1132
ANALYTICAL CHEMISTRY
degree 6f substitution values clicckcd closely those obtained by other standard methods. It may also be concluded that duplicate determinations should agree within 0.5 t o 1.0% sodium carboxymethylcellulose for assay value, and within about 0.03 unit for degree of substitution. The method is not particularly suited for control work becausc of its length. It is estimated that approsimately 6 to 8 hours are required to run a set of up to four det'erminations. However, it has a definite use as a referee method or in other analytical work not of the control type. The assay values obtained by the copper salt precipitation method on purified grades of sodium carboxymethylcellulose agree reasonably well with those found by an alcohol-washing procedure currently employed in the trade. The chief source of error in this method is an appreciable loss of sodium carbosymethylcellulose through solubility in the ,wash liquor. The estent of this loss varies and is mainly dependent on the degree of s u b stitut,ion, degree of degradation, and uniformity of the product. Salt impurities insoluble in the alcohol will, of course, interfere. Further evaluation of the copper salt method may result in improvement in time required an'd accuracy. If an organic base, which does not give a precipitate with copper ions, could be found for the adjustment of the final pH, the danger of precipitat.ing copper hydroxide could be obviated and the solution made more basic to ensure complete precipitation. It may also be possible to precipitate the copper salt in a suitable physical form in some organic solvent-aqueous copper salt solution which would elimi-
nAte pH adjustment. These possibilities will be considered in future investigations. ACKSOW LEDGMENT
The authors wish t o express their appreciation to R. T. Hall of the Analytical Methods Development Group, Hercules Powder Company, for his assistance and advice throughout the course of this work. LITERATURE CITED
(1) Brown, C. J., and Houghton, .4. A,, J . SOC.Chem. I d . ,60,254T (1941). (2) Chowdhury, J. K., Biochem. Z., 148, 76 (1924). (3) Eyler, R. W., and Hall, R. T., Paper Trade J., 125, 59 (1947). (4) Eyler, R. W.,Klug, E. D., and Diephuis, F., ANAL.CHEM.,19, 24 (1947). (5) Hoffpauir, C. L., and O'Connor, R. T., A m . Dyestuf Reptr., 31, 395 (1942). (6) Hollabaugh, C. B., Burt, L. H., and Walsh, A . P., Ind. Eng. Chem., 37, 943 (1945). (7) Penn, W.S..Point M Q ~ u J '16, . , 127 (1946). (8) Pierce, IT. C., and Haenisch, E. L., "Quantitative Analysis," 2nd ed., New York. John Wiley & Sons, 1940. (9) Reid, J. D., and Dad, G. C., Tertile Research J . , 17, 554 (1947). (10) Ibid., 18, 551 (1948). (11) Sakurada, I., J . SOC.Cht.77~Ind. Japan, 31, 19 (1938). (12) Samsel, E. P., Bush, 5. H., Warren. R. L., and Gordon, A. F., ANAL.CHEM.,20, 142 (1948). (13) Shax, E. H., Jr., Proc. S . Dakota Acad. Sci., 25, 57 (1945). RECEIVED .4pril 12, 19.50.
Spectrophotometric Determination of Anthracene in Crude Anthracene Cakes Method of Correcting for Extraneous Background Absorption FRANK P. HAZLETT, ROY B. HANNAN, JR., AND JOSEPH H. WELLS Mellon I n s t i t u t e , P i t t s b u r g h , P a . An ultraviolet method for determining anthracene in anthracene cakes is described that is at least as accurate as any known chemical method and much more convenient. The results are within about 1% of the known concentration in synthetic cakes. Spectrophotometry has also been applied to the analysis of mixtures of anthracene, phenanthrene, and carbazole. A method of correcting for extraneous background absorption is outlined which assumes only ,that the background is linear. In particular, it is unnecessary to know the intensity of the extraneous absorption. The method is similar to that of Banes and Eby.
