You are on page 1of 61

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Physics applied to archaeology I. Dating

This article has been downloaded from IOPscience. Please scroll down to see the full text article. 1970 Rep. Prog. Phys. 33 941 (http://iopscience.iop.org/0034-4885/33/3/303) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 134.58.253.57 The article was downloaded on 12/03/2013 at 14:14

Please note that terms and conditions apply.

Physics applied to archaeology I. Dating


M. J. AITKEN
Research Laboratory for Archaeology and History of Art, University of Oxford

Contents

. 1. Introduction . 2. Radiocarbon dating 2.1. Introduction . 2.2. Man-made fluctuations in the 14Cconcentration of the exchange
3. Potassium-argon dating and uranium series dating . 3.1, Potassium-argon dating . 3.2. Uranium series dating (including ionium dating) . 4. Thermoluminescent dating . 4.1. Introduction . 4.2. Evaluation of the accumulated radiation dose . 4.3. Radiation dosimetry . 4.4. Application . . 5. Fission track dating 5.1. Introduction . 5.2. Archaeological application . 6. Dating by chemical change . 6.1. Dating of bone by fluorine, uranium and nitrogen contents 6.2. Obsidian dating . 6.3. Glass layer counting . . 7. Archaeomagnetism 7.1. Introduction . 7.2. Remanent magnetism . 7.3. Directional measurements . 7.4. Intensity measurements .
References

Page

942 943 943 948 950 955 959 960 96 1 963 963 965 970 973 975 975 975 976 976 976 977 978 978 979 983 992 994

2.3. Checks with known-age samples. . 2.4. Causes suggested for the 14Cvariations

reservoir.

Abstract. An account is given of the three main chronological applications of physics in archaeology: radiocarbon dating, thermoluminescence dating and archaeomagnetism. I n addition the following techniques are outlined: potassiumargon dating, uranium series dating (including ionium dating), fission track dating and dating by chemical change (including hydration of obsidian and weathering layers on glass). Some mention of dendrochronology and varve chronology is included in the section on radiocarbon. Radiocarbon dating, applicable to wood and some other organic remains, is predominant in this field and forms the basis of most prehistoric chronologies. Its techniques of measurement are now well established and the main research emphasis is concerned with the fluctuations in the radiocarbon content of the atmosphere that have been inferred from small systematic errors shown up in the dating of known-age samples. Possible causes for these fluctuations are past variations in climate, in sunspot activity and in geomagnetic field intensity. There is evidence from archaeomagnetic studies that the latter has been significantly different from its present day value, possibly varying sinusoidally with a maximum about 2000 years ago and a minimum about 6000 years ago. Archaeomagnetism, which is based on the magnetization fossilized in clay when it is fired, also
35

Rep. Prog. Phys., 1970, 33, 941-1000

942

M . J. Aitken
provides data on the past secular variation of the geomagnetic direction and to a limited extent this can be used for dating; the geophysical interpretation of the data is also of interest. Thermoluminescence is a fairly recently developed method and the main research emphasis is still concerned with various complications of technique. I t is important because it is directly applicable to pottery and pottery styles are the grammar of most archaeological chronologies. Potassium-argon dating is a well developed geological technique and is notable in the present context for its dating of volcanic material associated with early hominid remains. Uranium series dating has the possibility of being applicable to bone and shell of the Paleolithic period, this being too early to be reached by radiocarbon. Fission track dating is primarily a geological technique but it holds promise for archaeology, particularly in early periods. Obsidian dating and glass layer counting are based on the hydration caused by weathering. The former needs to be calibrated against radiocarbon but in a limited number of cases the latter appears to be absolute although the difficulty of understanding the physicochemical mechanism involved casts some doubt on its validity.

1. Introduction
An early involvement of physics in archaeology was at the end of the nineteenth century when the Italian scientist Folgheraiter (1899) measured the direction of the remanent magnetism in a number of Etruscan and other vases ranging in age from AD 100 to 800 BC; by noting the most likely position in which the vase would have been placed for firing he sought to establish the variation of the angle of inclination of the earths magnetic field over the period in question. A few years later Mercanton (1907) published a paper entitled La mCthode de Folgheraiter et son rale en gbophysique. This title emphasizes that the point of view of these early archaeomagnetists was that such archaeological objects carried a record of the past behaviour of the earths field which was of considerable geophysical interest; they were not primarily concerned with the idea of physics as an aid to archaeology. Archaeological remains also carry information about physical processes in materials. Again archaeomagnetism provides an early example-the long term stability of remanent magnetism is proved by the differing magnetic directions found in the bricks of an ancient wall or, as by David (Brunhes 1906), in the blocks of basalt used in the Roman temple of Mercury on the Puy de Dame. Thus ancient man sometimes unwittingly started off experiments which need centuries to complete and for which knowledge of when t = 0 relies on archaeological evidence. A short term example of this is the determination of the spontaneous fission rate of uranium by counting the fission tracks in glass of known date of manufacture. It was only with the advent of radiocarbon dating that archaeologists really became aware of the tremendous importance to themselves of the stored information which physics could unlock. Radiocarbon dating now provides the backbone of most archaeological chronologies, having dramatically lengthened several of them. Its dominating role was appropriately underlined by the recent symposium held jointly by the Royal Society and British Academy to mark twenty years of growth. Thus sometimes physics is an aid to archaeology, and sometimes archaeology is an aid to physics (or geophysics), I n addition there is a range of technical aids, particularly in respect of detection of buried remains (to be discussed in part I1 of this article), that the special needs of archaeology has stimulated; sometimes geophysical techniques can be adapted for use on a smaller scale, sometimes new ones must be devised.

Physics applied to archaeology I . Dating

943

Inevitably, the treatment given to the various topics will reflect some personal bias and the author apologizes to those who may feel their work has been underemphasized. T h e depth of treatment varies too with the state of development of a project and its established relevance to archaeology. T h e references are not intended to be comprehensive but to provide an up to date entry into the subject. Lest this article is taken to imply that physics is an all-powerful force that can make conventional archaeology redundant, let it be emphasized at the outset that there would be little progress without the understanding collaboration of archaeological colleagues and that for many archaeological problems the trained eye of the excavator is an unrivalled instrument!

2. Radiocarbon dating 2.1, Introduction


Radiocarbon dating is possible because of the presence of a minute concentration of the radioactive isotope 14C in all living plants and animals, and in the dissolved carbonates of the ocean. T h e concentration is about 10-l2 relative to natural carbon (12C) and is effectively uniform ; the uniformity obtains because of comparatively rapid mixing throughout the carbon exchange reservoir which consists of the atmosphere, the biosphere and the oceans. The concentration stays approximately constant, because it represents the equilibrium level between the radioactive decay of 14C and its production from 14N in the upper atmosphere by cosmic ray neutrons. On the other hand, in animals and plants that are preserved after death (and thereby removed from the exchange reservoir) the concentration decays exponentially with the 5730 year half life of l4C, and by comparing the concentration found in such material with the concentration for living material the time that has elapsed since death can be determined. The presence of 14C is detected by measuring the p activity. For living material this is approximately 15 disintegrations per minute (that is about Ci) per gram of natural carbon. T h e idea of age determination by 14Carose in the course of a study of the effects of cosmic rays on the earths atmosphere by W. F. Libby and his group at the University of Chicago, in the late 1940s. Libby (1955) has described the method and its inception and has reviewed subsequent developments from time to time elsewhere (for example, Libby 1967, 1970 a, b). Much of the following sections is derived from these publications.

2.1.1. The production of radiocarbon. Radiocarbon is produced in the atmosphere by the thermal neutron reaction:
(2.1) The cross section for this reaction is about 1.7 x lO-Z4 cm2, It is the dominant way in which neutrons interact with nitrogen and for oxygen the neutron cross section is lower by a factor of a thousand. T h e neutrons concerned are secondary particles produced by the incidence of the primary cosmic ray flux on the atmosphere. Although they may have high energy on formation, the neutrons rapidly become thermalized by collision and because of the dominance of reaction (2.1) the effective fate of each neutron is to produce a 14Catom. It is known from high altitude balloon measurements that the average neutron production rate is about 2/s/cm2 of the earths surface; this yields a global production rate of 7.5 kg of 14C per year.
14N +n
=

14C+IH.

944

M . J . Aitken

The neutron intensity builds up from zero at the outer limit of the stratosphere to a maximum at a height of around 15 km, and then falls off again reaching 5 % of maximum at about 4 km. T h e primary cosmic ray particles are charged and they are deflected by the magnetic field of the earth unless they are travelling parallel to the lines of force. Consequently the cosmic ray intensity is a maximum at the poles and a minimum around the geomagnetic equator ; the corresponding neutron intensities in the stratosphere are in the ratio of about 5 to 1. However, the mixing of 14C in the atmosphere is sufficiently rapid and thorough for these inhomogeneities to be unimportant when considering the uniformity of the I4C concentration. It is presumed that the 14Catoms combine with oxygen to form heavy carbon dioxide, which, except in respect of radioactive decay (and isotopic fractionation effects), is indistinguishable from the ordinary carbon dioxide of the atmosphere and so circulates in the same pattern through the carbon exchange reservoir. Although the latitudinal differences in 14C production do not matter, the dependence of the overall global production on the earths magnetic field intensity is important in view of the archaeomagnetic evidence that in the past substantial changes in the latter have occurred. A stronger magnetic field means a greater degree of shielding of the earth from the primary cosmic ray flux and therefore a lower production of 14C. This is discussed further in $2.4.2. It should be noted that although the constancy of the rate of production is of critical importance, the determination of age does not require quantitative evaluation of this rate.

2.1.2. Half life and equilibrium level. Radioactive decay is accompanied by the emission of a weak B particle (maximum energy, 160 keV) according to
(2.2) T h e presently accepted value for the half life is 5730 years. This corresponds to a mean lifetime of 8290 years and it is convenient to remember that in approximate terms the disintegration rate is 1% per 83 years. The equilibrium amount of 14C on earth is simply obtained by equating the decay rate to the production rate of 7-5 kg per year. T h e answer is 62 tons and this is distributed throughout the carbon exchange reservoir. When death occurs (or, for example, when shell is deposited from ocean carbonate) and the material happens to be preserved, the atoms in it are no longer in exchange with the reservoir and the concentration of I4C decays according to
1 4 c= 1
4

+p-. ~

NoexP

0.693t (-m)

where N is the concentration at time t years, and No is the value of N when death occurred (at t = 0). Experimentally the concentration is determined by measuring the p activity per gram of natural carbon and the date calculated from

where C is the observed ,B activity and COis value of C at death. Obviously the value of COcannot be measured directly and it has to be assumed that it is the same as for recently grown material (but see $2.2). This basic assumption that the 14C concentration in the exchange reservoir has been constant over past time was a

Physics applied to archaeology I. Dating

945

reasonable one to make and from check measurements on known-age samples it appears that to within a few per cent the assumption is valid back to 1500 BC. Before that, however, there is evidence of a significant long term excess of up to 10% lasting for several millennia. T h e causes of concentration variations, and the corrections that are now applied to radiocarbon dates, are discussed in $5 2.3 and 2.4. 40, T h e half life of 5730 years is the mean of three determinations: 5695 L5715 ? 50, 5780 i- 65 (Hughes and Mann 1964). This value is 3% higher than the value 5568 which was used by Libby at the inception of the method on the basis of the best determinations then available. However, for the time being, laboratories continue to calculate and publish dates on the basis of the Libby half life so as to avoid confusion and risk of multiple correction; this was agreed at the 1962 Cambridge Radiocarbon Conference (Godwin 1962a) and at the same time the convention was recommended that such dates should be quoted in years BP (meaning Before Present, but for convenience defined as before AD 1950) and regarded as the age in radiocarbon years rather than in calendar years. When a date is quoted in terms of AD or BC it is implicit that the half life correction-and also the bristlecone pine correction (see $ 2.3.1)--may have been applied, and whatever correction has been made should be stated. 2.1.3. The carbon exchange reservoir. Plant life grows by photosynthesis of atmospheric darbon dioxide and, in turn, animals live off plants; consequently 14C spreads throughout the biosphere. Also, atmospheric carbon dioxide enters the oceans as dissolved carbonate, so this too contains 14C. T h e carbon dioxide withdrawn from the atmosphere by plant and animal life is eventually returned to it by the decomposition that follows death-except for the very small proportion that is locked up in well preserved organic remains. T h e carbonate in the ocean is formed by an exchange reaction so that carbon dioxide leaves as well as enters the ocean-again except for a small proportion which is locked up for long periods, this time in shells and other carbonate deposits. Thus the atmosphere, biosphere and ocean form an exchange reservoir throughout which the carbon atoms circulate comparatively rapidly, so that the residence time in any one component of the reservoir is short compared to the mean lifetime of a 14C atom. T h e amount of natural carbon in the reservoir is 40 x 10l2tons; 937& of this is in the deep ocean, 2 % in the surface (mixed) ocean, 1.67; in the atmosphere, 0.8% in the terrestial biosphere and 2.6% in humus. T h e mixing into the reservoir of the equilibrium 62 tons of 14Cgives an average weight fraction of about 1.5 x 10-l2. This concentration determines the value of CO of equation (2.4) and the assumption that CO has always been the same could be upset by a change in the production rate of 14C(as already mentioned), by a change in the overall size of the reservoir or, more subtly, by a change in the mixing rates which determine small nonuniformities in the 14C concentration in different parts of the reservoir. I n considering possible nonuniformities the most obvious would be a greater concentration at high latitudes than at low on account of the deflection of the primary cosmic ray flux by the earths magnetic field. However, as mentioned in $2.1.1, the mixing of carbon dioxide within the atmosphere is sufficient to remove any latitude dependence of I4C concentration; this was shown by an early test on recently grown wood samples from various latitudes (Libby 1955) and has been confirmed subsequently. Since the atmospheric circulation systems of the northern and southern hemispheres are separate there is also the possibility of a

946

M . J . Aitken

difference between the concentrations in the two hemispheres, but measurements indicate that the concentration in the southern hemisphere is only about 0.5 % lower than in the northern hemisphere (see, for example, Lerman et al. 1970); this is attributed to the greater (by 40%) area of ocean in the former compared to the latter. There are, however, two well established types of nonuniformity. T h e first arises because the interchange of carbon dioxide between atmosphere, surface ocean and deep ocean is not infinitely rapid and may be understood by reference to the box representation of the exchange reservoir given in figure 1. T h e average
7.5 kg of per,year
I4C

c
Humus 2.6 10

y FlBO
Atmosphere Terrestrial biosphere

I
I

DeeD ocean

9 3 lo

Figure 1. The carbon exchange reservoir. The heavy arrow at the top represents the production of 14C in the atmosphere by cosmic ray neutrons. The yearly transfer of natural carbon between compartments is expressed in units of lo9 tons. The percentages refer to the natural carbon in each compartment; the total natural carbon in the reservoir is 40 x 10I2tons.

residence time of carbon atoms in the atmosphere and surface ocean before reaching the deep ocean is some tens of years but the residence time in the deep ocean is at least several hundred years. Consequently there is time for a significant amount of radioactive decay and relative to the atmosphere the carbon of the deep ocean shows an apparent age; for example, measurements by Bien et al. (1963) on various deep water samples from the Pacific Ocean showed apparent ages of up to two thousand years. T h e surface ocean is intermediate since the carbon in it is a mixture of reinvigorated carbon from the atmosphere and old carbon from the deep ocean (see, for instance, Nydal 1967). T h e second type of nonuniformity is due to isotopic fractionation. T h e mechanism of the exchange reaction between atmospheric carbon dioxide and ocean carbonate slightly reduces the ageing effect just referred to (by 1.2%);conversely the photosynthesis of atmospheric carbon dioxide into terrestrial plant life results in a lower (by 3.7% on average) 14C concentration in the latter. It is now common practice in most laboratories to evaluate the effect for each individual sample that is dated; the percentage enrichment of I4C due to fractionation is twice that for the stable isotope I3C and it is possible to measure

Physics applied to archaeology I . Dating

947

the fractionation effect for the latter isotope with a mass spectrometer since its concentration in natural carbon is about 1%. T h e effect for I4C is obtained by doubling the percentage enrichment found for I3C. 2.1.4. Suitability of sample material. T h e essential requirement is that at death the sample becomes isolated from the exchange reservoir. T h e requirement must be obeyed stringently since in a sample which is, say, 17 000 years old (three half lives), the 14C concentration is one eighth of the concentration in the reservoir and the addition of only 1% of reservoir carbon will increase the apparent sample activity by 8 % ; this will give an apparent age that is too recent by about 600 years. I t follows also that great care must be taken to rid the sample of any extraneous contamination by modern carbon such as might arise from humic acids in the soil being absorbed in the sample; to avoid this samples are usually washed in alkali before measurement. Conversely the deposition of calcium carbonate from ground water may contaminate the sample with old carbon and this can be guarded against by washing in acid. Exchange with the reservoir after death is least likely when the molecular structure is large; in addition, the sample should be in a well preserved condition since decomposition may be associated with the assimilation of fresh carbon. Charcoal is one of the best types of material and this is fortunate since it is widely associated with human occupation. Other suitable materials are wood, cloth, peat, leaves, hair, skin, antler, leather, and paper. Bone and shell can also be dated but special pre-treatments are necessary. With charcoal and wood there is the possibility that the sample represents inner rings of a slow-growing tree and in such a case the radiocarbon date can be several hundred years earlier than the archaeological event of felling the tree; this is because when a ring is formed the carbon atoms are fixed in it and the 14Cactivity found in it is determined by the decay that has taken place since the date of formation despite the presence of higher and lower activities on either side.

2.1.5. Measurement techniques. Accurate measurement is difficult and tedious for


several reasons. Because the specific radioactivity is very low, each measurement must continue for a long time (at least 24 hours) in order to average out statistical fluctuations ; this implies good stability both in electronics and in background. Stringent precautions must be taken to ensure that constructional materials are free from contamination. T h e equipment must be shielded from cosmic rays as otherwise the effect of the sample itself will be insignificant; it is necessary to employ an anticoincidence shield in addition to ten or twenty tons of steel. Besides being weak in terms of number per minute, the /3 particles, having a maximum energy of only 160 keV, are weak in terms of penetrating power (their intensity is roughly halved by a 10 pm thickness of aluminium); it is therefore necessary that the sample should be inserted in the sensitive volume of the counter rather than separated from it by a window. Initially Libby used the solid carbon technique (Libby 1955). I n this the sample is converted to carbon dioxide by combustion and then reduced to carbon black by heating with magnesium. T h e carbon black is then painted onto the inner wall of a modified Geiger counter. One disadvantage of the method is that the painting process entails exposure of the highly absorptive carbon to the atmosphere with risk of radioactive contamination. Another disadvantage is that only a proportion of the

948

M.

r. Aitken

p particles

emitted give rise to a count (due to poor geometry and self-absorption) and the sample size required is about 10 g of carbon-unpleasantly large in many archaeological circumstances. T h e most commonly employed method today is to convert the carbon into carbon dioxide, methane or acetylene and to use this as the counting gas of a proportional counter. I n addition to the higher counting efficiency this has the advantage that pulse height discrimination can be used to reject large pulses arising from 01 contamination. T h e volume of the counter is typically several litres and it is filled to a pressure of one or two atmospheres; the sample size is of the order of 1 or 2 g of carbon and depending on the amount of dead volume in the counter, the counting rate for modern carbon is between 10 and 20 counts per minute. With good shielding a typical background would be of the order of 1 count per minute. A more recent technique which is now being adopted (and is in use for instance in the British Museum Research Laboratory) is to convert the carbon into an organic liquid (such as benzene) that is suitable as a liquid scintillator solvent. One advantage is that a given amount of carbon is contained in a much smaller volume than for gas counting and so shielding against cosmic ray background is much easier. Another, more important advantage is the ease with which counting of sample, standard and background can be alternated (automatically), thus eliminating error due to drifts in detection sensitivity or background. I n both of the last two techniques the chemical preparation and purification calls for considerable effort and care. Apart from the obvious need to avoid contamination in reagents, it is important to secure 100% conversion of the samples carbon into the gas or liquid; this is because of the possibility of isotopic fractionation in the various chemical processes. T h e level of the counter background sets a limiting age of about 50000 years for a good installation; for such samples the radioactivity is down by a factor of 500 (corresponding to 9 half lives). T o measure older samples it is necessary, before measurement, to enrich the 14C concentration by deliberate isotopic fractionation (requiring a very large amount of sample) ; this was in fact the procedure used in the initial detection of 14Cin nature when only high background counters were available ; but used in conjunction with a low background counter some years later it was possible to detect the radioactivity corresponding to a 70000 year old sample (Haring et al. 1958). However, problems of sample contamination and sample integrity then become formidable and in practice the limit remains at around 50 000 years. 2.2. Man-made fluctuations in the 14Cconcentration of the exchange Yeseraoir 2.2.1. The Suess OY fossilfuel effect. T h e combustion of coal and oil releases into the atmosphere large quantities of carbon dioxide in which the 14C has decayed long ago, because the coal and oil mere removed from the exchange reservoir millions of years before. This old carbon significantly dilutes the 14Cconcentration in the atmosphere and the activity of wood samples, grown say in 1950 (prior to hydrogen bomb testing-see next section), is in fact lower than samples grown in 1850 (prior to the industrial revolution), despite the decay that has occurred in the latter. T h e effect was first observed by Suess (1955) using samples of recent wood, and further work (Houtermans et al. 1967) has indicated that the variation observed is unlikely to be one of the natural fluctuations discussed in 52.3. I t has

Physics applied to archaeology I . Dating

949

also been found that the 14C concentration is lower in the neighbourhood of large cities than elsewhere. T h e 14C concentration of samples grown in 1954 is about 2% lower than what would be expected from measurements on wood grown before 1850 (that is to say, after making allowance for decay). This is about ten times the effect expected on the basis that the estimated amount of old carbon released in this way (from 1860 to 1954) amounts to only 0.2% of the total carbon in the exchange reservoir. T h e observed effect is much stronger because of the so-called hold u p in the atmosphere; the finite mixing rate between atmosphere and ocean means that the fossil carbon has not had time to mix uniformly throughout the reservoir and an above average proportion of it is in the atmosphere. From the observed strength of the effect it can be estimated that the average residence time before transfer to deep ocean is between 10 and 30 years (Revelle and Suess 1957).

