further useful information could be obtained using different filament materials, thereby evaluating the importance of surface effects and particularly the role played by the filament material in the reduction of metal oxides. The nature of the shield gas also warrants further consideration although its effect on other parameters such as quenching and cooling of the filament would have to be accounted for. Recently ( 4 ) , a method has been devised, whereby it is possible to measure by spectroscopic means, the temperature of atoms in the atomic cloud above the filament. Although the number of elements to which this method can be applied is few, by combining this technique with the temperature measurements on the rod, it should be possible to obtain a complete thermal history of an analyte species. This would be of great value in diagnosing signal profiles, particularly in cases such as Ge where vapor phase reactions are mainly responsible for the atomization process (8).
LITERATURE CITED T. S.West and X. K. Williams, Anal. Chim. Acta, 45, 27 (1969). B. V. L'vov. Spectrochim. Acta, Part B, 24, 53 (1969). H. Massman, Spectrochim. Acta, Parts, 23, 215 (1968). M. P. Bratzel and C. L. Chakrabarti, Anal. Chim. Acta, 63, 1 (1973). G. Torsi and G. Tessari, Anal. Chem., 45, 1812 (1973). B. V. L'vov, "Atomic Absorption Spectrochemical Analysis", Adam HiC ger, London, 1970. E. A. Moelwyn-Hughes, "Physical Chemistry", 2nd revised ed., Pergamon, Oxford, 1965. D. J. Johnson, T. S.West, and R. M. Dagnall, Anal. Chim. Acta, 67, 79 (1973). R. G.Anderson, I. S. Maines. and T. S. West, Anal. Chim. Acta, 51, 355 (1970). R. M. Dagnall, D. J. Johnson, and T. S. West, Anal. Chim. Acta, 66, 171 (1973). D. J. Johnson, Ph.D. Thesis, University of London, 1973.
RECEIVEDfor review December 2, 1974. Accepted February 28, 1975. We thank Imperial Chemical Industries Ltd., (Agricultural Division), Billingham, Teesside, for supplying financial support for D.J.J.
Atomization in Graphite-Furnace Atomic Absorption Spectrometry-Peak-Height Method vs. Integration Method of Measuring Absorbance: Carbon Rod Atomizer 63 R. E. Sturgeon, C. L. Chakrabarti,' 1. S Maines, and P. C. Bertels Department of Chemistry, Carleton University, Ofta wa, Ontario, K1S 586,Canada
Oscilloscopic traces of transient atomic absorption signals generated during continuous heating of a Carbon Rod Atomizer model 63 show features which are characteristic of the element being atomized. This research was undertaken to determine the significance and usefulness of the two analytically significant parameters, absorbance maxlmum and integrated absorbance. For measuring Integrated absorbance, an electronic integrating control unlt consisting of a timing circuit, a lock-In amplifier, and a dlgltal voltmeter, which functions as a direct absorbance X second readout, has been designed, developed, and successfully tested. Oscilloscopic and recorder traces of the absorbance maxlmum and digital display of the integrated absorbance are simultaneously obtained. For the elements studied, Cd, Zn, Cu, AI, Sn, Mo, and V, the detection limits and the preclslon obtained are practically identical for both methods of measurement. The sensitivities by the integration method are about the same as, or less than, those obtained by the peak-height method, whereas the calibration curves by the former are generally linear over wider ranges of concentrations.
Oscilloscopic traces of transient atomic absorption pulses can be characterized by at least two quantities, namely, the peak absorbance, N p e a k , corresponding to the peak of the pulse and the integrated absorbance, QN,representing the pulse area. The integrated absorbance is obtained by summing the absorbance values over the entire period of time during which free atoms are present in the optical path (the analysis volume). The most widely used methods of 1 Author
1240
to whom all correspondence should be addressed.
