Computer Analysis of Photographed Optical Emission Spectra*

A. W. Helz, F. G. Walthall, and Sol Berman U. S. Geological Survey, Washington, D. C. 20242

(Received 21 April 1969; revision received 29 May 1969)

A recording system and computer can be used for the complete spectrochemical analysis of photographed spectra. Transmission values, taken at equal intervals of travel Mong the spec- trum are t ransferred to magnetic tape with high precision, high speed, and in a form suitable for computer processing. This processing may include wavelength determination, line identi- fication, and plate calibration. INDEX HEADINGS: Emission; Computer; Microphotometry; Spectrochemistry; Silicates.

I. INTRODUCTION

The spectrochemist analyzing a complex material for all minor and trace elements for practical expedi- ency limits his observations to a small fract ion of the information recorded in the spectrum. By transfer- r ing a large amount of such recorded data to a com- puter, much more detailed considerations become feasible: inter-element effects, background correc- tions, source parameters, and the composition of the standards. Some of the problems of automatic plate reading and computer interpretat ion of spectrochemi- cal analysis have been discussed by Helz. 1

A microphotometer moves the spectrum through a scanning beam at a velocity of 5-mm per see. At 5- /~ intervals of distance traveled, the analog photo- mult iplier output is digitized and stored in a buffer. When a predetermined number of readings have ac- cumulated, they are t ransfer red to magnetic tape as a group without in ter rupt ing incoming readings. The number of the reading (not recorded) determines the wavelength coordinate. Transmission values are recorded in three digits. By giving careful attention to instrumental details and using " i n t e r n a l " spec- tral-fiducial lines, accurate rapid scanning is realized.

Computer line-finding starts with a known strong line and an approximate value of dispersion. The position of the line in the spectrum expressed as a reading number is determined automatically and is stored with the related transmission value. Other fiducial lines are then picked out, sequentially accord- ing to wavelength, by searching for a minimum trans- mission in the projected position, examining trans- mission values, calculating wavelength, comparing with the " k n o w n " value, and determining the actual dispersion.

Lines for chemical analysis are " f o u n d " by the computer using a wavelength list of such lines, the dispersion curve, and fiducial line positions. The dis- persion curve and fiducial line positions have been determined earlier from an emulsion-calibration ex-

~Publ icat ion authorized by the Director, U. S. Geological Survey.

posure in the manner described above. The location of an analytical line is calculated up and down the spectrum from the bracketing fiducial lines. The transmission at the calculated spectral position or one reading on either side, whichever has less trans- mission, is selected for fu r ther processing.

Conversely, f rom the same kind of recorded data the wavelengths of all observable lines may be cal- culated and listed along with corresponding trans- mission values.

An IBM 360-65 computer was used in this work with Fo r t r an IV language.

II. RESULTS

The performance of the entire system is summed up by (1) a s tudy of calculated wavelengths for evidence of length measuring capability, (2) trans- mission repeatabil i ty data for evidence of the feasi- bility of taking such readings rapidly, and (3) a spectrochemical analysis of a rock sample for 18 ele- ments to demonstrate line-finding capability and all the spectrochemical calculations required for a quan- t i tative report .

To demonstrate transmission repeatabi l i ty and wavelength accuracy, the following experiment is reported. The spectral region between 2940 • and 3110 A of an iron-argon spectrum was scanned 40 times at the maximum rate of 1000 readings per sec. All spectra re fer red to in this repor t are photo- graphically recorded with a reciprocal linear dis- persion of approximately 5 £ / m m . The spectral range contained seven fiducial lines: 2941.343, 2965.255, 2999.512, 3024.033, 3047.605, 3067.244, and 3100.- 666 4. These lines are f rom a larger list prepared for the 2250-3500 ~_ spectral range. The computer was programmed to pick out all lines between 0% and 89.0% transmission and calculate wavelengths in four ways, which were combinations of the fol- lowing factors: (1) the first and last fiducial lines in the above list, (2) all seven fiducial lines, (3) the position of minimum transmission, and (4) a l inearly interpolated position of minimum transmis-

S08 Volume 23, Number 5, 1969 APPLIED SPECTROSCOPY

RUN NO° 17 . L I N E N O . READ. NO. FRACL P E A K FR&CR CWLU GMLC CWLUFL C~LCFL UiSP

FIOL 1 e * 2 0 2 ~ 0 . 0 7 2 7 2 3 0 . g 2 8 2 9 6 | . 3 A 3 2 9 6 1 . 3 6 3 2 9 4 1 . ~ 6 3 ~ 2 9 6 1 . 3 6 3 3 9 , 6 Z 9 2 262 0 . 8 7 0 9 0 . 1 3 0 2 9 6 2 . 8 6 6 2 9 4 2 . 9 ~ 5 2 q ~ 2 , 8 6 7 2 9 6 2 . 8 7 7 3 9 , ~ 5 2 3 659 O .5&q 39 0 . ~ 5 | 2967o861 2 9 6 7 . 8 7 3 2 9 6 7 . 8 7 3 2 9 6 7 . 8 1 5 $ 9 . 3 5 2 4 698 O.6~& 76 0 . 5 4 4 2 9 5 3 . 9 2 2 2 9 ~ 3 . 9 3 2 ~ 9 5 3 . 9 4 ~ 2 9 5 3 . 9 ~ 5 J9 .3~2

833 0 . 2 S 0 2 | 1 0 . 7 5 0 2 9 5 7 . 3 4 6 2 9 5 7 . 3 5 1 2 9 5 7 . 3 7 7 2 9 5 7 . 3 7 l 3 9 . J ~ 2 6 q67 0 o 2 3 & 796 0 . 7 6 6 2 9 6 0 . 2 ] 7 2 9 6 0 . 2 6 1 2 9 6 0 . 2 7 ~ 2 9 6 0 . 2 6 8 3 9 . $ 5 2

EIOL 7 * e l [¢3 ** 0.879 376 0.121 2965.208 2965.229 2965.255 *$ 29h5.255 ~9.352 1208 0 . 2 8 6 5 0 . 7 1 ~ 2 9 6 6 . 8 5 7 2 9 6 6 , 8 6 2 2 9 ~ 6 . 9 0 2 2 9 0 6 . ~ 9 1 ~ 9 . ~ 7