2.2.2. EfSect o f nuclear weapon tests. It has been estimated that the neutrons released by fission and fusion explosions up to 1962 have caused the formation of about 2 tons of 14C (Walton et al. 1967). Distributed uniformly throughout the reservoir this would increase the 14C concentration by about 3 yo. However, as with the fossil fuel effect there is a hold up in the atmosphere and the observed effect is much stronger; at present the atmospheric 14C concentration (as determined from recently grown plants) is approximately double the level in the pre-bomb era. As the 14Cfrom the bombs gradually circulates into the ocean the level will fall to the 3 yo excess representing uniform distribution-assuming hopefully that there are no further explosions. T h e sharp injection of I4C into the atmosphere in this way allows useful study of short term parameters, such as mixing rates between different parts of the atmosphere, and as the years go by the longer term parameters as well (see, for example, Nydal 1967, Munnich and Roether 1967, Bien and Suess 1967). I n particular, explosions at high altitudes in high latitudes as in 1961 and 1962 introduce 14Cinto the same regions as are most important naturally and the movement of the excess level through the various components of the reservoir allows the route taken by natural I4C to be traced. Another use of these sharp rises in the atmospheric 14C concentration is in confirming the integrity of samples. By definition a reliable sample is one for which the carbon atoms are firmly fixed and do not exchange with the atmosphere. Hence it can be checked that, in living trees, the inner rings that were grown in the pre-bomb era do not show contamination from the bomb 14C. Botanical considerations suggest that there may be some continued exchange for rings within the sapwood layer which is about twenty rings thick for most trees.
2.2.3. Standards for recently grown material. As explained in 92.1.2 the age calculated for a sample depends basically on the ratio ( C 0 / C ) where COis the ,B activity that the sample is assumed to have had at death and C is the activity measured today. On account of the fossil fuel effect and the bomb effect it is clearly invalid to assume that COis given by the activity found in present day living material. An early alternative adopted was to calculate the value for COby measuring the activity of wood grown prior to the onset of the fossil fuel effect in 1850 and make allowance for the subsequent radioactive decay. However, in view of the natural short term fluctuations in 14C concentration (see figure 2) this procedure has pitfalls, and

950

M . J. Aitken

discrepancies between different laboratories are likely to arise unless all choose wood of the same age. It is now common practice to refer all activities to the oxalic acid standard stockpiled for this purpose by the National Bureau of Standards, Washington, D.C. T h e value for COis taken as 95% of the measured oxalic acid activity; the latter is in excess because it was prepared subsequent to the beginning of nuclear weapon testing. 2.3. Checks with known-age samples Although radiocarbon dates are calculated solely from laboratory measurements their validity depends critically on the assumption that CO, the activity of living material, has remained constant over past time. This is quite apart from either of the recent fluctuations mentioned in the last section which are more in the nature of technical difficulties. T o within the accuracy then attainable, the validity of this assumption was confirmed in the early days of the method by Libby et al. (1949) using known-age samples back to 3000 BC. Later as precision was increased it became apparent firstly, that short term fluctuations in COof one or two per cent have occurred during most periods, and secondly, that for at least several millennia before 2000 BC there was a significant long term excess in Co. T h e quantitative evaluation of these variations in COand an understanding of their geophysical causes have been major preoccupations of workers in this field for a number of years and these topics were the subject of a Nobel Symposium (No. X I I : Radiocarbon Variations and Absolute Chronology) held in Uppsala in August 1969. T h e following sections draw heavily on the papers presented at that meeting. 2.3.1. Tree-ring samples. As mentioned earlier, the radiocarbon age for wood indicates the date of formation of the wood rather than the death of the tree. Consequently, the annual rings of a long lived tree provide a valuable source of well dated samples. In a recent tree the date is established merely by counting backwards from the outermost ring, but by using the techniques of dendrochronology it is also possible to date the rings of trees that died in antiquity. This applies to trees grown under stress conditions in the sense that their annual. growth is mainly determined by climate. Then, within one climatic region the pattern of varying ring widths is the same for all trees and it is possible to match the pattern of the outer rings of an old tree with the pattern of the inner rings of a younger tree whose life span just overlapped the older tree. T h e overlapping is applied to successively younger trees until the present day is reached. Care must be taken to avoid error due to missing rings and multiple rings-by cross checking against contemporary trees grown in conditions that cause them to be sensitive to climate to differing degrees. Tree-ring studies are going on in various parts of the world and the species and regions which are suitable for dendrochronology are gradually being established. Measurements on reliably dated samples by various workers show that fairly continuous short term fluctuations (over a few hundred years) of 1 or 2% have occurred in the atmospheric 14C concentration during most periods. Evidence for this was first obtained by de Vries (1958) using European and American trees spanning the past 500 years and the short term fluctuations are often referred to as the de Vries effect. This type of measurement was extended back a further

Physics applied to archaeology I . Dating

95 1

800 years by Willis et al. (1960) and subsequently to earlier periods by a number of others. T h e bristlecone pines of the White Mountains of California are of unique importance in this context; these can grow to ages of more than 4000 years and are well preserved after death. Intensive study at the Tree-ring Research Laboratory (University of Arizona) has resulted in the establishment (Ferguson 1968) of a 7100 year chronology, with expectation of eventual extension back to 9000 years. Measurements on rings from the 7100 year sequence (in 10 year bundles) have been made by the radiocarbon laboratories of the Universities of Arizona, California (La Jolla), and Pennsylvania. T h e results indicate that in addition to the short term fluctuations, there was a long term excess of about 8% in the atmospheric 14C concentration from 5000 BC to 3000 BC, reducing to zero by the first millennium

s
v

i
t 't
I t

Figure 2. Fluctuations in atmospheric 14C concentration over the last 12 000 years. The continuous and dashed curve represents the tree ring data of Suess (1970), the large full circles are the Scandinavian varve measurements of Tauber (1970) and the dots with vertical lines are the lake-varve measurements of Stuiver (1970). The fluctuations are inferred from the differences between the radiocarbon ages and the ages obtained by visual counting of rings or layers. The value obtained by measuring upwards from the conventional horizontal axis X indicates the excess obtained if the radiocarbon age is calculated on the basis of the Libby half life of 5568 years. Measuring from the inclined axis X' gives the excess on the basis of the revised half life of 5730 years. There is as yet no explanation of the discordance, before 5000 BC,between the two sets of varve measurements. The lake-varve measurements subsequent to 4500 BC (which are not shown in order to avoid confusion) are in good agreement with the tree ring data.

and changing to a small deficit of 1 or 2% until around about 600 years ago; it is convenient to remember here that an excess or deficit of 1% corresponds to an error in age of 83 years. T h e most comprehensive results are those obtained at La Jolla by Suess (1970 a) and these are shown in figure 2. They are based on measurements from upwards of three hundred samples, each being counted successively in each of two counters to a statistical standard error of less than 0.5%; consequently the effort in carrying out a programme of this type is formidable. Results from Arizona and Pennsylvania have been published by Damon (1970) and Ralph and Michael (1969) respectively. Damon compares his results with those of Suess and concludes that any overall systematic difference must be less than 0.5%, discounting the suggestion by Walton and Baxter (1968) that there may be a more significant discrepancy. T h e bristlecone pine results are effectively a calibration by which 'radiocarbon years' can be converted to calendar years. An important consideration is confirmation that the fluctuations are indeed worldwide, though in view of the effective
BC

952

iM. J . Aitken

uniformity found for present day concentration it would be surprising to find that they were not. This point of view is reinforced by studies of the movement of bomb 14C within and between components of the exchange reservoir; the only appreciable delay is some tens of years before entry into the deep ocean (see $5 2.2.1 and 2.2.2) but this is irrelevant to the question of uniformity within the atmosphere and biosphere besides being barely significant in the context of dating. Additionally, from a study of tree rings from different latitudes in the northern and southern hemispheres Lerman et al. (1970) conclude that, between AD 1400 and AD 1900 at any rate, the fluctuations are synchronous in the two hemispheres and of the same amplitude. Detailed consideration of tree ring samples grown in different latitudes in AD 1835, AD 1695 and AD 1450 show only the steady 0*50/, excess of northern hemisphere over southern already noted (3 2.1.3). T h e evidence just mentioned is relevant primarily to the short term fluctuations. However, the considerations of rapid mixing argue even more strongly that the long term excess of the fourth and fifth millennia BC must be a worldwide effect. Nevertheless, because of the archaeological consequences of the corrections to dates that are implied by this excess-and stoutly resisted in some quarters-the direct indications from glacial varves (see $ 2.3.2) and Egyptian archaeological material (see 5 2.3.3) are important. These confirm the reality of the long term excess but are not as yet sufficiently detailed to confirm (or dispute) the short term fluctuations. Hence at the moment the evidence for the short term fluctuations in the early millennia is solely from the bristlecone pines of California; however, since they are well established for recent millennia it would indeed be surprising if they were not present for earlier times. Although the long term excess has been confirmed as occurring in several parts of the world, it is of interest to note an effect (suggested by Rama and referred to in Berger 1970) that could in principle give rise to an abnormally high 14C concentration in the bristlecone pines. T h e altitude at which these trees grow is 3000 m and the cosmic ray neutron intensity at that height, being an order of magnitude greater than at sea level, may be sufficient to give significant in situ 14Cproduction from 14N present in the wood (to the order of 1y). However there are a number of parameters in this calculation whose value can only be roughly estimated and it is by no means certain that the effect is of importance. T h e short term fluctuations represent a more serious drawback to the method than the straightforward need to apply a correction. I n some periods the decrease in I4C concentration is so rapid that in principle an older sample may give a more recent radiocarbon age than a younger one. I n practice such reversals tend to be obscured by the standard deviation of measurement and the effect is rather that, in such periods, a given radiocarbon age may correspond to a range of calendar dates that is several times wider than the span indicated by the standard deviation. T o take one extreme example, calculation from figure 2 will show that a radiocarbon age of 4100 I 5 0 BP corresponds to calendar dates from 2900 BC to 2550 BC. On the other hand, in certain special circumstances the short term fluctuations can be utilized to give a much more precise date than would otherwise be available. This is when a large beam of wood at the site being dated provides a floating tree-ring sequence of a hundred years or more ; the technique is to find where the 14Cfluctuation pattern in the floating sequence matches the pattern in the bristlecone pine sequence. I n this way Suess (1970 c) has dated the Neolithic site of Auvernier to 50 years. an accuracy of i

Physics applied to archaeology I. Dating

953

T h e bristlecone pine corrections have some important effects on various archaeological chronologies of which the most notable is probably on the dating of the beginning of the Central European Middle Neolithic period (Quitta 1967, Neustupny 1968: see $4.4.1). An example in a more familiar context is in the dating of Stonehenge 111-the sarsen structure with its trilithons-which are usually assumed to be linked chronologically with Mycenae. Renfrew (1968) has pointed out that application of the bristlecone pine corrections to the relevant radiocarbon dates places the construction of Stonehenge I11 before 1800 BC and probably before 2000 BC. On the other hand the date of the first phase of the Mycenaean culture, being based on links with the Egyptian historical calendar, remains at around 1600 BC. Hence the assumption that the Wessex culture associated with Stonehenge I11 was strongly influenced by Mycenae is no longer tenable. 2.3.2. Checks against varve chronology. Analogous to tree rings are the annual layers (varves) found in clays that originate in the beds of lakes dammed up by glaciers, During the summer, melting is more rapid than in winter and the sediment deposited is thicker and coarser; conversely to tree-rings, a hot, dry summer is represented by an exceptionally thick layer. Varve sequences are recognizable in the same way as with tree-rings and the method has been extensively applied by Baron de Geer (1940) to the varves left behind as the glaciers retreated the length of Scandinavia at the end of the last Ice Age, establishing an absolute chronology beginning in 11500 BC. There is insufficient organic matter in these varves to allow direct radiocarbon dating but by the changing spectrum of pollen types found in them (as different botanical species successively became dominant) they can be related to peat bogs, from which satisfactory radiocarbon samples are obtainable. A recent study by Tauber (1970), together with a reassessment of the de Geer chronology in the light of further varve investigations (see also Fromm 1970), firstly gives confirmation of the long term 14C excess indicated by the bristlecone pine measurements (see figure 3). I n addition, Taubers results, reaching back to 10 000 BC, indicate that the period of excess did not begin until after 6500 BC and that before then there had been a deficiency of about 3%. T h e latter is in effective agreement with earlier checks made against the Swedish varve chronology. T h e long-term excess indicated by the bristlecone pine measurements is confirmed more comprehensively in direct measurements made by Stuiver (1970) on organic matter in a long core from the Lake of Clouds, Minnesota; this core contains 9500 thin laminations which are reasonably assumed to be annual layers. On the other hand (see figure 2), these results indicate that there was a excess of about 10% as early as 8000 BC, in strong disagreement with the Swedish varve results. I n work on other lake sediments that are not laminated, Stuiver (1970) plots radiocarbon age against thickness of sediment and notes an apparent inversion of the time scale around 20 000 radiocarbon years ago, corresponding to a 20% change in 14C concentration. He tentatively suggests that this might be associated with the Laschamp geomagnetic polarity inversion (see $7.3.6). 2.3.3. Checks against the Egyptian calendar. T h e basis of the absolute chronology of European, Near and Middle Eastern archaeology of the second and third millennia BC is the Egyptian calendar ; this is via stylistic relationships and artefact

954

M . J . Aitken

links. Therefore in checking radiocarbon dates against established archaeological chronology, reliance on sometimes rather circumstantial evidence is avoided by using Egyptian material itself-as was done by Libby when he used wood from the tombs of Sesostris 111,Zoser, Sneferu and Hemaka in his initial test of the technique. Egyptian ~ chronology is firmly based on recordings of the Back to 2 0 0 0 ~ astronomical event of the heliacal rising of the bright star Sirius in 1470-1450 BC, in 1540-1520 BC and in 1870 BC. T h e 1870 BC rising, in fact, occurred during the seventh year of the reign of Sesostris 111. Earlier than 2000 BC the chronology is by extrapolation using the Turin canon (a list of kings that was not compiled until the thirteenth century BC) and incomplete fragments covering the first and fifth dynasties ; differing interpretations place the start of the first dynasty-the beginning of historic times in Egypt-within the limits 3000 BC 100 (Save-Soderbergh and Olsson, 1970). T h e discrepancies between radiocarbon dates for Egyptian material and the Egyptian chronology has been considered by a number of investigators (e.g. Suess 1967, Michael and Ralph 1970, Save-Soderbergh and Olsson 1970) and within the uncertainties of archaeological context there is general agreement with the bristlecone pine results. One frequent type of uncertainty occurs when the sample is from a beam, since in such cases the actual growth of the wood may precede the archaeological context by up to a century or more. T o avoid this a special collection of short lived materials (for instance, straw in bricks) has recently been sponsored by the British Museum and made available to various radiocarbon laboratories ; preliminary results (Barker et al., 1969, Berger 1970) substantiate the trend of the bristlecone pine results but, assuming the Egyptian chronology to be exact, it is possible to infer that the bristlecone pine results give a small overestimate of the long term I4C excess (Berger 1970). This overestimate could arise from exchange within the sapwood layer (which in bristlecone pine is probably substantially thicker than the twenty or so rings that is usual for most trees) or from the in situ production effect mentioned earlier (5 2.3.1). 2.3.4. Comparison of other techniques with radiocarbon. T h e techniques outlined in $0 3 and 4-potassium-argon, uranium series and thermoluminescent dating-do not depend on cosmic ray intensity and so a comparison with the results of these techniques is of interest. However, the time range of radiocarbon is too recent for the first of these, potassium-argon dating, to give results of sufficient precision for this purpose. It is only recently that the lower age limit of the method has been reduced so as to overlap with radiocarbon at all and it is much more a matter of using radiocarbon as a check of potassium-argon rather than vice vesa. With uranium series dating, although the time ranges covered by ionium (230Th)and protactinium (231Pa) have a much better overlap with radiocarbon, various geochemical uncertainties make the situation similar. However, the results of Kulp and Volchok (1953) do indicate that variations in the radiocarbon production rate (assuming this to be the dominant cause of the long term I4C fluctuations) have not exceeded 10-20% over the past 35 000 years. T h e accuracy obtainable with thermoluminescence dating should soon be good enough to confirm (or dispute) the reality of the long term excess; preliminary comparative results (see 0 4.4.1) are not inconsistent with the bristlecone pine results. Thermoluminescence holds the prospect of carrying comparison back beyond the period covered by the bristlecone pines and even before the period covered by varve chronology; there is then the possibility of more marked long

Physics applied to archaeology I. Dating

955

term variations in 14C concentration because of climatic variations associated with the Ice Age. T h e preliminary results for the 30 000 year old Czechoslovak figurine site of Dolni Vestonici are in reasonable agreement with radiocarbon, though possibly about 10% too old. It is unlikely that thermoluminescence dating will ever be sufficiently precise to have bearing on the short term fluctuations.

2.4. Causes suggested f o r the I4C variation As with $2.3, the discussion of the present section is based mainly on the proceedings of the Uppsala Symposium. Although there seems good reason to think that the long term excess is associated with change in the geomagnetic moment and that the short term fluctuations reflect solar activity, it appears likely that climate also plays some part in both aspects. Opinions differ as to the relative importance of the various possible mechanisms and a definitive stage has not yet been reached.

2.4.1. Response of the reservoir to a change in production rate. Assuming the natural carbon content of the reservoir to remain fixed, the average 14C concentration N is determined by the differential equation

_ dN - - A"
dt

Q(t)

where X the radioactive decay constant is equal to the reciprocal of 8290 years (the mean lifetime) and Q(t) is the normalized production rate of 14C (expressed per gram of natural carbon). If Q(t) changes from a fixed rate Qo to a new fixed rate Qo+ SQ, the concentration responds slowly, with a time constant of 8290 years, according to

S Q 1 - exp (- At). SAT = X


Thus the effect of transient changes in Q will be greatly attenuated relative to the sin u t , effect of a long term change. If the time variation of Q is expressed as Qo+ Q1 then the amplitude of the resulting sinusoidal component in N is h(u2+A2)--l/~ times the steady state change in N that would result from a long term increase of Q1 in the production rate. Because the reservoir consists of several compartments (atmosphere, biosphere, surface, ocean and deep ocean-see figure 1) between which the mixing rates are finite, the actual situation is much more complex. A useful understanding may be obtained by considering (de Vries 1958) the electrical analogy to the system-a number of leaky capacitors connected in parallel via small resistances ; the capacitors represent each compartment, the leakage current corresponds to radioactive decay, and the currents flowing in the interconnecting resistors represent the net flow due to mixing between compartments. More recently the problem has been studied using computers (see, for instance, Houtermans 1966, Venkatavaradan 1970) ; the general conclusion remains valid-that there is severe attentuation of short term changes in production rate (typically by a factor of ten for periods of the order of a few hundred years). This does not contradict the amplifying effect due to 'hold up ' in the atmosphere noted earlier ($2.3.2) in respect of bomb-produced 14C. T h e latter effect corresponds to the dumping of a finite amount of charge on a capacitor-the voltage change is immediate and depends on the size of the capacitor, not on the time constant determined by any parallel resistors.