A
STHRACENE occurs in coal tar in conrentrations of less than 1%. When coal tar is distilled, the anthracene concentrates in the anthracene oil fraction distillihg between 300 ' and 400" C. This fraction on cooling deposits a mixture of solids which may be centrifuged to yield an anthracene cake containing from 10 to 25% by weight of anthracene (the so-called "crude cake"). Other major constituents of this cake are carbazole and phenanthrene. T o evaluate properly a commercial method for the recovery of anthracene, the analysis of the starting material, the treated material at intermediate stages, and the purified product must be knowp. The chemical properties of the many compounds present are so similar that a chemical analysis is difficult, Two chemical procedures are available for the determination of anthracene. I n the Hochst test the anthracene is oxidized t o anthraquinone aLid the amount of anthraquinone is determined
gravimetrically (9). In the dienometric or maleic anhydride method anthracene reacts with a known amount of maleic anhydride and the excess anhydride is then determined by titration ( 4 , 5, 8 ) . Both procedures require esperienc,ed technicians and a considerable amount of time. Furthermore, their absolute accuracy is open t o question when they are used in the analysis of crude anthracene cakes containing less than 40% anthracene. I n the maleic anhydride method, for example, low results obtained are apparently due to acidic components of the cake. There is no good chemical proc'edure available for the determination of phenanthrene. Carbazole in anthracene cakes may be estimated by a Kjeldahl nitrogen analysis. The use of ultraviolet absorption spectrophotometry for the analysis of anthracene overcomes the disadvantages of the chemical procedures and is of equal or greater accuracy. I n addition to the determination of anthracene, the ultraviolet method permits
V O L U M E 22, NO. 9, S E P T E M B E R 1 9 5 0
1133 samples of these compounds were determined from the LambertBeer law :
.4 = logla Io/I = Ed
(1)
where A = absorbance, Io/I = ratio of radiant intensity incident on sample to that transmitted by the sample, E = extinction coefficient, c = concentration in grams per liter, and E = cell length (1.000 cm. in all cases). Table I shows these extinction coefficients a t various wave lengths for this instrument. The validity of the Lambert-Beer law for anthracene was tested by plotting concentration against absorbances for the 358 and 376 mp bands. There is a slight deviation from linearity when A > 0.9, which is almost certainly due to too u-ide slit widths. Consequently, all analyses were made at absorbances below this vdlue.
--- 0 0 2 5 g p l A n t h r a c e n e 0 0 2 5 9 p l Phenanthrene - 00209' p I Carba7ole
--
PURIFICATION OF REFERENCE COMPOUNDS
-__ I
'L.
-./--
I
290
I
310
\ 1
33 0
1
350
1
I
I
'\,
370
390
W A V E L E N G T H rnp
Figure 1. Absorption Spectra of Anthracene, Phenanthrene, and Carbazole Chloroform Solutions
the identification and in many cases the determination of the impurities present. The progress of thb separation of anthracene from crude anthracene cakes in an industrial process can be followed quickly and routinely by means of the general picture of the constitution contained in the ultraviolet spectrogram. APPARATUS AKD SPECTROSCOPIC PROCEDURES
A Cary double-beam ultraviolet spectrophotometer was used throughout this investigation. This instrument records absorbance vs. wave length. Tests on standard solutions of potassium dichromate gave extinction coefficients which agree within 1% with those given by Hogncss, Zscheile, and Sidwell (3). Absorbance readings are reproducible within +0.01 unit. At the absorbances used in this work the error in the analysis due to photometry is estimated to be within +=2% of the amount present. The spectra of anthraccne, phenanthrene, and carbazole in chloroform solution are shown in Figure 1. It will be observed that the spectra of carbazole and phenantherene interfere seriously with one another in this limited spectral region, but not a ith that of anthracene. Extinction coeffisients for highly purified
Table I.
