Inductively coupled plasma mass spectrometry analysis of wines Mercedes Yolanda Pe�rez-Jorda�n, Jose Soldevila, Amparo Salvador, Agustý�n Pastor and Miguel de la Guardia* Department of Analytical Chemistry, Faculty of Chemistry, Doctor Moliner St. 50, Burjassot, 46100-Valencia, Spain Received 8th May 1998, Accepted 22nd October 1998 An inductively coupled plasma mass spectrometry (ICP-MS) procedure has been developed for the determination of major elements, such as Mg, Na, K, Ca and Fe, minor elements, such as Al, Cr, Mn, Cu, Zn, Se, Sr, Br and Rb, and trace elements, such as Li, Ti, Ni, As, I, Ba, Pb, Sc, V, Co, Y, Zr, Mo, Sn, Cs, Ga, Nb, Pd, Cd, Sb, Hf, W, Hg, Tl, Th and U, in wines.The results obtained for Na, Mg, Al,Mn, Fe, Co, Zn, Rb and Cs were compared with those found by neutron activation analysis (NAA). Two ICP-MS calibration methodologies were used and results evaluated from spike recovery studies from which an average recovery of 102±20% was found for quantitative mode measurements.Multi-determination, using Be, Ge, In and Bi for the calibration of the ICP-MS sensitivity in the whole mass range and Rh as the internal standard, provided fast and accurate results, whereas the quantitative mode, using a series of external standard solutions, needs more time and consumes more reagent. A great number of articles can be found in the analytical Ba, La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm, Yb, Lu, Hf, Ta, W, Re, Os, Ir, Pt, Au, Hg, Tl, Pb, Bi, Th, U and literature on the chemical characterization of wines, including Pb being monitored using In as internal standard, with limits investigations of their adulteration, studies of changes that of detection from 0.005 to 8 mg l-1 and recovery percentages take place during vinification and other researches focused on for all reported elements within 100±25%.The rest of the clearly establishing wine origin. papers were focused on the determination of Pb,15,18,19 with Mineral contents of wines depend on several factors, includdetection limits of 0.2 mg l-1.18 ing soil, area of production, type of grape, weather or environ- The accuracy of ICP-MS determinations has been evaluated mental conditions and viticultural practices.1–3 The by intercomparison using diVerent nebulization systems,2 or determination of the concentration of some elements is also from recovery studies,16,20 or by the analysis of reference of interest due to their toxicological or physiological materials of a diVerent nature than wines,15 which entails a characteristics.1 correct evaluation of the ICP-MS methodology through com- In the period from 1980 to 1997, 217 papers have been parison with a reference method also applied to the analysis published on trace element determination in wines.Atomic of natural samples of wine. spectrometric techniques were the most commonly used but The main objectives of this paper were the development of also other techniques such as molecular spectrometry, eleca methodology suitable for routine analysis of trace elements troanalytical methods or chromatography were employed for in wines by ICP-MS using a semi-quantitative procedure, this purpose.The main body of articles were focused on the based on an internal calibration, and the development of analysis of only a single element or just a few, because no another methodology using multi-calibration for quantitative multi-element techniques were used.However, modern analytanalysis, involving the ICP-MS data of Na, Mg, Al, Mn, Fe, ical methods, such as those based on inductively coupled Co, Zn, Rb and Cs validated by a comparison with those plasma atomic emission spectrometry (ICP-AES),4–14 ICPfound by NAA, and those obtained for additional elements MS2,15–20 or NAA,21,22 were also used to carry out multievaluated from spike recovery studies. element analysis of wines. Neutron activation analysis is a multi-elemental technique widely used in reference material certification, which has several advantages for direct sample Experimental measurement, although it suVers serious interferences caused Apparatus by major activation products.23 ICP-AES permits multi-element analysis but ICP-MS pro- An Elan 5000 ICP-mass spectrometer from Perkin-Elmer vides higher selectivity and sensitivity and a lower limit of SCIEX (Thornhill, Ontario, Canada), with a cross-flow nebudetection than ICP-AES, so that it could be preferable for the