956

M. J. Aitken

T h e insensitivity to short term changes is fortunate since variations in the production rate of up to 20% may occur during periods of high solar activity (see $2.4.3). If the same percentage variation was shown by the I4C concentration, radiocarbon dating would barely be viable since 20% corresponds to an error of about 1600 years.
I Or
I

0.6I

2.4.2. Variation o f the geomagnetic moment. There is good archaeomagnetic evidence (see $7.4.2) that substantial variations have occurred in the earth's magnetic moment, M. T h e resulting variation in the I4C production rate was first considered by Elsasser et al. (1956) who concluded that

Q oc M-0.62.

Physics applied to archaeology I. Dating

957

T h e value of M determines, for a given latitude, the minimum momentum of cosmic ray particle that can reach the atmosphere, and consequently (2.7) is based on the observed spectrum of cosmic rays. T h e effect has subsequently been considered by others, including Ramaty (1965) and Lingenfelter and Ramaty (1970) who are in agreement with relation (2.7) as a good approximation. When the archaeomagnetic data were extended back into early millennia (Bucha 1967), it became evident (Bucha and Neustupny, 1967) that there was good corelation with the long term 14C concentrations variations, in line with relation (2.7). T h e correlation has subsequently been considered by Damon (1970), La1 and Venkatavaradan (1970), Lingenfelter and Ramaty (1970) and Suess (1970 b). Figure 3 is taken from the work of the last-named; the lower part shows the production rates calculated from the various archaeomagnetic data points according to relation (2.7) and the best fit sine wave to those points has a peak to peak amplitude of 65% of the minimum value and a period of 8000 years. T h e upper part of the figure shows the best fit sine wave to the observed 14C concentration; this has a peak to peak amplitude of 10% and a period of 10300 years, lagging in phase by about 700 years with respect to the production rate sine wave. To within the limits of uncertainty the two periods are in agreement, and the phase lag and the ratio of the amplitudes are consistent with the response characteristics expected for the reservoir (Houtermans 1966). Suess (1970 b) further points out that if part of the 14C variation is accounted for by a steady decrease in concentration following the last Ice Age (as in Suess 1968) the best fit sine wave to the residual variation has about the same amplitude but the period decreases to 8000 years in close agreement with the magnetic prediction, However, as emphasized by La1 and Venkatavaradan (1970), it would be premature to accept the long term 14C variations as being definitively explained in this way since the archaeomagnetic data are not yet on a sufficiently worldwide basis, particularly in the earlier millennia. It is possible that the observed variations are due to regional disturbances and do not truly reflect a change in the geomagnetic dipole moment (see 9 7.4.2). 2.4.3. Correlation with solar activity. Indications that low 14C concentration is correlated with high sunspot activity led Stuiver (1961) to postulate that the associated intensification of the weak interplanetary magnetic field carried by the solar wind caused a reduction in the cosmic ray flux in the earths vicinity. More direct evidence of this was obtained from cosmic ray data over the past two 11-year solar cycles, particularly in respect of the low energy component responsible for I4C production. An analysis of the data and knowledge of the 14C yield from the various components of cosmic radiation led Lingenfelter (1963) to suggest the empirical relation
= 2.64--

B
337

where Q is the average I4C production rate (expressed as atoms/s/cm of the earths surface) and 8 is the average sunspot number. T h e variations in Q to be expected during a typical solar cycle are less than 6% but during periods of high activity the decrease might reach 20%. However, for the 11-year cycle the effect on the 14C concentration in the reservoir will be unobservable because of the very strong attenuation of short periodicities referred to

958

M . J . Aitken

earlier. On the other hand, there are various longer periodicities in solar activity, and from a comparison of 14Cvariations (using dated tree ring samples) and sunspot numbers over the past 400 years (from recorda back to AD 1750 and estimates by Shove (1955) for earlier periods), Houtermans et al. (1967) have found significant correlation for periodicities of 50 years and 200 years. Also, from a detailed study of the short term 14C variations over the past 7000 years, Suess (1970 b) notes a tendency for minima in the concentration to occur at a spacing of 400 years, and a cycle of this periodicity is among those postulated by various authors for solar activity. Suess further observes that the short term variations are less pronounced during periods when the geomagnetic moment is high which is to be expected if the cause is indeed modulation of the low energy component of the cosmic rays, since it is the lower energy cosmic rays that are deflected away by the geomagnetic moment. An alternative explanation of the 14Cvariation/sunspot correlation is that solar activity is reflected in climatic variations and the short term 14Cvariations are caused by the effect of the latter on the exchange reservoir (see Q 2.4.4). Lingenfelter and Ramaty (1970) have pointed out that there is another effect during high solar activity which may obscure the decrease due to interplanetary field intensification; this is the production of 14C by particles associated with solar flares. I t is estimated that the increase in atmospheric 14Cdue to a single event may be of the order of 1yo and consequently solar flares may be responsible for some of the sudden short term increases. These authors also consider possible increases in 14C production that might arise from a supernova explosion. 2.4.4. Climatic eflects. I n commenting on the tree-ring measurements which gave the first indication of short term fluctuations in 14C concentration, de Vries (1958) noted that excess concentration occurred during periods when the climate was severe (as indicated by glacier advance), and vice versa. Since then correlation with climate has been considered by a number of authors (for example, Damon 1968, Suess 1968), but, as with sunspot activity, the time span over which there are reliable records is limited, though the general trend in early millennia is known from paleotemperature measurements using the isotopic ratio, ls0/l6O. The most obvious climatic effect is the change in the size of the oceans due to gradual melting of glaciers following the end of the last Ice Age. Carbon that is locked up in glaciers is not part of the exchange reservoir and so it is old carbon, deficient in 14C. Consequently its release into the reservoir would cause a fall in the 14C concentration. The normally accepted figure of 2.5% for the change in ocean volume would make the effect too small to explain the observed long term decrease of 872, although Schell et al. (1967) suggest that 2.5% may be an underestimate. An additional complicating effect is the temperature dependence of carbonate solubility; this produces a transient effect because a warming of the oceans causes the release into the atmosphere of carbon that is to a small degree relatively deficient in l4C. I t is to be noted that sharp transient effects due to climatic change can only all be such as to produce a decrease in atmospheric 14C concentration-since components of the reservoir have a lower I4C concentration than the atmosphere. Hence it is difficult to explain sharp increases in terms of climate. Most mechanisms considered for climatic effect are in terms of changes in mixing rates between reservoir compartments. If there was infinitely rapid mixing between atmosphere and ocean then there would be no 14C deficiency in the latter (other than arising from isotopic fractionation). In fact the mixing is finite and this

Physics applied to archaeology I . Dating

959

means that on the average the I4C in the atmosphere is a little newer than that in the ocean, giving rise to a slight excess of 14C concentration in the atmosphere. If the mixing increases then the excess decreases and vice versa. One way in which the mixing rate between atmosphere and deep ocean could be altered has been emphasized by La1 and Venkatavaradan (1970); this is the decrease in the area of the southern surface outcrop of deep ocean water that would occur during glacial and colder periods, due to coverage by ice. T h e detailed behaviour can only be understood in terms of models of the reservoir (such as shown in figure 1) using parameters that are to some extent conjectural. 2.4.5. Indications f r o m other radioisotopes. Data about the constancy of the cosmic ray flux are available from studies of other radioisotopes produced by cosmic rays, though in practice the precision at present attainable is barely sufficient to make the results relevant to radiocarbon dating. Oeschger et al. (1970) have recently discussed various aspects of such studies and emphasize the importance of the rare gas isotopes 3Ar, s9Ar and 31Kr in this context. T h e respective half lives are 35 days, 270 years and 210000 years, and the present concentrations in the atmosphere (when compared with the expected concentration ratios as deduced from production cross section data) reflect the cosmic ray intensity averaged over time periods of the order of these half lives. Thus, the ratio of 39Ar to 81Kr is an indication of the average intensity over the last few hundred years relative to the average over the past few hundred thousand. A record of the past atmospheric concentrations is available in ice from deep bore holes in Greenland and Antarctica which has been deposited over the past l o 5 to lo6 years. After allowance for radioactive decay subsequent to deposition, the concentration ratios at a given level in the ice give indication of the cosmic ray intensity averaged over time periods (as determined by the respective half life) immediately prior to deposition of that level. T h e concentrations of these rare gas isotopes will not be affected by climatic changes in the same way as radiocarbon since only small percentages dissolve in the ocean. Hence, if when precise measurements become available the pattern of fluctuation matches that shown by the 14Cvariations, this will be definitive evidence that the cause of the variations lies in cosmic ray fluctuations. There will still be the question of distinguishing between the effect of a varying geomagnetic moment and effects due to solar or other extraterrestrial activity. Evidence about this can be obtained from study of meteoritic and lunar material since the geomagnetic moment does not affect the cosmic ray flux experienced by such samples. For example, Schaeffer et al. (1970) have measured the T 1 (half life, 308 000 years) and 39A (half life, 270 years) concentrations in a number of recently fallen meteorities and conclude that the average cosmic ray intensity over the past million years was the same (to within 10%) as the average value during the past thousand years. T h e observed 14Cvariations are in fact within this limit of accuracy but the result is illustrative of the type of information that is likely to be useful when greater precision is attainable.

3. Potassium-argon dating and uranium series dating


These are essentially geological techniques and except for the use of the latter one in dating bone and shell they are not directly applicable to archaeological materials. However, when remains are associated with a geological event (such as a volcanic eruption) archaeological application becomes possible.

960

M . J. Aitken

3.1. Potassium-argon dating This is an example of the classic technique for radioactive dating in geology; it is simpler in concept than the radiocarbon method and does not utilize isotopes produced by cosmic rays. Its basis is the build-up of argon formed by the radioactive decay of 40K in volcanic rocks, the amount present being a measure of the time that has elapsed since the rocks last cooled down. Natural potassium is present in rocks in the range 0-10% and in addition to the stable isotopes 39K and 41K contains 0.012% of the weakly radioactive isotope 40K. This has a half life of 1-3x lo9 years and decays either to argon,
40K

110 O

>

40A

or to calcium, 40K
%

, 40Ca,

T h e age determination is based on argon rather than calcium because the former can be detected with very high sensitivity in a mass spectrometer. Also, because argon is a gas, it is more likely not to exist at the time of the event being datednamely, the volcanic eruption. Thereafter the argon builds up according to

N,

exp ( . ; (1exp (-

At) At)

where Nl and N, are the numbers of 40Kand 40Aatoms in a given amount of sample and X is the disintegration rate. The *OK content is determined either by measuring the total potassium content chemically (by flame photometry) or by measuring the 39K content using neutron activation (Merrihue and Turner 1966, Crasty and Mitchell 1966). I n either case knowledge of the abundance ratio for the potassium isotopes allows derivation of the 40K content itself. The difficulty of using the method on rocks that are young enough to be of archaeological relevance is due to the very slow rate of formation of argon. For example, rocks with a potassium content of 2%, formed 10000 years ago, have a O A content of only about Beside the need for high fractional radiogenic * instrumental sensitivity there is also the difficulty of contamination by 40A from the atmosphere; this occurs by absorption on the sample material and on the walls of the measuring apparatus. The minimum age that can be measured depends on the condition of the minerals present as well as on the potassium content. The limit to which the method is considered generally useful is around 100000 years though in special cases meaningful ages have been obtained for much younger rocks; for instance, Dalrymple (1967) obtains ages in the range 6000 to 10000 years for some potassium-rich lava flows (the measurements being made on potassium feldspar separated out from the lava). The most notable examples of archaeological application are the dating of hominid remains in Africa. The Olduvai man (Zinjanthropus) has been dated to 1.75 million years by Evenden and Curtis (1965). Hominid remains and artefacts found near Lake Rudolf in Kenya have been dated to 2.6 million years by Fitch and Miller (1970). I n both cases the dating is of the volcanic pumice with which the remains are associated, Mention of some other archaeological applications, as well as a fuller account of the technique, will be found in Gentner and Lippolt (1969). T h e applicability of potassium-argon dating to volcanic lava makes it important for paleomagnetic studies of the remanent magnetization recorded in the lava; in

Physics applied to archaeology I . Dating

961

particular it has been vital in establishing that the reverse magnetization of rocks is synchronous on a wordwide basis and does truly represent a reversal of the geomagnetic field (see 3 7.3.6).

3.2. Uranium series dating (including ionium dating) T h e uranium series consists of a dozen or so radioactive isotopes and in a rock that is old enough these are in equilibrium, i.e. A,Nl = A,N, = &AT3 ... (34 where A,, A,, ... represent the disintegration rate of the successive decay products, and N,, N,, ... their equilibrium quantities. Thus the equilibrium amounts are universely proportional to the decay constants-that is, directly proportional to the respective half lives. If the equilibrium is disturbed by the chemical removal of all products subsequent to the nth member of the chain then the re-establishment to equilibrium of the (n f 1)th member is determined by the equation

of which the solution is

This means that the (n+ 1)th member builds up at a rate determined by its own half life rather than that of the preceding member. Alternatively if the upset to equilibrium consists of removal of all members preceding the (n + l)th, the ( n + 1)th member is said to be unsupported, and it decays according to

eventually reaching the new equilibrium value of zero. Uranium series dating is based on measurement of the degree to which equilibrium has been re-established (according to either 3.4 or 3.5) following an event that has upset the equilibrium. T h e successive long lived components of the uranium series are
Uranium I Uranium I1 Ionium Radium Radium G

238U 234U
230Th zzaRa z06Pb

half life, 4.5 x lo9 years half life, 2.5 x lo5 years half life, 75 200 years half life, 1622 years stable

For simplicity the intervening short lived members have been omitted; their short half lives mean that the equilibrium amounts are very small and the rapidity with which they re-establish equilibrium makes them uninteresting from the geological point of view. T h e long lived components of the actinium series are:
Actinouranium Protactinium Actinium D

235U
231Pa z07Pb

half life, 7 x lo8 years half life, 32 500 years stable

T h e thorium series has only two long lived components :


Thorium Thorium D 232Th losPb half life, 1.4 x 1O1O years stable

962

M . J . Aitken

T h e isotopes of interest in the present context are ionium (230Th),radium (226Ra) and protactinium (231Pa). T h e basis of the ionium method (230Th)of dating ocean sediments is that because of its chemical nature, this isotope tends to be precipitated whereas the parent uranium remains in the sea water in solution. Hence in deposited sediment the ionium is no longer supported and its concentration decays according to the half life of 75 200 years. Consequently in sediment that has been deposited at a uniform rate the 230Thconcentration decreases exponentially with depth and the time that has elasped since the sediment at a given depth was deposited can be determined. Error due to variation in rate of deposition can be avoided by measuring the ratio 230Th/232Th. 232Th has a very much longer half life but is chemically identical so that its concentration is a measure of the deposition rate of the 230Th. I t is assumed of course that the uranium and thorium concentrations in sea water remain constant and that the inherent concentrations of each in the sediment are negligible. These and other assumptions are not often strictly met in practice and more reliable results are obtained (Rosholt et al. 1961) by measuring the ratio of ionium to protactinium (IPa). This isotope has a half life of 32500 years and is similarly deposited in ocean sediments. Reasonable agreement has been obtained with radiocarbon dating. Such agreement was also obtained with the straightforward ionium technique for cores selected as best fulfilling the necessary requirements (Volchok and Kulp 1957). Of more direct archaeological relevance is application to shell, stalagmite and bone (Cherdynstev et al. 1965, Broecker et al. 1968, Fornaca-Rinaldi 1968, Szabo and Rosholt 1969, Szabo et al. 1969, Hansen and Begg 1970, Turekian et al. 1970). If it is assumed that uranium is incorporated at formation (or shortly after) without any accompanying 230Thand 231Pathen measurement of the degree to which these isotopes have built up towards equilibrium can be used to determine the age on the basis of (3.4). This equation assumes that after formation the system was closed, so that there was no subsequent incorporation of uranium, nor any leaching out. In this case the ages obtained independently from measurement of each isotope (relative to its parent uranium) should be concordant. T h e frequent occurrence of discordant ages indicates that, for shell, the closed system assumption is not valid and it is necessary to use the more complex open system model (Szabo and Rusholt 1969) in which the age is determined from the 231Pa/230Thratio; it is assumed that there is a mobile uranium component which migrates through the shell and leaves it whereas the decay products from this component are retained. Direct evidence for the mobile component is the occurrence of 231Pain excess of equilibrium (when compared with the amount of retailzed uranium). T h e uptake of uranium during burial is a chemical method (see 56.1) of distinguishing old bones from new. Hence one might expect that the ages obtained for bone by the 230Thand 231Patechniques will be too young. However, the results of Szabo et al. (1969) for seven bone samples from Mexico gives ages for four of them that are unacceptably old from the archaeological point of view-since they imply the presence of hominids in America more than 200000 years before the generally accepted date for the Old World. T h e results of Hansen and Begg (1970) for nine bone samples from California indicate that when ground water has been present the uranium rapidly reaches its final value, but that in drier situations the residence time of the uranium is less than the true age. T h e average age (103 000 6000 years) for three samples that gave concordant 230Thand 231Pa ages was in

Physics applied to archaeology I. Dating

963

agreement with geological evidence. T h e number of measurements is as yet too limited to judge the validity of the method's application to bone but it is a development of the highest potential importance for the Paleolithic period since this is mostly beyond the range of radiocarbon dating. Turekian et al. (1970) have attempted to date bone by measuring the amount of helium that has accumulated from the CY particle decays in the uranium series, a technique which when applied to shell has given ages concordant with 230Thdating. T h e results for nine bone samples (estimated to be several million years old or more from other evidence) give ages that are much too young; this is taken to indicate that there has been continuous loss of helium from the samples. Measurement of the various isotopes relevant to uranium series dating is by the techniques of mass spectrometry, a particle spectrometry and neutron activation, in combination with chemical separation.

4. Thermoluminescent dating 4.1. Introduction


This method of dating pottery (and some other ceramic materials) is a combined application of solid state and nuclear physics. I n radiocarbon dating the quantity measured is the amount by which the 14C concentration has decayed below its initial value and in potassium-argon dating one measures the build up of radiogenic argon. I n thermoluminescent dating one also measures a build up but this time it is an accumulating solid-state effect in the mineral constituents of the clay fabric of the pottery. T h e effect results from exposure to the nuclear radiation coming mainly from radioactive impurities that are present in the fabric too-a few parts per million of 238L? and 232Th(plus their radioactive decay chains in equilibrium) and 232Thand a few hundred parts per million of 40K. Because the half lives of 238U, 40K are all very long (log years or more) the radiation dose rate they provide is effectively constant; typically it is of the order of one radt per year. Part (about 20%) of the dose is provided by y rays from radioactive impurities in the soil in which the pottery is buried; the contribution from cosmic rays is 2 or 3%. T h e ionization produced by this radiation gives rise to the trapping of charge carriers (electrons and holes) at defects in the crystal lattices of the minerals concerned, such as quartz (see figure 4). Some of the electron traps are deep enough (1.7 eV, according to Fleming 1969) for there to be negligible leakage from them over the period of nine to ten thousand years during which man has been making pottery, so that the trapped electron population should grow linearly with age. Although the trapped electrons (for this particular trap in quartz) have a long lifetime at an ordinary temperature, the lifetime is reduced to a fraction of a second if the temperature is raised to about 400 "C, and the traps are emptied. This of course happens when raw clay is fired into pottery (involving temperatures of 700 to 1000 " C ) and so a newly made pot starts off with zero trapped electrons. T h e electrons released when the temperature is raised recombine with holes in various ways. If this happens at a luminescent centre, light is emitted (thermoluminescence). T h e wavelength, in the region of 420 nm for quartz, is usually characteristic of the luminescent centre concerned and in a simple case the process is as follows. When charged with a hole the centre is a singly ionized atom and if an attracted free electron (after release from a trap) is captured by it, it becomes a
1 rad = 100 ergs of absorbed energy/g.

