ANALYST JANUARY 1986 VOL. 111 23 Determination of Nitrogen-I5 by Optical Emission Spectrometry Using an Atomic Absorption Spectrometer Michael H. Timperley Taupo Research Laboratory Division of Marine and Fresh water Science Department of Scientific and Industrial Research P. 0. Box 4 15 Taupo New Zealand and John C. Priscu Department of Biology Montana State University Bozeman MT 59717 USA The determination of 15N percentage abundance by optical emission spectrometry (OES) using an atomic absorption spectrometer (AAS) with an inexpensive radiofrequency generator (27.5 MHz 20 W) and a specially designed cavity is described. Sample tubes for micro-Dumas combustion are easily constructed as required and nitrogen spectra in duplicate for each sample are recorded in 20 s for two-peak or 30 s for three-peak spectra.Because of the cavity design the technique tolerates masses of N per sample tube between 1 and 110 pg. The precision coefficient of variation <6% is equivalent to that reported for other OES techniques and statistical analysis suggested that a large proportion of the analytical variability is introduced during sample preparation. Many laboratories have access to AAS and this together with the cavity and inexpensive r.f. generator makes 15N technology widely available at low cost. Keywords Nitrogen-I5 determination; optical emission spectrometry; atomic absorption spectrometer; micro-Dumas combustion Optical emission spectrometry (OES) is now a widely used technique for measuring the percentage abundance of 15N, particularly in applications involving only a few micrograms of N and also in laboratories for which the cost of mass spectrometry the alternative technique is prohibitive.Mass spectrometry has the advantage of high accuracy and sensi-tivity but requires several hundred micrograms of N in each sample a feature which is a disadvantage for example in studies with phytoplankton. Both OES and mass spectro-metric methods have been reviewed in considerable detail by Fiedler and Proksch.1 Analysis by OES involves two distinct operations. In the first sample N is converted into N2 gas and in the second the N2 either alone or mixed with an inert gas is excited by microwave or radiofrequency energy to emit a spectrum that is analysed to determine the intensities of the emission lines from the 28N2 29N2 and 30N2 molecules.The relative intensities are proportional to the relative amounts of the three molecules in the sample N. Converting sample N usually organic N into N2 can be achieved by the Kjeldahl - Rittenberg procedure in which ammonium produced by Kjeldahl digestion is subsequently oxidised with alkaline hypobromite solution. 1 Alternatively a more direct technique the Dumas method can be used.2.3 In this method sample N is combusted with CuO (plus Cu if N03- is present) and the gases produced with the exception of N2 are absorbed by CaO. The Dumas method is commonly used alone but has also been applied to ammonium produced from sample N by Kjeldahl digestion46 and catalytic ammoni-fication.7 These latter procedures are necessary for large heterogeneous samples5 and where subsamples e.g.for Kjeldahl N analysis are required but for small samples these procedures with their potential for contamination do not offer any advantage over the direct Dumas method.* Purchase of a 15N emission spectrometer was not an option available to the Taupo Research Laboratory but an atomic absorption spectrometer is part of its instrumentation. When preliminary trials confirmed that identical principles were involved in generating an N2 emission spectrum and in generating a spectrum from an electrodeless discharge lamp (EDL) a light source common in atomic absorption spec-trometry,g the possibility of using an atomic absorption spectrometer for emission spectrometric analysis of 15N was recognised.Despite being one of the most widely distributed analytical instruments their use for 15N OES has not been previously reported. In this paper we describe methods for sample preparation, the r.f. generator and design of the r.f. cavity recording the N2 spectrum using the atomic absorption spectrometer and the performance of the system. Experimental Sample Preparation Sample tubes are made as required by sealing one end of a borosilicate glass tube 8 mm o.d. 5 mm i.d. and 200 mm long. The tubes are combusted at 500 "C overnight to remove any nitrogenous substances from the glass surfaces1 and stored in a desiccator until required. Our samples for 15N analysis include phytoplankton on glass-fibre filters liquids either in capil-laries or dried on to glass-fibre filters and finely ground solid matter.Each sample is placed in