Inductively coupled argon plasma atomic emission spectrometry with

P.S.C. van der Plas , E. Uitbeijerse , M.T.C. de Loos-Vollebregt , L. de Galan. Spectrochimica Acta Part B: Atomic Spectroscopy 1987 42 (9), 1027-1038...
0 downloads 0 Views 910KB Size
Download PDF
Anal. Chem. 1984, 56,2329-2335

Table 1. Detection Limits with Ultrasonic Nebulizer detection limit,

detection limit,

element

pg.L-1

element

/.kg.L-'

A1

5.1 0.091 0.55 0.42 0.91 0.35

Mn Mo Ni Pb V Zn

0.042 0.50 0.46

Cd co Cr

cu

Fe

0.94 0.35 0.12

because the dry spots are alternately formed and then covered with sample. The relationship between power and ICAP response for various sample flow rates is shown in Figure 3. The ICAP response shown is the reading in the manganese channel. The responses for other metals were similar. It is seen that at each flow rate there is an optimum power level and that the optimum response increases with sample flow rate up to about 7 ml.min-l. We attribute the flattening of response at the higher flow rate to the inability of the desolvation system to transport the aerosol. The sample flow rate and power levels of 3.0 ml-min-l and 40 W, respectively, were chosen as representing an acceptableICAP response with extended transducer life.

DRYING AND DESOLVATION A large amount of aerosol is produced by the nebulizer. Under the "routine analysis" conditions (3 ml-rnin-l, 40 W) approximately 0.5 mL.min-l of water leaves the nebulizer chamber and is collected via the condenser system. From measurements similar to those previously described (3), we estimated that the equivalent of 0.3 ml-min-' of sample reaches the plasma under these conditions. We believe that the largest potential loss of aerosol is by agglomeration of the wet particles. The small volume nebulizer chamber and the annular drying chamber permit the rapid removal and drying

2329

of the particles to minimize this agglomeration. The condenser provides fast and efficient condensing of the water vapor with a minimum of dead volume. The volumes of the nebulizer chamber, drying chamber, and condenser are 26, 25, and 35 mL, respectively. These small dead volumes also give fast sample washout characteristics;e.g., when a sample containing 1mg-L-' of manganese is followed by a 0.2% HNO, wash, the ICAP signal falls to 1% of the sample reading within 30 s of the changeover.

SENSITIVITY AND DETECTION LIMITS The detection limits of the system were determined in a manner similar to that previously described (3). The detection limits observed for a number of elements in water from Lake Ontario as is, i.e., with no preconcentration,are shown in Table I. These are about 6 times better than those obtained with the pneumatic nebulizer/heated spray chamber system. The improvement is caused by a 3-fold increase in sensitivity and a 4-fold reduction in the variance; i.e., the standard deviation of replicate analyses decreased by a factor of 2. This lower variance obtained with the ultrasonic nebulizer, as compared to the pneumatic nebulizer/heated spray chamber system, is a measure of the greater stability of the nebulizer within the analytical period of a day. The long term stabilities of the two systems are about the same. Analyses carried out on the same standards over a period of 3 months give "millivolt readings" that do not vary by more than 10% for either of the systems. LITERATURE CITED (1) Olsen, K. w.; Haas, W. J.; Fassel, V. A. Anal. Chem. 1977, 49, 632-637. (2) Taylor, C. E.; Floyd, T. L. Appl. Specfrosc. 1981, 35, 408-413. (3) Goulden, P. D.; Anthony, D. H. J. Anal. Chem. 1982, 5 4 , 1678-1681.

RECEIVED for review March 7,1984. Accepted June 25, 1984.

Inductively Coupled Argon Plasma Atomic Emission Spectrometry with an Externally Cooled Torch Peter A. M. Ripson, Liesbeth B. M. Jansen, and Leo de Galan* Laboratorium voor Analytische Scheikunde, Technische Hogeschool, Jaffalaan 9,2628 B X Delft, The Netherlands An evaluation is presented of a torch for inductlvely coupled plasma atomic emlsslon spectrometry that is coded externally by either air or water. The total argon consumption Is 1 Umln and the air-cooled torch requires substafitlaliy less rf power than elther the water-cooled torch or a conventional ICP. The optimization study shows that only the Incident power, the argon flow rates, and the tip diameter of the sample Introduction tube are critical parameters. When run under compromise conditions, the externally cooled I C P Is easy to operate and accepts hlghly satted aqueous solutions as well as organic solvents. All designs tested show excellent anaiytkal dynamic range, precision, and stablilty. However, in terms of detection power and matrix Interferences the air-cooled torch with an outer tube diameter of 16 mm yields superior performance, fully Comparable to that of a conventional I C P torch.

