Spectrochimiea

Actn, 1954,Yol.6,pp. 198 to 210. Pergamon PressLtd., London

The spectrochemical determination of impurities in copper and copper alloys by means of a spark-ignited arc-like discharge
FREDERICK V. SCHATZ
Research Department, Revere Copper and Brass, Inc., Rome, New York (Receivecl13 November 1953)

Summary-A spectrochemical method of analysis of copper and copper alloys with a sensitivity capable of detecting lead, tin, iron, nickel, silicon, bismuth and aluminium at O.OOl%, tellurium, arsenic and phosphorus at 0.0047& and zinc at 0.01% is described. This method differs from the conventional arc analysis in that it is based on the excitation in a selected region in the vicinity of the cathode when using a special triggered discharge of the Multisource. If the circuit constants of the Multisource are adjusted to give a heavily, over-damped discharge, and if the sample is made negative, a narrow region of enhanced sensitivity exists for certain elements in the vicinity of the cathode. This region has been demonstrated to be equivalent to the cathode layer in D.C. arc. Variations in the relative sensitivities of certain elements, as well as variations in the excitation temperature, have been investigated for various points across the discharge gap.

1. Introduction
Recent trends in spectrochemical analysis, especially on the part of the large metal producing industries, have been toward fast automatic methods of analysis, utilizing the direct-reading spectrometer. This trend has introduced problems which, though not serious in a spectrographic laboratory where non-routine analyses predominate, have become increasingly important to the dire$ reading analyst. Most of these problems centre around simplifying and standardizing analysis procedures, and at the same time increasing precision and sensitivity. Applied specifically to the copper and brass industry, the problem involves combining into a simplified procedure a method of high precision, capable of determining 40% zinc with an error of 1% of content or less, and a method of extreme sensitivity, capable of determining impurity elements present in amounts of O.OOlo/o or lower. In a previous paper [l] the author described a procedure which handled the above problem with considerable success. Briefly, this procedure utilized several point-to-plane discharges, with a flat cast-metal specimen as the plane and a graphite electrode as the point. Different discharge characteristics were obtained by adjusting the capacity, inducta.nce and resistance in the power circuit of the Applied Research Laboratories Multisource [a]. By means of these spark igriited discharges, 30% zinc in 70-30 brass was analyzed with an error of &0.35% of content, and impurity elements were detected down to concentrations of 0.01-0.02 in yellow brass and 0.02-0.05/0 in red brass and copper. This original work was repeated by others [3] and has provided the basis for a successful direct-reading method of analysis. Although these sensitivities were adequate for the analytical control of a large number of alloys, an equally large number of high purity alloys and refined copper remained to be analyzed by other methods. 19%

The spectrochemical

determination

of impurities

in copper and copper alloys

These other methods, based principally on D.C. arc excitat,ion, use self-electrodes, solutions, oxides, or metal globules. Results based on D.C. arc excitation in general are very good. The procedures, however, are often very unsatisfactory or even impossible when viewed from the standpoint of fast analytical control. In addition to disrupting working schedules, they mult,iply the basic analysis cost per sample by large and unpredictable factors. This situation is aggravated to the extreme when direct-reading spectrometers are used. The method described in this paper utilizes a spark-ignited, arc-like discharge of high sensitivity, and is based on an unusual experimental arrangement. The components of the arrangement, though not new, do yield unique results when combined in the experiment to be described. Also once installed and adjusted, t,he method has the same simplicity and speed of operation as the conventional analyses using solid samples and point-to-plane sparking techniques. Applied to copper alloys it has an inherent relative sensitivity for some elements equal to all but the most refined D.C. arc type of analysis. The sensitivity of detection of phosphorus in copper, for example, is better than any previous method investigated by the author. Elements investigated in detail were lead, tin, iron, nickel, silicon, bismuth and aluminium, all detected at 0.001%; tellerium, arsenic and phosphorus, detected at 0.004$1/,; and zinc, detected at 0.01%.

