Equivalent dose determination using single aliquots

Nucl. Tracks Radiat. Meas., Vol. 18, No. 4, pp. 371-378, 1991 Int. J. Radiat. Appl. lnstrurn., Part D

0735-245X/91 $3.00 + .00

Pergamon Pressplc

Printed in Great Britain

EQUIVALENT DOSE DETERMINATION USING SINGLE ALIQUOTS G. A. T. DULLER

Institute of Earth Studies, University College of Wales, Aberystwyth, SY23 3DB, U.K. (Received 20 February 1991; /n revised form 5 July 1991)

Abstract--Two methods for the dctermination of equivalent dose using single aliquots of potassium feldspar are described. The non-destructive nature of infrared stimulated luminescence (IRSL) has made it possible to make repeated luminescence measurements on a single aliquot without apparently changing the sensitivity of the material. Such an approach has been made practicable using the IRSL add-on to the Riso automated thermoluminescence (TL) reader (Bztter-Jensen et al., Nucl. Tracks 18, 257-263, 1991). The hardware and software that have made this possible are described. The method based upon the regeneration technique does appear to suffer from problems of changes in sensitivity when bleaching between doses. It is hypothesized that this is due to partial emptying of the trapped charge population. In the case of the additive dose technique, no such problem is seen. T/le precision of both techniques is far superior to normal TL and IRSL techniques. The accuracy appears good for the additive dose technique, but not for regeneration.

1. INTRODUCTION T ~ STANDARD methods of equivalent dose (ED) determination for sedimentary material [total-bleach method (Singhvi et al., 1982), regeneration method (Wintle and Proszynska, 1983), and partial bleach method (Wintle and Huntley, 1982)] involve the preparation of many aliquots of the sample. This is inevitable given the changes of sensitivity that occur upon heating during thermoluminescence (TL) measurements. Each aliquot can only be measured once. Problems arise concerning normalization between these different aliquots, and the large number that is required necessitates the processing of large amounts of sample. Southgate (1985) attempted to use TL measurements on single feldspar grains to determine the equivalent dose of beach and dune sands, but no final results were presented, and only one measurement could be made on each grain because of the changes in sensitivity that occurred upon TL readout. The use of optically stimulated luminescence (OSL) offers a less destructive method of sampling the trapped charge concentration (Huntley et ai., 1985) and of determining the equivalent dose using a single aliquot of material. Smith et al. (1986) mention the use of a single aliquot of quartz or zircon for ED determination using OSL (stimulating with an argon-ion laser), but give no details or results. Rhodes (1990, p. 90) pursued the possibility, but stated that such a method was impracticable because of the change in sensitivity that occurred upon exposure to the laser. The large amount of labour involved was also prohibitive since no automation was available.

The Rise automated TL reader, together with an infrared stimulated luminescence (IRSL) add-on unit (Bztter-Jensen et al., 1991), provides all the facilities that are required to carry out ED determination. Computer control of a radioactive source, a heating strip, and an infrared (IR) diode array means that all measurements and treatments can be carried out automatically, without the need for constant operator supervision.

2. THE ADVANTAGES OF EQUIVALENT DOSE DETERMINATION USING ONLY A SINGLE ALIQUOT Despite the problems found in earlier attempts, using the argon-ion laser to stimulate quartz, the advantages of using only a single aliquot are clear. (a) Precision. Since all measurements for the equivalent dose (ED) are made on a single aliquot there is no need for normalization between measurements due to variations in sample weight or sample mineralogy. Although IR will only stimulate luminescence from feldspathic minerals, variations in potassium feldspar mineralogy are a problem. A potassium feldspar separate from a sedimentary sample contains a variety of minerals (mainly orthoclase, sanidine and microcline). These are unlikely to behave similarly in terms of dose response and in normal luminescence methods of ED determination (either TL or IRSL) some means of normalization is required. If the saturation doses of the individual minerals were different then standard ED determinations would be seriously compromised. In particular when high doses

