JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1987, VOL. 2 27 Studies of a Low-noise Laminar Flow Torch for Inductively Coupled Plasma Atomic Emission Spectrometry Part 2.* Noise Power Studies and Interference Effects John Davies Trace Analysis Laboratory, Department of Chemistry, Imperial College of Science and Technology, London SW7 ZAY, UK Richard D. Snookt Chelsea Instruments Ltd., Avonmoor Business Centre, Avonmoor Road, London W14, UK Noise power spectra of emission signals from the laminar flow torch (LFT) designed in our laboratory and a conventional tangential flow torch (TFT) are presented and described. It is shown that the fundamental frequency of rotation in a conventional torch is completely removed in the laminar flow torch. Moreover, the use of an extended torch is essential in removing air entrainment effects and also for the maintenance of undisturbed laminar flow.The interference effects of phosphate and sodium on calcium emission are assessed in the laminar flow torch and compared with those in a conventional torch. Phosphate interference is not observed in the laminar flow torch while the effect of sodium on calcium emission is shown to be of the same magnitude in both torches. Keywords : Inductively coupled plasma atomic emission spectrometry; laminar flow torch; noise power; interference effects The inductively coupled plasma atomic emission spectrometry (ICP-AES) technique, like all UV - visible spectrometric techniques, is source-noise limited. Above the background equivalent concentration (BEC) level the ICP is flicker noise limited and therefore detection limits and precision are governed by the inherent optical noise observed in the plasma.This noise is caused by two phenomena: firstly, the rotation of the plasma gases induced by vortex stabilisation (thought necessary in conventional torches to produce a stable dis- charge) and secondly, the fluctuations in the analytical signal caused by the nebuliser and sample introduction system. The effect of the rotation of the rapidly swirling plasma gases is to impose a tangential force upon the injector channel perpendi- cular to its motion, giving rise to two possibilities. The first is simple, the rotating gas merely destabilises the injector channel and introduces a higher level of random noise, and secondly, there are tangential forces on the injector channel that cause rotation of the emitting species.It has been observedl-3 that emission is spatially distributed along the boundary regions of the injector channel where these tangen- tial forces are likely to be greatest. The removal of the former effect would be characterised by a lowering of the white noise level and the removal of the latter effect by the removal of particular frequencies in the noise power spectrum. While the removal of turbulence acting on the injector channel has brought an order of magnitude improvement in detection limits4 the removal of principle frequencies is important for use with Fourier transform spectrometry (FTS).5?6 The reason for this is simple. The principle of operation of a UV - visible FT spectrometer, based on the Michelson interferometer, leads to an interferogram that contains all the information in the form of intensity versus mirror displacement. Thus each spectral element in the source produces a signal at the detector modulated at an audio- frequency proportional to the spectral frequency. Therefore any noise at audio-frequency in the source contributes to the * For Part 1 of this series, see reference 4.t Present address: Department of Instrumentation and Analytical Science, University of Manchester Institute of Science and Technology, PO Box 88, Manchester M60 lQD, UK. interferogram and after Fourier transformation appears as side bands to each spectral line. Thus Belchamber and Horlick7 realised the importance of audio-frequencies in the source in UV - visible FTS when they measured the noise power spectra of optical and acoustic emission signals from an ICP, because their spectrometer required that ICP emission signals be observed with measurement system band widths ranging from 0 to 20 kHz.Thus the absence of audio- frequencies in the ICP source is important for FTS. In Part 1 4 we described the fundamental characteristics of a plasma sustained in a laminar flow torch (LFT) employing an extended outer (coolant) tube extending 40 mm above the load coil (ALC). Noise power spectra of an LFT and a