9 1305 0 . 6 5 ~ 50q 0 . 3 6 6 2 9 6 9 . 3 1 7 2 9 6 9 . 3 3 2 2 9 6 ~ . 3 6 1 Z 9 6 9 . 3 ~ 9 ~ 9 . 6 ~ 7 I0 1336 0.9|0 56 0.090 2970.052 2970,076 2970 .09R 2 9 7 0 . 1 0 2 ~9.¢37 | I 1656 0,333 17 0 o 6 6 7 2973.167 2973.153 2973.191 2 9 7 3 . 1 8 ~ ~ 9 . ~ 7 12 1687 0 ° ¢ 1 5 58 0 . 5 8 5 2979.005 2 9 7 9 . 0 1 6 2 9 7 9 . 0 6 9 2 9 1 9 . 0 6 1 ~ 9 . 6 3 7 13 178 l 0 . 8 6 2 62 O.13S 2 9 8 1 . 3 8 9 2 9 8 1 . 6 0 9 2 9 8 1 , 4 3 2 2 9 8 1 . 6 3 6 ~ 9 . ~ $ 1 l ~ 1865 0.~1~ 53 0 . 5 8 6 2 9 8 3 . 5 Z 0 2 9 8 3 . 5 2 ~ 2983.562 2 9 ~ 3 . 5 5 ~ 3 v . 6 ~ 7 15 191~ 0o0 850 1 . 0 0 0 2 9 8 6 . 7 6 2 2 9 8 4 . 7 6 1 29~6 . f l 0S 2 9 o 4 . 7 ~ 5 ~9o~37 16 229~ 0 . 0 5 ~ 12 0 . 9 5 0 2 9 9 6 . 6 0 0 2 9 ~ 6 . 3 9 9 2 9 9 6 . 4 3 9 2996.6Z2 3 9 . 6 ~ 7

FIDL 17 * * 2~9~ * * 0 . 5 6 9 133 0 . 4 3 1 2999.6T2 2 9 9 9 . ~ 5 2999.51Z * * 2 9 9 9 . 5 1 2 ~ 9 . 6 3 7 I8 2531 0.632 ~60 0.368 3000.611 3000.425 3000.6h3 3000.~59 39.6~7 l~ 2550 0.976 20 0.026 3 0 0 0 . 8 9 3 3 ~ 0 0 . 9 1 6 3000.963 3 0 ~ 0 . 9 6 9 ~ .6~7 20 2517 0.511 597 0.699 3 0 0 2 . 5 9 2 3 0 0 2 . 5 0 3 3 0 0 2 . 6 6 3 3OUZob61 ~9.637 21 2~00 0 . 8 5 6 199 0 . 1 4 6 3 0 0 7 . 2 3 3 3 0 0 7 . 2 5 3 30NT.2R3 3001.2~5 ~ 9 . ~ 1 22 283~ 0.321 51 0.679 3 0 0 8 . 9 9 5 3008.102 3008.1~5 300~.l$~ 39.6~ 23 2890 0.735 ~36 0.265 3009.516 3009.533 30,t9.5S~ 3009.56~ ~9 .~7 24 3208 0.195 515 0.805 3017 .~ l 3017.58~ 3017.629 3 0 L 1 . b 1 5 3 9 . 6 ~ I 25 3261 0.808 711 0.192 3019,925 3 0 1 ~ . 9 6 6 3 0 1 P . 9 7 3 3 0 1 ~ . 9 1 5 39.q3~ 26 3327 0 ° 0 7 l .OOO 3 0 2 0 . 5 9 9 3 0 2 0 . 5 9 7 3 0 2 0 . 6 6 6 3020°627 ~9o~$I 27 33~3 0 . ~ 6 7 6 0 . 5 3 3 3 0 2 1 . 0 0 5 3 0 2 1 . 0 1 5 3 0 2 1 . 0 5 3 3 0 2 1 . 0 ~ 5 3 9 . ~ 7

F IDL 28 ~* 3~60 *¢ 0.930 66 0.066 3023.972 3023.996 3026.033 ¢¢ 302~.033 3~o631 29 3525 0.356 57L 0.664 3 0 2 5 . 6 2 1 3025.628 3025.678 3 ~ 2 > . 6 1 2 39.696 30 3532 0.615 49 0.585 3 0 2 5 . 7 9 8 3025.807 3( ;25 .855 $025.852 $9.6v~ 31 355& 0.706 863 0.296 3026.407 3026.423 3026.663 3026o465 39°~96 32 356! 0.829 677 0.171 3026.6~6 3026.705 3026.769 3026.768 39.~96 33 3653 0.655 699 0.365 3028.861 3028.882 3028.918 3328.920 39.~96 36 3702 0 . 5 3 8 787 0.662 3 0 3 0 . 1 1 0 3 0 3 0 . 1 2 2 3 0 3 0 . 1 6 0 3030.15~ ~ 9 . ~ 9 ~ 35 3766 0 .727 84l 0.213 3031.175 3031.192 3031.223 3 0 3 1 . 2 2 1 39.~9o 36 3835 0 . 2 5 2 177 0 . 7 6 8 3 0 3 3 . 4 8 3 3 0 3 3 . 6 8 ~ ~ 0 3 3 . 5 2 7 3 0 3 3 . 5 1 8 3 ~ . ~ 9 6 37 3988 9 . 5 6 7 14 0.633 3037.363 3~37.376 3037.~00 3031.600 3 9 . ~ 9 6 38 ~108 0.255 811 0.766 3060.60~ 3060.611 3060.639 30~0.630 ~9.696 39 4156 0.719 607 0 . 2 8 1 3041.626 3061.661 3061.656 3061°65~ 39.69b ~0 ~291 0.762 862 0.258 3065.068 3065.065 3065.072 3o65.016 ~ t .~gb

FIDL 41 ** 4391 ¢* 0 . 2 3 1 9 0.769 3067.586 3067.588 3067.605 ** 3d~1.60~ 3 9 . ~ 6 6Z 6693 0 . 3 3 3 860 0 . 6 6 7 3055.266 3055 .~50 3055.271 3055.2o6 ~9.~96 63 4779 0 . 6 0 3 57 0 . 5 9 7 3 0 5 7 . 6 2 5 3 0 5 7 . 6 3 3 3 0 5 7 . 4 ~ 9 3 0 5 7 . 6 6 5 3 9 . ~ 9 6 66 ~863 0.778 9 0,222 3059.06~ 3059,066 ~059.068 ~059.07~ 39.~96

