Accurate Analysis with Infrared Double-Beam Spectrophotometer

Robert. Schnurmann and Edward. Kendrick. Anal. Chem. , 1954, 26 (8), pp 1263–1269 ... John A. Perry. Applied Spectroscopy Reviews 1970 3 (1), 229-26...
0 downloads 0 Views 808KB Size
V O L U M E 2 6 , NO. 8, A U G U S T 1 9 5 4

1263 -

and of t8he olefin portioii of the chromatographically separated gasoline are (Ilifferent olefin types are given in mole % of t h e olefin portion. T o t a l olefin portion given as volume sh0w.n. The marked changes of absorp yo of total gasoline) Total tratrscistion in the RR'C=CHR" and RCH= Olefins. KClIRR'C= RR'C= RCH= RCH-= CHR' (cis) regions are to be noted. (;asdine 70 ( ' I { > , "'I CHI. % CHR". '7c C H R ' , 070 C H R ' . 070 Polyior~udistillate 27 "I 6 33 12 16 111 any concentration procedure, of Tllrrmnl cracked 37 32 10 30 14 14 which chromatography is an example, C a t . blending stock 45 9 13 42 20 16 errors made in the sDectral analvsis of t'he concentrated produce are reduced by the concentration factor when related 1)ac.k t o total sample. Thus in Table I X the absolute absorbances a t the paraffin points was rather poor and resulted error in the determination of any class of the polyform distillate in poor paraffin results. I t was evidrrit that, the neglect of modis reduced by the factor of 0.27 n-her1 r t h t e d bark to total sample. erate paraffin contamination did not seriously affect the olefitr This assumes that there i p no error i n the determination of the anulysis. Table VI11 shows the olefin results of a synthetic sample in which paraffins are present, to the extent of 38% t)y concentration factor. volume, and of an olefir] concentrate from a dehydrogenation ACKNOWLEDGhlENT experiment in which the paraffin content was of the order of 20% by volunir. A group-type olefin-specific paraffin matrix was The aid of the Analytical Section in providiiig the &4STM used in one instance, and the ca1culat)ionwas repeated using t,he .irralyres of particular samples is gratefully a c k r i o ~ledged. grouptype olefin matrix alone. T:ible VI11 shows t'hat moderate paraffin contamination in tliv LITERATURE CITED olefin cuts of a chromatographic separation will not seriowly (1) .-h. doc. Testing Rlateriala, "Tentative Method for Test for afTecst tlie ordinary accuracy of t,he olefin analysis. However. Olefins and Aromatics in Petroleum Distillates," (Issued i n order to obtain bett,er information on the paraffins, chroruato1946), Designation ASTRI D 875 46T, Petroleum Book D-2. graphic separatioii was used. It was then possible to improw (2) sa^. CHEM.,22, 850-81 (1950). the paraffin accuracy so that the average error was approsimat,elJ( 3 ) Inderson, J. A., Jr., and Seyfried, R'.D.. Ibid,. 20, 998 (1948). (4) Criddle, D. W., andLe Tourneau, H. L., Ibid., 23, 1620 (1951). il% of tot,al sample. The rrsu1t.s for specific paraffins befont ( 5 ) Francis, S. il., J . Chern. Phys., 19, 942 (1951). sild after reaction permitted the evaluation of any isomerizatioii ( i ; ) Hastings, S. A , , rt al., ~ ~ x - A L CHEX., . 24, 612 (1952). :tc.c:onipanying the dehydrogenation. (7) Hibbard, R. R., and Cleaves, d. P., Ibid., 21, 486 (1949). In the :Application to gasolines, only the olefinic distributioris (X) Johnston, R. W. B., Applehy, W.G., and Raker, hl. O., Ibid., 20, 805 (1948). were determined spect,rally. The chromatogram permitted the (9, Mair, €3. J., private communication. drtermination of total saturates, olefins, and aromatics. The (10) Hose. F. W., J . Research Xatl. Bureau Standards, 20, 129 (1938) twults for three gitsolines are presentrd in Table IX. H W K I Y E Dfor review February 18, 1954. Accepted April 29, 1954 111 1;igitre 5 the spectra of' :i total thermal cracked gasoline Table IX. Olefin Data for Three Representative Gasolines

Accurate Analysis with an lnfared Double-Beam Spectrophotometer ROBERT S C H N U R M A N N

and

E D W A R D KENDRICK

Physics Department, Manchester O i l Refinery, Ltd., Manchester, England

The purpose of this work was to examine the effect of instrumental factors on the anal) tical accuracy attainable with an infrared double-beam spectrophotometer. For xylene-ethylbenzene mixtures analytical results well within 1% accuracy ran be obtained if instrument corrections are applied and averages of at least four recordings are taken. The technique of "difference spectra" can yield mean values with a deviation of not more Corrections for finite slit widths must be than f O . l % . made. Empirical corrections are preferable to assumptions concerning distrihii tion of spectral impurity and of response coupled w i t h geonletrical approximations LO the shape of the absorption band. Over an appreciable range of concentrations the tangent to the absorbance-concentration riir) e in the optimum transmittance region is a closer approximation to the observed relationship than the chord drawn through a point on the curie. The technique of difference spectra i5 speedy and accurate, compared with the cell in-cell out method, where the spectra of t w o samples are ver> similar.