ANALYTICAL CHEMISTRY, VOL. 47,
NO. 8,
JULY 1975
measuring signals in atomic absorption spectrometry are the measurement of an equilibrium absorbance in the flame technique ( I ) and a peak absorbance in the nonflame technique (2).An alternative method of measuring absorbance signals in either of these techniques is to measure the integrated absorbance. As far back as 1965, a method of integrating absorbance pulses with samples vaporized from a graphite cuvette was reported by Massmann ( 3 ) .The theoretical advantages of this method of measurement were clearly shown by L'vov ( 4 ) and L'vov et al. ( 5 ) . Because of the severe limitations of the peak method, i.e., of measuring peak absorption with sample vaporization in graphite atomizers, leading to loss in sensitivity and to systematic errors, the possibilities offered by the integration method have been explored by several groups of workers. L'vov and Plyushch (6) found that the integration method, employing a technique of sample evaporation from a tungsten wire heated with an air-acetylene flame, was applicable only to solutions of highly volatile elements and not to solid samples. In a later publication, Borzov et al. (7) used the integration method and vaporized both liquid samples (solutions) and solid samples into a flame by means of a graphite furnace independently heated by an alternating current. In all cases, a graphical method of integration (cutting and weighing the recorder chart tracings) was used. An attempt to use an integrating circuit for atomic absorption measurements with an arc atomizer was made by Belyaev et al. (8).These authors employed accumulation of the signal expressed in terms of an intensity difference, not of absorbance values and, because of this, their method of measurement was less satisfactory. Recently, Pickford and Rossi (9) reported signal integration with a nonresonance line correction system as applied to a graphite-furnace atomizer and commented on the im-
proved precision obtained. Minor increases in precision were reported also by McIntyre et al. (10). Schramel (11) reported increases in the linear ranges of the calibration curves drawn with integrated signals over those drawn with peak signals. Despite the expected theoretical advantages (4, 5 ) in using an integration method, the only applications reported to date are those mentioned above. Unfortunately, apart from the work reported by L’vov and coworkers, data reported by other workers for the integrated signals are in arbitrary units only and, therefore, are not useful for quantitative comparisons (9-12). This paper critically examines the results of a preliminary investigation into the potential of an electronic integration method with a direct absorbance X second readout vis-a-vis the peak method of measuring atomic absorption signals generated by a modified Carbon Rod Atomizer 63 (CRA).
THEORY L’vov (13, 1 4 ) and Katskov and L’vov ( 1 5 ) have considered a simplified mathematical model for the transport of a sample through the analytical cell and have obtained expressions which describe the kinetics of change in the number of atoms in the analysis volume. Using this simplified model for sample transport, these authors have shown that correct choice of the recording method and proper evaluation of the parameters of the recording device enable one to improve both precision and accuracy. The following assumptions are made in constructing the mathematical model: the element to be determined is completely atomized, all of the atoms of the element enter the analysis volume, and the removal of atomic vapor from the analysis volume is determined solely by diffusion under a concentration gradient, i.e., exponential. The following notations are introduced. No is the number of atoms of an element in the sample, N is the total number of atoms entering the analysis volume at any instant t , 7 1 is the atomization time, 7 2 is the mean length of time spent by an atom in the analysis volume, Le., the residence time, and 73 is the length of time during which the signal is recorded. 7 2 is defined as the time taken for the absorbance (atom population) to decrease from its peak value, A,,,, to a value e times smaller, based on measurements taken from the “tail” of the pulse. L’vov et al. ( 5 ) determined 7 2 from the “tail” of the pulse by measuring the time for the absorbance to decrease from its peak value to a value e2 times smaller (2~2)-both the definitions give the same value for 72.
Peak Method. At the moment the sample is completely atomized, ( t = 7 1 ) , N reaches its peak value Npeak: (See Appendix)
If 7 1 / 7 2 1, 472.
The sensitivity of the peak method is highest when the condition 7 2 / 7 1 >> 1 is satisfied. In the integration method, provided 7 3 1 7 1 4 7 2 , the following relation can be shown to be valid within an accuracy of 2 %
+
or when 7 2 / ~ 1>> 1. Thus, the magnitude of 7 2 governs the attainment of high or low sensitivity, and the sensitivity of the integration method is related to that of the peak method by the factor 72.
EXPERIMENTAL Apparatus. A Varian-Techtron atomic absorption spectrophotometer, model AA-5, fitted with a Carbon Rod Atomizer, model 63 (CRA 63) was used throughout. Modifications made to the power supply of the CRA 63 to obtain a faster rate of heating and a higher final temperature (ca. 3800 K) have been described in a previous paper from this laboratory (16). Signals were detected with a Hamamatsu R446-NR photomultiplier tube which was enclosed within a magnetically and electrostatically shielded housing to reduce interferences. All temperatures of the CRA 63 tube were measured with an automatic optical pyrometer (Ircon Inc., Niles, IL, series 1100) calibrated by the manufacturer and rechecked by measurements with thermocouples and by the melting points of selected pure metals to cover the entire range of temperature. Single-element hollow-cathode lamps were used as narrow line sources. Amplifier-Integrator Assembly. For measuring integrated absorbance, an electronic integrating control unit consisting of a timing circuit, a lock-in amplifier, and an integrating digital voltmeter (DVM) which functions as a direct absorbance X second readout has been designed, developed, and tested in the present study. A simplified block diagram of the amplifier integration assembly is presented in Figure 1. The operation of this unit is as follows. The signal from the photomultiplier tube is fed through a pre-amplifier into a lock-in amplifier, the reference signal to which ANALYTICALCHEMISTRY, VOL. 47, NO. 8, JULY 1975