2.

Experimental arrangement

The method in its ultimate form was a natural outgrowth from an investigation of the distribution of certain element line intensities across the discharge. The electrode system was a conventional point-to-plane type consisting of a machined metal surface and a pointed graphite counter electrode. The spectrograph was a large Bausch and Lomb Littrow model using quartz optics. The excitation source was an Applied Research Laboratories Multisource [2]. In order to examine line intensities emanating from the vapour cloud at various distances from the sample, the discharge was considered to be composed of a set of lamellae parallel to the sample plane, each lamella emitting radiation characteristics of the state of affairs within its small volume. The optical arrangement for isolating the lamellae along the length of the discharge employed two crossed cylindrical condensing lens combined in a manner suggested by HANSEN [4]. One lens with its axis horizontal was placed at the spectrograph slit to focus the image of the discharge vert,ically on the collimator opening with a magnification of 6.5. The second lens with its axis vertical was placed in front of the discharge to focus the image of the discharge horizontally on the slit with a magnification of 4.0. If the axis of the spark discharge is made horizontal, t,he radiation received by the system will be from the lamellar volume described above. The resolution of the system, i.e. the thickness of the lamella, is defined by the slit width. In the present instance a 40 p slit was used, defining a discharge cross-section 10 ,u in width. Consequently, horizontal adjustment of the discharge position will enable the system to analyze 10 p sections across the discharge from the sample plane to the counter electrode. A horizontal arrangement of the conventional Petrey stand was devised to
199

FREDERICK

V. SCHATZ

facilitate the handling of flat samples (see Fig. 1). To prevent the projecting surface of the sample from partially blocking the horizontally divergent rays. the vertical plane of the sample was placed at an angle of 54 to a vertical plane along the optic axis. The stand and electrode holder were mounted on a piece of t.ransite, a hard, easily machinable, asbestos board. The whole assembly was then mount,ed on the optical bench in a stand with a fine screw thread for accurate horizontal adjustment,. The vertical adjustment was relatively coarse. To facilitate horizontal adjustment and to furnish convenient reference point)s. an auxiliary lens was placed on the optic axis at the rear of the system and an enlarged image of the counter electrode was projected on a screen. Exact placement of the discharge gap was accomplished by aligning the image of the counter electrode after the position of its point with reference to the sample plane had been accurately set by means of a 2.5 mm gap spacer, placed in the position of the analysis sample. The spark stand was split horizontally for this purpose. Initial adjustments of the optical system were made with a line-filament microphotometer lamp placed with t.he filament vertical at the location of the spark gap. The horizontal and vertical images were then centred at the slit and the collimator respectively. The position of the discharge gap with respect to the optic axis can be maintained if fiducial marks and an appropriate scale are placed on the auxiliary screen once the init,ial alignment is accomplished.

3.

Investigation

of intensity variations of selected elements across the discharge gap

Three types of discharges commonly used in copper alloy analysis were studied for line intensity variations across the discharge gap (see Table 1). With the sample posit,ive none of the discharges showed any unusual variation in line intensities
Table 1.
j Sample

Types

of discharge

polarity

~ ;

Precision

I
I

Sensitizdy

Typical

setting

I
2.0% Error 10 O/cError 2.0% Error Poor Excellent Good 10 ,uF., 200 /H.,* 10 Cl 60 /IF., 400 ,uH.,* 20 R 10 /LF., 200 /H.,* 3 fl

1. Slightly overdamped 2. Heavily overdamped 3. Slightly underdamped

/ i

+ +

* Settings include 25 ,uH distributed inductance.