A

RIGOROUS application of simultaneous equations prrsupposes an exact knowledge of every one of the coefficients. In spectrophotometry this condition is not fulfilled. The degree of analytical accuracy which can be achieved with a fully

:sutomatic infrared double beam spectrophotometer depends upon instrumental factors and the availability of nonoverlapping absorption bands. The instrumental factors include random errors and systematic corrections. I t is shown for the example of mixtures of the xylene isomers with ethylbenxene that analytical results ciin be obtained with an accuracy of within f 0.4% and even nithin f 0.2% of the total if the random errors are reduced by Nveraging over four or more recordings and if the systematic corrections are applied to the readings taken from the records. T h r equations are then solved using the corrected experimental data. By the method of difference spectra a t fixed wave length8 I I ~ P Y I Ivalues can be obtained with a deviation of not more than + O.lyo from the composition of test samples. EXPERIMENTAL FACTORS

Random Errors. With a Prrkin-Elmer Model 21 double beam instrument records in the rock salt region commonly repeat with an accuracy of the order of the width of the trace when no adjustments are made between consecutive runs. Figure 1 shows for a number of recording speeds three neighboring pyridine bands in the 5-micron region. These bands were scanned six times a t each speed. It is clear that the autosuppressor has a slight effect upon the recorded position of the peak of each band. The two single scans are with and without autosuppression and a slight shift toward longer wave lengths is apparent when the autosuppressor is not applied. Throughout this investigation the autosuppressor

ANALYTICAL CHEMISTRY

1264

is used and the reproducibility of the wave-length positions of the bands is e x t r e m e l y good. There are, however, small random fluctuations of intensity similar to those shown by White and Liston (21). With a recording speed of 1 minute per micron and slits of 0.05-mm. width the largest of these v a r i a t i o n s mere 1.7% w h e n t w o values of the amplifier gain were selected; a t each amplification t h e bands were scanned four times. That these were irregular f l u c t u a t i o n s was confirmed by scanning these bands 10 times and numberi n g e a c h t r a c e on the chart consecu4-5p 5-Or 5-5p tively. Thus, for Figure 1. Six Consecutive Scans one band the tenth with Autosuppression of Three scan gave the Neighboring Pyridine Bands in the smallest absorption 5-Micron Region intensity, for the Recordifg speeds. 0.5,1,3. and 8 minutes next band the ninth per micron Slit width. 0.05 mm. and the tenth, and Amplifier gain setting No. 5 used with Perkin-Elmer Model 21 for the third band the n i n t h scan. Similarly. for one band the second and fourth records gave the highest absorption, For the next band the first, second, and third, and for the third band the first record. It is considered that frictional forces bctween the pen and the recording paper as well as frictional force. in the servomechanism have significant influence on such errors; this is borne out by close examination of the motion of the pen. Lack of stability in the output stage of the amplifier also contrihUtes to these errors, since better agreement between consecutive recordings is obtained when the pen is set to the zero transmittance line before each scan. This agrees with recent findings of Boaman and Tarpley ( 5 )and of Hausdorff, Sternglanz, and Williams (9). The balance control of the amplifier is used to move the effective zero transmittance point of the optical attenuator and the amount by which the amplifier is biased is referred to as the off-balance signal. Systematic Corrections. OPTICAL .~TTENUATOR. The PerkinElmer double-beam instrument uses a comblike shutter which is moved in or out of one optical path with the object of reducing its intensity until it is equal to that of the beam containing the absorbing sample (8, 16, 23). From the position of this comb along its direction of motion normal to the reference beam, the percentage transmittance of the sample is recorded. It is sometimes said that the aperture of the optical attenuator must vary in direct proportion to its displacement relative to the position of the reference beam. Since the comblike shutter of the Perkin-Elmer Model 21 instrument shows deviations from linearity, appropriate corrections must be made to the absorbance scale. (Details concerning the nonlinearity of the “optical wedge” will