tangential flow torch (TFT) at a viewing height of 25 mm ALC were presented and demonstrated that the LFT effectively removed a frequency component at 117 Hz (presumed to be that due to the rotation of the coolant gas in the TFT).Moreover, the white noise (random noise) level was also shown to be reduced by an order of magnitude in the LFT in comparison with the TIT. In Part 2 we present further noise power spectra of both the LFT and TFT using either short or extended coolant tubes to demonstrate the importance of the use of a torch extension. The interference effects of phosphate and sodium concentration are assessed and compared in the LFT and TFT. Experimental Instrumentation The LFT employed in this study has been described fully in Part l.4 In this study the dimensions of the torch base and quartz tubing in the LFT and TFT were the same. The noise power spectrum analyser used in this study was a Solatron 1200 Digital Signal Processor.Signals from the photomultiplier tube (EM1 6256B) were processed by this unit and presented as decibels, dB, (where dB = -20 logloV) versus frequency on a Hewlett-Packard 7470A X - Y plotter. For the noise power studies a Plasma Therm HF 1500 RF generator and matching network were employed with a Spex 1-m monochromator while for the interference studies an28 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1987, VOL. 2 _ _ _ _ ~ ~~ Table 1. Operating parameters R.f. forward powerkW . . . . . . 1.0 Entranceslitheight/mm . . . . . . 3 Coolantgasflow-ratellmin-1 . . . . 12 Injector gas flow-rate11 min-1 . . . . 0.5 Photomultiplier tube voltage/kV . . 1.4 Entrance/exitslitwidths/pm . . . . 35 Auxiliarygasflow-rate/lmin-l . . 0 International Plasma Corporation generator connected to a capacitively coupled manual matching network was used.Both types of generators have been described fully pre- viously.4 Sample introduction into the ICP was facilitated using a concentric glass nebuliser (Meinhard Associates, Model T-230 A3) combined with a Scott-type double-pass spray chamber. In the short torches the coolant tube extended 2 mm ALC while the long (extended) torches extended 40 mm ALC. Procedure All noise power spectra were obtained whilst nebulising a 10 p.p.m. calcium solution and monitoring the Ca I1 393.366-nm line. The operating parameters for the noise power studies are shown in Table 1. For the interference studies on the calcium emission intensities the operating parameters were the same as those in Table 1 except that the PMT voltage was set at 1.0 kV and intermediate and injector gas flow-rates of 0.2 and 0.8 1 min-1, respectively, used for the LFT; a short coolant tube was employed.For all spatial profiles of emission intensity presented in this study measurements were made at 2,5,8,10, 12, 15, 18, 20, 22 and 25 mm ALC. Reagents All reagents used were of AnalaR grade (BDH Chemicals, Poole, Dorset, UK). De-ionised, distilled water was used throughout the experiments. Results and Discussion Noise Power Studies A series of noise power spectra were recorded of plasmas supported in either an LFT or a TIT fitted with either a short or an extended coolant tube. A comparison was made of two 'types of noise: (a) random noise, or white noise, and ( b ) the presence or absence of noise at discrete frequencies thought to be due to rotation of the plasma in the TFT.Fig. 1 shows a comparison of the magnitude of white noise present in the LFT and TFT (determined on the base line between 500 and 1000 Hz) with viewing height. At all viewing heights the white noise level is less in the L f i than the TFT. This was shown to be a real improvement, rather than to be due to an over-all decrease in sensitivity of the LFT, by comparison of the Ca I1 393.366-nm absolute line intensity in both torches and comparison of the d.c. level that is transformed into the noise power spectrum. The Ca I1 line intensity has been shown4 to be slightly greater for the LFT. We think that this improvement is simply due to the absence of instabilities in the injector channel boundary caused by the rotating plasma gases.To investigate further the source of noise at discrete frequencies the noise power spectra of the TFT and LFT were recorded using extended and short torches at viewing heights of 5,15,30 and 50 mm ALC (Figs. 2 and 3). It is apparent from these figures that the extension of the coolant tube has a marked effect on the noise power spectra obtained. Fig. 2(a) shows