ANALYST JANUARY 1986 VOL. 111 29 Studies on the Determination of Mercury in Human Beard Shavings by Neutron-activation and y-Ray Analysis J. G. Pritchard and S. 0. Saied* Department of Physical Sciences North East London Polytechnic Stratford London E 15 4LZ UK The determination of mercury by instrumental neutron-activation analysis has been studied with particular emphasis on hair as the matrix. The well known analytical standards orchard leaves and Bowen's kale have been compared and the value 0.164 k 0.013 p.p.m. found for the level of mercury in Bowen's kale. Hence the likely accuracy is _+ 8% when matrices similar to Bowen's kale are analysed for mercury by the method described. All factors affecting the precision are discussed. Powdered hair and the standards were shown to be easily homogenisable; however beard shavings were found to be generally inhomogeneous with respect to mercury distribution especially when the mercury content was much in excess of ca.2 p.p.m. owing to exposure of subjects to mercury in the professional working environment. The beard shavings of non-occupationally exposed human beings contained 0.1-1.5 p.p.m. of mercury. Difficulties with the establishment of a mercury mass standard for use with the method are discussed. Keywords Mercury determination; hair; beard shavings; neutron-activation analysis; y-ray analysis In a previous report on the determination of mercury in hair by cold vapour atomic absorption spectrometry,lJ it was noted that the instrumental neutron-activation analysis method had been widely applied to the determination of mercury in hair but had given some results that appeared much too high and was unattractive as too few experimental details were available from which the procedure might be judged.2-6 In the course of continued work on the determi-nation of mercury in human hair we have taken the opportunity to examine the neutron-activation method in detail and to report on its reliability in this application.In principle the neutron-activation method seems simple. Samples and standards should be irradiated under the same conditions in the high neutron flux of an atomic reactor then the gamma radiation from both should be compared quantita-tively by means of a germanium crystal scintillator.7-9 Nat-urally occurring mercury consists of a mixture of seven (non-radioactive) isotopes but only the excitation of 196Hg and 202Hg yields products that involve y-ray half-lives long enough to allow a convenient analytical procedure.10 The radio-chemical reactions for 202Hg the major isotope of mercury of natural abundance 29.7% are as follows: Activation 2gtHg + An += 2ziHg ~ ~ 4 7 d ' 2z:T1* + P- (210 keV) 203 * 2:T1 + y(279 keV) 'lT1 T ' z 3 x 10-los ' Here An indicates a neutron p- an electron (of energy 210 keV) y a gamma ray an asterisk a nuclear excited state and T4 the decay half-life.The P-rays are mainly absorbed in material located between the sample and the detecting crystal and essentially only the y-emission is measured in this instance at 279 keV. However a background is picked up by the detector, owing mainly to scattered X-rays (bremsstrahlung) that arise from the absorbed P-rays.Although of only 0.2% natural abundance the 196Hg is far more easily activated12 to 197Hg in both the nuclear ground state and an excited state.11J2 The complicated decay scheme of these species results in ';;Au as the final product with the emission of a principal y-ray of energy 77 keV the over-all decay half-life being ca. 3 d.ll * Present address Department of Mathematics and Physics, University of Aston Gosta Green Birmingham B4 7ET UK. During the early part of the decay period the y-ray spectrum of a mercury-containing substance shows a 77-keV peak that is much higher than the 279-keV peak but the much shorter half-life of the former is not conducive to a convenient analysis in our view when large batches of samples are involved.Rather the 279-keV peak with Ti = 47 d allows relatively easy measurement scheduling and the necessity for decay corrections is rare. Moreover preliminary experiments showed that the accuracy and precision of measurement of the 77-keV signal were seriously compromised by a "noisy" sloping background whereas the 279-keV signal appeared against a much clearer and flat background. Hence the 279-keV peak was used for the analysis in this work. In this instance a small' correction was necessary to the lowest mercury levels observed owing to interference from sel-enium trace amounts of which almost always occur in the hair of healthy persons.13 In this method accuracy requires (a) that analytical samples be irradiated at the same time and in the same location as aliquots of material that has been standardised with respect to mercury content so that variations in neutron flux within the reactor can be compensated and (b) that standards and samples be of the same physical form and be sufficiently homogeneous in bulk with respect to mercury so as to yield reproducible aliquots for the analysis.Samples of cut hair strands taken at random in aliquots of 0.1-0.2 g differ seriously in mercury content and are therefore unsuitable, whereas similar sized aliquots of powdered hair taken from thoroughly mixed batches have been shown to be uniform in mercury content within an experimental error of 4 5 % .2 Moreover thorough testing of the powdered and easily homogenised standard orchard leaves has shown that con-sistent results can be obtained provided that aliquots of not less ,than 0.25 g are used for the analy~is.1~ We therefore undertook to compare 0.25-g samples of powdered hair with very similar samples of analysed standard