964

M . J . Aitken

neutral atom in an excited state; subsequent de-excitation gives rise to the luminescence emission. T h e phenomenon has been studied for several hundred years-for example, by Sir Robert Boyle (1664)-but it is only in the past decade that, stemming from the work of Daniels et al. (1953), the effect has been utilized in practical applications. T h e most important of these has been in measuring the accumulated exposure to radiation of patients undergoing radiotherapy and of research workers subject to nuclear hazard. For these purposes, highly sensitive natural minerals such as calcium fluoride and artificially prepared phosphors such as lithium fluoride are used (see, for example, Cameron et al. 1968). T h e thermoluminescence (TL) observed is a measure of the cumulative dose of radiation to which the phosphor has been exposed since the last previous heating.

ITi
I I
1

CO)

(61

Figure 4. T h e thermoluminescence process in quartz : (a) Ionization by natural radioactivity gives rise, during antiquity, to the accumulation of trapped electrons. For dating purposes only traps deep enough to retain electrons over thousands of years are of interest. T h e lifetime of an electron in the trap shown is estimated to be of the order of lo7 years (Fleming 1969). (b) On heating in the laboratory the trapped electrons are released at about 375 C (for the trap shown) and some of these are captured at luminescent centres with the emission of light.

I n application to pottery dating the natural TL carried by the sample is compared with the artificial T L acquired from exposure to a known dose of radiation from a radioisotope source, and, in this way, the equivalent radiation dose that has accumulated in the pottery since it was fired is determined. By measuring the amounts of uranium, thorium and potassium present in the pottery and in the soil, the radiation dose received by the fragment each year can be calculated. In principle the age is then directly obtained as: Accumulated radiation dose Age = Dose per year though in practice various factors (see $4 4.2 and 4.3) make the determination much more complicated.

Physics applied to archaeology I . Dating

965

Early work on the T L dating of ancient pottery was carried out by Kennedy and Knopff (1960) and by Grogler et al. (1960). Since then reports have been made by Aitken et al. (1964), Fleming (1966), Ralph and Han (1966), Tite (1966), Z'immerman (1967), Aitken, Zimmerman and Fleming (1968), Mejdahl(1969), Z' immerman and Huxtable (1969) and Fleming (1970). T h e outline now to be given is based primarily' on the developmental approach of the author's group at Oxford. At present an absolute accuracy of somewhat better than + l o % can be achieved; eventually it is hoped to improve this to + 5%, but for various reasons that is likely to be the limit of precision. This is somewhat poorer than is obtainable with radiocarbon dating, though when the full implications of the short term fluctuations in 14C concentration are worked out this may not be the case in some periods (see g2.3.1). However, the TL method has the advantage that the event dated is the actual firing of the pottery by ancient man, whereas in the case of radiocarbon dating of wood or charcoal the event dated is usually the felling of a tree and this may precede involvement with ancient man by one or two hundred years. Also, because of their durability, pottery fragments are relatively abundant on most archaeological sites.

4.2. Ee.aluation of the accumulated radiation dose 4.2.1. Apparatus. Three considerations dominate the design of the measuring equipment. Firstly, the light intensity is low-in typical circumstances it is of the order of 105 photons s-lsr-l, roughly lmsr-l. Secondly, when the sample temperature exceeds 300 "C there is a significant amount of incandescent light (i.e. thermal radiation), which masks the TL signal unless colour filters are used to discriminate against it. Thirdly, in addition to the TL that is induced by exposure to radiation there is also a 'nonradiation-induced' or 'spurious' TL. One type of this is the tribo-TL that results from crushing pottery into a form suitable for measurement; there are other more subtle types too (see, for example, Aitken, Fleming, Reid and Tite 1968) and the phenomenon is far from well understood. If the TL measurement is made with the sample in air the spurious TL is liable to swamp the true radiation-induced TL. Fortunately, however, it has been found by making the measurement with the sample in an atmosphere of dry, oxygen-free nitrogen (or argon) that the spurious TL is reduced by several orders of magnitude. I t is not possible to use vacuum because of the difficulty that then occurs of con'ducting heat to the sample. A suitable form of apparatus is outlined in figure 5. As will be discussed in $4.3 the sample for measurement may be in the form of thin layer of fine grains deposited on a 1 cm diameter aluminium disc, or it may consist of about 10 mg of quartz (or other mineral) grains which have been separated out from the pottery. I n either case the sample is placed on a 0.3 mm thick strip of nichrome that is heated electrically. T h e current is servocontrolled, so as to give a uniform rate of rise of temperature-usually 20 "C per second. Rapid heating is advantageous since the TL intensity is proportional to the heating rate whereas the thermal radiation is independent of it, but if the heating is too rapid thermal lags in the sample will result in poor temperature resolution and the important 'plateau' test (see later) will be upset. T h e TL is detected by a low noise photomultiplier with a high blue to red sensitivity (for example, EM1 type 9635Q, or a specially selected 6255Q). T o

966

M.j ? Aitken

improve the discrimination against thermal radiation further (and also against spurious TL-see Fleming 1968 a), blue colour filters are used, a suitable combination being a Corning 5-60 or 7-59 together with a Chance-Pilkington HA3. T h e pulses at the anode of the photomultiplier, each of which corresponds to the arrival of a photon at the photocathode, are fed through a fast amplifier to a ratemeter, the output of which actuates the Y axis of a co-ordinate plotter. T h e X axis of this is driven by a thermocouple attached to the heater plate. A typical natural glow curve obtained from ancient pottery is shown in figure 5.
Pulse omplif ier

5 50

X-Y p l o t t e r

High voltoge supply ( = l 2 0 0 v ) 7 !


/ / / / / /

\\\\\\\
//////

hotomultiplier

,
Light guide Colour f i l t e r

\\\\\\\

//////

\\\ \\\\

///////

Nitrogen Servo heating control

I
Heater current of 100 A t o give heating rote o f 20 C per sec

Nichrdme heater plate

Sample

Figure 5. Block diagram of apparatus for thermoluminescence measurements. The glow curve drawn by the X-Y recorder is typical of the natural TL from ancient pottery. Also shown is the light observed from a second heating of the same sample; only thermal radiation (TR) is emitted, all the TL having been released in the first heating.

It is also possible to use the photomultiplier in dc mode; that is, the pulses at the anode are integrated into a current which is fed to coordinate plotter through a picoammeter or dc amplifier. Comparison between the two modes indicate that the pulse system is advantageous at low light levels (Aitken, Alldred and Thompson 1968). 4.2.2. Theglow curve. For a single trapping level the glow curve consists of a broad peak; in the case of the quartz level mentioned earlier this occurs at 375 C and has a width at half height of about 100 C (Fleming 1969). T h e kinetic processes which determine the shape of a TL peak have been discussed by Randall and Wilkins (1945), Garlick and Gibson (1948), Garlick (1949), Louchtchik (1953) and Halperin and Braner (1960). I n the simplest case (first order kinetics or monomolecular decay) it is assumed that the rate at which the traps empty is given by
dn -= ns exp dt

$1

Physics applied to archaeology I . Dating

967

where n is the number of filled traps at time t , s a frequency factor which expresses the number of times the electron presents itself in an attempt to escape from the trap, E the trap depth (in electron volts), k Boltzmann's constant and T the absolute temperature. This relationship assumes that there is negligible probability of an escaped electron being retrapped. If it is also assumed that the probability of an electron giving rise to luminescent emission is independent of the number of available luminescent centres, the light intensity at time t is given by

I ( t ) = -c-

dn dt

(4.3)

where c is a constant. Even with these simplifying assumptions the equation for the shape of the glow peak is rather complex since the number of trapped electrons remaining at time t is given by
n
=

n,exp

Snf

sexp

(-&I

dt] .

(4.4)

T h e temperature ( T * ) of the peak of the glow curve is an implicit function of E according to the relation

However, Curie (1963) has pointed out that the relationship between T * and E is comparatively insensitive to the values of (dT/dt) and s ; to a rough approximation T" is given by 500E for S E lo9 s-l, falling to 350E for S N 1013s-1. Fortunately, age determination is not quantitatively dependent on the foregoing. T h e glow curve observed is usually a composite of a number of overlapping peaks resulting from the presence of several types of trap within a single mineral, or from the presence in the sample of several different minerals. T h e dating information is obtained essentially by comparison of the natural TL with the TL induced by a known dosage of artificial radiation under conditions of identical TL measurement. T h e critical question in this comparison is the extent to which the artificial TL is associated with traps deep enough to have retained their electrons without leakage over archaeological times. Obviously traps that are shallow enough to be emptied at one or two hundred degrees are hardly likely to retain electrons over thousands of years and this is evidenced by the absence of TL below 250 "C in the glow curve shown in figure 5 . Substituting typical values in the relations of the previous paragraph one finds that the traps emptying at about 250 "C should have a lifetime at 17 "C of the order of 10 years whereas for 350 "C the lifetime is 30000 years. However, it is not necessary to rely on theory to determine at what glow curve temperature the TL becomes 'stable'. An experimental test that can be conveniently applied to each sample is readily to hand. This is the plateau test referred to earlier and illustrated in figure 6; the shape of the natural glow curve is compared with that of the artificial glow curve by plotting the ratio of the light outputs as a function of glow curve temperature (T,).This ratio is effectively zero below 250 "C, but, as Tgreaches a value for which the lifetime of the corresponding traps is comparable with the time over which the natural TL has been acquired (typically in the range 350 "C to 400 "C), the ratio rises and quickly reaches a constant level. I n all the discussion that follows it is to be taken for granted that the term 'natural TL' refers to the level in this stable plateau region.

968

M . J. Aitken

This same plateau test also checks against the possibility of electron leakage from traps by wave mechanical tunnelling, that is leakage which is independent of temperature. If such a process had in fact occurred it is extremely unlikely that it w7ould be to the same degree for all the traps concerned and consequently there would be no plateau.

.
yl y c l

0 c

GI w

IO00

v L

500-

0) c

0 > a

I n the case of quartz extracted from pottery the glow curve is usually a broad peak formed by the superimposition of the 385 C peak already mentioned, and a peak at 325 C (Fleming 1970). T h e lifetime associated with the latter is comparatively short whereas the lifetime for the former is some tens of millions of years (Fleming 1969). Consequently the onset of the plateau in this case can be taken to indicate that a temperature has been reached at which the TL is predominantly from the higher peak; the evidence for stability is then the experimental determination of the lifetime of that peak. A complicating factor in the interpretation of the plateau graph is the differing degrees of supralinearity exhibited by the two peaks. 4.2.3. Supralinearity. It has been assumed so far that the TL grows uniformly with radiation dose. This is only approximately true and figure 7 shows a typical growth curve observed for a pottery sample irradiated with ,&particles. T h e obvious way to deal with this complication would seem to be to plot out the growth curve of each sample after the natural TL has been drained from it in the course of the natural glow curve, and to determine the dosage necessary to induce an amount of artificial

Physics applied to archaeology I. Dating

969

T L equal to the natural TL. However, Tite (1966) found that the drainage of the sample often caused a significant change of TL sensitivity, either because of a transparency change associated with mineralological alterations due to the heating of the sample or by more subtle effects arising from the radiation dose received before drainage (the so-called pre-dose phenomenon). A more reliable procedure is to determine the TL sensitivity by the additive procedure : the additional TL induced by a known artificial dose is found by subtracting the natural TL measured with one portion of the sample from the natural plus artificial TL measured with a second portion. As long as the level of the natural TL is well above the supralinear portion of the growth curve the error due to ignoring the initial curvature is not too serious; also, it has been found by Thompson (1970) that reliable correction can be

Figure 7. Typical supralinear dependence of TL on radiation dose for pottery samples irradiated with or y radiation.

made by assuming that the dose intercept (see figure 7) is unchanged by drainage (but not necessarily assuming that the sensitivity is unchanged). Studies by Fleming (1970) on quartz extracted from pottery and on fired geological quartz show that for the 375 C (benign) peak the growth curve is the same before and after drainage but that for the 325 C (malign) peak it changes both in slope and intercept. There are several possible explanations of supralinear growth. T h e simplest is that additional traps are being created in the early stages of the irradiation thereby giving rise to enhanced sensitivity. However, there are various reasons for rejecting this explanation and Tite (1966) favours a competition model in which it is assumed that there is a second set of traps (which do not give rise to TL) competing for electrons. These are assumed to have a higher cross section and to saturate earlier than the TL traps, and so as the competing traps gradually approach saturation the T L sensitivity increases until when the competing traps are fully saturated, the incremental sensitivity becomes constant. Another explanation is in terms of an enhancement in the probability that an electron freed from a trap will produce luminescence. This could arise, for example, because of the increase with dose of the number of activated luminescent centres (Claffy et al. 1968, Thompson 1970).

970

M. J. Aitken

Supralinearity is exhibited only when the tracks of the ionizing radiation are sufficiently numerous for their effective volumes to overlap. For the range of dosages relevant to archaeology this occurs for ,8 and y radiation but not for a: radiation. 4.2.4. Dose rate dependence. T h e artificial dose is administered at a rate that is about ten million times the rate at which the natural dose was acquired. It is therefore reasonable to ask whether the efficiency of acquiring TL is at all dependent on dose rate. This is unlikely to be the case until the rate is so high that there is a probability of track volumes overlapping within the lifetime of the ionized carriers. Experiments with lithium fluoride (Tochilin and Goldstein 1966) give no evidence for dose rate dependence until levels of 1O1O rad s--l are reached-this is about lo9 times the highest rate used in dating work. Dose rate dependence might also occur through one of the supralinearity mechanisms mentioned earlier. For instance, if the competing traps are shallow ones their equilibrium population, and hence also the TL sensitivity, will be dependent on the rate at which they are being filled. However, there is as yet no experimental evidence that this occurs (Fleming 1969). 4.2.5. Dependence on density of ionization. It has been found experimentally that, for a given amount of absorbed energy, irradiation by a: particles is less efficient in inducing T L than /3 particles and y radiation. T h e ratio is usually called the k value and values range from 0.05 to 0.3, averaging 0.15 for pottery (Zimmerman 1971). Studies by Zimmerman (1970) establish that the ineffectiveness is due to saturation of the T L by the high local ionization density in the core of the track, rather than to radiation damage. Substances with a low k value also show early saturation in the TL growth induced by /3 irradiation. T h e consequence for dating is that when evaluating the dose rate which a sample has experienced it is necessary to calculate separately that due to ct particles. This component is then multiplied by the measured k value for that sample, thereby giving the (effectivedose rate from cy particles. Since the ionization density depends on particle energy it is also necessary to take into account the a: particle spectrum received by the sample (Zimmerman 1971). 4.3. Radiation dosimetry 4.3.1. Measurement. It is usual to evaluate the dose rate from uranium and thorium by a counting (Turner et al. 1958) and to determine the potassium content by chemical analysis (flame photometry). a: counting is preferable to chemical determination of uranium and thorium because of the gross errors which would be introduced by the latter procedure if the decay chains were not in equilibrium. this may be the case for some glacial Because of the 267000 year half life of 234U and postglacial clays. Another cause for the disequilibrium in the uranium series is the escape of the gas radon (n22Rn). T h e typical dose rates received by a fragment of pottery buried in soil are shown in figure 8. I t will be seen that when account is taken of the reduced effectiveness of 01 particles in inducing TL, the and y dose rates become important, and because of their greater range this means that the dose received by the pottery is determined not only by its own radioactive content but also by environmental radiation. T h e major environmental contribution is the y dose from the burial soil; for the ,8 dose

Physics applied to archaeology I. Dating

97 1

there is a transition layer of between 1 and 2 mm at the surface of the pot in which the dose is partially determined by the radioactive content and partially by that of the soil. T o avoid uncertainty from the latter effect the material for TL measurement must be taken from the inner part of the fragment. T h e y dose rate is estimated by laboratory determinations on a sample of the relevant soil or, more directly, by in situ TL dosimetry (see $4.3.3).

layer ( 9 2 m m )

Th K
Total

T I L effectiveness
Effective dose

860 855 0 1815 x0.2 363

39 26 86 151
XI

36 63 26 I25
XI

151 24 '10

I25

57 010

19 OIO

Figure 8. Typical annual radiation dose (in millirad) for fragment of pottery buried in soil, both having U 3 ppm; T h 12 ppm; K 1%.

T h e laboratory evaluation of radioactive content is made on a sample after it has been dried in the course of preparation. When buried the pottery fragment carries a significant amount of water and this has the effect of diluting the radioactivity. T h e saturation water content of most types of pottery is between 10 and 20% by weight; this can be measured for each sample and for the wet climates of northwestern Europe it is fairly safe to assume that the fragment has been permanently saturated. Where the climate is intermediate, uncertainty about the past climate significantly worsens the limits of accuracy of the date obtained, though the saturation content at least indicates the upper limit of the appropriate correction.

4.3.2. Radiation attenuation due to finite size of grains. T h e mineral inclusions in pottery are often up to several hundred microns in diameter. Since the uranium and thorium is carried predominantly in the fired clay matrix in which these inclusions are embedded rather than in the inclusions themselves, the a particles, having ranges between 20 and 50 pm, reach only a thin surface layer of the inclusions. T h e result is that the average dose received by an inclusion is less than that received by fine grains, to a degree dependent on the size of the inclusion. T h e TL sensitivity of the inclusions is usually several orders of magnitude higher than that of the fired clay and consequently the effect is important, even though their percentage by weight may be small. Two techniques for overcoming this difficulty have been

972

M . J . Aitken

developed-the inclusion technique (Fleming 1966, 1970) and the fine grain technique (Zimmerman 1967, 1971). I n the former the T L measurements are made on grains for which the a particle contribution is negligible, whereas in the latter the measurements are made on grains small enough for the a particle contribution to have its full value. As explained earlier these TL measurements yield the accumulated radiation dose. Then in using the relation (4.1) only the 13 and y contributions to the dose per year are included in the former technique, whereas for the latter the a contribution (multiplied by the k value-see 84.2.6) is included in addition. It is evident straightaway that the inclusion technique is much more affected by environmental radiation, but on the other hand it does avoid the complication of the k value. For the inclusion technique the crystalline components of the sample are separated from the clay matrix using a magnetic separator, and grains in the size range 90 to 105 p m are selected by sieving; larger grains are avoided because of attenuation effects. Grains of the selected size still carry a skin of a-irradiated material and the major part of this is removed by etching with hydrofluoric acid. I n the fine grain technique it is essential, in crushing the fragment, to avoid breaking up any inclusions. This is to ensure that the fine grains obtained are genuine rather than fragments of larger ones. A satisfactory crushing techique is to squeeze the fragment in a vice. After this, grains in the size range 1 to 5 pm are selected by their settling rate in acetone and deposited on a number of aluminium discs (1 cm in diameter and 0.45 mm thick). Each disc carries a 1 mg layer of fine grains, about 4 p m thick and these are used for the TL measurements, the reproducibility between discs being about 5 oh.

4.3.3. The environmental dose rate. It will be seen from figure 8 that for fine grain
dating the y ray dose is typically about 20% of the total effective dose, and for inclusion dating, about 50%. If the radioactivity of the soil is stronger than that of the pottery (particularly in respect of the thorium series, on account of its high y activity) then the contribution becomes more important still. T h e effective attenuation coefficients for the y rays from uranium, thorium and potassium are not dissimilar and calculations indicate (Fleming 1969) that 9576 of the dose originates from the soil within a 30 cm radius. I t is therefore important, firstly that the pottery fragment should have been buried to a depth of at least 30 cm for the major portion of the burial time, and secondly that the material lying within 30 cm of the fragment should be homogeneous. T h e ideal context is the middle of a pit or ditch. Heterogeneous contexts, such as destruction layers and tombs, are less satisfactory. Even in homogeneous situations there are a number of difficulties associated with the evaluation of the y dosage obtained by laboratory measurements on an extracted soil sample. Firstly, as with the pottery fragment itself the radioactive concentration is diluted by the presence of ground water (which is low in radioactivity) and this will vary seasonally and possibly on a longer term basis. Secondly, the y dosage from the uranium series is strongly affected by the emanation of the gas radon and it is well known that this is an appreciable effect, being enhanced in dry weather and attenuated in wet. T h e loss of radon in this way means not only the loss of the dose rate due to radon itself but also that due to the daughter products that follow radon in the uranium decay chain. I n fact 98% of the total y ray energy associated with the uranium chain lies beyond radon. T h e uncertainties due to this emanation would be far more important were it not for the fact that the uranium series supplies quite a modest part of the y dose rate.