lizer from Perkin-Elmer (Norwalk, CT, USA) and a Minipuls multi-elemental analysis of wines.peristaltic pump from Gilson (Middleton, WI, USA), was In the literature concerning ICP-MS analysis of wines some used for ICP-MS measurements. The experimental conditions papers focused on the evaluation of rare element concen- employed in both the semi-quantitative and quantitative mode tration,2,16,17 but a few also studied other trace elements.17,20 are summarized in Table 1.Pb, Cd, Cr, V, Bi, Li, Ba, Rb, Mn, Fe, Cu, Ni, Sr, B, Cs, As The irradiation of samples was carried out in the swimming and Se were determined in wines, obtaining detection limits pool reactor Berlin II of the Hahn–Meitner Institut (Berlin, of the order of 0.3–3 ng l-1 and a recovery percentage of Germany), and detection was made by gamma-ray spec- 99–109% for the analysis of spiked samples. On the other trometry using an HPGe (Berlin, Germany) detector, the hand the determination of the authenticity of wine has been experimental conditions indicated in Table 1 being employed.carried out from its elemental composition,20 Li, Be, Al, Sc, Polyethylene vials (3 cm height and 1 cm id) and quartz Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Ga, Ge, As, Se, Br, Rb, ampoules (4 cm height and 0.5 cm id) were used to introduce samples into the nuclear reactor. An AE240 analytical balance Sr, Y, Zr, Nb, Mo, Ru, Rh, Pd, Ag, Cd, In, Sn, Sb, Te, I, Cs, J.Anal. At. Spectrom., 1998, 13, 33–39 33Table 1 Operating conditions for inductively coupled plasma mass spectrometry analysis and neutron activation analysis. Isotopes monitored by ICP-MS were 7Li, 23Na, 24Mg, 27Al, 39K, 44Ca, 55Mn, 56Fe, 58Ni, 59Co, 63Cu, 64Zn, 69Ga, 75As, 85Rb, 88Sr, 114Cd, 133Cs, 138Ba, 202Hg, 208Tl, 209Pb in the quantitative mode, all isotopes being measured in the mass range from 6 to 238 in the semi-quantitative mode.Isotopes monitored by NAA were 24Na, 27Mg, 28Al, 42K, 56Mn, 59Fe, 60Co, 65Zn, 86Rb and 134Cs ICP-MS operating conditions Parameter Plasma conditions RF power 1100 W Plasma gas flow 14–15 l min-1 Nebulizer gas flow 0.85–0.95 l min-1 Auxiliary gas flow 0.75–0.85 l min-1 Sample flow rate 1 ml min-1 Mass spectrometer settings Parameter Semi-quantitative mode Quantitative mode Vacuum pressure 9.26×10-6 Torr 9.26×10-6 Torr Dwell time 250 ms 75 ms Sweeps/reading 1 5 Readings/replicate 1 1 Number of replicates 2 3 Total time 103 s 135 s NAA conditions at Berlin II nuclear reactor Parameter Long-time irradiation Short-time irradiation Irradiation time 42 h 2 min Waiting time 18 d – Measuring time 2 h 4 min Thermal neutron flux 2×1014 n cm-2 s-1 1×1013 n cm-2 s-1 Epithermal neutron flux 6.1×1012 n cm-2 s-1 1×1011 n cm-2 s-1 Fast neutron flux 5.7×1013 n cm-2 s-1 6.2×1010 n cm-2 s-1 from Mettler (Zu� rich, Switzerland) was used for weighing Dorset, UK), and ethanol 95–96% (v/v), Prolabo R.P.Normapur (Fontenay Sous Bois, France), were added to samples inside the irradiation containers. Polystar 100GE (Rische-Herfurth, Berlin, Germany) welding scissors and a standard and blank solutions in order to provide a final concentration of 1% (v/v) and 6% (v/v), respectively. home-made N2O–O2 welding torch were used to seal the containers. Argon C-45, high purity 99.995%, provided by Carburos Meta�licos (Barcelona, Spain), was used as the plasmogen.A freeze-drier Lyovac GT2 from Leybold Heraeus and an electric heater from Heraeus Instruments (Hanau, Germany) High purity water with a conductivity higher than 18.3 MV cm-1 was obtained using a Milli-Q water system were used in the initial treatment of samples. (Millipore Inc., Paris, France). A 10 000 mg ml-1 Ti amg ml-1 V standard solu- Reagents and samples tion, both prepared in 5% (v/v) HNO3, and an Mg, Al, Ti, A multi-elemental 1000 mg ml-1 standard solution containing K, Na and Mn multi-elemental standard solution containing Li, B, Al, Cr, Mn, Fe, Ni, Co, Cu, Zn, Sr, Ag, Cd, Be, Tl and 28, 28, 2.4, 20, 7 and 0.4 g l-1, respectively, were used as Pb, dissolved in 13% (v/v) HNO3 (Alfa, Karlsruhe, Germany), standards for short irradiation experiments. A Cs, Co, Fe, Rb was used to prepare diluted solutions of these elements, and a and Zn multi-elemental standard solution containing 2, 20, multi-elemental 1000 mg ml-1 standard solution containing Sc, 2000, 250 and 200 mg l-1, respectively, was used as a standard As, Se, Rb, Hg and Cs prepared from Sc2O3, As2O3, H2SeO3, for