FIOL 45 ** 5165 *¢ 0.859 196 0 , | 6 1 3067 . ? | 5 3 0 6 7 , ~ 6 3067,266 *¢ 3 0 6 1 . 2 ~ $9.~96 66 5500 0.508 388 0.692 3075.711 3075.727 3075.738 307~.736 3~.696 67 5788 0.585 780 0,615 3082.966 3082.977 3082.~79 3082.911 39.~96 68 5816 0.687 612 0,313 3083.725 3083.761 30~3.738 3083.738 ~9.~96 69 5992 0.935 883 0.065 3088.189 3088,211 3088.195 3088.201 39.696 50 6126 0.186 819 0,816 3091.587 3091,590 3091.588 3091.576 39.~96 51 6171 0.263 888 0.737 3092.729 30q2.733 3092.727 3092.711 ~9.696 52 6197 0.925 232 0,075 3093.388 3093.410 3093.385 3093,39l 39.~96 53 6656 0.265 362 0 , 7 3 5 3 0 9 9 . 9 5 7 3099.902 3099.963 3099,932 39.6~b 56 6670 0.215 609 0.785 3100.312 3100.316 3100.297 3100.2~5 39.696

FIOL 55 ** 6686 ** 0.695 688 0,505 3100.666 3100.666 3100.666 *~ 3[00.666 3~.696

F~G. 1. Computer tabulat ion of wavelength results. Run No. 17 of the 40-run experiment calculating all wavelengths of lines with t ransmissions between 0% and 89.0%. The second to the last column, CWLC~L, is the best value of calculated wave- length based on an interpolated line posit ion and the use of all seven fiducial lines. The four th column labeled C'peakJ ' lists the corresponding t ransmissions on a 0-999 scale. The data in the columns on either side of this, FRACL and FRACR, are used for in terpola t ing line position. The wavelengths shown for lines numbered 1~ 7~ 17~ 28, 42~ 46, and 56 are handbook values. These are the fiducial lines. The spectral source was an i ron -a rgon hollow cathode.

sion. The reason for the la t ter is readi ly apparen t when two equal transmission values occur as a mini- mum. The line position is assumed to be half way between. Computer interpolat ion is based on three readings in the neighborhood of the minimum.

The above experiment produced 40 listings of 53- 58 lines each of calculated wavelength (except those used as fiducial lines), and the observed min imum transmissions. A variable number of lines were re- corded because of five borderline cases relative to the 89.0% transmission limit. Seven thousand read- ings for each run are recorded to include the spectral range under test, or a total of 280 000 readings. The computer t ime required for processing these data,

i n c l u d i n g p r o g r a m c o m p i l a t i o n a n d a l l o f t h e a b o v e

i n d i c a t e d c a l c u l a t i o n s , w a s 3 .51 r a i n .

F i g u r e 1 i s a c o p y o f o n e o f t h e f o r t y r e p l i c a t e

r u n s c o m p l e t e d b y t h e c o m p u t e r . T h e s e c o n d to l a s t

c o l u m n , C W L C F L , is t h e b e s t r e s u l t f o r c a l c u l a t e d

w a v e l e n g t h s b e c a u s e a l l s e v e n f i d u c i a l l i n e s ( F I D L )

w e r e u s e d as w e l l a s a n i n t e r p o l a t e d m i n i m u m t r a n s -

m i s s i o n p o s i t i o n a n d i n t e r n a l l y c o r r e c t e d d i s p e r s i o n

v a l u e s . T h e c o l u m n to t h e l e f t o f t h e a b o v e , C W L -

U F L , d i f f e r s f r o m C W L C F L o n l y i n t h a t t h e p o s i t i o n

o f t h e m i n i m u m t r a n s m i s s i o n r e a d i n g w a s u s e d f o r

c a l c u l a t i n g w a v e l e n g t h r a t h e r t h a n a n i n t e r p o l a t e d

p o s i t i o n . T h e n e x t c o l u m n t o t h e l e f t , C W L C , w a s

c a l c u l a t e d n o t o n l y w i t h t w o f i d u c i a l l i n e s b u t w i t h

APPLIED SPECTROSCOPY 5 0 9

fractional reading number interpolation, and for the column to the left of that, CWLU, one step further simplification is obtained because whole reading num- bers were used. The column labeled PEAK gives the minimum transmission (zero for opaque to 999 for clear plate). The corresponding reading number is shown under READ. NO. The column to the far right shows the dispersion in readings per ~, recal- culated for each pair of fiducial lines. These lines are indicated at the extreme left in Fig. 1.

Table I summarizes results of the above experi- ment for five spectral lines. The average wavelengths of 40 replicate runs are compared with handbook values, the maximum spread in wavelengths are

shown , and the standard deviations are calculated. These are shown for calculations based on the mini- mum-transmission-reading numbers as well as calcu- lations based on interpolated minimum transmission positions. Table I also shows the average percent transmissions for the same five lines, the maximum spread in percent transmission for 40 replicate runs, standard deviations, and coefficients of variation.

These results indicate that the performance of the system is limited by factors of microphotometry which were arbitrarily selected and are deemed suf- ficiently precise for the stringent requirements of trace element (chemical) analysis of complex spectra. Figure 2 is a computer-prepared report of analysis of a typical geologic material (granite, sample G-l) ~ for 18 elements. For comparison the "best values" from the above reference are hand written in Fig. 2. The computer input consisted of magnetic tape re- cordings of iron spectra for dispersion and emulsion calibration purposes, spectra of an 18-element, 19- member reference series in which the concentrations of each of the elements varied from 0.0001%-0.1%, a spectrum of the base material used for making the reference standards, and two spectra of G-]. The computer input also consisted of the program on IBM cards and a wavelength list of fidueial, emulsion calibration, and analytical lines. The production of the analytical report (Fig. 2) by the computer from the above input is the culmination of many factors and operations: the preparation of accurate mag-

E M I S S I O N SPECTROGRAPHIC ANALYSIS

J l l l H I S I l l l t l S J J l Q S I J S R U 8 B I 8 8 1 8 8 8 1 1 1 B 8 t i l l 1 8 8 8 1 8 1 8 8 1 J t l 1 8 6 8 8 1 8 1 1 1 1 1 J l J J 8 1 ! 1 5 t R 8 1 8 8 t l J J 8 8 t 8 8 R 8 8 1 8 8 1 t B q t 8 B 8 8 1 0 8 8 B J g i J B I J l l

PROGRAM NO. 10 -REQUESTOf l - REPORT NO. PLATE NO. | 2 3 4 5 JOB NO. GROUP NO. 23 SAMPLE MO. SPECTRUM NO. 26 DATE: 6 1 2 T / 6 8 F I E L D ~ 0 .