371

372

G. A. T. D U L L E R

are given, with some minerals approaching their saturation dose and others not, the scatter between measurements increases. This effect will not be a problem using a single aliquot. (b) Multiple determinations. By running many single aliquots the true variability in the ED of the material can be assessed. Small variations in the ED of the sample (from grain to grain for example) can be seen and an average value calculated together with the standard deviation. Using standard techniques it is very time-consuming to make repeat ED determinations, and where only one determination is made one has to estimate the error purely in terms of the mathematical fit of the data to a linear or exponential regression. The errors obtained in this way are highly dependent upon the exact method of error analysis and are not representative of the variability in mineral response (Berger and Huntley, 1989). (c) Effort. Standard luminescence techniques require at least 30 measurements to be made in total to ensure reproducible behaviour. In many cases, where greater precision is required, this number may easily exceed 60, each measurement needing a separate aliquot. Using a smaller number of aliquots would mean that smaller samples could be used (particularly important in archaeological sites and where field access is difficult), less effort is required to extract sufficient potassium feldspar, and less time taken to mount the sample. The greatest economy in effort is in the ED determination where all sample treatment--irradiation, pre-heating, bleaching, measurement and data manipulation--is undertaken automatically within the Riso reader. 3. EQUIPMENT The IRSL add-on unit for the Rise reader uses 31 TEMT 484 IR emitting diodes run at 40 mA. The IR power at the sample is 32 mW cm -2, as measured using a photodiode calibrated at Rise. Luminescence is detected by an EMI 9635QB photomultiplier tube with a 1 mm thick Schott BG-39 filter for IR rejection and a Coming 5-58 filter to isolate the blue emission. Irradiations were performed using a 9°Sr/~Y beta source delivering a dose of 1.24 Gy rain- ~. A pre-heat of 220°C for 10 rain was used to remove any thermally unstable component of the luminescence signal (Li, 1991). The IRSL measurements were made at a constant temperature of 50°C. 4. THE REGENERATION M E T H O D 4.1. Methodology This is intrinsically the simplest method to use on a single aliquot. The means by which an equivalent dose can be determined is that described originally by

Wintle and Proszynska (1983) and relies upon the ability to bleach the sample to its residual level, mimicking its signal at deposition. Using a single aliquot the sample is pre-heated at 220°C for l0 rain to remove the relatively unstable potassium feldspar TL peak at 270°C. The IRSL signal is then measured for 100 s to produe a decay curve. Further IR exposure is used to reduce the IRSL signal from the aliquot to a residual level. During this time the IRSL signal is constantly monitored. The sample can be exposed to IR for a fixed time, or until the IRSL signal falls below a preset count rate. Once a residual level has been achieved, the aliquot can be irradiated, pre-heated and re-measured. Bleaching back to the residual level can then be followed by as many irradiations as required to build up a growth curve. 4.2. Testing the method Material from a New Zealand dune sand (laboratory designation G D N Z 5) with a 14C date on associated charcoal of 22,590 + 230 yr BP (Wilson et al., 1988) was chosen for some basic experiments on the method described above. A grain size of 180-211/am was chosen, and the potassium feldspar separated using sodium polytungstate to select material lighter than 2.58g cm -3. Sample masses of about 8 mg were mounted on aluminium discs using Silkospray. Figure 1 shows the IRSL decay curves and associated plateau for a single disc of G D N Z 5. Six discs were run at the same time, all yielding plateaux similar to that in Fig. 1. The mean ED for the six discs was 32.9 +0.5 Gy, that is a scatter of 1.5% about the mean. After measurements of the natural IRSL the ED was determined measuring the residual level first, and then giving successively larger irradiation doses. Samples were exposed to IR between each measurement until their IRSL signal fell below 600cps. Since the sample is bleached between each measurement there is no special reason why the regeneration curve has to be built up in this order. Figure 2 shows another disc of G D N Z 5 measured using similar parameters as for the sample in Fig. 1. The main difference is that after measuring the natural signal the largest irradiations were given FIRST. The differences in the shape of the growth curve is obvious. For the same disc the response to the largest dose and the residual level were re-measured after the points in Fig. 2 were taken (Fig. 3). Clearly, as the number of measurements taken increases, so does the signal. This is not simply an unbleachable component that is building up since the signal increase from the sample after it has been bleached and given 37.2 Gy is far larger than that for the residual. As more measurements are taken there is a progressive