Extinction Coefficients of Pure Compounds (in Chloroform Solution)
Wvve Length, M#
(Liters per gram-cm.) Anthracene, Carbazole, E E 39.0 0.0 42.7 0.0 18.4 0.4 22.9 16.7 12.4 19.6 11.9 16.7 8.8 22.5 2.3 67.0 1.9 101.0 2.9 €2.7 6.0 23.7
Phenanthrene, E 0.0 0.0 1.5 1.3
0.8 1.7 0.9 70.0 39.9 58.8 78 9
Anthracene. A Reilly Tar and Chemical Company samplr of anthracene was recr stallized from pyridine and msde to react with maleic anhydriz, and the anhydride adduct was hydrolyzed to the potassium salt. The alkaline-water solution of the acid adduct was washed with xylene at 90" C., and the adduct was recovered in the form of the anhydride. This step was repeated until the ultraviolet spectrum of the adduct was transparent in the 270 to 400 mp range. The adduct was then heated under vacuum to recover the anthracene. The anthracene was chroniatographed on alumina and the eluent evaporated to yield the p6ifikd anthracene. Phenanthrene. A Reilly Tar and Chemical Company sample of Dhenanthrene was recrystallized from dichloroethane and made to react with maleic "anhydride. The anthracene-maleic anhydride adduct was removed by repeatedly washing the benzene solution of the reaction mixture with 5% aqueous potassium hydroxide a t 90" C. The benzene was then cvaporat,ed and t,he Phenanthrene chromat,ographed on alumina. Carbazole. A Reilly Tar and Chemical Gompany sample of 977, carbazole was recrystallized from a benzene-acetone mixture ( 6 to 1) and the 9-acetyl carbazole compound formed. The 9acetyl compound was repeatedly rerrystsllized from acetic acid and hydrolyzed to regenerate carbazole. ANALYSIS OF CRUDE AUTHRQCENE C.AKE
The crude anthracene cake obtained by centrifuging anthracene oil is a mushy mass containing 10'to 25% anthracene and 35 to 50% of oily matter in addition to carbazole and Phenanthrene. The spectrum of a typical crude cake (Figure 2) shows the presence of background absorption in the 350 to 390 mp range. This absorption, which is caused by substances other than carbazole and phenanthrene, necessitates correction of the spectral data before analyzing for anthracene. Because the background absorption increases sharply a t shorter wave lengths, it is impossible to determine carbazole or Phenanthrene in the crude cake. Base-Line Method of Background Correction. The simplest type of background correction is to assume that the extraneous absorbance is linear. In what is often termed the "base-line" method, illustrated in Figure 3, a straight line is drawn between points on the absorbance curve a t two fixed wave lengths ( 1 , 2, IO), which are conveniently taken as the positions of minima for the pure compound on each side of the band to be used. I t will now be shown that the height, A , , from this base line to the band maximum is directly proportional to the concentration of the material present. Furthermore, the concentration of the material can be obtained without knowing anything about the background except that it is linear-in particular, without knowing its absclute amount. These results have appeared in a paper by Banes and Eby ( I ) , which seems to have escaped general notice. The present authors first derived them in a slightly different form that is somewhat easier to use with their instrument. I n view of this fact and of the utility of the method, it seems worth while to give the mathematical proof. A paper by Morton and Stubbs ( 6 , 7) also describes a method that is similar in principle, but much more complicated in practice,
ANALYTICAL CHEMISTRY
1134
Consider first the case of one pure component (substance D). From Figure 3 it can be seen that the slope of bhe base line may be eypressed rn
AA= - -At - Aa _ AX XJ - At
- Ai - (A, - A,) XI -
Substituting these expressions in Equat.ion 2 : 8.
=
(n - I)(klcD
+ gxi f h )
f ( h C D
f BXz
+ h ) +- gx: + h )
n(4c~
Separating terms:
Rearranging:
A=
=
[(?I - 1)klCD
A+ = [(n - 1)kl Let
$- kzCD - nkZCD] f [(n -. l)(gXt h)
+ + gXn + h - 4 g X I + h)l + kz - nkt] + (0) CD
But this equation is identical with Equation 3. Because n and the k's are constants, this shows that the value of A, is independent of the background as long RS the background is linear. One can also write:
Then and
kerf. = ( n
. CD
=
-
A*
( n - 1)k
l)kl
+ k,
- nkt-
Az
+ k, - nk3 - -k f r ,
(4)
(5)
where k,ff. is an effective extinction coeffirient. It follows from Equation 5 that the concentration of the desired component (substance D ) can be determined by the measurement of the absorbance of the mixture a t three fixed wave lengths, without knowing anything about the background exrept that it is linear. (Although only one absorbance appears explicitly in Equation 5, two more are needed to fix the base line for the measurement of A , . See also Equation 2 . ) From the foregoing considerations the following points may be noted :
I
I 310
I
I
I
330
I
I
I
350
370
I
I
J
390
WAVELENGTH rnA
Figure 2.
