16 Analyst January 1979 VoL. 104 PP. 16-34 Evaluation of a Method for Determination of Total Antimony Arsenic and Tin in Foodstuffs Using Measurement by Atomic-absorption Spectrophotometry with Atomisation in a Silica Tube Using the Hydride Generation Technique W. H. Evans F. J. Jackson and Dorothy Dellar Department of Industry Laboratory of the Government Chemist Cornwall House Stamford Street Londo?z, SEl ONQ A method is described for the determination of antimony arsenic and tin in foodstuffs in which organic matter is destroyed using a wet-oxidation pro-cedure except for arsenic in samples of marine origin in which organic matter is destroyed by the dry-ashing technique. Each element is obtained in the highest valency state and converted into the respective normal hydride with sodium tetrahydroborate( 111) prior to atomisation in a flame-heated silica tube and atomic-absorption spectrophotometric measurement.The optimum conditions for this procedure are discussed and direct and indirect interference effects are described. The accuracy of the procedure is assessed for each element and where possible the accuracy of the method in application is considered Standard deviations of the results for levels normally found in foodstuffs have been calculated and derived limits of detection and confidence intervals are given. Keywords A ntimoizy determination ; avsenic determination ; tin determination ; foodstufls analysis; atomic-absorption spectrophotometry Antimony and arsenic are metalloid elements that occur naturally in foodstuffs and tin may be present at high levels originating from the canning of foodstuffs.None of the three elements has a known essential function in human physiology although it has been reported that tin is an essential element in animal nutriti0n.l Organic arsenicals have been used as additives in feeding stuffs to promote growth in farm animals. The toxic nature of arsenic in inorganic form and lower valency state is widely recognised. In the UK the level of total arsenic in foodstuffs is controlled under the Arsenic in Food Regulations 1959.2 Traditional recommended methods of analysis for these three elements in foodstuffs have used spectrophotometric end measurement^.^*^ Since the beginning of this decade attention has been directed to methods involving evolution of metalloid elements as their hydrides, followed by measurement with atomic-absorption spectroscopy.Various sources of nascent hydrogen have been proposed for the preparation of the hydrides. Of these sodium tetra-hydr~borate(III)~-~ is now generally accepted as the most suitable provided that for certain of the end measurements each element is first reduced to the lower valency state using for example iodide, Arsenic is normally measured at a wavelength of 193.7 nm and it was recognised at an early stage that the use of an argon - hydrogen entrained air flame was necessary in order to reduce flame absorption in this wavelength region.8 Subsequently the direct use of a flame was dispensed with and the arsine produced was entrained with argon in an electrically heated tube.g From combinations of these variables a number of systems have been evolved and applied to metalloid elements in foodstuffs for example by Fiorino et aZ.1° using a ternary acid wet oxidation followed by a pre-reductant stage with iodide evolution of the hydride with sodium tetrahydroborate(II1) in an automated system and measurement in a nitrogen -hydrogen entrained air flame.Crown Copyright EVANS JACKSON AND DELLAR 17 In this laboratory estimates are obtained for a number of elements for the total diet survey of the UK including the three elements considered here. The ability to use a common digest for the determination of ten or more elements reduces the cost of analysis. The extreme sensitivity of the heated silica tube method proposed by Thompson and Thomerson,ll together with the ability to measure when possible in either valency state and coupled to a digestion procedure slightly modified from that previously described,12 permits measure-ments to be made on small volumes of the common digest.An evaluation of this combina-tion is presented in this paper. Method Reagents be prepared with distilled water. Ultrar or equivalent grade acid be used. All reagents must be of the grade indicated or of analytical-reagent grade; solutions should Nitric acid sp. gr. 1.42. It is recommended that BDH Aristar Hopkin and Williams Sulphuric acid sp. gr. 1.84 (1 + 19) and (1 + 99). Recommended grade as for nitric acid. Hydrochloric acid (1 + 1). Hydrogen peroxide 30% m/V. Sodium hydroxide solution 5% m/V.Antimony( V ) chloride. Arsenic( V ) oxide. Sodium tetrahydroborate(III) solutions 1 .O% 1.5% and 3.0% mlV. Magnesium nitrate hexahydrate solution 50% m/V. Cz@ferron (ammonium N-nitrosophenylhydroxylamine) solution 5% nzj V . Chloroform . Antimony(V) standard solution. Freshly prepared. Dissolve 1.228 g of antimony(V) chloride in 1 1 of hydro-chloric acid (1 + 1) to give a solution containing 500 mg 1-1 of the element. Dilute 10 ml of this solution to 1 1 with sulphuric acid (1 + 19) to give a solution containing 5 mg 1-1 of the element. Immediately before use dilute 10 ml of the latter solution to 250 ml with sulphuric acid (1 + 19). Dissolve 1.534 g of arsenic(V) oxide in 100 ml of sodium hydroxide solution (5% m/V) and dilute to 600 ml with water.Add 30 ml of hydrochloric acid (1 + 1) and dilute to 1 1 with water to give a solution containing 1000 mg 1-1 of the element. To 10 ml of this solution add 10 ml of hydrochloric acid (1 + 1) and dilute to 1 1 with water to give a solution containing 10 mg 1-1 of the element ; immediately before use take 5 ml of the latter solution and dilute to 250 ml with sulphuric acid (1 + 19) giving a solution containing 0.2 mg 1-1 of arsenic. Tin ( I V ) standard solution. A primary standard solution specially prepared for atomic-absorption spectrophotometry and containing 1000 mg 1-1 of tin can be purchased from commercial sources. Dilute 5 ml of the primary standard solution to 100 ml with hydro-chloric acid (1 + 1) to give a solution containing 50 mg 1-1 of the element and immediately before use dilute 5 ml of the latter solution to 500 ml with sulphuric acid (1 + 99) to give a solution containing 0.5 mg 1-1 of tin.Prepare working standard solutions for arsenic and antimony separately taking 0 1 2 4 6 8 and 10 ml of the 0.2 mg 1-1 standard solutions and diluting to 100 ml with sulphuric acid (1 + 19) to give solution concentrations of 0 0.002 0.004, 0.008 0.012 0.016 and 0.020 mg l-l respectively of arsenic and antimony. For tin dilute 0- 1- 2- 4- 6- and 8-ml aliquots of the solution containing 0.5 rng 1-1 of tin to 100 ml with sulphuric acid (1 + 99) to give solutions containing 0 0.005 0.01 0.02 0.03 and 0.04 mg 1-1 of tin respectively. Store all solutions in calibrated flasks; in the absence of reducing atmospheres these solutions are stable for up to 1 month.This solution contains 0.2 mg 1-1 of antimony. Arsenic(V) standard solution. Working standard solutions. Apparatus All glass apparatus should be kept full of 1 N nitric acid when not in use; the glass apparatus used for preparing diluted sample solutions for measurement of tin should be kept full of 1 N hydrochloric acid 18 EVANS et al. DETERMINATION OF ANTIMONY ARSENIC AND TIN Analyst Val. 104 Silica T-piece tubes. Those used for the procedure described are 150mm long with internal diameter 5 mm and with a side-arm 70 mm long. A separate tube must be kept specifically for each element and must be pre-conditioned for that element by repeated applications of the most concentrated working standard solution until a constant response is obtained.In normal use for foodstuff digests such a tube has a limited life expectancy before the internal walls become poisoned or excessively coated with sodium sulphate. The useful life expectancy is about 400 injections. Nitrogen is supplied at a flow-rate of 0.5 1 min-l together with an air bleed of variable flow-rate of 5 10 and 50 ml min-l for antimony arsenic and tin respectively. The gas supply is connected to the hydride generator chamber. That used is 11 cm long with external diameter 2 cm with a septum entry point 3 cm from the base which is fitted with a drainage tap. The gas entry tube to the chamber reaches to within 5 cm of the base and the exit tube is connected by a 60 cm long silicone-rubber tube to the silica tube which is mounted on a 10-cm single-slot acetylene - air burner; the silica tubes should be readily interchangeable.Disposable (polypropylene syringes and needles. These should be of 1-ml capacity fitted with 21-gauge needles for standard and sample solutions