G A . . I O . O O O t ) . . . 0.00266 ( o O O l ~ SM . . . . . . . . . . . . . 0 .0 THESE ARE APPROX. ANOUNT GO . . . . . . . . . . . . . O.O S N . . I O . O 0 0 1 ) . . . 0 . 0 0 0 2 0 4 ( , 0 0 ~ )

USUALLY D E T E C . PRESENT I N GE. . . . . . . . . . . * * 0 . 0 SR . . . . . . ° . . . . . . 0 . 0 M&JOMS L I M I T S PER CENT H F o . . ° . . . . . . . . . 0 . 0 TA° . . . . ° ° . . . . . . 0 . 0

• . . .w= • i . = .~ . =" - - - - • = : -= - HG* * . . . . . . . . . . . 0 .0 T B ° . . . . . . . . . . . . . 0 . 0

ST. . . . . . . * * . . . . 0 .0 HO. . . . . , . , . . . . . 0 .0 TE. . . . . . . . . . . . . 0 .0 ~ L . . . . . . . . . . . . . 0 .0 IN . . . . . . . . . . . . . 0 .0 T H . , . , . . . . . . . . . ~ . 0 FE . . . . . . . . . . . . . 0 . 0 IR . . . . . . . . . . . . . O.O - _ _ TL . . . . . . . . . . . . . O.O MG°, . . . . . . . 0 . * * O.O LA* . IO.O00681)* O.Ol03 ~. 0 # ~ TM . . . . . . . . . . . . . 0 . 0 C A t . . . * * * * . . . * * 0 .0 L | * . * * * * . * . . . . * 0 .0 U . . . . . . . * * . . . . . ~ o 0

NA . . . . . . . . . . . . . 0 .0 LU . . . . . . . . . . . . . o .o v . . . ( o . ~ o l ) . . , n.oozos ~ 0 0 / ~ K . . . . . . . . . . . . . . o.o MD..(O.O001).., 0.000~07 ( , 0 o 0 7 ) w . . . . . . . . . . . . . . o .o T | , . . . . . . . . . . . . O.O NB.. (O.OOOZI5) . O , O 0 1 q 3 (,#0~14~) Y . . . . . . O.O P . . . . . . . . . . . . . . 0 . 0 ND . . . . . . . . . . . . . 0 .0 Y B . . ( O . OOOB~I). N (,~001# MN ............. 0 .0 NI . . I 0 , 0 0 0 2 1 5 ) . N (,O0#lJr) ZN ............. O.O

z R . . , o . o o ~ | . . . ~ . o z ~ (.o~0~

THESE ARE R O U T I N E L Y LOOKED FOR

m= = l w m 2 1 ~ g ~s wsss I w 8 ~mm s s l l

AGq. l O . O 0 0 ~ 1 . . . N " ( . ~ 0 ~ ) A S . . . . . . . . , , . . , O.O A U . . . . . . . . . , , , . 0°0 B . . . 1 0 . 0 0 0 4 6 4 ) . N ~dPOOlg) B A . . I O . O O 4 b 4 | . . O.OTOT ( , l ~ )

B E * * I O . O 0 0 | ) . . . 0 . 0 0 0 2 3 4 ( . 0 0 0 3 ~ CnMMENTS: BI * * . . . * * . . * * * . 0 . 0 C D * * . * * * * * . * * . . 0 . 0 C E * * * * * * . . . . . * * 0 . 0 C O . . l O . O 0 0 1 l . . . 0 . 0 0 0 3 0 7 ( , 0 0 0 ~

, . . , o . o o o , , , , . o . o o , , C U . . ( O . O 0 0 | I * * , 0 . 0 0 1 2 3 D Y . * * * * * * * * * * * * 0 . 0 E R * . * * * * * * * * * , * 0 * 0 ANALYSTS E U * . . * * * * . * , * * , 0 . 0

0 S , * * * * * * . * $ * * ~ 0°0

ps,.co.ooo21~J. (.ooqf) P D Q e . . . . ~ . . . . . , P R Q . o . * * o . * * . . , P T ~ . * * * * , . . . . . .

R E e . . . . * * * , . . . , R H . . o . . o . . * * . . , R U . . , * * . * * . . . . . SB . . . . . . . . . . . . . S C . . ( O . O O O I ) . . ,

O .0q504 f].O 0 .0 0 , 0

0.0 0.0 0 .0 0.0 0.000357 (oO~3)

NOTES: G " GREATER THAN UPPE~ L IMIT .

N = LESS THAN LDW~q L I M I T .

0 .0 = N ~ a N A L Y T I C A L CLJ~VE.

APPROVEOI (PROJECT LEADER|

FFIG. 2. Computer report of analysis. Geologic sample G-1. The numbers in parentheses following the element symbol are either the nominal limits of detection (calculated concentration of the related s tandard) or the lowest nominal concentra- tion in the series of standards. Where no detection limit is shown the s tandards for analytical curve da ta were not yet available. The concentration is reported as 0.0 for such cases. The best values f rom Ref. 2 are entered in the figure, hand writ ten for comparison.

510 Volume 23, Number 5, 1969

Table I. Precision studies and accuracy observations of some line wavelength calculations and their transmissions. [The five iron lines selected were free from interferences, were well distributed, and had a wide range of transmission. The averages, the maximum spread of values, and standard deviation were determined from 40 runs, the data having been recorded at the rate of 1000 readings per see.]

Calculated wavelength based on position of min imum transmission

Observed In terpola ted

Std. Std. Published Av. of Max. .De~. Av. of Max. Dev.