Absorption of Crude Cake Before and After Removal of Anthracene
Aksumingthat the Lambert-Beer law is obeyed, and stttting k =!,??I, there is had a t each wave length:
Ai = Elcl = klc A2 = kic
AB Sullstituting in Equation 2,
kgc
+
A , = [ ( n - 1)kl kz - nkalc (3) Xow consider the case of this same com nent plus a linear background absorption. The background a G r b a n c e can be expressed by the equation for a straight line, A B = gX h
+
where g and h are constants. The absorbance at any wave length will now be the sum of the absorbances of the sample (substance D ) and the background Therefore
and
+ XI + h
A1 =
~ I C D
A1
kic~ g h
=:
+ +h
1. Only two assumptions have beem made-name1 that the Lambert-Beer law is valid and that the background is k e a r . 2. Measurements are always made a t the same wave lengths, even though in a mixture the absorption minima may no longer occur at these wave lengths. Furthermore, it is not necessar that' the two extreme wave lengths he a t absorption minima. T l e absorbance measurements will usually be more accurate if they are at minima; but if the minima are far apart the chance of the background's being linear between them is lessened. Consequently, with a broad band it may be advantageous to choose wave lengths that are on the sides of t-he band. One of the wave lengths may also be on the tail-off portion of an absorption curve, such as point I11 ofPigure 3. 3. It is possible to evaluate the effective extinction coefficient, kefr,, in two different ways. (1) Determine kl, k p , and k 3 for pure D a t the three wave lengths, and insert them in Equation 4. (2) Determine keff. directly for pure D by measuring the height, A,, above the base line and using this datum in Equation 5. The latter method was followed in this work. 4. Banes and Eby's results, in the present authors' notation, are :
The numerator is an analytical expression for A , (Equation 2). The results may be easier to use in this form for a nonrecording instrument, such as the Beckman Model DU spectrophotometer. With the Cary spectrophotometer, Equation 5 is simpler, as the base line can be ruled on the recorded curve and height A , above this read directly. In Table I1 results obtained by the baseline method are compared with chemical analyses by the maleic anhydride procedure. The results agree fairly well for cakes containing more than about 25y0 anthracene, but are not so satisfactory for lower concentrations, The chemical results are probably low for these crude cakes because of the presence of acidic components such as phenols. This table therefore does not constitute a critical test of eithw method. More disturbing is the fact that the ultraviolet analysis based on the 358 mp band is always higher than that based on 376 mp, indicating that the background may not be linear as assumed. The reason for this discrepancy waa shown by removing the anthracene from several crude cakes. Direct Determination of Background Absorption. Anthracene was removed from four typical cakes by reacting them individually with maleic anhydride in benzene to form 9,lO-dihydro-
1135
V O L U M E 22, NO. 9, S E P T E M B E R 1950
maxima. It is now apparent why a higher analysis was obtained with the 358 mp band than with the 376 mp band by the base line method of calculation. Tests of, Spectrophotometric Analysis. The possession of these deanthracened cakes enabled a critical test of the accuracy of the analysis.
Table 11. Anthracene in Crude Anthracene Cakes
Test No.
(Weight %) Ultraviolet (Base-Line Method) 358 nip (A) 376 nip (B) Diff. (A-B)
Chemical
10.9 11.1 11.3 16.3 18.8 24.1 35.4 40.0 41.8 71.0 90.6 95.0
9.2 0.4 10.4 13.4 19.0 23.3 35.0 38.4 41.1 70.6 91.1 93.0
13.5 12.1 13.4 17.9 19.8 25.0 36.2 40.7 42.1 71.5 91.1 96.0
6 7 8 9 10 11 12 13 14 15 16 17
3.6 1 .o 2.1 1.6 1.0 0.9 0.8 0.7 0.3 0.5 0.5 1 .o
anthracene-9,lOendo-a,fl-succinir anhydride, which was extracted from the reaction mixture by washing the benzene solution a t 71)”C. ivith a q u e o u s W u pot assium h yz droxide. T h e q m alkaline extracts a 0 were c o m b i n e d u) m and acidified, and a the precipitated adduct of anthra-’ cene and maleic anhydride was removed by filtration. The acidified extracts were then washed with WAVE L E N G T H benzene and these Figure 3. Base-Line Method of Backbenzene extracts ground Correction were added to the solution of anthracene-free cake. After the benzene was completely evaporated on the steam bath, the absorption spectrum for the anthracene-free cake was obtained, whirh is compared with the spectrum of the original crude cake in Figure 2. The concentration of the material contributing the background absorption is nearly identical in these two curves. This same type of background was obtained with all four cakes. The background absorbance is nQt linear, but, on the contrary exhibits two small maxima in the general region of the anthracene