and of 5-ml capacity for sodium tetrahydroborate(II1) solutions. Separate 1-ml syringes should be used for standards and samples and discarded after each series of measurements. This should be of the desired sensitivity preferably with an enclosed burner chamber. The instrument should be fitted with an integrator having a response time not exceeding 1 S. Electrodeless discharge emission sources are preferred for both arsenic and antimony. Gas supply. Hydride generator chamber. Atomic-absorption spectvoplzotometer. A hollow-cathode emission source can be used for tin.Procedure Preparation of sample digests Destroy the organic matter from 5-20 g of foodstuff depending upon the nature of the foodstuff according to Method (1)C of the Analytical Methods C0rnmittee.1~ Ensure that nitric acid (sp. gr. 1.42) is added prior to 5 ml of sulphuric acid (sp. gr. 1.84) at the com-mencement of the oxidation and take suitable precautions to avoid an excessively violent reaction. When oxidation is complete cool the mixture dilute with 10ml of water cool and add 1 ml of hydrogen peroxide (30% m/V); boil until white fumes of sulphur trioxide are evolved. Cool the mixture dilute with 10ml of water add a few drops of nitric acid (sp. gr. 1.42) (to remove any peroxides) and boil gently to fumes.Repeat the boiling to white fumes twice according to the method described previously. Cool the mixture and dilute to 100 ml with water to give a nominal 5% V/V sulphuric acid digest which is colour-less and contains no suspended solids. At the same time prepare two reagent blank solutions, from volumes of acid used in the sample oxidation and treat these in a manner identical with that for sample digests. Preparation of digests for arsenic in samples of marhe origin For any material of marine origin a method based on that of Leblanc and Jackson14 must be used. Weigh up to 1.00 & 0.01 g of wet sample (up to 0.25 g of dry sample) into cleaned silica dishes and add 8 ml of magnesium nitrate hexahydrate solution (50% m/V). Disperse the sample well and evaporate to dryness in an oven at 105 "C.Transfer the sample into a muffle furnace warmed to 100 "C and increase the temperature in steps of 50 "C allowing the heating to continue for 30 min at each higher temperature until a temperature of 500 "C is attained. Continue the ashing overnight for convenience. When the sample is cool dissolve it in sulphuric acid (1 + 19) and dilute the solution to 100 ml with sulphuric acid (1 + 19). Prepare simultaneously two reagent blank solutions. Additional stage in the preparation of sample digests in the presence of excessive amounts of t i n When measuring antimony or arsenic if prior indication is that the sample digest solution contains tin at concentrations that interfere i.e. greater than 0.5 mg 1-1 and 0.2 mg 1-1, respectively this tin must be removed by extraction with cupferron solution according to the following procedure January 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY 19 Cool a diluted digest to 4 "C.Preferably digest a separate sample to prevent excessive loss of sulphuric acid from fuming. Add 2 ml of cupferron solution (5% m/V) and 10 ml of chloroform and shake vigorously for 2min. Discard the chloroform layer repeat the cupferron addition and extraction and extract with a further 10 ml of chloroform. Evaporate to white fumes cool and dilute to 100 ml for a separate digest or to the volume originally taken. Measurement If the elemental concentration in the digest exceeds the calibration range dilution must be made with sulphuric acid (1 + 19). For determination of tin 2 ml of the digest must be diluted to 10 ml with water and if the calibration concentration range is exceeded further dilution made with sulphuric acid (1 + 99).When dilution of the solution is required the actual concentration measured should always be greater than 0.008 mg 1-1 for antimony and arsenic and 0.01 mg 1-1 for tin. Adjust the emission sources in a suitable atomic-absorption spectrophotometer to give a maximum sensitivity to noise ratio according to the maker's instructions at wavelengths of 217.6nm for antimony 193.7nm for arsenic and 224.6nm for tin Insert the burner assembly fitted with the silica tube into the radiation beam and adjust the position of the tube vertically horizontally and on the horizontal axis to obtain maximum response.Ensure that the nitrogen and air flows into the hydride generator chamber are as described and heat the tube with a normal acetylene - air flame. After heating for 10 min ensure that the base line is free from drift on a millivolt recorder. By means of a syringe introduce a volume of sodium tetrahydroborate(II1) solution, appropriate to the element being measured into the hydride generator chamber. The volumes will be 2 ml of lyo 2 ml of 1.5% and 1 ml of 3% solution for antimony arsenic and tin respectively. Inject 1 ml of the standard solution of highest concentration and repeat renewing sodium tetrahydroborate( 111) and standard solutions until a uniform response is obtained washing the generator chamber with distilled water between measure-ments. Subsequently obtain a constant response for the zero standard solution.Arrange the sample standard and blank solutions in random order and record for each element in turn the response from each solution in at least duplicate. Measure the responses for each sample and standard calculate the average and subtract the mean of the relevant blanks. For each element calculate the response at each of the standard concentrations as a measure of the lowest concentration and calculate the average. If any individual standard concentration falls outside the range &20% of this average for arsenic or antimony and -+lo% for tin that response must be rejected and a new average calculated. This should be an infrequent occurrence except for the lowest standard in the series. If wz is the mass of sample taken then for a net sample response rl within the calibration range, that sample will contain 100 cr,/mr mg k g l for arsenic and antimony and 500 cr,/mr mg kg-l for tin.The digest can be used as prepared for measurement of antimony and arsenic. Let the average response be r for the lowest concentration c of one of these elements. Experimental Each of these elements is normally present in foodstuffs at low concentrations and it is these levels which dictate the range of measurement and the experimental conditions of the described method. Arsenic may be concentrated in animal offal and in samples of marine origin but a moderate dilution permits measurement within the calibration range. Tin may be present at elevated levels in canned foodstuffs but provided a limiting coefficient of variation is accepted such levels can be determined similarly following dilution.The requirements in the development of a complete analytical method have been described e1~ewhere.l~ The most important requirements for any method are that the variables in each stage of the method be considered in order to obtain the maximum response for the low levels present and the response and the variation in response for each stage of the method be investigated to obtain conditions for which no systematic bias exists. There will inevitably be exceptions to the application of the resulting procedure and these must be defined and if important alternative procedures recommended 20 Analyst vol. 104 Destruction of Organic Matter Gorsuch16 originally showed that no loss of inorganic antimony or arsenic would be expected during wet oxidation with nitric and sulphuric acids while such a wet oxidation has been recommended for the determination of tin in foodstuffs.* Subject to the stability of these elements in sulphuric acid solution this would appear to be a satisfactory procedure to adopt.In the preparation of these digests however several aspects require consideration. While the chemical states of these elements in foodstuffs may remain unknown the possibility must be considered that elements exist as volatile forms or in forms intractable to destruc-tion. There are no recorded instances for antimony and tin but arsenic in samples of marine origin may exist in different inorganic valency states as specific methylated arsenic compounds or as complex organic compounds of high relative molecular mass and of unknown composition.17J8 It is difficult to release arsenic from the organically bound states by normal wet oxidation and alternative methods of destruction have been recom-mended.l*@ This inability of wet oxidation to release all the arsenic can be illustrated by the results obtained during this investigation on a wet fish homogenate sample using several different analysts.Levels of 1.4 1.4 3.5 7.0 7.7 11.4 and 14.3 mg kg-l were returned to give an average and standard deviation of 6.7 and 5.0 mg k g l respectively. Clearly even without definition of the method these results reflect variation that is not under contr01.l~ Dry ashing in the presence of magnesium nitrate was chosen as the most suitable means of releasing the total arsenic from samples of marine origin.Because levels are generally high relatively small masses of sample are acceptable for the range of measurement adopted, provided that such samples are homogeneous. It is preferable not to overload even this system as is exemplified in Table I. There is a limiting mass of sample that may be taken, which could depend upon the nature of the organoarsenical or the sample itself. EVANS et al. DETERMINATION OF ANTIMONY ARSENIC AND TIN TABLE I ARSENIC CONTENT (mg k g l ) FOUND IN SAMPLES OF MARINE ORIGIN USING DRY ASHING OF DIFFERENT MASSES Mass of sample ashed/g I A \ Foodstuff 0.1 0.25 0.6 1 .o 2.0 NBS tuna . . . . 