a tube and oven-dried overnight at 75 "C. Using a small spatula CaO (approximately 15 mg) and a CuO - Cu mixture (approximately 15 mg) are then placed in each tube. Before use the CaO powder is heated at 900 "C for 2 h and stored in a minimum volume desiccator. Re-heating is necessary once each week while the container is being frequently opened. The CuO powder is heated at 500 "C for 2 h mixed with Cu metal granules (4 + 3 m/m) and stored in a desiccator. A small plug of cotton-wool is inserted approximately 20 mm into the open end of each tube to prevent CaO - Cu powder from being carried into the vacuum manifold and the tubes are then immediately attached to the vacuum line.The manifold incorporates six sample tube attachment ports with a tap for each port a vacuum sensing head (Pirani G5C2) a detachable cold trap (liquid N2) an oil diffusion pump (Edwards E203D) and a backing pump (Jayvac JDX 60). Sample tubes are attached to ports by short (25 mm) lengths of neoprene vacuum hose ( 5 mm i.d.). Evacuation of the sample tubes is hastened by brief application of a high-frequency discharge (e.g. Teslar coil) to each tube. When the pressure is less than 2 x 10-4 Torr each tube is sealed by a small gas torch at a point 110 mm from the closed end. The sealed tubes are then combusted at 500 "C for 6 h an 24 ANALYST JANUARY 1986 VOL. 111 Beam aperture 11 3 7 5 I - . Base plate (Al) Coil 9 0.d. Quartz tube 28.7 0.d.25.2 i.d. Coil coated Cu wire swg 20 .Shield (Al) 45.0 0.d. 42.0 i.d. Coaxial connector Supports (All 10 ,+ Beam aperture 38 *I- 7 i Fig. 1. Detail of the cavity designed for optical emission spec-trometry of N contained in straight-sided sample tubes. Dimensions in millimetres allowed to cool slowly. Immediately before analysis a Teslar coil is again briefly applied to each tube. Instrumentation The r.f. energy source (Scientific Associates Hamilton New Zealand) has a maximum output of 20 W at 27.5 MHz with controls for power output and for matching the output to the cable cavity and tube. A shielded coaxial cable leads the signal to the r.f. cavity (Fig. 1) which was specially designed to match the generator output and the size of the sample tubes.Although the structure of this cavity is relatively simple, materials often vary in composition and precise dimensions depending on the supplier and the optimum dimensions of the coil need to be matched (Le. with a grid dip meter) to the generator. A sample tube is placed in the cavity the r.f. activated and the matching controls are adjusted to give maximum emission intensity. Cooling is achieved by an air jet directed into the top of the cavity. The cavity is fitted into an atomic absorption spectrometer (Perkin-Elmer Model 4000 reciprocal dispersion 0.65 nm mm-1) at the position normally occupied by the light source. The cavity can also be placed in the flame position but the physical arrangement is less convenient. The PE 4000 has push-button control of all functions necessary for recording the N2 emission spectrum including wavelength setting scan speed gain forward and reverse scan base-line zero and recorder expansion.The spectrometer is operated in the emission mode with a band pass of 0.07 nm a signal integration constant of 0.2 s and a scan rate of 5 nm min-l. The chart recorder is set to run at 10 cm min-1. Recording the N2 Spectrum Immediately the sample tube begins to emit the following operations are performed: 1. Set the wavelength to 299.3 nm. Zero the base line (AZ on the PE 4000). 2. Set the wavelength to 297.7 nm if there is <35% or to 298.3 nm if >35% of 15N adjust the gain to give a reading of approximately 1.2 units (this is achieved automatically on the PE 4000 by pressing GAIN).Set the chart expansion to full scale for an emission signal of 1.5 units (1500 RECEXP on the PE 4000). Set the wavelength to 297.4 nm. Turn on the chart drive. Scan towards higher wavelength (SCAN + control on the PE 4000). The 28N2 emission line will be recorded at 297.7 nm. Immediately the recorder pen reaches the base line ( i e . , ca. 298.0 nm) after the 28N2 peak change the recorder expansion as required (e.g. 50 RECEXP on the PE 4000 for <1% of 15N) to record the 29N2 emission line at 298.3 3. 4. 5. 6. > v) C a, C t 0 v) c .-c .-.-.