E Q U I V A L E N T DOSE D E T E R M I N A T I O N U S I N G S I N G L E ALIQUOTS

373

1200 S0"

a

................................

(1) ~

{O)

800

==__.

'~"

,oo

~10-

": ~'...

,'o

o

',, , , ~ . . . .

U) ¢v

_,~==

~o Dose (Q

z

"~

131

4oo

200 \

"'-..~

.....

---:--=-_

[

__:=~

t 10

. . . . . . . . I o . 24.e Qy

2o 40 so 6o Time (seconds)

40

Natural

' .......................

~ .......................

!,o "

C o

20

40 60 Time (seconds)

80

1oo

1:](3. 1. IRSL decay curves (a) for construction of the regeneration curve for the IRSL signal integrated over 100 s for sample GDNZ 5 (b) and its associated plateau test (c). The smallest regeneration doses were given first, followed by successively larger ones. increase in sensitivity. This explains why the shape of the two growth curves in Figs I and 2 are so different. This increase in sensitivity has been seen for almost all samples that have been studied. The significant a

...........................................

q

20-

\ m ~'10O)

",\

_. -- Noetural

2'o

,'o

s'o

6'o

i 30

40

FIG. 3. The same aliquot as in Fig. 2, but showing replicate measurements of the response to the highest dose and the residual. The order in which measurements were made is indicated by the numbers in brackets. The progressive increase in sensitivity of the material as measurements are made is clear.

loo .............. lo + 37.2 Gy

.....................................................................

I 20 Dose (Gy)

exception to this is a sample whose natural signal was close to saturation. Several repeat measurements were made using a large dose, close to that of the ED, and very little change in signal was seen. To see whether the increase in sensitivity was due to the use of IR bleaching, three samples of Dutch cover sands [Lutte 2, 8 and 12, Dijkmans and Wintle (1991)] were analysed. They were exposed to the SOL2 solar simulator for 24 h between each measurement. The growth curve (Fig. 4) was linear and none of the samples showed any signs that the IRSL signals were anywhere near saturation. Measurements of the signal induced by the largest and smallest doses were repeated three times. F o r sample Lutte 2 the largest dose showed an increase in signal of 4.2% ( + 3 . 0 % at one S.D., n ffi 12) and the residual an increase of 1.4% ( + 3 . 2 % at one S.D., n ffi 12). This is less than when the samples were bleached using IR. This procedure requires the operator to transfer the samples from the Rise reader to the SOL2 between each measurement. This is time-consuming and so somewhat negates the advantages of the method over standard equivalent dose techniques. The change in sensitivity is believed to be caused by a progressive build-up of the trapped charge within

...... 1o + 12.4 Gy . . . . . . . . 1o + 24.8 Gy 100 .............. Io + 37.2 Gy

1600 1400

~ f Progreeel~IrmreaseIn sensitivity?

Time (seconds)

1200

111

(11) (7) iS) (;D

1000 40 ............................................................

i

........................

600

==

600

c

~20-

0

~ =o

i ,'o

~ 6'o

Time (seconds)

(el I 5

c 6o

400 200 (10)

too

FIG. 2. An aliquot of GDNZ 5 identical to that in Fig. 1.

