Replicability of Analyses for Ethylene-Propylene Copolymers a t Three Different Film, Sample 1 Film. Samule 2 Foln.. ThickSoh, ThickSoh, Mol. % ness, Mol. yo ness, A8.,/ ?viol. 7 0 Mol. yo A8.,/ Mol. To mils A13.8 CI CI Ca mils AI3.9 CB C3 33.7 4 4 0 395 33.5 48 0 3 9 0 895 45 8 58.1 4 1 33 7 0 392 33.2 4 4 0 872 48 0 45 4 54.8 35.4 2 5 0 334 31.0 48 4 4 5 0 852 45 1 56.5 33.7 48 2 2.7 0 389 33.9 57.6 6 0 0 897 45 8 4 4 0 424 34.2 34 6 47 9 46 0 897 45 8 57.1 31.2 2.3 0.340 33.1 ... 4 9 0 898 57.1 45 8 ... 2.3 0.337 31.0 . . 6 2 0 901 46 0 1.9 0,319 ... 30.2 6 8 0 854 45 2 1.8 0.338 31.0 4 9 0 895 45 8 3.2 0.360 32.4 5 0 0 857 45 2
Table 1.
..
... ;IV.
2.3 3 8
34.0
0.345 0.383
31.5 33.0 32.3
...
4 4 6 8
0 871 0 902
48.1
45 3 46 0 45 6
...
Composiiion levels
Thickness, mils 3 4 2 6 2 1 2 4 4 4 3 5 4 3 2 9 3 2 3 2 2 9 4 2
Film, Sample 3__AS.,/ A13.9 2 423 2 410 2 230 2 170
2 420 2 320 2 480 2 200 2.120 2.400 2 390 2.220
Fj6.9
Mol. yo Ci
60.8 60.7 59 5
.w
1
A0 8 80 2 61 1
59 ,58 80 60 59 60
4
8
6 5
6 1
Xhs. std.
dev.
0.784
1.405
precisions of the two methods are practically the same, and t h a t the film thickness has no significant effect on precision within the thickness range examined. Polypropylene gives rise to many infrared absorption peaks, and of these the symmetrical C-H bending vibration of the methyl group a t 7.25 microns is much too intense for use in film analysis. Many of the other peaks characterize polymer crystallinity and are likewise useless for the present purpose. On the other hand, the propyl skrletal vibration a t 8.7 microns has moderate intensity and is not due to crystallinity ( I ) . It also gives good mc3asurement and was therefore selected for the measurement of copolymer propylene content. Figure 2 shows the infrared spectra for four different rthylene-propylene copolymers, correFponding to different compositions and conditions of synthesis. Polyethylene
0.200
0.333
has very few characteristic infrared absorption peaks. The most suitable peak occurs a t 13.9 microns; this represents the block methylene rocking vibration, for blocks containing a sequence of three or more methylene units. This sequence length decreases as the copolymer propylene content increases, and the intensity of the 13.9-micron peak falls off exponentially. This explains why the linear relationship shown in Figure 1 requires a semilogarithmic plot. The least mean squares line of Figure 1 is given b y the equation log,, AI,.^) = 0.0286 (C,) - 1.3730 where (CJ is the mole per cent of propylene units in the copolymer. The analyses from this method represent, of course, the average compositions of the various copolymer species present in the sample. If one is dealing m t h copolymer fractions. or with physical mivtures of polyethvlene and polypropylene, the
1.045
0.703
values of the numerical parameters given in our correlation equation may require modification. ACKNOWLEDGMFNT
I express m y indebtedness to John Rehner, Jr., for his personal inspiration and many fruitful suggestions. I a m also very grateful to Herbert F. Strohmayer for kindly supplying the copolymers. LITERATURE CITED
(1) 4be, K., Yanagisawa, K., J . Polymer Sci. 36, 539 (1959).
12) Bua. E.. Manaresi. P.. ANAL. CHEM. 31.2022 ri959i. (3) Natta, ~ G .Mazzanti, , G., Valvassori, A., Pajaro, G., Chim. e ind. (Milan) 39, 733 (1957). I
,
RECEIVEDfor review May 11, 1960. hccepted October 31, 1960. Presented before the Division of Polymer Chemistry, 137th National Meeting, ACS, Cleveland, Ohio, April 9, 1960.
