments should beavoided. Iron-stained sediments, especiallywhere there is evidence for significantmovement of groundwater, and samples withevidence for precipitation of calcite should alsobe avoided. However, when working inlimestone caves such problems may beubiquitous.Where disequilibrium is suspected,high-resolution gamma spectrometry (anemission counting technique described below)may be appropriate.

4.2 Emission countingInstead of measuring the concentration of U,Thand K and using these to calculate the doserate, a more direct method is to count thealpha, beta and gamma radiation.

Alpha particles can be counted by using either athick source alpha counter (TSAC) or alphaspectroscopy.Where disequilibrium is suspectedin a sample (see discussion in section 4.1), alphaand gamma spectroscopy are the best ways ofstudying the U and Th decay series.

Fig 23 The decay chain of 238U: arrows show the sequence of isotopes produced by emission of a beta particle (horizontalarrows) or an alpha particle (diagonal), with half-lives varying from hundreds of thousands of years to less than a second (Z =atomic number, the number of protons in the nucleus and hence chemical properties; A = atomic mass (sum of protons andneutrons), which differentiates isotopes).

ZA

82 83 84 85 86 87 88 89 90 91 92

238 U4.47x109 yr

234 Th24.1 d

Pa1.17m

U2.45x105 yr

230 Th75.4x103 yr

226 Ra1600 yr

222 Rn3.83 d

218 Po3.05 m

214 Pb26.8 m

Bi19.9 m

Po162 s

210 Pb22.2 yr

Bi5.01 d

Po138.4 d

206 Pbstable

Fig 24 Environmental sources of radiation, showing ranges for alpha (α), beta (β) and gamma (γ) radiation and the decrease incosmic rays with increasing depth below ground.

18

TSAC is able to measure the total number ofalpha particles, and by exploiting the rapiddecay of 220Rn (radon) in the Th decay series,the relative proportion of the alpha particlesthat come from the U and Th decay series canbe calculated separately.

Alpha spectroscopy is more complexanalytically, requiring the sample to bedissolved; but by using this method the energyof each alpha particle can be measured, whichis characteristic of the isotope undergoingdecay. In the same way, the energy of gammaradiation emitted from a sample ischaracteristic of the isotope undergoing decay.High purity germanium (HpGe) detectors, keptat liquid nitrogen temperatures (–196°C), areable to measure this energy, and produce aspectrum for each sample (Fig 25).

Beta particles, unlike alpha particles andgamma rays, are emitted at a range ofenergies, and thus there is no point inspectroscopy. However, simply counting thebeta particles is perfectly feasible, for exampleby using a Geiger-Müller based system, and thisdirectly assesses the beta dose rate.

One radioactive element, radon (Rn), is a gasand thus there is the potential for it to diffuseout of sediments. Rn isotopes occur in the232Th, 235U and 238U decay chains, but they havedifferent half-lives – ranging from 3.96s (219Rn)to 3.83 days (222Rn).The longer-lived isotope222Rn is part of the 238U decay series (Fig 23),and because of its longevity there may be timefor the gas to move through the poresbetween sediment grains and escape into theatmosphere. If this occurs, the daughterisotopes will be produced away from thesample, thus reducing the radiation dose rateand causing disequilibrium in the decay chain.This situation will not cause difficulties ifemission counting methods are used, but couldbe an issue if only parent concentrations (eg Uand Th) are measured using chemical methods.These methods enable alpha, beta and gammaradiation flux to be measured, and alpha andgamma spectroscopy enable assessment of theextent of equilibrium in the decay chains.

Owing to the distances they travel, laboratorymeasurements of the alpha and beta flux on asmall (c 10g) sub-sample are likely toaccurately measure the natural flux of thesample. However, gamma radiation travels upto 0.3m in sediments. Laboratorymeasurements, based on emission counting orchemical methods, will only accuratelydetermine the gamma dose if the materialaround the sample is homogeneous over ascale of c 0.3m in all directions. In manysituations this will not be the case and then in

Fig 25 Part of a spectrum obtained from gamma spectroscopy: gamma rays emitted by specific isotopes have fixed energies;some of the more easily identifiable isotopes are shown (modified from Aitken 1985).

Fig 26 (above) Measurement wasundertaken for one hour and producedthe energy spectrum shown (peaks usedto estimate K, U and Th concentrationsare marked – the energy peak of1460keV from K is particularlyprominent). (left) A portable gammaspectrometer being used to make an insitu measurement of the gamma doserate for a sample collected from a buriedsurface preserved within Silbury Hill,Wiltshire.

Energy (MeV)

Cou

nts

Energy (KeV)

0

102

103

104

10000

1000

100

10

10 1000 2000 3000

0.1 0.2 0.3C

ount

s

19

situ measurements need to be made.

4.3 In situ measurementsThere are two approaches. One is to useartificial phosphors that are sensitive toradiation – for example, specially treatedcalcium fluoride or aluminium oxide.These canbe buried – at least 0.3m deep in order to get acomplete gamma field – at a site for between afew months and a year (or more), thenrecovered and their luminescence measured.Thisapproach has the advantage that only smallquantities of artificial phosphor need to beburied, in capsules c 10mm in diameter, thusminimising disruption to the site.Thedisadvantage is the need to visit the site twice.

The second approach, and the one used morecommonly, is to use a portable gammaspectrometer. In laboratory measurementsHpGe crystals used in high-resolution gammaspectrometry must be at liquid nitrogentemperatures. As this is impracticable in thefield, sodium iodide (NaI) crystals that can beoperated at room temperature are used.Thequality of the spectrum obtained from NaIcrystals is much poorer than that from HpGe,and so it is not possible to detect the samenumber of isotopes. However, peaks in thespectra obtained from a NaI crystal can beassigned to the U and Th decay series, and toK (Fig 26).The advantage of such in situmeasurements is that they accurately capturethe gamma dose rate, even if there isheterogeneity in the radiation field.Thedrawback is that such measurements requirethe probe to be inserted into a position asclose as possible to the exact location ofsample collection. Additionally, most NaIcrystals used in luminescence researchers are c50mm in diameter, requiring a hole at least thiswide and 0.3m deep to ensure that acomplete gamma field is measured.To obtain acomplete spectrum such as that shown in Fig26 requires approximately one hour persample, although this will vary depending onhow radioactive the site is. Smaller NaI crystalsare available, but require commensuratelylonger counting times. More rapidmeasurements can be made by integrating thearea under the curve in Fig 26; however, suchmeasurements will yield only the total gammadose rate, rather than information about theindividual U,Th and K concentrations.

4.4 The impact of water contentExcept in situ measurements, all the methodsdescribed above are made on dried sub-samples.Water between mineral grains in theenvironment absorbs some of the radiation,meaning that only a proportion of theradiation emitted by the U,Th and K isabsorbed by the mineral grains making up the

sample.The larger the amount of water, theless radiation is absorbed by the minerals.Calculations can compensate for this effect, butthey require an estimate of the water contentthroughout the burial period.This is a difficultparameter to estimate, but some constraintscan be put upon the value.

The upper limit can be determined bymeasuring the saturation water content(typically no more than 20–30% for sandysediments), and by making measurements ofthe modern-day water content. A reasonablerange of values can then be estimated by thesite researcher and luminescence researcher.The precise impact of water on dose rate, andhence the age, will vary from sample to sample,but typically, a 1% increase in water content willincrease the calculated age by c 1%. Particularlyfor sediments, uncertainty in the water contentis commonly one of the largest sources ofuncertainty in the final age estimate derivedusing luminescence, so careful consideration ofthis is essential. For caves, dunes and lakes, theuncertainty is likely to be fairly small, while foralluvial deposits it may be larger.

Samples with a high proportion of peat shouldbe avoided because of the potential fordisequilibrium (see section 4.1). However, ifanalysed, particular care is needed in estimatingthe water content, as it may have changedthrough time due to compression anddewatering. Additionally, organic matterabsorbs radiation differently from water, andthus specific attenuation factors need to beused in dose rate calculations.

4.5 Cosmic ray contributionCosmic rays are an additional source ofradiation.The cosmic ray dose rate dependsprimarily on three parameters: latitude, altitudeand the thickness of any sediments or othermaterials that overlie the sample.The shieldingprovided by the Earth’s magnetic field varieswith latitude, and thus the dose rate increasescloser to the poles, where the shielding is least.However, this effect is normally only significantfor latitudes greater than c 60°N or S.Shielding from the atmosphere depends on itsthickness; thus dose rate increases withaltitude.

Standard calculations for the cosmic dose rateare appropriate for samples taken close to sealevel. In practice the altitude normally onlyneeds to be considered at heights > c 500mabove sea level.The most significant factor isthe thickness of any sediment, rock (especiallyin cave sites) or built structures overlying thesample. For sites near sea level in mid-latitudes,at a depth of 0.3m the cosmic dose rate is c0.2Gy/ka, while at 10m it falls to c 0.07Gy/ka.

The effect of overlying sediments or buildingscan be calculated accurately when thethickness of this shielding is known. In manycases samples might have been buriedrelatively rapidly following deposition, and thethickness of the overburden has remainedeffectively constant; but this will not always bethe case. For many samples, the cosmic raydose is a small proportion of the total dose.Typical silts and sands have combined beta andgamma dose rates of c 2 to 3Gy/ka, and thecosmic dose rate will be c 0.1 to 0.2Gy/ka –less than 10% of the total.

Calculating the cosmic dose rate isstraightforward for most samples. As itconstitutes only about 5% to 10% of the totaldose it makes any uncertainty in the valuerelatively unimportant. However, whereconcentrations of U,Th and K are low, cosmicdose rate becomes proportionately moreimportant. In carbonate-rich sediments – asare commonly found in East Anglia and partsof south-east England – beta and gamma doserates may be closer to 1Gy/ka or less. Forthese samples cosmic dose rate may be 20%or more of the total.

The situation is more complex if the thicknessof the overburden has changed dramaticallythrough time – for example when sedimentshave accumulated steadily through time,increasingly shielding the sample from cosmicrays. If there is some indication of the timing ofthese changes, the impact can be modelled.

5 Limits of luminescence dating5.1 Suitability of materialThe quality of luminescence dating results arecrucially dependent on the suitability of thesamples, and centrally, whether a usefulluminescence signal can be obtained.

Predicting where such samples will beencountered is not always simple, and this iswhy a pilot study can be important. In general,quartz from heated materials tends to givebrighter signals, so pottery, burnt flints andbricks generally give bright signals. In unheatedsediments, a major control on the brightnessseems to be the initial geological origin of thequartz. General observations indicate thatquartz derived from granites and fromhydrothermal veins are dim. Once the grainshave been through many cycles of reworking(eg to form a sandstone) there is a higherpercentage of bright grains.

5.2 Resetting of the trapped electronpopulationIn all luminescence dating, the event beingdated is the last time that the relevant trappedelectron population in the crystal (Figs 2 and4) was reduced to a low level.Thus thearchaeological value of the age obtained willbe related to whether that resetting eventcoincided with the archaeological event ofinterest, and the completeness of that resettingevent.

For heated materials, the crucial issue iswhether the sample was heated to a highenough temperature, and for a long enoughperiod of time, for the trapped electronpopulation to be removed.The plateau test(Fig 22) undertaken during TL measurementscan be a useful indicator, and furtherconfidence can be given if replicate samplesyield consistent results.

For unheated samples, the critical factors arethe duration and intensity of the light to whichthey were exposed at deposition. Inadequateexposure to daylight at deposition leaves aresidual population of trapped electrons(leaving the sample incompletely bleached). Ifallowance for this is not made, the age of thesample will be overestimated.

Methods to explicitly test this for each sampleare described in section 3.

5.3 Upper age limitsThe range of ages over which luminescencedating is applicable varies from one site toanother, depending on the nature of theluminescence signal and the dose rate fromthe environment.The upper age limit isnormally controlled by saturation of theluminescence signal. Fig 27a shows that the

luminescence signal, whether OSL or TL,increases as the size of the prior radiationdose increases. At first, this increase with doseis almost linear, but at some stage the trapswithin the crystal where electrons can bestored become full. As this happens, theluminescence signal grows more slowly, until allthe traps become full, whereupon theluminescence signal ceases to increase despitecontinued exposure to radiation.This is knownas saturation.This saturation imposes an upperlimit to luminescence dating.

The dose at which the sample saturates variesfrom one sample to another, as does the rateat which it receives radiation.Thus it isimpossible to give a precise upper limit to theage that can be obtained; instead the limit willbe the maximum value of De that can bemeasured.

It is for this reason that reports onluminescence ages should always contain anexample of the dose response curve obtained,as this indicates how close the sample is tosaturation.The OSL signal from quartznormally saturates at approximately150–300Gy (although this range also variesbetween samples); and with typical dose ratesof 2Gy/ka this gives an upper limit of 75–150ka.