3.0 3.0 3.0 2.5 2.7 2.7 1.7 1.6 1.2 1.2 Fish meal .. - 11 12 6 4 -During the measurement of the total levels of these elements in the hydride formation stage it is necessary to start the sodium tetrahydroborate(II1) reduction with the element entirely in one valency state. Most workers have preferred the lower valency state attained by means of pre-reduction with for example iodide. Despite a loss of sensitivity (the responses obtained in the lower valency states are 2.0 and 1.4 times that given in the higher valency states for antimony and arsenic respectively) we prefer the higher valency states, which are more readily and simply obtained. Tin in dilute solutions can exist only as tin(IV), while application of wet oxidation or dry ashing to the destruction of organic matter yields arsenic in the pentavalent form.This was found not to be the case for antimony which invariably remains in the trivalent state during wet oxidation with nitric acid. The con-version of antimony into the pentavalent state however is accomplished satisfactorily by the hydrogen peroxide stage described in the procedure. Finally it has been reported that no loss of antimony or tin occurs in solution at a pH less than It would be expected that both antimony and arsenic would be stable in 3-5% sulphuric acid in the higher valency state in the absence of reductants. Tin can however, hydrolyse and precipitate in similar circumstances. No losses of tin were noted for solutions of concentration up to 10 mg 1-1 (equivalent to a determination of 100 mg k g l in foodstuffs) in 5% sulphuric acid for 1 month; losses were observed within days for tin solutions of con-centration greater than 1 mg 1-1 in 1% sulphuric acid.Conditions for Evolution of Hydrides I t is well known that the thermal breakdown of these hydrides (taken to be of compositions SbH, ASH and SnH,) is greatly affected by the heated surface with which they are i January 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY 21 contact. The pre-conditioning of the silica tubes described in the method indicates the need for a catalytic film of the element on the silica surface before consistent response is achieved. The fragile nature of this catalytic film can be shown by the observation during this investigation that whenever the response from pure solutions was disturbed for some reason to give a higher or lower response the following injection compensated to give a lower or higher response thus maintaining the average response.It follows that poisoning of this catalytic film would be expected from evolved hydrides of other elements deposited on the silica tube surface or from carryover from the hydride chamber in particular of sodium sulphate. This would have the effect of reducing the sensitivity during repeated use. It has also been observed during this evaluation that variation increases as a tube ages with use. It was impossible to restore the silica tubes to their original performance by means of acid cleaning or scouring, and it was considered necessary to replace each tube after the use described in the procedure. To enable facile replacement therefore an electrically heated silica tube was not considered.Conditions that can vary in the hydride evolution stage for each element are as follows: silica tube dimensions; the volumes of sample digest solution used; the flow-rate of the carrier gas and any auxiliary oxidant; the volume and strength of sodium tetrahydro-borate(II1) solution; and the acidity of the sample digest solutions. The length of the silica tube is determined by the size of the enclosed atomic-absorption burner chamber and it was found that maximum response was obtainable with tubes of internal diameter 5 mm. The residence time for the evolved hydrides in such tubes is short and precludes the use of instrument integration times exceeding 1 s. As it has been reported that 1 ml of an acidified solution is adequate for hydride evolution,ll the adoption of such volumes maximised the use of the prepared foodstuff digests.When using such a system it was found that the flow-rate of the carrier gas nitrogen was most sensitive for the range 0.4-0.6 1 min-l for each element. It has been reported that the addition of air into the carrier gas stream improves the relative response,21 but for the present procedure no such improvement was observed. It was particularly noticeable, however that additions of air decreased the variation in response for a series of injections for each element e.g. at 0.04 pg of tin with 0,lO and 50 ml min-1 of air added to the nitrogen flow relative standard deviations of 0.10 0.07 and 0.04 respectively were obtained for replicated injections.The optimum additions of air described in the procedure were there-fore adopted. While it is desirable to maximise the amount of sodium tetrahydroborate(II1) used to overcome consumption of this reagent by other elemental species present the amount that can be used in the present procedure for antimony(V) and arsenic(V) is limited by the need to maintain acidic conditions for the establishment of an equilibrium between the pentavalent and the lower valency states in sit% prior to hydride evolution,22 and the need to avoid an over violent evolution of hydrogen and ignition of gases leaving the silica tube. Amounts of sodium tetrahydroborate(II1) between 20 and 30 mg for antimony(V) and 30 and 40 mg for arsenic(V) were found to give a relatively constant response and the lower amounts were subsequently used with the proviso that these amounts were contained in a 2 ml-volume of solution.For tin(1V) the acidity of the solution used for measurement is much reduced and constant response is attainable for the range 10-30 mg of sodium tetrahydroborate(II1) ; 1 ml of 3% m/V reagent could be used without danger of ignition. In all instances it was found advisable to add the sample solution to the sodium tetrahydroborate(II1) solution in order to ensure thorough mixing. It was also noted that these solutions of sodium tetra-hydroborate(II1) gave consistent responses for a period of at least 3 h. The prepared digest solutions are 3 4 % V/V in sulphuric acid and for the range 2.540% there is no change in response from hydride evolution of either antimony(V) or arsenic(V).The response for tin(1V) is constant for acid concentrations of 0.5-1.0% V/V sulphuric acid, and hence it is necessary to dilute the digest solution 5-fold with water to fall within this range and make any further dilution with 1% V/V sulphuric acid to ensure a constant acid concentration not exceeding 1%. It was also found that blank responses for each element were constant within the respective ranges of sulphuric acid concentrations. For atomic-absorption measurements stable base lines with minimum but detectable noise levels are advantageous and electrodeless discharge emission sources are preferable for both arsenic and antimony. Comparison of results measuring with and without background A separate tube was therefore retained for each element 22 EVANS et al.DETERMINATION OF ANTIMONY ARSENIC AND TIN Analyst VoZ. 104 correction on foodstuff digests displayed no significant difference and results described in this evaluation were obtained without background correction measurement being made on a Pye Unicam SP1900 spectrophotometer. It may be noted that the condition of the optical system of any spectrometer when measuring arsenic at 193.7 nm is important. The results contained in this evaluation were obtained over a period exceeding 12 months with the optical system in average condition and so reflecting normal practice. A new instrument would be expected to have lower noise levels that could be reflected in slightly lower relative standard deviations (R.S.D.) than those listed.Using the procedure described the average R.S.D.s of a single injection for a series of 10 replicates for each element after blank correction by one analyst are shown a t the top of Table 11. TABLE I1 RELATIVE STANDARD DEVIATIONS OF SINGLE INJECTIONS FOR DIFFERENT EXPERIMENTAL CONDITIONS Degrees of Parameter freedom Replicate injections. . 