-E w 0.366% A A A A A 1.91% A A A Wavelength Fig. 2. Examples of N emission spectra (recorded in duplicate) for samples containing different atom-% 15N contents.A 28N2; B 29N2; C 30N2. A Change of chart expansion; A change of wavelength scan direction; 3 0 ~ = chart expansion relative to 28N2 peak. Broken lines are base lines for the determination of peak height (see text) nm. If required ( i e . at >12.5% of 15N) continue the scan to record the 30N2 emission line at 298.9 nm. Immediately the scan is complete reverse the direction of scanning (SCAN - on the PE 4000) and re-record the spectrum in reverse by changing the recorder expansion at the appropriate wavelengths. This procedure gives two scans for one sample in 20 s if only the 28N2 and 29N2 peaks are recorded or in 30 s if the 30N2 peak is also required. By recording two scans reliable results can be obtained from tubes with decreasing emission intensity (a common feature of tubes containing <1 pg of N) because the two spectra are symmetrical about the centre point and the peak-height averages are therefore unaffected by signal drift.Examples of the spectra obtained for different concentra-tions of 15N are given in Fig. 2. The positions of the operational changes described above are shown. 7. Calculation of Atom- % 15N Irrespective of peak size and spectral interference from other gases the base lines for the 28N2 and 30N2 emission peaks are easily found as shown in Fig. 2. The low wavelength side of the 29N2 peak coincides with a small peak possibly due to CO,8 which confuses the 29N2 base-line position. Comparisons of the measured per cent. of 15N with the actual per cent. of 15N in a series of standard samples using a variety of methods for determining the 29N2 peak base line showed the following procedure to give the most consistent accuracy.The recorder trace was marked at positions 0.17 nm on both sides of the highest point of the 29N2 peak and the line between these two marks was taken as the correct base line. For given scan and chart speeds the required positions were conveniently expressed in chart units or millimetres. Examples of these base lines are shown in Fig. 2. Peak heights were taken from the highest point vertically down to the base line and then multiplied (or divided) by the chart expansion. At low 15N concentrations these base lines gave smaller peak heights than those calculated from our data by other methods.5.8 Standard equations1 were used to calculate 15N concentra-tions: 30N2 + 0.5 29N2 15N (atom-%) = 30N2 + 29N2 + 28N2 x 100 .. (1) 100 2 x 28N 2 * 1 15N (atom-%) = I 1 . . . . . *9N2 where the symbols e.g. 30N2 represent the peak heights. Equation (1) gave more precise results and was preferre ANALYST JANUARY 1986 VOL. 111 25 whenever the 30N2 emission signal was significant i.e. >2% of the 28N2 peak height. This corresponds to approximately 12.5% of 15N. Equation (2) was used when 15N < 12.5%. At 12.5% of 15N the difference between the values given by the two equations was about 0.2% of 1jN. Method Performance Tests The major disadvantage of OES compared with mass spec-trometry for measuring 15N abundance is its lower precision, and accordingly part of this study involved an examination of factors contributing to precision.This examination considered the measured 15N concentration at natural 15N abundance (0.366%)1 in sets of tubes containing different masses of N per tube each set prepared on a different day and replicate analyses of these tubes on the same and different days. A solution containing 1 mg 1-1 of N [(NH4)2S04 BDH Chemicals Aristar grade] was prepared in distilled de-ionised water. One glass-fibre filter (Whatman GF/F 25 mm, combusted at 500 "C for 2 h washed in 1 M HCl overnight, rinsed and stored in distilled de-ionised water) was added to each of six tubes. The filter in each tube was dried over a small flame and aliquots (10 pl = 10 pg of N) of the (NH4)2S04 solution were applied to the filter to give a set of six tubes containing 10 20 30 40 50 and 60 pg of N.The tubes were dried at 75 "C sealed and combusted as described above and analysed for 15N. 'l'he first experiment to evaluate the effects on measured 15N concentration of re-analysing each tube on the same and different days involved analysing one set of six tubes containing between 10 and 60 pg of N with six determinations per tube on each of four days. For the second experiment, designed to assess the effect of tube preparation a new set of tubes was prepared and analysed with six determinations per tube on each of four days. At low 15N abundance (i.e. <1% of ISN) the values determined by OES are usually higher than the true Val-ues,7,8JO probably because of residual C02 in the tubes dissociating to CO in the discharge and producing an emission peak coincident with the 29N2 line.This problem is usually overcome by applying a calibration graph to the values determined by OES. This graph was determined from 11 standard solutions containing differing 15N concentrations prepared by mixing appropriate aliquots of two stock solu-tions one prepared from (NH4)2S04 (BDH Chemicals, Aristar grade) and the other from 15N-enriched (NH4)$04 (99 atom-% 15N) (Amersham International). Both stock and standard solutions contained 200 pg ml-1 of N. For each standard solution three sample tubes were prepared as described above using 100 pl of solution per tube. Six determinations of the 15N concentration in each tube were made.Upper and lower limits on the amount of N per tube necessary to give a measurable spectrum were determined using seston (mostly phytoplankton) from a culture of lake water. Water from Lake Rotorua a eutrophic lake in the centre of the North Island of New Zealand was collected from 1 m below the surface and screened (0.25 mm) to remove zooplankton. Eight litres of this water were supplemented with 3000 pg of 15N [(NH4)2S04 99 atom-% lsN] and 300 pg of P (KHZPO,) and cultured with occasional agitation at 18.5 "C under fluorescent light (100 pE m-2 s-1),* 18 h light 6 h dark. After 5 d the total N in the seston was measured11 and different volumes of the culture were filtered (GF/F filter 25 mm) ranging from the volume giving a just discernible green colour on the filter (approximately 3 pg of N) to the maximum volume filtered in 20 min (approximately 110 pg of N).Triplicate tubes were prepared for each mass of N and each tube was analysed six times for 15N. * E = einstein; 1 E = 1 mol of photons. Our research includes studies on the kinetics of nitrogen uptake by phytoplankton and the following experiment was conducted to assess the precision of the 15N technique for this type of experiment. Water from Lake Rotorua was cultured exactly as described above except that the volume was 9 1 instead of 8 1 and continuous light was used. After each time interval a sample (300 ml) of the culture was taken and six subsamples (50 ml) were filtered (GF/Ffilter 25 mm). At time zero before 15N was added 12 subsamples were taken.Six determinations of 15N were made for each subsample giving 36 measurements at each sampling time except at t = 0 when 72 measurements were made. Results In the first experiment one set of six tubes containing between 10 and 60 pg of N was re-analysed with six determinations per tube on each of four different days. The total data set 144 values had a mean and standard deviation of 0.468 and 0.033% of 15N respectively. Analysis of variance (ANOVA) showed that replicate analysis of sample tubes on the same day (main effect A) did not make a significant contribution to the total sum of squares (ss) and also that re-analysing the tubes on different days (main effect B) and the first-order inter-actions involving A and B (A:B A:C B:C) were not significant.These effects and interactions represent the non-random influence of the spectrometer and this ANOVA showed that neither short nor longer term changes in spectrometer performance were identifiable contributors to the data variability. These non-significant effects and inter-actions were combined with the second-order interaction (A B C) to give the error ss. This (ANOVA) is summarised in the upper part of Table 1. The other main effect the mass of N per tube (C) was a highly significant contributor (31%) to the total ss but as C also included the influence of tube preparation unambiguous interpretation of effect C was not possible from this experiment. The second experiment consisted of four sets of six tubes containing between 10 and 60 pg of N per tube each set prepared and analysed with six determinations per tube on each of four days.This data set had the same dimensions as the data set from the first experiment and the mean and standard deviation 0.456 and 0.031% of 15N respectively were similar. A summary of the ANOVA for this experiment is given in the lower part of Table 1. As was found in the first experiment main effect A the influence of daily changes in spectrometer performance was not significant. Preparing and analysing tubes on different days (D) was the most significant main effect and this contributed 12.9% of the total ss. Although D included the influence of daily changes in spectrometer performance the first experiment showed that this was not significant. The mass of N per tube (B) included the effect of tube preparation and made a small contribution (8.7%) to the total ss but its significance was low.The interaction between B and D (B D) was significant. This experiment showed that tube preparation was an identifiable contributor to data variability but its importance depended on how much of effect B could be attributed to the mass of N per tube. The means of the six determinations made on each tube prepared in the second experiment were not correlated with the mass of N per