Rapid Fluorine Analysis by Wide-Line Nuclear Magnetic Resonance HERBERT RUBIN Schlumberger Well Surveying Corp., Ridgefield lnsfrumenf Group, Ridgefield, Conn.
ROBERT E. SWARBRICK Analytical Research Division, Esso Research and Engineering, linden, N. J.
b The salient aspecis of the instrumentation and method for the determination of fluorine in fluorocarbon liquids by wide-line NMR are presented. By means of an integrator, fluorine determinaiions may b e quickly made on single compounds and mixtures. The influence of dilution and chemical shifts in mixtures was studied.
With measuring times of 1 minute, the fluorine content of 20-ml. samples was measured with a relatibe error of about 1%. The limit of detection in peak measurements was about 12 mg. of F. Samples of 0.2 ml. were measured with a relative error of about 2% and have a deiection limit of 3 mg. of F.
CONSIDERABLE WORK has been done with nuclear magnetic resonance in elucidating molecular structure and motion. this method has been little used in quantitative analysis. The specificity of NMR for a particular isotope gives i t a strong position in the field of elemental analysis. I n wide-line measurements, fine spectral HILE
VOL. 33, NO. 2, FEBRUARY 1961
217
' O r -
INTEGRATOR R E A D - OUT
k 8 t
0
FIELD
-
I
Figure 1. Spectral display illustrating overmodulation and integrator readout from 20 ml. of benzotrifluoride Modulation amplitude, 200 milligousr
details are not obtained nor are they necessary. For this reason, the instrumentation and technique are considcrably simpler than in the high resolution method. The major applications of wide-line KMR in analysis have been the measurement of hydrogen in protoncontaining liquids (9>4 , 5 ) arid moisture determinations on hygroscopic solids (3, 6). Fluorine and boron analyses in liquids ( 2 ) have been investigated only briefly. The present study was undertaken to establish the feasibility and analytical details of this method for elemental fluorine analysis in fluorocarbon liquids. The effects of dilution in carbon tetrachloride and the role of chemical shifts in mixtures are also considered. APPARATUS
The instrunient is basically a Schlumberger SAIR analyzer, Model 105 (4) 11ith a n accessory electronic intclgrator, Model 1042. It uses, as part of a resonant circuit, a single roil in scveral standard sizes so that samples bctn cen 40 and 0.2 ml. may be accommodated. The magnetic field is modulated nith a square nave a t 33.1 c.p.s. and the technique of phase-sensitive detection is employed. I n this method, the modulatrd magnetic field is slonly s w p t across the resonance value and variations of the radio-frequency coil resistance are produced. These are measured by a circuit including a radiofrequency amplifier and detector, and an audio amplifier and demodulator. Segative feedback from the output of the audio amplifier to the resonant circuit serves to give stability t o the gain of the system. Changes in the effective resistance of the resonant circuit which occur during iesonance are partially compensated by feedback
218
ANALYTICAL CHEMISTRY
to a barreter, which is a resistive coniponent. The signal applied to the barreter serves to maintain constant t'he Q of the resonant circuit. The resulting signal and integrator read-out are presented on a 10-mv. chart recorder. -4 typical recorder display is shown in Figure 1. This spectrum vias obtained from an 18.48-gram sample of benzotrifluoride using a 200-milligauss modulation amplitude. The integrator readout appears as a bar trace at the end of the measuring cycle. l'he permanrnt magnet has a field strength of 1717 gauss and a homogenrity better than 1 part iii 106 over t'he sanip!e volume. The gal) b e t w e n pole faces is 2 inches. The radio-frequency field strength iiiay be varied in steps over a 60-db. range. Pcak-to-peak modulation amplitude may be varied in ten steps betn-een 0.005 and 5 gauss. The magnetic