Difference of the Absorbance of Calcium in the Various Flames.
Resonance absorption has been caused a t shorter and longer wavelengths rather t h a n a t the central wavelength of t h e resonance emission line from t h e lamp 1% he? t h e self-absorption of the emission line was increased, or t h e line was reversed. Therefore, the absorbance indicated approximately the mean value of the absorbances or the absorption coefficient k , on the sides of the resonance absorption line. Consequently, the difference between the curved lines in Figure 11 is mainly due to the asymmetry, the shift and the broadening of the resonance absorption line of calcium in the flame. I n Figure 11, the second derivative of this cuived line in the oxyhydrogen flame is negative, but the others are poqitive. This fact is considered to mean the following. Since the shift of the absorption line in the Oxyhydrogen flame is smaller than that in other flamey, the center of the resonance emision line from the lamp is in the
wavelength region in which the second derivative of the absorption line profile is negative. Accordingly, the second derivative of the curved line in the oxyhydrogen flame has a negative value, but the curved line in the other flames have a positive value so that the absorption line is shifted considerably. Since calcium is affected by a carbon radical which has an energy level close to one in calcium, it can be understood why the shifts in the oxyacetylene, oxypropane flames, and oxyhydrogen flame into which the organic solvent (EtOH-H20) was sprayed, become larger than the shift in the oxyhydrogen flame in which water was sprayed. ACKNOWLEDGMENT
The author expresses his gratitude to ll.Shimazu of Hitachi Central Research Laboratory, K. Xakamura of Hitach Perkin-Elmer, W. E. L. Grossman of Iowa State TJniversity and I. Alakino and H. Okagaki of Hitachi S a k a Works for their helpful suggestions in conducting a series of the present experi-
ments and also to S. llatsudaira, K. Kurita, and K. Cchino for their kind assistance in the experimental work. LITERATURE CITED
( 1 ) David, D. J., Analyst 85, 779 (1960). ( 2 ) Gaydon, A. G., Wolfhard, H. G., “Flames,” p. 243, Chapman and Hall, Ltd., New York, London, 1953. ( 3 ) L’vov, B. V.,. Spectrochim. Acta 17, . 761 (1961). (41 Menzies. A. C.. ANAL.CHEM.32. 898 (1960). ‘ ( 5 ) Mitchell, A. C. G., Zemansky, AI. ’
W., “Resonance Radiation and Excited Atoms,” pp. 169, 31.2, 101, 99, 173, 174, 175, Cambridge University Press, London, 1934. (6) Robinson, J. W.,ANAL. CHEM.33,
1067 (1961). ( 7 ) Russel, B. J . , Shelton, J., Walsh, A., Spectrochim. Acta 8, 317 (1957). (8) Shimazu, M., Hashimoto, A., Science of Light (Tokyo) 1 1 , 131 (1962). ( 9 ) Walsh, A., Spectrochim. Acta 7, 108 (1955). (10) Winefordner, J. D., A p p l . Spectry. 17, 109 (1963).
RECEIVEDfor review June 12, 1964. Resubmitted September 10, 1965. Accepted December 23, 1965.
Effect of Pulse Amplitude Shifts on Electron Probe I ntensity Ratios D. R. BEAMAN’ Paul D. Merica Research Laboratory, The International Nickel Co., Inc., Sferling Foresf, Suffern, N. Y. Serious errors in electron probe microanalysis can arise as a result of the dependence of the pulse amplitude output of a proportional detector upon the incident x-ray intensity a t relatively low counting rates. Such errors can b e avoided when utilizing pulse height selection techniques b y controlling either the pulse height selector position or the location of the intensity distribution. Pulse height discrimination eliminates these errors a t the expense of sensitivity. In this experiment pulse amplitude shifts and the associated errors are determined using magnesium, silicon, chromium, and a dilute iron-silicon alloy. Large errors are encountered, but it is shown that they can b e eliminated and it is concluded that it is possible to simultaneously obtain accurate intensity ratio measurements and high peak/background ratios.