It should also be noted that as samplesapproach saturation (c 200Gy in Fig 27a),errors in the De value increase, for tworeasons: because as the dose response curveflattens, small uncertainties in the value of thenatural signal being projected have a largerimpact on De values (visible in the asymmetricerror given on individual De values); and, moredifficult to assess, because of the impact ofsystematic uncertainties in the measurement ofthe dose response curve – thus ages obtained

with De values close to the limit of saturationshould be regarded with additional caution.At present there is no consensus on howclose to the limit of saturation one can reliablygo.Wintle and Murray (2006) suggest that it isonly prudent to work in the range where thenatural signal (LN /TN) is 85% or less of themaximum luminescence signal obtainable. In Fig27a the maximum signal on the y axis is 20,while the natural signal is 18.Thus the naturalis 90% of the maximum value and is beyondthe range suggested by Wintle and Murray(2006). However, the reality is often morecomplex, for two reasons. First, the doseresponse curve shown in Fig 27a is for a singlealiquot, and it has been commonly observedthat the behaviours of different aliquots of asingle sample are different at high doses.Thus,as higher De values are approached, somealiquots fall below the 85% limit suggested,while others do not.To discard only thosealiquots that fail this test would lead tosystematic bias. So should any sample in whichone aliquot fails this test be rejected? Secondly,the dose response curve in Fig 27a increasesto a maximum value (c 20 on the y axis), anddoes not increase further.This response iswhat would be expected if a single type oftrap in the mineral were being filled, and isfitted with a mathematical expression knownas a saturating exponential.

However, it is not uncommon for somealiquots, and some samples, to have doseresponse curves where the luminescencesignal does not reach a maximum value, butinstead seem to consist of two signals – one ofwhich saturates, while the other appears tocontinue to grow to higher doses (Fig 27b).Here the criterion suggested by Wintle andMurray (2006) cannot easily be applied as thesample appears not to saturate.

20

Fig 27 (a) At low doses the luminescence signal grows almost in a straight line, but at higher doses it grows more slowly. Ultimately the signal stops growing, when the sample is said to besaturated.This limits the age range to which the method can be applied. For this sample, it is effectively saturated at c 200Gy; if the dose rate were 2Gy/ka, the dating limit would be < 100ka.(b) Other aliquots or samples show patterns where the signal begins to saturate, then continues to grow linearly.The reliability of using this additional growth at high doses is currentlyunknown.

(a) (b)

However, the cause of the continued growth of the luminescence signal at high doses is at present poorly understood, and it is not known whether this can reliably be used indating. Current results are equivocal.There areexamples in the literature where old ages were obtained using dose response curves with forms such as that in Fig 27b, and whichagreed well with independent age control (eg TL ages of Huntley et al 1993). A usefulanalogy for the present situation is theapplication of radiocarbon dating to oldsamples in the 1980s and 1990s. Radiocarbonages ranging from 35ka to 40ka were obtainedby laboratories for a number of samples,including ones that were key to archaeology(Roberts et al 1994), and these results yieldedanalytical errors suggesting that they hadrelatively small uncertainties. However, theintroduction of trace amounts of young carbon – either during sampling or in samplepreparation – had caused some of these agesto be erroneous. In the light of subsequent

improvements in techniques, it has becomeclear that not all the radiocarbon ages in therange 35–40 ka were wrong; but equally not all were correct. It is likely that a similarsituation prevails in the luminescence analysis of old samples.

5.4 Lower age limitsThe lower age limit is also difficult to define.This is controlled by two factors.The first ishow well the luminescence signal was reset atthe time of the event being dated.The secondis the luminescence sensitivity of the mineralbeing studied. For young samples, De is likely tobe small, and so the luminescence signal maybe weak. Recent OSL studies show that infavourable conditions – for example inapplication to coastal dunes – ages in the lastfew centuries can be achieved (eg Bailey et al2001; Sommerville et al 2003); and for heatedmaterials, such as bricks, results for the last fewcenturies can be obtained (Bailiff 2007). Foryoung samples it is particularly important to

undertake luminescence measurements to determine the most appropriate preheattreatment (see section 3.1.2, Fig 15).This isbecause thermal transfer can erroneouslyincrease the apparent age of sediments if too high a preheat temperature is used.

5.5 Accuracy and precisionThe accuracy of luminescence ages dependson a number of parameters. As discussed inPart B, it is essential that as muchmeasurement information as possible isincluded in any report or publication dealingwith luminescence ages. A number ofcompilations of ages measured on samples ofknown age have been produced. Forsediments dated using quartz OSLmeasurements and the SAR protocol, acomprehensive review was given by Murrayand Olley (2002), demonstrating goodconcordance with independent age estimatesfor the range from a few decades to morethan 200ka. Barnett (2000) undertook aprogramme of TL dating of pottery from laterprehistoric Britain that had been used todefine the typological framework for thatperiod. She found that where diagnostic formand surface decoration were present, theagreement between the luminescence ages onthe pottery and the ages from independentmethods was excellent (Fig 28). A recentcomparison of luminescence ages measuredon bricks with independent architecturalevidence (ranging from AD 1390 to 1740)showed that the mean difference between theluminescence age and the assigned ages was5±10 years (Bailiff 2007;see Part C, section 14).

Limitations on the precision of luminescenceages have been described (section 5.3 and 5.4 above).When uncertainties in themeasurement of the dose rate – oftendominated by issues of water content (seesection 4.4) – and De are combined, errorsquoted on luminescence ages normally rangefrom 5 to 10%, including both random andsystematic sources of error. Luminescence ages are normally quoted at one standarddeviation; that is, at the 68% confidenceinterval. Luminescence ages are obtaineddirectly in calendar years, and thus do notrequire further adjustment.Table 1 illustratesthe calendar ranges for luminescence ages with a 5% error.The example is shown forages determined in AD 2000.There is nocommonly accepted datum for luminescenceages, and so ages are expressed in yearsbefore the date of measurement.The term BP (before present) should never be appliedto luminescence ages as BP has specificmeaning relevant only to radiocarbon ages.These issues and how to quote luminescence

21

Fig 28 Comparison of TL ages on pottery and its archaeological age assessed on form and surface decoration (based onBarnett 2000): luminescence ages (red), with error bars (1σ, including random and systematic errors).

Table 1 Luminescence ages typically have errors between 5 and 10%. Illustrated here areexamples of a 5% error, expressed in years (annum, a; or kiloannum, ka)

luminescence age central date 5% error calendrical calendrical (years before (±1σ) bandwidth bandwidth AD 2000) (68% confidence) (95% confidence)42,000 ±2100a 40000 BC 2,100 years 42100–37900 BC 44200–35800 BC22,000 ±1100a 20000 BC 1,100 years 21100–18900 BC 22200–17800 BC12,000 ±600a 10000 BC 600 years 10600–9400 BC 11200–8800 BC5,500 ±275a 3500 BC 275 years 3775–3225 BC 4050–2950 BC4,500 ±225a 2500 BC 225 years 2725–2275 BC 2950–2050 BC3,800 ±190a 1800 BC 190 years 1990–1610 BC 2180–1420 BC2,600 ±130a 600 BC 130 years 730–470 BC 860–340 BC1,900 ±95a 100 AD 95 years AD 5–195 90 BC–290 AD1,000 ±50a 1000 AD 50 years AD 950–1050 AD 900–1100300 ±15a 1700 AD 15 years AD 1685–1715 AD 1670–1730100 ±5a 1900 AD 5 years AD 1895–1905 AD 1890–1910

22

ages are discussed in section 10.Part B Practicalities

6 Project management underMoRPHEThe potential for luminescence dating tocontribute to a project should be identified asearly as possible. Under Management ofResearch Projects in the Historic Environment(MoRPHE) guidelines (Lee 2006), it shouldform part of the project design document.Advice may be required on whether thearchaeological question under considerationcan be resolved by luminescence dating and, ifso, whether the available samples are suitablefor analysis.This advice should be sought in thefirst instance from an English Heritage RegionalScience Advisor (see Appendix 1). Moretechnical advice may be obtained from theEnglish Heritage Scientific Dating Team or fromthe collaborating luminescence laboratory (seeAppendix 2).

Where possible, a pilot project to assess thesuitability of the material is advisable.This maybe complicated by the turn-around time oflaboratories (typically six months), but if thisfits within the framework of the project then itis a valuable exercise. It is often difficult, evenfor someone with many years experience ofluminescence, to be confident that samplesfrom a specific site will be suitable for analysisunless a pilot study is carried out.

Luminescence dating requires the assessmentof many variables, and while many of these donot require specific site information, there aresome issues that do require this. Closecollaboration between the project director andthe luminescence specialist throughout theprocess will maximise the probability ofobtaining high-quality chronologicalinformation. It is likely that personnel from aluminescence laboratory will want to collecttheir own samples, and that they will want tomake a measurement of the gamma radiationdose rate in the field. At the same time theywill want to assess the water content relevantto the sample, and with the project directorconsider whether this is likely to have changedthrough the period being dated.

The sampling approach, the materials to besampled, the likely number of samples neededand details of how sampling is to beundertaken should be discussed with theluminescence specialist and the field team aspart of assessing the dating potential.

The results obtained from luminescence datingmust be archived. A large number ofmeasurements are undertaken as part of

obtaining a luminescence age, and manydifferent approaches may be taken dependingon the precise nature of the material beingdated and the methods used.The reportshould give this information in full, both toenable the excavator to judge the validity ofthe results, and for future researchers to assessthe results in the light of subsequentdevelopments in knowledge.The specificationsfor reports are laid out in section 9.

7 Sampling considerationsThe most important part of usingluminescence dating effectively is designing anappropriate sampling strategy.This will beconstrained by the suitability of materialavailable at the site, but ultimately it is essentialthat the strategy is driven by a clear idea ofthe archaeological question that is being asked.Where possible, luminescence dating shouldbe used in conjunction with otherchronological information: other datingmethods (eg radiocarbon), documentarysources and stylistic information from artefacts.Samples for luminescence dating can broadlybe classified into two groups: those includingsamples that were heated in antiquity andthose made up of sediments whose grainswere exposed to daylight during transport ordeposition.

Samples that were heated in antiquity, leadingto removal of trapped electrons in the mineralgrains, date the heating event.This includespottery, burnt flints and bricks. For all heatedsamples, a crucial issue is whether the mineralgrains were heated to a sufficiently hightemperature for their luminescence signal tobe reset. Generally, a sample must reach c300–400°C, for a duration dependent on itssize.

Samples from sediments whose grains wereexposed to daylight during transport ordeposition date the time of transport ordeposition.The most suitable sediments arethose that were exposed to the most daylightduring transportation, including wind-blownsands and silts.The optically stimulatedluminescence (OSL) signal from minerals ismore rapidly reset by daylight than is thethermoluminescence (TL) signal (Fig 8). OSLmeasurements have made it feasible to look atfluvial and colluvial sediments as well, but careneeds to be taken in these cases to assesswhether they were exposed to sufficientdaylight at deposition to reset theluminescence signal being measured.

For all samples, care should be taken to avoidexposure to any strong source of radioactivity.This includes X-rays (as from XRF corescanners or hand-held XRF devices), as these

will increase the trapped electron populationand make the sample appear older than itreally is. Fortunately, current evidence is thatX-ray systems used for security checks oncommercial air flights or in customs for postaldeliveries result in only a small dose. However,if the use of such devices has already occurred,or is likely, it should be discussed with alaboratory, as it may become significant for theanalysis of young samples.

Where feasible, it may be prudent to collectduplicate samples: one for analysis by theluminescence laboratory and one for archiving.This is particularly useful in rescue archaeology,where sites are destroyed, or where sites arelikely to become inaccessible due to building.Archived samples should be kept in cool, dryconditions, and their existence should benoted in the site report. If they are rich inorganic matter they may require cold storageto prevent degradation.

7.1 Sampling strategyA number of questions can be used to helpdesign a sampling strategy.The first decision iswhether luminescence dating is appropriate fora specific site:

� Are there materials at the site suitable forluminescence dating?

� Would a luminescence age for thesesamples be clearly related to thearchaeological event that is of interest?There may be tension between selecting thesample that is closest to the event ofinterest and the optimal site for sample

Fig 29 Selecting the optimal location for sampling isimportant.The best position is from the centre of the infill,where the material surrounding the sample ishomogeneous. One consideration is the complexity of thegamma dose rate environment – dose rate will havechanged through time near former ground surfaces, causingdifficulties in assessment.The rock at the base of this pit mayhave a radionuclide concentration different from itssurroundings, so calculating the gamma dose rate forsamples collected near it may also be difficult (redrawn fromAitken 1990).

collection (eg Fig 29).� Are the samples likely to be in the age

range of luminescence dating for thatmaterial? (see sections 5.3 and 5.4)

� Will the precision of a luminescence age(typically 5–10% at ±1σ) be sufficient toanswer the archaeological question anddistinguish between different possibleanswers? (see Table 1)

If it is concluded that luminescence might beapplicable, then the details of sampling shouldbe considered:

� How many samples are required to answerthe archaeological question(s)? Single agesfrom sites have limited value; multiple agedeterminations generally yield much moreinformation.

� Are there clearly defined stratigraphicand/or chronological relationships betweensamples? This may significantly enhance thevalue of the ages. For example, Bayesianmodelling of the luminescence ages may bepossible to refine the dating of the site.