field may be sn-cpt in any of (Jight spans from 0.1 to 20 gauss. I. 1 lie center of the sn-eep span is atljustable continuously over 10 gauss by a IO-turn n-ire-wound potcntioniet'er. ' h i e of measurement cycle can be varied from 0.5 to 4 minutes in four stcps. The radio-frequemy oscillator frcquency is regulated a t 6.580 Mc. per second by a quartz crystal. l'he electronic intcgrator is esscritially a conventional operational amplifier and programmer. Provision is made for automatic control of the integration, reset, and read-out functions. ?'he 1-oltage fed to the intcgrator conies from a liiglily linrwr potentiometer whose moviiig arm is couplpd to the recorder servomotor by a precision gear train. 130th pnds of the potentiometer are negative n-ith rcspect to t,he ccntrr tap of thc potmtiomr%er which is a t ground potcntial. 13y this mc'nns deflwtions on rither side of rcntcr fwd signals of the saiiir polarity to tlic. intrgrator. The base line of the spc~truin corresponds to a mol-ing arm position coiricirlcnt n.ith the cc,ntc,r tap and a zero input signal to th(, intogrator. The intcgr:itor has scvc~althrcdiolil liiuits nhic,h s r r w to rc>ject tlir. sign:il area bet w e n dcsj gna t t d IC v(,Is B bovc. and bi~low the bas(. line. Hy this means, base line noise and rrrors tlur to shifts of tlw spwtrum v r t~i d y and liorizont,ally may hc avoirlcd. This introduces a small nrgati1.e m-or in thc integrator road-out. In ac>tual practic~c~,this is incorporatcd in thc cdibration curve and appcars as a sm:ill niyytiw iiitcrwpt in the plot of wad-out signal 1's. 111 fluorine. l'he complctc Ineasuring cycle inrluding sprctruni disp1:Ij- and intrgration read-out, is initiatcd hy on(' start s\\-itc,li. DESCRIPTION OF METHOD
I t is profitable to make comparisons between hydrogen and fluorine resonance since the former, being more familiar, serves as reference. The significant differences brtn een hydrogen and fluorine analysis by the n ide-line S N R technique are due to the relative size of the chemical shifts. In proton resonance, the shifts are small and the
-
.iOG1,SS
I-
O
W -I LL W
n lc
d
,I
\i
-I
I f
I1
'J
I GA~SS-I
FIELD
Figure 2. Influence of modulation amplitude on spectral presentation from binary mixture with relative chemical shift of 1 13 p.p.m.
resolution of the instrunient is such that most proton spectra from liquids are included in the envelope of one spectral line whose width is determined by the field inhoniogeneit,y. I n the case of fluorine resonance. the chemical shifts are larger b y one order of magnitude or more and one may obtain spectra which have a single line or multiple lines n-ith various degrees of overlapping. By using large modulation amplitudes, the spectral presentation is transfornied to one in which the positive and negative arches are separated as in Figure 1. Each has the approsiniate shape of a n absorpt'ion spectrum. If this is intqratcd. thc magnitude of t h r read-out obtained bears a direct relationship to the total number of nuc-lti being measured. Thc restriction of this twhnique to liquids is ncccssary because for liquids the effectivc line widths are the same, namely, that of the niagnet inhomogeneity. I n this rase, each spectral line is ovcrniodulated to the saniv eztcnt. Overmodulation is defined a s the use of a modulation amplitudt. which is much larger than the observed line width. K h e n a sample n-ith a single line is to be analyzed, a ratio of about ten of modulation amplitude to observed line width is appropriat'e. This is also applicable t o samples haring closely spaced lines suck as niistures and compounds n.ith nonequivalent fluorines. As the chemical shifts get largcr, the modulation amplitude is increased. I n such caws, and with unknown mixtures containing a n indefinite number of constituents, the general principle is to use a modulation amplitude of sufficient size so that