(See Fig. 1 for captions.) The procedure only differs in that the largest dose was given first, followed by successively smaller ones.

i 10 Dose (Gy)

I 18

20

FIG. 4. Regeneration curve for sample Lutte 2 using a 24 h SOL2 bleach between each measurement. The order in which measurements were made is shown by the numbers in brackets. The magnitude of the change in sensitivity is smaller than that seen in Fig. 3.

374

G.A.T.

the sample which affects the rate of subsequent trap filling when a sample is irradiated. It is known that IR cannot remove most of the TL signal [about 95%; Duller (1991)] even when the sample is held at elevated temperatures (Duller and Wintle, 1991). Use of the SOL2 solar simulator reduces the severity of the problem because it removes a greater proportion of the signal, but apparently the problem still persists to some degree. The effect of a sensitivity change when bleaching with IR can be reduced if the ED of the sample is approximately known. Giving the dose that is nearest the ED first means that this will have the smallest error. However, the shape of the growth curve will be incorrect. Since the regeneration method relies upon interpolation, rather than extrapolation, the form of the growth curve away from the intercept with the natural is unimportant. However, the size of the systematic error, though small, is still unknown. 5. THE ADDITIVE DOSE M E T H O D The IRSL signal bleaches so rapidly for material on the surface of a sand dune or sand sheet that there is effectively no residual signal at deposition. Hence the residual signal for a buried sample may be taken to be zero and an additive dose procedure adopted. The additive dose method involves no exposure to light (except for the very short IR exposure during readout) and hence no opportunity for changes in sensitivity. Because there is no bleaching between irradiations the measurement time is greatly reduced. 5.1. Methodology The sample is pre-heated at 220°C for 10 rain prior to each measurement to ensure that only a stable signal is measured. After measurement of the natural signal, a beta dose is given and the sample is pre~heated again. This is repeated for as many doses as required to generate a growth curve for a single aliquot. The drawback with this procedure is that every pre-heat removes a part of the stable signal as well as

;

DULLER 800 Too

000

+

S coo

i

400

(el

(a)

$00 =oo 100(

J,Preheal CyCleS Or*t¥

t 50

t 100

t 180 Added

J 200 Do8e (Gy)

t: 250

FIG. 6. Additive dose IRSL growth curves for sample GDNZ 5 using (a) no correction for losses due to pre-heating (a), a correction based upon the assumption that the IRSL signal will decay in the same way irrespective of previous irradiation (b), and a correction assuming that for each dose given to the sample the IRSL will decay independently of the IRSL due to previous irradiations (¢). all the unstable component. After adding the first beta dose, the pre-heat will remove all the unstable component induced by that dose, and a part of the stable natural signal. It is necessary to correct for this loss of signal. The loss of signal due to pre-heating can be assessed for each sample by taking a single aliquot and repeatedly pre-heating it and measuring its IRSL without adding any artificial dose. Figure 5 shows the loss of signal for sample G D N Z 5. The scatter is thought to be due to the effect of small variations in the mineralogy of the samples on the aliquots. The final calculation of the error in the growth curve incorporates the uncertainty associated with these figures. Measurement of the IRSL signal can be made using very brief exposure to IR stimulation (e.g. 0.1 s). This causes negligible loss of signal. However, the uncertainty in such a measurement may be relatively large, especially for samples with a low signal. A half-second IR stimulation improves the reproducibility of the measurements and a correction for the small loss of the IRSL signal during measurement is incorporated in the correction for pre-heat loss. 5.2. Pre-heat correction

0.9

×',

~ 0.8

~. 0.7 i 0.6 0.5 0.4 [ 1

2

3

4

5

6

i

t

t

7

8

9

10

Preheat number

Fro. 5. The effect of successive pre-heats at 220°C for 10 rain on the natural IRSL signal of sample GDNZ 5. The IRSL measurements were taken for half-a-second while the sample was held at 50°C. The results from five duplicate runs are shown.