long irradiation experiments using the operating conditions RbCl, HgO and CsCl from Merck (Darmstadt, Germany), indicated in Table 1.dissolved in HCl 0.3 M, HCl 0.6 M, a mixture of 0.18 M HCl No reference wine materials can be found for trace element and 0.08 M HNO3, water, HCl 5 M and water, respectively, analysis in wines.Therefore, a biological reference sample was also used to prepare diluted standards. Seronorm Second Generation from Nycomed Pharma Sero Standard solutions (1000 mg ml-1) of Na, K, Mg and Ca A/S (Oslo, Norway), with a similar level of concentration to were prepared from their corresponding salts NaCl, KCl, that found in wines for the elements to be studied, was used Mg(NO3)2 and CaCO3, all from Merck, dissolved in water, in NAA experiments as a reference.except CaCO3 which was dissolved in 1 M HCl, and diluted Three red and three white Spanish bottled wines from standards were then prepared from these. diVerent areas (Rioja, Utiel-Requena and Valencia) were Rh 10 mg ml-1 standard solution prepared from RhCl3, employed as real samples. Merck, dissolved in 5 M HNO3, was used as the internal standard. Standard solutions (1000 mg ml-1) of Ge and Bi were pre- General procedures pared from GeO2 (Fluka, Buchs, Switzerland) and Bi(NO3)3 (Merck) dissolved in 1 M HCl and 0.5 M HNO3, respectively.ICP-MS analysis. For the semi-quantitative mode 5 ml of sample were introduced into a flask, and then 100 ml of the A 996 mg ml-1 Be standard solution was prepared from Be(NO3)2·4H2O (Merck) dissolved in 0.5 M HNO3 and a 10 mg ml-1 standard solution of Rh, 100 ml of the 10 mg ml-1 multi-elemental standard solution of Be, Ge, In and Bi and 1000 mg ml-1 standard solution of In was dissolved in 5% nitric acid (Alfa).These solutions were used to prepare a 150 ml of 70% HNO3 were added, before a final dilution to 10 ml with distilled water was made. 10 mg ml-1 solution for calibration of the whole mass range for the measurements made in the semi-quantitative mode. Blanks were prepared under the same conditions as samples but including 600 ml of ethanol and without adding Be, Ge, Nitric acid, BDH AnalaR 69.0–70.5% (Merck Ltd., Poole, 34 J. Anal. At. Spectrom., 1998, 13, 33–39In and Bi because blank corrections were made after adjusted to be at the same intensity level while that of 208Pb was set as low as possible.measurement. The parameters employed for data acquisition in the semi- For the quantitative mode, samples were prepared by 50% quantitative mode were fixed to reduce the number of sweeps (v/v) dilution with distilled water and addition of Rh and per reading and readings per replicate and to improve the HNO3. Standards of all elements to be determined were dwell time, without increasing excessively the time of analy- prepared in the concentration range from 0.050 ng ml-1 to sis,25 whereas those used in the quantitative mode were fixed 3 mg ml-1.The same concentrations of Rh, nitric acid and to maintain a short measuring time but averaging several ethanol as employed in the semi-quantitative mode were used replicates in order to be able to obtain the standard deviation for this kind of measurement. of the ion counting process.The operating conditions selected Samples, standards and blanks were aspirated into the throughout this study are summarized in Table 1. ICP-MS system by a peristaltic pump under the experimental Previous studies on wine analysis by ICP-MS reported that conditions shown in Table 1. the best sample dilution level, in order to obtain a low organic The data acquisition system allows one to obtain the concenmatter content, corresponded to 10% (v/v).18 However, in the tration of samples from the ratio between the signals provided paper of Baxter et al.,20 a 151 dilution was recommended.To by each element considered and the Rh internal standard, be able to determine elements at very low concentrations, using, in the semi-quantitative mode, the signals of Be, Ge, In throughout this study a 50% (v/v) dilution of samples was and Bi to calibrate the sensitivity of the instrument over the used. In the aforementioned conditions, and taking into whole mass range considered.For the use of external caliaccount that the wine samples analyzed contain between 11 bration the appropriate interpolation was carried out for each and 12.5% (v/v) alcohol, standards were prepared with 6% element in the corresponding calibration