in the region adjacent to the sample electrode. However, with the sample negative, an enhancement of line intensities appeared in the region adjacent to the cathode. This cathode enhancement increased from a small value in the slightly underclamped discharge to a very pronounced effect in the heavily overdamped discharge. As the heavily overdamped discharge had the most pronounced effect in combination with a 20 to 1 increase in overall sensitivity, it was selected for study. In the heavily overdamped or Type 2 discharge line intensit,ies of certain element,s were found to be greatly enhanced in a narrow region close to the negative
200

sample electrode. Visually, the image of the discharge at the slit also presented an unusual appearance in that in this same narrow region there was a pronounced greenish glow, due to copper emission. The image of Ohis region was easily discerned since it. was spread out in the verGea direction. This glow was at its narrowest, when using a copper sample and t<ended to spread across the discharge as the zinc The region of enhanced line int,ensities also tended to content was increased. broaden with increased zinc content but not to as great an extent as the visual This glow appeared shortly after the discharge was appearance would indicate. initiated and reached its maximum extent within approximately t)he first ten seconds of excit,at,ion. Type 2 excitation is a series of ~~nidirect,ional, heavily o~Ter~la~~ped pulses. Computation shows each pulse to have a peak current of 45 amperes and a duration of 6.3 milliseconds, i.e. time required for t,he current to drop to 10% of its maximum value. As t.here is an 8.3 millisecond interval between charging cycles, this pulse length is about the maximum permissible with 60 cycle initiat,or control. Basically, Type 2 excitation can be considered an interr~lpted D.C. arc. Since discharge conditions reassert themselves each cycle with the condenser voltage returning to its peak value of 940 volts each time, the average arc voltage drop remains higher than in the continuous D.C. arc. Also, unlike interrupted arcs of the Pfeilsticker variety, where the input mains remain connected to the circuit and current is limit,ed only by the arc and a series ballast an accurately measured packet of charge is delivered t,o the are at each cycle by the ~ultisouree. The result is a very reproducible arc spectrum in which t,ransit,ions from high energy levels are encouraged. This is evident from the large number of ion lines in addition to the usual atom lines. Selected brass and copper samples containing various elements in small amounts were excited, and spectra were photographed from sections of the diseharge 0, 0.25, 0.5, I-0, l-5 and 2.5 mm from the sample elect.rode in a 2.5 mm discharge gap. R,elat,ive sensitivity is defined as the concentrat,ion where the line merges into background, provided sufficient sample is excited to produce background. It also may be considered proportional to the effective line-to-background ratio when background is present, and the line intensit&y is corrected for background intensity. Consequentsly, to obtain a comparative picture of the change in relative sensitivity of impurity element, lines as the dist,ance from t,he sample increases, effective line-to-background intensity ratios were comput,ed and plotted on a logarithmic scale. These results for a sensitive iron and lead line in both a 70-30 matrix and a copper matrix are shown in Fig. 2. Inspection of the curves shows an increase in cathode enhancement wit)h increasing copper content, changing from 4.5 to 1 t,o 15 to 1 in the case of lead. from I to 1 to 6 to 1 in t,he case of iron. This e~haneement varies to some extent, with the line select~ed, but in the ease of low impurity elements the number of lines available for examination is usually restricted to the most sensitive arc lines. Elements behaving like lead, in that there is a pronounced cathode enhancement regardless of the zine content, of the matrix, are tellurium, nickel, tin and bismuth. Tellurium and nickel have a slightly redueed cat,hode enhancement in copper in comparison to that, in t,he brasses. Tin is similar t,o lead in t,hat it,s cathode enhancement, is greatly increased
201

FREDERICK SCHATZ V. in copper. Bismuth was checked only in brasses where its cathode enhancement is very pronounced. Elements, behaving like iron, in that cat)hode enhancement becomes very slight when zinc content increases, are arsenic and antimony. Phosphorus and zinc have a pronounced cathode enhancement in copper. Aluminium showed Wle or no cathode enllancement in either 70-30 brass or copper. In Fig. 3 the intensity ratios of a few copper atom and ion lines are plotted as atom-atom and atom-ion combinations versus distance across the gap to demonstrate the variation in cathode enhancement for different lines of the same element. The above