The details of these decisions will vary fromone site to another, and a good samplingstrategy is probably best designed indiscussions between the project manager andthe luminescence specialist. Regional ScienceAdvisors and the English Heritage ScientificDating Team (see Appendix 1) also haveexperience that may be of great value in thisprocess. Some examples of different samplingstrategies are given in Part C.

7.2 Heated samplesPottery generally comprises a mixture of claysand coarser-grained materials used to temperthe pot. Providing that the pot was fired to asufficiently high temperature when it wasmanufactured, luminescence measurementscan be made on either silt particles or, morecommonly, sand-sized quartz inclusions in thetemper.The dose rate to the sample is derivedfrom both the pot itself and the surroundingsoil.

The outer 2mm of the pot is removed in thelaboratory so that the beta dose rateoriginates entirely from within the sample: thicksherds are needed for this, and mostlaboratories prefer ≥ 10mm thick and 30mmacross. Ideally, several sherds should becollected from each context, including a varietyof fabrics to provide the laboratory with arange of material to work with. Because thesurface is removed in the laboratory, the sherddoes not have to be protected from light; thusit can be washed and recorded in the normalway prior to submission. As with allluminescence samples, the best samples are

from at least 0.3m from any major change incontext, so that the gamma dose rate ishomogeneous (Fig 29).

A sample of the soil surrounding the sampleshould also be collected, for measuring thegamma dose rate and water content.The soilsample need not be kept in the dark, butshould be sealed in an airtight plastic bag assoon as it is excavated so that the watercontent is preserved. Sites should be avoidedwhere the gamma dose rate is likely to havechanged through time, such as near current orformer land surfaces.Where samples must becollected near a context boundary (egbetween an infill and the undisturbedsurroundings) in situ measurements of thegamma dose rate is essential (eg using aportable gamma spectrometer).

Burnt flints can also providethermoluminescence dates (Valladas 1992;Richter and Krbetschek 2006). Although flint isnot common, at sites where it has been usedfor tool making it provides good material fordating. However, flints can only be dated if theywere heated during antiquity, for examplewhen pieces accidentally fell into a fire or if theflint was deliberately heated to improve itsknapping properties. It is not always possible inthe field to identify whether flints have beenburnt, but surface features such as conchoidalfractures and colour may help to select piecesmost likely to have been burnt; therefore whensampling, it is advisable to collect as many flintsas possible.

Flint is a form of silica similar to quartz and hasexcellent luminescence properties: it gives abright TL signal but no OSL signal. Itsradioactivity is normally low, so the majorcontribution to the dose rate is the gammadose rate from the surrounding soil. As withpottery, it is therefore essential to remove theouter 2mm of a sample in the laboratory usinga diamond saw, to remove the betacontribution from the surrounding soil (Fig 30).Consequently, to get the required c 1g samplethe initial sample needs to be c 10g. Suchsamples are, unfortunately, quite rare.Additionally, it is important to assess whetherthe sample has been heated throughout itsmass to a sufficiently high temperature (c300–400°C) to remove any pre-existingtrapped electrons.Therefore for largersamples, the duration of heating would have tohave been longer.

Thermoluminescence measurements are stillroutinely undertaken on flints, dating the lastheating of the sample in antiquity. Additionally,by comparing the TL signal from the naturalsignal to that from an artificial irradiation, one

23

can measure a TL plateau (Fig 22). As forpottery, accurate assessment of the gammadose rate is essential – requiring carefulselection of the sample locations – from eitherin situ measurements or collection of a sampleof surrounding soil (Fig 29).

Fired clay bricks have been a commoncomponent in building construction over thelast two millennia, and potentially providesuitable material for luminescence dating. Aswith pottery, either silt-sized grains in the brickmatrix or coarser quartz sand inclusions fromthe temper can be analysed. More recent workhas focussed on quartz sand inclusions, usingOSL measurements, as they provide greatersensitivity than TL (Bailiff 2007). Samples canbe taken from a brick using a diamond-tippedcore drill. Bailiff (2007) used a 50mm diameterdrill, and sampled to depths of 100mm.Thisprovided an adequate length after discardingthe outer 1–2mm of the core affected by thedrilling.

Accurate assessment of the gamma dose rateis essential, as the gamma field may becomplex.This can done in situ using a gammaspectrometer or capsule dosimeter (seesection 4.3).

7.3 Sediments (unheated samples)In theory almost any sediment can be datedusing luminescence, but three major

Fig 30 Removing the outer surfaces of a burnt flint with adiamond saw – the inner core is crushed and the materialused to measure luminescence (diagram courtesy of Dr NDebenham).

constraints reduce the applicability.The first is the physical nature of the sediment.Luminescence analysis requires the separationof quartz (or sometimes feldspar) grains.Quartz is the most ubiquitous mineral at theEarth’s surface, but it is not common in someareas – for instance in areas where theunderlying geology is carbonate rich (eglimestone or chalk), or clay rich (eg slates).The second constraint relates to measuringthe sample dose rate. Luminescence analysis isundertaken either on silt grains (typically4–11μm diam) or on sand (90–300μmdiam).Therefore, clay or sediments dominatedby gravels and larger grain sizes are unlikely tobe suitable.

The third constraint is whether the mineralgrains were likely to have been exposed todaylight at the time of deposition, orimmediately before this.The most suitablesedimentary environments are wind-blownsands or silts (eg coastal dunes or loess). Inrecent years, measurement of the OSL signal,which is reset by exposure to daylight muchmore rapidly than TL (Fig 8), has made ispossible to date other sediment types. Fluvialsediments, and sediments derived by slopewash, may be suitable. However, in these casesit would be important to make replicatemeasurements on samples to assess the

reliability of the ages: the approach (see section3.3) relies on reducing the number of grains ineach aliquot so that variations in the De fromone sub-sample to another become visible,and thus it is only applicable to sand grains, asit is impossible to reduce the number of grainssufficiently and still obtain sufficient signals fromsilt.

Dates for the construction of standing stonesand other prehistoric monuments have beenattempted by measuring the luminescence ofthe last exposure of sediments underlying thestones or features (eg Rink and Bartoll 2005).Such an approach requires careful samplingfrom underneath the stones or structures, andit assumes that the ground surface wasbleached prior to construction.

When taking sediment samples it is importantnot to expose them to daylight.The bestmethod for achieving this depends on thenature of the sediment.The simplest approachis to use an opaque sample tube – plastic ormetal, sufficiently thick-walled to excludedaylight, and mechanically strong enough to bepushed or hammered into the section.The sizeof the sample required, and hence the tube,will vary depending on the nature of thesediment.Typical sample tubes are c 70mm indiameter and 0.2–0.3m long. It may be possible

to use smaller sample tubes, but this should bediscussed with the luminescence specialistbefore collection. Immediately before pushingthe tube into the face of the section, a fewmillimetres of sediment should be removedfrom the surface, as these grains will have beenexposed to daylight. After filling the tube withsediment, it should be wrapped in thick blackplastic. Any spaces left in the tube should bepacked with plastic bags to secure the samplein transit, sealed with tape and clearly labelledwith a black marker (red marking is invisible inluminescence laboratory lighting conditions).Where sediment is too hard or stones in itprevent the use of a tube it may be necessaryto excavate a sample shielded by a blacktarpaulin to exclude daylight (Fig 31).Thesampler can use a dim red light (eg a bicyclerear light) when collecting, then put the sampleinto an opaque black plastic bag. Black ‘buildersplastic’ made into bags with a heat sealer, orbags used for X-ray plates by hospitals aresuitable; but black bin liners are not sufficientlythick. Luminescence laboratories may be ableto supply black bags.

The volume of sample will depend on thenature of the sediment, but a good guide is500g. In some circumstances the material maybe sufficiently hard to carve out an intact blocksample using a trowel or similar implement.The block must be wrapped in black plastic,aluminium foil or other opaque material, andsupported to prevent it from disintegrating intransport to the luminescence laboratory, forthe outer surface of the block, which will havebeen exposed to daylight during its excavation,has to be removed.Thus the block must belarge enough – typically c 100mm in eachdirection – to allow for this.

When it is especially difficult to excludedaylight during sampling or a block sample isimpracticable, a last resort is to sample atnight, using a red light.

Whatever sampling method is used, the holecan be used for making in situ gammaspectrometry measurements.

A number of options exist for sampling bycoring below the ground surface.The bestsolution is to take samples using dedicatedopaque core liners. Such core sections can betaken directly to the laboratory, where theycan be sub-sampled under controlledlaboratory lighting conditions. If this procedureis not possible, sub-sampling can done on siteunder a black tarpaulin as soon as the corebarrel reaches the surface.

In all cases, samples must be taken formeasuring the water content. Sub-samples

24

Fig 31 Sampling fluvial gravels near Ripon,Yorkshire: a black plastic tarpaulin was used to exclude light and the sample wascollected in a thick black plastic bag (photograph by G A T Duller).

collected for luminescence analysis, andcompletely sealed, can be used, or separatesamples (c 20g) can be collected in air-tightplastic bags (which do not have to be keptdark).

7.4 Health and safetyWhatever type of sampling is beingundertaken, it is important to assess healthand safety risks. Such issues are unlikely to beunique to the sampler, but the difficulties ofsampling without exposing samples to daylightmay increase potential risks.

The greatest risks are probably thoseassociated with sampling a section.The stabilityof the section should be checked, and whereappropriate suitable shoring provided. A hardhat should be worn where appropriate.Thedangers associated with sections are especiallysignificant when using a tube to take a sample,as this will normally require hammering intothe section, possibly dislodging material fromthe section. Equally, where the sampler isrequired to be underneath a tarpaulin, he orshe will not be able to get away from thesection rapidly should it fail, or any part of itbecome dislodged.

8 Laboratory considerationsOnce submitted to a luminescence laboratory,a sample undergoes a complex sequence ofpreparation procedures and requires themeasurement of a wide range of parameters.Preparation is carried out under subdued redlighting, similar to a photographic laboratory,whose light intensity and wavelength does notdamage the trapped electron population. Anumber of chemical steps remove organicmatter and carbonates, a specific grain size isseparated out, and a specific mineral is isolated(Aitken 1985).The common approach forquartz uses density separation, which removesheavy minerals and some feldspars; andhydrofluoric acid, which etches away feldspars.These processes take time. Generally, resultstake six months, but this will depend on thelaboratory workload and can be confirmed atthe time of submission.

At the date of writing there is no formalaccreditation scheme for luminescencelaboratories or any recognised body to whichthey would be expected to belong; and noformal laboratory comparisons have beenmade. In the United Kingdom almost allluminescence laboratories are in universities orresearch institutes, and most are active inacademic research and publish inarchaeological or geographical journals.Appendix 2 is a list of laboratories and contactdetails.Their web sites have publication lists,including recent dating projects in which they

have been involved.9 Reporting specificationsThere are presently no specifications for thereporting of luminescence ages. A luminescenceage calculation requires two parameters: De anddose rate. In turn, each parameter requiresmany specific measurements, and a variety ofmethods may be used for each. Lists of ages, oreven summary De and dose rate values alone,provide insufficient information to assesswhether the ages are meaningful.

Laboratory reports should include sufficientinformation

� to document the work that has beenundertaken;

� enable the field project director or relevantperson to assess the reliability of the agesproduced and any difficulties that wereencountered;

� and provide an archive of the methods usedand the results obtained, so that the resultscan be assessed in the future in the light ofsubsequent developments in understandingof methods.

Reports from specialist luminescencelaboratories should include information underthe following headings.

9.1 Sample details

� A description of what was sampled and theprecise context of the sample, giving asmuch information as possible. For multiplesamples, stratigraphic or contextualrelationships should be made clear.This ispossible only if the luminescence specialistcollects the samples, or has preciseinformation from the excavator. Given theimportance of estimating the sample watercontent and understanding the radiationcontext, having the specialist on site is ofgreat value.

� For pottery and burnt flint, both the sampleand the surrounding sediment should bedescribed.

� A brief description of the collection methodand, where appropriate, details of samplestorage should be given.

� Photographs or detailed field drawingsshowing sampling locations should be made.

� The depth below the current groundsurface should be recorded for thecalculation of cosmic ray dose rate.

9.2 Luminescence measurements

� Describe the laboratory procedures used toprepare samples, indicating the grain size andmineralogy selected for luminescencemeasurements.

� Describe the equipment used to makeluminescence measurements, specifying thefilters used to detect the luminescencesignal, and, in the case of OSL dating, thewavelengths and power used for opticalstimulation.

� State the calibration of the artificial radiationsource used during measurements.Thisshould include the date of calibration andthe external source used (eg NationalPhysical Laboratory, or Risø NationalLaboratory), including details of the sourcesused (eg the 60Co gamma source used tocalibrate a 90Sr beta source). Include theuncertainty associated with the calibration(usually 2–3%), and state that thisuncertainty has been propagated into thetotal uncertainty on the luminescence age.

� State the method used for measuring De andthe nature of the aliquots used formeasurement (eg fine grains (4–11μm diam)settled on a 9.8mm aluminium disk, or c 100grains (180–211μm diam) on a 9.8mm steeldisk).