1241
Table I. Instrumental Settings PHOTO-MULTISTRIP CHART RECORDER REFERENCE
I SIGNAL
I 1 ;J-Lr
1 OSCILLOS-
SUPPLY
CRA POWER SUPPLY
INTEGRATION CONTROLLER PERFORMS FOLLOWING FUNCTIONS
TRIGGER
I. CONTROLS DELAY AN0 INTEQRA*TION TIMES 2.CORRECTI FOR DRIFT 3. INTLORATES THE ABSOROANCL
Figure 1. Amplifier-integrator assembly
is the same 285 Hz trigger voltage used to modulate the hollow cathode lamps. The amplified signal is then passed through a logarithmic converter. The resulting absorbance signals are then fed into an integrating digital voltmeter (Dymec, model 2401C) where they are summed over a preset time interval and displayed. In its basic form, the digital voltmeter consists of a voltage-to-frequency converter and an up/down counter. The input signal voltage is converted into a series of pulses whose frequency is proportional to the magnitude of the signal voltage. The pulses are then added in the counter stage, thus giving a total number of counts relative to the area under the absorbance curve. The integrated output is displayed simultaneously with the peak absorbance reading on a fastresponse (250 msec) strip chart recorder (Honeywell Elektronik 194) as well as an oscilloscopic display of the absorbance pulse (Tektronix, model 549). As has been done by L’vov (13, 1 4 ) , the values of integrated absorbance are expressed in absorbance X seconds, and the absolute sensitivity in terms of 0.0044 absorbance in one second, Le., 0.0044 absorbance X seconds. The integrated output from the DVM is in mV-second; with this unit, an absorbance of 1.0000 corresponds to a 1000.0-mV output from the log converter. Thus, an absorbance of 0.1000 for a period of 1.0000 second corresponds to 100.00 mVsec. The correctness of this measurement was verified by the graphical method (by cutting and weighing a photograph of an oscilloscope trace of the pulse area), and by electronically integrating a simulated square-wave absorbance pulse of known value for exactly one second. The results obtained in this manner agreed to within f3% of the mV-sec digital display. Reagents. All chemicals used were ACS grade or of the highest purity commercially available. The stock solutions of metal standards were prepared as follows for each metal separately and contained 1000 pg/ml of that metal only. Solutions of Cu, Sn, Al, and Zn were prepared from the metals, Cd from cadmium carbonate, and Mo and V from molybdenum trioxide and vanadium pentoxide, respectively. The above metals or their compounds were dissolved in pure acids or bases where required and diluted with ultrapure water, obtained direct from a Milli-Q water system (Millipore Corporation). All test solutions were prepared immediately prior to their use by diluting with ultrapure water. The special laboratory where this study was done was clean enough to ensure very small and reproducible blanks. Gases. Hydrogen gas of 99.995% purity and nitrogen gas of 99.95% purity were used (Roncar Oxygen Co.). Procedure. Test solutions were injected into the graphite tube with an Eppendorf syringe fitted with disposable plastic tips. Sample volumes of 5 fil were used. A blank was run with every test solution, and its value was subtracted from the gross value to get the net value, which was reported. Nitrogen gas at a flow rate of 4.8 l./min was used to sheath the CRA 63. The effect of a hydrogen diffusion flame on the sensitivity was also studied and found to be important in the determination of aluminum and tin. In all determinations, an optical stop with internal diameter of 3 mm was inserted between the graphite tube and the condensing lens to minimize emission from the sides of the incandescent tube. The wavelengths, spectral bandpasses, and lamp currents used in this study are presented in Table I. Integration Control Unit. The sequence of operations for the integration control unit of the amplifier-integrator assembly (Fig1242
ANALYTICAL CHEMISTRY, VOL. 47, NO. 8, JULY 1975
Element and wavelength, nm
Lamp c m e n t ,
Cu 324.7 Cd 228.8 Zn 213.9 Al 309.3 Sn 224.6 Mo313.3 V 318.5
9 8
mA
7 6 8
15 15
Spectral bandpass, nm
0.082 0.082 0.082 0.082 0.165 0.033 0.033
ure 1) is described below. The initiation and duration of the integration period is controlled by an automatic timing circuit which is incorporated into the integration control unit. The timing circuit is triggered by a pulse from the CRA 63 power supply. First, the entire absorption pulse is displayed on the storage oscilloscope. The time parameters which require optimization are shown in Figure 2. It should be noted that no visible distortion occurred in the absorption pulse as a result of its passage through the log converter. All reported pulse characterization times are identical with those obtained by feeding the unmodulated dc pulse directly from the photomultipler tube to the oscilloscope. This proved that the electronics had a sufficiently fast response time so that even the absorption pulses of the most volatile element examined (Cd) could be reliably recorded. Second, the Td&y is measured from the oscilloscopic trace. This value is then manually introduced into the delay cycle of the timing circuit of the integration control unit. The duration of the absorption pulse is then measured, after which it is manually introduced into the integration cycle of the timing circuit. Both these times are selected so that integration over unwanted base-line noise is avoided. The times can be set accurately and reproducibly to within fl msec. In this study, integration has been extended over a period of a few msec before and after the pulse occurs (i.e., between points A and B on Figure 2) to ensure that integration is carried out over the entire pulse width in all cases. Thus, possible uncontrollable temperature variations in the graphite tube during heating cycles which would lead to errors in the pulse characterization times are avoided. Except where otherwise stated, all peak absorbance recordings on the strip chart recorder were done using a system time constant of 250 msec; all integrated measurements and oscilloscopic peak recordings were made with a system time constant of 18 msec. Atomization Voltage. Atomization voltages and times were so adjusted as to give maximum values for the peak and the integrated absorbance. In all cases, atomization was continued over a time which extended beyond the duration of the absorbance signal, as can be seen from the continuous rise in the temperature of the inner surface of the graphite tube shown in Figure 2.