be published elsewhere.) These empirical corrections are plotted in Figure 2. The correction curve in this figure was obtained by varying the slit width by known amounts when the shutter of the sample beam was closed and a test signal was injected to balance the energy in the reference beam. Very wide slits were used to avoid apparent wedge errors caused by the entranoe and exit slits not closing simultaneously. FINITE SLIT WIDTH. Comparatively wide slits are used in infrared spectrophotometers. Finite slit width leads to spectral impurity and deviations from LambehBeer’s law. Proposals have been made ( 7 , 18, 19) for approximate corrections of the absorbance measured with finite slit widths. Since, however, empirical corrections can be employed, these are much to be preferred to assumptions concerning distribution of spectral impurity and response which are usually combined with geometrical approximations to the shape of the absorption band. For each componerit of a mixture an analytical wave length is selected a t which this component shows appreciably stronger absorption than any of the other components. To amess the deviat,ionfi from Lambert-Beer’s law, measurements of absorbance must be made with various concentrations of each component of a mixture a t the analytical wave lengths. For a given error in transmittance the fractional error in measured concentration is l e u t a t 37% transmittance. Since the minimum of the fractional concentration-per cent transmittance curve is very broad and flat, a tangent to the absorbance-concentration curve may hr drawn a t any transmittance between 25 and 50% in order to correct the measured absorption coefficient for deviations from Lambert-Beer’s law. T o make the measurements which are necessary for the plotting of absorbance-concentration curves, a sealed cell of carefully calibrated path length must be used. [The thickness of the cells was determined by interference fringes (la).] EXPERIMENTAL RESULTS

Xylene-Ethylbenzene Mixtures. Convenient analytical wave lengths for xylene-ethylbenzene mixtures (14, 15, 2 2 ) are 12.59 microns for p-xylene, 13.02 microns for m-xylene, 13.48 microns for o-xylene, and 14.36 microns for ethylbenzene. The analyses considered in this paper were made with “slit program 3”-i.e., a slit width of 0.156 mm. a t 10 microns and 0.412 mm. a t 14 microns. Attention must be paid to the relative proportions of the components of any particular mixture, as there is a comparatively weak m-xylene band a t 14.45 microns and a weak and broad ethylbenzene band a t 13.4 microns. However, the resolution of the instrument is good enough to show an absorption peak a t 14.36 microns when the proportion of ethylbenzene is

ADD.

Figure 2. Corrections for Nonlinearity of Optical Wedge of Perkin-Elmer Model 21 Spectrophotometer Trend of experimental points is approximated by the heavy lines which are used for correcting recorded data

I

1265

V O L U M E 26, NO. 8, A U G U S T 1 9 5 4

shown in Figure 3 were constructed from measurements a t the analytical wave lengths for earh of the xylene isomers and ethylbenzene covering the range of concentrations given in Table I. Over an appreciable range of concentrations the tangent to the absorbance-concent,ration curve in the optimum transmittance region is a closer approximation to the observed relationship than the chord drawn through a point on the curve. This is illustrated hy the inset of Figure 3 where the crosses show the insignificant variation of the difference of concentration values between the t,angent and the absorbance-roncentratioll curve. whereas the dots show the rapid change of the corresponding differences for the chord. Therefore, the tangent method was adopted. The slope of the tangent is, for the particular component of a mixture, its absorbance per unit of the concentration scale a t the analytic%l wave length of the material under consideration. T o correct the absorbance of the mixture a t the analytical wave lengths the intercepts of the corresponding tangents with the absorbance axie were subtracted. Each corrected absorbance must be multiplied by the dilution factor of the solution. This factor is determined by the desirability of working in the optimum transmittance region. The concentration values calculated from equations containing the absorption coefficients were further corrected for the difference on the concentration scale between the empirical absorbance-conrentratian curve and ib, tangent (Figure 4). The final results for seven synthesized mixtures are given in Tables I1 and 111, which are comprised of data relat,ing to one four-component, four three-component, and two two-component mixtures. It,is apparent from these tables that the totals fall between 99 and 101 %, apart from one case of l02.9%, for a mixture which contains predominantly ethylbenzene (misture F). I n a case like this thr base-line method gives higher accuracy than the setting up of simultaneow equations. The agreement between the measured and thc :Lctual concentrations of ethylbenzene is best in the c’ase of the four-componrnt mixture .4 where thrrr was good resolution :it all the analytical wave lengths (Figure 5 ) . Mixt,ure E offrrs a uarful illustration of the pitf:111~ when components have overlapping absorption bands. Because the analytical wave length of ethylbenzene lies on the side of the m-xylene band a t 14.45 micronsone derives from a set of four equations a small amount of ethylhenzene (0.3%),although mixture E had heen synthesized from the threr xylene iaomers without the addition of any ethylbenzene (Table 111). To average over the random errors four scannings were taken of the spectra of the mixtures A to G. The importance of maintaining a constant off-balance signal for accurate measurements was mentioned previously. Two industrial samples of xylene-ethylbenzene mixtures were examined without and with adjustment of t1ii.C signal (Tables I P and V). M e m ( w o w of 0.49 :ind 0.0:j % (depending on the number of