In a previous publication we have described a torch for inductively coupled plasma (ICP) atomic emission spectrom0003-2700/84/0358-2329$0 1.50/0

etry that uses a total argon consumption of only 1L/min by virtue of external cooling with pressurized air (I). The analytical data provided were promising but did not allow a complete assessment of the analytical performance of the torch. Since then a similar torch utilizing external cooling with water has been designed. Because a liquid is a much more efficient coolant medium than a gas, the use of water cooIant would enlarge the range of practical operating conditions (2). In a later publication an analysis of the power requirements of both externally cooled torches (3) demonstrated that the air-cooled torch runs on less than 400 W of incident power, whereas the water-cooled torch requires about 800 W. In order to compare the two designs with existing ICP torches, an evaluation of their analytical performance is necessary. Nearly all investigators who have constructed a low consumption ICP either by modified internal or by external cooling (4-20) indicate that the detection limits are roughly comparable to those of a conventional ICP. If matrix interferences are reported, they occasionally show less promising results (5-7,I.l). Other analytical properties are only rarely @ 1984 American Chemlcal Society

2330

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

Table I. Instrumentation

1.5

observation

rf generator work coil

Philips PV 8490, 1.5 kW, 50 MHz two-turn copper plates (I),inner diameter 23 mm torch dimensions see'Figure 1 and Table I1 sample introduction narrow bore (100 pm) Babington nebulizer fitted to a 20-mL single-wall chamber (15), force feeding of the sample solution by argon pressure or a peristaltic pump monochromator McPherson Model 2051, 1 m Czerny-Turner with a 10-cm grating, 1200 grooves/mm, blazed for 250 nm and adjustable slits, set to 25 pm photomultiplier EM1 62565 operated at 700 V amplifier PAR Model 120, lock-in amplifier operated at 400 Hz with a time constant of 0.3 s specified. In two cases the linear range of calibration curves has been studied and shown to be comparable with that for conventional plasmas ( 4 , 6 ) . Finally, fair accuracy has been reported twice (6,1I), and-somewhat surprisingly-for the 5 L/min torch reported recently by Rezaaiyaan et al. (12) this is the only analytical property mentioned. Little, if any, quantitative information is available on precision, stability, maximum permissible salt uptake, and the application of organic solvents. In this study we evaluate the analytical performance of an air-cooled and a water-cooled ICP with respect to the properties mentioned above. Attention is also given to practical aspecta such as ignition and ease of operation. The two plasma systems are compared mutually and with a conventional ICP.

EXPERIMENTAL SECTION Table I lists the instrumentation used for the analytical measurements. Two different systems have been evaluated: (a) an air-cooled ICP with a torch consisting of two concentric quartz tubes contained in a coil made from two copper plates (1)(outer quartz tubes with inner diameters of 13.5 and 16 mm have been investigated)and (b) a water-cooledtorch, contained in the same flat-plated coil, but with a cooling jacket after Kawaguchi et al. (7) to replace the outer quartz tube (because of the space taken up by the jacket, only one inner diameter of 13.5 mm has been evaluated). The incident power was determined by a method described previously (3),which relies on data for the anode current and for the power dissipated in the work coil, both with and without the plasma burning. Argon flow rates were measured with a soap bubble meter, the air flow rate with a wet gas meter and the cooling water flow rate by timing with a 2-L volumetric flask. The latter procedure, but with a 10-mL flask, was also used to determine the rate of sample delivery to the nebulizer. The observation height (OBH) was varied using a periscope arrangement of one fixed and one rotatable plane mirror in the optical path. The torches have been optimized for signal-to-backgroundratio, which is the usual optimization criterion in ICP analysis. In order to evaluate the influence of the various parameters, a univariate optimization procedure is used. The precision of the signal to background ratio (SBR) values is a few percent. Possible interdependences are accounted for by iteration. The parameters considered in this study are shown in Figure 1 and can be distinguished in two categories. Continuously variable parameters are the flow rates of the coolant medium (F&, Fwak,), of the plasma argon (F ) and sample carrier argon (Fc), and of the sample solution (F,f, the distance from the rf coil to the top of the outer torch tube (HT), to the sample introduction and to the center of the observation window (OBH),and tip (HJ, the incident power (pin).Discontinuously variable are the tip diameter of the sample introduction tube (dJ, the inner diameter (din),and the width (w)of the torch tube. Detection limits were derived corresponding to a 3u criterion and a time constant of 15s as usual in ICP literature (13). Matrix effects are quantified as the percentual difference between the net signals of 5 mg/L of analyte with and without interferent. The short term precision was determined by making six rep-

Fwater I Fair

I-

FP FS

FC Flgure 1. Schematic drawing of the experimental system with the

investigated parameters. The dotted lines denote the water-cooiihg jacket after Kawaguchi ( 7 ) . If this Is present Fa,,is zero, and if not, Fwa,w is zero. etitive wavelength scans over the analyte line of interest within a period of 5 min and calculating the standard deviation in the net peak signal. This was repeated 10 times at concentrationlevels of at least a hundred times the detection limit with a time constant of 0.3 s. The mean standard deviation and the 95% confidence interval were determined by using standard statistical methods (14). The long term stability was determined at the same concentration levels by repetitive wavelength scans at selected time intervals, e.g., every half hour over 6 h. The ICP systems were allowed to warmup half an hour before starting the measurements. The 16 elements studied are listed in Table IV together with the wavelength of their most sensitive spectral line. These transitions are used throughout the paper and agree with those used in a conventional ICP.