tube ( r = 0.090 n = 24) and, although this does not preclude the existence of a non-linear relationship it does suggest that the mass of N per tube was not an important component of effect B. The extreme possibility is that the mass of N per tube at least between 10 and 60 pg had no influence on data variability (this was supported to some extent by the results given in Table 3 and discussed below) and this leads to the conclusion that tube preparation contributed 31% and 49% of the total ss in the first and second experiments respectively.The remaining data variability represented by the error ss can be considere 26 ANALYST JANUARY 1986 VOL. 111 Table 1. Analysis of variance (ANOVA) of data matrices obtained by analysing atom-% l5N in samples of (NH4),S04 (0.366 atom-% 15N). First ANOVA is for data from one set of six tubes containing 10 20 30,40,50 and 60 pg of N per tube each tube analysed six times on each of four days. Second ANOVA is for data from four sets of six tubes containing 10 20 30 40 50 and 60 pg of N per tube with six determina-tions per tube each set prepared and analysed on a different day.A = Replicate analyses on the same day; B = analysis on different days; C = mass of N per tube; and D = preparation and analysis on different days Sum of Degrees of Mean Variance Dimension squares freedom square ratio Significance First ANOVA: Main effect: . . . . . . . . . . . . . . . . 12.54 0.001 0.04846 5 0.00969 c Error: A + B + A B + B C + A C + A B C . . 0.10661 138 0.00077 Total sum of squares 0.15507 143 B . . . . . . . . . . . . . . . . 0.01193 5 0.00239 4.12 0.01 D . . . . . . . . . . . . . . . . 0.01771 3 0.00590 10.17 0.001 B:D . . . . . . . . . . . . . . 0.03804 15 0.00254 4.37 0.001 A + A B + A D + A B D Total sum of squares 0.13741 143 .. . . . . . . Second ANOVA: Main effects: 1st-order interaction: Error: . . . . . . 0.06972 120 0.00058 . . . . . . . . TaMe 2. Data used for establishing a calibration graph. Each measured value is the mean of six determinations on one sample except for standard 1 which is the mean of 24 samples with six determinations per sample. Each coefficient of variation (C.V.) was calculated from three tubes six determinations each i.e. 18 values except for standard 1 for which 24 tubes were used 15N Yo Measured Standard 1 2 3 4 5 6 7 8 9 10 11 12 True 0.366 0.588 0.751 1.14 1.91 3.45 6.56 12.9 25.0 49.7 74.4 99.0 1 0.644 0.886 1.18 1.96 3.53 6.53 -12.8 25.9 50.9 75.7 95.9 2 0.634 0.857 1.18 1.95 3.46 6.47 -12.7 24.8 50.6 76.4 96.0 3 0.662 0.835 1.18 1.93 3.45 6.55 -12.8 23.5 50.8 76.3 96.2 Mean 0.456 0.647 0.859 1.18 1.95 3.48 6.52 12.8 24.7 50.7 76.1 96.1 C.V.Yo 6.8 4.8 4.5 3.1 3.5 2.5 3.0 2.5 4.9 1.1 0.6 0.2 Table 3. Effect of different amounts of seston N per tube on measured 15N concentration. Samples taken from Lake Rotorua water cultured with added (15NH4)2S04 (see text). Each mean and standard deviation are for triplicate samples tubes six determinations per tube 15N Yo Volume filtered/ml 5 10 50 100 150 200 Particulate N/ 2.8 5.6 28 56 84 110 Standard Mean deviation 40.8 1.25 43.6 0.67 47.6 0.89 47.5 0.47 47.4 0.70 46.7 1.27 as the analytical noise and this accounted for 51-69% of the total ss.Calibration data are given in Table 2. The measured value for natural 15N abundance 0.456 atom-% was that obtained from the second experiment described above. The means of both measured and true values were transformed by adding 1 to the base 10 logarithm of each value to give symmetrical distributions of positive values for stepwise multiple regres-sion analyses. The calibration equation derived explained 99.99% of the total ss in the measured values and was log (7') = 1.1055 log M - 0.1473 [ln(l + log M)]2 - 0.0332 where T = true value and M = measured value. This equation predicted the 15N concentration in the 12 standards (0.366-99 atom-%) with average and maximum errors of 0.007 and 0.017% of 15N respectively.The results obtained for the measured per cent. of 15N with different amounts of seston per tube are shown in Table 3. Masses of N per tube between 28 and 110 pg had no significant effect on the measured per cent. of 15N. The intensity of the discharge decreased with increasing amount of N per tube and at 110 pg of N the discharge was only 10 mm long and confined within the small coil. Despite this the discharge was perfectly stable as it was for all tubes in this experiment. The tubes containing 2.8 and 5.6 pg of N showed evidence of contamina-tion by 14N. The amount of contaminant 14N was calculated to be 0.50 pg from the tubes containing 2.8 pg of particulate N and 0.46 pg from tubes containing 5.6 pg of particulate N.A third estimate made from the