Wavelengtha 40 in A in ~ m A 40 in A in A in A

2957.365 2957.363 0.051 0.011 2957.363 0.021 0.005 3009.570 3009.566 0.025 0.007 3009.567 0.031 0.005 3018.982 3018.979 0.043 0.010 3018.978 0.027 0.005 3021.073 0321.047 0.032 0.009 3021.045 0.034 0.006 3083.742 3083.744 0.026 0.009 3083.745 0.014 0.005

Percent Transmission

Published Av. of Max. Std. Coeff. of Wavelength 40 Spread Dev. Var. %

2957.365 20.9 1.7 0 4 2.0 3009.570 43.0 1.9 .5 1.2 3018.982 71.0 1.8 .4 .6 3021.073 0.61 .2 .05 8.2 3083.742 61.1 1.9 .5 .8

G. R. Harrison, Wavelength Tables (The M.I .T. Press, Cambridge, Mass., 1939).

netic tape recordings of spectra, photographic plate calibration, automatic line finding and background selection, relative intensi ty calculation and back- ground compensation, analytical curve determina- tions, calculation of concentrations of the elements in the samples, averaging replicate analyses, and prepar ing a repor t in a prescribed format. One-half minute is given as a crude estimate of computer t ime required for analyzing a single sample recorded f rom a spectrum 25 cm long. A more exact t ime evaluation cannot be made until the completion of the analytical scheme for all elements listed and elimination of many details in the p rogram which are unnecessary for a routine analysis, but very useful for inter- mediate studies. A careful s tudy of all details of

Fro. 3. The mierophotometer. Part numbers are the same as for Fig. 5.

FiG. 4. Rear of mierophotometer. Part numbers are the same as for Fig. 5.

the analysis shown in Fig. 2 by conventional manual methods indicates that the automatic scanning and computer reduction of the data have not degraded the results either in limits of detection or accuracy.

I I I . E X P E R I M E N T A L D E T A I L S

A. Microphotometer

Figures 3 and 4 are two views of the micropho- tometer and Fig. 5 is a schematic of the optical sys- tem. The optical system is similar to that used by Steinhaus, Engleman, and Briscoe2

A commercially available comparator of the 1-~ accuracy class, having a high-precision lapped screw, supplies the basic measur ing and t ranspor t functions. The l-ram pitch screw of the comparator is driven by a 1/125 I IP , reversible, synchronous motor having a double-ended output shaft rota t ing at 300-rpm. An Oldham type coupling (less than 5 rain of arc backlash) connects the motor and comparator and a helical coupling (also low backlash) to the encoder. Conlmercially available couplings and encoder are used. The encoder is a photoelectric device provid- ing 200 accurately spaced pulses per revolution which, in connection with a six-digit, bidirectional counter supplies a visual measure of length (position in the spect rum). I t also supplies a square wave t iming signal for the purpose of t r igger ing discrete phototube readings to be made at accurately spaced points (5-~ separat ion) along the scan line of the photographic image.

The lamp housing of the microphotometer is iden- tiffed as No. 1 in Fig. 4. This l ight source, a 50-ram fl lens, and the adjustable slit are mounted in a cap (shown in Fig. 4, pa r t No. 3) of a tube with a 5.4-cm diam and 20 cm long. Fo r focusing and al ignment purposes this tube, which also carries the permanent ly mounted, 45-deg front-surface mirror , may be ad- justed longi tudinal ly as well as rotated. The re-

APPLIED SPECTROSCOPY 511

mainder of the optical system is on a vertical axis. Par t s 6-9 (Fig. 3) are rigidly mounted on a micro- scope-type tube. Lens 4 (Fig. 5) is mounted, axis vertical, just below the photographic plate. The up- per par t of the optical system, including the fixed slit, may be rotated, the position of the fixed slit is adjustable perpendicular to the optic axis, and the phototube horizontal position as well as orientation are adjustable-- to align and adjust focus, parallel- ism, superposition of images on the entrance slit, and spectral lines on the exit slit. A magnified view of the under surface of the exit slit (supplied by lens and mirror No. 10, Fig. 5) is essential for these pur- poses. The viewing lens (close to the objective--No. 11, Fig. 5) aids in locating position in the plate and initial spectrum alignment.

A fixed exit slit 0.15-mm wide and 12.7-mm long is used. For the optical system described, this slit, equivalent to 0.011-ram by 0.941-mm at the plate, is about half the width and half the length of the geometrical size of the spectral lines to be observed. These dimensions are a compromise: small for better definition and large for less relative electrical and photographic image "no i se" . Using the above slit width (0.011-ram, or 11 ~ at the plate) and taking readings for every 5 ~ of motion, results in about a 50% overlap of readings, i.e., each new reading" cov- ers an area of one-half new readings and one-half of the previous readings. The decision to take a

9

/ \

l__L_

th i j

II

A / \

Fro. 5. Optical system. (1)--Light source; (2)--50-ram fl lens; (3)--first silt, adjustable; (4)--32-mm fl lens; (5)-- photographic plate; (6)--16-mm fi microscope objective lens; (7)--second slit, ~xed width, adjustable position; (8)--54- mm fl lens; (9)--phototube; (10)--eyepiece for viewing underside of second slit (7); and (ll)--eyepiece for viewing image of first slit on the photographic emulsion.

Fro. 6. ¥iew of entire recording system.

reading for every 5 t~ of travel as did Steinhaus, Engleman, and Briseoe, was based oll considerations of the above line, slit, image sizes, resolution and line profile tests, and the desire to keep the total number of readings from becoming cumbersome.

A 1P21 multiplier phototube is used with cathode and dynodes supplied through ten 20 000-~ resistors in the conventional series arrangement. Highly regu- lated commercial dc power supplies are used through- out. These include a high-potential supply for the phototube, a low-potential supply for dark current and ground potential compensation, and a lamp- power supply. Two ten-turn poteutiometers in the phototube output circuit and a ten- turn potentiom- eter in the control circuit of the lamp power supply provide precise maximum and minimum settings for the transmission scale. These controls are shown in the center panel of Fig. 6, between the mieropho- tometer on the left and the tape recording system on the right.

B. Magnet ic Tape Recorder

The recording system was designed and built by an engineering company specializing in data aequisi- tion systems. ~ Commercial stock items included the preamplifier, analog-to-digital converter, sample-and- hold module, coincident current ferr i te core memory system, and an ineremental digital tape recorder.

A block diagram of the reeording system supplied is shown in Fig. 7. Discrete readings (samples of the multiplier phototube output having values from 0-999) may be recorded at any rate up to 1000/sec. This recording mode, permit t ing any speed up to the maximum, is necessary because of the variable speeds with start ing and stopping the scanning process. The square-wave output of the mierophotometer described above triggers each sampling-digi t izing-recording cycle. The multiplier phototube output is smoothed (averaged) throughout the sampling phase. The out-

512 Volume 23, Number 5, 1969

FIG. 7. Block diagram of the recording system.

put of the system is an IBM compatible, 7-track mag- netic tape recorded at a density of 800 bits per in. BCD conversion with sign is used. Manual data for identification, etc., may be entered between runs with 12 thumb-wheel switches. End-of-record and end-of- file gaps are automatically inserted. Recording may be started and stopped with manually operated switches or from remote contact closures at the micro- photometer or counter.