There are two ways of applying the correction for the effect of pre-heating.They rely upon different assumptions concerning the way in which charge is distributed in a crystal when it is irradiated. The two procedures have been tested empirically using the same data set. Aliquots of G D N Z 5 were used to provide the data set. A single aliquot was given doses of 15, 16, 31, 62 and 124Gy, and pre-heated and measured between each dose. This resulted in cumulative doses at each measurement stage of 15, 31, 62, 124 and 248 Gy. From these a growth curve can be constructed [Fig. 6(a)]. The aliquot was then pre-heated again, with no dose added, and re-measured. This

EQUIVALENT DOSE D E T E R M I N A T I O N USING SINGLE ALIQUOTS was repeated four times to give five measurements in total for the N + 248 Gy point. The IRSL signal became progressively smaller with the successive preheats [Fig. 6(a)]. Application of a successful correction procedure should be able to make these last five figures equal. The first method of correction assumes that after irradiation the distribution of trapped charge will be similar irrespective of what pre-heats or irradiations have occurred previously. In this case each pre-heat will cause the previous signal to decay by the ratio between pre-heat I and 2 on Fig. 5. That is by a factor of 0.858. Table 1 contains a numerical example of the corrections applied to the raw data using this assumption (columns 3 and 4). The natural signal (SB, =/1), when the sample is pre-heated after the first beta dose is added, will decay to 0.858 of its initial level. When this value is subtracted from the luminescence measured after 15 Gy had been added to the natural (12) it gives the signal induced just by the last beta dose that was added (S~, column 3). This is then added to the natural signal (SB~) obtained in the first measurement. To calculate the signal induced by the second added beta dose (16Gy) the contribution of the natural and the first beta dose have to be subtracted. These are found by multiplying the previous measurement taken at N + 15Gy (12) by 0.858. This process was repeated for all 10 measurements and is summarized in the expression. S ~ = In -- (.I._

j X

0.858).

The corrected data are plotted in Fig. 6 (curve b) and clearly show that the correction procedure is not valid. The form of the growth curve up to N + 2 4 8 G y is reasonable, but the five repeat measurements of the largest dose are different and increase monotonically with successive pre-heating corrections. The second method of correction assumes that the trapped charge induced by each dose behaves

independently of the others. For instance, the natural signal decays in precisely the same way as shown in Fig. 5 whether beta doses are added on or not. Similarly, when the first beta dose is added, the trapped charge population that it produces will decay following the same pattern as in Fig. 5. To correct for this type of effect, one has to isolate each component of the signal and then use the decay curve in Fig. 5 to calculate the loss of signal. By subtracting the contribution that all the previous doses make to the measured signal, one can calculate the part of the signal that must be due to the last added beta dose (column 6). This can be described mathematically as n--I

s,. = t . -

~(S~xF~._,+,) 1

where for measurement (n) the luminescence signal (S) due to that beta dose (Bn) can be calculated by subtracting the sum of the components of the previous signals (Sin) (column 5, Table 1) multiplied by the appropriate correction factor (F(,_~+~), Fig. 5) from the measured luminescence In (column 2, Table 1). Column 6 of Table l has been corrected using this second assumption, and the data are plotted in Fig. 6 (curve c). This is the more successful of the two correction methods. The five repeat measurements of the N + 248 Gy point are not identical but are well within the error limits associated with the uncertainties in the pre-heat correction factors. This experiment was repeated 19 times and in all these cases a small, but consistent, decrease in signal was seen for these five measurements. On average, the corrected signal had dropped by 4.2% from the first to the last of these repeat measurements. Such a set of high precision results would be ditficult to achieve without use of an automated system such as the Rise reader. The form of the growth curve [Fig. 6(c)] thus obtained is also very similar to that obtained using standard TL and IRSL techniques. This is the correction method used in all subsequent analyses.