RESULTS AND DISCUSSION The Influence of the Operating Conditions upon the Signal to Background Ratio, Coolant Flow Rates. Air and water are the obvious choices for external cooling by gas and liquid, respectively. In the case of air, the cooling capacity and hence the incident power that can be accepted increases with the flow rate. Therefore, the flow rate of the air, taken from the laboratory's pressurized air system, was set to the maximum value of 62 L/min attainable with the present design. The cooling capacity of water is much larger. The flow rate is not critical and was set to 1.3 L/min. A value over 1.0 L/min is needed to prevent the formation of bubbles from boiling water. The Outer Torch Tube. Theoretically, the transfer of rf power from the coil to the plasma proceeds more efficiently the closer the diameter of the plasma torch approaches the inner diameter of the rf coil (23 mm). However, in the case of water cooling, 9 mm is needed for the cooling jacket, so that the maximum inner diameter of the plasma torch, &, is 13.5 mm. The air-cooled torch can be made wider. Heat transfer through the quartz is facilitated by thin walls, but to obtain acceptable rigidity the wall thickness was set at 1.5-2 mm. This leaves room for a maximum inner diameter of 18.5 mm. However, a previous report has shown that such a wide tube yields an unstable plasma (1). Therefore, tubes with an inner diameter of 13.5 mm and 16 mm were selected. The minimum extension of the outer tube above the rf coil needed to prevent air entrainment is 20 mm for the air-cooled torch and 15 mm for the water-cooled torch. Consequently,

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

d

2331

, , :.* , *: . , , , , I a5 lA

-dc.mm

Figure 2. Influence of the dlameter of the carrier argon introduction orifice upon the slgnal to background ratio: (0)air-cooled torch (din = 13.5 mm), 2 mg/L Zn; (*) water-cooled torch, 1 mg/L Mg.

Flgure 4. Influence of the plasma argon flow rate upon the slgnal to background ratio: (0)air-cooled torch (d,,,= 13.5 mm), 2 mg/L Zn; (0)air-cooled torch (din= 16 mm), 1 mg/L Mg; (*) water-cooled torch, 1 mg/L Mg. Shaded regions denote overheating of torch tubes. SIB

t15

-

10

5-

0'

xx)

=Op,w

Figure 9. Influence of the carrier gas flow rate upon the slgnal to background ratio: (0)alrtooled torch (dh = 13.5 mm, d , = 0.48 mm), 2 mg/L Zn; (0)aircooled torch (dh = 16 mm, do = 0.48 mm), 1 mg/L Mg; (*) water-cooled torch ( d , = 0.57 mm), 1 mg/L Mg.

Flgure 5. Influence of the Incident power upon the signal to background ratio: (0)air-cooled torch ( d , = 13.5 mm), 2 mg/L Zn; (0) air-cooled torch (din= 16 mm), 1 mg/L Mg; (*) water-cooled torch, 1 mg/L Mg.

all observations must be made through the outer tube and the water jacket. Sample Introduction. Neither the solvent delivery flow rate, F,, nor the distance from the sample introduction tip to the rf coil, H,,proved to be very critical. In agreement with the virtually constant nebulization rate reported previously for sample delivery rates between 1and 1.7 mL/min (15),the SBR varies less than 10% as long as the delivery rate exceeds 1.2 mL/min. Optimum SBR values are obtained when the sample introduction tube is raised close to the rf coil. Between 0 and 3 mm the distance is not critical, although it influences the optimum observation height somewhat. Over 4 mm the SBR rapidly decreases, and the plasma becomes unstable when the inner torch tube is more than 10 mm below the rf coil. All data were collected with a separation of 1to 2 mm. The carrier gas flow rate, F,, and the tip diameter of the sample introduction tube, do,are more critical and interdependent. Therefore, for each different tip diameter the argon carrier gas flow rate was optimized for SBR. In this and in the following studies no significant differences were observed between various elements. The results presented in Figures 2-5 refer to zinc and magnesium, that serve as typical pilot elements. The influence of the tip diameter is shown in Figure 2. The two air-cooled torches show the same variation with an optimum tip diameter of 0.48 mm. In the water-cooled torch the optimum is shifted to 0.57 mm. For these tip diameters the influence of the argon carrier gas flow rate is shown in Figure 3. Its influence is not very critical for the water-cooled