A single spectrum is recorded as one " f i l e " and may consist of 50 000 readings for 25-cm length. This is subdivided into groups, or records, of any number of readings between 100 and 1000 as deter- mined by a set of thumb-wheel switches. When the number of readings selected for the record length has accumulated in the buffer they are t ransferred to the tape without in ter rupt ing the storage of readings arriving from the microphotometer. The insertion of the buffer permits fulfilling computer format re- quirements while maintaining constant scanning velocity. Because of the above capability it is not necessary to record the wavelength coordinate, which would require three times more data recording and processing.

The subsequent discussion is restricted to one recording speed, 1000 readings per sec, and to record lengths of 500 readings each. A 25-cm length of spectrum (10-in.) will require about 100 records which as a group constitute a file. Each file, in addi- tion to the transmission data, has one or more short records, manually inserted with the thumb-wheel switches, of information for identification and for program direction.

C. Data Handling

1. Line Finding

For applications to spectrochemical analysis of geologic material, 5,6 spectra are recorded with grat ing spectrographs (5.0 A/ram) on two 4 - × 10-in. plates covering the spectral region 2250 to 4750 A. Only the lower half of this spectral region is discussed in this report. In principle the handling of the data for the other half is identical.

A single plate may have recorded on it as many

as 40 spectra, including two iron spectra for plate calibration which are exposed simultaneously through a two-step filter. All spectra have superimposed on them a single cadmium line, 2288.02 A. This provides a convenient mark for setting the start ing point for the microphotometer scan and also the initial fiducial line for the computer program. Accurate setting of the microphotometer on the line is not required but ra ther at a conveniently observed point in the spec- t rum just preceding the line. To conserve computer data storage a complete spectrum (a file) is processed by using groups of 12 records simultaneously and moving through the spectrum by adding four new records and dropping the four oldest as processing continues.

The desired cadmium line is obtained with a dis- charge lamp (Osram type) and is isolated by a nar- row band filter. The lamp and filter are mounted on a swivel on the optical bench between the slit of the spectrograph and the arc source. Preceding an ana- lytical exposure, the operator swings the cadmium source into the optic axis for a 10-sec exposure.

In addition to the superimposed cadmium line, iron lines are used for fiducial points in the line- finding program. One percent iron added to the elec- trode charge supplies a large number of such lines. This is accomplished by including iron oxide with the graphite or matr ix-type diluent of the samples. Many geologic materials contain iron. This method is workable even if the initial iron content is as high as 15% (fiducial line images not too broad).

Trace-element analysis requires positive determina- tion of the presence or absence of certain lines. The line-finding procedure therefore is critical and iron is used because it supplies a good fiducial line every 10 or 12 A. Although an element other than iron could be used, the selection would depend upon the spectral region, required accuracy, etc. A list of wavelengths of these fiducial lines is par t of the input data, just prior to the tape data. The identification of the first line, and an initial dispersion value accurate to one par t in 400 completes the basic start ing requirements.

The line-finding logic for fiducial lines and emul- sion calibration lines is different f rom that used for finding analytical lines and is applied first. The first line in a file having a transmittance under 80% is the Cd 2288.02 A line. The position of the next fidu- cial line, Fe 2320.26 A, is estimated by mult iplying the wavelength difference between the cadmium and iron lines by the initial dispersion and adding this product to the cadmium line position, reading five transmission readings on each side of the estimated position, and selecting the number of the reading which represents the minimum transmission. The next fiducial or plate-calibrating line in the list is then located in a similar way from the accurate posi- tion of the preceding fiducial line. The initial dis- persion is replaced by a calculated dispersion a number of times as this line-finding procedure ad- vances through the spect rum-- the number of times

APPLIED SPECTROSCOPY 5"13

being determined by the dispersion curve and re- quired accuracy.

In tegra l with the line-finding procedure for fidu- eial lines and plate-calibration lines is the definition of a " l i n e " beyond the requirement of a simple transmission minimum. Fo r this purpose spectral background is defined as the average of the trans- missions of the nearest dip (maximum transmission) on each side of the peak (minimum transmission). I f the difference between the peak and background is less than 1.5%, the line is not used as a fiducial line and is skipped over. The estimated position of the next line, using the appropr ia te dispersion value, is referenced f rom the last good line, and this proce- dure continues unti l the last line is located. All fidu- cial and calibration lines are found f rom the two-step plate-calibrat ion spectra before processing the spec- t ra for chemical analysis. The plate-calibration lines are selected on the basis of freedom f rom interfer- ence, repeatabil i ty, and distribution of transmissions.

A different line-finding procedure, is used for lines of the elements sought in the chemical analysis. This is done to enhance reliabil i ty at or near the l imit of detection. The element lines are included in the same list re fer red to in the above section. They are dis- t inguished f rom fiducial and plate-recalibrat ion lines by an accompanying pr ior i ty number, which also serves to designate pr ior i ty if mult iple lines are listed for a single element. The computer starts the processing with cadmium line as it does for calibra- tion spectra. Fiducial lines are located as described above, skipping over the lines designated as " a n a l y - s i s" lines by the pr ior i ty number. Af te r locating two good fiducial lines bracketing any analytical lines, the computer then determines the spectral positions for the analytical lines. These are calculated f rom both the leading and following fiducial lines. The two results are averaged, rounded off, and one reading on each side of this estimated position is the range ex- amined for a " p e a k " . Thus, only three readings are examined, tha t is, a wavelength range of 0.075 A. The min imum transmission, the peak, in the group of three readings is selected for fu r the r calculation.

Spectral background may be defined in many ways. The following method is used to i l lustrate the feasa- bili ty of a complex procedure ra ther than to suggest a specific procedure. Background is evaluated by de- termining the coefficients of the s t ra ight line connect- ing the max imum transmission on one side of the line within 100 readings (2.5 A) with the maximum transmission on the other side within the same dis- tance and solving for the background value at the position of the peak (min imum transmission). Da ta f rom the tape are stored within the computer in such a way as to assure a min imum of 100 readings before the first fiducial line and 100 readings af ter the last fiducial line. Fo r every analysis-element line, a background transmission associated with the line transmission is determined. These two transmissions

are converted to intensities before taking the differ- enee.