the noise power spectra obtained from a TFT using a coolant tube extension of 40 mm. The noise power spectra show one main frequency component at 126 Hz, and the 50-Hz 0 I I 1 I I I I 1 10 20 30 40 50 60 Viewing height/mm ALC Fig. 1. Comparison of the white noise level in A, the LFT and B, the TFT 50 126 B - 50 126 C D 126 126 C D 100 200 300 400 500 600 700 800 900 1000 FrequencyfHz Fig.2. Noise power spectra of (a) an extended (40 mm ALC) and ( b ) a short TFT (2 mm ALC) at viewing heights of: A, 5 ; B, 15; C, 30; and D, 50 mm ALC artefact that is due to the mains electrical frequency. The former frequency component was also found in the noise power spectra of the short TFT, Fig. 2(b). In order to ascertain whether this 126-Hz component was due to the rotation of the coolant gas (no intermediate gas was used) the coolant gas flow-rate was varied in both the short and extended TFTs. Fig. 4 shows the effect of coolant gas flow-rate on the 126-Hz feature frequency present in the noise power spectra of Fig. 2 at a viewing height of 15 mm ALC and using the operating parameters given in Table 1. The frequency was indeed found to shift with change in coolant gas flow-rate.As the flow-rate was increased from 8 to 20 1 min-1 (Fig. 4) the feature frequency increased to a maximum and then decreased. This was also observed by Belchamber and Horlick.7 Presumably this is because as flow-rate increases the angular velocity of the gas also increases. The absence of this frequency component in the L m and the frequency shift with flow-rate clearly indicates that this frequency component is due to the rotation of the coolant gas in the TFT. Furthermore, the maximum frequency at which this occurs appears to depend upon theJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1987, VOL. 2 29 160 140 > S P) g 120 F w- II - - - A 1 75 150 C 75 150 D I 1 I I 1 I 100 200 300 400 500 600 700 800 900 1 FrequencyJHz 0 Fig.3. Noise power spectra of (a) an extended LFT (40 mm ALC) and ( b ) a short LFT (2 mm ALC) at viewing heights of: A, 5 ; B, 15; C, 30; and D, 50 mm ALC 80 1 I I I I I 8 10 12 14 16 18 20 Coolant gas flow-rate/l min-1 Fig. 4. Effect of coolant gas flow-rate on feature frequency in the TFT. A, Torch extension 40 mm ALC; B, torch extension 20 mm ALC; C, torch extension 2 mm ALC. Viewing height 15 mm ALC length of the coolant tube extension. Fig. 4 also shows that when the coolant gas flow-rate is varied in coolant tubes of different lengths the maximum frequency at which the rotational feature occurs increases with increasing extension. These are curious phenomena that are difficult to explain. One possible explanation, however, is that the density of the plasma gas decreases as the torch extension increases, and hence, for a given gas flow-rate the angular velocity of the gas increases, which increases the rotational frequency, (provided the pressure does not change).Indeed, when we look at a plasma sustained in an extended torch it appears less dense than one sustained in a short torch. Further evidence for the decreased density of the plasma with torch extension can be obtained from electron density measurements. Using the H(3 Stark broadening method previously reported4 we have measured the electron density at different viewing heights in both the short and extended TFT m 2.5 c N z 1 2.0 i ? 1.5 C w .- U S 2 tj 1.0 al w - I I I 1 1 I 25 0.5 5 10 15 20 Viewing heightimm ALC Fig.5. Comparison of electron densities in A, a short and B, an extended TFT (Fig. 5). At viewing heights of less than ca. 18 mm the electron density is lower in the extended torch than in the short torch. The lower electron density in the extended torch below 18 mm is because an extended torch sustains an extended and less diffuse plasma. Above 18 mm the situation is reversed because the plasma recombines earlier with respect to viewing height in the short torch (i.e., it is cooled by its emergence into the surrounding atmosphere) whereas in the extended torch the plasma is sustained for a greater distance above the load coil, the torch extension preventing cooling from the surrounding atmosphere until much later. The cross-over in electron densities is expected to be dependent on the relative length of the coolant tube and operating parameters such as injector gas flow-rate and coolant gas flow-rate.The effect of decreasing the density would be to decrease the viscous drag between the plasma and its surroundings. Thus