Table 1. Pre-heat correction using two assumptions. Assumption 1: that the signal will decay in the same way irrespective of previous irradiations and pre-heats. Assumption 2: that each component of the trapped charge population behaves independently of the others. All values in columns 2-6 are luminescence measurements in counts sAssumption 1 Total dose (Gy) Natural N + 15 N + 31 N + 62 N + 124 N + 248 N + 248 N + 248 N + 248 N + 248

Measured IRSL 88,944 110,950 134,514 193,288 306,690 498,718 439,592 396,688 364,412 337,916

IRSL due to last added dose (88,944) 34,636 39,319 77,875 140,849 235,578 11,692 19,518 24,054 25,251

375

Corrected IRSL 88,944 123,580 162,899 240,774 381,623 617,201 628,893 648,411 672,465 697,716

Assumption 2 IRSL due to last added dose (881944) 34,636 36,154 72,424 133,938 225,072 -4246 - 6032 - 3804 - 2566

Corrected IRSL 88,944 123,580 159,722 232,146 366,084 591,156 586,910 585,124 581,320 578,754

376

G.A.T.

Table 2. Equivalent dose determinations for sample GDNZ 17 using a variety of methods Method of determination Regeneration Additive dose Single disc

Equivalent dose (Gy)

TL IRSL TL IRSL IRSL

n.d. 23.7 + 2.4 21.6 + 1.6 22.0 + 0.8 22.5 ± 1.6

5.3. Testing the method Four experiments were used to test the accuracy and precision of the method outlined above. The first test was aimed at examining the efficacy of the correction technique. A second sample of dune sand ( G D N Z 17) from New Zealand with a geological age estimate of 6.5-13 ka (Palmer et al., 1988) was used. Equivalent doses were determined using standard, multi-aliquot, TL and IRSL methods (Duller, 1991) and the results are given in Table 2. Five added doses were given to each disc in a set of 20 discs of this sample to generate growth curves. Another set of six discs were just given a single dose (93 Gy), equivalent to the total dose received by the other 20. The ratios between the signals for the natural and the N + 93 Gy were calculated for both sets of discs. One set needed six "phases" of correction, the other only one. If a problem exists with the correction method then there should be a significant difference between the two ratios. F o r the 20 discs given a whole sequence of doses, the ratio between the natural signal and that for N + 93 Gy was 5.14 + 0.27, while for the other six it was 4.89 + 0.22. The two ratios lie within each other's error limits and suggest that there are no problems in the correction method that can be seen within the precision of this experiment (_+ 5%). The second test involved repeat measurements of G D N Z 5 and G D N Z 17 to determine the precision

DULLER of the analysis. Nineteen samples of G D N Z 5 were given the same doses as those in Fig. 6 and corrected using assumption 2. Equivalent dose determinations were calculated using an exponential fit to the data and only the first of the measurements of the N + 248 Gy dose was included. The mean ED was 36.0 Gy, the standard deviation of the means being 1.0 Gy, i.e. a precision of 2.8%. Twenty samples of G D N Z 17 were given doses of 10, 21, 31, 52 and 93 Gy. A linear fit was found suitable for each growth curve and an average ED of 22.5 G y was calculated. The standard deviation of the means was 1.6 Gy, i.e. a precision of 7%. The third test looked at the validity of the assumption that an IRSL signal induced by a beta dose decayed during pre-heating in the same way as the natural signal. This is a central assumption of the correction method. Twenty-one aliquots (split into four groups) of sample G D N Z 5 were exposed to 10 cycles of preheating. After each pre-heat, the IRSL signal was measured. One group still retained their natural signal (ED of about 40 Gy) while the other three groups had been bleached in a SOL2 solar simulator for 24h and then given beta irradiations of 15, 191 and 1143 Gy respectively. The results for all four groups of discs (Fig. 7) lie within error limits. The implication is that for this pre-heat there is no significant difference in the way that the relevant natural and beta-indueed signals decay. However, in Fig. 7 it can be seen that the decay of the signal from samples which have had a beta dose show a consistent, but small, dose dependence, with the samples that have received the larger dose decaying more slowly than those that received the smaller dose. This runs contrary to intuition, which would suggest that the sample with the smaller signal should