Physics applied to archaeology I . Dating

973

; 1superior technique for the evaluation of the environmental gamma dosage is direct on-site measurement using a TL dosimetry phosphor (Aitken 1969, Mejdahl 1970). This involves placing a small capsule of the phosphor in as similar a burial situation as possible to that from which the pottery fragment has been removed. Phosphors are available with sufficient sensitivity for accurate measurement to be made within a few weeks but, to average out seasonal variations in wetness and in radon emanation, the capsule is left in position for a year when possible. Doubt about the y dosage is then centred on the extent to which short term measurements correctly reflects the situation throughout the burial period of the fragment. Radioactivity measurement by TL dosimetry can also be used in respect of the fragment itself (Fleming 1968 b, Mejdahl 1969). Besides the advantage of avoiding the need for the facilities necessary for a counting and potassium analysis, there is the advantage of principle that the calculated age is no longer dependent on a calibrated radioisotope source (Aitken 1968). T h e importance of environmental radiation imposes a serious restriction on TL dating, both in respect of accuracy and in respect of the circumstances from which samples are acceptable. At some stage in the future it may be possible to remove this limitation by using the fine grain technique and the inclusion technique in conjunction ; by subtracting the equivalent radiation dose measured by the latter from that measured by the former, one is left with the effective dose due to a radiation thereby avoiding the need to evaluate the environmental dosage. However, the measurement of the equivalent radiation dose is not yet sufficiently accurate to make this subtraction technique viable.

4.4.Application
4.4.1. Archaeological. As outlined in the foregoing sections the TL method is an absolute one, giving ages that are independent of any other chronology. Nevertheless it is important to confirm the validity of the technique and also that, in practice, there are no unforeseen interfering factors. Good material for this purpose is provided by Romano-British pottery which can usually be dated on archaeological grounds to better than k 50 years; comprehensive tests using such material have been reported by Aitken et al. (1968), Fleming (1970), and Zimmerman (1971). A test using ancient Danish ceramics has been made by Mejdahl (1969). T h e two more recent of those, employing respectively the inclusion technique and the fine grain technique, show that at any rate in the good circumstances obtaining for the : 10%. tests, an individual fragment can be dated to appreciably better than i T h e interest of comparing TL dates with radiocarbon has been mentioned earlier (5 2.3.4) in connection with the systematic deviations shown to be present in the latter method by comparison with bristlecone pine tree ring counting. A highly suitable type of pottery for this purpose is the Linear Bandkeramik of the Central European Middle Neolithic period. On the one hand it has a distinctive type of decoration so that there is no ambiguity about the material being dated, and on the other hand it was fairly short lived and is associated with radiocarbon dates on many sites (Quitta 1967). T h e conventional radiocarbon dates are in fact some 1500 years earlier than the date range of 3000-2500 BC to which it has been ascribed on the basis of connections with the Aegean and Egypt-the so-called short chronology. When the radiocarbon dates are corrected by means of the bristlecone pine data the discrepancy is increased by a further 600 or 700 years; they are then
30

974

M . J . Aitken

consistent with the long chronology proposed by Neustupny (1968) from a reinterpretation of the archaeological evidence. T h e preliminary TL dates obtained by Zimmerman and Huxtable (1970) provide additional strong support for the long chronology though (because the material obtained initially was not fully satisfactory) they are not sufficiently precise to confirm (or dispute) the reality of the bristlecone pine corrections. Even so the T L evidence is important since, being from a completely different technique, it should help to dissuade from their continued adherence to the short chronology the small but important minority of archaeologists who reject radiocarbon dates (whether corrected or not). 4 longer range comparison with radiocarbon has been made, using burnt clay fragments from the Upper Paleolithic site of Dolni Vestonici in Czechoslovakia ; this site is of particular interest because of the discovery there of human figurines and because of the early radiocarbon date of dates obtained for it of around 28 500 BC; (after adjustment to the new half life). T h e T L date obtained (Zimmerman and Huxtable 1971) was 31 000 I3000 BC; this used material collected from an excavation some years earlier and consequent environmental uncertainties do not allow a more precise determination.

4.4.2. Authenticity testing. It has been stressed that, because of the y dose from the
soil, it is only possible to obtain reliably accurate results from material specially collected from current excavations. However, for testing authenticity it is usually a question of distinguishing between an age of less than a hundred years and one of upwards of several hundred. I n such circumstances it is adequate to assume a typical value for the y dose. Consequently T L provides museum curators with a very powerful independent judgement of what on their shelves is imitation and what is genuine. Besides pottery, application is being made to terracotta figurines and statues (see, for example, Fleming et al. 1970, Fleming and Roberts 1970). 4.4.3. Extension to other materials. One essential of T L dating is that the level of stored T L in the specimen is zero at the time of the event being dated. For pottery this is achieved by the heating to above 500 C that occurs during firing; glass, burnt flint, burnt stone and volcanic lava also have their TL clock set to zero in this way. For shell and possibly bone the T L level may be zero on formation. However, there are other important requirements that must be met. T h e T L properties of the substance concerned must be satisfactory: the level of spurious TL must be small, the growth of the TL must be reasonably linear and the temperature of the T L glow peak must be sufficiently high for electrons to remain stably trapped over the time period concerned. Encouraging preliminary measurements have been obtained in the dating of lava as old as 70 000 years (Aitken and Fleming 1971), and in studies of the T L of shell (Johnson and Blanchard 1967). Another requirement is that the radioisotopes responsible for the radiation dosage must be long lived, or supported by a long lived parent, and this aspect requires further investigation for both lava and shell. I n respect of glass and flint a likely difficulty is the lightinduced TL acquired by the sample before burial. With bone the difficulty is in the removal of organic matter from the sample; it is difficult to do this without the application of heat and this of course drains the TL. However for really old bones, sufficient time may have elapsed for the removal to have taken place naturally and encouraging preliminary measurements have been reported by Jasinska and Niewiadomski (1970) for dinosaur bones.

Physics applied to archaeology I. Dating

975

5. Fission track dating 5.1. Introduction


In minerals and glasses that contain traces of uranium, substantial damage to the lattice is caused by recoiling fragments from the spontaneous fission of the 238U nucleus. These damage tracks can be made visible under the microscope by etching with hydrofluoric acid, the damaged regions being less resistant to attack. T h e tracks so revealed are between 5 and 20 pm long and can be distinguished from etch pits due to dislocations. T h e tracks are sufficiently stable (a temperature of the order of at least 500 "C typically being needed to anneal them) for the number present to be used as a measure of age-in a similar way to the build up of thermoluminescence. T h e uranium concentration is conveniently evaluated by measuring the increase in track density that results from exposing the sample to a known dose of thermal neutrons in a nuclear reactor. Thermal neutrons induce fission in 235U and with the cross section for this, the present day abundance ratio of 235U/238U and the spontaneous fission decay constant all being known, the age may be determined. T h e technique is primarily of use for geological dating (Price and Walker 1963), the event dated being the formation of the mineral concerned. Some archaeological application has been made but lack of sensitivity limits this technique to materials with a fairly high uranium content. I n absolute terms the accuracy of the geological ages obtained is limited by the precision to which the decay constant for spontaneous fission is known. T h e value for this used by Fleischer and Price (1964) is 6.9 x years-l and this gives concordance between fission track ages and potassiumargon ages. An alternative way of evaluating this constant is by measuring the track density in fairly recent uranium-rich man-made glasses, the age of which is known exactly from documentary evidence. T h e values quoted in the literature from radiochemical determinations show a considerable spread (Kaufhold and Herr 1967). T h e contribution to fission events in natural materials by fission induced by the ambient neutron flux is negligible.

5.2. Archaeological application Because of the very slow rate at which 238U undergoes spontaneous fission the number of tracks per cm3 for a 1000 year old sample having a uranium content of 1 ppm is only about 500. I n scanning a surface for tracks only those originating within a few microns of the surface are seen and the corresponding number of tracks seen per cm2 is about 0.3. I n practice it is difficult to work with samples containing less than 10 tracks per cm2 (Walker et al. 1969). Consequently for the method to be used on 1000 year old samples it is necessary for the uranium concentration to be at least 30 ppm. Although the uranium contents of ancient man made glasses is too low, in more recent glasses (of the past 150 years) uranium is sometimes used as a colorant at a level of a few per cent. Brill et al. (1964) have reported good agreement between the historical age and the fission track age for five such specimens. ,Although the average uranium content of ancient ceramics may be too low, there are sometimes uranium-rich mineral inclusions. As long as the firing temperature of the ceramic is sufficient, the tracks acquired by these during geological time are annealed and the date obtained corresponds to the archaeological event. This has been demonstrated (Nishimura 1971) for some Japanese hearths, pottery and tiles of the past two thousand years from which zircons (having uranium contents in the range 1000 to 10 000 ppm) have been extracted. Using the same principle Watanabe

976

M. 7. Aitken

and Suzuki (1969) have dated a 4000 year old ceramic vase by counting the tracks in an obsidian spearhead that was embedded in it. I n this case the uranium content was only 3 ppm and diligent counting of repeatedly repolished surfaces was necessary. Another way of obtaining an archaeological date is when skeletal remains are buried in volcanic material. T h e date of the geological event is then the same as the archaeological one. Fission track dating (Fleischer et al. 1965) of volcanic pumice from Bed I of the Olduvai Gorge has given an age of 2.03 k 0.28 million years, in agreement with the potassium-argon age of 1.76 million years. T h e geologic age of formation of volcanic material used for prehistoric tools could also be used as an additional facet in identifying the geographic origin of the material (Walker et al. 1969). I n particular this may be applicable in respect of obsidian-on the basis that all obsidian of the same source should have the same age-and complement the existing techniques of trace element identification (Renfrew et al. 1966). T h e barrier to using fission track dating in archaeology is, as already emphasized, the low rate at which spontaneous fission occurs. T h e observation by Huang and Walker (1967), using phase contrast microscopy, of fossil tracks in mica produced by the recoiling heavy nuclei that accompany the a-particle decay of uranium and thorium suggested that a much more sensitive technique might be available, since the a decay rate is many orders of magnitude higher. However, the tracks are very much shorter (0.01 pm) and consequently rather difficult to identify with reliability, particularly against a natural background of similar pits from some other cause (see, for example, Turkowsky 1969). There is as yet no evidence that such fossil tracks occur in glass.

6. Dating by chemical change 6.1. Dating of bone by jluorine, uranium and nitrogen contents
T h e fluorine and uranium contents of buried bone gradually increase because of slow incorporation of these elements in the phosphatic mineral (hydroxyapatite) of which bones are mainly composed. On the other hand, the nitrogen content decreases due to the gradual disappearance of protein (collagen). Because of dependence on external conditions these methods are restricted to relative dating of bones found in the same deposit (Oakley 1969). Old bones can be distinguished from later intrusions and these techniques were used in showing that the Piltdown kIan was a hoax. T h e rate of acquisition depends on the contents of uranium and fluorine in the ground water which has percolated through the deposits; in limestone caves, for example, the fluorine content of ancient bones is anomalously low. T h e disappearance of nitrogen is severely attenuated by freezing conditions or by the exclusion of bacteria and air (for example, by burial in clay). Associated with the incorporation of fluorine there is a change in the crystal lattice constants of the apatite. This can be detected by x ray diffraction studies and correlation pattern spacing, and age has been reported (see, for instance, Tanabe and Watanabe 1968). 6.2. Obsidian dating Obsidian, a variety of volcanic glass, was widely used for prehistoric tools. A fresh surface (as made by ancient man in chipping the material to form the tool)

Physics applied to archaeology I . Dating

977

slowly absorbs water and in the course of time there is a measurable hydration layer. T h e rate of growth depends on temperature, and consequently, in using the thickness of the layer as a measure of age, it is necessary to take this into account. Typical growth rates for different climatic regions have been established ; rather surprisingly the rate of growth is independent of the wetness of the burial circumstances. T h e chemical composition influences the hydration rate but fortunately most obsidians used for artefacts do not differ very much in this respect, T h e actual thickness of the layer for a 4000 year old specimen (as an example) is about 7 p m for a hot region such as Egypt but only about 1 p m for the Arctic. On theoretical grounds the thickness is proportional to the square root of the age. T h e hydrated layer contains about 3 o/o water which is about ten times the water content of fresh obsidian. Extensive application has been made in Mexico by Michels (1967) who points out that for a single site or region (for which the growth rate has been established by comparison with radiocarbon at one or two levels) the cheapness of the technique makes it extremely advantageous. It is particularly useful in testing the validity of the stratigraphy deduced for a site, or for segregating materials from a poorly stratified site into chronological groups. Michels has found that for similar samples buried in similar conditions nearly all the observed variation in hydration rim thickness is attributable to measurement error (of which the typical standard deviation is about 0.07 pm). A complication with the technique is the possibility of archaeological re-use of old material, though it is sometimes possible to detect the separate hydration rims corresponding to separated periods of burial. An account of the method has been given by Friedman et al. (1969). 6.3. Glass layey counting Some ancient glass exhibits irridescence and studies of glasses from Nineveh by Brewster (1863) showed this to arise from diffraction effects associated with thin weathering layers on the surface. For glasses which have a poor corrosion resistance and which have been buried in moist conditions, the weathering may be sufficient to give rise to a crust which is sometimes several millimetres in overall thickness. Study of some well dated specimens by Brill and Hood (1961) revealed the remarkable fact that the number of layers within the crust was equal to the age, suggesting the possibility of layer counting as a powerful method of dating. I t is however only applicable to a rather restricted number of samples (Brill 1963, 1968). Roman and Byzantine glasses are generally too resistant to become sufficiently weathered and Egyptian glasses are usually found in too arid an environment; on the other hand, when the weathering is heavy the poor quality of the glass often gives rise to layer structures that are too irregular for reliable results. T h e validity of the technique has been questioned by Newton (1966, 1969, 1971) who points out that layered crusts can also be produced in accelerated weathering experiments in which the conditions (temperature, humidity) are kept constant. Also, layers are exhibited by ancient glasses which have been subject to negligible annual variation in environment, such as obtains on the bed of the sea. I n some examples quoted by Newton the number of layers varies from place to place in the same piece, and although the number is equal to the age in one or two places, elsewhere it is less. Newton suggests that the layers are produced at a fairly constant rate which is generally rather less than one per year and that the examples of apparently annual rings occur when the rate happens to be as fast as one per year.

978

M . J . Aitken

T h e mechanism that gives rise to the layers is not at all understood, but it seems possible that they result from internal strains set up by chemical change in which there is an associated expansion or contraction. Early studies (Newton 1971) showed that in the weathering crusts the alkali of the glass had been replaced by water, the degree of hydration being about 200,:. Periodic variations in the silicon and calcium concentrations have been observed with an electron probe microanalyser (Shaw 1965), the peak to peak spacing being approximately 6 to 8 pm. T h e layer thicknesses reported from microscopic examination by various workers in this field in general lie in the range 0.5 to 20 pm.
7. Archaeomagnetism

7.1. Introduction T h e magnetic declination D (the angle between magnetic north and geographic north) changes slowly; at the present time, for London, D becomes less westerly by roughly 1" every decade. T h e angle of dip or inclination I (the angle between the geomagnetic lines of force and the horizontal) also changes and this secular zariation of the geomagnetic direction has been recorded for London over the past four centuries from observations on suspended magnetized needles (see figure 9). This

Figure 9. Secular variation for London as recorded from direct observations over the past four centuries.

information is taken from Bauer (1899), who collated early magnetic data from various parts of the world, and from later records (for instance, Kaye and Laby's Tables of Physical and Chemical Constants). Besides the written record, this information is also stored in baked clay which has remained in position on cooling down from firing by means of the thermoremanent magnetism (TRM) which is induced by the geomagnetic field in iron oxide minerals in the clay as it cools from 700 "C. This means that the baked clay acquires a weak but permanent magnetization in the same direction as the field at the time of cooling (and in the case of successive firings it is the last cooling that is effective). Clay has been used for thousands of years in the construction of hearths, ovens and kilns. Hence whenever such a structure can be dated by reference to archaeological chronology, or by radiocarbon or thermoluminescent dating, there is the opportunity to obtain otherwise irretrievable geophysical information. Conversely,

Physics applied to archaeology I. Dating

979

once the secular variation curve has been thus extended backwards into archaeologically interesting periods, the direction of remanent magnetization found in a structure can be used to establish its date. It is also possible to determine the ancient intensity of the geomagnetic field by measurements on baked clay. This because the strength of the magnetization acquired is proportional to the field intensity. If a sample is reheated and allowed to cool in the present day field, the ratio of the original magnetic moment to the new moment gives the ratio of the ancient intensity to the present day intensity. Since records of direct measurement of intensity are only available over the past two centuries this type of archaeomagnetic data is of particular interest geophysically as well as in connection with radiocarbon dating (see $2.4.2). Reliable results are not easy to obtain owing to the possibility of mineralogical changes occurring during the reheating. T h e necessary techniques and criteria for reliable results were established by Thellier and Thellier (1959) and their first results showed that substantial variations had indeed occurred in the past-they found that in Roman times the intensity in France was some 60% higher than its present day value. For intensity measurements it is not essential that the baked clay should have remained undisturbed. Bricks and tiles can be used, also fragments of pottery. It is also possible to use bricks and tiles for determination of the angle of dip since the requirement of stacking in a kiln usually ensures that one face or edge is horizontal (and irregularities in the stacking can be averaged out by measuring a large number of samples). T h e angle of dip can similarly be found from whole pots if the form is such as to necessitate that they should be baked standing upright. This is the case, for example, with Chinese Yueh ware on account of the heavy glaze and ornamentation, but unfortunately the more usual technique in other cultures was to pack the pots in the kiln rather irregularly. T h e present state of archaeomagnetism derives mainly from the work of Professor E. Thellier at LInstitut de Physique du Globe in Paris from 1933 onwards (and now at L e Laboratoire de GComagnCtisme du Parc-Saint-Maur). Archaeomagnetism has developed in parallel with and as part of the much wider field of paleomagnetism. T h e former is usually taken to signify some involvement with ancient man whereas the latter covers all geophysical and geological applications. T h e fossilized magnetic record in rocks is used to give information about the past behaviour of the geomagnetic field on a geological time scale and also to make deductions about the formation and drifting of the earths crust (see, for example, Irving 1964; for shorter accounts, Blackett 1956 or Doell 1969 a). 7.2. Remanent magnetism Most clays contain the order of 5:/, of iron oxide dispersed in the form of fine grains, and it is usually assumed that these are sufficiently far apart for mutual interaction to be unimportant. T h e resulting magnetic properties are conveniently discussed in terms of the type minerals, magnetic (Fe,O,) and haematite (a Fe,O,); for details the reader is referred to articles by NCel(1955), Stacey (1963) and Dunlop and West (1969). T h e last named discuss in particular the effect of grain interaction. Although the main features of remanent magnetism can be understood by regarding these minerals as ferromagnetic, magnetite is in fact ferrimagnetic and haematite has an antiferromagnetic lattice structure with a weak (parasitic) ferromagnetism. I n a ferromagnetic, such as iron, exchange interactions cause

980

M. J. Aitken

parallel alignment of the atomic magnetic moments, whereas in a ferrimagnetic there are two interwoven sublattices, usually designated A and B, which interact negatively. Thus the atomic moments on A sites are aligned parallel to one another but antiparallel to those on B sites; however, if the magnetizations of the two sublattices are unequal there is a net magnetic moment. For magnetite the saturation magnetization is 92 emu 8-l (corresponding to 4pB per molecule). I n an antiferromagnetic the sublattices also interact negatively but as the magnetizations are equal in this case there is no net moment and the material behaves as a paramagnetic. However, in haematite there is a parasitic ferromagnetism even when it is pure and it exhibits weak ferromagnetic properties with a saturation magnetization of around 0.5 emu g-l. -4 grain of magnetite is not spontaneously magnetized as a whole because of the strong demagnetizing fields that would result; instead the grain divides into domains, within each of which the magnetization is uniform, and which arrange themselves so that the magnetic flux forms closed paths as far as possible within the magnetic material. T h e domain size and arrangement are determined by the condition of minimum total energy (to which the contributing types are crystalline anisotropy energy, domain wall energy and magnetostrictive strain energy, in addition to the exchange interaction energy and the magnetostatic energy already mentioned). I n magnetite the domain size is upwards of 0.03 pm. I n haematite, because the saturation magnetization is smaller by a factor of 1000, the domain size can be as large as a millimetre. Consequently, whereas the grains of magnetite are 'multi-domain grains', those of haematite are ' single-domain grains'. T h e net magnetization of a multi-domain grain in an applied field arises from movement of domain walls such that domains in the direction of the applied field grow at the expense of domains which are opposed to it; for magnetite the coercive force is the order of ten to a hundred oersteds (being largest for small grains). For an assembly of single-domain grains the net magnetization in an applied field arises from domain rotation from one direction of easy magnetization (as determined by the crystalline anisotropy and by the shape of the grain) to another; for haematite coercive forces of the order of several thousand oersteds are commonly observed.