materials of which two were found suitable orchard leaves and Bowen's kale.14J5 Both of these contain mercury at the lowest level likely to be found in hair 0.1-0.2 p.p.m.and were free from interferents except for a small amount of selenium for which a correction could easily be applied. (River sediment ,I6 contain-ing multifarious interferences and heterogeneous owing to its broad distribution of particle sizes and high density and bovine liver,17 with far too high a selenium content were rejected.) Loss of mercury by evaporation during the analytical process is an ever present problern.lJ8 Polythene capsules 30 ANALYST JANUARY 1986 VOL.111 which contained no serious interferences were selected to enclose the samples and standards. A relatively cool atomic pile was selected together with a modest degree of excitation to ensure the integrity of the analytical specimens. (The sealing of hair samples in quartz tubes was rejected because of suspected loss of mercury during the hot-sealing process and serious interference from selenium in the silica.) Within the above context this paper reports an exploration of the neutron-activation method and aims to provide an assessment of its precision internal consistency and accuracy with respect to atomic absorption spectrometry.2 The range of mercury concentration in the hair samples examined extends from the normal range 0.1-2 p.p.m.towards 100 p.p.m. by inclusion of hair from dentists occupationally exposed to mercury vapour. Experimental Materials Batches of electric-razor hair shavings usually in amounts from 0.5 to 1 g were collected from several dentists and other persons. Their accumulated daily shavings in the compart-ments behind the cutting rotors of the electric razors were simply tipped out into clean paper or polythene envelopes and sealed until taken for the analysis. In some instances large accumulations of up to 10 g of shavings were collected. The dentists were all in active practice with mercury and amalgam continuously in the immediate working environment whereas the other persons had no unusual contact with mercury other than from their normal daily food and a usual number of dental restorations.Standard Reference Materials Nos. 1571 (orchard leaves) , 1645 (river sediment) and 1577a (bovine liver) were supplied by the Office of Standard Reference Materials National Bureau of Standards Gaithersburg MD USA (supplies of orchard leaves were no longer available after 1983). Bowen’s kale was supplied by Professor H. J. M. Bowen University of Reading Berkshire UK.15 One millicurie of 203Hg was obtained in the form of a solution of ca. 10 mg of mercury(I1) acetate in 2-3 cm3 of dilute acetic acid at a pH in the range 2-3 from Amersham International Amersham Buckinghamshire UK.Nitric and acetic acids were of Aristar grade and sodium sulphide selenium dioxide mercury(I1) oxide and thio-acetamide were GPR-grade reagents from BDH Chemicals, Poole Dorset. Equipment Polythene capsules of pill-box shape with tightly fitting separate heat-sealable lids external dimensions 5 x 15 mm diameter capacity 0.50 cm3 (“size-l”) were purchased from Precision Machining Engineering (Harrow) Ltd. South Harrow. Polythene canisters external dimensions 75 x 25 mm diameter capacity 30.0 cm3 for containment of batches of pill-boxes were obtained from Gallenkamp Griffin London. Filled canisters were irradiated in core tubes of the University of London Research Reactor at Imperial College of Science and Technology Silwood Park Sunninghill Ascot , Berkshire UK.This reactor is water cooled fuelled by 235U, operates at a maximum power of 100 kW for 7-8 h during the normal working day and generates maximum thermal neutron fluxes of approximately 10’2 neutrons cm-2 s-1 and maximum temperature ca. 60 “C in the core tubes.19 The detection - recording system for measurements of the y-ray spectra of irradiated samples was as follows. A Canberra Model 7226 lithium-drifted germanium crystal detector was used in conjunction with a Canberra Model 3002 3-kV stabilised d.c. power supply (supplying 2.5 kV). The 4-ps pulse signal from the detector pre-amplifier was fed into a Canberra Model 2011 shaping amplifier slotted into the front panel of a 4096-channel Canberra Model 8524 pulse-height analyser.The detecting crystal was a coaxially drifted hollow cylinder of 22.7 cm3 active volume with one closed end of 12.6 cm2 area facing upwards to the sample and situated 5 mm below a 51 mm diameter beryllium window recessed by 1 mm into the top of the detector housing. The crystal was cooled by conduction from a copper finger protruding from the bottom of the detector and maintained in liquid nitrogen. The nominal resolution of the crystal was reported as 1.97 keV at 1332 keV, and its efficiency as a nominal 3.4% measured arbitrarily with 1332-keV y-ray emission from 6OCo with respect to a high-efficiency sodium iodide crystal detector of low resolution.20 The analyser incorporated a video-screen display automatic timing and numerous programmable control facilities and was interfaced to separate output units, namely a Centronics Model 739 dot-matrix printer a Hewlett-Packard Model 7470A graphics plotter and a Radio Shack Realistic Minisette-9 Model 14-812 compact cassette recorder (Canberra Model 5421MK).The analyser energy scale was 0.45 keV per channel. The sample holder was an 85 mm diameter Perspex disc of 2.5 mm thickness ruled with concentric circles to allow precisely reproducible positioning of the pill-box sample capsules. The disc was attached to a brass annulus that could be adjusted up and down along the length of a rigid vertical brass rod calibrated