(n =5) (n=6) (n=5) (n=5)

10t i

i Natural(ED~40Gy) I Io+15Gy ! Io+ 191 Gy i Io+ 1143Gy

c/) 0.8ev-

E i Z 0.6-

i

0.4~ 2

3

4 5 (~ "l Number of preheat cycles

¢=

9

1'0 ~"

FIG. 7. The effect of successive pre-heats at 220°C for 10 rain upon samples of GDNZ 5. Group (a) still retain their natural dose (about 40 Gy), while the other three groups have been bleached for 24 h in the SOL2 solar simulator, followed by beta doses of 15Gy (b), 191 Gy (¢) and l143Gy (d). IRSL measurements were taken for half-a-second while the sample was held at 50°C.

EQUIVALENT DOSE D E T E R M I N A T I O N USING SINGLE ALIQUOTS Table 3. Equivalent dose determinations for sample GDNZ 5 using a variety of methods Method of determination Regeneration TL IRSL Additive dose TL IRSL Single disc IRSL

Equivalent dose (Gy) 44.7 -l- 1.7 48.0 + 3.6 35.8 + 2.6 45.6 + 1.5 36.0 + 1.0

decay more slowly, proportionately, than that with the larger signal. Further work is required to see whether this phenomenon is generally true, and to look for a mechanism. The final test compared the EDs calculated using this method resulting from 20 measurements with those determined by standard, multi-aliquot, IRSL and TL techniques. For G D N Z 17 previously published EDs (Duller, 1991) all overlap those determined here (Table 2). For G D N Z 5 the four previously published EDs show considerable scatter (Table 3) and that determined using the single disc additive dose method lies at the lower end of these values. A sample from the Mojave desert was also used to compare the single aliquot method with standard methods of ED determination. Sample K24 was taken from a linear dune ridge system in the Kelso dune field, California. The preparation and measurement of the sample were identical to that for the two samples from New Zealand. Table 4 shows that although the single aliquot result is slightly lower than that for the standard methods, all three methods are in agreement within errors. The results of these checks on the method are encouraging but not yet conclusive. Further tests are required, both into the internal consistency of the method, and in its comparison with other methods.

6. SINGLE GRAIN EQUIVALENT DOSE DETERMINATION The work described above allows far greater precision in ED determination than is available using standard techniques. Additionally, the advantages of speed and the lack of operator time required for such procedures make them very attractive. The techniques outlined above are capable of working upon a single aliquot. The size of this aliquot is not important. In theory there is no problem extending this to the limit and using a single grain of feldspar on a disc. This would have the advantage of avoiding the problems of mixed mineralogy within Table 4. Equivalent dose determinations for sample K24 using a variety of methods Method of determination Equivalent dose (Gy) Additive dese TL 20.2 + 1.0 IRSL 19.5 + 1.6 Single disc IRSL 19.2-1- 1.7

377

potassium feldspar separates from sedimentary materials. Mejdahl (1983) was able to produce an isochron for a sample by dating sub-samples with different potassium contents, giving different EDs, but similar ages. An identical method could be followed using single grains. Such an approach would require the potassium content of a single grain to be determined. When using the additive dose method for a single grain the loss of the signal from the grain as a result of the pre-heating must be established. This is likely to vary from one grain to another. However, from the work shown in Fig. 7 it is clear that following the additive dose procedure it is possible to bleach a sample, give it a beta dose, and then determine the correction factors relevant for that grain. 7. CONCLUSIONS Use of single aliquots for ED determination has obvious and appealing possibilities. Of the two methods described above, the regeneration method seems to have several deep-seated flaws, associated primarily with changes in sensitivity. The additive dose method appears more promising, although it is a two-stage process involving the characterization of the pre-heat. From the preliminary tests described above, the method appears to offer high precision, internal consistency within the error limits, and good agreement with standard TL and IRSL techniques. The possibility of extending this type of work to single grains to gain further advantages in dosimetry and precision is exciting and warrants further effort. Acknowledgements--The software that made this work