RESULTS AND DISCUSSION Effect of Atomization Voltage. The effect of atomization voltage on the peak and integrated absorbance is presented in T a b l e 11. For the elements shown, t h e peak absorbance decreased with decreasing atomization voltage whereas the integrated absorbance remained essentially constant for fairly volatile elements (Zn, Cd) and Cu but decreased with decreasing atomization voltage for the med i u m volatility elements (A1 a n d Sn). T h e effect on Mo and V could n o t be studied as incomplete atomization resulted for voltages below a setting of 9, thus leaving very little room for varying t h e voltage setting upwards, (the highest setting available on t h e CRA 63 power supply is 10). The above mentioned effect is explained as follows. In the peak method, with the modified power supply, i t has been found that 71/72 1 1 a n d in some cases < 1, a s shown in Table 111. It can b e seen from Equation 1,that t h e smalle r the ratio 7 1 / 7 2 , the more closely t h e value of N p e & approaches No. Since t h e electrical power (i.e., the r a t e of heating) = E 2 / R (where E = voltage a n d R = resistance), t h e r a t e of heating increases rapidly with increased voltage a n d , hence, t h e r a t e of atomization increases rapidly (i.e., 71 decreases rapidly). T h e ratio 71/72 therefore becomes progressively smaller with increasing voltage a n d results in
Table 11. Effect of Atomization on P e a k and Integrated Absorbance Absorbance Voltage setting
Element
9 8 7 9 8 7 9
Cd
Zn cu
8 7 9 8 7 9 8 7
A1 Figure 2.
Pulse characterization times on an oscilloscopic trace
7my= The time elapsed from the start of the atomization cycle to the first appearance of a signal: T p k = The time from the start of the atomization cycle to the peak (maximum absorbance, Amx): T~,,, = The time from the start of the atomization cycle to the return of the pulse to the base line: 7d$phy = ( ~ ~ n d7 . 4 i.e., the total duration of the pulse; 7, is the atomization time = ( r p e a k 7*hy). 7 p is the residence time
-
Sn
Peak
Integrated absorbance x seconds
0.304 0.257 0.163 0.140 0.125 0.115 0.040 0.032 0.021 0.102 0.056 0.032 0.071 0.053 0.026
0.160 0.157 0.162 0.042 0.039 0.043 0.007 0.007 0.006 0.023 0.014 0.008 0.023 0.014 0.008
-
higher peak absorbances for all the elements studied. Up to this point, no consideration has been given to the factor 72, it being taken as a constant in the above discussion. In reality, it too is temperature-dependent as can be seen from the following equation (13, pp 204-220). Tq =
Table 111. Values of T I and T Z and Their Ratios for Elementsa Element
Cd
Zn CU
10312~ nrm
A1 Sn Mo
“01
where 1 = length of the graphite tube, P = pressure of the inert gas inside the graphite tube, Do is the diffusion coefficient of the analyte atoms a t S.T.P., and the value of n varies between 1.5 and 2 for different combinations of gases. The CRA 63 is an open system and because of the short length of its graphite tube, loss of analyte atoms from the inside of the graphite tube occurs both by diffusion (due to a concentration gradient) and by convection (due to a temperature gradient). The effect of increasing rate of heating is to make 71 progressively much smaller without substantially increasing 72, thereby making the 71/72 progressively smaller, and hence higher values of peak absorbance. Therefore, the highest rate of heating (i.e., the highest atomization voltage setting) gives the highest peak absorbance for all the elements studied. The results of the integration method may also be explained in a similar manner. From Equation 7 , it can be seen that the integrated absorbance is equal to the product of the peak absorbance and the residence time. As the atomization voltage is increased, 71 decreases but, as the graphite tube is heated at a faster rate, the temperature of the tube is also increased, which makes 7 2 smaller (Equation 8). Only the diffusional loss is accounted for by Equation 8-the loss due to convection makes 7 2 smaller still. For the fairly volatile elements (Cd, Zn) and Cu, the increase in peak height is compensated for by a decrease in the residence time. These two factors oppose each other, resulting in constant values for the integrated absorbance as the atomization voltage is varied. For elements of medium volatility (Sn, Al), it can be seen from the results in Table I1 that, as the atomization voltage is decreased, the increase in 7 2 is not large enough to compensate for the decreased peak absorbance. As a result, the integrated absorbance is not constant with respect to the atomization voltage. Sn and A1 show the atomization characteristics of elements of relatively low volatility such as Mo and V, because Sn and Al, during atomization, form intermediate
a
V Obtained
‘1, msec
r 2 , msec
110 230 400 340 160 550 450 under the atomization
Tl’T
2
120 0.92 80 2.90 320 1.25 130 2.60 105 1.50 500 1.10 725 0.62 conditions shown in Table
IV.