in millimetres. At the lowest point in its adjustment the sample holder was in contact with the top of the detector housing. The multi-channel analyser recorded automatically a net integral count and its statistical variation.2I Typically chan-nels 526-536 ( N channels) accumulating in the range 276.8-281.3 keV were pre-set to cover the whole of the 203Hg band.Also typically channels 523-525 and 537-539 (2n channels), symmetrically situated either side of the 276.8-281.3 keV band were pre-set to calculate a mean background count per channel rn. Then a total background count of ( N + 2n)rn = B may be projected over the entire 2n + N channels considered. If the total background count ( B ) is subtracted from the total count over the 2n + N channels (T) then the count attributable to the 203Hg is T - B = A and its standard deviation is calculated as [A + B + (2n + N) B/2n]&. In this expression in addition to the total number of counts recorded, an additional number of background contributions is added according to the ratio of the total number of channels considered to the number of channels used to estimate the mean background.Clearly if the latter is not a large number compared with the former then additional uncertainty (apart from the random nature of the natural radioactivity compris-ing the total number of counts) is introduced by virtue of the method used to estimate the unknown background incorpor-ated in the 267.8-281.3 keV band. The analyser automatically recorded a percentage error function approximately equiv-alent to the 90% confidence limits of the net count A due to 203Hg which were calculated as k1.65 x 100 [ A + B + (2n + N)B/2n/VA.The relative standard deviation is this function with the factor 1.65 omitted which could be equally well programmed into the analyser output. Preliminary Experiments and Calibration The complete spectra of irradiated blank pill-box capsules, several encapsulated 1-pg samples of selenium dioxide and 0.25-g samples of hair and the standard materials were recorded. The 277-281 keV band net integral due to 203Hg was measured for a particular hair sample containing ca. 5 p.p.m. of mercury over a period of 36 d and several hourly counts were repeated systematically over a 48-h period. Aliquots of 0.25 g were selected from certain large batches of powdered beard shavings in some instances after thorough mixing of the batch and in other instances on a take-it-as-it-comes basis and irradiation and counting analysis was carried out with respect to 0.25-g aliquots of either orchard leaves or Bowen’s kale a ANALYST JANUARY 1986 VOL.111 31 the only y-ray component in the range of interest is a peak at 237 keV which does not interfere with an analysis for standards. Several aliquots of each standard were tested against one another. The mercury(I1) acetate solution containing 203Hg was suitably diluted to a level well below 0.1 pg p1- in dilute acetic acid to give an easily manageable level of radioactivity. A Jencons Finnpipette was used to transfer aliquots of the diluted 203Hg solution in multiples of 5 p1 on to 13-mm circlets of Whatman No. 4 filter-paper fitted into the bottoms of the polythene capsules. These transfers were made on to filter-paper that was previously (a) untreated (b) treated with like aliquots of a 2 pg pl-1 solution of sodium sulphide and (c) treated with like aliquots of a 2 pg pl-l solution of thi0acetamide.1~ This was repeated with a 0.2 pg p1-1 solution of non-radioactive mercury(I1) oxide in 0.2 M nitric acid.The radioactive capsules were closed with lids and counted. The non-radioactive capsules were closed with lids sealed, irradiated and counted. the measurement laboratory in lead containers of thickness not less than 25 mm. (Any vehicle employed for this purpose need not be specially labelled in the UK when the radiation dose-rate does not exceed 0.5 mrem h-l at any part of the exterior surfaces of such lead containers and when no more than 80 mCi of 203Hg are so transported.22) Handle and open the canisters by remote control and handle the capsules with 6-in long forceps.Place each capsule in the sample holder of a suitable y-ray detecting and scaling system and count the radiation for a time sufficient to generate not less than 3000 counts in the 277-281 keV band. Apply corrections for the counting dead-time if appreciable the contribution of 75Se if any to the 277-281 keV peak and for the decay of the 203Hg if counted a significant period of time after counts with which comparison is to be made. The corrected count rates for analytical samples and standards may then be directly compared. Recommended Procedure Results Dry the orchard leaves in an oven at 85 "C for 4 h before use as prescribed.14 Dry the Bowen's kale in an oven at 90 "C for 20 h before use as recommended.15 Weigh exactly 0.25 g of powdered hair or dried powdered vegetable standard into a pill-box capsule filling it almost completely and put the lid into place immediately.Heat-seal each capsule by running a hot iron around the tight join between pill-box and lid. Install 10 or 11 pill-boxes in each canister so that they remain immobile. (Samples may be identified by numbers written on the polythene capsules with Staedtler water-proof permanent-ink pens.) Allow not less than three standards per canister. Irradiate the canisters in a thermal neutron flux of ca. 1012 n cm-2 s-1 for ca. 24 h. Allow not less than 7 d for the radioactivity of the canisters to decay to a manageable level before transportation or measurement of the samples is attempted.Transport the canisters from the reactor area to y-Ray Spectra Fig. 1 shows the y-ray spectrum of