all the positive arches are recorded before the negative arches appear. By this means, cancellations due to overlapping are avoided. I n Figure 2, the effect of varying the modulation amplitude on a 20-ml. mixture of CCl,F and fluorobenzene containing a total of 4.110 grams of fluorine is shown. Thc,se compounds h a w a rclative chemical shift of 113 1i.p.m. This is equivalent to 194 milligauss in the 1717-gauss field. K i t h 0.0c50 gauss, the lines are resolved and distinct. l’licy are on the verge of overlapping when a modulation of 0.100 gauss is used. Overlapping and partial cancellation occur with 0.200 gauss arid, finally, with 0.500 gauss, the positive and negative arches have been scparatecl and the requirements for proper integration have been met. T h e deercase in resolution accompanying large modulations is caused by the magnifiration of the deviations from sharpnc.ss of t,he squartl ~ ~ ’ a v eMix. tures containing largrr clieriiical shifts give similar presentations n-it11 larger modulations. If there are two or more set? of spectral lines n-hich are sep:trat,cd by large chemical shifts, each sct containing one or more lincs may be :innl~-zcdseparately. The s w e p center adjustnicnt :illon-s any field value n-ithin 1717 = 5 gauss to be taken as center. The n-itle-line technique niakes field hcrniogmeity fairly noncritical so that s:tmplcs may be nicmurcd without searching or manipulation of the fields. The a ~ n o u nof t radio-freyucncy power :ijqdierl to snmples must be kept beloiT a c’ertain level or else the K l I R signal will he attcnuated ( 1 ) . This attenuation, c:illrd radio-frequency saturation, will dvpend upon the rf>laxatioiitimes and is not, generally, the same a t a given radio-frequency power for different substances. For this reason, a n opcrating condition of negligible saturation is u s d . To drtcrmine this condition, a nunibcr of liquids having difforent relaxation times are nic.asurrd over a n-ide range of radio-frequmcy lovels. Because of the fwd-back circuit in the n-ide-line spectronicter, the ncpligihic Gaturation region corrc,sponds to a constant signal level. ‘This occurR a t low radio-frequency p o ~ ~ - eand r s as the r:itlio-fiquency level is incrcased the signal attcnuates rapidly. A convcwient value of radio-frequcncy is then sc~lcc*tc~tl in the platcau region and if s:itur:ition has bccn avoided, a plot of the, signal against the inass of fluorine jr-ill be a smooth straight line. This nwthotl has b w n i i s ~ din this study. Figure 3 is a saturation plot for a w i i p l e of hciizotri luoriclr. For canlibration purposes. all samples w r e prepared by accurately weighing c w h component into the S l I R sample t u h . T o assure that a11 materials n-ill he exposed to the same static,
radio-frequency, and modulation magnetic fields, sample volumes should be approximately the same. The use of a n ordinary calibrated syringe is convenient for sample transfer and easily meets the volume requirements. The major part of this study n a s made with 20-ml. samples sealed into flat-bottomed glass tubes. These tubes nhich have an inside diameter of 31 mni. are Kinible S o . 32500 cold test jars necked don n to capillary opening.. After filling by a syringe and needle, the tubes are sealed to avoid evaporation. Samples of 0.2-nil. arc icalcd
R A 3 I OF R EQ U E N C Y S I T U RAT1 3 h OF B E N Z O T R I F L U O R I D E IFLU3RIhE RESONIhCEl
25 3
9
r 0
$10 3
c
zb-
z
9
- 50
L
-0.0 - z
c
->0
Rt3lOF’ECJEI
~-10 CY LEVE-
4 x ~~
0 IN
iB
Figure 3. Effect of radio-frequency saturation on integrator read-out from 20 ml. of benzotritluoride