species of significantly lower volatility than the pure metals. Evidence of this is presented in Figures 3 and 4. The presence of a hydrogen diffusion flame showed no effect on the determination of Cu, Cd, Zn, Mo, and V but showed large effect on the determination of Sn and Al. Maximum absorbance for Sn and A1 by both methods occurred with hydrogen flow rates of 0.5 and 2.0 l./min, respectively. The increased absorbance, a result of an increase in the atomization efficiency, is probably due to a change in the chemical environment within the graphite tube. There has been much evidence that Sn and A1 are susceptible to oxide formation during atomization ( I 7-19). The hydrogen reducing flame provides an increasingly reducing environment, particularly in the gas phase where the solid, incandescent carbon of the heated graphite tube is less effective as a reducing agent. The presence of the hydrogen reducing flame creates a more favorable balance of metal over the metal oxide(s). In the absence of a hydrogen reducing flame, the oscilloscopic traces of the absorption pulse of Sn and A1 (figures not presented) showed multiple peaks, indicative of the formation of more than one intermediate species during the atomization. Further evidence substantiating these conclusions will be presented later in this paper. Absorption Pulse Characteristics. The oscilloscopic traces obtained for atomization of the elements under optimum experimental conditions are presented in Figures 5-11. The time coordinate in these figures begins after an elapsed time, called delay time, Tdelay, and therefore corresponds to the pulse characteristics between the points A and B of Figure 2. Taking into account the different time scales for different elements, some generalizations can be ANALYTICALCHEMISTRY, VOL. 47, NO. 8. JULY 1975
1243
-0.200
B
In x
mvi
4 8 2
-0.1so
3a
z m a
-0.00
n
W
k
0.1
h
I
0
L W om05
I
Figure 6. Oscilloscopic trace showing absorbance by 5.0 X zinc
g
Vertical scale: absorbance, O.l/scale unit. Horizontal scale: sweep speed, 100 msedsdle unit 0 W v)
-0.14
I
vi
P
-0.13
8 z a
iq/
-0.12
2m
0.20
l l , , , l j t
0.10
0
1
9
3
4
I
6
0.10
I
HYDROGEN FLOW RATE, L l M l N
Figure 4. Effect of hydrogen diffusion flame on absorbance by Sn
Figure 7. Oscilloscopic trace showing absorbance by copper
(0) Peak absorbance; (A)Integrated absorbance
2.5 X
g
Vertical scale: absorbance, O.P/scale unit. Horizontal scale: sweep speed, 200 mseclscale unit
Figure 5. Oscilloscopic trace showing absorbance by 5.0 X cadmium
g
Vertical scale: absorbance, 0. llscale unit. Horizontal scale: sweep speed, 50 mseclscale unii
drawn from these figures. Absorption pulses of fairly volatile elements are characterized by sharper and narrower curves having higher peak absorbance values than the elements of medium or low volatility or those elements which form compounds of low volatility in graphite furnaces, e.g., Mo and V, which form refractory carbides with incandescent carbon of the heated graphite tube. The pulse characterization times for each oscilloscopic trace are presented in Table IV, which also shows the optimum atomization conditions under which the oscilloscopic traces were obtained, and a comparison of the peak and the integrated absorbance values. The following conclusions can be drawn from Table IV with respect to the elements studied. For elements of high volatility, peak absorbances obtained from the oscilloscopic traces are considerably greater than those given by the 1244
ANALYTICAL CHEMISTRY, VOL. 47, NO. 8, JULY 1975
Flgure 8. Oscilloscopic trace showing absorbance by aluminum
1.25 X lo-’
g
Vertical scale: absorbance, O.P/scale unit. Horizontal scale: sweep speed, 100 msec/scale unit
strip chart recorder. This can be explained by the fact that slight distortion of the curves drawn by the recorder pen results from purely mechanical limitations of the recording device and the finite speed of travel of the pen. The distortion in peak amplitude recordings with a recorder pen must increase as the peaks themselves increase, but the net result fs also dependent upon the value of 71 (Le., speak 7delay). Pen-recorded peaks of every element, therefore, suffer variable degrees of distortion, a8 each element is characterized by a different value of 71. Large values of 71 (e.g., Mo and V), result in the absorbance values rising more slowly with time. As a result, the comparatively slow-response pen recorder is able to follow such pulse traces more accurately. Generally, the largest differences in the oscillo-
Table IV. Pulse Characterization Times under Optimum Conditions A tom i La tion conditions0 Voltage setting, arbitrarv
Characterization times, msec
Time.