Table I. Data Relating to the Construction of Absorbance-Concentration Diagrams Range of Concn. in Iso-octane, G./hll. 0.00102-0.02455 0.00158-0.01857 0.00045-0.01432 0.00106-0.03476

Component p-Xylene wXylene o-Xylene Ethylbenzene

2-c

D

2

0.005

0.010

0.015

0. 3

CONCENTRATION IN p / d .

Figure 3. Absorbance-Concentration Diagram for Ethylbenzene with Tangent and Chord Drawn at a Point in Optimum Transmittance Region ¶ m e t shows deviations from experimental data i n the neighborhood of the selected concentration when tangent, x, i s drawn and when chord, O, is drawn

about one fifth of that of m-xylene. It appears that in spite of the clow neighborhood of the weak ethylbenzene band to the analytical wave length of o-xylene (13.48 microns), this isomer is the one which is deterniined with the relatively highest degrce of accuracy, because thr o-xylene band a t 13.48 microns IS by cornpalwon very strong The ethylbenzene band a t 14.36micions is romp:iratively weal\ and it ni,lT therrfore hnpprn that the a n a l y t i t a l v, a v e l e n g t h foi ethylhenzene falls on t!ir ride -___ _ _ _ ~ _ _ _ of the m-xylene band at 14.45 Tahle TI. ihsorption Coefficients and Intercepts on Ahsorbance Axis of Xylene niicroiis. This tends to inIsomers and Ethvlbenzene crease the uncertainty of the Ahsorption Coefficients in a 0.272-hlin. ____ _ _Cell _ ethylben zene deterniination, p-Xylene m-Xylene o-Xylene Ethylbenzene Deviations from LambertKaYe-Length, .Ibsorption Absorption Absorption Absorption P roefficient Intercelit roefficient Intercept coefficient Intercept coefficient Intercept Beer’s Law. Absorbance-con12 9; 72.4 0.016 1 070 0.477 . .. 2.88 ... centration diagrams of t.he type 13 02 0.866 67.0 01633 0.883 5.04 ... 13.48 0.175 ... 0 46,; ... 112.3 o:oi4 11.87 14.36

0.033

...

...

i 63

0.24B

...

35 7

0:033

Table 111. Analysis of Synthesized Mixtures of Sylenes and Ethylbenzene by Infrared Spectrophotometry P-Xylene, % m-Xylene, % o-Xylene, y? Ethylbenzene, % Mixtun A B C D E F

c;

Actual 34.3 1.39 1.15 35.5 39.5 0.00 0.00

Meas. 33.6 1.27 1.17 39.8 40.1 1.49 0 83

Diff. -0.7 -0.12 +0.02 -0.1 4-0.2 f1.49 + O 83

Actual 4.07 4.65 20.1 2.79 2.79 27.1 51.6

Meas. 4.07 4.78 20.4 2.50 2.70 27.4 00 4

Diff. 0 00 f0.13

+0.3 +O.ll -0.09 +0.3 --I

2

Actual

3Ieas.

Diff.

50.3 54.2 78.8 56.9 56.9 0.00 0.00

50.3 93.7 78.4 57.4 58.0 0.88 1 17

0.0

-0.5 -0.4 +0.5 fl.1

f0.88 4-1.17

Actual

Meas.

Diff.