T
dependence of the pulse amplitude output from a proportional detector upon the x-ray intensity at relatively low counting rates has been HE
well established. Birks ( 2 ) first mentioned this behavior and more recently a thorough investigation of the phenomenon has been completed by Bender and Rapperport ( I ) , who have noted that this behavior can lead to errors in electron probe microanalysis when pulse height selection is utilized. Because the limit of detection decreases with increasing peak/background ratio, in analyzing low concentrations i t is necessary to measure the intensity ratio using pulse height selection in which pulses from a narrow amplitude range are counted as a result of adjusting the base line and channel width of a single-channel pulse height analyzer. The pulse amplitude dependence on intensity complicates the use of pulse height selection since the base line and channel adjustments selected to accept the intensity from a pure standard will not be correct for an alloy. I n the present investigation the existence and magnitude of the pulse amplitude shift is established for a flow and sealed proportional detector. The errors caused by the amplitude
shifts are measured over a range of intensity ratios from 0.008 to 0.500. Techniques of correcting for the amplitude shifts are described and the accuracy of such procedures is then determined. The purpose is to illustrate that it is possible to obtain accurate intensity ratios concurrent with high peak/background ratios when using pulse height selection techniques regardless of changes in counting rate, anode potential or x-radiation wavelength. EXPERIMENTAL
A xenon-filled sealed proportional detector and a flow proportional detector (argon 10% methane at 30 cc./minute) were used in conjunction with a standard electronic counting circuit. All measurements were made on a Cameca electron probe microanalyzer operated a t 21.6 kv. I n the measurement of pulse amplitude shifts, pulse amplitude positions
+
Present address, The Dow Chemical Co., Metallurgical Laboratory, Midland, Mich. VOL. 38, NO. 4, APRIL 1966
0
599
i
I
i
1000
2000
3000
I
I
I
I
I
I
5000
6000
7000
8000
9000
0
(1%) ints., and ( 1 3 ) inips are measured (third method) using the integral mode. Finally, the beam currents il, i p , and i3 are remeasured to confirm current stability. The results are analyzed by considering the following ratios: i2/il and i 3 / i l are beam current ratios that indicate the true intensity ratio values; 12/11 and 13/11illustrate the large intensity ratio errors encountered from the pulse amplitude shifts; (I)GA,!I1, (I)PG/~I, and ( I ) A ~ /are l ~ the intensity ratios obtained when the base line voltage, anode potential, and amplifier gain, respectively, are varied; (IJlntg./ (11) Intg. and ( 1 3 ) ,Dtg./(ll) Intg. are the intensity ratios measured when use is made of the integral mode and like G/i1 and i 3 / i l correspond to the true intensity ratios. RESULTS AND DISCUSSION
0
4000
INTENSITY
( COUNTS /SECOND)
.