7.2.1. Thermoremanent magnetism (TRM). According to the classic exposition of NCel (1955) the acquisition of thermoremanent magnetism (TRM) is to be understood in terms of the blocking temperature, T,. Above TB the magnetic domains are able to realign as dictated by the direction of the external magnetic field, while below it they are 'frozen'. T h e value of TB depends on the material and grain size; TB is of course lower than the Curie point (565 "C for magnetite and 675 "C for haematite), the temperature at which the alignment of the atomic moments in domains disappears. For the single-domain grains of haematite the T R M may be understood in comparatively simple terms. I n the case of ellipsoidal grains the direction of easy magnetization is along the major axis and the grain is spontaneously magnetized either parallel or antiparallel to that axis, Reversal from one state to the other becomes possible when the thermal agitation energy of the grain is comparable with the potential energy of the constraints tending to keep the magnetization along the preferred axis of the grain. T h e probability of reversal is a very sharply rising function of temperature and, by definition, above T, there is a high probability of reversal within a few seconds, while below T, reversal is only likely within a very

Physics applied to archaeology I . Dating

981

long time. Above TB the behaviour of the grains is paramagnetic. I n the absence of an external field there is no preferential direction but when a field is applied the probability for each grain to be in the state that has a component in the same direction as the field is higher than for the reverse situation. Hence there is a net magnetization in the same direction as the field. When the temperature falls below TB this magnetization is retained although the applied field may be altered. Because the probability of reversal is dependent on the size of the grain (the relaxation time being exponentially dependent on the volume of the grain) a sample of baked clay exhibits a continuous spectrum of blocking temperatures. This gives rise to Thellier's law of partial T R M : the remanent magnetization acquired within a given temperature interval (q to T,) resides in a group of grains having blocking temperatures in that interval, and, on demagnetization by heating and cooling in zero field (see 57.3.3) the same group is unaffected if the temperature does not reach but totally demagnetized if the temperature reaches T,. Thus it is possible for a single specimen to carry more than one T R M direction. Suppose that after acquiring an initial T R M by cooling from 670 "C or above, the specimen is subsequently reheated to say 400 "C with a different field orientation. Grains having blocking temperatures below 400 "C will realign to the new field direction but grains with higher blocking temperatures will retain the primary direction. T h e net T R M is then the vector sum of the primary and secondary components but by successive partial demagnetizations (cooling in zero field from successively increasing temperatures) each may be evaluated both in magnitude and direction-as the demagnetizations proceed the net T R M direction gradually swings round to that of the primary component. Archaeological examples of composite TRNI are provided by pots that have been reheated in the process of cooking and bricks that have been part of a building destroyed by fire. Typical magnetization and demagnetization curves for baked clay are shown in figure 10. Using the parameters suggested by Stacey (1963) grains which become demagnetized by heating for 1 hour at 200 "C (for example) have a relaxation time of greater than 200000 years at 20 "C. Hence for the specimen of figure 10 at least 90% of the T R M is stable over archaeological times; in practice any unstable component can be removed by successive demagnetization to increasing temperatures until the remanent direction stays the same. Although the intensity of magnetization within a grain is unaffected by the applied field, the net T R M acquired by an assembly of dispersed grains is proportional to it. This is because, although there is a high probability of domain reversal at TB, the fractional alignment achieved by the field is far from completethe difference in magnetic potential energy, @', between the parallel and antiparallel states being small compared to kTB. T h e fractional excess of domains with a component in the field direction is effectively equal t o (pF/3kTB) where p is the magnetic moment of the grain and F the applied field. Thus the T R M of a sample of baked clay is proportional to the field intensity in which it cooled down and archaeomagnetic measurements can be used to determine the ancient geomagnetic field in strength as well as in direction. T h e actual magnitude of the T R M is usually in the range emug-l and is large compared to the instantaneously to induced magnetization. T h e theory of the T R M acquired by multi-domain grains has been developed by Stacey (1963) and by Dickson et al. (1966). It is less straightforward but it leads to the same pattern of behaviour as does the single-domain theory, though with

982

M . J . Aitken

quantitative difference. O n the other hand, according to Verhoogen (1959) and to Kobayashi and Fuller (1968), the T R M of multi-domain grains really resides in restricted regions displaying properties analogous to those of single-domain grains. 7.2.2. Other types of remanent magnetization. Grains with low blocking temperatures are not sufficiently stable at ordinary temperatures to retain their original T R M over thousands of years; these short relaxation time grains tend to follow changes in the ambient field direction and this component of the magnetization is called aiscous remanent magnetism (VRM). It can be removed by thermal demagnetization (as mentioned above) or by the application of alternating field demagnetization (see $7.3.3). Experimentally, it is found that the growth (and the decay) of VRM is

i
io
80 60

--_-Demagnetization

.-

.-

2 F

k 40

5 c

\ \
\

20t

Magne t izat ion \

Figure 10. Acquisition of TRM-idealized curve for baked clay. The full curve shows the magnetization (measured at 20 "C) acquired on cooling in a magnetic field from the temperature indicated by the horizontal scale. After final cooling from above 675 "C the sample has been gradually demagnetized by heating to successively increasing temperatures in zero field-as indicated by the broken curve. The shape depends on the composition of the sample but as long as there is no mineralogical change during heating the demagnetization curve is always the mirror image of the magnetization curve; that is, the T R M gained in a certain temperature interval during magnetization is equal to that lost in the same temperature interval during demagnetization, as illustrated for the interval 500 to 675 ' C .

proportional to the logarithm of time; this has the fortunate consequence that the long term VRM can be estimated from fairly short term observations. For instance, the change observed between t = 10 minutes and t = 14 days is equal to the change between t = 14 days and t = 90 years or between t = 90 years and t = 200000 years. Thus as an alternative to the demagnetization procedure it is possible to estimate the likely strength of any VRM in a sample by observing the change in net magnetization during a few weeks of laboratory storage. Fortunately, as indicated in figure 10, the VRM component in typical archaeological samples is nearly always less than 10% of the total and usually less than 50/,. VRM grows slowly with time, being governed by the statistical probability of the thermal agitation energy of the domain exceeding the potential barrier of the constraining forces. This is at ordinary temperatures and weak fields (a few oersteds or

Physics applied to archaeology I. Dating

983

less). On the other hand, in strong fields of a hundred oersteds or more, a sample may acquire remanent magnetization because the applied field exceeds the coercive force of some of its grains. This isothermal remanent magnetism (IRM) is acquired instantaneously and does not decay when the field is removed. For the singledomain grains of haematite, fields of several thousand oersteds are necessary but for the multi-domain grains of magnetite several hundred oersteds are enough to produce an IRM. An I R M can be produced in baked clay by the magnetic field associated with a flash of lightning but fortunately, like VRM, I R M can be removed by application of alternating field demagnetization. Lightning may also produce anhysteritic remanent magnetization (ARM), the effect of a strong alternating field in the presence of a weak steady field. ,4RM may also be inadvertently produced in the course of alternating field demagnetization if the earth's field is not precisely annulled. If a chemical change occurs in the magnetic mineral at a temperature below the blocking temperature of the newly formed grains, then a chemical remanent magnetism (CRM) may be acquired. This is an interfering effect in paleomagnetism but fortunately it does not appear to be of importance in archaeomagnetism. Detrital or depositional remanent magnetism (DRM) occurs in sediments deposited in calm water. T h e magnetic particles present inherit a remanence from the rocks from which they have been eroded and tend to be aligned by the geomagnetic field as they settle. As mentioned in 9 7.3.4, the D R M of Swedish glacial varves has been considered as a source of archaeomagnetic data.

7.3. Directional measurements 7.3.1. Sample collection. I n assessing the suitability of a baked clay structure (such as a kiln, oven or hearth) the prime consideration is whether any movement has occurred since baking. Thus in a pottery kiln reliable samples would be expected to come from the floor and the lower part of the walls rather than from any of the superstructure. Before any sample is extracted for measurement it is of course vital that its exact orientation should be marked on it. This is done by partially encasing it in gypsum plaster (plaster of Paris) before detaching it from the structure. T o do this a miniature square 'ditch' is dug leaving the sample as a stump in the middle, about 10 cm by 10 cm. A square metal frame is placed in the ditch and levelled. Plaster is then poured in and the top surface smoothed off flush with the top of the frame. When the plaster has set, a line is sighted from a nearby theodolite and marked on the top surface. T h e theodolite orientation is found by shooting the sun at a known time and using suitable astronomical tables. Hence the azimuth of the line of the plaster surface is found with respect to true (geographic) north; the sample's orientation with respect to the horizontal is provided by the plaster surface, When the marking is complete the sample is detached by cutting away underneath the frame and transported to the laboratory for measurement. Although the precision of measurement (and of marking) is better than 0.5", the scatter in the individual directions found in samples from the same structure is often of the order of 5" or 18". This makes it important to take at least a dozen samples from each structure. T h e most obvious cause of the scatter is that irregular subsidence has taken place, but there is also evidence that it can be due to distortion of the geomagnetic field by the magnetism of the structure itself (Harold 1960, Weaver 1961, 1962, Aitken and Hawley 1971).

984

M . J . Aitken

7.3.2. qeasurement techniques. T h e astatic magnetometer consists essentially of two bar magnets of equal strength fixed rigidly at either end of a rod about 10 cm long (figure 11) and suspended by a fine fibre of phosphor-bronze or quartz. Since the two magnets are antiparallel the net torque due to the earth's field is zero. On the other hand, the torque on the lower magnet due to a small sample positioned beneath it is substantially greater than that on the upper one and the resultant deflection is a measure of the horizontal component of the magnetization of the sample. T h e classic description of this type of instrument, and consideration of its fundamental sensitivity capabilities, has been given by Blackett (1962). A good quality instrument can detect samples having an intensity of magnetization of the order of 1O-' emu ~ m - ~ .

S
N

Sample

Figure 11. Principle of the astatic magnetometer. Restoring torque due to the earth's field is eliminated by using magnets of exactly equal moment. The sample causes a net torque, however, because it is nearer to the lower magnet than to the upper one.

For archaeomagnetic work the astatic magnetometer has the drawback that for nonuniformly magnetized samples the sample dimensions need to be small compared to the distance from the lower magnet. Samples extracted from archaeological structures are usually inhomogeneous and it is not practical to obtain small samples because of their weak mechanical strength and because of the difficulty of marking an accurate enough direction on a small sample. On account of these considerations Thellier (1938) developed a magnetometer based on electromagnetic induction; the sample is rotated inside a coil system and its magnetization is determined by measurement of the induced voltage. Early measurements were made by observing the throw of a ballistic galvanometer consequent on 180" rotation of the sample but greater sensitivity is obtained by using continuous rotation (typically at 5 Hz). Essentially, the amplitude of the induced voltage serves to determine the component of the TRM in the plane of rotation and the phase of the voltage with respect to the driving shaft yields the azimuth of that component. T h e continuous rotation system is now usually known as the 'spinner magnetometer'. I n its more sophisticated forms fairly complex coil configurations are used. T h e objectives of these can be best illustrated by the straightforward system shown in figure 12. Firstly, in order to avoid interference from external magnetic fields due to moving iron objects (such as passing cars) a pair of compensating coils is connected in series-opposition with the detecting coils ; both sets have the same number of area-turns so that the voltages induced in each set by a distant source

Physics applied to archaeology I. Dating

985

annul one another. As in the case of the upper magnet of the astatic pair, the outer set is comparatively remote from the sample, and the reduction in sensitivity due to the compensating coils is small. Secondly, the detecting coils form a Helmholtz pair so that there is a region within which the detection sensitivity is uniform (a sphere of 20 cm diameter in the case of the 60 cm diameter coils used by Thellier). This uniformity of sensitivity is important when intensity measurements are being made although the direction is determined correctly without uniformity as long as the electronics is such that only the fundamental frequency is accepted.

Compensating coils

Spherh of uniform sensitivity

Figure 12. Simple compensated Helmholtz coil system for spinner magnetometer. The two outer coils are wound in opposite senses to the two inner ones, thus eliminating the effect both of uniform magnetic disturbances, and of nonuniform disturbances of constant horizontal gradient.

A spinner magnetometer suitable for archaeomagnetic samples and capable of accommodating samples having dimensions up to 30 cm has been described by Aitken et al. (1967). An alternative to detection by pick-up coils is to use fluxgate detectors and a spinner magnetometer of this type suitable for small samples (2.5 cm cores) having a minimum detectable moment of order of emu has been described by Foster (1966). Magnetometers based on nuclear magnetic resonance (NMR) are not suitable for this type of measurement because they require the field due to the sample to be fairly uniform over the volume of the detector. This means that the sample would have to be at an appreciable distance from the detector and in consequence the sensitivity obtainable would be poor. At the other end of the scale of sophistication is the technique in which the sample is suspended on a fibre (or floated on a pool of mercury) and allowed to orientate itself in a uniform magnetic field; although simple the procedure is rather slow. A comprehensive account of measurement techniques will be found in Methods in Paleomagnetism (edited by Collinson et al. 1967).
7.3.3. Washing procedures. As has been mentioned in $7.2.2 a sample may carry other remanent magnetizations than the true T R M and before measurement takes place it is necessary, for reliable results, to rid the sample of any such additional magnetizations or at any rate to evaluate their importance. A small viscous (VRM)

986

M . J: Aitken

component is to be expected in most samples and, since this arises from grains with low blocking temperatures, the obvious way to remove it is the thermal demagnetization procedure in which the sample is heated to the blocking temperature corresponding to the 'hardest' VRM and cooled in zero magnetic field. A suitable nonmagnetic oven for use with the bulky samples of archaeomagnetism has been described by Aitken et al. (1967); the earth's field is annulled by a single set of coils aligned along the field direction and designed to reduce the field to within 1O 0 of its normal value over a volume of one cubic foot. T h e question of what temperature corresponds to the blocking temperature of hardest VRhI has been studied by Weaver (1964) who, in respect of archaeomagnetic samples of up to 2000 years old, finds that the major part of the VRM is removed by 100 " C but that there is a small further removal by heating to 200 "C. I n practice, unless there is a dramatic shift of direction on 'washing' to 100 "C, this temperature is satisfactory for routine measurements, samples being treated in batches of a dozen or more. T h e hardness of the VRM increases slightly with age but on account of the logarithmic law mentioned in 7.2.2 the contribution of the hard VRRI is relatively unimportant and the main part of the VRM is acquired in the last few hundred years. Viscous magnetization can also be removed by alternating field demagnetization; the earth's field is annulled as before and the sample is subjected to a strong alternating magnetic field which is smoothly and slowly reduced to zero. This randomizes domains having a coercive force less than the maximum value of the alternating field. T h e normal procedure is to use successively increasing maximum fields until the remanent direction ceases to shift. Because a fairly wide range of blocking temperatures may correspond to a given coercive force (Everitt 1961, 1962, Doell and Smith 1969) there may be appreciable removal of 'true' T R M before a field sufficiently strong to remove all the VRM has been reached; with thermal washing a much better separation of the two can be achieved. On the other hand, the alternating field technique is more effective in removal of IRA1 (due, for instance, to strikes of lightning-see Rimbert 1958). 7.3.4. Results and geophysical discussion. T h e first task in archaeomagnetism is the establishment of the secular variation curve for angle of dip ( I ) and declination (D) for periods when it is not available from directly recorded measurements. T h e curve so far obtained for England by the Oxford Research Laboratory for Archaeology is shown in figure 13; the curve is based on the results from some eightly archaeological structures, mainly pottery kilns, of which the details have been published elsewhere (Aitken and Weaver 1962, Aitken et al. 1963, Aitken and Hawley 1966, 1967). T h e data have been accumulated over a period of about eight years, the rate of progress being dictated by the limited number of relevant and suitable structures that are excavated each year. For the period AD 400 to AD 1000 and prior to AD 50 the number of structures so far available has not been sufficient to allow a reliable curve to be drawn. T h e portion of the curve that is based on Bauer's collation of recorded data (see figure 13) approximates to an ellipse. This led to the belief that the secular variation was a well behaved cyclic phenomenon with a period of about 450 years and this idea has persisted for a long time (indeed it still persists in some quarters today). Geophysical explanations of the cause of the secular variation are at variance with any notion of cyclic behaviour; according to the dynamo theory of the

Physics applied to archaeology I . Dating

987

origin of the earths magnetism the main field is due to a selfmaintained magnetohydrodynamic current system in the earths fluid core, and the secular variation arises from transient irregularities in the pattern at the core-mantle interface. A possible way in which these eddies in the current pattern may be generated is by relative motion through the core of small protruding bumps on the mantle at the interface (Hide and Malin 1970). T h e archaeomagnetic data of figure 13 give direct evidence that the belief in a well behaved cyclic variation is erroneous-in agreement with geophysical ideas-though, as we shall see later, a comparison of archaeomagnetic data from different parts of the world suggests that there may be some systematic features present.

Figure 13. Secular variation for London. T h e archaeomagnetic data are shown with a dashed line. All dates indicated are AD.