9 Triplicate injections in presence of non-interfering element . . 8 Triplicate standard injections in a series containing standards and non-interfering elements . . * . 8 All standard injections in the above series 11 Triplicate injections in presence of inter-Triplicate standard injections in a series containing standards and interfering elements . . 8 All standard injections in the above series 11 fering element .. 8 Element concentration/mg 1-1 r Antimony(V) Arsenic(V) Tin( IV) *- 0.004 0.020 0.004 0.020 0.01 0.04 0.13 0.06 0.15 0.06 0.11 0.04 A \ 0.16 0.03 0.14 0.08 0.09 0.05 0.10 0.05 0.18 0.06 0.11 0.04 0.12 0.03 0.27 0.07 0.14 0.06 0.08 0.06 0.15 0.11 0.10 0.05 0.10 0.04 0.11 0.09 0.04 0.05 0.17 0.04 0.12 0.08 0.10 0.05 Calibration These R.S.D.s are higher than those reported for automatic systems of measurement of these elements e.g. 0.018 for antimony and arsenic and 0.022 for tin,23 and this would be expected. The latter were obtained with unknown solution volumes at concentrations of 0.05 0.10 and 0.05 rngl-l respectively which may mean the determination if not the measurement at one time of larger amounts of element than the 440 ng tested in this evaluation (Table 11).The R.S.D. of single injections is only one contributor in practice to the variation of the hydride evolution and measurement stages. In routine use contributions to the variation may include variation between series of readings the variation inherent in the blank and also that from transient interferences from different sample digests. A better expression of the R.S.D. would be that obtained in actual routine use. This can conveniently be obtained by consideration of the calibration lines used in determining the results described in this paper. The series of standards were measured 20-28 times by three analysts for over 12 months using the defined procedure of duplicate injections. Expressing the average net response for each calibration line in terms of the lowest con-centration taking the over-all average as unity the range for these average responses together with the standard deviations are as shown in Table 111.This gives an indication of the likely range of response between series of measurements. The ratio of the average of two reagent blanks to the zero standard for each series of measurement is also displayed for each element. While the magnitude of this reagent blank for each element is normally about 2 ng it would appear that wet oxidation can reduce this response. Table I11 also illustrates the ratio of the sum of the responses at each standard concentra-tion compared with the sum of the average responses for each element together with the R.S.D.of the differences of each response with the average response for the individual calibration lines. The ratios obtained for antimony (V) are variable for the lower standard January 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY 23 and exceed the 95% confidence limits (95 C.L.) imposed by the R.S.D.s but it is assumed that response within the range measured is linear and this is supported by the decreasing magnitude of the R.S.D. as the measured concentration increases. It is concluded from similar considerations that the response for arsenic(V) is linear each ratio obeying the statistical limits imposed by the R.S.D.s while for tin(1V) linearity diverges markedly above 0.04mg.1-1 from consideration of both the ratio and the elevated R.S.D. at 0.05 mg 1-l.TABLE I11 CALIBRATION DATA Parameter Range of average response per unit concentration. . Relative standard deviation (R.S.D.) of average response . . Ratio of reagent blank with standard blank . . Ratio of standard to average response f R.S.D. : 1st standard . . 2nd standard . . . . 3rd standard . . . . 4th standard . . . . 5th standard . . * . 6th standard . . Antimony (V) 0.75-1.20 h0.13 0.83 1.14 f 0.40 0.90 f 0.21 0.94 f 0.10 1.08 f 0.12 1.00 f 0.05 1.01 f 0.05 Arsenic (V) 0.82-1.24 h0.12 0.88 0.89 5 0.41 0.95 f 0.15 1.05 f 0.11 0.99 f 0.07 1.02 & 0.06 1.00 f 0.08 Tin(1V) 0.7 7-1.26 If 0.16 0.92 1.02 & 0.10 1.03 f 0.07 1.02 & 0.04 0.99 & 0.05 0.98 f.0.06 (0.93 & 0.14)* * This refers to a 0.05 mg 1-1 standard not used in the calculation of the mean. For comparison with succeeding tables the coefficient of variation is 100 x R.S.D. To obtain the variation for single injections of pure solutions during application of the procedure to food digests R.S.D. values listed in Table I11 must be multiplied by d2 to give in the region of the second and sixth standards (fifth for tin) values which are effectively 0.29 0.07 for antimony(V) 0.21 0.10 for arsenic(V) and 0.10 0.07 for tin(IV) respectively, for 20-28 degrees of freedom. These can be compared with values for replicate injections for pure solutions alone at the same concentrations defined in Table I1 for 9 degrees of freedom. The 95 C.L. for the ratios of standard deviations can be obtained from tables2* and for 24 against 9 degrees of freedom will be 0.53-1.64.If it is accepted that at the higher concentration of tin(1V) the slightly exceeded upper limit is further evidence for incipient curvature of the calibration range there is agreement between the two sets of results within these limits except for the lower concentration of antimony(V) .* This agreement suggests equivalence of analysts absence of undue between-series variation and little interference from the food-stuffs examined in this evaluation that could conceivably indirectly increase the variation of measurement. Each value is an estimate governed by the asymmetric x2 distribution. Interferences With a system of measurement such as atomic-absorption spectroscopy with atomisation in a silica tube interference may be of four distinct types (a) directly upon the response of a measured element in the presence of an interfering ion (b) that upon the variation of measurement of an element in the presence of interfering ions (c) that upon the variation of subsequent measurements during an extended series of measurements in routine use and (d) that upon the response of subsequent measure-ments in an extended series of measurements.The last statement is taken somewhat out of context. * It must be accepted that when measuring low concentrations of an element similar in magnitude to the standard blank e.g. the first standard in Table 111 the standard deviation derived for these standards could be affected by the variation inherent in the blank giving a value up to 2/2 of that obtained for single injections within a series.Similar but diminishing relative contributions could occur as the measured concentration increases 24 EVANS et aZ. DETERMINATION OF ANTIMONY ARSENIC AND TIN Analyst VoZ. I04 Direct interference is usually considered only and necessarily so to ascertain occasions when the procedure is invalidated through systematic bias. Significant direct interferences have been documented for measurement by the argon - hydrogen entrained air flame25 and with graphite furnace atomisation.26 No interference has been reported for cationic species for a fully automated system using a heated silica tube provided that concentrated hydro-chloric acid is added to the sample solutions of arsenic(II1) and selenium(1V) prior to the sodium tetrahydr~borate(III).~~ The present method is different to all three and hence these effects have been considered for two concentrations of each element for the ionic species listed in Table IV the latter reflecting amounts that could be encountered in foodstuff digests.The differences are those obtained for the net response for triplicate injections, with and without interfering ion each contained within the same series of readings. The 95 C.L. are those calculated from Table I I (first row) for the same concentrations of antimony(V) arsenic(V) and tin(1V) investigated and are reduced by a factor of 0.58 to account for the triplicate injections.