Figure 1 Pulse amplitude as a function of intensity for chromium (sealed proportional detector) and silicon (flow proportional detector) (curves with data points; scales on left) O p e n and closed circles differentiate between two different experimental runs. Percentage shift in pulse amplitude from a low intensity count of 400 c.p.s. as a function of intensity for silicon ond chromium (lower two curves; scale on right). The sealed and dow proportional detectors were operated a t anode potentials of i 800 and 1475 volts, respectively
were determined from base line sweeps (3 volts/minute) recorded on a chart recorder. Pulse amplitude shifts were created by varying the electron beam current from 8 to 222 na. and the x-ray intensity (300 to 11,000 c.P.s.) was measured using the integral mode. Each element was studied a t a constant amplifier gain which was measured with a calibrated oscilloscope. Following are the three methods that have been employed to overcome the difficulties that arise because the mean amplitude of a low intensity distribution (alloy) is greater than that of a high intensity distribution (standard) : The low intensity distribution is moved to the location of the high intensity distribution by decreasing the amplifier gain (amplifies output of preamplifier) or the detector anode potential (alters gas gain which is proportional t o pulse amplitude). Such a procedure shifts the low intensity distribution under constant conditions of pulse height selection. The pulse height selector is moved to the mean amplitude of the low intensity distribution by increasing the base line voltage until maximum intensity is noted while maintaining a constant channel width. Pulse height discrimination (integral mode) is used in place of pulse height selection in order that all pulses above a minimum amplitude will be counted regardless of their amplitude. All three methods will yield the correct intensity ratio but the first method requires a correction of the measured ratio that will be explained later. I n studying the effects of the pulse amplitude shift on electron probe in600 *
ANALYTICAL CHEMISTRY
tensity ratios, alloy compositions were simulated by using pure elements and reducing the electron beam current t o avoid the difficulties involved in producing homogeneous alloys. This is possible since the characteristic x-ray intensity is directly proportional to beam current. The following sequence of measurements was made on pure silicon, pure magnesium, pure chromium, and a dilute iron-silicon alloy. Three values of beam current are preselected so that il > i2 > i3. The pulse height selector is adjusted on the pure element for a given anode potential and amplifier gain using a beam current of il, and the intensity 11 is measured. Then the intensitiesl?and I 3 corresponding t o beam currents of iz and i 3 are measured under identical conditions of base line voltage, channel width, amplifier gain, and anode potential. The numbered subscript associated with the intensity I corresponds t o the beam current a t which the intensity was measured and the lettered subscripts correspond to the variables held constant with G , A , and B representing the amplifier gain, anode potential, and base line voltage, respectively. A constant channel width of 3 volts is used throughout. Next ( 1 2 ) ~and ~ ( 1 3 ) Q A are measured by adjusting the base line voltage (second method) until maximum intensity is noted. Then (12&3)80 and ( 1 2 & 3 ) A B are measured (first method) by reducing the anode potential and amplifier gain, respectively, to obtain maximum intensity. 11,12, and I 3 are then remeasured to check on stability. Immediately following this check the intensities (ZI),Dtg.,
Pulse Amplitude Shifts. The decrease in pulse amplitude with increasing detector intensity is shown in Figure 1. The shift with the xenon-filled sealed detector (chromium radiation) is substantially less than with the flow detector (silicon radiation). I n the same figure the percentage amplitude shifts are shown. Intensity Ratios for Silicon. The results for pure silicon are shown in Table I. If no correction is made for the amplitude shift (column 4), high and moderate concentration materials would yield intensity ratios which would be on the average 10 and 36y0 low, respectively. (Percentage deviations are indicated in parentheses in Table I.) As expected, the error increases as the intensity ratio decreases because of the increased amplitude shift. Repositioning of the base line (column 5) leads t o the measurement of the correct intensity ratio. Variation of the anode potential yields results (column 6) that are