torch and a value of 170 mL/min was chosen. Cleary defined optima are found for the air-cooled torch, and the wider torch appears to require a smaller carrier gas flow rate: 120 mL/min for the 16-mm tube and 145 mL/min for the 13.5-mm tube. The combination of these tip diameters and flow rates leads to linear velocities for the cold carrier gas ranging from 11to 13.5 m/s, which is remarkably close to the value of 14 m/s derived for our conventional plasma that uses 1.15 L/min of argon through a 1.3-mm tip. For lower linear velocities the carrier gas does not penetrate the plasma ring sufficiently and for higher velocities the residence time of the sample falls off. Plasma Argon Flow Rate. The influence of the observation height, the incident power, and the plasma argon flow rate is mutually interdependent. A systematic variation of the plasma argon flow rate therefore required subsequent adjustment of the other two parameters. In the air-cooled plasma Fp appears to be a very critical parameter and the SBR increases sharply with decreasing Fp values. No optimum is reached, since the plasma cannot be operated below either 0.40 or 0.65 L/min for the 13.5-mm and 16-mm torch, respectively. Below these values the outer tube starts glowing visibly and plasma flickering occurs. For the present system a safety margin was built in and plasma argon flow rates of 0.50 and 0.80 L/min were selected. The water cooled torch behaves differently. The cooling capacity of water is so large that the torch never starts to glow. However, the influence of Fpupon the SBR is only weak and an optimum is found at 1L/min, which value was selected for further testing. It is tempting to speculate about the

2332

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

Shl

Table 11. Operating Conditions

-

tI I

0

/

-

OBH, mm

Figure 6. Influence of the observation height upon the signal to back ratio: (0)air-cooled torch (d,, = 13.5 mm), 2 mg/L Zn; (0)alr-cooled torch (dh = 16 mm), 1 mg/L Mg; (*) water-cooled torch, 1 mg/L Mg.

difference between the two torch designs. The heat sink provided by the water-coolingjacket probably influences the temperature distribution in the plasma, so that only the central region dominated by the carrier gas is of interest. In the air-cooled torch the outer tube becomes quite hot (3))so that the plasma gas may be heated more evenly and contribute to the emitted intensities. A reduction of the plasma gas flow would then prolong the residence time of the analyte atoms and increase the intensities. Incident Power. The influence of the incident power on the SBR is shown in Figure 5. The measurementswere carried out at the plasma argon flow rates selected above, and for every data point the OBH was adjusted to yield maximum intensity. Again, a significant difference between air-cooled and water-cooled plasmas occurs. For the air-cooled plasma, the SBR first rises sharply with the incident power and then reaches an optimum after which it slowly decreases again. For the water-cooled plasma, on the other hand, the SBR tends to increase slowly but continuously with decreasing incident power, and no optimum is reached, because the plasma extinguishes below 500 W. The difference in the lower limit for the incident power is due to the fact that the water-cooling jacket acts as a heat sink that draws away power (15). Indeed, air-cooled plasmas can be sustained at incident power levels as low as 150 W, although they are not analytically useful then. On the other hand, water-cooled torches can be run a t high power without fear of overheating (10, 13),whereas for aircooled torches the upper power level, above which torch glowing occurs, is about 325 W for the 13.5-mm torch and 425 W for the 16-mm torch. The difference in upper limit can be explained in terms of power balances previously described (3). The larger area of the heat conducting surface in the wider torch accounts for some 75 W of additional conductive heat transfer. The remaining 25 W can be explained by the increase of convective heat transfer by the 60% higher argon flow rate (Figure 4). Thus, the incident power level is typically 100 W higher for the 16-mm air-cooled torch, when compared with its 13.5 mm equivalent, but the power delivered by the rf generator is virtually the same for both systems, because of a gain in coupling efficiency at larger plasma diameters. At the selected optimum incident power levels of 270 and 400 W, the generator has to deliver 450 and 500 W, respectively. By comparison, the selected incident power level of 600 W for the water-cooled torch (a safety margin of 100 W was built in) corresponds to 1100 W of generator power. Hence, the application of smaller (solid state) rf generators seems entirely possible for air-cooled plasmas at optimum incident power levels. Obseruation Height. Up to this point the OBH has always been set to the point of maximum line intensity, which lies