The spectrum of the deanthracened cake was matched with that of the crude cake in the wavelength region where anthracene is relatively transparent. Subtraction of these curves yields analyses a t each of the two maxima which are almost the same, as brought out in Table 111. These data show that a true background absorption spectrum was subtracted. The magnitude of these analyses may not be correct. Critical tests were performed in which the deanthracened materials were used to make up synthetic crude cakes by adding known amounts of anthracene. The spectra of the synthetic cakes were then analyzed by the base-line method. The results on four cakes are summarized in Table IV. In the latter tests the base-line results obtained with either band are, with one exception, correct within 235 of the known concentration, and their average is correct within about 1% of the known concentration. After the nature of the background was known, it was observed that the results could be slightly refined by drawing the base line between points I and I11 (Figure
2 .o
12.4% Anthracene
1.5
W V
z
a
m
a 1.0 0 cn m
a
0.5 L4 I , I % Ant hrocene
I
Table 111. Experimental Tests of Spectrophotometric -4nalyses in Analysis of Crude Cakes Subtracting
NO.
376 m p 17.1 22.5 17.7 34.9
1 2 3 4
Table IV.
358 mp 16.9 22.8 17.7 35.0
1
12.3
2 3
22.9 12.4 35.6
22.8 12.1 36.2
AY.
Illustrating successive stages of purification
17.0 22.6 17.7 35 0
-
14.0
13.0
fO.7
11.3
13.0
12.1
-0.2
25.2 14.1 37.0
24.0 13.1 36.6
+1.1 f0.7 f l . O
21.8 11.6 35 3
23.8 13.2 35.6
22.8 12.4 35.4
-0.1 0.0 -0.2
_ _ _ c
~.
, 330 WAVELENGTH m y
Usual Base-Line Method, Wt. % ’ Modified Base-Line Method, W t . % Av. Diff. A v. 3 5 8 ~ (B) (A B) 376 m,, 358mp (C) ( C Diff. - A)
376mr 12.0
1
310
Figure 4. Absorption Spectra of Crude Anthracene Cakes
Experimental Tests of Spectrophotometric Analyses in Analysis of Synthetic Cakes
Known Sample Addition NO. (A), Wt. 7%
4
Approximate
Background, W t . % -___
Sample
I
290 -
3) rather than between I and I1 and between I1 and 111. The two bands then gave results which always bracketed the correct value, and their averages were a little more accurate. These results are included in Table I\’ under “modified base-line method.” This latter procedure is advantageous only for the cruder cakes (up to about 35% anthracene), where the nonlinear background
A N A L Y T I C A L CHEMISTRY
1136 absorption could cause large errors if not taken into account. For purer cakes the usual base-line method using the 376 mp band is satisfactory. A rather valuable by-product of the ultraviolet analysis of crude anthracene cakes is illustrated in Figure 4. Along with the determination of anthracene there is obtained a general picture of the constitution of the crude anthracene cakes. As the purification process proceeds, the ultraviolet spectrum loses its sharp upward trend a t shorter wave lengths, and the carbazole peak a t 292 mp becomes evident. By noting the change in the spectrum a t 340 mp, the removal of an impurity can be followed. The peak a t 337 mp gradually shifts to 342 mp, where it coincides with a characteristic anthracene peak. ANALYSIS OF MIXTURES OF ANTHRACENE, CARBAZOLE, AND PHENANTHRENE
Some preliminary work has been done on the analysis of mixtures of anthracene, phenanthrene, and carbazole. An examination of the ultraviolet spectra of the three compounds shows that there are suitable analytical wave lengths for each of them. In this work the best band for carbazole could not be employed because the studies were made in chloroform solution, and this band occurs a t a wave length which is below the cutoff in the transmittance of chloroform. (Chloroform’was chosen because other conventional solvents did not completely dissolve the crude cakes that were the object of the original studies.) In spite of this handicap, results were obtained which compare favorably with chemical analysis. Figure 5 demonstrates that small amounts of carbazole (1.5%) in anthracene and anthracene (0.18%) in carbazole, respectively, can be determined. Similarly, traces of phenanthrene in anthracene, and the converse, can be determined. These analyses are difficult or impossible’by chemical means. I t may be seen from Figure 1 or Table I that the determination of anthracene is independent of the amounts of the other two compounds, inasmuch as they do not absorb a t the 376 mp anthracene band. Therefore when all three substances are present in the same sample, the anthracene content is readily and accurately obtained. I n chloroform solutions, however, the bands that must be used for carbazole and phenanthrene interfere with each other to such an extent that their analysis is somewhat uncertain. Table V contains the results of several of these three-component analyses which are typical of the experience in this laboratory. The chemical analysis given for comparison is based on the maleic anhydride method for anthracene and a Kjeldahl nitrogen determination for carbazole. Everything else is assumed to be phenanthrene. I t is evident that even in chloroform solution the ultraviolet analysis compares favorably with the chemical analyTable V.