� Include graphs showing typical examples ofthe luminescence signal measured, anyunusual behaviours observed and anexample of a dose response curve,particularly for older samples that may beclose to signal saturation.

� State the mathematical method used forderiving De (eg interpolation, ormathematical fitting of a dose responsecurve) and the method used to calculatethe uncertainty on this De value (eg jackknifing, Monte Carlo methods, etc). Include astatement of the instrumental reproducibility(typically 1–2%) and how this has beenincluded in the calculation of De.

� Note where samples are close to the limitsof the method, either because they are veryyoung or very old, and discuss theimplications for the resulting ages.

� If multiple measurements of De have beenmade on each sample, state the number ofreplicates, including the number of aliquotsrejected during analysis.

� Show an example of the distribution of Devalues; if there are complex distributions,show them all.

� State method used to combine multiple Deestimates together in order to calculate theDe value used for age calculation (eg finitemixture model, central age model, etc.).

� Describe details of quality control checksundertaken, and their results, to ensure thevalidity of the method of De determination.For example, using the SAR protocol onquartz routine checks should include arecycling ratio measurement, a doserecovery test, a zero-dose (recuperation)check and the results of varying the preheat

25

temperature.9.3 Dose rate measurements

� Give a brief statement of the methods usedfor dose rate measurement, differentiatingclearly between in situ measurements andlaboratory based measurements.

� Give information about any outsourceddosimetry measurements (eg neutronactivation analysis, or ICP-MS), including thelaboratory used and the lower detectionlimit of the determination of each element.

� State whether an internal dose rate hasbeen included; if so, state its value and howthis was obtained. For quartz such a value isnormally small (< 0.05Gy/ka), but may belarger if other minerals are used formeasurement.

� State if any explicit measurements havebeen made to assess whether radioactivedisequilibrium is present.

� Include a table of results, including watercontent value, grain size and separateenumeration of the components making upthe total dose rate.

� Give a justification for the choice of watercontent used for dose rate calculation.

� Give a brief justification for the choice ofoverburden used for cosmic-ray dose ratecalculation, and any adjustments made forsite altitude and geomagnetic latitude.

� Where relevant, state how the alphaeffectiveness (a-value) was assessed.

9.4 Age calculation

� State the dose rate conversion factors andthe beta-dose attenuation factors employed.

� State the software (commercial or writtenby the laboratory itself) used to calculatethe age.

� Make it clear if errors quoted for the ageincludes the total uncertainty (bothsystematic and random).

� Round up or down ages to an appropriatelevel: for ages < 1000 years round the ageand its uncertainty to the nearest 5 years;for ages > 1000 years round to the nearest10 years (eg 12891 ±768a suggestsunfounded precision, and should be roundedto 12890±770a, which could also be written12.89 ±0.77ka).

� State the datum used for calculating ages,normally the year of measurement.

9.5 InterpretationThe detail in this section will be depend onthe relationship between the luminescencespecialist and the project.While manyluminescence ages will be straightforward,there will be occasions where unexpectedresults are found. Such issues are mosteffectively dealt with through a close

collaboration and discussion between theluminescence specialist and the projectdirector.Where possible, the outcomes ofthese discussions should be included in thefinal report so that the ages are put into theirappropriate contexts.

Where available, the results should becompared with other lines of evidence.

9.6 IndexingLuminescence reports should be indexed usingone or more of the preferred terms adoptedby the Forum on Information Standards inHeritage (FISH).These should describe the typeof analysis and material analysed (Tables 2 and3; NMR Archaeological Sciences Thesaurus), andthe archaeological object or site being dated(MDA Object Type Thesaurus; NMR Thesaurusof Monument Types; http://thesaurus.english-heritage.org.uk/thesaurus.asp?thes_no=1).Any chemical methods of analysis used todetermine the dose rate should also beindexed.

9.7 Non-technical summaryTo aid readers in their use of the luminescencereport it is helpful to provide a non-technicalsummary of the key findings of the study.

10 Quoting ages and disseminatingresultsThe following is a guide for quotingluminescence ages in reports and otherpublications.

Luminescence ages are calculated in yearsbefore the date of measurement.The unit ofmeasurement is the annum (abbreviation ‘a’).Thus an age of 110 ±10a measured in 1997 isdating the same event as one of 120 ±10ameasured in 2007.There is no agreedconvention about a datum year from whichages are measured, although some laboratoriesare starting to report ages in years before AD2000.Thus it is important to make clear whatdatum is being used – normally this is the yearof measurement.

The uncertainty in luminescence ages shouldalways be quoted, indicating whether itrepresents one sigma (1σ) (68% confidenceinterval) or 2σ (95% confidence interval).Luminescence laboratories commonly quoteerrors at 1σ, but it should be noted that thiserror normally includes both random andsystematic errors. All luminescencelaboratories will assign unique sample codes(eg Aber29/RG6, X2451, GL03008,Dur05OSLqi-301-1), and these should bequoted with the age (eg ‘…the sample gave aluminescence age of 130 ±10a (Aber29/RG6)and demonstrates…’).

Luminescence results presented as calendardates should be given as age ranges.Thus aluminescence age of 1000 ±50a measured in2000, could be expressed as a calendar rangeof AD 950–1050 (eg ‘…aeolian sand collectedfrom level 3 gave a luminescence date of AD950–1050 (X2451) …’). It is also correct toquote ages as calendar dates with theirassociated errors.Thus a luminescence age of955 ±12a measured in 2005 could be quotedas AD 1050 ±12.

Recording and disseminating the results fromluminescence analysis must be done. Reportsproduced by the specialist luminescencelaboratory form part of the final projectarchive, and a copy of them should be lodgedwith the local Historic Environment Record(HER).Their publication, either in a monographresulting from the project, or in a separateresearch paper is also encouraged so theresults have wider dissemination and undergopeer review.

Wherever luminescence ages are reported, itis crucial that information associated withgenerating those ages, is included, or at least isavailable.The report must include a brief

26

Table 2 Preferred terms used in theNational Monuments Record (NMR)Archaeological Sciences Thesaurus for typesof luminescence dating(http://thesaurus.english-heritage.org.uk/thesaurus.asp?thes_no=560).PDF versions of the thesaurus are availableon request from the Data Standards Unitof English Heritage Luminescence Dating

• thermoluminescence (TL)• optically stimulated luminescence (OSL)• infrared stimulated luminescence (IRSL)

Table 3 Preferred terms used in the NMRArchaeological Sciences Thesaurus for thematerials analysed (http://thesaurus.english-heritage.org.uk/thesaurus.asp?thes_no=144&thes%20name= MDA%20Object%20Type%20Thesaurus)

• brick• burnt flint• quartz• feldspar• zircon• polymineral• biogenic carbonate• pottery• geological sediment

description of the analytical methods used, thematerials analysed and summary tables, andgive a reference to the original laboratoryreport.Where possible, the form of theluminescence signal measured (TL glow curveor OSL decay curve), the dose responsecurve, and the distribution of De values for atypical sample should be shown in figures.Thisis essential for readers to judge the quality ofthe age estimates.

Publication should normally involve personnelfrom the luminescence laboratory, to presentthe technical data involved in themeasurements and to advise on interpretationof the results.

Summary tables should include:

� the laboratory code of the sample;� the water content used in calculations of the

dose rate;� either the elemental concentrations of U,Th

and K, or the alpha, beta and gamma doserates;

� the cosmic dose rate and the total doserate used for calculation of the age;

� De, and if appropriate the number ofreplicate measurements that were used inthe calculation of the final De;

� and the age of the sample.

Where possible, give the uncertainties in thesevalues (eg water content 7 ±3%, or beta doserate 1.44 ±0.07Gy/ka).Where some of thesevalues are the same for all samples, a footnotemay help to simplify the table.

When it is not possible to include all theprimary information, a reference should bemade to a report or some widely accessibledatabase where such information has beendeposited.The Archaeology Data Service(ADS) (ads.ahds.ac.uk) or a specialist journalsuch as Ancient TL (www.aber.ac.uk/ancient-tl)may be an appropriate place to do this (seefor example Barnett 1999).

27

28

Part CCase Studies

11 Wat’s Dyke, GobowenWat’s Dyke is an important feature in thehistoric landscape of the northern borderregion between Wales and England.Whenconstructed, the ditch was up to 8m wideand 4m deep and ran for c 65km fromsouth of Oswestry to the Dee Estuary.Thedyke is thought to have been constructedsometime in the post-Roman period, butthe exact date of its construction is poorlyknown and previous excavations have failedto uncover any significant artefacts to helpdetermine its age.

Close to its southern limit,Wat’s Dyke runsnear Gobowen on the Welsh border. Anarea given planning permission for a housingdevelopment by Fletcher Homes included aclearly visible part of Wat’s Dyke.ThereforeGifford Ltd undertook archaeological surveyin 2006 before construction started.Theprimary objectives were to record themorphology of the dyke at this location andto retrieve samples for absolute dating.

Two trenches were excavated, revealing anin-filled ditch c 2.5m deep and 8m wide.Near the ditch there was a bank c 0.5m highand 5m wide. Excavation of two trenchesacross the ditch revealed a sequence ofinfilling sediments, but these had limitedorganic material suitable for radiocarbondating (Fig 32).

The only sample directly associated with theditch and suitable for radiocarbon datingwas a charred twig found in the lowermostfill. It yielded a calibrated age of 1120–890cal BC (at 2 standard deviations; SUERC-12826; 2855 ±40 BP), significantly older thanthe assumed post-Roman age.

It was decided to obtain an independentdate using luminescence. A simple strategyfor sampling would have been to collect asample from as low down in the fill aspossible, close to the undisturbed site. Sucha sample would be difficult, if not impossible,to date using luminescence because there isoften rapid infilling immediately after diggingof this type of ditch, and the grains areunlikely to have been bleached.Thematerials into which the ditch was dug arefluvial sands and gravels which gave aluminescence age of 58.9 ±5.8ka (X2840).Moving away from the boundary and takingsamples from areas that are more likely tohave been deposited gradually, and to have

Fig 32 (above) Section drawing showing positions of the five samples, their laboratory codes and the ages obtained fromthem (range = 1σ or 68% confidence) (modified from Malim and Hayes, submitted). (below) Gamma spectrometrymeasurement in situ at Wat’s Dyke, Gobowen – five samples (holes in excavated section) were collected for luminescencedating (photograph by J-L Schwenninger).

Table 4 Summary of luminescence data for samples from Wat’s Dyke (from Malim and Hayessubmitted): De for each sample is the average of 12 determinations

lab code depth equivalent dose dose rate age (a) date (AD)

(m) (De) (Gy) (Gy/ka) (before AD 2007) (1σ range)X2833 2.10 2.73 ±0.19 2.47 ±0.15 1110 ±105 792–1002 X2834 1.90 2.79 ±0.10 2.37 ±0.15 1170 ±90 747–927 X2835 1.55 2.72 ±0.16 2.34 ±0.15 1160 ±105 742–952 X2836 0.95 1.58 ±0.06 2.32 ±0.17 680 ±60 1267–1387 X2837 0.46 1.50 ±0.06 2.44 ±0.26 610 ±70 1327–1467 X2838 2.30 2.89 ±0.06 2.33 ±0.14 1240 ±85 682–852 X2839 0.40 2.88 ±0.08 2.59 ±0.29 1110 ±130 767–1027 X2840 0.60 185.5 ±10.8 3.15 ±0.24 58900 ±5800

29

trench, five samples were collected at depthsfrom 2.10m to 0.46m. In all cases, the sampleswere from contexts where evidence indicatedslow infilling. Selection of samples from twotrenches made it possible to assess thereproducibility between sites. Detailedsampling at one trench made possible achronology for the infilling.

Samples for OSL dating were collected byhammering opaque black plastic tubes into theexcavated section.

Sampling was done by RLAHA personnel inone day, at the same time making in situgamma dose rate measurements with agamma spectrometer (Fig 32), and takingsamples for the measurement of sedimentwater content (which ranged from 11% to22%, increasing with depth).The water contentvalues were used to calculate the sample doserate.

In the laboratory, quartz sand grains(180–255μm) were extracted and aliquots

4–5mm diam (c 100 grains per aliquot) wereprepared.The OSL signal from these aliquotsdropped rapidly in the first few seconds ofmeasurement, then showed a slightly slowerdecrease from c 5s to c 20s (Fig 33a). Devalues were determined using the SARprocedure and the dose response curvesshow a smooth increase in OSL signal withdose. As normal, one of the regenerationdoses (in this case 2.5Gy) was given asecond time, to assess the ability of thesample to recycle a dose, giving a recyclingratio of 1.00 ±0.02, well within theacceptable range (Fig 33b).Twelve replicatemeasurements of De were made on eachsample.Those for X2836 are similar, andalmost all fall within the shaded band in Fig33c.