2. Wavelength Calculation

A test of the overall wavelength accuracy of the system consists of calculating the wavelength of all observed lines (a rb i t ra r i ly l imited to lines between 0 and 89.0% transmission) in a selected spectral region and compar ing them with accepted vMues. A table of fiducial lines is located as previously de- scribed. All readings between each pa i r of good fiducial lines are examined and whenever a min imum transmission is found it is kept t emporar i ly as a line. The nearest max imum transmission on each side is then found and the lesser of the two is also stored temporar i ly . I f the difference between this t rans- mission and the one above, t emporar i ly kept as a line, is 1.5% or greater, a line is defined. The wave- length is repor ted as the average of the two calcu- lated f rom both fidueial lines.

3. Calibration Curves

The calibration curves include the dispersion curve, element-concentration calibration curves, and the plate-emulsion calibration curves.

Ten of the fiducial lines used in a two-step iron ex- posure (placed on each plate for calibration pur- poses) are used to eMculate a dispersion which is expressed as the number of readings per A. Asso- elated with each of these dispersions is a wavelength which is the average wavelength of the two lines used in the calculation. Since there are two spectra, 20 points are obtained which are fitted as a s t ra ight line giving coefficients for a dispersion-wavelength relationship which is good for jumps up to a maxi- mum of about 50 A. Af te r the second iron spectrum, the coefficients of the dispersion curve are determined and are used for all remaining spectra calculating a new dispersion every 50 A. Wi th in the distance there are 8-10 fiducial lines at least half of which will be good. Therefore, the actual jump will rarely, if ever, be over 25 A. In actual practice these jmnp distances are kept to a reasonable min imum so tha t the estimated position of the next line will be accu- rate, thus taking full advantage of the entire length of the search area. The first dispersion used is the initial dispersion which is i n t roduced ' within the fixed data information. This is used in the unfiltered step of the two-step iron spectrum unti l the first dispersion is calculated. At this point, the newly calculated dispersion is used unti l the next dispersion is calculated, and this procedure of using the newly calculated dispersion continues through the spectrum.

The coefficients for one analytical curve are deter- mined for each spectral line which will be used as an analytieM line for samples. The abscissa is the nat- ura l logari thm of concentration expressed in percent of the element and covers a range up to three orders of magnitude. The ordinate is the na tura l logar i thm

514 Volume 23, Number 5, 1969

Table II . Analyt ical lines selected and used for the example de- scribed in the text.

Element Wavelength in £

Ag 3280.68 3382.89 B 2497.73 Ba 3071.57 Be 2348.61 Co 3453.51 Cr 3021.56 Cu 3273.96 3247.54 Ga 2943.64 La 3245.12 Mo 3170.35 Nb 3163.40 3194.98 Ni 3414.76 Pb 2833.07 2663.17 Sc 3353.73 Sn 2839.99 V 3183.41, 3183.98, 3185.40, 3102.30 Yb 3289.37 Zr 2752.21

of the ratio of the intensi ty of the peak minus its backg round and the intensi ty of the internal stan- dard minus its background. I f no internal s tandard is used the denominator of this ratio is set equal to 1.0. All these curves are fitted as first or second deg curves, whichever is better, by polynomial regression. One of these curves is associated with each line which is uniquely identified by a pr ior i ty number and its atomic number. Each line also has predetermined upper and lower concentration limits assigned to it.

Five plate calibration curves ( P E C C ) are used, covering five areas along the plate, to represent the plate emulsion characteristics. Each curve is used for a distance of 250 3~. The abscissa of the pre l iminary PECC is the na tura l logari thm of the transmission f rom the unfiltered step, and the ordinate is the natu- ral logari thm of the transmission f rom the filtered s tepJ The pre l iminary P E C C in each case is fitted as a quadrat ic with the two " e n d s " being weighted. The coordinates of the weighted points are (99, 100) and (1.0, 1.5). Several of the raw data points overlap in each section. A total of 115 lines are used to obtain 26 points for each of the five pre l iminary PECC ' s . These points are fitted by polynomial regression methods and an analysis of variance is made in every case. The second degree equation thus obtained where y = l n TFiL and x = l n Tv~piL is of the form

y----Bo-t-B1 ( x ) + B 2 (x) ~

Points obtained f rom this curve serve as y-compo- nents for points on the final PECC and the logar i thm of the line intensi ty values serving as x components.

The first point is chosen in such a way tha t it is the inflection point of the " S - s h a p e d " final P E C C and lies on the s t ra ight lines connecting the nearest point on each side of it.

Al though the final P E C C ' s are all " S - s h a p e d " they cannot be adequately represented by a cubic equation. More fitting would have to be accomplished which means the addit ion of more residuals and more estimates. Problems of control are also involved. The weighting of the added uppe r and lower points of the pre l iminary curve and the selection of a large number of lines having well-distributed transmissions assure a proper ly oriented and well-controlled pre- l iminary curve but one which is still representat ive of the data. Weight ing of the individual t ransmis- sions under these conditions does not appea r to offer any advantages. A method has beeen devised to handle this set of points for the final P E C C which results in a continuum of points having an exact fit. The fit t ing and smoothing errors are thus lef t (as they should be) for the p re l iminary PECC, which is always closely represented by a quadrat ic equation.

Ln relative intensity

I I I I 0 .00046 percent

I I I I

f - -

g

0.046 percent --'NO 1 t 1 I I I I J

FiG. 8. Analytical curve for B 2497.73 A between concen- trat ions of 0.00046% and 0.046%. Circles represent computer selected readings with conversion to relative intensi ty and background correction. Solid dots represent computer-calcu- lated quadratic equation results.

APPLIED SPECTROSCOPY 515

Ln r e l a t i v e i n t e n s i t y

I I I [ I I I I I I

f -

o

0 ~ 010001 p e r c e n t

\ o

0

0 0

0.0032 p e r c e n t

I I I I I 1 I I . I I

FIG. 9, Analytical curve for Co 3453.51 X between concen- trations of 0.0001% and 0.0032%. Circles represent computer- selected readings with conversion to relative intensity and background correction. Solid dots represent computer-calcu- lated quadratic equation results.

Complete details on computer programming and all calculations have been prepared by F. G. Walt- hall (unpublished data) .