Total, %

11.4

11.4 0.00 -0.75 f0.20 +0.30 73.1 48.4

0.0 0.00 -0.78 fO.20

99.4 99.7 95.2 100.3 101.1 102.9 100 8

0.00 0.00 0.00 0.00 72.9 48 3

fO.30

+0.2 +O 1

ANALYTICAL CHEMISTRY

1266

7

It is important for analytical purposes to ascertain whether the infrared spectrum indicates additional constituents. Reference has already been made to the presence of paraffinic hydrocarbons in n mix\ ture of Cs aromatic compounds. hleasurements of two further industrial sample9 of xylene-ethylbenzene mixtures reveal the presence of additional aromatic hydrocarbons Sample 3 shows a weak band \ at 13.74 microns which might be caused by 1/2yo of toluene. If the contribution of this amount of toluene to the absorption i p allowed for a t 14.36 microns, it reduces the ethylbenzene concentration from 15.1 to 14.9%. The analysis of sample 3 would therefore yield 24.9% p xylene, 44.4% m-xylene, 6.6% o-xylene, 14.9% ethylbenzene, approximately 0.570 toluene, and 9.0% paraffins, making a total of 99.3%. The spectrum of samplc Figure 4. Empirical Corrections, Ac, of Concentrations Derived from 4 s h ~ i v ca weak band a t 14.83 micron\ Simultaneous Equations for Xylene Isomers and Ethylhenzene and, qince the origin of this band is uncertain, the provisional analysis of this sample is subject to slight revision. The present analysis is individual recordings) were obtained without adjustment of thr 7.8% p-xylene, 27.4% tu-xylene, 49.9% o-xylene, 2.8% ethyloff-balance control (Table VI); a mean error of 0.2% was obhenzene, and 12% parafins, iiiaking a total of 99.9%. tained when such adjustment was made before each scanning (Table VII). The content of paraffins in these two samples was estimated from their refractive indices on the assumption that 1% DIFFEREU(:E SPECTRA of paraffinic material reduces the refractive index by apBy difference spectra is meant spectra obtained with a spcciproximately 0.001. The paraffinic material was also separated men of known composition in the reference beam. This techfrom a xylene cut and its absorption spectrum was recorded PO nique is adopted to reveal differences in the spectra of the samples that allowance could be made for its contribution to the absorbin the two beams, hence the name difference spectra. In the ance at the analytical mive lengths. American literature (2-4,6,10-15, $0, g 4 ) this technique is som('times referred to as differential spectra. A double-beam instrument permits the direct recording of difference spectra and this may be useful in determining with :L Table IV. Repeatability of Analysis of Two Industrial Xylene-Ethylbenzene Mixtures by Infrared Spectrophotometry

(ntt)

(Without adJustment of the off-balance control of the instrument) Tntnl

4.

--"-I, ,"I

AnalySample si8 1 la lb le

Xylene,

%

% 51.1 50.2 47.8 f9.0 01.2 49 4 49 7 49 0 48 5

%

yo

2.96 3.01 3.05 3.12 2.95 3.32 3.10 3.13 3.14

14.6 13.3 13.6 13.0 12.7 12.6 13.1 13.9 14.2

:

za

2.5 7 25.6

Id

Ig

Mean 2

mXylene,

24.0 24.7 25.2 25 9 25.4 26 0 26.2 26.0 26 3

le If

2h

2c

2d 2e 2f

Mean

Ethylbenzenc.

PXylene,

27 4 24 1 25.4 24.9 26.1 256

4'3 5 so 7 56 1 48.4 52.9 56.5 ;i3 8

53.1

0-

3.09 4.14 3.66 3.63 3.66 3.84 3.80 3.79

13.3 12.6 13.1 11.6 12.4 11.7 12.4 12 3

Including 9.0% Paraffins 101.7 100.2 98.6 100.0 101.3 100 3 101 I

100 6 102

o

109.3 96.7 103 4 105 7 105 1 103 7

T.~ ntkl

lk 11

Mean 2

Mean

;; 2g

25.0 25.2 25.0

49.0 48.4 47.9

2.90 2.92 2.87

14.0 13.8 13 3

25 1

48 I 51.4 Sl.2 ,570 8

2 90 3.52 3.41 3.49 347

13 i

9'3 1

12.1

101 3 101 4 101.3 101 3

%

25.3 25 9 25 4 2.5.;

tnXylene,

%

3 1 1

0-

Xylene,

%

Ethylbenzene.

67,

Incluhi~g 9.0% Paraffins 99.9 99.3 98.2

P-

So

11.9 12.6 12 2

1

Analysis la lb IC

Id le

If

k

lh li

101.0

( W i t h adiiistment of the off-balance control of the instriinient)

Xylene,

Sample

101.0

Table V. Repeatability of Analysis of Two rndustrial Xylene-Ethylbenzene Mixtures by Infrared Spectrophotometry

.4naly.Sample 91s 1 1j

Table VI. Deviations from \lean of Data Given in Table IV

Mean deviation

2

p-Xylene -1 7 -1 0 -0

5

+o 2 -0 3 +o 3 to5 +o 3

+o

5

,n-Sylene +I

+o

.F

7 -1.7

-0 R +1.7

-0.1 2 -0.5 -1 0

+o

o-Xylene - 0 13 - 0 08 - 0 04 +O 03 - 0 14 +O 23 +o 01 + O 04 +O 05

06

0 8

0 08

2a

0 0

+o 35

2h 2c 2d 2e 2f

+18

-2 4 7 3 0 -4 7 -0 2 +3 4 1 0 7

-i 5 -0.2 - 0 , +o 3

Mean deviation

0.8

2.4

Ethylhmaene +I 3 0 0 +o R -0 3 -0 -0

6

7

-0 2 +o 6 +0 9

n

.i

+O 3

+o

-0 13 -0 16

-0

8 7

-0

R

r o

-0.13 f0.05 t0.01

1

+n 1 n-L

0.14

Mean error for sample 1 over all ninr determinations, 0 49% Mean error for sample 2 over all siu drterniinations, 0.93%