* When the direct effect of an ionic species was repeated, these limits should be reduced by a further factor of 0.71.Consideration of the interfering elements that exceed the 95 C.L. suggests that in the examination of foodstuffs by the described procedure only arsenic tin and copper could occur a t levels that directly interfere in the individual determination of the three elements of interest. I t would appear that each interference may be of a consistent level irrespective of the measured concentration. Copper may exceed 10mgkg-l in liver offal or tomato products while total arsenic in marine samples could frequently exceed 5 mg kg-1. The most common interfering species is however tin at levels exceeding 5 and 2 mg kg-1 for antimony(V) and arsenic(V) respectively. When canned produce is examined for these elements the tin must therefore be removed according to the procedure detailed in the method.I t is interesting to note that metaphosphate partly precipitates tin(IV) another example of systematic bias but in practice although metaphosphate must be obtained during the digestion procedure] dilution of the cooled acid digest with water ensures con-version into the orthophosphate. There remain possible indirect interferences of types (b) to (d). These can be assessed for the concentrations shown in Table I1 according to the following pattern separately for non-interfering species chosen at random and for species known to interfere directly. For the latter with antimony(V) and arsenic(V) 2 pg of copper(I1) and 1 pg of tin(1V) were used as direct interfering amounts and for tin(IV) 1 pg of copper(I1) and 0.2 pg of nickel(I1).Alternate triplicate injections of standard solutions (S) and these solutions plus interfering (non-interfering) species (Ill I,) were made in the sequence S I, S I, S I, S I, S each standard concentration constituting an individual series of injections. The R.S.D. of the net response for a single injection within triplicate injections for I, I may reflect indirect interference of type (b) for 8 degrees of freedom. The R.S.D. of the net response for a single injection within triplicate injections for the standard S excluding the first triplicate] may reflect indirect interference of type (c) for 8 degrees of freedom. The R.S.D. of the net response for a single injection for all injections for S excluding those contained in the last triplicate may reflect indirect interference of type (d) for 11 degrees of freedom.The R.S.D.s obtained are tabulated (Table 11) and can be compared with the R.S.D. of repli-cated single injections of standards (Table 11 first row) by consideration of the 95 C.L. of the ratios of standard deviations for 8 or 11 degrees of freedom against 9 degrees of freedom. These limits will be approximately 0.5-1.9 and none of the ratios exceed these limits. The evidence presented suggests therefore that the direct interference of single species is important only for the levels investigated and for these it is desirable in the application of the method to have a knowledge of the levels of the interfering species. Cumulative effects direct and indirect (b) of both statistically significant and non-significant single interferences can only be observed during the application of this tested method to foodstuffs.* A full statistical treatment has been reporteds8 for assessment of interferences upon both response and variation in the presence of interfering ions. This treatment takes account of both error of the first kind, a i.e. a hypothesis is rejected when in fact i t is true and of the second kind i.e. a hypothesis is accepted when i t is false in application of the t-test for differences in response and the F-test for differences in variation. Irrespective of the magnitude of the R.S.D. to obtain a clear indication of either type of interference equivalent to that of the more empirical study undertaken in this text would require 13 replicates.This is a number we feel unable to undertake because of the many possible interferences. These tests are applied one-sided Jawary 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY Results 25 The accuracy of the total method in the presence of foodstuffs was assessed by recovery experiments on selected total diet homogenates containing negligible or low amounts of the three elements evaluated and no level of species known to interfere according to a design TABLE IV DIRECT INTERFERENCE EFFECTS ON HYDRIDE EVOLUTION AND MEASUREMENT Antimony( V) Arsenic( V) Tin(1V) Parameter - 1 - 7 Concentration/mg 1-1 . . . . 0.004 0.020 0.004 0.020 0.01 0.04 95% confidence limits. . . . . . f0.17 j 0 . 0 8 f0.20 k0.08 k0.14 f0.05 ,Added ion Ag+ A l 3 + As5+ Bi3+ Ca2+ Cd2+ cu2+ co2+ Cra+ Fe3+ Hg2+ K+ Mg2+ Mn3+ Moa+ Na+ Ni2+ Pb2+ PO,-Po43-Sb5+ Se4+ Sn4+ Sr2+ Zn2+ Amount/ K* 0.1 100 1 0.5 0.2 0.1 0.2 5 000 5 2 1 0.5 0.2 0.1 5 1 5 100 20 5 000 6 000 100 20 5 5 000 1 0.5 0.2 0.1 0.05 5 2 000 1000 20 2 000 500 0.1 0.2 0.1 1 0.2 1 0.5 0.2 100 100 20 Relative difference in response with and without added element: 7 .A -0.20 -0.06 +O.Ol - 0.03 ---0.15 (- 0.05) (-0.08) (-0.15)t (- 0.08) ---- 0.29t -0.16 -0.03 -0.14 -- 0.14 - 0.08 -0.16 0.00 -0.03 -0.06 +0.10 0.00 --- - + 0.04 f0.04 -0.01 + 0.05 ----( - 0.12) __ ( - 0.05) (- 0.07) -0.06 -0.14t -( + 0.04) -0.03 - 0.06 + 0.07 + 0.02 +- 0.02 -0.06) - 0.08) -0.11) -0.05) ------0.17 -0.03 0.00 0.00 f0.03 f0.04 - 0.06 - 0.01 - 0.02 - 0.03 0.00 - 0.03 -----fO.01 -0.03 -0.06 --0.06 ---(0.00) -( - 0.14) (-0.07) +0.01 (- 0.04) --0.15 Over-all average non-significant differences .. . . . . . . -0.05 -0.02 - 0.08 -0.14 ----- 0.23 - 0.08 + 0.08 (- 0.11) t (- 0.07) ---- 0.51 t ( - 0.15) + 0.02 - 0.14 - 0.06 (-0.11) (- 0.03) -0.17 - 0.17 - 0.18 (+ 0.04) - 0.09 - 0.04t +0.11 -0.06 --- 0.08 +0.17 + 0.14 + 0.08 --- 0.04 + O . l l + 0.04 (-0.35)t (-0.10) ( - 0.03) 0.00 - 0.367 -(0.00) - 0.04 - 0.04 +o.o2 ---- + 0.01 - 0.02 - 0.07 (- 0.17) (- 0.03) ---- 0.24 (0.00) + 0.03 + 0.03 + 0.01 (- 0.05) (-0.02) - 0.02 - 0.09 - 0.03 (- 0.03) i o .0 1 - 0.14 0.00 - 0.06 -- + 0.01 - 0.05 - 0.Oi + 0.02 - 0.05 0.00 - 0.03 (- 0.23) ---(-0.11) (- 0.02) + 0.07 - 0.35 (- 0.06) - 0.02 -0.01 -0.04 +0.09 +0.06 - --0.30t -0.11 (-0.01) (-0.04) (-0.20)t (-0.11) 0.00 -0.06 S0.04 +0.02 +0.16 -0.03 (-0.29)t (-0.24) (-0.05)f (-0.10) (- 0.06) ( - 0.02) -0.52t +0.01 +0.05 -0.02 - -(-0.ll)f (-0.10) (-0.11) (-0.01) +0.01 +0.08 - 0.05) - 0.07 -+0.04 --0.03 + 0.10) - --0.01 0.00 ( + 0.02) +0.01 -+ 0.03 + 0.06 -(+0.01) ---0.19)t (-0.17) - 0.06) t (- 0.12) ,-0.06) (-0.04) -0.07 -0.04 - -(- 0.50) t (- 0.44) -0.46t -0.46 +0.05 -0.02 -0.06 +0.01 - -+O.ll +0.03 - --0.03 -0.06 - -- --0.03 +0.02 (+0.03) (-0.03) - -0.00 -0.01 * For 100 ml of digest prepared from 10 g of foodstuff the concentration of interfering species as t ~ 5 % significance.$ ( mg kg-l can be obtaimd from x 10 for antimony and arsenic and x 50 for tin. ) Implies a duplicated set of readings 26 EVANS et al. DETERMINATION OF ANTIMONY ARSENIC AND TIN AnaZyst VoZ. 104 previously de~cribed.1~ Amounts of all three elements were added in the inorganic form and in the lower valency state to 20 g of beverage infusion and milk 5 g of fat homogenate and 10 g of the remaining homogenates (cereals meat fish preserves root vegetables and green vegetables) and in amounts dictated by that noimally found in foodstuffs.Recovery was determined in duplicate each base level being determined simultaneously in duplicate and all values for a particular diet homogenate being contained within a series of determina-tions. For arsenic in fish a separate recovery exercise was made using dry combustion of 1 g (wet mass) of a fish homogenate sample for reasons already discussed. Separate recovery experiments were made for 40 and 100 pg of added tin (to meat preserves and green vegetable homogenates reflecting canning practice) o avoid interferences when measuring the other two elements under consideration. The results after subtraction of the mean base levels, are summarised in Tables V and VI.From these results standard deviations pertaining to the method in the presence of foodstuffs can also be calculated and where applicable confi-dence limits deduced (Table VI). The varying base levels in each food homogenate cannot be allowed for in these calculations and the standard deviations reflect variation only in the determination of added species with subsequent measurement as a different valency species. TABLE V AVERAGE RECOVERIES OF ADDED ANTIMONY ARSENIC AND TIN Each value is the average of six results a t three added levels. Recovery yo Food homogenate Cereal . . Meat . . . . Fish . . Fats . . . . Fruit . . . . Root vegetables . . Green vegetables . . Beverages . . . . Milk . . Average . . Antimony 0.021 pg* 108 89 103 103 102 102 103 106 103 102 Arsenic 0.095 pg* 99 88 92 99 102 92 97 84 94 -Tin 0.69 pg*t 95 97 104 93 95 99 100 98 91 97 I Average 101 91 104 96 99 101 98 100 93 98 * Average content in the homogenate.t The average content of homogenates used for recovery a t 40 and 100 pg was 5 pg of tin. For the method in