approximately 7 and 16% high for the high and low ratios, respectively, while reduction of the amplifier gain gives similar results (column 7 ) . When using the amplifier gain or anode potential (first method) there is an overcompensation. Increases in anode potential and amplifier gain broaden the amplitude distribution and consequently result in a reduction in the area included under the fixed channel width. Thus, as the anode potential or amplifier gain is reduced to reposition the distribution only part of the increase in intensity is due to the actual repositioning. Correction for this undesirable intensity increase can be made from data obtained on the pure material. The value of the quantity (dl/drl)~, can be found by measuring the intensity variation with anode potential a t constant amplifier gain, beam current, channel width, and kilovoltage. Re-
cause the intensity is proportional to the beam current, a correction for the increase in intensity accompanying the reduction in pulse amplitude can be made by using the following expression : (1)BGcorr. = (1)BGmeas. - (bI/bA)G% A A (i/i,,,,,) where i k n o w n and i are the respective beam currents a t which ( b Z / d d ) ~ and , (I)B~meas. are measured and A.1 is the reduction in anode potential required to give the low intensity distribution the same mean amplitude as the high intensity distribution. .A similar correction can be made when using the amplifier gain. In this case (bl/bG)A,is evaluated by measuring the intensity variation with amplifier gain for various channel widths a t constant anode potential, beam current, and kilovoltage. The corrected intensity is obtained uqing the following expression : (Z)mcorr. = I A B ~ ~~ S .( b l / b G ) ~ s A G (i/iknown) where AG is the required reduction in amplifier gain. -4pplication of the above procedures to selected eyierimental data resulted in ratios quite close to the measured current ratios. To utilize such correction techniques in the analysis of an alloy material would require that intensity variations with amplifier gain and anode potential be collected for different compositions. Due to the large number of materials analyzed, and the use of various kilovoltages, channel widths, amplifier gains, and anode potentials, such a procedure is not recommended, particularly since correct measurements can be easily made by adjusting the base line. Intensity Ratios for Magnesium. T h e results for magnesium (Table I ) are similar t o those for silicon, and illustrate how t h e percentage error in t h e intensity ratio increases rapidly, as expected, with decreasing intensity ratio when no correction for t h e pulse amplitude shift is made. Once again t h e use of aniillifier gain and anode potential lead to intensity ratios which are 1 to 12% high, depending on composition. The average error in intensity ratio encountered when adjuqting the base line was 1.5% in six experimental measurements. Intensity Ratios Obtained with Sealed Detector. I n t h e case of chromium t h e pulse amplitude shift ia small enough so t h a t no significant eriors are encountered. Hence, in this case there is no necessity for adjusting t h e base line. Similar results are obtained with nickel. Intensity Ratios for Iron-Silicon Alloy. These results (Table I) illustrate t h e effects described in t h e case of a practical material. I n using a beam current of 48 na. a n approximate error of 79y0 in t h e silicon intensity ratio of 0.008 results when no account of t h e pulse amplitude shift
Table 1.
Element
Intensity Ratios Obtained under Various Experimental Conditions for Pure Silicon, Magnesium, Chromium, and Iron-Silicon Alloy
i
Run
z1
1
0.481
Si
1
0,159
Si
2
0.486
Si
2
0.158
Si
3
0.491
Si
Si
3
0.159
Si
4
0.492
Si
4
0.161
Mg
1
0.492
hk
1
0.160
R k
2
0,500
m
2
hIg
3
0.168 0.213
Mg
3
0,035
Cr
1
0.167
Fe-Si
1
Fe-Si Fe-Si
2
Z
I1
11
0.491
0.506 (5.2) 0.180 (13.2) 0,517 (6.4) 0.181 (14.6) 0.535
0.510 (6.0) ‘0.183 (15.1) 0.514 (5.8) 0.181 (14.6) 0.535 (9.0) ‘0,190 (19.;) 0.018 (5.3) 0.190 (18.0)
O142S (11.0) 0.098 (38.4)
0.433 (10.9) 0.101 (36.1) 0.441 (10.2) 0.100 (37.1) 0.447
(9.1) 0.107 (33,5) 0.462 (6.1) ‘0.130 (18.8) 0 481 (3.4) 0,133 (18.9) 0.194 (9.3) 0.023 (30.3) 0.170 (1.8) 0.0017 (78.7) 0.0017
(OAB _
I 1
(0.)
0.161 (1.3) 0.491 (0,2)
0.159 (1.2) 0.490
(0.41
0.ln3 (4.4j 0.497 (0.2) 0.165 (0,6) 0.213 (0.5)
0,032 (3.0) 0.169 (1.2) 0.0079 (1.2)
179.8)
3
(OBG ~
~ (I)GA
I 1
‘ 0.0006 (92.3)
0,523 (6.3) 0,190 (18.0) 0.500
0,170 (6.3)
0,505
0.509 (2.2)
0.177
13.0)
(7.9)
‘0.222
(3.7)
0.037 (12.1) 0.170 0.0115 (36.9) 0.0121 (55.1)
0.491 0.159
(0.4)
(1.4) 0.169
(h::I33
_
0.490
(1.6) 0.113 (8.1)
166.3’1
Iintg. (I1)intg.