D.C. copper
demonstrated
6.0

arc, as described

material suggests a similarity bet,ween Type 2 discharge and the by MILBOURN [5] [S]. In his papers, MILBOURN qualitatively the validit,y of the following points:

Cu2634.91

04

0.5

I.0 mm -

I.5

2.0

2.5

0.21 0

05

I.0

I.5

2.0

25 mm

DISTANCE ACROSS DISCHARGE CAP-

Fig. 2. Variation in relative sensitivity of lead and iron mross t,he discharge gap.

Fig. 3. Variation in intensity ratios of selected copper line pairs amoss discharge gap.

1. In a D.C. is untouched. 2. Oxygen

copper arc the negative

electrode is consumed

while the anode with

is necessary

for the maintenance

of the arc, the reaction

copper occurring at the cathode. 3. Impurity lines behave generally like copper lines in that their arc radiations

are strongest near the cathode and diminish in intensity across the arc. 4. Ion-line radiations are emitted only near to the electrodes. 5. When the size of the sample at the cathode is reduced to a small globule, vapour concentration in the arc rises, the cathode layer disappears and the line enhancement spreads across the arc gap. 6. The principles applying to copper are valid in general for it,s alloys. 202

Fig. 1. View of spark stand and sample.

Fig.

4.

Horizontal

D.C. arc, showing pin samples.

The spectrochemical

determination

of impurities

in copper

and copper

alloys

In order to make a direct comparison between the two kinds of discharge, a horizontal D.C. arc was devised. Some experimental difficulties were encountered, due to the t,endency of a D.C. arc to either hang on a point or to wander erratically. However, &-in. diameter pin samples with hemispherical tips were found to be adequate for the purpose of comparison, in that the arc wandering could be kept reasonably within bounds in the optical system and still not hang at one point (see Fig. 4). The gap distance was adjusted to 2.5 mm and a 3.5 ampere arc current, equivalent to the average current, in the Type 2 discharge, was used. An 85-15 brass alloy containing 0.016% lead was used for bot#h anode and cat,hode,

mm Fig. 5. Comparison of the relative sensitivities of Pb 1833 across the discharge in Type f and the D.C. arc.

and a series of spectra at varying distances from the cathode was made, using both types of excitation. The variation in relative sensitivity of the Pb 28338 line versus distance across the gap is shown for both cases in Fig. 5, where they are superimposed for the sake of comparison. As indicated in the figure, t,he cathode layer is present in both discharges and in the D.C. arc shows a relative sensitivity approximately ten times greater than the corresponding layer in the Type 2 discharge. This suggests the use of cathode layer in the above described D.C. arc excitation for t,he detection of impurities at extremely high sensitivities. The increase in relative sensitivity at the anode in each case is not due t#o an increase in the effective intensity of the lead line, but to a decrease in background which became quite low and erratic in the region of the anode. In the Type 2 discharge, the standard deviation for the intensity ratio Pb 2833/Cu 3030 was 203

3.X0,/ in the centre of the discharge, and ranged from approximately 107; at the cathode to 157 and greater at the anode. In the D.C. arc, standard deviations were not computed but appeared to be somewhat larger. Proceeding from the cathode to the anode, the reduction in effective intensity of the lead line is approximately 10 to 1 in the case of the D.C. arc and 5.5 to 1 in the case of the Ion lines, which in both cases are strongly enhanced at the Type 2 discharge. cathode, extended across the Type 2 discharge. Also in both cases the cathode was consumed and the anode was relatively untouched. The similarities in the spectra, in the electrode attack, and in the cathode enhancement of the lead line, indicate that Type 2 excitation is close to that of a D.C. arc, and that the cathode layers are similar in both types of discharge. 4.