application to foodstuffs a similar exercise on standard reference materials and retail foodstuffs was instituted without regard to possible interfering species. The effects of the latter on the results are discussed later. Duplicate total analyses were made by each of the analysts each duplicate result being obtained within the same series of measure-ments and these results are summarised in Tables VII and VIII.Values obtained by wet oxidation for arsenic in two fish samples are included for comparison as a matter of interest. For the sample of NBS tuna it has been reported that the first can examined gave inconsistent results for lead12 and similar inconsistency was observed for tin in the range 2-7 mg k g l . All results given in Tables VII and VIII for this material were obtained on a second can and consideration of the variation for tin results suggest this can to be normal. It has been mentioned that levels of these three elements in foodstuffs are normally low, but exceptions have been described that require dilution of digests to enable measurement to be made in the concentration range described. As there is the possibility that most or all of the variations inherent in the total method arise from the hydride evolution and measurement stages it is necessary to consider the standard deviations (and factors derived from them) in two ways.To maintain a constant practice the standard deviations in Tables VII and VIII do not take account of these dilutions and for comparison of concentra-tions measured necessary allowances must be made. The footnotes to Tables VII and VIII are therefore brought to the attention of readers TABLE IT1 Element Antimony Arsenic Tin . . Total No. of diet groups 9 9 9 8 8 8 119 1:s I f 7 6 6 9 37 37 RECOVERY OF ANTIMOWY(III) ARSENIC(II1) AND TIN ADDED TO TOTAL DIET Amount Mean Significance* Repeatability No.of No. of added/ recovery Range r-*-, results analysts p g Y O % Matrix Analyst -1 1s 3 0.2 106 70-152 NS NS 25 0.053 0.0053 18 3 0.4 103 56-154 NS NS 22 0.092 0.0092 18 3 1.0 96 56-109 NS NS 10 0.10 0.010 16 3 0.2 91 67-116 NS NS 18 0.032 0.0032 16 3 0.4 92 78-115 NS 1% 8.0 0.030 0.003 16 3 1.0 99 78-124 NS NS 8.9 0.088 0.0088 6 3 2.0 101 60-125 - NS 22 0.44 0.44 6 3 4.0 103 94-112 - NS 5.4 0.22 0.22 6 3 10 101 94-113 - NS 7.6 0.76 0.76 12 3 2.0 100 67-134 NS NS 18 0.36 0.036 12 3 4.0 95 76-113 NS NS 9.4 0.36 0.036 18 3 20 96 68-113 576 KS 7.2 1.4 0.14 6 2 40 96 91-104 NS NS 5.4 2.1 0.21 6 2 100 102 95-108 NS NS 4.3 4.4 0.44 * NS = not significant. t These intervals are for a single result the first value derived from repeatability and the second derived from $ For these samples the results represent recovery from 1 g of a fish homogenate using dry ashing for destruction Measurement was made on prepared solutions diluted by a factor of two.3 Measurement was made on prepared solutions diluted by a factor of five. on a 10-g sample mass. method the confidence intervals are calculated for a 1-g sample mass Element Antimony Tin . . TABLE VII REPLICATE ANALYSES FOR TOTAL ANTIMONY AND TIN ON SINGLE FOODSTUFFS AND Foodstuffs material* Dried milk . . Flour . . Apricot puree . . NBS tuna . . NBS liver . . Tuna . . . . Pig kidney . . Bowen’s kale . . Spinach . . NBS liver . , Flour . . Dried milk . . Pig kidney . . Bowen’s kale . . NBS tuna . . Apricot purees . . Spinach .. . . Tunas . . Mean Range of Repeatability Sample content/ content/ Amount/ (-Ap, mas& 2.5 10 10 5 5 10 10 10 2 5 10 2.5 10 10 2 5 10 10 mg kg-l 0.002 0.002 0.002 0.004 0.005 0.003 0.005 0.010 0.082 0.01 0.01 0.05 0.02 0.02 0.16 1.39 7.5 8.3 mg kg-i 0-0.007 0-0.005 0-0.005 0-0.01 1 0-0.012 0-0.011 0-0.016 0.002-0.03 1 0.047-0.12 0-0.03 0-0.05 0-0.21 0-0.04 0-0.04 0-0.30 1.15-1.56 6.7-8.4 7.4-9.1 Pg 0.005 0.017 0.017 0.020 0.027 0.027 0.052 0.10 0.16 0.05 0.10 0.13 0.15 0.18 0.33 7.0 75 83 % Pg - 0.008 - 0.017 - 0.015 - 0.026 - 0.021 - 0.042 - 0.054 - 0.11 33 0.053 0.06 - 0.20 - 0.17 - 0.16 - 0.12 67 0.22 6.1 0.43 4.8 3.6 5.8 4.8 -mg kg-l 0.003 0.002 0.002 0.005 0.004 0.004 0.005 0.011 0.027 0.012 0.020 0.067 0.016 0.012 0.11 0.085 0.36 0.48 * Each row represents six results obtained by three analysts.t These intervals are for a single result the first value derived from repeatability and the second derived from For repeatability and reproducibility concentrations are based on sample mass taken. Results are calculated on a dry-mass basis; recorded consensus means for Bowen’s kale are antimony 0.069 There was no analyst significance for any sample. on 10-g sample masses. Measurement was made on prepared solutions diluted by a further factor of five to that normally used TABLE VIII Method of destruction of organic matter Wet oxidation Dry ashing REPLICATE ANALYSES FOR TOTAL ARSENIC ON SINGLE FOODSTUFFS AND STANDARD Foodstuffs material* Dried milk Apricot puree Coffee Flour Bowen's kales NBS livers7 Spinach Tuna7 NBS tuna37 Pig kidney71 Sample mass/g 2.5 10 5 10 2 5 10 10 6 10 Mean content/ mg kg-l 0.007 0.002 0.01 1 0.011 0.091 0.041 0.024 0.47 2.55 1.50 Range of content/ mg kg-1 0-0.023 0-0.008 0-0.022 0.007-0.01 3 0.066-0.12 0.027-0.050 0.020-0.030 0.45-0.53 2.42-2.97 1.29-1.78 Amount/ Pg 0.017 0.023 0.055 0.11 0.18 0.20 0.24 4.7 12.7 15 Signifi-cance of analystst NS NS NS NS NS NS NS NS 5% 5 Yo Repeatability Pg mgkg-- 0.027 0.011 - 0.037 0.004 66 0.037 0.007 8.2 0.009 0.001 16 0.030 0.015 9.0 0.019 0.0037 15 0.035 0.0035 6.4 0.30 0.030 8.6 1.1 0.22 8.8 1.3 0.13 T---Bowen's kales 1 0.093 0.005-0.168 0.093 5% 24 0.022 0.022 Tuna 1 0.54 0.40-0.62 0.54 NS 12 0.063 0.063 NBS tuna 0.25 2.88 2.38-3.12 0.72 NS 10 0.072 0.29 Pig kidney 1 1.33 1.16-1.58 1.33 NS 6.7 0.089 0.089 Fish flourlr 0.20 11.8 10.6-12.4 2.35 NS 4.5 0.11 0.54 * Each row represents six results obtained by three analysts.t NS = not significant. These intervals are for a single result the first value derived from repeatability and the second derived For repeatability and reproducibility concentrations are based on the sample tj Results are calculated on a dry-mass basis; recorded consensus mean for Bowen's kale is 0.14 mg kg-1.2g 7 Measurement was made on a prepared solution diluted by a factor of five for tuna and ten for NBS tuna based on a 10-g sample mass.matter by dry ashing intervals are calculated for 1-g sample mass. the fish flour 30 EVANS et al. DETERMINATION OF ANTIMONY ARSENIC AND TIN Analyst VoZ. 104 Discussion Tin To avoid confusion information that can be derived from these results will be considered separately for each element and for convenience tin will be considered first. The accuracy of the total method in the presence of representative foodstuffs can be accepted by reference to Tables V and VI. To test whether the recovery from any total diet homogenate is abnormal each duplicate mean at each level must be tested for 95% confidence intervals (C.I.) of -!&s/l/2 where s is the reproducibility from the over-all mean at each particular level.A single low recovery of 68% for the fat homogenate at 20 pg of added tin is responsible and this is the source of the matrix significance listed in Table VI. (Discards are not allowed because of the low levels being monitored.) It can also be noted that the average recovery of 91% from milk at the three added levels is the lowest obtained. No reference foodstuff material exists with authenticated levels of tin and the accuracy of the method in application to foodstuffs must depend upon the consensus mean for Bowen’s kale and any certified future value of NBS tuna. (On the basis of the evidence in Table IV the copper content of NBS liver does not permit the determination of tin in that material by the method.) The variation of the results for the method and the method in application according to the design previously described,15 are each defined by two standard deviations at each estimated level of the element.The repeatability so is the standard deviation of analytical uncertainty obtained by known analysts on representative foodstuffs and is unencumbered by other sources of variation other than the homogeneity of the samples examined. The reproducibility s reflects that obtained by any analyst on any foodstuff to which the method is applicable. s will also contain the variation between series of determinations and the variation inherent in the blank and these sources of variation will be included within the matrix variance for recovery experiments and contribute to analyst variance for the indi-vidual foodstuffs examined.Each value for so and s is an estimate subject to variation; for 6 degrees of freedom the 95 C.L. for so will be 0.64-2.20 so. Within this context values listed for so expressed as the coefficient of variation in Tables VI