an irradiated hair sample containing ca. 6 p.p.m. of mercury. The energy range shown here is from 25.4 to 480.3 keV the cooling-off period ca. 20 d and the count time 1208 s. The band due to 203Hg appears well defined in the 277-281 keV range. The 90% confidence limits calculated automatically for the likely variation of the net area measurement of the band are +3.5%. No peaks due to selenium are in evidence. The only other isotopes that might conceivably give rise to y-rays in the 277-281 keV range are either extremely rare of short half-life or both.23 The 77.6-keV peak due to 197Hg is evident as the right-hand member of the irregular quadruplet located on the characteris-tically raised and distinctly spikey background to the left-hand side of the spectrum and the accompanying mercury X-ray peaks are at 68.9 and 72.9 keV.The completely blank background counted for 10 000 s showed no peaks of any consequence above 100 keV. Fig. 2 shows the y-ray spectrum of a typical irradiated hair sample containing ca. 0.5 p.p.m. of mercury. The energy range shown is 48-645 keV the decay period ca. 20 d and the count time 72 000 s. The 203Hg band is still very clearly defined with respect to its background. The mercury X-ray region is off-scale at the far left of the spectrum. Peaks due to 75Se are clearly evident at 122 136 and 265 keV and there is also a small peak at 400 keV; the 203Hg peak needs to be corrected for a small contribution due to selenium in instances such as this as discussed below.When counted for equally long periods irradiated blank capsules gave peaks of variable intensity due to gold 198Au at 412 keV (Tj= 2.7 d), chromium 51Cr at 321 keV (Th = 28 d) and antimony 124Sb at 604 keV (Tr = 61 d).24,25 Such peaks could also be seen i 32 ANALYST JANUARY 1986 VOL. 111 peak at 145 keV is probably a combination peak due to simultaneity of the 77-keV 197Hg y-ray and 69-keV mercury X-ray events.9 The separately run spectrum of selenium (irradiated as the dioxide) gave well defined peaks at 121,135,265,280 and 400 keV in the approximate ratios 0.9 2.6 1.0 0.4 0.2 (uncor-rected.for variation of crystal efficiency with energy). The ratio of net count rates for the 265- and 280-keV 75Se peaks, measured on 12 different samples gave a mean of 0.42 with a standard deviation (s.d.) of 0.03. This ratio was used to correct the net counts for the 203Hg bands obtained from certain hair samples that contained less than 2 p.p.m. of mercury as discussed later. The y-ray spectrum of orchard leaves was similar to Fig. 2. The 203Hg peak stood equally well above the background. In the earlier stages of the decay however orchard leaves showed particularly strong peaks due to 48Sc and 124Sb and also 122Sb at 565 keV (7’4 = 2.8 d). The y-ray spectrum of Bowen’s kale was relatively free from background peaks, except those due to Sc and Cr. When counted after decay periods of 15 d the correction to the 203Hg band due to 0.08 p.p.m.of selenium in orchard leaves was about -4% and for 0.14 p.p.m. of selenium in Bowen’s kale it was ca. -6%. Stability of the Instrumentation and Half-life of 203Hg A set of ten 60-min counts taken over a 48-h period for the 203Hg band of a strongly radioactive hair sample (yielding net counts of ca. 13 500 associated with backgrounds of ca. 2500) had after correction for an assumed half-life of 47 d a relative standard deviation (r.s.d.) of 1.6%. The r.s.d. values com-puted by the multi-channel analyser for the same set ranged from 1.0 to 1.2%. The difference from 1.6% is not significant and the results reflect essentially only the random nature of nuclear disintegrations.The stability of the detector and electronic amplifying and counting system was thus demon-strated at least over a 48-h period. The ratio of pulse counting time to count time the detector “dead time,” was zero (<0.1%) throughout the work owing to the low level of radioactivity involved. Four plots of log(net count rate) against mean clock time for the count obtained from four independent sets of hourly count data taken with high-activity irradiated hair samples over a 36-d period yielded least-squares lines with correlation coefficients of ca. 0.99 and a mean value of 47.7 d (s.d. 0.5 d) for the apparent half-life of 203Hg in our system. The half-life of 203Hg is variously quoted,ls127 26-30 the most reliable value being 46.91 k 0.14 d.11728 The lower 95% confidence limit of our range is equal to this standard value and confirms the identity and quality of the 203Hg peak in our system.Our value for the half-life may have been slightly raised by a small contribution to the 277-281 keV band from a trace amount of 75Se of half-life 122 d. The stability of the apparatus over a period of weeks was hence established. Homogeneity of Powdered Hair Samples The position of the capsules on the counting platform of the detector could be reproduced to within +_ 0.3 mm horizontally and vertically and the results in Table 1 show that no measurable error could be introduced by any slight misadjust-ment. When counted right way up and upside down typical sample capsules containing irradiated stir-mixed powdered hair gave identical results within the replicate count error discussed above.Ten similarly irradiated 0.25-g aliquots taken from a several-gram amount of well mixed beard shavings containing ca. 6 p.p.m. of mercury gave results with an r.s.d. of 5.4%. A repeat of this exercise with 11 aliquots from shavings containing ca. 30 p.p.m. of mercury yielded an r.s.d. of 2.9%. The shavings therefore appeared reasonably homogeneous in bulk. This variation also includes any non-uniformity of the neutron flux in the atomic reactor and any haphazard Table 1. Effect of vertical and horizontal displacement of sample from the centre of the counting platform Effect of displacement on the count rate* Displacement/mm 0 4 8 12 20 30 40 Vertical? 