the frequency would be expected to increase, as is observed. This is only a tentative suggestion as there are doubtless several other parameters that need to be measured, e.g., surrounding temperature and axial gas velocity. From Fig. 2(b) it can be seen that the noise power spectra obtained in a short TFT show more frequency components when compared with the noise power spectra obtained in an extended TFT. The most likely cause of the appearance of these additional frequency components in the short TFT is air entrainment effects.As the torch extension was decreased so the appearance of these frequencies occurred lower in the plasma. Moreover, none of these additional frequency com- ponents were found when viewing inside the torch extension. In summarising the spectra obtained for the TFT, it can be seen that the noise power spectra obtained for both the extended and short TFTs possess a fundamental frequency component due to the rotation of the coolant gas. In the short TFT, air entrainment effects produce further frequency components present in the noise power spectra. Fig. 3(a) shows the noise power spectra obtained for an extended LFT. No frequency components appear in the noise power spectra except that due to the mains electrical frequency. However, when a short LFT was employed frequency components were found to appear, Fig.3(b). The difference between the appearance of these frequency com- ponents in the LFT compared with those in the TFT is that in the LFT there is a fall off in intensity of these components between 0 and 15 mm ALC, then there is an increase in intensity up to 30 mm ALC and then a final fall off. In the TFT there is a simple increase in intensity up to 30 mm ALC and then a fall off in intensity. These marked additional frequency components in the short LFT are also due to air entrainment effects, which serve to breakdown the laminar flow establi- shed in the torch, thus increasing the magnitude of the noise and negating the advantage of using the LFT. As might be expected, the values of these additional frequency com-JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1987, VOL.2 I 1 I I I 5 10 15 20 25 Viewing height/mm ALC Fig. 6. Effect of phosphate concentration on calcium emission in the TFT, for 1 p.p.m. of Ca. (a) Ca 11,393.366 nm; and ( b ) Ca I, 422.673 nm. A, 0; B, 10; C, 100; and D, 1000 p.p.m. of phosphate ponents are different in both torches due to the difference between the laminar and tangential flow patterns. Thus in order to maintain laminar flow in the plasma sustained in the LFT it is essential to employ an extended torch. One of the possible problems that might arise from the use of extended torches is the gradual devitrification of the torch, which makes viewing through the torch extension undesirable. Certainly with the TFT “fogging” of the tubes in this way decreases the analytical signal after a few weeks.The LFT does not appear to suffer from this problem presumably because of enhanced cooling efficiency obtained by using laminar flow, or because laminar flow prevents heat transfer from the plasma to the quartz. Indeed we have been using the same torch extension daily for three months without signs of devitrification or fogging. Interference Studies Fig. 6 shows the effect of phosphate concentration on the spatial profiles of emission intensity of the Ca I1 393.366-nm line and the Ca 1422.673-nm line in the m. The Ca I1 spatial emission profile is seen to be shifted very slightly away from the load coil as the concentration of phosphate increases. The magnitude of emission is also seen to be reduced by ca.10% for a phosphate concentration of 1000 p.p.m. No shift in the spatial emission profile is observed for the Ca I line but a decrease in emission intensity of the same order as that for Ca I1 is seen. However, in the LFT no effect was seen on the spatial profiles of Ca I and Ca I1 emission intensities or the magnitude of emission. (Moreover, when an end-on viewing configuration was employed8 the presence of phosphate up to 1000 p.p.m. did not affect the emission intensities either. However, in the TFT phosphate interference has been shown to be worse in an end-on viewing configuration.8) A figure for the effect of phosphate concentration on calcium emission in the LFT is not shown as all four lines would be superimposed upon each other.The effect of sodium concentration on the emission intensities of the Ca I and