and VII display a suitable gradation as the estimated amount increases. If allowance is made for the additional &fold dilution of digests to give a new sequence for the amount measured this gradation continues to apply and further, similar values are evident at similar measured levels whether the results originate from recovery experiments or from individual foodstuffs. Each individual result will be obtained from at least duplicate injections and the R.S.D.s of single injections of standards are defined in Table 11 first row.Consideration of the 95 C.L. of the ratio of the coefficients of variation obtained in the recovery and single foodstuff exercises with that for similar levels of dupli-cated injection of pure solutions at 0.01 and 0.04 mg 1-1 (equivalent to measurements of 5 and 20 pg of tin in 100 ml of original digest) of 8 and 3y0 respectively indicate only one ratio that falls outside these confidence limits namely for recovery a t 20 pg of tin and the reason has been isolated. Similar observations can be made for s and values compared with duplicate injections of pure solutions for between-series measurement in routine use (Table 111). In the region of 0.01 and 0.04 mg 1-1 the latter coefficients of variation are 7 and 5y0, respectively and the ratios at equivalent amounts measured do not exceed 95 C.L.with the exception of recovery for 20 pg of tin. The agreement of so from the recovery exercise with that for injection of standard solutions suggests that there is little contribution to variation from digestion of different foodstuffs. Similar agreement for the relevant individual foodstuffs indicates that tin in these samples is homogeneously distributed. The agreement for s with duplicate injections of standard solutions between series measurement suggests cumulative indirect interference type (b) is not a problem for the range of foodstuffs and foodstuff homogenates examined. The general agreement between recovery and foodstuffs exercises for so and s further signifies not sur-prisingly for tin the absence of variation from different chemical states in foodstuffs.Finally, the agreement for both exercises with injections of pure solutions suggests that the overriding variation in the determination of tin originates from the hydride evolution and measurement stages. Table VII indicates the difficulty of obtaining meaningful lower levels of tin in non-canned Only one out of 27 means exceeds this limit which is statistically acceptable January 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY 31 foodstuffs. Even though the standard deviation expressed as an amount is constant the skew distribution caused by zero results makes it impractical to compare derived limits of detection of results defined by so or s for the relevant degrees of freedom from the two exercises.Those derived from so and s for the recovery exercise at 2 and 4 pg of added tin will be 1.3 and 1.2 pg respectively which for the described procedure implies detection limits of 0.13 and 0.12 mg kg-l. The so and s values for recovery at 2 4 and 40 pg of tin (the last amount measured as Spg) and for NBS tuna each expressed as amounts are effectively constant and this is reflected in the 95 C.I. for each. This suggests that for subsequent exercises undertaken single results for total tin in the range 0.2-0.8 mg k g l should be reported to a 95 C.I. of 0.2 mg kg-l and above 0.8 mg kg-1 to a 95 C.I. of 25%. Antimony Recovery of antimony( 111) added to representative foodstuffs with hydride evolution from digests containing antimony as antimony(V) is illustrated in Table VI.The figures are justification for the decision to avoid a pre-reducing stage before hydride evolution and also exemplify the efficiency of the additional hydrogen peroxide stage in completely oxidising the antimony present to the higher valency state. The added antimony(II1) originated from a commercial primary standard solution specially prepared for atomic-absorption spectroscopy while the antimony(V) calibration standards were prepared from a reagent of purity not less than 97%. The over-all mean (from 54 results) reflects this. No significance from either matrices or analysts is apparent for this exercise but closer inspection of the duplicate means reveals that 95 C.L.are exceeded for addition of 1 pg to the meat homogenate. This is caused by a single low recovery of 56% the remaining recoveries at this level being in the range 88-109%. No standard reference material exists containing low authenticated levels of antimony. The summary of results in Table VII, with the exception of Bowen’s kale are included only for interest reflecting the low levels found in practice. It is impossible therefore to consider the variation of results in a manner similar to tin. The significance of the ratio of the R.S.D.s within and between the series of measurements on standard solutions has already been noted (p. 23) and this significance could arise from insufficient care with blanks from excessive instrument variation between series of measure-ment or from transient indirect interferences singly or in total.Consideration of the ratio of standard deviations indicates that there is clear agreement for s at each recovered level and for Bowen’s kale expressed as the coefficient of variation with between-series measure-ments on standard solutions (Table 111) and clear disagreement at measured concentrations of 0.004mg1-1 (equivalent to the recovery of 0.4pg of antimony) for so for within-series duplicate measurements on standard solutions (Table 11). This increased variation for so to a level similar to s could not be reflected by blanks or between-series measurements. As none of the foodstuff homogenates or single foodstuffs (except NBS liver) contain levels exceeding those in Table IV that directly interfere and it has been shown that indirect inter-ference on variation from single ionic species does not occur (p.24) a tenuous conclusion is drawn that at these low determined levels transient indirect cumulative interferences cause this increase in variation for so. Even if this conclusion is accepted it cannot be said, on the basis of the evidence that there is little contribution to variation from digestion of different foodstuffs even though the single sample of Bowen’s kale returns values for so and s that are in a suitable gradation for the amounts determined in recovery experiments. It can be said however that cumulative direct interference is not a problem (from the accuracy) and that indirectly the principle source of variation in the determination of antimony is the hydride evolution and measurement stages.Values for so and s expressed as the amount for Bowen’s kale and recoveries at 0.2 and 0.4 pg of added antimony(II1) are relatively constant. The derived limits of detection from the recovery exercise will be 0.23 and 0.21 pg for so and s respectively. The 95 C.I. for these three examples suggest that single results for total antimony can be reported as 0.02 and 0.04 mg kg-l and above the latter to 95 C.I. of 50%. This is not effective for the levels of this element likely to be present in foodstuffs. Arsenic The evidence presented in Table VI confirms that following both wet oxidation and dr 32 EVANS et al. DETERMINATION OF ANTIMONY ARSENIC AND TIN Analyst Vol. 104 ashing of a fish homogenate as a procedure for the destruction of organic matter arsenic added as arsenic(II1) will be obtained in the resulting digest in the higher valency state The exercise reflecting the accuracy of the total method in the presence of representative food-stuffs does display analyst significance for a recovery of 0.4 pg of arsenic(II1) and this ori-ginates from single high recoveries for the cereal and root vegetable groups.The recovery a t 1.0 pg of arsenic(II1) added while displaying no significance also has a single low recovery, which causes the 95 C.L. for that duplicate mean to be exceeded. In general recovery from the meat homogenate appears to be low and the same remark applies to milk. For recovery experiments using dry ashing a single non-significant low result of 60% occurs at the 2.O-pg added level.While the accuracy of the method can be considered to be acceptable for the method in application to foodstuffs again few standard reference materials of known arsenic content exist. NBS liver possesses a certified arsenic content but unfortunately with the present method (and for similar methods) copper would be expected to interfere each 1-ml injection of acidified digest containing 10 pg of copper (see Table IV). The bias obtained for this material is 0.041-0.055 which is equal to -0.014 mg kg-l with a 95 C.I. for this