100 66.5 54.5 44 30 21 15 Horizontal$ 100 97 90 83 60 35 24 * Normalised to 100% for zero displacement.t Effect on the 1332-keV peak of a 6OCo point source. $ Effect on the 279-keV peak from a capsule containing Z03Hg in a typical hair sample. Table 2. Losses of mercury(I1) acetate containing 203Hg from filter-paper soaked in sulphur-containing reagents Relative count in presence of reagent Timeld None Na,S Thioacetamide 0 100 100 100 1 72 80 100 4 45 60 100 evaporation of mercury from the hair in the capsules which are evidently unappreciable. All effects are within +lo% at the outside limits. Linearity of the Detection System with Mass of Mercury and Mercury Losses The accurately measured volume increments of mercury(I1) acetate solution containing 203Hg that were placed at room termperature on filter-paper impregnated with thioacet-amidel3 gave count rates accurately proportional to mass of mercury within the counting error.The range of linearity tested was the equivalent of up to 50 p.p.m. of mercury in a 0.25-g sample of hair. However omission of the thioacet-amide in this test revealed a characteristic and seemingly mysterious loss of mercury (Table 2). Experiments on the irradiation and subsequent counting of accurately measured aliquots of non-radioactive mercury solutions even in conjunction with thioacetamide showed the disappearance of 3&90% of the mercury from “sealed” capsules and experiments on the establishment of an absolute standard for mercury were not continued (see further discus-sion below).Calibration of the Detector and Standardisation of the Method Table 3 details the results of repeated calibration of the instrumentation with the selected standards. The count rates and their instrumental r.s.d. values are listed for five separate runs in which each orchard leaves capsule was compared with two Bowen’s kale capsules. The count rate listed is integral count in the 276-282 keV band minus background per 105 s. The percentage recovery is calculated as net count rate per nanogram of mercury for Bowen’s kale divided by the same for orchard leaves times 100 with respect to the reported mean mercury content (and standard deviation) values of 0.155 k 0.008 pg g-1 for orchard leaves,l4 and 0.174 k 0.030 yg 8-1 for Bowen’s kale.15 The mean recovery of mercury in Bowen’s kale with respect to orchard leaves is 94.2% with an s.d.of 6.2% on 14 results (only just significantly different from Analysis of eight 0.25-g aliquots of homogenised beard shavings at different times with respect to different samples of Bowen’s kale yielded a mean value of 0.54 p.p.m. with an s.d. of 0.05 p.p.m. (r.s.d. 9%). The result of the analysis of the 100%) ANALYST JANUARY 1986 VOL. 111 33 Table 3. Calibration of y-ray detector and direct comparison of 250-mg lots of orchard leaves and Bowen’s kale Orchard leaves Bowen’s kale Count rate 4447 4447 4948 4948 4124 4124 4236 5659* 4704 4704 4183 4183 2095t 2095t Relative standard deviation, O/O 7.9 7.9 6.7 6.7 9.1 9.1 9.1 10.3 7.3 7.3 9.7 9.7 12.7 12.7 Count rate 4537 4509 4652 4862 4661 4840 4423 6293 * 5131 5209 4048 4403 2169t 23781-Relative Recovery of standard mercury from deviation Bowen’s kale, YO YO 9.1 90.9 7.9 90.3 9.7 83.7 8.5 87.5 9.1 100.7 8.5 104.6 12.7 93.0 10.3 99.0 7.9 97.2 7.9 98.6 12.1 86.2 9.1 93.7 17.6 92.1 8.5 101.1 * Amplifier adjustment in this single instance.t Substantially longer decay time elapsed before measurement in these instances. Table 4. Results demonstrating the heterogeneity of dentists’ beard shavings with respect to mercury Dentist Mercury content of consecutive aliquots * p.p.m. A . . . . . . 4.9 4.9 5. I 2.4 B . . . . . . 2.9 3.2 3.8 4.6 c . . . .. . 3.3 4.0 6.3 8.6 D . . . . . . 6.3 11.1 15.9 40.0 E . . . . . . 13.8 19.1 20.9 26.9 F . . . . . . 42.4 56.8 59.0 68. I value are in italic type. * Values within +lo% of their mean with a nearest neighbouring same hair by the previously developed atomic absorption method with respect to a weighed mercury standard was 0.5 p.p.m. with a likely error of k 0.05 p.p.m. ,2 corroborating the accuracy of both methods. Analysis of Beard Shavings from Control Subjects and Selected Dentists Analysis of the individually homogenised beard shavings from 13 randomly selected control subjects with respect to Bowen’s kale as standard gave the following spread of results: 1.33 1.0 0.7 0.65 0.5 0.45 0.4 0.4 0.4 0.3 0.2 0.15 and 0.1 p.p.m. These values are the means of duplicate determina-tions taken to the nearest 0.05 p.p.m.Five of the values below 0.6 p.p.m. were for vegetarian or modest fish- and meat-eating subjects. The range obtained previously for the scalp hair of non-occupationally exposed persons was 0.35-2.7 p.p.m.2 Results obtained on several aliquots taken at random from unmixed hair samples obtained from certain dentists are shown in Table 4 and demonstrate the heterogeneity of this material. In general more than two determinations are necessary to demonstrate the heterogeneity as duplicates may be misleading. Thus a span of up to +lo% can be expected in populations of results obtained on homogeneous hair by this neutron-activation technique when the mercury level is several parts per million as demonstrated above.Hetero-geneity is demonstrated if duplicate results differ by more than this span. If for case A in Table 4 the determinations were made in the order left-to-right homogeneity would be assumed if no more than three determinations