0.498 0.164 0.214 0.033
146.3) 0.0114 (35.7) 0.0126 (61.5) ~
0.0080 0.0084
0,0078
The figures in parentheses below each intensity ratio indicate the percentage deviation from the true intensity ratio where the true intensity ratio is taken to be ( Z ) l n t g . / ( I l ) i n t g . wherever this quantity was measured but otherwise is i/il.
is taken while adjustment of the base line leads to an error of only 1%. The effect of beam current is illustrated by comparing experimental runs 1 and 2 (48 na.) with run 3 (119 na). When employing a high beam current, a greater intensity drop is encountered for a given change in composition than for a lower beam current. I n experimental run 3 the silicon ratio (0.0078) is almost obliterated (0.0006 detected) by the amplitude shift but a base line shift of 4.3 volts yields the correct ratio. Decreasing the anode potentisl and the amplifier gain result in ratios that are 55 and 62% high, respectively. The differences in I,,, /(lJlntg for the three experimental runs are due to inhomogeneities in the iron-silicon alloy. CONCLUSIONS
Pulse amplitude shifts in excess of 25% were noted for intensities normally encountered in our instrument \\-ith silicon when using the flow proportional detector, while shifts of less than 5% were observed for normal chromium
intensities when using a sealed proportional detector. d s discussed elsewhere ( 1 ) the magnitude of the shift will vary from one detector to another and will be dependent upon operating conditions. Experiments using the flow proportional counter revealed errors in measured intensity ratios as high as as compared to negligible errors encountered in the use of the sealed proportional counter. Consequently, it is unnecessary to consider the pulse amplitude shift when working with a sealed counter, and the problems associated with the flow counter can often be alleviated by avoiding the use of pulse height selection. K h e n a low detection limit is sought, a narrow channel width should be employed. Increasing the channel width to minimize the effects of a pulse amplitude shift is not feasible because of the rapid decrease in peak/background ratio which accompanies the increased channel width. The base line rather than the amplifier gain or anode potential should be adjusted to account for VOL. 38, NO. 4, APRIL 1966
601
the pulse amplitude shift. I n very dilute materials it will be necessary to make a base line scan to accurately locate the pulse amplitude corresponding to the peak intensity. It is concluded that accurate intensity ratios can be easily obtained simultaneously with high peak/background ratios when utilizing pulse height selection with narrow channel widths by adjusting the base line to
correct for the observed pulse amplitude shifts. ACKNOWLEDGMENT
The author expresses his appreciation to S. L. Couling and V. B. Kurfman of The Dow Chemical Co. and C. R . Cupp and R. J. Raudebaugh of The International Nickel Co. for their helpful comments on the manuscript.
LITERATURE CITED
(1) Bender, S. L., Rapperport, E. , J., Fall meeting of the Electrochemical
Society, Oct. 11-13, 1964, Washington, D. C., Extended Abstracts of Electrothermics and Metallurgy Division, T’ol.
2, N o . 2, p. 104. (2) Birks, L. S., “Electron Probe Microanalysis,” p. 104, Interscience, Xew
York, 1963. RECEIVEDfor review October 4, 1963. Accepted February 11, 1966.
The Influence of Phosphoroscope Design on the Measured Phosphorescence Intensity in Phosphorimetry T. C. O’HAVER and J. D. WINEFORDNER Deparfmenf o f Chemistry, University o f Florida, Gainesville, Fla. 32603 The unique feature of all spectrophosphorimeters is the shutter mechanism, commonly called the phosphoroscope. Mathematical expressions are derived which relate the measured and instantaneous phosphorescence intensities to the decay time of the phosphorescent species and to the characteristic parameters of the phosphoroscope used in the spectrophosphorimetric measurement system. These expressions are valuable for evaluating any given phosphoroscope as well as for the design and construction of new phosphoroscopes. Numerical results of the equations are given for a commercial instrument using a rotating can phosphoroscope and a laboratory constructed instrument using a rotating disk phosphoroscope.