selected value parameter dintmm d,, mm

F,, mL/min F,,, L/min Fcoolant,

L/min

F,, mL/min

H,,mm HT, mm observation window, mm OBH,mm pint

w

w , mm

air

air

water

13.5 0.48 145 0.50 62 2 1-2 20 2 12 270 1.5-2.0

16 0.48 120 0.80 62 2 1-2 20 2 11 400 1.5-2.0

13.5 0.57 170 1.00 1.3 2 1-2 15 2 10 600 1.0

between 8 and 9 mm above the work coil for both externally cooled plasmas. The results in Figure 6 indicate that a changeover to SBR does not lead to major changes in optimum OBH. Indeed, for the water-cooled torch and for the 13.5-mm air-cooled torch the SBR is remarkably insensitive to the OBH. It is only for the 16-mm air-cooled torch that a shift in OBH from 8 to 11mm increases the SBR for zinc by 40%. Even then, the variation is smooth. A similar behavior was found for all elements investigated. The results in Figure 6 contrast with the significant lateral variations observed in a conventional ICP torch (16). The minor influence of the observation height means that an optimum value for one element entails little loss in detection power for other elements. The operating conditions derived from the preceding optimization study are collected in Table I1 for all three torch designs. It is recalled that these optimum values vary little or not at all with the element or the transition considered. Consequently, the values presented in Table I1 represent what is known in ICP literature as compromise conditionsand they have been used throughout the evaluation of the analytical performance of the externally cooled plasmas. Operational Characteristics. Ignition. Ignition of the air-cooled plasma is straightforward, provided that the sample introduction tube is carefully placed in the center of the torch and the torch itself is well centered in the work coil. The plasma argon, carrier argon, and solvent delivery rates can be set to their operational values (Table 11) prior to ignition, but the incident power must be within the range of 350-400 W. After ignition with a Tesla coil, no readjustment of operating conditions is necessary for the 16-mm air-cooled torch, but for the 13.5-mm torch the incident power level must be quickly reduced to its operational value to prevent torch deformation from overheating. Ignition of the water-cooled plasma is somewhat more elaborate. Again, the torch assembly must be well centered, but, furthermore the plasma argon and the carrier argon flow rates must be set to 1.5 L/min and 75 mL/min, respectively, and no sample should be delivered to the nebulizer prior to ignition. After ignition at an incident power level of 800 W, all flow rates and the incident power level should be readjusted to their operational values (Table 11), but in view of the large cooling capacity of water the time is not critical. Working Range of Operating Conditions. Beyond the limits reported in Table I11 either the plasma vanishes (low incident power, high flow rates for plasma argon and carrier argon) or the torch starts to glow, eventually leading to deformation (high incident power, low flow rates for plasma argon and occasionally even for carrier argon and solvent delivery). For each parameter the limits were determined at constant values of the other parameters, as stated in Table 11. The results in Table I11 indicate small working ranges for the incident power in the air-cooled plasma. This observation can be

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

Table 111. Working Range of Operating Conditions

Table IV. Detection Limits in Externally Cooled Plasmas

working range air cooled din = 13.5

din = 16

water

mm

mm

cooled

conventionala

500-1000 400-1750 175-425 Ph,w 150-325 0.5-2.5 7-40 Fp,L/min 0.4-3.0 0.6-3.3 0-500 0-5000 0-500 F,, mL/min 50-500 arbitrary arbitrary arbitrary F,, mL/min >0.5 Measured with a conventional three-tube concentric torch combined with a 2.5-turn copper tubing coil connected to a LINN FS 4 generator.

explained by the limiting cooling capacity of air, as described previously (3). In the 13.5-mm air-cooled plasma there is even a lower limit to both sample introduction flow rates (Fc,F,). The situation is more favorable for the 16-mm air-cooled plasma, where the sample introduction flow rates can be turned down to zero and where, in general, the working ranges are somewhat wider. As expected, the working range for the incident power is larger in the water-cooled plasma. At the lower limit of the plasma argon flow, the water-cooled plasma vanishes before the torch starts to glow In general, water cooling ensures safe working ranges for all important operating conditions. Continuous Operation. With the parameters adjusted to within their working range, air-cooled torches have been operated 8 h/day for several weeks without noticeable damage. Occasionally, however, cracks in the quartz tube were noticed that eventually led to torch rupture. Possibly, the cracks arise from tensions in the quartz or from nonuniform air cooling. In the present design the pressurized air is blown perpendicularly onto the torch through five tubes positioned around the outer quartz tube ( I ) . Cooling is thus not radially homogeneous. Perhaps a tangential flow of the cooling air, as described by the Robin et al. (17)or the use of more rigid quartz tubes might prove effective. In the 13.5-mm air-cooled plasma solvent should be supplied continuously to the plasma during operation. The time of sample interchange is not critical, but if no solvent is aspirated for a period longer than 30 s, torch deformation can occur. If, for some reason solvent delivery must be stopped for an elongated period, the incident power level must be decreased to a value below 250 W. This limitation does not occur in the 16-mm air-cooled plasma. The water-cooled plasma has been operated full days for over a year and no damage to the torch was observed. In general, this plasma is as easily operated as a conventional ICP and in this respect water cooling should be favored above air cooling in the present design. Highly Salted Solutions, Organic Solvents. Concentrations up to 5% (w/v) of sodium chloride and sodium tetraborate have been aspirated in the air-cooled and the water-cooled plasmas, without any visible instability of the plasma. The incident power need not to be adjusted. In the air-cooled plasma no salt deposits on torch and tip were observed at the operational incident power level (Table 11). At power levels, which were 50 W lower, deposits on the torch tube could be seen. In the water-cooled plasma some salt deposit was found on the torch tube at the suggested operational power level of 600 W, and to analyze highly salted solutions a higher incident power (800-900 W) is desirable. For the introduction of organic solvents the carrier gas flow rate must be reduced to 75 mL/min and the incident power increased to 330 W for the air-cooled plasma (dh = 13.5 mm) and to 900 W for the water-cooled plasma. When the distance from the sample introduction tip to the work coil, H,, is set