~Table VIL. Deviations frnin \Iean of Data Given in Table v Sample 1

Analysis 1j

lk 11

Mean deviation

2

2g 2h 21

Mean deviation

p-Xylene -0.1

+o. 1 -0

1

0.1 -0

+o -0

2 4

1

0.23

rn-Sylene +0.6 0.0 -0 5

0.37 3

t o +o -0

1 3

0 28

o-Xylene

Ethylheiiaene

0.00

+0.02 -0.03 0.017

+o 05 -0 06 +o 02 0 04

Mean error for sample 1 over all three determinations. 0.18% Mean error for sample 2 over all three determinations. 0.19%

+o x +o

i -0 3

0 23 -0

1

-0 3

to

4

0 2;

1267

V O L U M E 2 6 , NO. 8, A U G U S T 1 9 5 4

by the ASTM (1)is to use four blends, each of which is prepared with an accuracy of a t least 0.02 volume % for each component. One blend, reference standard A, contains 59.00% iso-octane and 41.00% n-heptane; reference standard C is made up to 61.00% iso-octane and 39.00% n-heptane. A test blend B contains 60.00% iso-octane and 40.00% n-heptane. With the aid of these three blends one is required to determine the iso-octane content of unknown blend D to within +0.2%. According to the ASTM handbook one measures in close succession the absorbances of the four liquids in the order: (1) one of the reference standards, (2) the test blend B, (3) the unknown blend D, and (1) the other reference standard. The investigator is requested T O “determine the absorbance of each of these liquids from four Po readings, resulting from incident radiant power, using the blank plate followed by four P readings, resulting from transmitted radiant power, with the liquid in the cell.” The further instructions are to ‘(reject the analysis unless the individual Po and P readings for each liquid agree with the corresponding average POand P values within plus or minus 0.2% of the average. . and to make a t least two additional analyses in the same manner checking the urave-length setting between analyses.” For the test blend the average values must agree with the known percentage to f0.1%. This rather searching requirement can be fulfilled with any type of infrared spectrometer provided that intensity measurements are made with the sample cell removed from the beam (Po measurements) before and after each absorp-

.

1 Pr

12igure 5.

1 3 ~

14a

15r

Spectrum of Mixture A in 12- to 15-Micron Region

Concentrations adjusted so that percentage transmittance was in region of 30 t o 50 at each analytical wave length Curve Concentration, G./M1. 0.1168 Bottom 0.0778 Second 0,0195 Third Fourth 0.0097 Too 0.0065 Short horizontal lines indicate percentage transmittance valuea used for odculation

i

I

high drglre of accuracy small ptwentages of one component in a mixture, a sample of which it placed in the reference beam. The technique can also save murh time in the accurate analysis of two-component mixtures. To give an example, a commercial sample of p-xylene revealed by its spectrum (Figure 6 ) a small amount of rn-xylene. If the concentration of a solution of this p-xylene material in iso-octane were adjusted to bring the absorption a t 13.02 microns, the analytical wave length of m-xylene. ivithin the optimum transmittance region (z in Figure 6), the ( orittibution of pxylene absorption a t this wave length would introduce a large error in the m-xylene determination. On the other hand, when this particular p-xylene material was measured against :i sample of pure p-xylene in the reference beam, the m-xylene I)and a t 13.02 microns (Figure 7 ) was prominent, although the m-.uylene content was only 0.64%; in addition a weak band appeared at 13.22 microns, indirating that there wa8 some further impurity present. ‘4good illustration of the time saving that can be achieved with diflerence spectra is given by the analysis within 0.20/, of blends of GOYoiso-octane and 40% n-heptane. The procedure laid don 11

1 %J

13r

14r

15r

Figure 6. Spectrum of Commercial p-Xylene Sample Containing Small Amount of rn-Xylene X. 0.2928 gram per ml. Cell length, 0.272 mm.