The ages for the two samples from near thebase of the ditch were 1240 ±85a (X2838;expressed as years before 2007) and 1110±105a (X2833) (Table 4); 1σ errors onthese ages give respective calendar dates ofAD 682–852 (X2838) and AD 792–1002

Fig 33 Luminescence data for a 4–5mm diam quartz aliquotfrom sample X2836,Wat’s Dyke, Gobowan: (a) OSL decaycurve for an aliquot measured using blue LEDs; (b) the doseresponse curve for the same aliquot, yielding a De of 1.45±0.07Gy – the regeneration dose at 2.5Gy was measuredtwice (repeated measurement shown in red) and therecycling ratio of 1.00 ±0.02 is within the acceptable range0.9–1.1; (c) radial plot of De values for 12 aliquots of sampleX2836 – only one De value falls outside the shaded band,suggesting that they all form a single population.

been bleached, would increase the accuracy ofthe luminescence age.

An additional means of confirming thatsamples were bleached before deposition is totake samples from contexts with clearstratigraphic relationships: ages consistent withthese relationships gives reassurance of theiraccuracy and allows a Bayesian analysis of theresults, giving additional information about theage model for the site.

At an early stage of excavation the projectmanager contacted the luminescencelaboratory at the Research Laboratory forArchaeology and the History of Art (RLAHA)in Oxford and, in collaboration with the EnglishHeritage Regional Science Advisor, theydevised a sampling strategy to collect eightsamples for luminescence dating: one from thefluvial sands and gravels of the underlying sitegeology and seven from the two trenches.

In one trench, a sample was collected at adepth of 2.30m and one at 0.40m; at the other

(a) (b)

(c)

30

(X2833).The dates of the sequence of fivesamples from one ditch show that withinthe uncertainties they agree with thestratigraphy, and thus document filling ofthe ditch between c AD 800 and the 14thcentury (Fig 32).This stratigraphicconsistency supports the notion that thesamples had been bleached beforedeposition. Further support for theluminescence chronology is provided byabraded sherds of a single wheel-thrown,lead-glazed vessel – discovered in anadjacent buried soil – dated to the 13th or14th century and thus contemporary withinfilling of the ditch by medieval ploughing.

The radiocarbon age of 1120–890 BCfrom the burnt twig recovered from thebottom of one trench cannot be easilyreconciled with the luminescence ages, andit seems likely that this was introduced intothe ditch fill from an earlier context whenWat’s Dyke was in use.

The luminescence ages imply that infillingoccurred in the early 9th century. Malimand Hayes (submitted) thereforepostulated that at this site Wat’s Dyke wasconstructed either in the reigns of Cenwulfand Ceolwulf (AD 723–796) or of Wiglafduring the 830s.

The excavation and analysis benefited fromthe enlightened approach of FletcherHomes to the work undertaken prior todevelopment of the site.

12 Dungeness Foreland, SussexDungeness Foreland is the third largestcoastal lowland in the United Kingdom, andhas a rich Bronze Age to modern dayarchaeological record.The absence ofarchaeological finds predating the Bronze Ageis consistent with the broad picture that hadbeen built up of the formation of this featureduring the last 4000 years. Drilling across theforeland showed that it comprised a series ofgravel ridges (Fig 34), with an underlying sandbody. Organic sediments had accumulated inthe swales between the gravel ridges.Theorientation of the gravel ridges, and earlierradiocarbon dating suggested that the featuregrew from the south, towards the current tipof the foreland. However, the detailedchronology and pattern of deposition of thesands and gravels making up the foreland wasunknown.

A project supported by an Aggregate LevySustainability Fund (ALSF) grant aimed to useOSL measurements to date the underlyingsands, and radiocarbon to date the overlyingorganic sediments, thus providing the age andpattern of formation of the underlying sands,and bracketing the timing of theemplacement of the gravel ridges.

A contractor drilled twelve vibracoreboreholes to recover sands from below thegravel ridges.Their sites were chosen to formtwo transects, running approximatelyeast–west and north–south.To providestratigraphic control, 37 sand samples werecollected: 3 samples from 11 of theboreholes, and 2 samples from the 12thborehole, plus 2 samples of the modernbeach sand to assess whether they werecompletely bleached.

As the majority of samples were collectedfrom several metres below the groundsurface, it was not possible to measuregamma dose rates using in situ gammaspectrometry. Instead, laboratorymeasurements were made based on sub-samples of the material collected for dating. Itwas likely that the gravels and the sandswould yield different gamma dose rates, so toensure that the laboratory-basedmeasurements gave accurate gamma dosevalues, sample locations were chosen thatwere at least 0.3m from the sand/gravelboundaries.

The water content of the samples was alsoconsidered. Most samples were collectedfrom below modern sea level, and wereexpected to be saturated (c 25% water

content). Roberts and Plater (2005) alsoconsidered the potential for changes in watercontent associated with changes in sea levelduring last 5,000 years, and generated a simplemodel for this.They concluded that the effectwould be negligible for all but two of theoldest samples.

The sand samples were collected in clearplastic liners (38mm diam), sealed in the fieldto prevent water loss and taken to theAberystwyth Luminescence ResearchLaboratory.

As the core liner was transparent, the outerpart of each core was exposed to daylightduring collection. Under red-light in thelaboratory a thin-walled, 20mm diam cylindricalsampling device was pushed up the length ofeach core liner to retrieve an inner sub-sample,unexposed to daylight. Quartz sand grains(150–80μm diam in this study) were isolatedand the OSL signals measured using blue LEDs(giving a power of 17 mW/cm2).

Aliquots consisting of c 200 grains weremeasured from the majority of samples; largeraliquots of c 800 grains were used for theyoungest samples, to increase the OSL signaland thus reduce the error on these samples(Roberts and Plater 2005).

Despite their young age, the sediments yieldedrelatively bright OSL signals (Fig 35); the signalsdropped rapidly during the first 5s of OSLmeasurement as would be expected forquartz.The SAR procedure was used toconstruct a dose response curve for eachaliquot and several tests were applied: arecycling test and a check to ensure that nofeldspar contamination was present wasapplied to each aliquot, and a preheat test wasundertaken for each sample (Fig 15). Onlythose aliquots that passed all of these testswere included for subsequent analysis.

A minimum of 21 aliquots were measured foreach sample, and between 11 and 44 aliquotspassed all the tests.The variation in the Devalues for each sample was analysed to seewhether there was any indication ofincomplete bleaching., but within each samplethe spread was small (Fig 36), implying that theluminescence signal in the sediments had beenbleached at deposition.

A sub-set of seven samples was also analysedusing medium aliquots (c 200 grains) in orderto assess whether any indication of incompletebleaching was being masked by the largeraliquots, but no evidence was found for this.

31

Finally, a dose recovery test was applied to thesamples.Together with the tests describedabove, favourable results gave confidence in theluminescence ages obtained.

The two samples collected from the modernbeach gave ages of 40 ±40a (Aber73/BH-USS)and 15 ±15a (Aber73/BH-SSR) demonstratingthat these beach sediments were completely

reset at deposition.The 35 OSL ages from thedrill cores ranged from 430 ±20a to 5120±220a (expressed as years before 2000)(Table 5). By collecting three samples fromeach core, the veracity of the ages could beconfirmed by observation of the stratigraphicconsistency (Fig 37). Several samples give agesthat are not in obvious order: eg in core 1, avalue of 4700 ±200a (Aber73/BH-1/1) overlies

ages of 4070 ±170 (Aber73/BH-1/2) and5120 ±220a (Aber73/BH-1/3). However,these results are consistent within twostandard deviations (ie 4700 – (2 x 200) =4300a and 4070 + (2 x 170) = 4410a) andsuggest that within the precision of theseages, typically 5%, the two samples are veryclose in age. Bayesian analysis of this suite ofdata revealed only one sample whose OSL

Fig 34 Gravel ridges across Dungeness Foreland, looking north-east (photograph by H M Roberts).

Fig 35 Luminescence data for a quartz aliquot from sampleAber73/BH1-1 from Dungeness Foreland: (a) OSL decaycurve using blue LEDs (17 mW/cm2); (b) dose responsecurve for the same aliquot, giving a De value of 3.9 ±0.3Gy;the regeneration dose at 1.85Gy was repeated and arecycling ratio of 1.06 ±0.04 calculated, within theacceptable range 0.9 to 1.1.

Fig 36 Radial plot of 31 replicate De values for Aber73/BH1-1: the majority of the data points lie within the shaded band,indicating that they are consistent with the average De

within the uncertainties of the individual De values.This isthe same data as shown in Fig 16a in the form of ahistogram.

(a) (b)

32

Fig 37 (top) Ages for the uppermost sands from boreholes drilled across Dungeness Foreland. (bottom) A west–easttransect (shown as a red line in the upper diagram) across Dungeness Foreland showing OSL ages (in years before AD2000) for the underlying sands (modified from Roberts and Plater 2007).

Table 5 Part (8 of 37 samples) summary of luminescence results from Dungeness foreland samples (Roberts and Plater 2005; 2007): dose rates adjustedfor water contents and calculated for grain size 150–180μm; ages expressed as years before AD 2000, rounded to nearest 10 years (or 10a)

lab code depth (m) % water equivalent no. of dose rate (Gy/ka) age (a)

content dose (Gy) aliquots beta gamma cosmic total

73BH-1/1 3.75 21 ±5 3.70 ±0.06 31 0.41 ±0.02 0.25 ±0.02 0.13 ±0.01 0.79 ±0.03 4700 ±200

73BH-1/2 5.95 24 ±5 3.11 ±0.05 33 0.39 ±0.02 0.28 ±0.02 0.10 ±0.01 0.77 ±0.03 4070 ±170

73BH-1/3 8.05 25 ±5 3.44 ±0.06 18 0.35 ±0.02 0.25 ±0.02 0.08 ±0.01 0.67 ±0.03 5120 ±220

73BH-2/1 11.25 25 ±5 4.31 ±0.09 18 0.49 ±0.03 0.37 ±0.03 0.06 ±0.01 0.92 ±0.04 4710 ±220

73BH-2/2 12.25 25 ±5 3.81 ±0.08 17 0.50 ±0.03 0.37 ±0.03 0.05 ±0.01 0.92 ±0.04 4170 ±190

73BH-2/3 13.60 25 ±5 3.55 ±0.09 12 0.51 ±0.03 0.29 ±0.02 0.05 ±0.01 0.84 ±0.03 4240 ±190

73BH-3/1 12.90 25 ±5 2.37 ±0.06 17 0.38 ±0.02 0.22 ±0.01 0.05 ±0.01 0.65 ±0.03 3630 ±160

73BH-3/2 13.9 25 ±5 2.42 ±0.06 18 0.36 ±0.02 0.23 ±0.02 0.05 ±0.01 0.63 ±0.03 3850 ±180

age was in poor agreement with the otherdata, and further investigation suggested thatthis was due to difficulties in calculating thedose rate.

A detailed, three-dimensional view of theevolution of this landscape could be madeby combining ages from both an east–westand a north–south transect (Long et al2006; Roberts and Plater 2007). Growth wasmostly to the south during the Bronze Age,but more easterly during the post-Romanera, then a switch to more northerlyaccretion since Saxon times.

To determine the stabilisation of the landsurface itself, radiocarbon samples werecollected from between gravel ridges.Although a smaller number of radiocarbonsamples were measured, they showed thatgravel deposition typically occurred within200–400 years of emplacement of theunderlying sand.Together these two datingmethods have provided a clear, three-dimensional model of landscapedevelopment that provides an importantframework for archaeological finds from thearea.

The sampling scheme was informed by priormodels of the development of thislandscape, and thus succeeded in maximisingthe information from the OSL ages.Takingthree samples from each core provided acheck on the reliability of the ages againstthe stratigraphic relationships. Luminescencewas ideal for dating the sands, but could notdate the gravels.Thus the combination ofluminescence dates with radiocarbon datesfor the overlying organic sediments gaveadditional information.

33

13 Fluvial gravels at Broom,DevonThe river terraces of the Axe River, nearBroom in Devon, have been known since the19th century to contain Palaeolithic tools.From 1932 to 1941 a distinguished amateurarchaeologist, Charles Bean, amassed morethan 900 implements from the gravelquarries actively mining the river terraces(Fig 38). In total more than 1800 handaxesare known from the site (Hosfield and Terry2001), making it one of the richestPalaeolithic sites in the United Kingdom(Hosfield et al in prep).

The sedimentary sequence at Broom canbroadly be subdivided into three parts:Lower Gravels, Middle Beds (sand, silt andclay), and Upper Gravels. Based on Bean’srecords of where artefacts were recoveredin the quarry, artefacts apparently occurthroughout this sequence.The majority ofthese handaxes (97%) are made of UpperGreensand chert; most of the remainder areflint.The age of these deposits was uncertain,only being indicated as Acheulian by thehandaxe typology and some pollen analyses.The limited archaeological evidence gave nofirm indication that the artefacts were in aprimary context, but the large proportion(81%) of tools with sharp or very sharpedges suggested that movement had beenlimited.Thus, although they were within asecondary context, an age for the depositionof the sediments would significantly improvechronological knowledge of the site and itstools. OSL dating was applied to the Broomsediments as part of The Archaeological

Potential of Secondary Contexts project,supported by the Aggregates LevySustainability Fund (ALSF).The projectillustrates some of the complexities in applyingOSL to such old materials.