‘peak
end
1000 90 9 200 490 1000 250 9 480 680 780 1600 2000 380 cu 9 1220 1460 A1 9 1500 880 1500 420 580 780 Sn 9 2600 1100 Mo 9.5 1650 3000 850 1300 2750 9.5 2000 V For both the peak and the integration methods. Cd
Zn
Figure 9.
Oscilloscopic trace showing absorbance by 7.5 X lo-’’ g
tin
‘display
400 430 1220 580 360 1900 1900
Vertical scale: absorbance, O.P/scale unit. Horizontal scale: sweep speed, 100 msec/scale unit
Table V. Effect of Analyte Mass on 71, Tpeak, and 1 2 Weight of Cu, ng
0.25 0.50 0.75 1.oo 1.25
Figure 10.
Oscilloscopic trace showing absorbance by 2.75 X
lo-’’ g molybdenum Vertical scale: absorbance, O.l/scale unit. Horizontal scale: sweep speed, 500 msec/scale unit
Figure 11. Oscilloscopic trace
vanadium
showing absorbance by 3.0 X
g
Vertical scale: absorbance, O.l/scale unit. Horizontal scale: sweep speed, 500 msec/scale unit
scopic and chart recorded peaks occur for those elements which are volatilized rather easily (Cd, Zn). The oscilloscopic mode of measuring absorbance does not suffer from the above mechanical limitations as this device has an infinitely fast response compared to that of the pen recorder. However, the reading error of this device is greater than that of the strip chart recorder because of the limited resolution of the oscilloscopic trace. Table IV also indicates that each analyte has a characteristic delay time (Tdelay) and time to peak absorbance (Tpeak). Within the limits of experimental error, the Tpeak values for the elements were independent of the weight of the element present. As an example, the Tpeak values for different weights of copper are presented in Table V. From
TI,
msec
400 420 410 390 400
T peak,
msec
780 780 740 770 780
T2,
msec
320 320 330 320 340
T1r2
1.25 1.32 1.25 1.22 1.18
the data in Table IV, the relative volatility of the analytes, based on the magnitude of their respective Tpeak values, may be predicted. The results, in the order of decreasing volatility (increasing Tpeak) are Cd < Zn < Sn < c u < A1 < V < Mo. This order of decreasing volatility parallels the order of increasing melting and boiling points except for the anomalous positions of A1 and Cu: Cd < Zn < Sn < A1 < c u < V < Mo. This correlation is expected since Tp& values obtained under the same atomization voltage setting of 9 (except for Mo and V where the minimum setting required for atomization was 9.5) reflect the increasing refractory nature of the elements. The reversal of the order for A1 noted above can be rationalized on the basis of evidence that atomization of A1 proceeds via formation of an intermediate compound or compounds having significantly different volatilities than that of A1 metal. The melting points, boiling points, and ionization energies of the elements studied are presented in Table VI. In addition, the temperatures of the inner surface of the graphite tube are tabulated a t all characterization times. These temperatures may be read off the time-temperature curve of Figure 12 (measured with the automatic optical pyrometer) and are accurate to within f5%. Most probably, these temperatures do not equal the temperatures of the atomic vapor but those at the delay time (Tdelay) are probably fairly close to the analyte temperature as the analyte is most likely still in intimate contact with the graphite surface at this point in time. In general, the temperatures at 7delay are basically of the same order of magnitude as the melting points for the elements, except for the large discrepancy evident in the cases of Sn and AI. The higher temperatures a t 7delay (Le., the temperature required for the. first appearance of the Sn and A1 signals), approximately 1000 K above the melting points, further support the hypothesis that, during atomization, these elements form intermediate compounds or at least pass through intermediate species of significantly lower vapor pressures than those of the pure elements. This hypothesis is in general agreement with the results presented by Aggett and Sprott (20). The physical significance of Tdelay is that it is a measure ANALYTICALCHEMISTRY, VOL. 47, NO. 8, JULY 1975
1245
Table VI. Pulse Characterization Temperatures Temperature, K Element
I
$9:f"$
594 693 1356 933 505 2890 2173
Cd
Zn
cu A1 Sn Mo V
1038 1180 2840 2740 2543 4885 3653
8.99 9.39 7.72 5.98 7.34 7.10 6.74
'delay
'peak
'end
730 1060 1310 2190 1390 2470 21 50
970 1510 2050 2610 1690 3010 2690
1570 1870 2960 2850 2030 3650 2580
~
Table VII. Absolute Sensitivity=
J
60 200
Mode of Measurement
400
800
1200
1600
2000
2400
2800
3200 3600
TIME, msec.