1268

ANALYTICAL CHEMISTRY

tioii measurement ( P measurement), but it is considered on the grounds of the present work that one series consisting of, say, six measurements of each sample would be sufficient to give the required accuracy, if two Pomeasurements were made separately for each P measurement, one before and one after. Certainly a great deal of time can be saved with a double beam instrument by taking difference spectra. In this case a cell filled with the reference standard A for the iso-octane determination is used in the reference beam or a cell filled with the reference c;tandard C is used for determining the n-heptane values. Fixed wave lengths of 8.54 and 13.82 microns, respectively, are chosen for iso-octane and n-heptane and again transmittance has to be adjusted to about 37%. A4typical set of results is shown in Figure 8 with average values of 60.00% iso-octane and 40.05% n-heptane. For the a-heptane determination a cell filled with reference standard C was placed in the reference beam; for the iso-octane determination the cell was filled with reference standard A. There was a slight difference in the length of the cells in the sample and in the reference beam; the sample cell was 0.117 mm. long and the reference cell was 0.120 mm. long. The relative transmittance of each blend was measured from the base line of the transmission wale.

ISO-OCTANE

N-HEPTANE

P

C

-+

-

Figure 8. Determination of Iso-octane and n-Heptane at Fixed Wave Lengths in Test Blend Relative transmittance of each blend measured from base line of transmittance male Iso-octane, 8.54 microna A. 59% C. 61% n-Heptane 13.82 microna 39% A. 4170

c.

Table VI11 contains eight individual iso-octane and eight individual n-heptane values for test blend B, yielding mean values of 59.94% iso-octane and 40.05% n-heptane. Both figures are well within 3 ~ 0 . of1 the ~ ~known composition. Two of the eight individual iso-octane values were 59.74%. Following the ASTM stipulation to reject individual values which are outside a deviation of &0,2%from the known composition, they would have to be omitted from the data considered for the arithmetical mean. The average of the six admitted values amounts then t o 60.00%.

60.02 59.90 ' 59.74" 59.85 59.74" 60.19 .

1Pr

Figure 7.

13r

14r

15r

Difference Spectrum of Commercial p-Xylene Sample

a. Obtained with sample of pure p-xylene i n reference beam b . Obtained with pure p-xylene in sample beam and commercial p-xylene in reference beam Broken line is a n extrapolation for case of pure p-xylene in both beams Commercial sample contains 0.64 970 m-xylene Weak band at 13.22 microns indicates presence of unknown contaminant

n-Heptane

60.00

V O L U M E 26, N O . 8, A U G U S T 1 9 5 4 Measurement of the same test blend a t 8.54 and 13.82 microns, respectively, using the instrument virtually aa a single beam instrument without another cell in the reference beam gave mean values of 60.00% iso-octane and 39.96% n-heptane, showing very close agreement within k0.1yo.

1269

Hence for an error in D of AD

=

=k0.2%-i.e.,

about 0.001

AP = 0.023 cni. for P = 10 cm. and AP = 0.016 cm. for P = 7 cm. Similarly,

IDISCVSSIOh

The elperimental procedure to obtain an accurate analysis by infrarcd spectrophotometry is t o employ the same slit widthi.e., the same slit program-foi scanning around the analytical w a w lengths of the spectra of a mixture and of its components. The absorbance values, D, of the mixture are read a t the analytical wave lengths and corrections, AD, for attenuator deviation AD. -4furfrom linearity are applied. This yields D' = D ther correction for deviations from Lambert-Beer's law is then applied with the aid of measurements at the analytical wave length for the respective component. In this way the absorption values A , = D' - A d , are obtained for the components of a mixture. The concentration of each constituent is evaluated by inserting in the equation for the concentration of each component the A , ' values of all components of a mixture. A,' = y , ( A , ) , where 7%is the dilution of the mixture to make the transmittance a t each of the analytical ware lengths between 25 and 50%. T o these results finally a concentration correction is applied for deviations from Lambert-Beer's law. When these three corrections are made the mean crrorq of the results for the example of mixtures of the xylene and ethylbenzene can be kept within 1%. It is of interest that a similar high degree of accuracy cannot be obtained by ultraviolet spectrophotometry ( 1 7 ) because here in spite of the smaller slit width there is more overlapping of bands with the result that thc differences of some of the cross products from the simultaneous equations assume comparatively small values. In the infrared case o-xylene is the isomer which can be determined R ith the relatively highest degree of accuracy, whereas in the ultraviolet case it is p-xylene. The importance of corrections for deviation from LambertBeer's law R as discussed by Williams, Hastings, and Anderson ( 8 2 ) . I t wan pointed out there that the same instrument and the same resolution must bc used for recording the spectra of mixtures and for plotting the absorbance-concentration curves oi the components of the mixtures. For the analysis of industrial samples it is considered that normalization of the results detracts from accuracy because of unknown components. Industrial xylene-ethylbenzene distillation cuts illustrate this contention; in such cases the contents of saturated hydrocarbons can be estimated from the refractive index of the mixture, whei eas normalization alone of the experimental results for the aromatic hydrocarbons would introduce a corresponding error. For measurements a t fixed wave lengths with the single beam technique the small random errors are easily averaged w-hen four or five consecutive recordings are made of pairs of Po and P measurements so that horizontal lines can be drawn through corresponding P O and P levels. It is then a matter of measuring with reasonable accuracy the distance between these levels and the base line of the recording chart. The measurements of P from the recordings have t o be accurate to less than 0.2 mm. if the most is t o be made of the accuracy of the spectrometer. From the definition of absorbance