Excavations since 2000 have yielded fewartefacts, so the major aim of the OSL datingwas to attempt to construct a chronology foraggradation of the fluvial gravels at the site.While all three parts of the sequence were ofinterest, the materials most suitable for OSLmeasurements were sands found in the UpperGravels and Middle Beds (Fig 38). Researchersfrom the Geochronology Laboratory,University of Gloucestershire collected ninesamples from various outcrops of these unitsat different locations in the area, using blackplastic pipes, or where this was not possible,carved out lithified blocks of material (c 75mmx 75mm x 50mm) and wrapped them inopaque plastic for transport to the laboratory.The outer surfaces of the blocks, which hadbeen exposed to daylight during sampling,were removed in the laboratory, and theresulting sub-sample prepared for OSLmeasurements.The innermost material wasselected for analysis from the samplescollected in opaque plastic tubes.

Quartz grains (c 0.1mm diam) were isolatedfrom most of the samples and the SARprocedure used to determine De for 12aliquots of each sample; each aliquot was c6mg of quartz (4,000 to 10,000 grains). Insteadof LEDs, a filtered halogen light system wasused to provide optical stimulation (16mW/cm2 at 420–560nm).The OSL signals

obtained were generally bright (Fig 39a); thesignal dropped rapidly during measurement,as expected for quartz.

Gamma dose rate measurements weremade in situ using a gamma spectrometer,complemented by measurements of K, U andTh concentrations by neutron activationanalysis (NAA) and ICP-MS in thelaboratory. Current water content for eachsample (between 7% and 29%) was used fordose rate calculation on the basis that sincedeposition they would have experienced awide range of climatic conditions.The doserate calculated for different samples variedfrom 1.08Gy/ka to 1.72Gy/ka (Toms et al2005).

The samples analysed from the UpperGravels and the Middle Beds gave De valuesbetween c 300 Gy and 550Gy, and exhibitedOSL signals showing growth to more than600Gy (Fig 39b).The natural OSL signals(LN/TN) intersected with this doseresponse curve, yielding finite De values, butthe natural signals were close to thesaturation level of the material (cf Fig 27a).

These results are a good example of thechallenges encountered when applyingluminescence dating to old samples (seesection 5.3).The samples at Broom yieldedfinite De values (and hence ages), but theirreliability may be questionable when they areclose to saturation. Small variations inexperimental uncertainties or systematiceffects will have a large impact on the Devalues. For the sample dose response curve

Fig 38 (left) Handaxe from the fluvial gravels at Broom,Devon. (right) Collecting samples at the boundarybetween the sand-rich Middle Beds and the UpperGravels for OSL dating (© Dr R Hosfield).

34

shown in Fig 39b, eleven replicatemeasurements yielded similar De values (Fig39c), but other samples show more variationbetween aliquots. Such variation can becaused by incomplete bleaching or variationsin environmental dose rate between grains;or by variations in the behaviour of grains asthey approach saturation.These possibilitiesare discussed by Toms et al (2005). Becausea large number of samples had beencollected for OSL analysis it was possible tocheck for consistency between results, and toattempt a Bayesian analysis, which impliedages ranges of 282–324ka for the MiddleBeds and 205–292ka for the Upper Gravels(Toms et al 2005).

Dating sites such as Broom is important, andOSL provides one of the few techniquesapplicable. However, interpretation of theages from this and other sites that lie nearthe limit of current luminescence datingmethods is complex. Publication of the agesfrom these sites requires the inclusion ofdetailed supporting luminescence data, sothat their validity can be assessed now and inthe future.

Fig 39 (a) OSL decay curve; (b) a dose response curvefor a quartz aliquot from sample GL03008, from theMiddle Beds (modified from Toms et al 2005); (c) radialplot of eleven replicate De measurements for GL03008 –the shaded bar shows the average De (353 ±19Gy) andpoints falling within this bar are within 2σ of this value.

(b)

(c)

(a)

35

14 Dating medieval bricksOnly in recent years has the potential ofluminescence dating been realised forincreasing our knowledge of the phasing ofconstruction of brick-built structures.Toevaluate and demonstrate this potential,samples of bricks were collected fromseveral buildings in England whose ages werewell known from documentary evidence ordendrochronology: Boston Guildhall,Tattershall Castle, Ayscoughee Hall,Doddington Hall and Fydell House.Theirages range from AD 1390 to AD 1737.Samples were collected by drilling intoselected bricks in situ to extract solid coresof material (Fig 40).The cores were wrappedin opaque plastic and taken to thelaboratory for sub-sampling under laboratorysafe-lights.

Measurements of gamma dose rates weremade in situ using two methods: using a NaIgamma spectrometer and by placingaluminium oxide dosimeters into the drillholes.The gamma dose rate measurementswere used as the basis for the gamma doserate.The beta dose for each sample wasmeasured in the laboratory using atechnique known as beta TLD Fig 40 Using a diamond tipped drill barrel to sample a brick for OSL dating.

Fig 41 (a) OSL decay curve for quartz grains separated from a brick from Fydell House,Boston, Lincolnshire (sample 311-6; Bailiff 2007); (b) De values from sample 311-6 as afunction of preheat temperature – blue points = data from individual aliquots, red pointsand bars = means and standard deviations; (c) De values (blue) means and standarddeviations (red) from sample 315-5 from Clarendon House,Wiltshire.This sample has adimmer OSL signal, which may contribute to the increased scatter observed in the De

values.

(b)(a)

(c)

36

(thermoluminescence dosimetry), whichuses a sensitive luminescence phosphor tomeasure the beta activity of the sampledirectly.The advantage for this study is thatthe beta dose will originate from entirelywithin the specific brick being sampled. Forthese samples the beta dose contributes c64% of the total dose (Table 6).

Further emission counting methods were used to measure the alpha activity(normally assumed to be zero) of thequartz grains used for luminescencemeasurements.The measurements heresuggested that the alpha activity was c 1%of the total dose rate (between 2.8 Gy/kaand 4.4Gy/ka). However, a small allowancewas made for this by adding 0.03±0.02Gy/ka to the dose rate calculated for each sample.

The cosmic dose rate was calculated using theposition of the bricks within each structure asa guide to their likely exposure.The watercontent of the bricks was less than 2% for allsamples except one, whose value was 3–4%.For calculation of the average dose rate, awater content of 3 ±1% was used for allsamples except for the one whose value washigher; for that sample 5 ±1% was used.

Coarse grains of quartz (c 0.1mm diam)extracted from the bricks were used for OSLmeasurements. OSL signals from aliquotscontaining 1–2mg of the quartz was measuredusing optical stimulation provided from eitherblue LEDs (c 50 mW/cm2) or a filteredhalogen light (c 30 mW/cm2).The brightness ofthe OSL signal varied substantially between the

different samples, and the dimmest sample hada signal to background ratio close to 1.0.These OSL signals were characterised by a rapid drop at the beginning of themeasurement, as expected for quartz (Fig 41a). Aliquots that had large statisticalfluctuations in their OSL signals afterbackground subtraction were excluded fromfurther analysis.

The samples were measured using a SARprocedure, giving between 6 and 16 De valuesfor each brick that could be combined for usein the OSL age calculation (Table 6).Theselection of an appropriate preheattemperature was done by measuring De at avariety of preheats for each sample. For somesamples the De values were reproducible,while more scatter was observed for others

Table 7 Comparison of luminescence ages from OSL measurements of quartz from bricks in medieval buildings: luminescence ages from Table6 given as years before AD 2005 and as calendar dates for comparison with independent assessment of the ages from archival and othersources (see Bailiff 2007 for details)

lab code building luminescence age (a) calendar date independent assessment(Dur05 OSLqi) or assigned date range

301-1 Alford Manor House, Alford, Lincs 450 ±27 1555 ±27 1611–1615310-1 St Mary’s Guildhall, Boston, Lincs 617 ±37 1388 ±37 1390–1395311-2 Fydell House, Boston, Lincs 278 ±17 1727 ±17 1700–1726311-4 Fydell House, Boston, Lincs 296 ±20 1709 ±20 1700–1726311-6 Fydell House, Boston, Lincs 284 ±17 1721 ±17 1724–1726315-4 Clarendon House,Wiltshire 330 ±18 1688 ±18 1667–1690315-5 Clarendon House,Wiltshire 273 ±18 1730 ±18 1717–1737317-1a Doddington Hall, Doddington, Lincs 419 ±24 1586 ±24 1593–1600317-1b Doddington Hall, Doddington, Lincs 427 ±27 1576 ±27 1593–1600318-1 Tattershall Castle,Tattershal, Lincs 550 ±33 1455 ±33 1445–1450318-2 Tattershall Castle,Tattershal, Lincs 552 ±34 1453 ±34 1445–1450319-1 Ayscoughfee Hall, Spalding, Lincs 558 ±32 1447 ±32 1450–1455

Table 6 Summary of luminescence data for brick samples from medieval buildings (see Table 7): ages expressed as years before AD 2005;uncertainties include random and systematic sources to one standard deviation.

laboratory preheat equivalent no. of dose rate (Gy/ka) age (a)(Dur05 OSLqi) temp (°C) dose (Gy) aliquots beta gamma total

301-1 220–250 1.27 ±0.02 16 1.89 0.93 2.82 ±0.06 450 ±27310-1 220–240 2.70 ±0.03 10 2.99 1.38 4.37 ±0.10 617 ±37311-2 220–250 1.05 ±0.02 11 2.48 1.31 3.79 ±0.08 278 ±17311-4 210–240 1.15 ±0.04 14 2.58 1.31 3.89 ±0.09 296 ±20311-6 220–240 0.97 ±0.01 14 1.96 1.26 3.42 ±0.12 284 ±17315-4 210–230 1.12 ±0.02 6 1.98 1.41 3.39 ±0.07 330 ±18315-5 200–240 0.95 ±0.03 8 2.11 1.35 3.46 ±0.08 273 ±18317-1a 200–240 1.36 ±0.01 8 2.05 1.19 3.24 ±0.07 419 ±24317-1b 210–230 1.56 ±0.04 9 2.42 1.21 3.64 ±0.08 427 ±27318-1 200–240 1.79 ±0.02 11 2.09 1.16 3.25 ±0.07 550 ±33318-2 200–240 1.84 ±0.03 8 2.26 1.07 3.33 ±0.08 552 ±34319-1 200–240 1.93 ±0.02 7 2.14 1.32 3.46 ±0.08 558 ±32

37

(Fig 41b and c).The comparison between the OSL ages andthe assigned dates was extremely good (Table 7 and Fig 42), and the typical error onthe OSL ages was only 25 years, even allowingfor systematic errors. Further analysis of bricksfrom another site, Alford Manor (301-1), gaveOSL dates some 55 years earlier than theassigned ages, and subsequent structuralanalysis showed that the part of the buildingthat had been sampled had been rebuilt,possibly re-using bricks from an olderstructure.

Although the use of OSL for dating bricks hasnot been common in England, the potential ofthe method is clear, both for dating structureswhose age is unknown, and potentially fordetecting areas that have been rebuilt, or

Fig 42 Comparison of ages obtained using OSL dating of bricks and dates assigned to parts of the built structures basedon documentary and dendrochronological evidence (Bailiff 2007).

Assigned date (AD)

1300 1400 1500 1600 1700 1800

Lum

ines

cenc

e da

te (

AD

)

1300

1400

1500

1600

1700

1800

SummaryLuminescence is the light emitted by mineralsafter they have been exposed to radioactivity. Itprovides a method for dating materials thatwere either heated in the past (pottery, burntflint and bricks) or that were exposed todaylight (geological sediments).The event beingdated is the last time the sample was heated orexposed to daylight.The method is applied tothe mineral grains within a sample, and themajority of analyses are undertaken on quartz.The method is based on the observation thatexposure of a sample to radioactivity increasesits luminescence signal and provides the basisfor the chronometer. Radioactivity is ubiquitousin the natural environment, originating fromuranium, thorium and potassium surroundingthe sample, and from cosmic rays that originatefrom beyond the Earth.

Luminescence measurements are used tocalculate the total radiation dose to which thesample has been exposed since the event beingdated.This quantity is known as the equivalentdose (De), and is measured in the units Gray(Gy).

The rate at which a sample is exposed toradioactivity in its natural environment can bemeasured either by chemical methods or bydirectly measuring the emission of radioactivity.This is termed the dose rate and has the unitsGray per year (Gy/year).The age of the sampleis calculated by dividing the equivalent dose bythe dose rate.

Two types of luminescence measurement canbe made in the laboratory: heating the sampleresults in thermoluminescence (TL); orstimulating the sample using light of a limitedwavelength, which results in the emission ofOptically Stimulated Luminescence (OSL).TL isused to date materials when the event beingdated is the last time they were heated. OSL isthe appropriate technique when dating the lastexposure to daylight. OSL can also be used forheated materials.