Ca I1 lines for both types of plasma 5 10 15 20 25 Viewing heightlmm ALC Fig. 7. Effect of sodium concentration on (a) Ca I and ( b ) Ca I1 emission intensity, in the TFT (solid line) and LFT (broken line), 1 p.p.m. of Ca. Sodium concentrations: A, 1OOO; B, 100; and C, 10 p.p.m. are shown in Fig. 7 as an enhancement factor, i.e., the ratio of the calcium emission intensity of a solution with 1 p.p.m. of Ca plus the sodium matrix to the calcium emission intensity of a solution with 1 p.p.m. of Ca without the matrix. The presence of such an interference is clear and in both plasmas the magnitude of the interference was found to be similar. The reason why there is an absence of a phosphate interference but the presence of a sodium interference is present is due to the nature of the interferences.The phosphate interference is a classical volatilisation interference effect.9 Optimisation of the conventional ICP source reduces the magnitude of phosphate interference9 and it would appear that in the LFT we have obtained conditions where no such interference occurs. Although the exact nature of the sodium interference is not fully characterised it is not a volatilisation effect but it is more likely to be due to a change in electron density and temperature. The similarity in magnitude of the sodium interference in both torches is due to the fact that the presence of similar amounts of matrix changes the excitation conditions in both torches to the same extent. Conclusions The replacement of tangential flow by laminar flow in the ICP torch has reduced the level of random noise associated with the injector channel by an order of magnitude and removed the fundamental frequency due to the rotation of the coolant gas on the analyte emission signal.Moreover, it is clear from these results that the use of an extended torch is essential to maintain the low noise capability of the laminar flow torch and prevent any disturbance from air entrainment effects that serve to break down the laminar flow established in the torch. The effect of the extended torch is to sustain a plasma, which, when compared with a plasma sustained in a short torch, is less dense near the load coil region because the plasma becomes extended, but decays less rapidly to equilibrium because of the exclusion of the surrounding cold environment until muchJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY, FEBRUARY 1987, VOL.2 31 further up the discharge. The use of torch extensions are by no means detrimental in AES as devitrification in the LFT is much less severe in comparison with the TFT. In our own laboratory one extended torch has lasted for three months of continuous daily use without showing any sign of devitrifica- tion. [Any problem of devitrification is, of course, overcome by the use of end-on (axial) viewing,g which maximises the advantage of simultaneous multi-element determinations with the low-noise laminar flow torch.] The interference effect of sodium on Ca I and Ca I1 emission intensities have been shown to be of the same order of magnitude in both torches. Phosphate interference is not observed in the LFT plasma while a slight depression is observed in the Tm. We have demonstrated that a plasma sustained in an LFT has no disadvantages compared with one sustained in a TFT. Its superior performance in terms of phosphate interference and reduction in noise makes it a superior alternative to the conventional TFT. We acknowledge the support of J. D. by the SERC and Chelsea Instruments Ltd. under the CASE Studentship Scheme. We also thank Dr. J. F. Alder of the DIAS, UMIST, Manchester, for the use of the noise power spectrum analyser and Plasma Therm HF 1500 ICP. 1. 2. 3. 4. 5. 6. 7. 8. 9. References Furuta, N. , and Horlick, G., Spectrochim. Acta, Part B, 1982, 37, 53. Eckert, H. U., and Danielsson, A. , Spectrochim. Acta, Part B, 1984,39, 15. Caughlin, B. L., and Blades, M. W . , Spectrochim. Acta, Part B, 1984,39, 1583. Davies, J., and Snook, R. D., J. Anal. At. Spectrom., 1986,1, 195. Thorne, A. P., Anal. Proc., 1985,22, 63. Horlick, G., and Yuen, W. K., Appl. Spectrosc., 1978,32,38. Belchamber, R. M., and Horlick, G., Spectrochim. Acta, Part B, 1982,37, 17. Davies, J., Dean, J. R., and Snook, R. D., Analyst, 1985,110, 535. Kornblum, G. R., and de Galan, L., Spectrochim. Acta, Part B, 1977,32,455. Note-Reference 4 is to Part 1 of this series. Paper J6l35 Received May 8th, 1986 Accepted August lst, 1986