Element
Peak (oscilloscope)
Peak (recorder)
Integration
Cd Zn
9.5 X 1.6 X lomi3 5.2 x 8.4 x 1.1 x 6.3 x Cu 2.7 x lo-'' 3.7 x lo-'' 6.7 x lo-'' A1 1.3 x lo-" 1.6 x lo-" 4.8 x lo-'' Sn 9.1 X lo-'' 1.1 x lo-'' 4.8 x Mo 7.5 x lo-'' 7.9 x lo-'' 7.7 x lo-'' V 3.1 x lo-" 4.5 x lo-'' 5.5 x lo-" Defined as the weight in grams of element which gives a peak absorbance of 0.0044 or an integrated absorbance of 0.0044 absorbance X second.
of the "induction time" which is the time required for the atomizer to reach a sufficiently high temperature for accelerated atomization to take place. L'vov (13, p 202) has estimated that, in order for a sample to be atomized rapidly, it is necessary for a temperature to be reached at which the saturated vapor pressure is greater than 0.1 Torr. This estimate was made under the conditions of one atmosphere pressure, a temperature of 3500 K, a surface area of the analyte deposited on the graphite surface -0.01 cm2, and an atomic weight of 50 for the element. There is, thus, a minimum temperature required before atomization may proceed rapidly. The magnitude of the delay time is dependent upon the atomization voltage. Measurement of temperature by the optical pyrometer has indicated that the larger the atomization voltage, the faster the rate of rise of temperature of the atomizer. Higher voltage settings therefore result in lower delay times as a result of the shorter time required to reach the initiation temperature of the pulse. In each case, for an element, the delay temperature is constant, within the limits of experimental error. The temperatures of the inner surface of the graphite tube which may be a few hundred degrees higher than the vapor temperatures at this point, indicate that relatively low temperatures are required for the atomization of even low-volatility elements such as Mo and V which may also form refractory intermediate compounds such as carbides with incandescent carbon of the heated graphite tube. The full significance of 7 2 values for Mo and V presented in Table I11 can now be seen in the light of the temperatures at 7pe& values for the elements of low volatility (Mo and V) as presented in Table VI. According to Equation 8, the T2 values for the above elements which require higher temperature for atomization should be shorter than those of the volatile elements (Cd and Zn) which require lower temperature for atomization. This prediction is in contradiction with the experimental results presented in Table 111, which shows that 7 2 values are much longer for Mo and V. This contradiction may be explained as follows. It is 1246
ANALYTICAL CHEMISTRY, VOL. 47, NO. 8. JULY 1975
Figure 12. Time-temperature characteristics of CRA 63 obtained with an automatic optical pyrometer
possible to evaluate T Z values from the tail of the absorption pulses provided that vaporization of the element is complete a t the time t = 7peak. For these elements of low volatility, vaporization was not completed at time t = Tpeak, so that the tail of the pulse was related essentially to the diminishing rate of vaporization for these elements (associated with the continuously reducing surface area of the analyte elements or the refractory carbides formed by them), rather than to the residence times for atoms in the analysis volume. The ionization energies are also listed in Table VI. The comparatively low ionization energies of A1 and V coupled with the high temperatures observed at the peak absorbance ( ~ 2 6 0 0K for the cell wall) would suggest that ionization of these elements may take place. However, careful experiments have established that there is no detectable ionization of any of these elements (21). Schuster's (22) studies of spectral excitation by electron impact in a flameless cell indicate that hot incandescent carbon may act as an electron buffer. The comparatively low work function of 4.6 eV for carbon (23) facilitates the thermionic emission of free electrons from the carbon surface a t these high temperatures thus suppressing ionization of the analyte. Sensitivity a n d Residence Times. The absolute sensitivities obtained for the elements by both methods of recording are presented in Table VII. Contrary to expectation, the integrated sensitivity, defined as the weight in grams of element which gives an integrated absorbance of 0.0044 absorbance X seconds, is lower in each case than the peak sensitivity, defined in the usual manner, by factors ranging from 1 to 5 . Only in the case of Mo and V does the integrated sensitivity approach that of the peak sensitivity. The lack of agreement of the peak recorded sensitivities obtained from the recorder trace and the oscilloscopic trace is the result of the slow response of the recorder pen and, hence, damped peak heights, as described earlier. The residence times, defined as the time required for the peak absorbance (A,,, in Figure 2) to decrease to a value e times smaller, based on measurements from the tail of the absorption pulse, are presented in Table VIII. Because of the open nature of the CRA 63 and its rapid temperature rise, large diffusional and convective losses of the analyte vapor occur. The residence time, TZ, is proportional to the square of the length of the absorbing medium (13, p 220) and, as such, the small length of the graphite tube of this atomizer is not conducive to large residence times. All 7 2 values are less than 1 second. As a result, the product N p e &
i.or
Table VIII. Residence Times, msecu
11.00 d W 0.900
0.80
0.800
0 0.70
0.700
m 0.60
0.600
72
Element
v)
120 80 320 130 105 500 725
Cd Zn
cu A1 Sn Mo V
W
z a a
0 v)
m
a Y
a W
P
Defined as the time taken for the peak absorbance to decrease to a value of l i e of the peak absorbance. a
Table IX. Upper Limit for Linear Working Curves, Gram 4 Ratio of columns 31 2
3
2
Peak
Integration
a.