Although the secular variation is considered to arise from transient irregularities, the westward drift phenomenon allows predictions about the past secular variation to be made from the present spatial irregularities of the geomagnetic field and archaeomagnetic data are relevant to checking these predictions. Analysis of worldwide magnetic data recorded over the past half century (Bullard et aE. 1950, Bullard 1958) shows that the main features of the geomagnetic field have been drifting westward at a rate of about 0.2 degrees of longitude per year. This is interpreted as due to relative motion between the outer part of the core and the mantle-the mantle rotates faster than the core so that magnetic features associated with irregularities in the current pattern in the core appear to an observer on the mantle to be drifting westward. Hence the secular variation observed at a particular station should be predictable from the spatial irregularities along the line of latitude going through the station. Obviously, if the form of the irregularity remained unchanged for long enough then the secular variation would repeat itself each time the mantle made one revolution relative to the core. However, the lifetime of the irregularities is expected to be only of the order of a few hundred years and so the correlation

988

M . J . Aitken

should not extend back for more than that length of time. For London there is in fact a credible correlation for the past few centuries, as was pointed out by Bauer (1899), but before that the archaeomagnetic data show-as expected-a complete breakdown of correlation (Aitken and Weaver 1964). These authors also point out that for the three centuries subsequent to AD 1000 the motion of the curve is anticlockwise rather than clockwise, and that this could be indicative of eastward drift. However, the obvious way to derive information about the direction of the drift is by comparison of archaeomagnetic data from different parts of the world. Comparative data is at present rather too sparse for a clear cut conclusion but the analyses of Yukutake (1967) suggest that for the past thousand years the drift has always been westward. This is also the viewpoint of the Soviet investigators Burlatskaya et al. (1969) who point to the observed retardation of the last maximum of the angle of dip as one moves from Japan (last minimum AD 1200-1400) to Britain (AD 1700). An alternative interpretation of the data is made by Kawai et al. (Kawai, Hirooka and Sasajima 1965, Kawai and Hirooka 1967, Kawai, Hirooka and Tokeida 1967). Although the archaeomagnetically observed secular variation curve observed for Japan bears little resemblance to that for Britain, when the data are expressed in terms of 'virtual pole positions' some common features become apparent (see figure 14). T h e virtual pole position for an observing station is the orientation of a dipole placed at the centre of the earth which would produce a magnetic field at the station having the observed values of I and D. Using the archaeomagnetic data for Britain, Japan, Arizona (Watanabe and D u Bois 1965) and Iceland (Brynjolfsson 1957) Kawai et al. (1967) have calculated the motion of the virtual pole that best fits all four sets of data. From this they propose as a working model that over the past 2000 years the representative dipole has executed a quasi-hypotrochoidal movement around the rotational axis of the earth which is the combination of an anticlockwise motion of approximate amplitude 11" and period about 1500 years and a clockwise motion of amplitude 7" and period 400 years. However, there is still the possibility that the approximate agreement between the pole position motions of the four stations may only be coincidence and more comprehensive data will be needed for a conclusive test of the hypothesis. T h e significance of the model is that it implies that a major part of the secular variation arises from a 'wobbling' of the main current pattern in the core, rather than from localized irregularities. T h e model is not supported by results obtained from volcanic lava flows on the island of Hawaii (Doell, 1969 b): these show a remarkably small secular variation and, although they probably do not extend into the period of the past 2000 years on which Kawai's model is specifically based (the flows are as yet undated), they do indicate that for a substantial period of time the main current pattern stayed steady in direction. T h e information referred to above was obtained from pottery kilns for Britain and Japan, Indian fire pits for Arizona and lava flows for Iceland and Hawaii. Archaeomagnetic studies have also been made on the lava flows of Mount Etna both by Chevallier (1928) and by Tanguy (1970). Comprehensive archaeomagnetic studies have been carried out by Thellier in France and Tunis (Thellier 1966). Other reported archaeomagnetic directional work has been done in Bulgaria (Kovacheva-Nozharova 1968), Arizona (Du Bois 1967)' Egypt (Du Bois 1969), Greece (BelshC and Cook 1963, Bammer 1969) and Japan (Watanabe 1959). Most of this work has utilized in situ baked clay thereby allowing the determination of both I and D. Because of the regular way bricks are stacked while being

Physics applied to archaeology I. Dating

989

baked, it is possible to use them for finding the ancient value of I as has been demonstrated by Thellier (1936, 1938). Extensive work using bricks has also been carried out in the USSR (see, for example, Burlatskaya et al. 1970); spectral analysis of this data suggests the probable presence of a 600 year periodicity with the possibility also of 100 year, 350 year and 1000 year periodicities. It is also possible to
Virtual pole Britain and J a m

270"

0" -Britain ---Japan

Figure 14. Virtual pole positions from archaeomagnetic data for Britain ( ) and Japan (- - - -). T h e virtual pole position for an observing station is the orientation of a dipole placed at the centre of the earth which would produce a magnetic field having the observed values of 1 and D for the station. Although the pole positions for the two countries do not superimpose, it is to be noted that there is a broad similarity in the sense of movement.

deduce the ancient value of I by measurements on a vase-as was first demonstrated by Folgheraiter (1899)-but only when the ornamentation or glazing suggests that it must have been baked standing on its base on a horizontal surface. This is not the case for most archaeological pottery but results have been reported by llitken (1958) for Chinese Yueh ware and by Clarke and Connah (1962) for British beaker pottery of the Late Neolithic and Early Bronze Ages.

990

M . J . Aitken

Another possible source of archaeomagnetic data is the depositional remanent magnetism (DRM) carried by glacial varve sediments. Since varved clays can be accurately dated by counting the annual layers they are potentially a very valuable record. Unfortunately the remanent direction appears to be influenced by factors other than the magnetic field direction (Griffiths et al. 1960, Rees, 1961, King and Rees 1966). However, the declination is faithfully preserved as long as the deposition occurs from fairly calm water; measurements of Swedish varves in the periods 1100 BC to AD 750 and AD 1300 to AD 1900 have been reported by Griffiths (1955). 7.3.5. Application to dating. As archaeomagnetism stands at present the emphasis is on the geophysical interpretation of the information obtained, rather than on archaeological dating. There are a number of reasons for this: (i) Because the secular variation is different in different places (except in so far as its broad features may be interrelated in the manner suggested by Kawai) the reference curve needs to be set up for each region of (say) 500 to 1000 miles across in which it is to be used for dating. Besides requiring time and effort, strictly speaking, this setting-up presupposes an adequate archaeological chronology and if this be the situation the eventual provision of a magnetic dating service is redundant {but see below). (ii) T h e recurrence in time of the same pair of values of I and D leads to ambiguities (as for England in the twelfth and fourteenth centuries-see figure 13). (iii) There are periods when the change in direction is rather slow (for example, Roman times in England). (iv) Although for some structures (and particularly when the samples are from a solid floor) the accuracy is good and the standard error of the mean direction may be less than one degree there are a significant proportion of cases for which the standard error is three or four degrees; this is comparable with the change in geomagnetic direction that occurs over a century. These errors are not due to lack of measurement precision but arise from the inherent scatter mentioned in $7.3.1. However, it would be wrong to give the impression that magnetic dating has no utility in archaeology, either actual or potential. The paucity of magnetic dates is to some extent due to unavoidable delay in getting the system established. T h e situation is not usually as clear cut as implied in (i) above. It may be that the archaeological chronology gives a few fixed points and magnetic dating can be used to interpolate, it may be that magnetic dating can be used to decide between alternative archaeological chronologies, or it may be that magnetic dating is a means of transferring the known chronology of one part of the region to another part in which the chronology is unknown. The development of a magnetic dating system calls for close cooperation and mutual understanding between scientist and archaeologist. T h e magnetic direction recorded in a pottery kiln should, to some extent, be regarded as an additional stylistic attribute which takes its place along with the more conventional typological evidence in the jigsaw of evidence which the archaeologist has to fit together in building his chronology. I n England attention is at present concentrated on the post-Roman period, particularly medieval and post-medieval. For a non-archaeologist it is surprising to find that the chronology of pottery types in these periods is far from well established (in contrast to the Roman period) and there is plenty of scope for the application of magnetic dating, as has been discussed by Hurst (1963, 1966).

Physics applied to archaeology I . Dating

99 1

Magnetic dating results in England will be found in excavation reports from the following sites : (i) Potovens, Wakefield (Brears 1967) (ii) Bagots Park, Staffordshire (Crossley 1967) (iii) Potterspury, Northants (Mayes 1968) (iv) Purwell Farm, Oxon (Arthur and Jope 1962-3) (v) Upper Heaton, Yorks (Manby 1964) (vi) Rainsborough, Northants (Avery et al. 1967) (vii) Potterton, Yorks (Mayes et al. 1966) (viii) Boston, Lincs (Mayes 1965) (ix) Grimstone, Suffolk (Smedley and Owles 1961) (x) Cox Lane, Ipswich (Smedley 1963) (xi) Blundens Wood, Surrey (Wood 1965). 7.3.6. Reversals. T h e evidence from paleomagnetism indicates that for certain periods of the geologic past the earths field has been reversed with respect to its present direction (Cox 1969, Doell 1969 a). T h e present polarity epoch is termed the Brunhes and this was preceded by the Matuyama reversed epoch, the transition occurring around 0.7 million years ago. The lengths of the last four polarity epochs (Brunhes, Matuyama, Gauss and Gilbert) have been of the order of a million years, but during an epoch there is the occasional occurrence of polarity events lasting for 10 000 to 1000 000 years. T h e act of reversing is estimated to take about 5000 years. T h e events having possible relevance to archaeomagnetism are the Olduvai (at 1.9 millions years), the Gilsa (at 1.6 million years), the Jaramillo (at 0.9 million years), the Blake (at 110 000 years) and the Laschamp (starting sometime later than 40000 years ago and ending sometime between 9000 and 20000 years ago). T h e Olduaai event is a short period of normal (that is, present day) polarity in the middle of the Matuyama; as implied by its name, evidence for this event was first obtained from lava underlying the volcanic ash in which the hominid remains at Olduvai were found. Subsequently the reversed magnetization corresponding to this event has been found in lavas in other parts of the world, the synchronism being established with potassium-argon dating (and as emphasized in 3 3.1 this dating technique has been vital in showing that the reversed magnetization really represents a reversal of the geomagnetic field rather than being due to the lava having abnormal magnetic properties). T h e .Blake event is estimated to have lasted for only 8000 years and it has been detected only recently, from measurements on deep ocean cores (Smith and Foster 1969). T h e Laschamp event, found in four lava flows in the Chaine-des-Puys in central France (Bonhommet and Babkine 1967, Bonhommet and Zahringer 1969), has not yet been detected elsewhere and so is probably of short duration also. I t is possible there are additional events not yet detected; an interesting extension of archaeomagnetism would be the detection of these events in Paleolithic hearths and their use as time markers. I t is to be noted that, in conflict with the end date proposed for the Laschamp event, the preliminary archaeomagnetic results obtained by Du Bois (1969) for hearths in Egypt of the period from 10000 years ago to 22 000 years ago all indicate normal polarity. Similar indication from the period of between 2000 years ago and 10000 years ago has been obtained by Du Bois from North American material.

992

M. J . Aitken

Reference has been made earlier (8 2.3.2) to some anomalous radiocarbon results in the dating of lake sediment (at around 20000 years ago) and the possibility that it is associated with the Laschamp event. A reversal in geomagnetic polarity is thought to occur by the decay to zero of the geomagnetic moment in one direction and its growth in the opposite direction, rather than by its keeping a fixed amplitude and swinging round. Consequently, during the act of reversal the earth loses its magnetic shield and the cosmic ray intensity temporarily increases, with associated increase in radiocarbon production. I t has also been suggested that the increased cosmic ray intensity during reversal might cause a rise in biological mutation rates with a consequent discontinuity in the evolutionary process (Uffen 1963). However, as pointed out by Harrison (1968), whereas the fairly low energy cosmic rays responsible for radiocarbon production in the upper atmosphere are strongly shielded by the geomagnetic field, the high energy component which reaches the surface of the earth is not much affected. For instance, whereas at the top of the atmosphere the cosmic ray intensity is reduced by a factor of about ten on going from pole to equator, at ground level the change is only 14%. I n addition, cosmic rays are responsible for somewhat less than half of the total radiation dosage received by an organism on earth so that the effect of any change tends to be diluted. Also, radiation is by no means the main agency responsible for mutations. Harrison suggests that reversals may affect evolutionary processes indirectly through climatic change, and that this may account for the evidence of correlation found in deep sea cores (Opdyke et al. 1966). 7.4. Intensity measurements 7.4.1. Technique. As outlined in $7.2.1 the T R M acquired by baked clay is proportional to the intensity of the geomagnetic field in which it cooled so that
(7.1)

where MA is the moment that the sample acquired by cooling in the ancient field F,, and ML is the moment acquired when the sample is reheated and cooled in the present day laboratory field FL. T h e fundamental difficulty with this technique is that during the laboratory reheating mineralogical changes may take place. If so, the effective sample that carries ML may have T R M properties different from the original sample carrying MA, and equation (7.1) is not valid. Such changes may be an actual chemical oxidation or reduction (for example, between magnetite and haematite) or merely an alteration in the condition of the magnetic grains. Heating in air or in an inert atmosphere appropriate to the oxidation state of the sample is helpful but is far from being a complete solution to the problem. Consequently, reliable results can be obtained only if careful tests for stability of T R M properties are incorporated in the experiment and samples thereby found unsatisfactory are rigorously excluded. I n the step by step heating method of Thellier and Thellier (1959) the sample is successively heated twice to each of a set of increasing temperatures (for instance, 100 "C, 200 "C, 300 "C, etc.). By reversing the orientation of the sample in between the first and second heating to each temperature it is possible to evaluate both the ancient and the new T R M corresponding to each temperature interval. I n the absence of any change in the magnetic characteristics

Physics applied to archaeology I. Dating

993

of the sample, the ratio of ancient T R M to new T R M is the same for all temperature intervals; if a significant change is observed the sample is discarded. Alternatively the first heating (and cooling) is carried out in zero magnetic field. Comment and discussion about these techniques has been given, for instance, by Smith (1967) and by Weaver (1966, 1970) and descriptions of practical utilization by Nagata et al. (1965), Du Bois and Watanabe (1965), Sasajima (1965) and Bucha (1967). It appears that successful application is most to be expected when highly oxidized samples are used. An encouraging development by Allen (1970) is a technique which incorporates alternating field demagnetization and avoids the need to heat above 400 C, thereby eliminating most of the risk of minerological change.
I I

2000

AD

BC

2000

Figure 15. Variation of geomagnetic field strength for Czechoslovakia and Central America, from archaeomagnetic measurements by Bucha (1970). The ordinate scale shows the ratio of the ancient field (FA) to the present day field (FL).

7.4.2. Results and geophysical discussion. T h e geomagnetic intensity at a given observing station may vary for any one of three reasons. (i) A change in the strength of the main current system in the core, i.e. a change in the magnetic dipole moment of the earth. (ii) A change in the orientation of the main current system, i.e. a wobble in the orientation of the dipole-as suggested in Kawais model (see $7.3.4). ( E ) An alteration in the local non-dipole field arising from irregularities of the current pattern near the surface of the core. These distinctions are of particular interest in considering the effect on radiocarbon production (see $2.4.2) since only (i) will produce such an effect. Reviewing intensity data published up to 1966 from France, Japan, Russia and Bolivia, Smith (1967) concludes that there is good evidence for (i) and that in the past 2000 years the dipole strength has decreased by SOY, of its present value-as had been suggested by the original Thellier and Thellier (1959) measurements. Recent results obtained by Bucha (1970) for Czechoslovakia and Central America are shown in figure 15. These are based on a larger number of samples than any of the earlier

994

M . J. Aitken

work and it appears that superimposed on the long term variation attributable to (i) there is a short term secular variation associated with (iii). I n particular Bucha points to the peaks that occur at 400 BC and AD 800 in Czechoslovakia, but not until AD 100 and AD 1200 for Central America, as evidence for the existence of westward drift of the non-dipole field. I n considering the effect on radiocarbon production it is usual to fit a sine wave to the intensity data (see figure 3 for example). The maximum around 2000 years ago is based on measurements well spread around the world and the evidence for a lower than at present value in the third millennium BC being worldwide is quite good (for a summary, see Xthavale 1969). On the other hand, the evidence for a higher than at present value in the seventh millennium BC (the belief that the variation is sinusoidal rests heavily on this) is based on a single site-namely Catal Hiiyuk in Anatolia (Bucha and ICIellaart 1967)-and so it is possible that the high value is due to (ii) and (iii) rather than representing a genuine increase in the geomagnetic movement, (i). However, there is supporting evidence in the results reported for lava flows from Hawaii (Smith and Doell 1969) and although these flows are as yet undated they do show an intensity variation that is not inconsistent with the other results. This is particularly significant because Hawaii is in a region of remarkably weak secular variation and this rules out attribution of the intensity changes to either (ii) or (iii). 7.4.3. Application t o dating. Since the overall variation in intensity is about = 500,d of the present value and the precision of measurement is in the range look to + 15% it is hardly to be expected that intensity dating will provide more than a rough estimate. However, the sharp changes observed by Bucha (figure 15) suggest that in some periods and localities the method may be quite useful. For the purpose of dating the cause of the variation (i), (ii) or (iii) is irrelevant. There is the advantage over directional measurements that the sample does not need to be found in situpottery fragments can be used-as long as the general locality of origin is known. There is also the possibility that intensity measurements would allow ambiguities in directional results to be resolved (to distinguish, for instance, between AD 1200 and AD 1400 in England-see figure 13).

References
1. Introduction BRUNHES, B., 1906, J . Phys. Paris, 4th series, 5, 705. G., 1899, Arch. Sci. phys. nut. (Genkve), 8, 5-16. FOLGHERAITER, P., 1907, Arch. Sci. phys. nat. (Genkve), 23,467-82. MERCANTON, 2. Radiocarbon BARKER, H., BURLEIGH, R., and MEEKS, N., 1969, Radiocarbon, 11, 278-94. R., 1970, Phil. Trans. R. Soc., Lond., A269, 23-36. BERGER, BIEN,G. S., RAKESTRAW, N. W., and SUESS, H. E., 1963, Radioactive Dating (Vienna: IAEA), pp. 159-73. BIEN, G., and SUESS,H. E., 1967, Radioactive Dating and Methods of Low-level Counting (Vienna: IAEA), pp. 105-18. V., and NEUSTUPNY, E., 1967, Nature, 215, 261-3. BCCHA; V., 1967, Archaeometry, 10, 12-22. BUCHA, CRAIG, H., 1957, Tellus, 9, 1-17. DAMON! P. E., 1968, Meteorological Monographs, 8, 151-4.

Physics applied to archaeology I. Dating

995

DAMON, P. E., 1970, Radiocarbon Variations and Absolute Chronology, Ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 571-94. ELSASSER, M7., NEY, E. P., and WINCKLER, J. R., 1956, Nature, 178, 1226-7. FERGUSON, C. W., 1968, Science, 159, 839-46. FROMM, E., 1970, Rudiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Xlmqvist and Wilrsell), pp. 163-72. DE GEER, G., 1940, Kungl. Sv. Vetensk. Akad. Handl., 3 Ser., Bd. 18, No. 6. H., 1962, Nature, 195, 984. GOODWIN, A. E., and DE VRIES, H., 1958, Science, N Y , 128, 472-3. HARING, A., DE VRIES, HOUTERAIANS, J., 1966, 2. Phys., 193, 1-12. HOUTERMANS, J., SC-ESS, H. E., and MUNK, W., 1967, Radioactive Dating and Methods of Lozcieael Counting (Vienna: IAEA), pp. 57-68. HUGHES, E. E., and MANN, W. B., 1964, Intern.J. Appl. Radiation Isotopes, 15, 97-100. KLLP,J. L., ~ ~ ~ V O L C H. HO Lo, K 1953, , Phys. Rea., 90, 713-4. LAL, D., and VENKATAVARADAN, V. S., 1970, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 549-70. LERMAN, J. C., MOOK, W. G., and VOGEL, J. C., 1970, Radiocarbon Variations and Absolute Chronology, Ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 275-302. LIBBY, W. F., ANDERSON, E. C., and ARNOLD, J. R., 1949, Science, 109, 227- 8. LIBBY, W. F., 1955, Radiocarbon Dating (University of Chicago Press). LIBBY, W. F., 1970 a, Phil. Trans. A. Soc., Lond., A269, 1-10. LIBBY,W. F., 1970 b, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 629-40. LINGEXFELTER, R. E., 1963, Rev. geophys., 1, 35-55. LINGENFELTER, R. E., and RAMATY, R., 1970, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 513-38. MICHAEL, H. N., and RALPH, E. K., 1970, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), p. 109-20. ~ ~ ~ J S N K. I C O., H , and ROETHER, W., 1967 in Radioactive Dating and Methods of Low-level Counting (Vienna: IAEA), pp. 93-104. XEUSTUPKY, E., 1968, Slovenska Archeologia, 16, 19-60. NYDAL, R., 1967, Radioactive Dating and Methods of Low-leael Counting (Vienna: IAEA), pp. 119-28. OESCHGER, H., HOUTERMANS, J., LOOSLI, H., and WAHLEN, M., 1970, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 471500. QUITTA, H., 1967, Antiquity, XLI, 263-70. H. N., 1969, Radiocarbon, 11, 469-81. RALPH, E. K., and MICHAEL, RAMATY, R., 1965, Magnetism and the Cosmos, ed. S . K. Runcorn (Edinburgh: Oliver and Boyd), pp. 66-78. C., 1968, Ann. Brit. Sch. Archaeol. Athens, 63, 277-85. RENFREW, REVELLE, R., and SUESS, H. E., 1957, Tellus, 9, 18-27. SAVE-SODERBERGH, T., and OLSSON, I. V., 1970, Radiocarbon Variations and Absolute Chronology, Ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 35-56. SCHAEFFER, 0. A., DAVIS, R., STOENNER, R. W., and HEYMANN, D., 1970, Brookhaven National Laboratory Report No. 7948. SCHELL, W. R., FAIRHALL, A. W., and HARP,G. D., 1967, Radioactive Dating and Methods of Low-leuel Counting (Vienna: IAEA), pp. 79-92. SHOVE, D. J., 1955, J. geophys. Res., 60, 127-45. hl., 1961, J. geophys. Res., 66, 273-6. STUIVER, STUIVER, M., 1970, Radiocurbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 197-214. SLESS, L E., 1955, Science, 122, 415-17. SUESS, H. E., 1967, Radioactive Dating and Methods of Lozc-leael Counting (Vienna: IAEA), pp. 143-52. SLESS, H. E., 1968, Meteorological Monographs, 8, 146-50. SUESS, H. E., 1970 a, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 303-12.