line using at least 20 (v/v) ethanol. standard solutions, covering the concentration range from 0.05 Table 2 shows, as an example, the eVect of the presence of to 3000 ng ml-1. ethanol on the slope of calibration lines obtained for some of the elements considered.As can be seen, it is absolutely Neutron activation analysis. Trace element analysis was necessary to match standards with a concentration of ethanol carried out using long- and short-time irradiation conditions, to avoid systematic errors, which can occur when aqueous according to the neutron cross section of the parent nuclides solutions are employed as standards for direct analysis of low of the elements to be determined.Thus, Cs, Cr, Co, Fe, Sc, diluted wine samples. Rb, Se and Zn were determined by lengthy irradiation of In fact, it can be seen from the data in Table 2 that for samples, whereas Al, Mg, Mn, K, Na and V were determined elements with a mass number lower than that of Rh, the using a short irradiation time. For long irradiation analyses presence of a 6% (v/v) concentration of ethanol reduces the 100 ml of each wine sample were weighed into a quartz value of the slope of calibration lines from 60%, in the case ampoule, which was heated at 50 °C for 24 h inside an oven, of Li, Co and Zn, to 90% in the case of Cd.On the contrary, and this procedure was repeated until a final mass of sample for elements with a mass number higher than Rh, like Pb, a of the order of 20 mg was obtained for approximately 1000 ml sensitivity increase of the order of 130% was found, this value of wine. Ampoules were then sealed with an N2O–oxygen being increased until reaching 170% for Bi.This eVect could flame. Two empty quartz ampoules, treated in the same way be overcome by using multiple internal standards, but it creates as those containing samples, were used as blanks. Standards the need for a series of elements absent in natural samples, were prepared in ampoules by the introduction of 2.5 mg of a thus increasing the complexity of these analyses. Because of stock standard solution of Cs, Cr, Co, Fe, Rb, Sc, Se and Zn, that it is preferable to match samples and standards with an using the same procedure as described above, or from a freeze- appropriate level of ethanol.dried serum. Two blanks, two reference materials and four The eVect of varying the ethanol concentration, from 4 to standards, prepared from inorganic salts, were introduced with 8% (v/v), on the relative signal of a 100 ng ml-1 Bi standard 16 samples into an aluminium container in two steps, with 12 produced variations of ±5%, thus validating the methodology quartz ampoules in each, and were then irradiated for 42 h employed for the range of alcohol contents of the wines under the conditions indicated in Table 1.After irradiation, analyzed. elements were determined by c-spectrometry, after a cooling time established as a function of the half-lives of the activation Comparative study of results obtained by semi-quantitative and products generated. quantitative mode in ICP-MS For short-time irradiation, samples, previously freeze-dried Semi-quantitative calibration is a very rapid methodology for until a viscous solution was obtained, were introduced into the determination of a large number of elements in the same polyethylene vials and then dried at 50 °C.A final mass of the order of 50 mg was employed. Two standards, one containing Table 2 Calibration lines obtained by ICP-MS in the quantitative a mixture of K, Al, Mg, Mn and Na, and another with V, mode for the diVerent elements considered with and without 6% were prepared in polyethylene vials from aqueous solutions of (v/v) ethanol inorganic salts.Four vials, two with samples and two with Calibration lines y=a+bx (r2) standards, were packed in a plastic bag and introduced in a polyethylene capsule, which was irradiated under the con- Elements Without ethanol With 6% (v/v) ethanol ditions indicated in Table 1, and then the elements were determined by c-spectrometry. Li y=-0.0388+0.0058x y=-0.0615+0.0038 x Additional details about the irradiation conditions can be r2=0.9983 r2=0.999 Co y=0.1141+0.0068x y=-0.0027+0.0042 x found in previously published documents.24 r2=0.9976 r2=0.9996 Zn y=0.0138+0.0015x y=-0.0007+0.0009 x Results and discussion r2=0.9988 r2=0.9999 Cd y=0.014+0.0019x y=0.0103+0.0018 x Experimental conditions employed in ICP-MS determination r2=0.9993 