2333

species A1 I

BI Ca I1 Cd I1 co I1 Cr I1 cu I Fe I1 La I1 Mg I1 Mn I1 Na I Ni I1 Pb I1 v I1 Zn I

wavelength, nm 396.1 249.7 393.3 226.5 238.8 267.7 324.7 259.9 408.6 279.5 257.6 588.9 221.6 220.4 309.0 213.9

detection limit, Gg/L air cooled din = 13.5 din = 16 water mm mm cooled 40 16 11 2 11

0.3 2 7 240 8

40 10 0.13 16 44 10 5 10 0.3 2 7 26 230 8 12

70 35 45 35 17 55 49 1 24 24

conven-

tional" 28 5 0.19 3.5 6 7 5 6 10 0.15 1.5 29 10 42 5 2

aFrom ref 19

to a value of 3 mm, methanol, ethanol, xylene, and MIBK can be aspirated continuously. The plasmas show a green tail flame. No carbon deposits onto the tip, but after aspirating MIBK for half an hour or more a carbon deposit was found on the torch tube. A few preliminary analytical measurementshave been made with organic solvents. As in the conventional ICP (18) the molecular background band spectrum increases strongly. Also, net signals tend to decrease, as expected from the change in operating conditions. As a result SBR values are much smaller than those for aqueous samples,and for both externally cooled plasmas the detection limit of calcium in methanol is 2 orders of magnitude higher that than in water. Therefore, in order to obtain acceptable SBR values modification of nebulizer design (e.g., spray chamber cooling), torch design, and operating conditions (e.g., oxygen doping) appears necessary. Analytical Peformance. Detection Limits. The data presented in Table IV reveal that the detection limits in the externally cooled plasmas are higher than those in a conventional ICP (19).The average difference is a factor of 2.5 for the air-cooled torches and 5.8 for the water-cooled version. All data refer to a single set of operating conditions. The results clearly indicate that compromise conditions can be formulated for externally cooled plasmas just as for a conventional ICP. Under such conditions the detection power of the air-cooled plasma comes close to that of a conventional ICP. The detection limits tend to become somewhat poorer with decreasing wavelengths due to the combined effects of higher noise levels and smaller net intensities at low wavelength in comparison with the conventional ICP. The higher noise level is predominantly due to the instrumentation used and will be improved with better mirrors, lenses, and detector. The smaller signals, however, can be correlated with the higher excitation energies of the transition used. Indeed the effect is more pronounced in the water-cooled ICP, where the cold torch wall influences the plasma temperature. In the aircooled plasma, for which we have previously reported an excitation temperature of 5500 K ( I ) , the net signals are only a little weaker than in the conventionalICP, for which similar temperatures have been reported (20). Precision and Stability. The short term precision, defined as the standard deviation over 5 min of a signal well over the detection limit, was measured for several elements. For the 13.5-mm air-cooled torch the precision varies between 3 and 4%, whereas this range is 1-2% for the 16-mm version and

2334

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

100

i“i”

A ~um.7nm

100

p

50

0

- 50 -100

-1

0

I

-1,

mg/L -50

’

-100

5

-lOgC,pg/L

Figure 7. Calibration curves for externally cooled plasmas: (0)air-

coded torch (dh = 13.5 mm), Mg slope = 1.02 (correlation coefficient = 0.999), Zn slope = 0.98 (0.9999); (0)air-cooled torch (dh = 16 mm), Mn slope = 0.93 (0.999), Zn slope = 0.94 (0.999); ( S r ) water-cooled torch, Mn slope = 0.96 (0.999), Cr slope = 0.95 (0.999).

B

A Mn257.6

100

nm

Mn257.6nm

loo

50

50

Wavelengths are given In Table IV, operating condltlons are given in Table 11. A denotes detectlon limits taken from Table IV.