tubes containing (NH4)2S04 (99 atom-% 15N) used for the calibration experiment (Table 2), gave a value of 0.60 pg for contaminant N. The origin of this 14N was not established but it must have been derived from the solid material (e.g. glass surfaces CaO or CuO - Cu mixture) in the tubes as the amount of residual atmospheric N in the tubes was only about 5 x The incorporation of 15N into seston from Lake Rotorua water over a period of 48 h is shown in Fig. 3. For the purposes pg ANALYST JANUARY 1986 VOL. 111 27 I I I Time/h Fig. 3. Change with time of measured 15N concentration in seston (retained by Whatman GF/F membrane 0.7 pm nominal pore size) from Lake Rotorua water cultured with added (15NH&S04 (see text for culture conditions).Vertical height of each point spans mean f 1 s.d. of six sample tubes with six determinations per tube except at t = 0 when 12 sample tubes were measured of demonstrating the precision of the 15N technique in this type of experiment the number of replicates taken at each time (i.e. twelve at t = 0 six at other times) was greater than would normally be taken ( e . g . triplicates at each time). The vertical dimension of each data point spans the mean k one standard deviation of the measured per cent. of 15N. The coefficient of variation for all data points lies between 1.3 and 4.7%. Discussion Several features of this OES procedure for determining 15N abundance offer distinct advantages over the attributes of other published techniques.Foremost among these advan-tages is that of cost. The main component of this technique is the atomic absorption spectrometer and although these instruments are expensive their cost is usually justified by their normal function i.e. the analysis of metals. Extending the use of these instruments to determine 15N abundance is a bonus at little extra cost. Further these instruments are widely distributed throughout the scientific world and although some may require the addition of a wavelength scanning device most could be readily adapted to the technique described here. The r.f. generator is a modified unit from a large radio transmitter and accordingly it was inexpensive. Many laboratories with atomic absorption spectrometers use EDLs and the r.f.generators for these need only have the special cavity attached to be suitable for 15N work. The vacuum manifold represents the major expense but this cost is essential irrespective of the type of spectrometer used to record the N2 spectrum. Most spectrometers for OES analysis use sample tubes with constrictions or other complex shapes but in this system simple straight-sided tubes gave the best results and were inexpensive and convenient to make. The main factor in the successful use of these sample tubes was the cavity design, which caused the discharge to concentrate in the small coil, giving the same effect as the narrow section in constricted tubes.8 The design also contributed to one of the most important attributes of the system.Often in experimental work e . g . with phytoplankton the amount of N retained on a filter or placed directly in a sample tube is either not known or cannot be controlled and for OES techniques that require a relatively narrow range of masses of N per sample tube2JJ2 this must be a substantial disadvantage. The method described here is not restricted in this way and in phytoplankton studies the upper limit is determined by the amount of seston that blocks a 25 mm glass-fibre filter and the lower limit at which contamination becomes important corresponds to a just discernible green colour on the filter. Smaller amounts of N can be determined e.g. 0.5-1 yg but allowance is then necessary for approximately 0.5 pg of contaminant 14N (see references 4 and 13 for a discussion on contamination).The precision (coefficient of variation) achieved by this method from 6.8% at natural 15N abundance to 0.2% at 99% of 15N can be compared with that reported by others for OES techniques e.g. 6.l-O.2% for 0.363-11% of 15N7778J2J4 However most of these precision data either were based on very few samples i.e. <lo or were for sample preparation techniques other than the micro-Dumas method used here. There have been no previous attempts to identify the causes of low precision in OES 15N analyses despite this being a major limitation of the method particularly when micro-Dumas combustion is used.8 In this study at least 13% and possibly 49% of the data variability at natural 15N abundance originated during the preparation of the sample tubes but careful attention to the procedures used did not substantially improve the precision.The type of glass used for sample tubes can influence the shape