mean value of &0.010 mg k g l based on six results which merely reflects the interference. NBS tuna, which does not contain levels of interfering species does not have an authenticated content, but several values are recorded in the literature e.g.3.3 mg kg-l with a standard deviation of 0.2 mg kg-1 based on three results.1° The mean value of NBS tuna in this evaluation, with destruction of organic matter by dry ashing is 2.88 mg k g l with a 95 C.I. of A0.31 Testing of the analyst means for the individual foodstuffs in Table VIII indicates 4 instances out of 45 that fall outside the relevant 95 C.I. and these are the sources of the analyst signifi-cance tabulated. The comparison of levels found in the two fish samples using different procedures for destruction of organic matter is of interest. In each instance the dry-ashing technique gives significantly higher mean levels confirming the previous literature informa-tion. The agreement for so and s for these samples within the meaning of variation is marked however which suggests that for NBS tuna itself the greater part of the arsenic exists in a form readily and consistently destroyed by the wet-oxidation technique.The variation defined by so and s separately follows an effective gradation expressed as the coefficient of variation for the total amount determined with the exception of the recovery of 2 pg of arsenic(II1) by dry ashing. This applies both for the recovery exercise and for the individual foodstuffs and whether so or s originate from foodstuffs treated by wet oxidation or by dry ashing of the sample. This gradation is undisturbed if allowance is made for the various digest dilutions described in the footnotes to Tables VI and VIII. Comparison of so and s for these latter measured levels with that for duplicate injections of standard solutions within and between series of measurements (Tables I1 and 111) by examination of 95 C.L.of the ratios indicates no single instance where these ratios are exceeded. The same conclusions obtained for tin therefore apply for the variation from digestion of different foodstuffs homogeneity cumulative indirect interferences and the principle source of variation in the determination of arsenic. Levels for so and s expressed as amount are relatively constant for low levels of measured arsenic and sufficient individual foodstuffs were examined for which zero results are absent. A comparison of the derived limits of detection from so and s can therefore be made for the two exercises. From the recovery of 0.2 and 0.4 pg of arsenic(III) values of 0.11 and 0.12 pg are obtained from so and s respectively and from the average of the two lowest standard deviations for the individual foodstuff's displaying no analyst significance values of 0.19 and 0.16 pg of arsenic are obtained; this implies detection limits for the procedure of 0.011-0.019 mg kg-l.From the 95 C.I. tabulated it does appear that single results for total arsenic could be reported as 0.02 and 0.04 mg kg-l and for higher concentrations to 95 C.I. of 50%; as for antimony this defines the effectiveness of the method. mg kK1. Conclusions A complete analytical method has been described for the determination of total antimony, arsenic and tin in foodstuffs. The digests obtained from the wet-oxidation procedure can also be used for the determination of a number of other elements (copper iron manganese Jcknaary 1979 IN FOODSTUFFS BY ATOMIC-ABSORPTION SPECTROPHOTOMETRY 33 zinc lead cadmium and nickel).In the application of this digestion procedure the deter-mination of arsenic in samples of marine origin is an exception and an alternative dry-ashing technique on a limited sample mass must be used. Each of the three elements will be in the highest valency state in these digests and as such can be measured using atomic-absorption spectroscopy with atomisation in a flame-heated silica tube after evolution as the normal hydrides. The optimum conditions for measurement are described and the effect of direct interferences from ionic species upon the measurement are assessed defining the occasions when the application to foodstuffs is invalidated.For these elements the most important invalidation is for antimony and arsenic by tin in canned foodstuffs. For such occasions it is advisable to include an additional procedure for the removal of tin. The effect of indirect interferences from ionic species to which atomic-absorption spectroscopy with atomisation in a silica tube is susceptible has also been considered and evidence is presented to show that such indirect interference from single ionic species does not occur for the levels investigated. The accuracy of the total method for the low levels of each element normally present in foodstuffs has been confirmed and the accuracy of application of the method to foodstuffs has been considered as far as it is possible at present.The standard deviations repeatability (so) and reproducibility (s) for hydride evolution and measurement have been obtained for standard solutions for both within- and between-series measurements. Similar standard deviations have been calculated from results for the method via recovery experiments and the method in application via individual foodstuffs. Testing the ratios of the series of repeatabilities and reproducibilities separately and against each other indicates no significant differences except for antimony at low concentrations. The absence of significance suggests that there is little contribution from the digestion of different foodstuffs; tin and arsenic distribution in individual foodstuffs tested are homogeneous cumulative indirect inter-ferences do not occur and the principle source of variation for each element originates from the hydride evolution and measurement stages.While the accuracy of low levels of antimony determined by the method is satisfactory the elevated variation at these levels may be caused by transient indirect cumulative interferences of ionic species. The standard deviations are relatively constant as amount for a very narrow range for the lowest amounts determined for each element but from these a practical limit of detection has been deduced from the results of each element. For results obtained above these levels, consideration of 95 C.I. suggests the effectiveness of the results for the three elements that can be obtained by this method and indeed by any similar manual method employing hydride evolution and atomic-absorption spectroscopy with atomisation in a silica tube.This paper is published with the permission of the Government Chemist. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. References Tech. Rep. Ser. Wld Hlth Org. No. 532 Geneva 1973. “Arsenic in Food Regulations 1959,” SI 1959 No. 1831 HM Stationery Office London. Hanson N. W. Editor “Official Standardised and Recommended Methods of Analysis,” Second “Official Methods of Analysis of the Association of Official Analytical Chemists,” Twelfth Edition, Schmidt F. J. and Royer J. L. Analyt. Lett. 1973 6 17. Fernandez F. J. Atom. Absorption Newsl. 1973 12 93. Pollock E. N. and West S. J. Atom. Absorption Newsl. 1973 12 6. Kahn H. L. and Shallis J. E. Atom. Absorption Newsl. 1968 7 5. Chu R. C. Barron G. P. and Baumgarner P. A. W. Analyt. Chem. 1972 44 1476. Fiorino J . A. Jones J . W. and Capar S. G. Analyt. Chem. 1976 48 120. Thompson K. C. and Thomerson D. R. Analyst 1974 99 595. Evans W. H. Read J . I. and Lucas B. E. Analyst 1978 103 580. Analytical Methods Committee Analyst 1960 85 643. Leblanc P. J. and Jackson A. L. J . A s s . OH. Analyt. Chem. 1973 56 383. Evans W. H. Analyst 1978 103 452. Gorsuch T. T. Analyst 1959 84 135. Lunde G. J . Sci. Fd Agric. 1973 24 1021. Braman R. S. and Foreback C. G. Science N.Y. 1973 182 1247. Uthe J. F. Freeman H. C. Johnston J. R. and Michalik P. J . Ass. Ofl. Analyt. Chem. 1974, Edition Society for Analytical Chemistry London 1973. Association of Official Analytical Chemists Washington D.C. 1975. 57 1363 34 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. EVANS JACKSON AND DELLAR Smith A. E. Analyst 1973 98 209. Goulden P. D. and Brooksbank P. Analyt. Chem. 1974 46 1431. Aggett J. and Aspell A. C. Analyst 1976 101 341. Schmidt F. J. Royer J . L. and Muir S. M. Analyt. Lett. 1975 8 123. Davies 0. L. and Goldsmith P. L. “Statistical Methods in Research and Production,” Oliver Boyd, Smith A. E. Analyst 1975 100 300. Pierce F. D. and Brown H. R. Analyt. Chem. 1977 49 1417. Pierce F. D. and Brown H. R. Analyt. Chem 1976 48 693. Maurice M. J. and Buiys K. 2. Analyt. Chem. 1969 244 18. Bowen H. J. M. J . Radioanalyt. Chem. 1974 19 215. Edinburgh 1972. Received June 21st 1978 Accepted July 25th 197