r2=0.9997 Pb y=0.0336+0.0028x y=0.0253+0.0038 x ICP-MS experimental conditions were investigated in order to r2=0.9983 r2=0.9992 obtain the maximum sensitivity and as short a measuring time Bi y=0.0723+0.0037x y=0.053+0.0062 x as possible.The 103Rh ion signal was adjusted to be as high r2=0.9967 r2=0.9984 as possible and, at the same time, 24Mg and 207Pb were J. Anal. At. Spectrom., 1998, 13, 33–39 35sample with an appropriate precision and accuracy for routine for pattern recognition analysis based on ICP-MS data, as has been explored by Baxter et al.20 analysis. To achieve this, Be, Ge, In and Bi, which are absent in the samples to be analyzed, were used in order to calibrate Concerning the repeatability of trace metal contents, it can be seen that an average coeYcient of variation of 6% was the signals in the whole mass range considered.To evaluate the applicability of this strategy in the analysis of natural wine obtained for all elements in all of the samples with just a few exceptions, these being elements with an extremely low concen- samples, these were analyzed also by using an external calibration and employing Li, Na, Mg, Al, K, Ca, Mn, Fe, Co, tration, such as Pd, Hg, W and Th, for which variation coeYcients of the order of 20% or more were found in some Ni, Cu, Zn, Ga, As, Rb, Sr, Cd, Cs, Ba, Hg, Tl and Pb as test elements.samples, and other elements, such as Sc, Ti, Cr or Se, for which molecular interferences have been reported.26 The results obtained for trace element determination in natural wine samples analyzed by both semi-quantitative and The results found for a selected number of 22 common elements, for which standards are easily available, also show quantitative ICP-MS modes are indicated in Tables 3 and 4 respectively.The standard deviation for 3 replicates is also (see Table 4) that major, minor and trace elements could be determined with an average coeYcient of variation of 5% by indicated as an estimation of the precision of these results. As can be seen in Table 3, 40 elements, at major, minor and ICP-MS using long concentration range calibration lines obtained from 0.05 to 3000 ng ml-1.trace levels, could be determined in a sample in less than 2 min, having observed that Na, Mg, K, Ca and Fe are, in A comparative study between the results obtained by both semi-quantitative and quantitative ICP-MS modes showed general, at concentrations higher than 1000 mg ml-1, Al, Cr, Mn, Cu, Zn, Se, Br, Rb and Sr between 100 and 1000 ng ml-1, that for 18 of the 22 elements determined by both methodologies, comparable results were found.For As, a systematic and the other elements at levels of a few ng ml-1, or parts of ng ml-1. average diVerence of +200% was found between the results obtained by the semi-quantitative and quantitative modes, In general, the content of trace metals varies with the nature and the origin of samples, thus opening up exciting possibilities while for Ga and Hg systematic errors of the order of -70% Table 3 Results obtained by ICP-MS analysis of wines using the semi-quantitative mode, expressed in ng ml-1 (*expressed in mg ml-1). Concentration values correspond to the average of 3 replicates±the corresponding standard deviation Samples (ethanol percentage content) 1 2 3 4 5 6 Red/Valencia Red/Utiel-Req.Red/Rioja White/Valencia White/Utiel-Req. White/Rioja Element LODb (11.5%) (12%) (12.5%) (11.5%) (12.5%) (11%) Li 3 13±1 13.73±0.01 28.8±0.2 13.0±0.5 10.1±0.8 29±1 Na* 3 18±0.3 13±1 17± 0.1 15±1 15± 1 12.9±0.8 Mg* 0.3 44±3 92± 4 89± 2 69± 3 21.3±0.4 66±1 Al 0.2 960±30 537±30 968±10 810±10 304±7 620±10 K* 2 460±40 1140±50 487±6 400±100 221±3 420±10 Ca* 0.6 109±6 62± 1 85± 3 103.2±0.9 68±1 92± 3 Sc 0.02 2.2±0.3 2.23±0.08 2.0±0.3 2.12±0.03 1.4±0.3 1.3±0.4 Ti 0.4 30±20 27±30 13±7 11± 2 9±6 7.1±0.2 V 0.3 7.3±0.3 12.4±0.4 55.1±0.9 12.9±0.2 13.6±0.5 36.5±0.7 Cr 0.04 300±100 290±20 340±20 290±40 220±10 220±20 Mn 0.1 186±5 540±20 700±40 280±10 274±9 662±5 Fe 18 2010±40 3490±40 4000±200 1720±60 900±20 2990±40 Co 0.03 2.04±0.07 2.4±0.1 3.3±0.15 1.65±0.03 2.70±0.04 1.90±0.08 Ni 2 13±1 13.3±0.2 18.9±0.5 9.1±0.3 11.7±0.9 13±1 Cu 0.05 130±1 170±10 79±4 183±5 490±10 28.4±0.6 Zn 0.7 390±10 350±20 470±10 194±9 627±10 231±6 Ga 0.008 0.76±0.03 1.4±0.1 1.5±0.2 1.2±0.1 2.60±0.06 0.69±0.06 As 0.6 24.4±0.3 12.2±0.4 20.3±0.6 12.2±0.3 15.6±0.3 12.4±0.7 Se 4 130±7 36± 9 40± 20 45±7 43± 5 20± 10 Br 40 400±50 300±10 280±10 