for the water-cooled torch. The latter figures are fully comparable to results with a conventional ICP without special precautions such as power stabilization (presently available in most commercial generators), mass flow controllers, highpressure nebulization, and internal standards (21). The long term stability, defined as the variation in 5-min averages over 6 h is similar for all three plasma systems. After half an hour of stabilization time the results varied between 3 and lo%, which is again comparable to conventional ICP values without special precautions (22). The instability is not due to unidirectional drift but to long-term fluctuations, that were found to be correlated with variations in the incident power. Thus, if better long-term stability is required, power stabilization is essential. Indeed, for the 16-mm air-cooled torch we found a stability of 1.5% over a particular 3-h period, where the incident power remained constant to within 0.3%. Dynamic Range of Calibration Curves. Figure 7 shows some calibration curves on a double logarithmic scale. The lower decade in these curves runs from the lowest data point measured to the detection limit (open triangle) stated in Table IV. The slopes of the least-squares fit are between 0.93 and 1.02, indicating linearity over at least 5 decades. Only for Mg, measured in the 13.5-mm air-cooled plasma, the highest data point (correspondingto 500 mg/L) deviates significantly from the straight line, indicating the onset of self-absorption a t a magnesium concentration of 100 mg/L. With respect to dynamic range the externally cooled plasmas are fully comparable to a conventional TCP with similar significance for multielement analysis capability. Matrix Effects. Previous reports on matrix effects in a conventional ICP have stressed a difference in behavior between so-called “hard” and “soft” lines (16-23). Since Boumans and Lux-Steiner (24) have found that especially the soft lines are matrix sensitive, we have taken two soft lines (Cu 1324.7 nm and Mn 1403.0 nm) and one hard line (Mn I1257.6 nm) for the evaluation of matrix effects. Four concomitants have been selected, representing classical volatilization and ionization interferences: phosphate, nitrate, sodium, and calcium. The measurements have been carried out at a constant analyte concentration of 5 mg/L and concomitant concentrations ranging from 10 mg/L up to a maximum of 10000 mg/L, which concentrations can be easily sustained by the externally cooled plasma. The results for the Cu I and the Mn I1 line in Figure 8 reflect normal ICP behavior. Matrix effects tend to increase continuously with increasing concentration of concomitant. Up to 1000 mg/L, all matrix effects are small and, in general,

-100

1

-100

1

Figure 8. Matrix effects in the alr-cooled plasma (d,”= 13.5 mm) (A) and In the water-cooled plasma (B): X, NO,-; A,Ca; 0,PO:-; A, Na. Operating conditions are glven in Table 11.

Table V. Matrix Effect, M (%), Caused by 10000 mg/L of Sodium analysis time

Zn I(213.9 nm) Mn I1 (257.6nm) Cu I (324.7nm) Mn I (403.0nm)

air cooled 13.5 mm

+loo +17 +90

+150

water cooled

-30 -55 -65 -55

air cooled 16 mm -5

+20 -10

+50

volatilization interferences (NO3-, Pod3-) are of minor importance. Also, especially for the air-cooled plasma, the hard line is less matrix-sensitive than the soft line. However, the interferences by calcium and sodium at the 10000 mg/L level are substantial, and in this respect the water-cooled plasma seems to perform better than the 13.5-mm air-cooled plasma. The incident power has a substantial influence on matrix effects in the ICP (25). In the air-cooled plasma the incident power can be increased from 270 W to 400 W by changing the 13.5-mm torch diameter to 16 mm. The improvement is clear in Figure 9, where matrix effects in the 13.5-mm air-cooled, the water-cooled, and the 16-mm air-cooled plasma are shown for the most matrix-sensitive line, Mn 1403.0 nm. For the 13.5-mm air-cooled and the water-cooled plasma the conclusions drawn in connection to Figure 8 are confiied. However, in the 16-mm air-cooled plasma, matrix effects are much smaller. Up to lo00 mg/L all effects are negligible and a t the 10000 mg/L level only sodium causes a significant enhancement, The excellent matrix behavior of the 16-mm air-cooled plasma is confiied in Table V, where the influence of 10000 mg/L Na is presented for four transitions. The 16-mm air-cooled torch shows the best results and performs similar to a conventional ICP (24-27). The observation height has a substantial influence on matrix effects in conventional ICP’s (26). Because in the externally cooled plasmas no significant loss in detection power is expected from a change in OBH, variation of the OBH is attractive. Figure 10 shows the influence of the OBH upon the matrix effect caused by 10 000 mg/L Na. A substantial in-

ANALYTICAL CHEMISTRY, VOL. 56, NO. 13, NOVEMBER 1984

because of the corrosive effect of acids on the stainless steel nebulizer applied.

7

100

B

-I.mg/L

-100

c

low or 1IM,”

C

P

50

0

-100 -50

-50

t

CONCLUSIONS With the aid of external cooling the total argon consumption (F, Fpin Table 11) in ICP analysis can be reduced from 10 to 20 L/min for a conventional torch to typically 1 L/min, without sacrificing the favorable analytical properties of the ICP. This value is lower than has been reported for other torches designed for the purpose of low argon consumption (4-12). This primary aim of our investigation has been achieved by applying either air or water as a coolant medium, and in both cases a two-tube torch is utilized instead of the conventional three-tube arrangement. To operate an aircooled plasma the rf generator need supply only 0.5 kW, while this figure is 1 kW for the water-cooled plasma (3). It has been shown that the good analytical performance of the ICP can be maintainted with externally cooled torches. The short-term precision, the long-term stability, and the dynamic range of the calibration curves observed for air-cooled and water-cooled plasmas are just as good as in conventional ICP’s. With respect to detection power and matrix effects there are minor differences between the torch designs studied and it seems that the 16-mm air-cooled torch yields the best overall performance. Registry No. Argon, 7440-37-1.