Ln r e l a t i v e i n t e n s i t y i i i I I I I v

o o.ooQI p e r c e n t

o

8"

o o

I I t I I I I I

FIG. 10. Analytical curve for Be 2348.61 A between concen- trations of 0.0001% and 0.01%. Circles represent computer- selected readings with conversion to relative intensity and background correction. Solid dots represent computer-calcu- lated quadratic equation results.

D. Anolysis of G-1

In establishing a procedure the computer may aid in the selection of analytical lines. In the example given (rock sample G-l) a list of 51 potentially useful lines was supplied by the computer, converted to relative intensities, corrected for background, and the analytical curves were determined as explained earlier. A print-out of all pert inent data includes for each line a table of actual relative intensities, as calculated from the final PECC, and calculated rela- tive intensities obtained from the equation of the analytical curve. These data when plotted supply a graphic means of judging the quality of usefulness of a line and the concentration limits of application. Other lines are removed because of interferences which will be automatically corrected after fur ther s tudy and completion of an analytical procedure. Table I I shows the lines selected as the result of such a study. In addition to specific concentration limits for each line, other general limitations were imposed on the system: (1) lines having transmissions less than two percent were not used, (2) data were used only if the line-background difference exceeded one percent transmission, and (3) the vertex of the pa- rabola could not be exceeded. Three analytical curves are shown in Figs. 8, 9, and 10. For most analytical

I 0 0

20C

30C

40C

500

600 i i

i

i i i i i

x

i

i i

i

/ i 8 O0 X/

900 /X

1000

FIG. 11. Fe 3024.03 X line recorded (1) stationary reading (solid line with circles) and (2) 1000 readings per sec (dotted line with × 's).

516 Volume 23, Number 5, 1969

curves the observed values (circles) show little scat- ter f rom the computer calculated points (solid dots) as i l lustrated for boron in Fig. 8. The cobalt line shown in Fig. 9 i l lustrates an extreme case of scatter and is shown as an interesting demonstrat ion of the results of the curve-fit t ing procedure. Measurements of all scattered points with conventional micropho- tomet ry g'ave assurance tha t the errors preceded the automatic scanning" and computer reduction of the data.

F igure 10, the analytical curve for beryl ium 2348= 61 A is included to i l lustrate a curvature inverted f rom tha t of Fig. 8 and 9. The analytical curve for Be also suggests tha t an equation of higher degree than the quadrat ic may result in a bet ter fit. I t is cur- rent ly be]ieved more expedient to retain the quad- rat ic equation and, instead, l imit the concentration range for a par t icular spectral line as required.

E. Photometric Response

The overall capabil i ty of the microphotometer and recording system is demonstrated in Figs. 1 and 2 and Table I. Af te r sat isfactory performance was indicated summary tests were made for (1) posi-

0

I 0 0

2 0 0

3 0 0

4 0 0

g 8 0 0

z

Ic $ 0 0

7 0 0

8 0 0

900

1 0 0 0

/:

*X X W ,X 'K "~'-X..~,K

FIG. 12. ~e 3025.64 X and Fe 3025.84 X lines recorded (1) stationary reading (solid line with circles) and (2) 1000 readings per sec (dotted line with X's).

5oc

z 6oc

m

/ o : x

FIG. 13. Fe 3100 X group of lines recorded (1) stationary observations (solid line with circles) and (2) 3_000 readings per sec (dotted line with ×'s) .

t ional accuracy, (2) retent ion of all readings, (3) heat ing effects f rom tu rn ing the precision screw at 5 rps, (4) spectral resolution, (5) line profile or defi- nition, (6) transmission precision, and (7) t ime re- sponse of the photometr ic system. The last item, of special interest because of the recording speed used, is i l lustrated in Figs. 11-13 in which the transmis- sions for several iron lines are plotted. Superimposed on values obtained at the ra te of 1000 readings per see are manual readings. The la t ter are determined by manual ly tu rn ing the motor shaft, one counter number at a time, and recording percent transmis- sion shown on the A - D converter. Manual reading cannot be precisely spaced relative to the convert signal. For this reason the manual readings plot ted in Figs. 11, 12, and 13 are averages of three deter- minations. The high speed values are f rom single determinations.

In Fig. 11 the difference in slopes for the two speeds both on the rising and fal l ing sides is evidence of some lag in photometric results for the high speed mode compared to the near ly static mode. The lag is also noted for par t ia l ly resolved lines such as 3025.64 and 3025.84 A shown in Fig. 12. The 3025.64 3~ min imum to max imum transmission distance be- tween the two lines is less for the high speed than for the stat ionary. These lags are small and are in- significant for comparat ive methods. The 3100 fk iron lines presented in Fig. 13 fu r the r i l lustrate resolution and definition of lines obtained with the system described, in which over lapping slit observations are taken every 5 ~, at the rate of ]000/sec.

APPLIED SPECTROSCOPY 517

1. A. W. ttelz, U. S. Geol. Surv. Profess. Papers 525-B, B16O (1965).

2. M. Fleischer, Geochim. Cosmochim. Acta. 29, 1263 (1965). 3. D. D. Steinhaus, R. Engleman, Jr., and W. L. Briscoe,

Appl. Opt. 4, 799 (1965). 4. Redcor Corp., Canoga Park, Calif. 91305.

5. It. Bastron, P. R. Barnett, and ~. J-. Murata, U. S. Geol. Surv. Bull. 1084-G, 165 (1960).

6. A. T. Myers, R. G. Havens, and P. J'. Dunton, U. S. Geol. Surv. Bull. 1084-I, 207 (1961).

7. C. E. Harvey, Spectrochewical Procedures (Applied Re- search Laboratories, Glendale, Calif., ]950), Chap. III.

COPYRIGHT AND PERMISSION

This Jou rna l is fu l ly copyrighted, for the protect ion of the authors and their sponsors. Permission is hereby gran ted to any other authors to quote from this journal , provided tha t they make acknowledgment, in- c luding the au thor ' s names, the Jou rna l name, volume, page, and year. Reproduct ion of figures and tables is likewise permi t ted in other art icles and books, provided tha t the same informat ion is p r in ted with them. The best and most economical way for the author and his sponsor to ob- ta in copies is to order the ful l number of r ep r in t s needed, at the time the art icle is pr in ted , before the type is destroyed. However, the author, his organization, or his government sponsor are hereby g ran ted permission to reproduce pa r t or all of his material . Other reproduct ion of this Journa l in whole or in par t , or copying in commercial ly publ ished books, periodicals, or leaflets requires permission of the editor.

518 Volume 23, Number 5, 1969