92±4 115±9 150±10 Rb 0.06 185±5 571±3 653±6 189±4 299±1 475±3 Sr 0.1 1000±10 860±20 1320±40 720±20 780±20 930±20 Y 0.005 1.07±0.04 0.37±0.03 2.00±0.07 0.57±0.04 0.37±0.06 0.75±0.05 Zr 0.03 2.2±0.1 3.7±0.3 4.2±0.3 5.20±0.08 2.7±0.1 3.4±0.1 Nb 0.001 0.13±0.01 0.29±0.04 0.20±0.02 0.08±0.02 0.30±0.02 0.39±0.01 Mo 0.04 1.6±0.1 5.6±0.3 1.68±0.05 2.3±0.1 8.2±0.3 1.6±0.4 Pd 0.02 NDa 0.03±0.01 ND ND ND ND Cd 0.06 ND ND 0.53±0.06 ND ND 0.35±0.06 Sn 0.08 1.7±0.2 3±1 0.6±0.1 6.6±0.3 15.5±0.2 0.99±0.09 Sb 0.002 0.9±0.09 0.8±0.1 0.60±0.02 1.31±0.04 1.9±0.2 1.27±0.03 I 0.2 22±2 10.0±0.3 8.0±0.6 13±1 8.8±0.4 6.9±0.4 Cs 0.01 0.21±0.002 1.31±0.02 3.51±0.04 0.04±0.004 0.81±0.02 3.45±0.05 Ba 0.06 66±2 87± 2 87± 1 40.9±0.7 47±1 39.4±0.5 Hf 0.005 0.04±0.02 0.15±0.01 0.16±0.02 0.13±0.01 0.16±0.02 0.11±0.05 W 0.01 0.25±0.06 0.09±0.03 0.55±0.05 0.25±0.01 0.11±0.01 0.6±0.4 Hg 0.005 0.29±0.04 0.20±0.04 0.19±0.01 0.17±0.01 0.31±0.04 0.31±0.08 Tl 0.001 ND 0.08±0.02 0.15±0.02 ND ND 0.08±0.02 Pb 0.02 43.8±0.3 52±2 39± 2 40± 2 61.5±0.8 26.2±0.8 Th 0.007 0.1±0.02 ND 0.06±0.02 0.03±0.01 ND 0.03±0.01 U 0.005 0.83±0.06 0.19±0.03 0.6±0.1 1.21±0.05 1.12±0.07 0.62±0.02 aND, not detectable values.bLOD, limit of detection values in ng ml-1. 36 J. Anal. At. Spectrom., 1998, 13, 33–39Table 4 Results obtained by ICP-MS analysis of wines using the quantitative mode, expressed in ng ml-1 (*expressed in mg ml-1). Concentration values correspond to the average of 3 replicates±the corresponding standard deviation Samples (ethanol percentage content) 1 2 3 4 5 6 Red/Valencia Red/Utiel-Req.Red/Rioja White/Valencia White/Utiel-Req. White/Rioja Element LODb (11.5%) (12%) (12.5%) (11.5%) (12.5%) (11%) Li 0.2 18.9±0.5 19.8±0.9 44±1 20.0±0.2 14.9±0.2 39.5±0.5 Na* 10 18±4 17.6±0.9 26±1 18.5±0.5 13.4±0.9 19.8±0.6 Mg* 100 77±2 94± 4 98± 2 69± 2 67± 2 73.7±0.7 Al 0.6 1360±30 800±30 1210±20 1000±20 240±10 840±10 K* 90 770±92 460±40 700±200 620±43 470±10 300±200 Ca* 30 81±3 57± 2 73± 2 99± 2 68.4±0.6 97±1 Mn 0.05 200±20 690±80 940±20 360±20 251±7 800±200 Fe 6 2500±100 3600±400 4680±80 1460±20 1001±50 3000±40 Co 0.004 1.99±0.001 2.5±0.1 4.0±0.1 1.56±0.03 2.79±0.04 2.14±0.08 Ni 0.003 9.6±0.2 13.9±0.6 17.8±0.3 11.0±0.2 13.66±0.09 16.2±0.4 Cu 0.003 112±2 190±2 76.0±0.9 229±4 609.2±0.2 36.6±0.5 Zn 0.6 320±2 309±9 385±5 199±4 575±1 330±80 Ga 0.02 NDa 5.4±0.2 5.4±0.2 3.05±0.07 3.13±0.03 2.79±0.05 As 0.5 ND 3.87±0.06 4.8±0.2 4.9±0.4 5.00±0.08 3.9±0.1 Rb 0.06 186±5 600±10 690±20 190±2 322±12 685±8 Sr 0.005 840±20 890±20 1280±30 810±9 889±6 1120±10 Cd 0.03 0.21±0.01 0.30±0.02 0.64±0.04 0.20±0.03 0.38±0.03 0.43±0.02 Cs 0.007 0.23±0.05 1.43±0.03 3.5±0.1 0.054±0.004 0.911±0.03 3.82±0.02 Ba 0.003 77±2 100±3 106±3 53± 1 59.8±0.3 51.2±0.6 Hg 0.5 13±9 1.7±0.4 0.7±0.1 4±1 1.0±0.1 0.58±0.08 Tl 0.0005 0.077±0.003 0.160±0.003 0.26±0.01 0.052±0.004 0.12±0.01 0.18±0.01 Pb 0.02 47±1 60± 1 48± 1 50.7±0.6 76.7±0.8 32.2±0.3 aND, not detectable values. bLOD, limit of detection values in ng ml-1.and -80%, respectively, were found. On the other hand, mode (x) after removing As, Ga, Hg and K for which noncomparable results were found. These points provide a results for K were diVerent without any systematic bias. These diVerences are probably due to the lack of an appropriate regression line, y=-0.00907+0.9959 x, with a regression coeYcient r=0.982 by weighted least squares fitting. Two calibration in the semi-quantitative mode, but additionally to problems found in the quantitative mode when extremely high statistical tests were applied to compare the aforementioned regression line with an ideal line y=0+1x.In order to ensure concentration ranges were employed for standardization, because the use of highly concentrated standards modifies the the comparability of the two data populations, values obtained for the intercept (a) and the slope (b) were compared with 0 confidence interval of the calibration line by changing the middle point of the line, thus aVecting the repeatability of and 1, respectively, using two tests: |b-1| must be lower than the product of tsb, t being the statistical parameter of Student signals and increasing the errors in the quantitative determination of trace elements.for a probability level of 95% (1.96) and sb the standard deviation (0.0024) of the slope of the regression line; on the Fig. 1 shows the plot of the mean values (in ng ml-1) obtained by use of the semi-quantitative mode (y), for each