were perfor-med. In case B any two consecutive determinations would be insufficient to demonstrate heterogeneity. Dentists C and D both have very heterogeneous shavings and any two of the four determinations in each instance would reveal some degree of heterogeneity. With dentists E and F the matter depends on which samples might be picked. Correction for Selenium and Limit of Detection No correction was necessary for selenium when the mercury content of the hair was above ca. 2 p.p.m. For 0.5-2 p.p.m. of mercury the contribution from 75Se to the net count rate of the 203Hg band was in the range 0-13% and for 0.1-0.4 p.p.m.the correction was variously 13-30%. The trend to lower corrections as the mercury level increases is due merely in our view to the relatively greater signals from mercury. In noting no correlation between the mercury and selenium levels in hair we are in agreement with Sakurai et ~1.31 (although a correlation has been reported32). The limit of detection is defined as three times the square root of the background count.33734 The longest count times were required for the standards and the occasional sample containing 0.1-0.2 p.p.m. of mercury and here the limit of detection was within the range 0.01-0.02 p.p.m. This is comparable to the level of correction usually required for selenium.Conclusion Suitability of Neutron-activation Analysis for the Determina-tion of Mercury in Biological Samples such as Hair and the Use of Standards During the early 1970s disparaging remarks were made about the accuracy of the neutron-activation technique by analysts who preferred an exacting micro-electroanalytical technique35 and in their masterly review Westermark and Sjostrandg suggested specifically that mysterious discrepancies of 70-90% in mercury recovery were general. The UK Atomic Energy Authority in 1978 noted that neutron-activation analysis gave an inexplicably high value for mercury in hair compared with a value only SOo% as great obtained by cold vapour atomic absorption spectrometry.36 Also studies in Italy have shown an apparent range of 0.3-32 p.p.m.for mercury in the hair of non-occupationally exposed people determined by neutron-activation analysis ,37 whereas the result by atomic absorption spectrometry is more plausibly 0.1-3.6 p.p.m.2.38 A low recovery of mercury in a standard would of course give too high a result for the mercury in an unknown sample. The experiments carried out in this work strongly suggest the disappearance of mercury salts by volatilisation during the irradiation process (however implausible this might seem). The use of thioacetamide to render the mercury salt non-volatile was successful at room temperature but not so when irradiation was involved. Japanese workers have stated that the thioacetamide should be kept wet during the irradiation process and then the standards should be stable,39 yet the ranges they have reported for mercury in the hair of non-occupationally exposed people are still suspiciously high, 0.99-13.2,39 0.6-1840 and 1-16 p.p.m.,31 and may not be explained wholly on the basis of a high content of fish in the Japanese diet.This recovery problem is not insurmountable but is unquestionably a difficult one. Indeed the National Bureau of Standards have reported that 18 man-months of effort were required in their laboratory to correct problems with the absolute standard in connection with their certified value for mercury in orchard leaves.41 However the consistency of the numerous experiments carried out with orchard leaves and Bowen’s kale in this work shows that there are no haphazar 34 ANALYST JANUARY 1986 VOL.111 gains or losses of the biologically combined mercury from these materials in the course of the neutron-activation process to which they were subjected. We therefore strongly advocate their use as secondary standards in this application. The NBS report mean values of (a) 0.155 pg 8-1 for mercury in orchard leaves by neutron-activation analysis on 0.25-g samples (b) 0.160 pg 8-1 by a cold vapour atomic absorption technique in which samples of 2-3 g were digested and (c) 0.145 pg g-1 by stable-isotope dilution spark-source mass spectrometry in which 5-g samples were digested.42 They quote 0.155 pg g-1 for general use with approximately two standard deviations 0.015 yg 8-1 representing the outside limits of all the results obtained.14 This exhaustive work leaves little doubt as to the proper value.(The recently reported range of 0.154-0.162 pg 8-1 based on atomic absorption work,43 offers some independent corroboration.) The neu-tron-activation method as tested in this work for the 0.5 p.p.m. level in hair has given results with this secondary standard that agree well with results obtained by the previ-ously reported atomic absorption method,2 and may be taken as accurate. As supplies of NBS orchard leaves are now exhausted we put forward our standardisation of Bowen’s kale against the NBS material. Taking 0.155 p.p.m. as the standard value for orchard leaves with r.s.d. 5% and the found value for Bowen’s kale as 94.2% of 0.174 p.p.m. with r.s.d. 6.2% (Table 3) we arrive at 0.164 p.p.m.with s.d. 0.013 p.p.m. for the level of mercury in Bowen’s kale. The standard deviation can be taken as the likely accuracy of individual determina-tions by the procedure described i.e. 8%. Matters Affecting the Precision of the Neutron-activation Method The best precision r.s.d. 1-2% was obtained with the large signals from the repetitive experiments designed merely to test the stability of the instrumentation. In the large-signal test of homogeneity of hair shavings and neutron flux carried out with respect to one