of fluorine in each component. The best straight line relationships were found by the least squares method and the mass of fluorine, experimentally found, is obtained by direct substitution of the observed signal in the equation. I n Table 11, the results from pure fluorocarbons are given. The fluorine contents vary betneen 13 and 50% by weight. TKOsets of serial dilutions in carbon trtrachloride were made to uncover any solvent and dilution effects. The results given in Table 111 are negative in this respect. I n Table IV, variations in the chemical shift, 6, are shon-n to cause no difficulties in analysis. With the 20-ml. sample size. a chemical shift of 5 p.p.ni. cannot be resolved whereas one of 55 p.p.m. i. easily resolved. The average absolute error of all the mcasuremente is 0.029 gram of F. The integration technique is gcnerally applicable to pure and mixed substancrs and has a lower limit of detection of about 50 mg. of F in a 20nil. samplr or 2.5 mg. per ml. I t is also poqsible t o use pcak-to-peak nieasurcments n hen single or n itlely separatr lines are present. In this case, a modulation amplitude of 0.05 gauqs is optimum. The loner limit of detection using peak aniplitudcs is about 12 mg. of F or 0.6 mg. per nil. This corresponds to a signal hright of about tnice thc noise level. The instrument conditions appropriate for pmk-topeak mcasurcnicnts rcwlt in Sharper
Modulation amplitude, 200 milligauss
into borosilicate glass tubes S 111111. in outside diameter. In both cases. a radio-frequency coil of appropriate size is used to avoid filling factor l o s s r ~ ~ The samples used in this study \\-ere of the highest purity conimcrcially available. Substances stored in nictal containers wrre filtered to wnio\-c: fine metal particles. RESULTS AND DISCUSSION
Four 1-minute mcasurcnicnts n ere mad(, on each sample using the integrator and an appropriate thrcdiold. The instrument paranietrrs and conditions used for most of the m(xawremcnts are g i w n in Table I. For the binary mixture nit11 6 = 113 p.p.m.. a 2-gauss scan n itith 11as e m p l o ~cd to make crrtain that both cnds of the spectrum ~ o u l dbe on the base liar. Since all other runs nere made with a 1-gauss scan nidth, a factor of t u o in this case serves to put all results on a common basis. The data gave linenr calibration curves n ith integrator readout plotted against the mass of fluorine. This latter is obtained from the total sample weight and the weight fraction
Table I.
Instrument Parameters and Conditions
Filter time-constant Radio-frequency level Sweep time Sweep amplitude Sensitivity LIodulation amplitude Integrator: Sensitivity multiplier Threshold Sample temperature Sample size 3lodulation frequency Field strength Radio-frequencj-
1 Pecontl -22 db. I minutc 1 gauss 100 0 . 5 gan;s 2 0 , 2 mv. 25O 20 ml. 33.1 C.P.S. 1717 gauss 6.880 N c . per
c.
second
spectra n ith higher signal-to-noise ratios and for this reason increase the scnsitivity. K i t h larger sample qizes the concentration detection limits may be furtlirr lowered. JYlicn lesser quantities of sample are analyzed i n the small radio-frequency coil, the precision is not as good; however, the detection limit in grams of F is lower. This low value results from exposure to a more homogeneous field VOL. 33, NO. 2, FEBRUARY 1961
219
with attendant increase in resolution and signal-to-noise ratio. If a signalto-noise ratio of two is taken as the smallest acceptable signal for analysis, the lower limit of detection is 3 mg. of F or 15 mg. per ml. using peak measurements. Integrated signals may also be made with the 0.2-ml. samples but the
Table 11.
Substance p-Fluoroanisole m-Fluorotoluene p-Fluorotoluene o-Fluorotoluene 1-Fluoro-4-nitrobenzene Fluorobenzene Trichlorofluoromethane m-Bromobenzotrifluoride Trichlorotrifluoroacetone Benzotrifluoride CCIzFCClFn Trifluoroacetic acid CF3CC12CFClCFS 0
Analysis of Pure Fluorocarbons
Mass of Fluorine, Grams Exptl.4 Absolute error -0,039 3.274 -0,004 3.365 -0,013 3.361 3.480 4-0.013 3.649 -to. 020 3.979 $0.023 -0,050 4.230 $0.004 8.192 8,999 +O 096 9.254 + O . 045 9.534 $0 ,039 15,095 -0 ,022 16.058 -0 ,037 Av.= 0 ,032
Theor. 3.313 3.369 3,374 3.467 3.629 4.002 4.287 8.188 8.903 9.209 9.495 15.117 16,095
Average of four determinations.
Table 111.