other hand |a-0| must be lower than the product of tsa, where sa is the standard deviation (0.0058) of the intercept of the element and sample, against those found by the quantitative regression line.The application of these tests shows that the obtained line is statistically comparable with the ideal one, thus demonstrating that both ICP-MS measurement modes give similar results for the elements and samples considered, therefore providing, with the semi-quantitative ICP-MS measurements, a good idea of the concentration level of as many elements as possible in wine samples in a short analysis time. Validation of ICP-MS procedure by comparison with results obtained by NAA Average values and the corresponding standard deviations of 3 replicate analyses obtained by NAA for Na, Mg, Al, K, Mn, Fe, Co, Zn, Rb and Cs are shown in Table 5.The only elements determined were those with a concentration level which could be determined by NAA using the procedure indicated in the experimental part.The results obtained by quantitative mode ICP-MS (y) and those obtained by NAA (x) have been compared for Na, Mg, Al, Mn, Fe, Co, Zn, Rb and Cs in the natural wine samples analyzed. K was removed for this comparison because of a severe lack of repeatability. Fig. 2 shows the regression between these two data populations. A correlation coeYcient of r= Fig. 1 Comparison between quantitative and semi-quantitative data obtained by ICP-MS in the analysis of 18 elements in six wine samples. 0.983 was obtained for a regression line, established by weighed J. Anal. At. Spectrom., 1998, 13, 33–39 37Table 5 Results obtained by neutron activation analysis of wine samples, expressed in ng ml-1 (*expressed in mg ml-1). Concentration values correspond to the average of 3 replicates±the corresponding standard deviation Samples (ethanol percentage content) 1 2 3 4 5 6 Red/Valencia Red/Utiel-Req.Red/Rioja White/Valencia White/Utiel-Req. White/Rioja Element LODa (11.5%) (12%) (12.5%) (11.5%) (12.5%) (11%) Na* 800 11.3±0.4 15.2±0.7 24±2 17.2±0.7 8.7±0.3 16.2±0.4 Mg* 50 40±2 81± 2 92± 7 61± 2 28± 5 59± 1 Al 6 681±0 730±60 1312±7 1010±80 230±40 810±20 K* 700 440±50 600±50 500±50 400±100 200±60 400±100 Mn 0.2 172±0.007 650±20 880±70 360±10 231±5 670±40 Fe 1 2400±100 3320±70 4600±600 1810±20 890±50 3100±100 Co 0.001 2.01±0.06 2.80±0.07 3.8±0.4 1.77±0.02 2.8±0.2 2.00±0.09 Zn 0.03 335±12 316±5 390±60 189±3 530±40 300±10 Rb 0.2 210±10 660±20 730±20 226±7 360±20 610±40 Cs 0.003 0.24±0.04 1.58±0.06 4.0±0.5 0.05 0.98±0.05 4.4±0.2 aLOD, limit of detection values in mg ml-1.Fig. 3 Relative accuracy errors obtained by comparison between quantitative ICP-MS and NAA data found for diVerent elements and samples. Fig. 2 Comparison between neutron activation analysis and quantitative ICP-MS determination of 9 elements in six wine samples.Table 6 Recovery percentages obtained for the analysis of spiked wine samples by ICP-MS in the quantitative mode. Recovery percentages minimum least squares fitting, of y=0.0014+1.008 x with sa= were established from data found on diVerent wine samples spiked with known amounts of elements from 50 to 200 ng ml-1 0.0039, sb=0.006 and t=1.96. It can be concluded that the results obtained by both techniques are similar and thus the Red wine Sweet wine White wine same accuracy level can be obtained for the elements evaluated, recovery recovery recovery providing a validation of the measurement process by ICP-MS Element percentage±s percentage±s percentage±s for these wine elements.In order to evaluate the comparability of the results found Li 143±4 119±7 130±6 Sc 108±5 99± 2 107±3 for each particular element and sample, relative accuracy error Mn 150±30 103±9 87± 8 percentages were calculated from the diVerence between Ni 118±3 118±2 111±4 ICP-MS and NAA.Fig. 3 shows the experimental results Co 120±2 120±2 116±5 obtained and it can be seen that for all the elements, except Cu 120±10 115±2 107±4 for Na and Mg in samples 1 and 5 and for Al in sample 1, Zn 130±10 134±4 108±5 for which the errors found were higher than 30%, the average Ga 106±2 111±1 105±4 As — 95±3 115±4 diVerences are in general lower than ±10%, with only a few 77Se 106±7 83± 6 — data with relative diVerences between ±10 and ±20%. 