standard the r.s.d. range was increased to >5%. If the variability introduced by analyses carried out in different runs with different standards is now added in particular for hair containing a low normal level of ca.0.5 p.p.m. of mercury the r.s.d. is increased to ca. k970 as noted in the calibration and standardisation section above (see also the population of r.s.d. values shown in Table 3). The precision is limited by the low level of mercury in the secondary standard. The r.s.d. of 9% can be apportioned as ca. 8% to the standard (which contains the minimum 0.1-0.2 p.p.m. of mercury) and 4% to the sample (containing 0.5 p.p.m.). Hence ca. 8% is the lower limit for the likely reproducibility of determinations on hair that contains 0.5 p.p.m. or more of mercury whereas the imprecision must rise to at least ca. 12% for samples of hair that contain only 0.1-0.2 p.p.m. of mercury. This is probably increased to say 15% by the need for high corrections for the contribution of 75Se to the 203Hg mercury band.This represents a span of 30% in standard deviation alone (albeit for the worst possible situations to be met with in practice) and we have therefore quoted the lower mercury levels only to the nearest 0.05 p.p.m. Several replicate determinations are required in order to ensure reasonable precision in the determination of these lower mercury levels. The precision may be raised by a number of measures that would give larger signals. Another homogeneous secondary standard that contains more mercury might be sought. Longer and more costly irradiation times and count times might be employed. A higher neutron flux with an attendant higher core temperature could be used with increased risk of chemical attack on the samples standards and containers.To take measurements earlier in the decay period however does not greatly improve the signal to noise ratio as then the general background is higher in addition to the signal of interest. Scope of the Neutron-activation Method and Alternative Methods World-wide interest in neutron-activation analysis continues, in particular for the determination of mercury in dentists’ hair,44,45 and in general for the simultaneous determination of several elements in hair.46-48 Studies emphasising accuracy and precision are rare although for example it has been reported that preliminary precipitation of mercury as the thiocyanate prior to activation can give a very low limit of detection (ca. 0.001 ~ g ) . ~ 9 Several modifications of the basic cold vapour atomic absorption technique2>50-59 and even some spectrophotometric methodsW62 have been preferred for the determination of mercury in biological samples such as hair, probably because by these means the analyst has the analysis under his complete control within the resources of his own laboratory.Radiometric recovery checks in such methods are rare63 and results may come into question. The neutron-activation method used carefully as described here with a suitable secondary standard is straightforward safe and reliable with regard to accuracy and precision and need not be considered only the province of the specialist. We thank Dr. Susan J. Parry and Mr. G. D. Burholt of the Imperial College Research Reactor Centre for freely discuss-ing their own analytical work and for making smooth arrangements at minimum cost for our programme.We also thank Dr. L. R. Day of NELP for his interest and support and Mr. Alan Wates for invaluable help with the maintenance of equipment and delivery of radioactive samples. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. References Pritchard J. G. McMullin J. F. and Sikondari A. H. Br. Dent. J. 1982 153 333. McMullin J. F. Pritchard J. G. and Sikondari A. H., Analyst 1982 107 803 and references cited therein. Iyengar G. Y. Kollmer W. E. and Bowen H. J. M. “The Elemental Composition of Human Tissue and Body Fluids,” Verlag Chemie Weinheim 1970 p. 52 and references cited therein. Nixon G. S . and Smith H. J. Oral Ther. Pharmacol.1965,1, 512. Coleman R. F.,. Cripps F. H. Stimson A. and Scott H. D., Atom 1967 Nos. 1-3 12. Sinclair P. M. Turner P. R. and Johns R. B. J. Prosthet. Dent. 1980 43 581. Bowen H. J. M. Adv. Act. Anal. 1969,1,101 and references cited therein. Girardi F. and Guzzi G. Adv. Act. Anal. 1969 1 137 and references cited therein. Westermark T. and Sjostrand B. Adv. Act. Anal. 1972,2, 57 and references cited therein. Aliev A. I. Drynkin V. I . Leipunskaya D. I. and Kasatkin, V. A. translated by Benny B. “Handbook of Nuclear Data for Neutron Activation Analysis,” Program for Scientific Translations Jerusalem 1970 p. 19. Dzhelepov B. S . and Peker L. K. translated by Allan D. L., “Decay Schemes of Radioactive Nuclei,” Pergamon Press, London 1961 pp. 606-607.The Radiochemical Centre “The Radiochemical Manual Part 1 Physical Data,” HM Stationery Office London 1962 p. 37. Sairenji E. Moriwaki K. Shimizu M. and Noguchi K., International Conference on Mercury Hazards in Dental Practice West of Scotland Health Boards Glasgow Scotland, September 2-4 1981 Proceedings and Discussion Volume, Paper 15 ANALYST JANUARY 1986 VOL. 111 35 14. 15. 16. 17. 18. 19. 20. 21. 22, 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. Cali J. P. “Standard Reference Material No. 1571 Revised Certificate of Analysis August 31 1977,” National Bureau of Standards Gaithersburg MD. Bowen H. J. M. At. Energy Rev. 1975 13 451. Uriano G. A. “Standard Reference Material No. 1645, Revised Certificate of Analysis November 16 1978,” National Bureau of Standards Gaithersburg MD.