possible was developed while the author was visiting the Rise National Laboratory, Roskilde, Denmark, as a guest of Dr Lars Bztter-Jensen. It would have been impossible to have carried out the work without the assistance of Dr Botter-Jensen and Henrik Christianscn. All ED analyses were carried out using Rainer Grun's software. The author would like to thank Dr Dijkmans for use of his samples and Stephen Edwards for all the ED determinations in Table 4. A preliminary draft of this paper was greatly improved by comments from Dr A. G. Wintle. This is publication No. 174 of the Institute of Earth studies, UCW, Aberyst. wyth. Single-aliquot methodology is supported by NERC grant GR3/8190. REFERENCES Berger G. W. and Huntley D. J. (1989) Test data for exponential fits. Andent TL 7(3), 43-46. Better-Jensen L., Ditlefsen C. and Mejdahl V. (1991) Combined OSL (infrared) and TL studies of feldspars. Nucl. Tracks 18, 257-263. Dijkmans J. W. A. and Wintle A. G. (1991) Methodologlal problems in thermoluminescence dating of Weichselian cover sand and young Holocene drift sand from the Lutterzand area, E. Netherlands. Geol. Mijnbouw 70, 21-33. Duller G. A. T. (1991) Comparison of equivalent doses determined by thermoluminescence and infra red stimulated luminescence for dune sands in New Zealand. Quat. Sci. Rev. (in press).

378

G.A.T.

Duller G. A. T. and Wintle A. O. (1991) On infrared stimulated luminescence at elevated temperatures Nucl. Tracks Radiat. Meas. 18, 379-383. Huntley D. J., Godfrey-Smith D. I. and Thewalt M. L. W. (1985) Optical dating of sediments. Nature 313, 105-107. Li S.-H. (1991) Removal of the thermally unstable signal in optical dating of K-feldspar. Ancient TL 9, 26-29. Li S.-H. and Aitken M. J. (1989) How far back can we go with the optical dating of K-feldspar? In Long and Short Range Limits in Luminescence Dating. RLAHA Occasional Publication No. 9. Mejdahl V. (1983) Feldspar inclusion dating of ceramics and burnt stones. P A C T J. 9, 351-364. Palmer A. S., Barnett R., Pillans B. J. and Wilde R. H. (1988) Loess, river aggradation terraces and marine benches at Otaki, southern North Island, New Zealand. In Loess: Its Distribution, Geology and Soils (Edited by Eden D. N. and Furkert R. J.). Balkema Press, Rotterdam.

DULLER Rhodes E. J. (1990) Optical dating of quartz from sediments. D.Phil. thews, University of Oxford (unpublished). Singhvi A. K., Sharma Y. P. and Agrawal D. P. (1982) Thermoluminescence dating of sand dunes in Rajasthan, India. Nature 295, 313-315. Smith B. W., Aitken M. J., Rhodes E. J., Robinson P. D. and Geldard D. M. (1986) Optical dating: methodological aspects. Radiat. Prot. Dosim. 17, 229-233. Southgate G. A. (1985) Thermoluminescence dating of beach and dune sands: potential of single-grain measurements. Nucl. Tracks 10, 743-747. Wilson C. J. N., Switsur V. R. and Ward A. P. (1988) A new date for the Oruanui eruption, New Zealand. Geol. Mag. 125, 297-300. Wintie A. G. and Huntley D. J. (1982) Thermoluminescence dating of sediments. Quat. Sci. Rev. I, 31-53. Wintle A. G. and Proszynska H. (1983) TL dating of loess in Germany and Poland. P A C T J. 9, 547-554.