The chelating resin disk (CD) used was the 3M Company Empore^TM Chelating extraction disk for environmental analysis with the diameter of 47 mm and with the iminodiacetic acid component as the functional group. It was immersed in 3 mol·dm⁻³ HNO₃ solution overnight prior to the application to separation of matrix components and pre-concentration of Ln’s. Water with the specific resistance of 18.2 MΩ・cm or higher at 25˚C (Milli-Q water) used throughout was produced with a Millipore water purification system. Stock standard solutions of Ln’s and indium (In) used as the internal standard substance and multielement standard solutions were purchased from Wako Pure Chemical Industries, Ltd. Working standard solutions were prepared by appropriate dilution and mixing of the stock standard solutions and addition of nitric acid (HNO₃). All other chemicals used were of analytical grade or better. The seawater sample used was kindly donated by Nihonkaisui Co. Ltd. It was collected off the coast of Onahama, Fukushima, Japan in 2009, and used after sand filtration.

The ICP-MS instrument used was a Seiko Instrument Model SPQ9000 ICP mass spectrometer. The operating conditions are summarized in Table 1. The value of pH of solutions was measured with a Horiba twin pH B-212 pH meter. The glass filtration apparatus used was a product of Toyo Roshi Kaisha Ltd.

In advance of treating actual seawater, we first checked the performance of the CD using a sample solution containing lanthanum (La), gadolinium (Gd) and lutetium (Lu). La, Gd and Lu were intended to be the representatives of the light, middle and heavy Ln’s, respectively. Since the Ln’s are mainly dissolved in seawater as carbonate complexes like except Ce which forms Ce(OH)₄ in seawater [4] , the sample solu-

^a: ¹⁴²Ce interference was corrected with the ratio of ¹⁴²Ce/¹⁴⁰Ce.

tion for the preliminary experiments was prepared from 1 dm³ of 0.1 mol·dm^3 sodium hydrogen carbonate (NaHCO₃) solution according to the pre-treatment procedures for separation of matrix components and pre-concentration of Ln’s in aqueous solutions with the CD [17] . The preparation procedure of the sample solution was as follows. The pH of the 0.1 mol·dm^3 NaHCO₃ solution was first adjusted to <2 with HNO₃. 5 mol· dm^3 ammonium acetate (CH₃COONH₄) solution was then added to this pH-adjusted solution so that the concentration of CH₃COONH₄ became 0.1 mol·dm^3, and the pH of the resultant solution was adjusted to 5 with HNO₃ and ammonia water (NH₄OH). Finally, a solution containing La, Gd and Lu was added to this solution to obtain the sample solution for the preliminary experiments. The contents of La, Gd and Lu in the sample solution were 49.5 ng, 50.0 ng and 49.5 ng, respectively.

The conditioning of the CD was carried out by referring to that of Maruta et al. [17] and Takaku et al. [18] . After set on a filter holder of the filtration equipment, the CD was wetted with 20 cm³ of Milli-Q water, washed with 20 cm³ of 3 mol·dm^3 HNO₃ and finally washed with 50 cm³ of Milli-Q water twice under suction. The chemical form of its functional group was then changed to the ammonium () form with 50 cm³ of 0.1 mol·dm^3 CH₃COONH₄, and the CD in the form was washed with a small amount of Milli-Q water. The CD thus conditioned was ready for use.

The whole volume of the above mentioned sample solution for the preliminary experiments was passed through the conditioned CD. After the CD was washed with 100 cm³ of 0.25 mol·dm^3 CH₃COONH₄ solution, the Ln’s (La, Gd, Lu) adsorbed on the CD was eluted out with 35 cm³ of 2 mol·dm^3 HNO₃ solution. The effluent containing the Ln’s was evaporated to dryness on a hot plate, and the evaporation residue was dissolved with HNO₃ solution. The volume of the solution was adjusted to 50 cm³ after 5.00 cm³ of an aqueous solution containing 100 μg·dm^3 In used as the internal standard and 2 mol·dm^3 HNO₃ solution, whose volume was determined so that the final HNO₃ concentration became 0.9 mol·dm^3, were added. The solution prepared in this way was subjected to the ICP-MS. The concentration factor was thus 20.

Three similar experiments were conducted. The mean recovery was 107% ± 8%, 105% ± 7% and 96% ± 1% (± representing the 1σ) for La, Gd and Lu, respectively. This result shows that the CD trapped the Ln’s completely and they were completely eluted out with 2 mol·dm^3 HNO₃ solution after surviving the wash with 0.25 mol·dm^3 CH₃COONH₄ solution within experimental uncertainties.

The analytical results are summarized in Table 3. Ln’s on the order of ng·dm^3 were found both by Procedures A and B. The standard deviations in Procedure B appear to be higher than those in Procedure A due to propagation of errors of c_{Ln, A}, c_{Ln (added)} and m_Ln. Procedure B yielded a higher concentration for every Ln except Eu and Tb than Procedure A, which strongly indicates that the CD adsorbed only a part of dissolved Ln. Although the preliminary experiments showed that the CD did adsorb all the amounts of the Ln’s in the sample solution, it is different from seawater in many ways. Seawater contains practically all naturally occurring elements, and the procedure using the CD can capture most multivalent metal ions except alkali earth metals. Although the decreasing order of the selectivity for trivalent ions on the imminodiacetic acid resin is reported as Fe³⁺ > La³⁺ > Al³⁺ [19] , many of divalent and trivalent metal ions are higher in concentration in seawater than Ln’s. It is thus reasonable that the CD could take up only a part of a Ln dissolved in 4 dm³ of the seawater sample. Consequently, the analytical results by Procedure B are considered to be more reliable than those by Procedure A.

In the standard addition method, the amount of the element to be added to the sample is important. In this study, the amounts of Ln’s added to the seawater sample in Procedure B were determined based on the results by Procedure A; that is, they mimicked those estimated from the concentrations in the seawater sample determined by Procedure A.

The recovery of a Ln in%, r_Ln, from seawater with the CD is calculable by Equation (3) and the results are summarized in Table 4.

It varies between 41% to 100% with the simple arithmetic average of 76%. The recovery of neodymium (Nd) and Sm seems somewhat too low. The average is 81% if the Nd and Sm data are omitted.

^a: Standard deviation (5 measurements).

Since there are no certified values of the concentrations of the Ln’s in the present seawater sample, the actual reliability of the data obtained by Procedure B is unknown. One way to check it may be to draw the Ln abundance pattern and compared it with those found in literature. The Ln abundance pattern obtained by normalization with the contents of the Ln’s in post-Archean average Australian shale (PAAS) [20] is shown in Figure 1 together with those by some other researchers [6] [18] [21] [22] . Our pattern is similar to the others as a whole. The negative Ce and positive Gd anomalies observed in our pattern seem to be reasonable. The negative Ce anomaly is commonly found in Ln abundance patterns of seawater due to the preferential scavenging of Ce(IV) with suspended matters like hydrous iron oxides. The determinant of the Gd anomaly in Ln abundance patterns of seawater is rather complicated. In the present case, it might be caused by the presence of anthropogenic organic complexes of Gd like Gd-DTPA used for a medical diagnosis, since the seawater sample used in this study was collected near the shore with moderate human activity. Such anthropogenic Gd anomalies in river waters and coastal seawaters were reported in several literatures [21] [23] - [25] . The alternative possible cause of the Gd anomaly is the differences in complexation behavior between Gd and its neighbors, Eu and Tb, in seawaters [6] [7] . Compared to the other determined Ln’s, Nd and Sm are obviously overestimated in our experiments, which must reflect the too low recoveries of Nd and Sm.