P
HOSPHORIMETRY as a means
of chemical analysis was first suggested by Keirs, Britt, and Kentworth ( 5 ) . Later, Parker and Hatchard (9) reviewed the possibilities of using phosphorescence measurements for the analysis of chemical constituents. Since these two articles, Winefordner and eo-workers (4, 6-8, 10-12) have applied phosphorimetry to the measurement of a variety of drugs or drug-like materials. hiost phosphorimetric studies have been carried out using the Aminco-Bowman Spectrophotofluorometer with phosphoroscope attachment (American Instrument Co., Inc., Silver Spring, >Id.). This instrument, called a spectrophosphorimeter, uses a rotating can shutter, commonly called a rotating-can phosphoroscope. The measured phosphorescence intensity for any species measured using a spectrophosphorimeter is a function of the characteristics and geom602
ANALYTICAL CHEMISTRY
etry of the phosphoroscope as well as the decay time, 7, of the phosphorescent species. I n this paper, mathematical expressions are derived lvhich relate the measurement and instantaneous phosphorescence intensities to‘the decay time of the phosphorescent species and to the characteristic parameters of a phosphoroscope. These expressions are valuable for evaluating any given phosphoroscope as well as designing new ones. PHYSICAL ARRANGEMENT OF COMPONENTS AND DEFINITIONS OF TERMS
The physical arrangements of the shutters in two typical spectrophosphorimeters are shown in Figures 1 and 2. The rotating-can phosphoroscope, shown in Figure 1, is the type used in the phosphoroscope attachment of the Aminco - Bowman spectrophotofluorometer. Radiation from the excitation monochromator passes through the excitation shutter aperture and onto the rotating can. When the can is positioned so that one of the openings is aligned with the excitation shutter aperture, the excitation radiation falls on the phosphorescent species in the sample cell and excites phosphorescence. At this instant, the emission shutter aperture is blocked by the phosphoroscope can so that fluorescence and scattered excitation radiation will not be detected. As the phosphoroscope can continues to rotate, the excitation shutter aperture is blocked. Fluorescence and scattered excitation radiation decay to a negligible amount almost instantly, and only phosphorescence remains. When the phosphoroscope can has rotated so that its other opening is aligned with the emission shutter aperture, the phosphorescence
radiation passes into the emission monochromator and photodetector. The dimensions of the phosphoroscope can are indicated in the insert in Figure 1. The arc lengths D and L , in em., refer to the closed and open portions of the can, respectively. The radius of the can is represented by r (in cm.). For simplicity, the excitation and emission shutter apertures are assumed to have the same width, given by T i ’ (in em.). Figure 2 shows the rotating disk, phosphoroscope (1-5’). Two notched disks are attached to a rotating shaft. When excitation radiation is passing through the excitation shutter aperture and the open portion of the excitation shutter disk, the other disk is aligned so that its closed portion is blocking the emission shutter aperture. Because the two disks rotate together, the sample is alternately excited and observed. The dimensions of the disks are also shown in Figure 2. They are exactly analogous to the dimensions of the rotating-can phosphoroscope shown in Figure 1. Figure 3‘4 shows, in detail, the operation of a phosphoroscope; it serves equally well for either the can or disk designs. All symbols and lines of Figure 3 are defined in the figure heading. Several assumptions which are generally valid will be made to simplify the following discussion. First, the excitation radiation intensity is assumed to be proportional to the extent of opening of the excitation shutter aperture. Second, the fraction of the instantaneous intensity of the phosphorescence radiation striking the emission shutter aperture which is received by the photodetector is assumed to be proportional to the extent of opening of the emission