Investigation

of excitation temperature

across the discharge gap

The variation in the arc-spark line pair in Fig. 3 raises the question of whether the discharge is in thermal equilibrium at all points across the gap. If t#emperature is the only parameter, the relative intensity of lines 1 and 2 is given by

- &) A,G,v, e - (4 KT_~_ 12 = A,G,v,

I, where I = measured intensity

A = transition probabilit*y
G = statistical v = frequency E = energy of init,ial transition K = Boltzmann constant level weight

T = absolute temperature
L_~KGSTROTH and MCRAE [7] have determined the transition probabilities of selected tin lines in the course of an investigation of temperatures existing in a D.C. and A.C. spark and the D.C. arc. Using values determined by them, we have for three tin line pairs: Sn 3262 - 5; = 3.06 e Sn 3034 Sn 2850 _ ??$?!? --__ = 0,95 e Sn 3034 Sn 2840 Sn 2706 = 1.90

A flat copper sample containing approximately 0.2% tin was excit,ed. using a graphite counter electrode, and relative intensities of the above tin line pairs, corrected for background, were determined at various locations across the discharge gap. The relative intensities of the same tin line pairs were also determined in t,he horizontal D.C. arc using pin samples of the dimensions described before. The t#emperature computed from the relat,ive intensibies of these tin line pairs 204

The spectrochernical

determination

of impurities in copper and copper alloys

are summarized in Table 2. It is apparent from the large divergences in temperature in the region next to t,he negative sample electrode for both the Type 2 discharge and the D.C. arc: that, in this region other factors in addition to temperature are operating. Consequently, a Boltzmann dist,ribution of the electron population among t,he energy levels is not possible in the region of the catjhode layer. On the other hand, at the centre of the arc, at about 1 to 1.5 mm from the electrode, temperatures computed from the two line pairs approach each other, and thermal equilibrium can be considered to exist.
Table 2. Temperatures across discharge gap
in mm

Distance across gap

RI Hn 326 Sn 3034

( TK

RI

TK

RI

) TK

RI

TK

~ RI 1 TK

RI

!
2.17 115,000+ 14,700 1.88 13,400 1.34 8,000 1.39 I 8,400 ( 1.46 8,900 _____~_ 1.29 i,600 1.27 7,500

i TK I
7,200

Type 2 DC &PC

1.97 I 14,700 1.88 i 13,400 2.0 , 15,000

, / 1.97

I 1.19 6,900 1 1.21 7,100

1.09 IF6,400 1.13 6,600

1.16

6,800

1.23 0.98 1.09 1.13 1.12

5,800 , 6,400 ( 6,600

6,500

sn 2850 Sn 3034 ---DC arc

Type 2

0.11 0.13 0.1 0.10 0.11

~ 6,500

6,000 6,300

0.12 0.15 0.14

6,300 7,000 6,700 6,000 6,000

0.13 0.14 0.14 0.11 0.12

6,500 6,700 6,700 6,000 6,300

~0.14

5,800 6,000

/ 0.11
0.11

6,700 6,700 0.15 7,000 __-___ 0.12 ) 6,300 0.11 ~6,000

0.14

0.12 0.11 0.12

6,300 6,000 ~6,300 0.11 0.11

I

! 0.11 6,000 i 0.10 1 5,800

I 2.02 2.34 ; 2.15

6,OO 6,000

2.56 2.32 2.17

1 ( 12.33 (
2.16 2.25

2.30 2.22 2.16

2.12

2.16 2.17 2.12 1.95 1.95

: I

2.17 2.33 2.48 ,16 2.16

, 2.06

2.08

2.03

BRACDO, CRAGGS, and WILLIAMS [R] observed divergences in temperature almost idenOica1 to those in the cathode region in the course of an investigation of t,he excitation temperatures in a low-voltage, high-power, triggered spark-arc source. They used Ba I, Ba II? and Mg I line pairs. Though not statSed explicitly, the t,otal light flux from the discharge was probably examined. As part of their explanation of this temperature discrepancy, reference was made to the possibility that different zones of the discharge produced different populations in different energy levels. However. this point was not examined in detail. Relative intensities of the tin line pair 2840/2706, though not temperature sensitive, were also determined to check departures from the theoretical value of I .9. This determination was not too successful for the spark-ignited arc-like discharge since the photographic density of Xn 2840 was greater than 2.0, and 205