996

M. J. Aitken

SUESS, H. E., 1970 b, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 595-614. SUESS, H. E., 1970 c, Antiquity, XLIV, 91-5. SUESS, H. E., and STRAHM, c . , 1970, Antiquity, XLIV, 91-9. TAUBER, H., 1970, Radiocarbon Variations and Absolute Chronology, ed. I. V. Olsson (Stockholm: Almqvist and Wiksell), pp. 173-98. lENKATAVARADAN, v. s., 1970, Proc. Ind. Acad. sci.,in the press. DE VRIES,H., 1958, Koninkl. Nederlandse Akad. Wetensch. Proc., Ser. B, 61, 1-9. WALTON, A., and BAXTER, M. S., 1968, Nature, 220, 475-6. W. J., and BAKER, M. J., 1967, Radioactive Dating and WALTON, A., BAXTER, M. S., CALLOW, Methods of Low-level Counting (Vienna: IAEA), pp. 41-8. WILLIS,E. H., TAUBER, H., and ~ I ~ ~ N N I K. CH O., , 1960, Amer. J . Sci., Radiocarbon Suppl., 2, 1-4. 3. Potassium-argon dating and uranium series dating BROECKER, W. S., THURBER, D. L., GODDARD, J., Ku, T. L., MATHEWS, R. K., and MESOLELLA, K. J., 1968, Science, 159, 297-305. CHERDYNTSEV, V. V., KAZACHEVSKII, I. V., and KUZMINA, E. A., 1965, Geochem. Internat., 2, 794-805. CRASTY, R. L., and MITCHELL, J. G., 1966, Earth and planetary Sci. Lett., 1, 121-2. DALRYRIPLE, G. B., 1967, Earth and planetary Sci. Lett., 3, 289-98. EVERNDEN, J. F., and CURTIS, H. G., 1965, Current Anthropology, 6, 341. J. A., 1970, Nature, 226, 226-8. FITCH,F. J., and MILLER, FORNACA-RINALDI, G., 1968, Earth and planetary Sci. Lett., 5, 120-2. GENTNER, W., and LIPPOLT, H. J., 1969, Science in Archaeology, eds D. Brothwell and E. Higgs (London: Thames and Hudson), pp. 88-100. HANSEN, R. O., and BEGG,E. L., 1970, Earth andplanetary Sci. Lett., 8, 411-9. G., 1966,J. geophys. Res., 71, 2852-7. MERRIHUE, C. M., and TURNER, ROSHOLT, J. N., EMILIANI, C., GEISS, J., K o c z ~ F. , F., and WANGERSKY, P. J., 1961,J. Geol., 69, 162-85. SZABO, B. J., MALDE, H. E., and IRWIN-WILLIAAIS, C., 1969, Earth and planetary Sci. Lett., 6, 237-44. SZABO, B. J., and ROSHOLT, J. N., 1969,J. geophys. Res., 74, 3253-60. TUREKIAN, K. K., KHARKAR, D. P., FUNKHOUSER, J., and SCHAEFFER, 0. A., 1970, Earth and planetary Sci. Lett., 7, 420-4. VOLCHOK, H. L., and KULP,J. L., 1957, Geochim. cosmochim. Acta, 11, 219-46. 4. Thermoluminescence
AITKEN,

R/I. J., 1968, Thermoluminescenceof Geological Materials, ed. D. J. McDougall (London

and NY: Academic Press), pp. 463-70. AITKEN, M. J., 1969, Archaeometry, 11, 109-14. AITKEK,NI. J., ALLDRED, J . C., and THOMPSON, J., 1968, 2nd Int. Conf. on Luminescence Dosimetry, eds J . A. Auxier, K. Becker and E. M. Robinson (available from US National Bureau of Standards as CONF-680920), pp. 248-65. S . J., 1970, Proc. Int. Cong. on the Volcano of Thera, Athens, AITKEN, M. J., and FLEMING, 1969, in the press. S. J., REID,J., and TITE, M. S., 1968, Thermoluminescence of GeoAITKEN, M . J., FLEMING, logical Materials, ed. D. J. McDougall (London and NY: Academic Press), pp. 13342. AITKEN, M. J., TITE, M. S., and REID,J., 1964, Nature, 202, 1032-3. M. J., ZIMMERMAN, D. W. and FLEMING, S . J., 1968, Nature, 219, 442-4. AITKEN, BOYLE, R., 1664, Experiments and Considerations upon Colours with Observations on a Diamond that shines in the Dark. (London: H. Herringham.) CAMERON, J. R., SUNTHARALINGAN, N., and KENNEY, G. N., 1968, Thermoluminescent Dosimetry (University of Wisconsin Press, USA). CLAFFY, E. W., KLICK, C. C., and ATTIX, F. H., 1968, 2nd Int. Conf. on Luminescence Dosimetry, eds J. A. Auxier, K. Becker and E. M. Robinson (available from US National Bureau of Standards as COh-F-680920), pp. 302-9.

Physics applied to archaeology I. Dating

997

CURIE, D., 1963, Luminescence in Crystals, Transl. G. F. G. Garlick (London: Methuen; NY: John Wiley). DANIELS, F., BOYD, C. A., and SAUNDERS, D. F., Science, N Y , 117,343-9. FLEMING, S.J., 1966, Archaeometry, 9, 170-3. FLEMING, S. J., 1968 a, 2nd Int. Conf. on Luminescence Dosimetry, eds J. A. Auxier, I(. Becker and E. M. Robinson (available from US National Bureau of Standards as CONF680920), pp. 266-80. FLEMING, S.J., 1968 b, 2nd Int. Conf. on Luminescence Dosimetry, eds J. A. Auxier, K. Becker and E. M. Robinson (available from US National Bureau of Standards as CONF680920), pp. 465-82. FLEMING, S. J., 1969, D.Phil. thesis, Oxford University. FLEMING, S . J., 1970, Archaeometry, 12,135-46. FLEMING, S . J., MOSS, H. M., and JOSEPH, A., 1970, Archaeometry, 12,59-68. FLEMING, S.J., and ROBERTS, H. S., 1970, Archaeometry, 12, 129-34. J. H., and SRIRATH, S., 1964, Archaeometry, 7,58-62. FREMLIN, G.F. J., 1949, Luminescent Materials (Oxford: Clarendon Press). GARLICK, GARLICK, G. F. J., and GIBSON, A. F., 1948, Proc. Phys. Soc., A60, 574. N., HOUTERMANS, F. G., and STAUFFER, H., 1960, Hela. phys. Acta, 33,595-6. GROGLER, HALPERIN, A., and BRAKER, A. A., 1960, Phys. Rev., 117,408-15. Y., 1965, Bull. Inst. chem. Res. Kyoto Univ., 43(1), 1-6. ICHIKAWA, JASI~~K A ,and NIEWIADOMSKI, M., T., 1970, Nature, 227, 1159. JOHNSON, N. M., and BLANCHARD, R. L., 1967, Am. Miner., 52,1297-310. G. C., and KNOPFF, L., 1960, Archaeology, 13,147-8. KENNEDY, LOUCHTCHIK, T., 1955, Crystallophosphors (Tartlu : Editions Acad. Fiz. and Astronomy), pp. 17-88. MAZESS, R. B., and ZIMMERMAN, D. W., 1966, Science, N Y , 152, 347-8. MEJDAHL, V.,1969, Archaeometry, 11, 99-104. MEJDAHL, V., 1970, Archaeometry, 12, 149-61. NEUSTUPNY, E., 1968, Slovenska Archeologia, 16,19-60. QUITTA, H., 1967, Antiquity, XLI, 263-70. RALPH, E. R., and HAN,M. C., 1966, Nature, 210, 245-7. M. H. F., 1945, Proc. R . Soc., A184, 366-407. RANDALL, J. T., and WILKINS, THOMPSON, J., 1970, D.Phi1. thesis, Oxford University. TITE, M. S., 1966, Archaeometry, 9, 155-69. E., and GOLDSTEIN, N., 1966, Health Phys., 12, 1705-13. TOCHILIN, TURNER, R. C., RADLEY, J. M., and MAYNEORD, W. V., 1958, Brit. J. Radiol., 31, 397406. ZIMMERMAN, D. W., 1967, Archaeometry, 10,26-8. ZIMMERMAN, D. W., 1970, D.Phil. thesis, Oxford University. ZIMMERMAN, D. W., 1971, Archaeometry, 13,in the press. D. W., and HUXTABLE, J., 1969, Archaeometry, 11, 105-8. ZIMMERMAN, ZIMAIERMAN, D. W., and HUXTABLE, J., 1970, Antiquity, XLIV, 304-5. D. W., and HUXTABLE, J., 1971, Archaeometry, 13, 53-8. ZIMMERMAN, 5. Fission track dating BRILL,R. H., FLEISCHER, R. L., PRICE, P. B., and WALKER, R. M., 1964, J . Glass Stud., 6, 151. FLEISCHER, R. L., LEAKEY, L. S . B., PRICE, P. B., and WALKER, R. hl., 1965, Science, N Y , 148, 72-4. FLEISCHER, R. L., and PRICE, P. B., 1964, J . geophys. Res., 69, 331-9. HUANG, W., and WALKER, R. M., 1967, Science, N Y , 155,1103-4. KAUFHOLD, J., and HERR, W., 1967, Radioactizv Dating and Methods of Low-lezwl Counting (Vienna: IAEA), pp. 403-13. NISHIMURA, S., 1971, Nature, in the press. PRICE, P. B., and WALKER, R. M., 1963,J. geophys. Res., 68,4847-62. RENFREW, C., DIXON, J. E., and CANN, J. R., 1966, Proc. prehist. Soc., 2, 30-7. TURKOWSKY, C., 19G9, Earth and planetary Sci. Lett., 5, 492-6.

998

IW.J. Aitken

~VALKER, R.M., MAURETTE, M., FLEISCHER, R.L., and PRICE, P. B., 1969, Publication KO. SPP-6 of the Laboratory for Space Physics, Washington University, St Louis, 3Iissouri. Ft'ATANABE, N., and SLZUKI, &I., 1969, Nature, 222, 1057-8. 6. Dating by chemical change BREWSTER, D., 1863, Trans. R. Soc. Edin., 27, 193-204. BRILL,R. H., 1963, Sci. Am., 209, 120-30. BRILL,R. H., 1968, Proc. 8th Int. Cong. on Glass 1968, pp. 47-68. H. P., 1961, Nature, 189, 12. BRILL,R. H., and HOOD, FRIEDMAN, I., SnwrH, R. L., and CLARK, Do,1969, Science in Archaeology, ed. D. Brothwe!l and E. Higgs (London: Thames and Hudson), pp. 62-75. MICHELS, J. W., 1967, Science, N Y , 158,211-4. KEWTON, R. G., 1966, Glass Technol., 7, 22-5. R. G., 1969, Glass Technol., 10,40-2. NEWTON, NEWTON, R. G., 1971, Archaeometry, 13,1-11. OAKLEY, K.P., 1969, Science in Archaeology, ed. D. Brothwell and E. Higgs (London: Thames and Hudson), pp. 35-45. SHAW, G., 1965, New Scientist, 27, 290-1. TANABE, G., and WATANABE, N., 1968, J . Fac. Sci., Univ. Tokyo, Sec. V, Vol. 111, Part 3, 1 99-2 16. 7. Archaeomagnetism AITKEN, 11.J., 1958, Archaeometry, 1, 16-20. M. J., HAROLD, M. R., WEAVER, G. H., and YOUNG, S. A., 1967, ,Wethods in PaleoXITKEN, magnetism, ed. D. W. Collinson, K. LT. Creer, and S. K. Runcorn (Amsterdam, London and NY: Elsevier), pp. 301-5. AITKEN, M. J., and HAWLEY, H. N., 1966, Auchaeometvy, 9,187-97. M. J., and HAWLEY, H. K., 1967, Archaeometry, 10, 129-35. AITKEN, AITKEN, M. J., and HAWLEY, H. N., 1971, Archaeometry, 13,in the press. M. J., HAWLEY, H. N., and WEAVER, G. H., 1963, Archaeometry, 6,76-80. AITKEN, AITKEN, M.J., and WEAVER, G. H., 1962, Archaeometry, 5,4-24. G. H., 1964,J. Geomag. Geoelect., 17,393-6. AITKEN, J., and WEAVER, ALLEN, R. W., 1970, Ph.D. thesis, University o f Liverpool. ARTHUR, B. W., and JOPE,E. M., 1962-3, Post-Medieoal Archaeology, 6-7, 12. R. N., 1969, Earth andplanetary Sci. Lett., 6,221-4. ATHAVALE, AVERY, NI., et al., 1965, Proc. prehist. Soc., XXXIII, 301. BAMMER, A., 1969, Gen. Sci. Assembly in Madrid of the Int. Ass. of Geomagn. and Aeronomy, paper 111-22. BALER, L. A., 1899, Phys. Rev., 3, 34-8. RELSHE, J. C., COOK, K., and COOK, R. hl., 1963, Annu. of Bri. Sch. at Athens, 58, 8-13. BLACKETT,P. M. S., 1952, Phil. Trans. R. Soc. Lond., A245, 309-70. BONHOMMET, N., and BABKINE, J., 1967, C. R. Acad. Sci. Paris, 264,92-4. J., 1969, Eavtla and planetary Sci. Lett., 6,43-6. BONHOMMET, N., and ZAHRINGER, BREARS, P., 1967, Post-Medieval Archueology, 1, 38. BRYNJOLFSSON, A., 1957, Adw. Phys., 6,247-54. BUCHA, V.,1967, Archaeometry, 10, 12-22. V., 1970, Radiocarbon Variations and Absoltrte Chronology, ed. I. V. Olsson (Stockholm: BUCHA, Almqvist and Wiksell), pp. 501-12. BUCHA, V., and MELLAART, J., 1967, Arclzaeometry, 10, 23-5. BULLARD, E. C., 1958, Yr. Bk. of the Phys. Soc., pp. 47-60. BULLARD, E. C., FREEDMAN, C., GELLMAN, H., and NIXON, J., 1950, Phil. Tvans. R. Soc. Lond., A243, 67-92. BLRLATSKAPA, S . P., 1970, Archaeometry, 12,78-88. BURLATSKAYA, S.,NECHAEVA, T., and PETROVA, G., 1969, Archaeometry, 11, 115-30. CHEVALLIER, R.,1928, Ann. Phys., 4, 1925. G., 1962, Antiquity, 36, 206-9. CLARKE, D. I,., and CONNAH, COLLINSON, D. UT., et al. (Eds), 1967, Methods in Paleomagnetism (London : Elsevier).

Physics applied to archaeology I. Dating

999

COOK, R. >I., and B E L S HJ. ~ ,C., 1958, Antiquity, 32, 167-8. Cox, A., 1969, Science, N Y , 163, 237. Cox, A., DOELL, R. R., and DALRYMPLE, G. B., 1968, Q.J. Geol. Soc. Lond., 124, 53-66. CROSSLEY, D. W., 1967, Post-Medieval Archaeology, 1, 81. DICKSON, G. O., EVERITT, W. F., PARRY, I,. G., and STACEY, F. D., 1966, Earth and planetary Sci. Lett., 1, 222-4. DOELL, R. R., 1969 a,J. appl. Phys., 40, 945-54. DOELL, R. R., 1969 b,J. geophys. Res., 74, 4857-68. DOELL, R. R., and Cox, A., 1963,J. geophys. Res., 68, 1997-2009. DOELL, R. R., and SMITH, P. J., 1969,J. Geomagn. Geoelect., 21, 579-94. DUBOIS, R. L., 1967, Space Magn. Explor. and Technol. Symp. (Reno: University of Nevada). DC'BOIS, R. L., 1969, Gen. Sci. Assembly in Madvid of the Int. Ass. of Geomagn. and Aeronomy, Paper 111-19. DUBOIS, R. L., and WATANABE, N., 1965,J. Geomagn. Geolect., 17, 417-23. DUNLOP, D. J., and WEST,G. F., 1969, Rev. Geophys., 7, 709-57. EVERITT, C. W. F., 1961, Phil. Mag., 6, 713-26. EVERITT, C. W. F., 1962, Phil. Mag., 7, 583-616. FOLGHERAITER, G., 1899, Arch. Sci. phys. nut., 8 , 5-16. FOSTER, J. H., 1966, Earth and planetavy Sci. Lett., 1, 463-6. GRIFFITHS, D. H., 1955, Mon. Not. R. Astr. Soc., Geophys. Suppt., 7, 103. GRIFFITHS, D. H., KING,R. F., REES, A. I., and WRIGHT, A. E., 1960, Proc. R. Soc., Lond., A256, 359. HAROLD, M.R., 1960, Archaeometry, 3, 45. HARRISON, C. G. A., 1968, Nature, 217, 46-7. HIDE, R., and MALIN, S.R. C., 1970, Nature, 225, 605-9. HURST, J. G., 1963, Archaeometry, 6, 81-2. HC'RST, J. G., 1966, Archaeometry, 9, 198-9. IRVING, E., 1964, Paleomagnetism (NY, London and Sydney: J. Wiley). K., et al., 1965, Ann. Geophys., 21, 574-8. KAWAI, KAWAI, S . , and HIROOKA, K., 1967,J. Geomagn. Geoelect., 19, 217-27. KAWAI, N., HIROOKA, K., and TOKIEDA, K., 1967, Earth and planetary Sci. Lett., 3, 48-50. KING,R. F., and REES, A. I., 1966,J. geophys. Res., 71, 561-71. KOB.ZYASHI, K., and FULLER, M., 1968, Phil. Mag., 18, 601-24. KOVACHEVA-KOZHAROVA, M., 1968, c. R. Acad. Bulgave des sci., 21, 761. MANBY, T. G., 1964, AntiquariesJ., CXXI,110. MAYES, P.,1965, J. Archaeol. Ass., XXVIII,104. MAYES, P., 1968, Post-Medieval Archaeology, 2, 82. MAYES, P., et al., 1966, AntiquariesJ., XLVI,276. "AGAT T., ~ KOBAYASHI, , K., and SCHWARZ, E. J., 1965,J. Geomagn. Geoelect., 17, 399-405. KEEL,L., 1955, Adz?.Phys., 4, 191-243. SPDYKE, K.D., GLASS, B., HAYS, J. D., and FOSTER, J., 1966, Science, N Y , 154, 349. REES, A. I., 1961, Geophys.J., 5, 235-51. RIMBERT, F., 1958, Thkse Faculte' des Sciences (University of Paris). SAS.4jIM4, S., 1965, J. Geomagn. Geoelect., 17, 413-23. SMEDLEY, N., 1963, Proc. Suffolk Inst. Archaeol., XXIX,327. SMEDLEY, N., and OWLES, E., 1961, Proc. Suffolk Inst. Archaeol., XXIX,71. SMITH, J. D., and FOSTER, J. H., 1969, Science, N Y , 163, 565-7. SMITH, P. J., 1967, Ge0phys.J. R.Astr. Soc., 12, 321-62. SMITH, P. J., and DOELL, R. R., 1969, Gen. Sci. Assembly in Madrid of the Int. Ass. of Geomagn. and Aeronomy, Paper 111-36. STACEY, F. D., 1963, Adz. Phys., 12, 45-133. TANGUY, J . C., 1970, Archaeometry, 12, 115-28. THELLIER, E., 1936, Ann. Inst. Phys. Globe, 14, 65-70. THELLIER, E., 1938, Ann. Inst. Phys. Globe, 16, 157-302. THELLIER, E., 1966, Nucleus, 7, 1-35. THELLIER, E., and THELLIER, 0.)1959, Ann. de Geophys., 15, 285-376. UFFEN, R. J., 1963, Nature, 198, 143. VERHOOGEX, J., 1959,J. geophys. Res., 64, 2441-9. WAT.%N.%BE, N., 1959,.7. Fac. Sci. Univ. Tokyo, Sec. 5, 2, 1-188.

1000

M . J . Aitken

WATANABE, N., and DUBOIS, R. L., 1965,J. Geomagn. Geoelect., 17, 395-7. WEAVER, G. H., 1961, Archaeometry, 4, 23-8. WEAVER, G. H., 1962, Archaeometry, 5, 93-107. WEAVER, G. H., 1964, BSc thesis, Oxford University. G. R., 1966, Archaeometry, 9, 174-86. WEAVER, WEAVER, G. R., 1970, Archaeometry, 12, 87-95. WOOD, E. S., 1965, Surrey Archaeol. Collections, LXII, 77, YUXUTAKE, T., 1967,J. Geomagn. Geoelect., 19, 103-16.