Unlike radiocarbon ages, luminescence ages donot require calibration.The unit annum(abbreviated to ‘a’) should be used instead of‘years’, or ka (thousands of years) whereappropriate.The term BP (before present)should never be used for a luminescence age; itshould only be used for radiocarbon ages.Theage range over which luminescence datingworks varies from sample to sample dependingupon the nature of the quartz and the doserate specific to that sample. In idealcircumstances the method works on samplesfrom as young as a few decades to as old asseveral hundred thousand years.

The precision of luminescence ages is typically5–10%; thus ages of 50 ±5a, 5,000 ±500a and50 ±5ka all have uncertainties of 10%.Thelargest single source of uncertainty is thewater content of the samples, which affects thedose rate.

Calculating luminescence ages involvesmeasurement of many parameters, andreports from luminescence laboratories shouldinclude all of this data and be included in theproject archive.When reporting dates inmonographs or academic papers it is essentialto incorporate sufficient data to enablereaders to judge the quality of the age.Particular problems occur for very youngsamples (< 1000 years) where lowluminescence signal levels may degrade theprecision, and for very old samples (> 50,000years) where saturation of the luminescencesignal limits the application of the method andintroduces larger errors as the sensitivity ofthe technique is reduced.When dating the lastexposure of materials to daylight additionalcomplications can arise if the sample was notexposed to daylight, for a sufficient period oftime to completely remove any pre-existingsignal.

Replicate measurements of De from a singlesample can be used to test whether thisoccurred, by looking to see whether the Devalues are consistent from one sub-sample toanother. Statistical methods are availabledesigned to calculate the age in situationswhere the signal was not completely removedat deposition.

A minimum requirement when reporting datesis to include diagrams illustrating theluminescence signal measured and the growthof the luminescence signal with radiation dose.Where there are concerns about whether thesample was exposed to sufficient daylight atdeposition to reset the signal the results ofreplicate measurements of De should beshown. A table of the results should also beincluded and at a minimum this should list thelaboratory code, the material analysed, doserate, De, the number of replicatemeasurements used to calculate De and thecalculated age. Luminescence ages arecalculated in years before the date ofmeasurement, and the date of measurementneeds to be stated.The text associated withthe table and the figures should also include abrief description of the analytical methodsused for De and dose rate measurements, anda description of how the water content wasassessed.

Given the complex nature of luminescencedating it is generally recommended that

publication is undertaken jointly with theluminescence laboratory involved with themeasurements so that they can provide expertinput into the presentation of the data.

In text, luminescence ages should be quotedwith their unique laboratory code, and thedate listed with its associated uncertainty. Agesmay be expressed in years before some datum(eg AD 2007), or as calendar dates BC or AD.

Luminescence provides a powerful techniquethat complements other dating methods.Given the complexity of the method it isrecommended that the luminescencelaboratory be consulted early in the projectplanning stage to provide advice about theoptimal sampling conditions and to assist indesigning a sampling strategy.

Further readingThe fundamentals of luminescence dating werespelt out in detail by Aitken (1985), and a briefversion was included in his later book (Aitken1990). Subsequently Roberts (1997) provideda comprehensive review of the use ofluminescence in archaeological applicationsworldwide.While these sources remainexcellent for discussions relating to the doserate and the use of thermoluminescence, theypredate the major developments in the use ofthe optically stimulated luminescence (OSL)signal from quartz, and the SAR procedurenow used. More recent reviews by Lian andRoberts (2006) and Duller (2004) focus ondevelopments in the use of OSL for estimationof De , and Jacobs and Roberts (2007) providean overview of the current state-of-the-art inapplying single-grain methods.Walker (2005)gives a broader summary of the use ofluminescence in Quaternary sciences.Wintle(2008) looks at the history of thedevelopment of luminescence dating withparticular attention to archaeological

38

applications.ReferencesAdamiec, G and Aitken, M J 1998 ‘Dose-rateconversion factors: an update’. Ancient TL 16,37–50

Aitken, M J 1985 Thermoluminescence Dating.London:Academic Press

Aitken, M J 1990 Science-based Dating inArchaeology. London: Longman

Bailey, S D,Wintle,A G, Duller, G A T and Bristow,C S 2001 ‘Sand deposition during the lastmillennium at Aberffraw,Anglesey, North Wales asdetermined by OSL dating of quartz’. QuaternarySci Rev 20, 701–4

Bailiff, I K 2007 ‘Methodological developments inthe luminescence dating of brick from English late-medieval and post-medieval buildings’.Archaeometry 49, 827–51

Barnett, S M 1999 ‘Date List 6: luminescencedates for Late Bronze Age and Iron Age potteryassemblages in eastern and northern England’.Ancient TL 17, 23-40

Barnett, S M 2000 ‘Luminescence dating ofpottery from later prehistoric Britain’.Archaeometry 42, 431–57

David, B, Roberts, R G, Magee, J, Mialanes, J,Turney,C, Bird, M,White, C, Fifield, L K and Tibby, J 2007‘Sediment mixing at Nonda Rock: investigations ofstratigraphic integrity at an early archaeologicalsite in northern Australia and implications for thehuman colonisation of the continent’. J QuaternarySci 22, 449–79

Duller, G A T 2004 ‘Luminescence dating ofQuaternary sediments: recent advances’. JQuaternary Sci 19, 183–92

Galbraith, R F 2005 Statistics for Fission TrackAnalysis. Boca Raton: Interdisciplinary Statistics,Chapman and Hall/CRC

Galbraith, R F, Roberts, R G, Laslett, G M,Yoshida,H and Olley, J M 1999 ‘Optical dating of single andmultiple grains of quartz from Jinmium rockshelter, northern Australia: Part I, Experimentaldesign and statistical models’. Archaeometry 41,339–64

Godfrey-Smith, D I, Huntley, D J and Chen,W H1988 ‘Optical dating studies of quartz and feldsparsediment extracts’. Quaternary Sci Rev 7, 373–80

Hashimoto,T, Koyanagi,A,Yokosaka, K, Hayashi,Yand Sotobayashi,T 1986 ‘Thermoluminescencecolor images from quartzs of beach sands’.Geochemical J 20, 111–18

Hosfield, R T, Green, C P and Chambers, J C (eds)in prep The Lower Palaeolithic Site at Broom,England. Oxford: Oxbow Books

Hosfield R T and Terry, R 2001 ‘Broom Palaeolithicsites’. Quaternary Newsletter 95, 1–5

Huntley, D J, Godfrey-Smith, D I and Thewalt, M LW 1985 ‘Optical dating of sediments’. Nature 313,105–7

Huntley, D J, Hutton, J T and Prescott, J R 1993‘The stranded beach-dune sequence of south-eastSouth Australia:A test of thermoluminescencedating, 0–800ka’. Quaternary Sci Rev 12, 1–20

Jacobs, Z and Roberts, R G 2007 ‘Advances inoptically stimulated luminescence (OSL) dating ofindividual grains of quartz from archaeologicaldeposits’. Evolutionary Anthropol 16, 210–23

Jacobs, Z,Wintle,A G, Duller, G A T, Roberts, R Gand Wadley, L 2008 ‘New ages for the post-Howiesons Poort, late and final Middle Stone Ageat Sibudu, South Africa’. J Archaeol Sci 35,1790–807

Lee, E 2006 Management of Research Projects inthe Historic Environment – The MoRPHE ProjectManagers Guide. Swindon: English Heritage

Lian, O B and Roberts, R G 2006 ‘Dating theQuaternary: progress in luminescence dating ofsediments’. Quaternary Sci Rev 25, 2449–68

Long,A J,Waller, M P and Plater,A J 2006 ‘Coastalresilience and late Holocene tidal inlet history: theevolution of Dungeness Foreland and theRomney Marsh depositional complex (UK)’.Geomorphology 82, 309–30

Malim,T and Hayes, L submitted ‘The date andnature of Wat’s Dyke: a reassessment in the lightof recent investigations at Gobowen, Shropshire’.Anglo-Saxon Studies in Archaeology

Murray,A S and Olley, J M 2002 ‘Precision andaccuracy in the optically stimulated luminescencedating of sedimentary quartz: a status review’.Geochronometria 21, 1–16

Murray,A S and Wintle,A G 2000 ‘Luminescencedating of quartz using an improved single-aliquotregenerative-dose protocol’. RadiationMeasurements 32, 57–73

Olley, J, Caitcheon, G and Murray,A S 1998 ‘Thedistribution of apparent dose as determined byoptically stimulated luminescence in small aliquotsof fluvial quartz: Implications for dating youngsediments’. Quaternary Sci Rev 17, 1033–40

Olley, J, Murray,A and Roberts, R G 1996 ‘The

effects of disequilibria in the uranium and thoriumdecay chains on burial dose rates in fluvialsediments’. Quaternary Sci Rev 15, 751–60

Richter, D and Krbetschek, M R 2006 ‘A newthermoluminescence dating technique for heatedflint’. Archaeometry 48, 695–705

Rink,W J and Bartoll, J 2005 ‘Dating the geometricNasca lines in the Peruvian desert’. Antiquity 79,390–401

Roberts, R G 1997 ‘Luminescence dating inarchaeology: from origins to optical’. RadiationMeasurements 27, 819–92

Roberts, H M and Plater,A J 2005 Opticallystimulated luminescence (OSL) dating of sandsunderlying the gravel beach ridges of Dungeness andCamber, Southeast England, UK. Centre forArchaeology Rep 27/2005. Portsmouth: EnglishHeritage

Roberts, H M and Plater,A J 2007 ‘Reconstructionof Holocene foreland progradation using opticallystimulated luminescence (OSL) dating: anexample from Dungeness, UK’. The Holocene 17,495–505

Roberts, R G, Jones, R and Smith, M A 1994‘Beyond the radiocarbon barrier in Australianprehistory.’ Antiquity 68, 611–16

Sommerville,A A, Hansom, J D, Sanderson, D CW and Housley, R A 2003 ’Optically stimulatedluminescence dating of large storm events innorthern Scotland’. Quaternary Sci Rev 22,1085–92

Stokes, S and Fattahi, M 2003 ‘Red emissionluminescence from quartz and feldspar for datingapplications: an overview’. Radiation Measurements37, 383–95

Toms P S, Hosfield R T, Chambers J C, Green C Pand Marshall, P 2005 Optical dating of the BroomPalaeolithic sites, Devon and Dorset. Centre forArchaeology Rep 16/2005. Portmouth: EnglishHeritage

Valladas, H 1992 ‘Thermoluminescence dating offlint’. Quaternary Sci Rev 11, 1–5

Walker M J C 2005 Quaternary Dating Methods.Chichester:Wiley

Wintle,A G 2008 ‘Fifty years of luminescencedating’. Archaeometry 50, 276–312

Wintle,A G and Murray,A S 2006 ‘A review ofquartz optically stimulated luminescencecharacteristics and their relevance in single-aliquotregeneration dating protocols’. Radiation

39

Measurements 41, 369–91

Glossaryaliquot a sub-sample of the material whoseluminescence signal is being measured.Typically,aliquots of sand-sized grains are 1–5mg of thesample. For fine-grain (0.004–0.011mm diam)samples an aliquot will weigh 1mg or less.Thegrains of quartz, feldspar or other mineralbeing measured are mounted on a steel oraluminium disc c 9.8mm diameter.

alpha particle one type of radiation emittedfrom atoms when they undergo radioactivedecay; an alpha particle is a helium nucleus andconsists of two neutrons and two protons thatcan travel up to 20μm.

anomalous fading a luminescence signal that isanomalously unstable compared withtheoretical predictions.This affects feldspars,and if unaccounted for would lead toluminescence ages that are too young.

beta particle one type of radiation emittedfrom atoms when they undergo radioactivedecay; a beta particle is an electron.They cantravel up to 2mm.

bleaching exposing a sample either to daylightor to an artificial light in order to remove thetrapped electron population.

conduction band a conceptual part of acrystal where electrons can move freely fromone part of the crystal to another. In mostminerals electrons need additional energy (egfrom heating or exposure to light) to reachthe conduction band, and electrons can onlyexist in the conduction band for a limitedperiod of time.

cosmic rays a type of high energy, penetratingradiation made up of charged particles thatoriginate beyond the Earth; can be deflectedby the Earth’s magnetic field.

dose rate (also known as the effective doserate) the total radiation dose to which asample has been exposed in a given period oftime; normally expressed as the radiation doseper year (Gy/a) or per thousand years (Gy/ka).

dose response curve the increase in aluminescence signal as a function of theradiation dose it has absorbed beforemeasurement.

equivalent dose the laboratory estimate of theradiation dose that the sample received duringburial; commonly, and preferably, abbreviatedDe; other terms sometimes used include

palaeodose (P).

gamma ray one type of radiation emitted fromatoms when they undergo radioactive decay; atype of electromagnetic radiation that cantravel up to 300mm.

glow curve the thermoluminescence signalfrom a sample as it is heated from roomtemperature to typical temperatures of c400°C or 500°C.