“Certified Standard Reference Material No. 1577A,” National Bureau of Standards Gaithersburg MD. Gorsuch T. T. “The Destruction of Organic Matter,” Pergamon Press Oxford 1970 pp. 79-84. University of London Reactor Centre “Brief Guide to the Irradiation of Samples in the University of London Reactor,” Imperial College Silwood Park March 1980. “Series 7000 Ge(Li) Spectrometer Systems Instruction Man-ual,” Section 7.3 Detector Specifications Canberra Industries, Meriden CT 1973. “Series 85 Multichannel Analyser Operator’s Manual,” Can-berra Industries Meriden CT 1981 Appendix A . l . Department of the Environment “Code of Practice for Carriage of Radioactive Materials by Road,” HM Stationery Office London 1975.Madson A. A. in Crouthamel C. E. Editor “Applied Gamma-ray Spectroscopy,” Pergamon Press London 1960, pp. 365 and 366. Parry S . J. Imperial College of Science and Technology, London personal communication 1983. Parry S . J. “Evaluation of Low Density Polyethylene for Irradiation Capsules,’’ Report ULRC/RES/34 University of London Reactor Centre Imperial College Silwood Park, March 1981. Cook G. B. and Duncan J. F. “Modern Radiochemical Practice,” Clarendon Press Oxford 1952 p. 384. Slater D. N. “Gamma-Rays of Radionuclides in Order of Increasing Energy,” Butterworths London 1962 p. 77. Eichholz G. and Krzyzewski A. Can. J. Phys. 1956 34, 1167. Thiry H. Bull. SOC. R. Sci. Li&e 1957 26 No.1 29. Crouthamel G. E. “Applied Gamma-ray Spectrometry,” Pergamon Press London 1960 p. 365. Sakurai S. Tsunoda H. Terai M. Yukawa M. Tomura, K. and Suzuki-Yasumoto M. in “Nuclear Activation Tech-niques in the Life Sciences 1978,” Proceedings of an Interna-tional Symposium on Nuclear Activation Techniques in the Life Sciences held by the International Atomic Energy Agency in Vienna 22-26 May 1978 International Atomic Energy Agency Vienna 1979 pp. 573-581. Mangal P. C. and Gulati N. Proc. Nucl. Phys. Solid State Phys. Symp. 1978 21B 356; Chem. Abstr. 1977 98 104. Currie L. A. Anal. Chem. 1968 40 586. Stranshiskii A. G. Khomyakov C . K. Shakun N. A., Serykh N. V. and Sevenn N. F. Biofizika 1975,4 621. Pillay K. K. S. Thomas C. C. Jr. Sondel J. A. and Hyche, C.M. Anal. Chem. 1971 43 1419. Collier T. R. “Experience of Using the Magos Atomic Absorption Method for the Determination of Mercury in Biological Materials,” Environmental and Medical Sciences Division Atomic Energy Research Establishment Harwell, Memo AERE-M2930 Reference HL 781399 (C5) February 1978. Clements G. F. Rossi L. G. and Santaroni G. P. in “Nuclear Activation Pechniques in the Life Sciences 1978,” Proceedings of an International Symposium on Nuclear Activa-tion Techniques in the Life Sciences held by the International Atomic Energy Agency in Vienna 22-26 May 1978 Interna-tional Atomic Energy Agency Vienna 1979 pp. 527-543. Pallotti F. Bencivenga B . and Simonetti T. Sci. Total Environ. 1979 11 69. 39. 40. 41. 42. 43. 44. 45.46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. Takeuchi T. Hayashi T. Takoda J. Koyama M. Kozuka, H. Tzuji H. Kusaka Y. Ohmori S . Shinogi M. Aoki A., Katayama K. and Tomiyama T. in “Nuclear Activation Techniques in the Life Sciences 1978,” Proceedings of an International Symposium on Nuclear Activation Techniques in the Life Sciences held by the International Atomic Energy Agency in Vienna 22-26 May 1978 International Atomic Energy Agency Vienna 1979 p. 545. Imahori A. Fukushima I. Shiobara S. Tomura K., Suzuki-Yasumoto M. Yukawa M. and Terai M. in “Nuclear Activation Techniques in the Life Sciences 1978,” Proceedings of an International Symposium on Nuclear Activa-tion Techniques in the Life Sciences held by the International Atomic Energy Agency in Vienna 22-26 May 1978 Interna-tional Atomic Energy Agency Vienna 1979 pp.563-571. La Fleur P. D. J. Radioanal. Chem. 1974 19 227. Alvarez R. National Bureau of Standards Gaithersburg, MD personal communication 1984. Lau 0.-W. Hon P.-K. Cheung C.-Y. and Wong M.-C., Analyst 1984 109 1175. Yatim S. Majalah BATAN 1983,16,60; Chem. Abstr. 1983, 101 105296h. Lin S . M. Chiang C. W. Tseng C. L. and Yang M. H., Radiochem. Radioanal. Lett. 1983 56 261. Kim N. B. Report IAEA-R-2 535-F (1982) INIS Atomindex, 1983 14 Abstr. 753711; Chem. Abstr. 1983 99 83323a. Das H. A. Hoede D. Nieuwendijk B. J. T. Van der Sloot, H. A. Teunissen G. J. A. and Woittiez J. R. W. Rep. ECN-131 Netherlands Energy Foundation 1983; Chem. Abstr. 1983,99 3558211. Cortes E. Cassorla V. and Munoz L. Radiochem. Radio-nal. Lett. 1981 50 177. Biso J. N. Cohen I. M. and Resnizky S. M. Radiochem. Radioanal. Lett. 1983 58 175. Roschig M. and Wuenscher R. G. Zentralbl. Pharm. Pharmakother. Laboratoriumsdiagn. 1982 121 893. Shiwastava A. K. and Tandon S . G. Int. J. Environ. Anal. Chem. 1982 11 22. Dmitriev M. T. Grasovskii E. I. and Shashchev A. Y. Gig. Sanit. 1983,50; Chem. Abstr. 1983,99 186595. Campe A. Velghe N. and Claeys A. At. Spectrosc. 1982, 3 122. Yamamoto J. Yoshida M. and Kaneda Y. Eisei Kagaku, 1983 29 418. Baba T. Ohmiya S. Hosokawa M. and Ishibashi T., Seikatsu Eisei 1983,27,258; Chem. Abstr. 1983,100,114127k. Yanagisawa M. Suzuki H. Kitagawa K. and Tsuge S., Spectrochim. Acta Part B 1983 38 1143. Robinson J. W. and Skelly E. M. Spectrosc. Lett. 1981 14, 519. Gardner D. and Dal Pont G. Anal. Chim. Acta 1979 108, 13. Peter F. and Strunc G. Clin. Chem. (Winston-Salem NC), 1984,30 893. Chung K. C. Chungang Uitaechi 1982 7 1; Chem. Abstr., l982,97,3945a. Voronin V. M. Lamentova T. G. Nikolaeve E. C. and Rubtsov A. F. Sud.-Med. Ekspert. 1981 24 47; Chem. Abstr. 1982,97 18505f. Krylova A. N. and Rubtsov A. F. Sud.-Med. Ekspert. 1982, 25 48; Chem. Abstr. 1982 97 176299~. Lendero L. and Krivan V. Anal. Chem. 1982,54,579. 62. 63. Paper A5lI 82 Received May I7th I985 Accepted August 6th 198