FREDERICK V. SCHATZ

accurate microphotometer readings were not possible. Using the values of the tin line pair 284012706 in the centre of the discharge as an emulsion contrast control, slight corrections were applied to the relative intensities of the other line pairs. When this was done, temperatures were found to be single valued in the region 1.5mm from the cathode to the anode in both cases, and to have a value of 6900"-7100Kfor the Type 2 discharge and 6300"-6500Kfor the D.C. arc. Since radiation from the cathode layer is several times more intense than that from the centre of the discharge, its contribution to the total light flux would tend to be dominant in any optical condensing system that collects radiation from the entire discharge. Consequently, any plate calibration or excitation standardization based on selected line pairs is of dubious value if the above factors are ignored. Errors arising from this situation are aggravated if slight changes in
IO

A7ttrgtt--_j

i/iii]]

% LEAD -

Fig. 6.

Determination of lead in brass and copper.

t,he adjustment of a critical optical system tend to throw the radiation from the cathode layer in or outside the ent,rance pupil of the optical system of the spectrograph. 5. A sensitive analysis of copper and copper alloys The enhanced relative sensitivity of many of the elements in the vicinity of the cathode or sample electrode, using the overdamped discharge, opened up the possibility of exciting large metal samples with a spark-ignited, arc-like discharge equivalent to the conventional D.C. arc in sensitivity. This possibility was encouraged by the fact that many of the highly enhanced elements are the same elements which need to be detected in low concentration in copper and copper alloys. However, if an analysis is to be practical for fast routine work, optical adjustments and the position of the sample should not be so critical that it is difficult to maintain reproducibility. In the investigation just described, line intensities with respect to background change with great rapidity in the cathode region. 206

The spectrochemical

determination

of impurities

in copper

and copper

alloys

Consequently, grave doubts were entertained regarding the suitability of a spectrochemical analysis based on enhancement at the cathode. However, in tests over a period of two years, during which the system was dismantled and reassembled many times, no difficulty was experienced in reproducing the w-orking curves. Day-to-day checks were easily accomplished not only by means of the auxiliary lens system. but also by means of spectrographic line pairs whose relative intensities change rapidly in the cathode region. Figs. 6 to 10 show working curves that have been included to demonstrate the various analyses possible on copper alloys. In many cases the most sensitive lines hare not been used. Since most of the curves are non-linear and approach

3.004

0 Cl

c,o:! IRON -

0.04

::
Fig. 7.

Determinat.ion

of iron in brass and copper

background relative int,ensity as a horizontal asymptote, background corrections are made. The auxiliary curves designated by primed letters, included in the various figures are obtained by correction of the background of the analysis line only. In all cases these curves are linear, and their slopes, wit,h few exceptions, are approximately 45. Once these auxiliary curves are obtained they can be used for extrapolating values on the uncorre&ed curves by adding the background relat*ive intensity to values along their extended range. This device is quite useful when a limit,ed number of standards are available and the uncorrected working curve is to be used for analysis purposes. However. for careful work near the limits of relat,ive sensitivity, background corrections should a,lways be made. Shifts of the working curves? due either to a change in concentration of the
207

FREDERICK V.