Gray (Gy) the SI unit used for absorbedradiation dose: one Gray is equivalent to onejoule of energy being deposited in eachkilogramme of a sample.

growth curve see ‘dose response curve’ (thepreferred term).

incomplete bleaching term applied tosediments that did not receive sufficientexposure to daylight before deposition toreset the trapped electron population, makingit more complex to date, and requiring manyreplicate De measurements and use ofstatistical methods.

infrared stimulated luminescence (IRSL) theluminescence signal emitted when a mineral isexposed to infrared radiation (typically c880nm, beyond the visible range); used fordating feldspars.

isotope one form of an element: differentisotopes of an element have the samechemical properties, but different atomicmasses due to different numbers of neutrons(eg 12C, 13C and 14C, or 235U and 238U, or 39K and40K – the number beside the element denotesthe atom’s mass).Thus, an atom of 12C has anatomic mass of 12, while 14C is heavier with amass of 14. Unstable isotopes – such as 14C,235U, 238U and 40K – emit radiation as theytransform into a different isotope or element.

luminescence a phenomenon exhibited bymany naturally occurring materials and used asthe basis for dating. Luminescence is lightemitted by some minerals when thermally oroptically stimulated following exposure toionizing radiation.The light is normally weak,invisible to the naked eye, but detectable in thelaboratory using a photomultiplier tube.

optical dating dating using the OSL signal.

optically stimulated luminescence (OSL) thelight emitted from a sample when it isstimulated by exposure to light.

photomultiplier tube (PMT) a highly sensitivedevice for measuring very small amounts of

light (eg luminescence).

photoluminescence (PL) another term foroptically stimulated luminescence.

preheat (cutheat) heating a sample prior tomeasuring its luminescence in order to removeunstable signals; samples are typically heated inthe range 150°C to 300°C.

radial plot a type of graph commonly used todisplay multiple De estimates, determined for asingle sample; each De value is shown as aseparate point together with the uncertaintyassociated with that measurement, enablingvisual differentiation between those pointsknown precisely and those known less well.

radioactivity the spontaneous disintegration ofatoms by emission of matter and energy,including alpha and beta particles, and gammarays.

recycling ratio or recycling test one of arange of tests commonly made as part of aSAR sequence. Recycling tests involve applyingthe same regeneration dose in two differentcycles, and giving the ratio of the sensitivitycorrected luminescence signals from the twomeasurements – recycling ratio. If the SARprotocol and the sensitivity correction areworking appropriately, then the recycling ratioshould be 1.Values from 0.9 to 1.1 areconsidered acceptable; values outside theselimits indicate that the SAR protocol is notworking as expected and that the data for thataliquot should be discarded.

single aliquot regenerative dose (SAR)protocol a sequence of laboratory operationscommonly used to measure De; normally usedon quartz, a main advantage of which is thatby using a test dose, any changes in theluminescence sensitivity of the sample areexplicitly monitored and corrected.

test dose a radiation dose, given to an aliquotin the laboratory, and used to monitorwhether the aliquot is undergoing changes inits luminescence properties during thesequence of measurements that make up aSAR protocol.

thermal transfer the movement of electronsfrom one trap to another in a crystal byheating; may cause problems when dating veryyoung sediments, so a preheat test should beused to determine the preheat temperature tominimize the problem.

thermoluminescence (TL) the light emittedwhen a crystalline material, which haspreviously been exposed to radioactivity, is

40

heated. It results from the release of energystored within the crystal and is different fromthe incandescence (black body radiation) thatis observed at higher temperatures (especially> 500°C). Incandescence is observed byheating the sample a second time, and issubtracted from the first TL measurement.

trap or trapping centre a site in a crystalwhere electrons can become lodged andremain stored for some period of time(varying from fractions of a second to millionsof years depending on the nature of the trap).Some traps are formed by chemical impuritiesin the crystal; others may relate to structuraldefects. Stimulating the crystal by heating it orexposing it to light can eject electrons from

41

42

traps into the conduction band.

Appendix 1 Sources of advice onScientific Dating from EnglishHeritageWithin English Heritage the first point ofcontact for general archaeological scienceenquiries, including those relating toluminescence dating, should be the EnglishHeritage regional science advisors, who canprovide independent, non-commercial advice.Such advisors are based either in universitiesor in the English Heritage regional offices.Please contact regional advisors currentlybased in universities at their university address.

North WestSue StallibrassUniversity of LiverpoolDepartment of Archaeology, Classics andEgyptology (SACE)Hartley Building, Brownlow StreetLiverpool L69 3GStelephone: 0151 794 5046fax: 0151 794 5057e-mail: Sue.Stallibrass@liv.ac.uk

North EastJacqui HuntleyDepartment of ArchaeologyUniversity of DurhamScience LaboratoriesDurham DH1 3LEtelephone and fax: 0191 33 41137e-mail: Jacqui.Huntley@english-heritage.org.uk

YorkshireAndy HammonEnglish Heritage regional office37 Tanner Row,York YO1 6WPtelephone: 01904 601 983fax: 01904 601 999mobile: 07747 486255e-mail: Andy.Hammon@english-heritage.org.uk

West MidlandsLisa MoffettEnglish Heritage regional office8th Floor,The Axis, 10 Holiday Street,Birmingham B1 1TGtelephone: 0121 626 6875mobile: 07769 960022e-mail: Lisa.Moffett@english-heritage.org.uk

East MidlandsJim WilliamsEnglish Heritage regional office44 Derngate, Northampton NN1 1UHtelephone: 01604 735451

mobile: 07801 213300e-mail: Jim.Williams@english-heritage.org.uk

East of EnglandJen HeathcoteEnglish Heritage regional office24 Brooklands Avenue, Cambridge CB2 8BUtelephone: 01223 582759mobile: 07979 206699e-mail: Jen.Heathcote@english-heritage.org.uk

South WestVanessa StrakerEnglish Heritage regional office29 Queen Square, Bristol BS1 4NDtelephone: 0117 975 0689e-mail:Vanessa.Straker@english-heritage.org.uk

LondonVacantEnglish Heritage1 Waterhouse Square, 138–142 HolbornLondon EC1N 2ST

South EastDominique de MoulinsInstitute of ArchaeologyUCLRoom 204A31–34 Gordon Square, London WC1H 0PYtelephone: 020 7679 1539fax: 020 7383 2572e-mail: D.Demoulins@ucl.ac.uk

Specific advice on scientific dating, includingluminescence, can be sought from the EnglishHeritage Scientific Dating Team (Alex Bayliss,John Meadows, and Isabelle Parsons).

Scientific Dating TeamEnglish Heritage1 Waterhouse Square, 138–142 HolbornLondon EC1N 2STtelephone: 020 7973 3299

fax: 020 7973 3001e-mail: Alex.Bayliss@english-heritage.org.uk

Appendix 2 Luminescencelaboratory contact detailsAn essential part of the successful applicationof luminescence dating is early discussionbetween the field project director and thespecialist undertaking the analysis. Contactdetails of laboratories in the United Kingdomthat are equipped to undertake luminescencedating are given below. Note that not alllaboratories undertake all forms of analysis nordo they all provide commercial services.Thelaboratories are listed in alphabetical orderbased upon their location or commercialname. Details were correct at the time ofwriting.AberdeenAlastair GemmellGeography & EnvironmentSchool of GeosciencesUniversity of AberdeenElphinstone Road, Aberdeen AB24 3UFtelephone: 01224 272337fax: 01224 272331e-mail: a.gemmell@abdn.ac.uk

AberystwythLuminescence Research LaboratoryInstitute of Geography and Earth SciencesAberystwyth UniversityCeredigion SY23 3DBtelephone: 01970 622606fax: 01970 622658e-mail: ggd@aber.ac.ukwebsite:www.aber.ac.uk/quaternary/luminescence

CheltenhamPhil TomsGeochronology LaboratoriesDepartment of Natural and Social SciencesUniversity of GloucestershireSwindon Road, Cheltenham GL50 4AZtelephone: 01242 714708fax: 01242 714826e-mail: ptoms@glos.ac.ukwebsite: www.glos.ac.uk/luminescence

DurhamIan BailiffDepartment of ArchaeologyDurham UniversitySouth Road, Durham DH1 3LEtelephone: 0191 3341100fax: 0191 3341101e-mail: ian.bailiff@durham.ac.ukwebsite: www.dur.ac.uk/lumin.dating/

LiverpoolAndreas LangDepartment of Geography

Roxby BuildingUniversity of LiverpoolLiverpool L69 7ZTtelephone: 0151 7942842fax: 0151 7942866e-mail: lang@liv.ac.ukwebsite:www.liv.ac.uk/geography/OSL/index.htm

LoughboroughHelen RendellDepartment of GeographyLoughborough UniversityLoughborough, Leicestershire LE11 3TUtelephone: 01509 223729fax: 01509 223930e-mail: H.M.Rendell@lboro.ac.uk

NottinghamMichèle ClarkeSchool of GeographyThe University of NottinghamUniversity Park, Nottingham NG7 2RDtelephone: 0115 9515446fax: 0115 9515249e-mail: michele.clarke@nottingham.ac.uk

OxfordJean-Luc SchwenningerResearch Laboratory for Archaeology and theHistory of ArtDyson Perrins BuildingSouth Parks Road, Oxford OX1 3QYtelephone: 01865 285222fax: 01865 285220e-mail: jean-luc.schwenninger@rlaha.ox.ac.ukwebsite: www.rlaha.ox.ac.uk

Oxford Authentication LtdDoreen StonehamOxford Authentication LtdBoston House, Grove Technology ParkWantage Oxfordshire, OX12 9FFtelephone: 01235 770998fax: 01235 771021e-mail: info@oxfordauthentication.comwebsite: www.oxfordauthentication.com

Richard BaileySchool of Geography, OUCE,South Parks Road, Oxford, OX1 3QYtelephone: 01865 284550fax: 01865 285220e-mail: richard.bailey@ouce.ox.ac.ukwebsite: www.ouce.ox.ac.uk/research/arid-environments/old

Royal HollowaySimon ArmitageDepartment of GeographyRoyal HollowayEgham, Surrey TW20 0EXtelephone: 01784 276124

fax: 01784 472836e-mail: simon.armitage@rhul.ac.uk

SheffieldMark BatemanDepartment of GeographyThe University of SheffieldSheffield S10 2TNtelephone: 0114 222 7929fax: 0114 2797912e-mail: M.D.Bateman@Sheffield.ac.ukwebsite: www.shef.ac.uk/scidr/luminescence

Scottish Universities Environmental Research CentreDavid SandersonScottish Universities Environmental ResearchCentreRankine Avenue, Scottish EnterpriseTechnology Park, East Kilbride G75 0QFtelephone: 01355 270110fax: 01355 229898e-mail: D.Sanderson@suerc.gla.ac.ukwebsite:www.gla.ac.uk/suerc/staff/sandersond.html

St AndrewsRuth RobinsonSchool of Geography & GeosciencesIrvine BuildingUniversity of St AndrewsNorth Street, St Andrews, Fife KY16 9ALtelephone: 01334 463996fax: 01334 463949e-mail: rajr@st-andrews.ac.uk

TL Quaternary SurveysNick Debenham19 Leonard AvenueNottingham NG5 2LWtelephone: 0115 9856785fax: 0115 9856785e-mail: n.debenham@qtls.globalnet.co.ukwebsite:www.users.globalnet.co.uk/~qtls/index.htm

43

Published September 2008Edited and brought to press by David M Jones,English Heritage PublishingDesigned by Adam Vines for English HeritageCreative ServicesPrinted by HawthornesMinimum of 50% recovered fibre, the remainderbeing from sustainable sources.

Product Code 51431

AuthorshipThese guidelines were compiled by G.A.T. Duller,Institute of Geography and Earth Sciences,Aberystwyth University, UK (ggd@aber.ac.uk).

These guidelines should be cited inbibliographies and references as follows:Duller, G A T 2008 Luminescence Dating:guidelines on using luminescence dating inarchaeology. Swindon: English Heritage

AcknowledgementsThe production of these guidelines has beenfunded by the Aggregates Levy SustainabilityFund distributed by English Heritage on behalf ofDEFRA.The case studies are included courtesyof Gifford Ltd and the RLAHA, University ofOxford (Gobowen),Aberystwyth and DurhamUniversities (Dungeness Foreland and medievalbricks), and Gloucestershire and ReadingUniversities (Broom).The luminescence analyseswere funded by Fletcher Homes (Shropshire)Ltd (Gobowen), and the Aggregates LevySustainability Fund (Dungeness Foreland andBroom).Antony Smith and Ian Gulley(Aberystwyth University) drew many of thediagrams. Comments on a draft of theseguidelines by many members of theluminescence and archaeological communitieshelped improve them substantially.

English Heritage is the Government’s statutoryadvisor on the historic environment. EnglishHeritage provides expert advice to theGovernment about matters relating to thehistoric environment and its conservation.

For further information and copies of thispublication, quoting the Product Code, pleasecontact:

English HeritageCustomer Services DepartmentPO Box 569Swindon SN2 2YP

telephone: 0870 333 1181e-mail: customer@english-heritage.org.uk

Front cover: Collecting luminescence samples under black plastic, Ripon,Yorks.

Micrographs from top to bottom:i. Measuring dose rate in Silbury Hill ii. Quartz grains and their luminescence signals iii. Laser beam for stimulating luminescence from single grains

Back cover: Luminescence sampling, Broom, Devon