+

log, p Po

D Therefore

loglo 0'1P

= -__

2.303

IItLncr, for A D = k 0 . 2 7 4 and Po = 25 cm.

PO = is of the order

of 0.6 mm.

CONCLUSION

The infrared spectra of a mixture and of its components mufit he measured with the same instrument and the same resolution. lhpirical corrections can be made for finite slit widths. The components of synthesized mixtures of xylenes and ethylbenzene can he determined within ~4~0.4%of the total. When industrial xylene-ethylbenzene distillation cuts are examined and their content of saturated hydrocarbons is taken into account, results are olitained with mean deviations of 0.2%. The technique of difference spectra is illustrated by the measurement of 0.64% of m-xylene in a commercial sample of pxylene. 1 lcasurcments of about 60-10 iso-octane-n-heptane blends a t lisvd wave lengths show t,hat an accuracy of 0.1% can be reached. 111 this lat.ter case the single beam technique can also be employed successfully if a check is kept on the PO values and an appreciably 1:irger number of measurements is made than with the doubleI ) w m twhnique. ACKNOW LEIK>lEX'I

The itut,liors express their thanks t o G. K. T. Conn for liis cwmnicnts and criticisms and t,o TYatiini Roman for the separation 01 thr paraffinic material from a x>.lcne cut. L I ' I E R A T L ~ K1.: CITEI)

Standards on Petroleuin Products and Lubricants," Tentative XIethod D 1095-5OT, American Society for Testing Materials, Coninlittee D-2, Philadelphia. Pa.,

"-iS'PAl

pp. 568-71,

1951.

Bastian, It., AKAL.CHEM.,21, 9 i % (1949); 23, 580 (1951). Bastian, R., Weberling, R., and Pallila, F.. I b i d . , 22, 160 (1950). Reroza, XI., Ibid., 25, 112 (1953). Rowinan, H. AT., and Tarpley, IT. I%.. A p p l . Spectroacopu, 7, 57 (1953).

Hammer, C. F., and Roe, H. I t . . .-\s.~7.. CHEM.,25, 668 (1953). Hardy, A. C . , and Young, F. 31., .I. Opt. SOC.Amer., 39, 265 (1949).

Hardy, J. D., and Ryer, A. I., Plryr. Rec.. 55, 1112 (1939). Hausdorff, H., Sternglanz, H., and Williams, V. Z., A p p l . Spectroscopy, 7, 63 (1953).

Hiskey, C . F., ANAL.CHEM.,21, 1410 (1919); Trans. .Y. Y. A c a d . Sci., 11, 223 (1949). Hiskey, C . F., and Firestone, D., ANAL.CHEM.,24, 34%(1952). Hiskey, C. F., Rabinowitz, I., and Young, I. G., I b i d . , 22, 1464 (1950).

Hiskey, C. F., and Young, I. G., I b i d . , 23, 1196 (1951). Kaye, W, I., and Otis, 11.V.. I b i d . , 20, 1006 (1948). Kendrick, E., M.Sc. Thesis, Manchester, 1953. Lehrer, E., 2. tech. Phys., 23, 169 (1942). lIaddams, W. F., M.Sc. Thesis, Shefield, 1948. Philpotts, A. R., Thain, W., and Smith, P. G., ANAL.CHEM., 23, 268 (1951).

Itamsay, D. A., J . A w . Chcm. Soc.. 74, 72 (1952). Robinson, D. Z . , ASAL. CHmf., 24, 619 (1952). White, J. E., and Liston, AI. D., J . O p t . Soc. Amrr., 40, 99 (1950).

Williams, R. B., Hastings, 9 . II., and Anderson. J. A , ,

.\s.AL.

CHEM.,24, 1911 (1952). Wright, S . , and Herscher, L. IT., .J. O p t . SOC.Amer., 37, 211

(1947).

Young, I. G.. and Hiskey, C. I'.. As.