+

0

I

2335

-100

-I,mg/L

LITERATURE CITED

Figure 9. Matrix effects for Mn I 403 nm in the alr-cooled plasma (dh

= 13.5 mm) (A), the water-cooled plasma (B), and the air-cooled plasma ( d , = 16 mm), (C): X, NO3-; A,Ca; 0, PO:-; A Na. Operating condltlons are given in Table 11. M ,%

o(b,

-OBH,mm

Flgure 10. Influence of the observatlon height on the matrix effect of 10 000 mg/L of sodium in the aircooled plasma (dh = 16 mm): X, Mn I 403.0 nm; A, Cu I 324.7 nm; 0, Mn I1 257.6 nm; 0, Zn I 213.9 nm. Operating conditions are given In Table 11.

fluence is observed only for the Mn I line, which itself is analytically uninteresting, but may be indicative for other soft lines. The two hard transitions (Zn I and Mn 11) show very little variation with the OBH. The behavior of the Cu I line is intermediary. As a further check on accuracy we determined manganese in a standard steel sample (BCS no. 230) using the Mn I1 line. One gram of sample was dissolved in a concentrated HC1/ HNOBmixture and diluted to a concentration of 250 mg of steel/L. The determination was based on calibration curves in distilled water. The results found of 0.51 f 0.02% Mn (w/w) for the 16-mm air-cooled torch and 0.46 f 0.02% for the water-cooled plasma, are close to the certified value of 0.50% Mn (w/w). More elaborate testa could not be pursued,

Ripson, P. A. M.; de W a n , L.; de Ruiter, J. W. Spectrochim. Acta, Part B 1982, 378, 733. Kornblum, Q. R.; van der Waa, W.; de Qlan, L. Anal. Chem. 1979, 5 1 , 2378. Ripson, P. A. M.; de Qlan, L. Spectrochlm. Acta, Part B 1983, 388, 707. Savage, R. N.; Hleftje, 0. M. Anal. Chem. 1979, 5 1 , 408. Allemand, C. D.; Barnes, R. M.; Wohlers, C. C. Anal. Chem. 1979, 5 1 , 2392. Welss, A. D.; Savage, R. N.; Hleftje, Q. M. Anal. Chim. Acta 1981, 124, 245. Kawaguchi, H.; Ito, T.; Rubl, S.; Mlzuike, A. Anal. Chem. 1980, 52, 2440. Lowe, M. D. Appl. Specfrosc. 1981. 3 5 , 126. Brltske, M. E.; Sukach, J. S.; Fillmonov, L. N. Zh. Prikl. Spectrosc. 1978, 25. 5 . Kawaguchi, H.; Tanaka, T.; Mlura, S.; Xu, J.; Mizuike, A. Specfrochim. Acta, Part B 1983, 388, 1319. Savage, R. N.; Hieftje, Q. M. Anal. Chem. 1980, 52, 1267. Rezaaiyaan, R.; Hleftje. Q. M.; Anderson, H.; Kaiser, H.; Meddlngs, B. Appl. Specfrosc. 1982, 3 6 , 627. Boumans, P. W. J. M.; de Boer, F. J. Spectrochim. Acta, Part B 1972, 278. 391. Green, J. R.; Margerlson, D. “Statlstical Treatment of Experimental Data”; Elsevier: Amsterdam, 1977. Ripson, P. A. M.; de Galan, L. Spectrochlm. Acta, Part 8 1981, 368. 71. Blades, M. W.; Horlick, 0. Specfrochim. Acta, Part B 1981, 368, 661. Robin, J. P.; Mermet, J. M.; Abdallah, M. H.; Batai, A.; Trassy, C. Proceedlngs XXII CSI, Pergamon: Oxford, 1982; p 75. Boorn, A. W.; Browner, R. F. Anal. Chem. 1982, 5 4 , 1402. Winge, R. K.; Petterson, V. J.; Fassei, V. A. Appl. Specfrosc. 1979, 3 3 , 206. Boumans, P. W. J. M.; de Boer, F. J. Specfrochim. Acta, Part 8 1977, 328, 365. Myers, S. A.; Tracy, D. H. Spectrochim. Acta, Part B 1983, 388, 1227. Broekaert, J. A. C.; Wopenka, 6.; Puzbaum, H. Anal. Chem. 1982. 5 4 , 2175. Boumans, P. W. J. M.; de Boer, F. J. ICP Inf. Newsletf. 1977, 3 , 228. Boumans, P. W. J. M.; Lux-Steiner, M. Ch. Spectrochlm. Acta, Part B 1982, 378, 97. Abdallah, M. H.; Mermet, J. M.; Trassy, C. Anal. Chlm. Acta 1976, 87. 329.

(26) Laisson, 0. F.; Fassel, V. A.; Scott, R. H.; Knlseley, R. N. Anal. Chem. 1975, 47, 238. (27) Maessen, F. J. M. J.; Balke, J.; de Boer, J. L. M. Spectrochlm. Acta, Part B 1982, 378, 517.

Received for review September 19,1983. Accepted May 25, 1984.