Analysis of Mixtures of Anthracene, Carbazole, a n d Phenanthrene
Teat N o . 18
19
Mixture Anthracene Carbazole Phenanthrene Anthracene Carbazole Phenanthrene
20
Anthracene Carbaeole Phenanthrene
21
Anthracene Carbazole Phenanthrene
22
Anthracene Carbazole Phenanthrene
Ultraviolet
Chemical
100.2 1.8 0.1 102.1
-
96.0 1.6 2.4
4.2 2.8 91.5
I
,
290
4
8
33009-
310
WAVELENGTH m#
Figure 5.
Absorption Spectra of Anthracene-Carbazole Mixtures
sis, and it is much more easily conducted. It also has the advantage of giving an independent check on all three components. SUMMARY
An umaviolet determination of anthracene in anthracene cakes is at least as accurate as any available chemical method, and is much more convenient. The results are within about 1% of the known concentration in synthetic cakes. Spectrophotometry is also suitable for mixtures of anthracene, carbazole, and p h e nanthrene. Accurate analyses can be made when there is extraneous background absorption, if this can be assumed to be linear. The amount of the background need not be known. The method is theoretically sound. ACKNOWLEDGMENT
The authors are grateful to Foil A. Miller of the Department of Research in Chemical Physics a t Mellon Institute for his advice and assistance throughout this investigation, and to the CarnegieIllinois Steel Corporation for permission to publish this work. Acknowledgment is made to S. G. Karan for his experimental aid. LITERATURE CITED
(1) Bane$, F. W., and Eby, L. T., IND.ENO.CHEM.,ANAL.ED.,18, 535 (1946).
(2) Heigl. J. J., Bell, M. F., and White, J. 11947) ~---.,-
U.,ANAL.CHEM.,19, 293
(3)
Hogness, T. R., Zscheile, F. P., Jr., and Sidwell, A. E., J . Phyu.
3.6 1.6 94.8
(4)
Kaufmann, H. P., Baltes, J., a n d Hartw-eg, L., Ber., 70B, 2559
(5)
Khmelevskii, V. I., and Postovskii, I. Ya., J. Apptied Chem.
0.2 100.0 2.2 102.4
0.5 96.5 3.0
(6)
2.1 52.1 46.2
2.0 49.1 48.9
53.2 3.4 44.2 100.8
51.0 3.1 45.9
ss.s 100.4
-
Chem., 41, 371 (1937). (1937).
U.S.S.R., 17, 463 (1944).
Lothian, G. F., “Absorption Spectrophotometry,” p. 68, London, Hilger and Watts, Ltd., 1949. (7) Morton, R. A., and Stubbs, A. L., A n a l y s t , 71, 348 (1946). (8) PostovskiI, I. Ya., and KhmelevskiK, V. I., J. Applied Chem. U.S.S.R., 10, 759 (1937) (in German). (9) Rhodes, F. H., Nichols, M. L., and Morse, C. W., Ind. Eng. Chem., 17, 839 (1925).
(10)
Tunnicliff, D. D., Rasmussen, R. S., and Morse, M . L., ANAL.
CHEM., 2 1 , 8 9 5 (1949). RECEIVED March 6. 1950. Work was performed under Multiple Fellowship sustained by Carnegie-Illinois Steel Corporation a t t h e Mellon Institute, Pittsburgh, P a .