Serial Dilutions in Carbon Tetrachloride
Solute m-Bromobenzotrifluoride
CC12FCClF2
(1
detection limit, as for larger volumes. is higher. The reproducibility of the nideline analysis was determined from a series of integrator measurements taken over a 2-week span. The relative error in fluorine content corresponding to 1 u limits was slightly under + I % for
Theor. 1.688 3,307 4.875 6.520 1.880 3.744 5.645 7.592
Mass of Fluorine, Grams ExptLa Absolute error 1.659 -0.029 3.289 -0.018 4.854 -0.021 6.452 -0.068 -0.038 1.845 3.750 + O . 006 -0.050 5.595 -0.019 7.573 Av. = 0.031
.4verage of four determinations.
Table IV.
Analysis
of Mixtures with Varying Chemical Shifts
Mass of Fluorine in Components, Grams Fluorobenzenea o-Fluorotoluenea 2.731 0.793 1.585 2.038 1.373 2,386 0.709 3.165 Benzotrifluoridee 1.847 3,901 5.592 7.284
o-Fluorotoluenec 2.712 2.079 1.371 0.710
Fluorobenzened 0.795 1.597 2.410 3.347
Trichlorofluoromethaned 3.318 2.499 1.700 0.851
Mass of Fluorine, Grams
Theor. 3.524 3.623 3.759 3.874
Exptl.* 3.544 3.665 3.790 3.908
Deviation +o 020 +O 042 tO.031 +0,034
4.559 5.980 6.963 7.994
4,559 5.995 6.990 8 014
0.000 10,015 t0.02i t o 020
20-ml. samples. With 0.2-ml. s a n i p l q reproducibilities of *2% relative were found over a similar period. The experimental precision on a shortterm basis was evaluated from ten sets of runs with four consecutive 1-minute measurements in each run. The sample used was 20 ml. of benzotrifluoride containing 9.209 grains of fluorine. The results give the 1 u instrumental error of a single measurement made during a 4-minute period. The absolute error under these circumstances is 10.025 gram of F and the relative error is 10.2i75. It is, therefore, possible to improve the long-term precision by using standard samples to correct instrumental fluctuations. While this work is primarily concerned with fluorocarbon liquids, it is clear that some types of solids may also be analyzed. T h e n solids arc dissolved or melted, the dipolar interactions which are responsible for the large spectral line widths are averaged to zero and the lines become narrow as in liquids ( 1 ) . Therefore, solution in a suitable solvent givcs the sharp-lined spectra which are appropriate for analysis by the technique described in this study. Aqueous solutions of electrolytes can be analyzed and this work will he presented at a future date. -4n example of a direct fluorine determination on a solid is the analysis of an ore containing more or less of one fluorine-containing species, CaF2. Because of the large line widths, the integration technique described for liquids cannot be used. However, peak-topeak measurements yield a linear calibration plot for a series of fluorspar ores in which the fluorine content rnngtd from 34 to 46% F by weight. Since line widths differ, data for calibration curves must be obtained for each compound. The technique of wide-line NMR has the general advantages of being specific, rapid, and nondestructive. Further, it is precise and does not suffer interference from nonmagnetic substances. LITERATURE
(1) Bloembergen, N., Purcell, E. M., Pound, R. I-., Phys. Rev. 73, 679
(1948).
(2) Ferrett, D. J., Trans. SOC. Instr.
Tech. 11,66 (1959). (3) Rubin., H.., Cereal Sci. Todau" 3.. 240 (1958). (4) Rubin, H., IRE Trans. I d . Electronics PGIE-I 1 , 9 (1959). (5) Shaw, T. M., Elsken, R. H., ANAL. CHEM.27,1983 (1955). (6) Shaw, T. AT., Elsken, R. H., J . A p p l . Phys. 26,313 (1955). \
4.113 4.096 4.110 4.198
-0.013 -0,041 + O . 048 +0 005
4.100 4.055 4.158 4.203 Av. =
6 = 5 p.p.m. Average of four measurements. c 6 = 55 p.p.m. d 6 = 113p.p.m.
a
220
ANALYTICAL CHEMISTRY
CITED
,
0.025
RECEIVEDfor review July 28, 1960. Accepted November 25, 1960. Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy, Pittsburgh, Pa., March 1960.