82Se 108±6 82± 4 — For the evaluation of the accuracy of the data obtained for Rb 110±10 60±10 110±7 those elements not determined by NAA, recovery studies on Sr 80±10 88±6 70± 10 spiked wine samples were carried out. Cd 97±3 96± 2 96± 3 Natural samples of red, white and sweet wines were spiked Cs 100±2 107±2 92± 3 Ba 96±5 107±1 94± 3 with known concentrations of Li, Sr, Mn, Ni, Co, Cu, Zn, Hg 75±2 78± 3 63± 3 Ga, As, Se, Rb, Sr, Cd, Cs, Ba, Hg, Tl, Pb and Bi of Tl 94±6 97± 1 88± 3 50–200 ng ml-1, and the samples analyzed by ICP-MS using Pb 89±7 94± 1 83± 3 the quantitative mode.The average results found are summa- Bi 86±8 103±3 85± 4 rized in Table 6, from which it can be concluded that all values 38 J. Anal. At. Spectrom., 1998, 13, 33–393 I. E. Frank and B. R. Kowalski, Anal. Chim. Acta, 1984, 162, 241. were within 102±20% in the case of quantitative 4 G. Thiel and K. Danzer, Fresenius J. Anal. Chem., 1997, 357, 553. measurements. 5 M.Bodyne-Szalai and P. Fodor, Magy. Kem. Foly., 1996, 102, 411. 6 M.Lo� pez-Artiguez, A. M. Camean and M. Repetto, J. AOAC. Conclusions Int., 1996, 79, 1191. 7 J. Yang, X. Zeug and B. Huang, Fenxi Huaxue, 1991, 19, 362. The procedure proposed for the ICP-MS analysis of wines is 8 H. Eschnauer, L. Jakob, H. Meierer and R. Neeb, Mikrochim. based on a direct 151 dilution with distilled water and does Acta, 1989, III, 291. not require any additional pre-treatment of samples.However, 9 S. Zhu, Z. Chen and L. Huang, Fenxi Huaxue, 1988, 16, 556. standards and blank solutions must be prepared in order to 10 L. Gu, Z. Zhou, R. Shen and H. Shi, Fenxi Huaxue, 1987, 15, compensate for the ethanol concentration of diluted samples 1140. to avoid systematic errors. 11 A. Voulgaropoulos and T. Soulis, Connaiss. Vigne Vin, 1987, 21, 23. It has been shown that semi-quantitative and quantitative 12 J. Zihlmann, Labor Betr., 1987, February, 15. ICP-MS oVer valuable alternatives for the multi-elemental 13 F.S. Interesse, G. D3Avella, V. Alloggio and F. Lamparelli, Z. determination of major, minor and trace components in wines. Lebensm.-Unters.-Forsch., 1985, 181, 470. Both methodologies can provide useful values, the semi- 14 F. S. Interesse, F. Lamparelli and V. Allogio, Z. Lebensm.- quantitative mode being a fast way of obtaining information Unters.-Forsch., 1984, 178, 272. about as many elements as possible in the same sample, thus 15 L.Moens, H. Vanhoe, F. Vanhaeke, J. Goossens, M. Campbell R. Dams, J. Anal. At. Spectrom., 1994, 9, 187. providing a good methodology for quality control, and the 16 A. Stroh, P. H. Brueneckner and U. Voellkopf, At. Spectrosc., quantitative mode requiring a careful standardization with 1994, 15, 100. appropriate solutions of the elements to be determined at 17 E. McCurdy, D. Potter and M. Medina, Lab. News, 1992, concentration levels comparable to those present in the September, 10. samples. 18 J. Goossens, T. De Smaell, L. Moens and R. Dams, Fresenius On the other hand, the comparability of data found for Na, J. Anal. Chem., 1993, 347, 119. 19 J. Goossens, L. Moens and R. Dams, Anal. Chim. Acta, 1994, Mg, Al, K, Mn, Fe, Co, Zn, Rb and Cs by ICP-MS and by 293, 171. NAA, employed as a reference procedure, indicates the accu- 20 M. J. Baxter, H. M. Crews, M. J. Dennis, I. Goodall and D. racy of the methodology employed, which has also been Anderson, Food Chem., 1997, 60, 443. validated for the other elements through spike recovery studies, 21 A. R. Byrne, M. Dermelj, L. Kosta and M. Tusek Znidaric, obtaining an average recovery of 102±20%. Mikrochim. Acta, 1984, I, 119. 22 S. May, H. Leroy, D. Piccot and G. Pinta, J. Radioanal. Chem., 1982, 72, 305. Acknowledgements 23 P. Bra�tter, Radiochim. Acta, 1983, 34, 85. 24 D. Gawlik and T. Robertson, Irradiation Devices at the Upgraded Authors acknowledge the financial support of the EU Program Research Reactor BER II, Hahn-Meitner Institutut, Berlin, ‘Human Capital and Mobility’ for the access to NAA in Berlin Germany, 1992. and are grateful to Dr. Virginia Negretti, Dr. Peter Bratter 25 J. Soldevila, M. El-Himri, A. Pastor and M. de la Guardia, and Dr. Dieter Gawlik for their help in the NAA. ICP-MS J. Anal. At. Spectrom., 1998, 13, 803. 26 S. E. Long and T. D. Martin, ICP Inf. Newsl., 1991, 16, 460. determinations have been supported by the project of the Generalitat Valenciana 2232/94. Paper 8/03476A References 1 C. Baluja-Santos and A. Gonza�lez-Portal, Talanta, 1992, 39, 329. 2 S. Augagneur, B. Me�dina, J. Szpunar and R. £obin� ski, J. Anal. At. Spectrom., 1996, 11, 713. J. Anal. At. Spectrom., 1998, 13, 33&n