If: z
2
a
8 m
0.50
0.500
0.4 0
0.400
0.30
0.300k
0.20
0.2008
0.10
0.100
; W
0
1 Element
A
0.90
a
I-
10 IS 20 25 WEIGHT OF ZINC x IO” GRAM
z
-
0 30
Figure 13. Working curve for zinc
(0) Peak absorbance: (A)Integrated absorbance
Cd cu
1.5 X 1.0 x 7.5 x 2.3 x 5.0 X 3.0 x 7.5 x
Zn
A1 Sn Mo
V
6.0 X 3.0 x 1.3 X 3.5 x 7.5 x 3.0 x 7.5 x
10-9 10-12 10-9 lo-’ 10-9
~~
4 .O 3 .O 16.7 1.5
lo-’‘ 10-9 lo-‘’ 10-9 10-10 10-9
1.5
1 .o 1 .o
10-9
~
Table X. Detection Limit, Grama Element
Cd
Zn
A1 Sn cu V
Mo
Peak
1.7 x 1.8 x 2.0 x 1.8 x 5.0 x 1.3 X 3.2 x
10-13 10-13 10-1’
lo-” lo-‘‘ lomii
Integration
3.0 x 2.7 x 2.0 x 3.1 x 1.3 x 7.7 x 1.5 X
1043
10-13 10-11 10-11 10-l‘ 10-12 10‘”
Defined as the weight of the analyte in grams that gives a signal equal to twice the standard deviation of 10 replicate measurements with the same solution containing a very low concentration of the analyte. a
T Z = QN, results in the integrated sensitivities being less than the peak sensitivities, as observed. Only when the T Z values approach one second, as in the case of Mo and V, do the integrated and the peak sensitivities approach one another. Working Curves. In theory, the integration method can give working curves which are linear over a much larger range of absolute weight of analyte. In the integration method, the rate of sample vaporization can be adjusted such that the absorption values do not exceed a value a t which the absorbance is no longer proportional to the mass of analyte. In this study, the vaporization rate was not adjusted to maintain a low vapor density in the graphite tube. The upper limit for linear working curves is presented in Table IX. The working curves by the integration method are linear over a larger range of absolute weight of analyte; this being most marked for Cd and Zn (a 16.7-fold larger range was obtained for Zn), and to a lesser extent for Cu, Al, and Sn. In addition, beyond the point of strict linearity, the working curve by the integration method has a more gentle curvature than that of the peak method-this can be seen from Figure 13; hence, the workable range is even longer for the integration method. By proper adjustment of the vaporization rates, it may be possible to extend the linear range much further. For Mo and V, the calibration curves by the peak meth-
od did not pass through the origin but intercepted the absorbance axis a t 0.02 and 0.05 absorbance, respectively. I t can be seen from the oscilloscopic traces of Figures 10 and 11 that, for both of these elements, the “blank” firing of an unloaded tube produces an absorption peak due to the scattering by carbon particles. However, this “blank” signal appears a t a difcerent point in time from that of the analytical signal. In addition, although there is a slight contribution from this “blank” a t the analyte absorption peak, the poor resolution does not allow for its accurate measurement. Absorbance values recorded with the strip chart recorder cannot therefore be corrected for blank signals as only the peak absorbances are recorded. This would make the analytical results wrong. This is of considerable practical importance since the peak method is usually used in the analysis of real samples under the assumption that blanks can be corrected for. Since the problem arises from the scattering by carbon particles emitted by the graphite tube a t high temperature, all other elements of relatively low volatility which would require high temperature for volatilization would experience the same problem. In the above cases, integration of the analytical signal followed by correction for the integrated blank signal would give the correct analytical result. Detection Limits. By considering the noise amplitude in terms of the statistical fluctuations in the photocurrent, Katskov and L’vov ( 1 5 ) have shown that the ratio of the signals in the peak and integration methods to the standard deviations due to the noise are equal (within 3%), provided that each is recorded under optimum conditions (Le., that the integral is recorded by a circuit with a time constant TRC = 0 and the peak, provided 71/T2