of the calibration graph but apparently does not affect the precision.15 One further source of data variability was thought to be inadequate recorder response. At the relatively high scan rate used of 5 nm min-1 the peak, i.e. 0.1 nm of the 28N2 emission signal was scanned in approximately 1 s which is close to the recorder response time of 0.5 s. This suggests that the precision would be improved by either a faster response recorder or direct computer acquisi-tion of the emission signal from the atomic absorption spectrometer. Measurements by other OES techniques of 15N at levels close to natural abundance are often higher than the true values and this was found to be so for the technique described here.The procedure used for finding the base line for the 29N2 emission peak gave smaller and more reproducible peak heights than those calculated by other methods and this reduced the amount of curvature of the calibration graph. Our technique for finding the base line is similar to that of Lloyd-Jones et aZ.5 Their method uses the dips on each side of the 29N2 peak but we found that the emission intensities at these dips relative to that at the 29N2 peak varied independently of the 15N concentration possibly because of emission from dissociation products of water16 not absorbed by the CaO. Some workers have advocated the use of a number of linear regression lines to span the calibration graph75,7J7 but this was unnecessary for the calibration data presented in Table 2.After the data had been transformed to give an approximately symmetrical distribution stepwise multiple regression analy-sis gave a single equation with an accurate fit over the entire range of 15N concentrations. A programmable calculator is used to calculate the true 15N concentration from peak heights. Recording the emission spectrum is a rapid operation, taking 20 or 30 s for duplicate records of two- or three-peak spectra. This is achieved by a high scan rate of 5 nm min-l, although at the expense of a small but tolerable loss of precision from the slow recorder response. The limit on the rate of analysis is sample preparation and with the vacuum line described 18 tubes can be evacuated and sealed in 1 h.After combustion these tubes can be analysed in 30-40 min. The system has been in use now for almost 2 years and has successfully analysed more than 1000 samples for marine productivity studies17 and many other samples from lakes.18 Conclusions The wide availability of atomic absorption spectrometers in scientific laboratories and the low cost of the generator an ANALYST JANUARY 1986 VOL. 111 cavity offer a simple and convenient facility for laboratories unable to acquire alternative instrumentation for 15N technol-ogy. This development has greatly expanded nitrogen research in the Taupo Research Laboratory and has the potential if adopted in other laboratories to advance substantially research on nitrogen in biological systems.The authors thank the following staff of the Taupo Research Laboratory Max Gibbs for contributions to the cavity design, Vaughan Wilkinson Rosemary Vigor-Brown and Paul Woods for conquering the vacuum line and for analysing the first 1000 samples and Jan Simmiss for typing the various drafts of this manuscript. 1. 2. 3. 4. 5. References Fiedler R. and Proksch G. Anal. Chirn. Acta 1975 78 1. Murphy T. P. Can. J. Fish. Aquat. Sci. 1980 37 1365. Lernasson L. and Pages J. J. Exp. Mar. Biol. Ecol. 1983, 67 33. Paasche E. and Kristiansen S . Estaurine Coastal Shelf Sci., 1982 14 237. Lloyd-Jones C. P . Hudd G. A. and Hill-Cottingham D. G., Analyst 1974 99 580. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. Ronner U. Sorensson F. and Holm-Hansen O. Polar Biol. 1983 2 137. Goeyens L. G. Stichelbaut L. W. Post E. J . and Baeyens, W. F. Analyst 1985 110 135. Fiedler R. and Proksch G. Plant Soil 1972. 36 371. Barnett W. B. At. Absorpt. Newsl. 1973 12 142. Karlsson L. and Middleboe B . in “Proceedings of a Symposium on Isotopes and Radiation in Soil - Plant Relation-ships Including Forestry,” IAEA Vienna 1972 p. 211. Priscu J. C. and Priscu L. R. Mar. Biol. 1984 81 31. Blackburn T. H. Appl. Environ. Microbiol. 1979,37 760. Lloyd-Jones C. P. Adam J. S. Hudd G. A. and Hill-Cottingharn D. G. Analyst 1977 102 473. Kanazawa S . and Yoneyama T. Soil Sci. Plant Nutr. 1976, 22 489. Lloyd-Jones C. P. Adam J. and Salter D. N. Analyst 1975, 100 891. Burridge J. C. and Hewitt I. J. Anal. Chirn. Acta 1980,118, 11. Priscu J. C and Downes M. T. 1985 Estuarine Coastal Shelf Sci. 1985 20 529. White E. Law K. Payne G. and Pickrnere S. N.Z. J . Mar. Freshwater Res. 1985 19 49. Paper A51207 Received June 1 Oth 1985 Accepted July 22nd I98