SCHATZ

matrix element or to the presence of other elements, were observed and are indicated on many of the curves. Studies of these shifts were hampered by an insufficient number of accurately analyzed standards, especially in the very low ranges of impurity concentration. Consequently, many of the curves are incomplete. In some cases shifts were so small that an average curve was drawn for the uncorrected curves. Such is the case in Fig. 8 where a single uncorrected curve is drawn for tin. The following list is a brief resume of the observed causes for curve shifts, many of which are illustrated in the figures: 1. Changes in the matrix element concentration. In many cases this shift occurs in a straightforward manner, but it is often complicated by the effects listed below.

o 67-33 0 70-30

BRASS, 0.5 % Pb BRASS

90-10 BRASS

0.3 0.01

0.02 Fig. 8.

0.04 % TIN Determination

0.1

0.2

0.4

of tin in brass and copper.

2. Changes in the length of the exposure time, Curve shifts due to different rates of evaporation with time of the element in question and the internal standard copper may result from a change in exposure time. 3. Changes in background intensity relative to the line intensity. Curve shifts of this type have been observed in copper where the presence or absence of minor impurity elements has an effect on the background. This effect can be eliminated by correcting both impurity and matrix lines for background. 4. Changes in relative intensity of an impurity line due to the presence of Changes of this nature, which are all out of proportion to the other elements. amount of their additions, have been observed when tin, silicon, and aluminium are added to copper. 5. Changes in relative intensity due to the effect of a change in concentration of the specific impurity on the intensity of the matrix element line. This rather
208

The spectrochemical

determination

of impurities

in copper

and copper

alloys

peculiar shift was observed in the determination of small amounts of iron in copper where increasing amounts of iron so enhanced the copper lines that the curve for relative intensity versus concentration was quite flat. [l] 6. Changes in the volatility of the brass alloy with increasing amounts of zinc.

5
z

0.1 001 Fig. 9. Determination

0.1 %ZINC of zinc in copper

I.0 and silicon-manganese bronze.

70 PHOSPHORUS

Fig.

10.

Determination

of phosphorus

in phosphor

bronze

and copper.

Since copper is quite different from zinc in its susceptibility to attack by the discharge, relative intensities of different elements vary differently with increasing zinc content. 7. Changes due to metallurgical effects. This extremely annoying curve shift
209
3

FREDERICK V.

SCHATZ

has its origin from causes which are often obscure and difficult to locate. Also it differs from the preceding shifts in that it is often random in its occurrence. In general the method has displayed the greatest improvement in sensitivity over conventional spark methods in the analysis of red brass, gilding metal and In the conventional spark methods, using the light emitted over the copper. entire discharge, background builds up with increasing copper content very rapidly, and is so heavy for pure copper that the entire spectrum is quite black. However, in the method described, which uses a selected portion of a special type of discharge from the Multisource, copper is handled with the same ease as the As an example, this method has been yellow brasses of high impurity content. Before especially noteworthy in the determination of phosphorus in copper. using the present procedure, no excitation conditions could be found that gave consistently good results for a phosphorus content of 0.01 to 0.05% in copper due to variations in the relatively strong background present. Any D.C. arc excitation methods were also complicated by the fact that when O.OOl~/, Fe was present, Fe 2535.6 interferes with P 2535.7, the most sensitive phosphorus line. However, in the present method Fe 2535.6 is suppressed and does not appear until Fe reaches O-O15o/o, well above the Fe content of most refined copper.

References
[l] SCHATZ, F. V.; J. Inst. Met. 1952 80 77. [2] HASLER, M. F., and KEI+~P, J. IT:.; J. Opt. Sot. Amer. 1944 34 21. [3] Applied Research Laboratories Newsletter 1951 4 2. [4] HANSEN, [6] MILBOURN, M.; G.; Z. Phys. 1924 29 356. [5] MILBOURN, M.; J. Inst. Met. 1943 99 441. Proc. Phys. Sot. Lond. 1947 59 273. [7] LANGSTROTH, G. O., and MCRAE, D. R.; Can. J. Res. 1938 16A 17. [8] BRAUDO, G. J.; CRAGGS, J. D., and WILLIAMS, A. C.; Spectrochim. Acta 1949 3 546.

210