1538
13
Anal. Chem. 1980, 52, 1538-1539
-12
-11
LOG
Figure 1.
-10 GRAMS
-3
8
7
5
INJECTED
Response of tetrapropyltin by two different mechanisms
Figure 2. Response of a quartz wool modified FPD to 100 fg of tetrapropyltin
addition to the other desired effects. The simple modification was to insert a loose plug of quartz wool right above the shielded flame. T o demonstrate the pronounced effect this produces, the regular quartz chimney of a Shimadzu FPD was replaced by a borosilicate tube of equal dimensions. This tube had slight indentations to hold a plug of quartz wool in place against
the detector gas flow. Other conditions were close to those described in our earlier paper (1). Figure 1 shows calibration curves of tetrapropyltin with and without the quartz wool inserted. Different emissions are observed under the two sets of conditions. With dark-adapted eyes it is easy to see them. Without quartz wool, tin compounds produce an elongated red glow (SnH band) right above the flame, surrounded by a grayish, diffused luminescence (sometimes attributed to SnO). When the quartz wool plug is in place, an intense blue glow appears a t its lower end right above the flame. The emitter responsible for this continuum, which peaks a t 390 nm, is unknown. Figure 1 shows calibration curves for tetrapropyltin with and without quartz wool. The improvement brought about by using the surface luminescence as opposed to the gas phase emission (with perhaps a minor amount of surface luminescence present) is evident. Despite the apparent simplicity of a FPD modification t h a t involves nothing more than the addition of some quartz wool, these results are better than those obtained with earlier models. These had given a slight deviation in the calibration curve, which now has vanished. Working with the quartz wool detector for some time, furthermore, demonstrated that it is less susceptible to contamination. The closeness of the flame to the quartz wool seems to dispose of certain impurities that would otherwise block its function. The detector is also more sensitive: the minimum deg, corresponding tectable amount of tetrapropyltin is 4 X to 5 X mol (about 2 million atoms) tin/s. Figure 2 shows a typical chromatogram of 0.1 pg tetrapropyltin. Note that due to the surface effect, the peak is slightly broader than one would expect from the regular gas chromatographic dispersion. We did not attempt to modify other commerical FPDs by insertion of quartz wool. We do assume, however, that similar improvements in tin response would result.
LITERATURE CITED ( 1 ) Aue, W. A,: Flinn, C. G. J . Cbromatogr. 1977, 142, 145-154
RECEIVED for review July 9, 1979. Resubmitted March 18, 1980. Accepted March 18,1980. This research was supported by NRC Grant A9604 and by a grant from the Department of Fisheries and Oceans.
Use of an Automated, Stepping Differential Calorimeter for the Determination of Molecular Weight J. Zynger” and A. D. Kossoy Lilly Research Laboratories, Eli Lilly and Company, Indianapolis, Indiana 46206
A recent report ( I ) from this laboratory describes an automated, stepping differential calorimeter which is used to determine purity through the analysis of the thermogram generated by stepwise melting of a sample. Assuming ideal behavior between the components of the sample, the inverse of the fraction melted of the main component when plotted vs. the temperature should yield a straight line that is defined by the van’t Hoff equation (2):
where T , = instantaneous sample temperature, To= melting 0003-2700/80/0352-1538$01 .OO/O
point of the pure compound, R = gas constant, AHf = heat of fusion, F = fraction melted a t T,,and X = mole fraction of impurity. In our laboratory, this technique has been used for over three years for about 3000 assays of mol 70 purity. Implicit in Equation 1 is the designation of two molecular weights which are incorporated into the mole fraction impurity term. Therefore, the data derived from the melting behavior of a mixture consisting of a known weight of a small amount of material of unknown molecular weight “impurity”, and a known weight of a relatively large amount of an appropriate host substance of known purity and molecular weight “main component” should yield the molecular weight of the unknown. Hence, by simply modifying the calculations in @ 1980 American Chemical Society
ANALYTICAL CHEMISTRY, VOL. 52, NO. 9, AUGUST
structure
name
I
+Y'
2
mol wt
benzoic acid 122 (Reagent grade )
1
amobarbital 226 (Lilly standard)
_-
\-,-
compoundno.
%a
betweensample error, avM.W. %
121 133 114 132
7.6 6.9 &.l 4.8
125
2.5
2(M.W. 2 2 6 )
1 2 3
7.5 18.5 4.4 s.7
223
1.3
4
213 240 208 231
3(M.W. 3 7 2 )
1 2 3 4
382 371 362 362
6.7 9.5 6 .5 1.o
369
0.8
4(M.W.437)
1
408 487 405 342
5.4 6.7 8.2 5.7
411
5.9
-3
2 3 4
flurandrenolide 43 7 (Lilly standard)
RSD,
1 2 3 4
rz-t-
4
Sam- withinple sample no. avM.W.
1 (M.W. 1 2 2 )
nabilone ( 3 ) 37 2 (Lilly standard) -
1539
Table 11. Results of the Analysis
Table 1. Compounds Chosen for Study compound no.
1980
Relative standard deviation. Table 111. Analysis of Variance Components (ANOVA) Compound Numbers
F
Equation 1, the data generated by the stepping calorimeter can be used not only to assay for purity, but also to determine the molecular weight of a n unknown compound. Data are presented which demonstrate the applicability of this procedure to a variety of compounds with a molecular weight range from 122 to 437. This procedure requires a minimal amount of sample (300 p g ) and exhibits an accuracy of 3 % .
EXPERIMENTAL The molecular weight of an unknown compound can be determined calorimetrically when the compound and the host component meet two essential criteria. First, the melting point of the host component must be depressed by the presence of the unknown. Second, the unknown cannot decompose when held at the melting temperature of the host component while in the presence of the host component. Owing to its chemical inertness, its relatively low melting point, and its solvent properties, phenacetin (NF grade) was used as the host component with a defined purity of 99.80% and with a defmed heat of fusion of 7610 cal/mol. Procedure. Phenacetin is normally weighed, between 3 and 6 mg, to an accuracy of 1 pg into aluminum sample pans. The compound to be analyzed is then weighed, between 100 and 300 pg, directly on top of the phenacetin. The pan is sealed using a Perkin-Elmer volatile sample sealer, and analyzed using the instrumentation and computer program previously described ( I ) . Weighing inaccuracies (less than 100 pg) at low impurity levels and the imprecision of the technique at high impurity levels (I), dictate an optimum impurity concentration of about 3%. All samples are prepared to contain about 3 mol 90 of the compound to be analyzed and all samples are pre-melted in the instrument before analysis. RESULTS AND DISCUSSION T h e compounds chosen to demonstrate the applicability of this technique are shown in Table I. The materials were chosen for their diverse molecular weights and for their differing chemical functionalities. Four different samples were prepared for each compound and each sample preparation was analyzed in triplicate. T h e results of these 48 experiments are tabulated in Table 11. Assuming a normal distribution for all the data, the average RSD is 7.3% for a single analysis of a single sample preparation. T h e between-sample, average molecular weights and their percent error are also shown in Table 11. The average percent
variance
1
2
between samples within samples
25% 75%
10049
0 %9
3
4
0% 100%
80% 20%
error for the four compounds is 2.6%. This is in good agreement with the expected error of 2.1% which one would obtain with a n RSD of 7.3% and 1 2 replications (4 samples-each run in triplicate). T o determine the relative contributions of the variance in preparation and the variance in determination, the data was analyzed by the technique of analysis of variance (ANOVA, 4 ) . The results of ANOVA are listed in Table 111. Compounds 2 and 3 indicate that essentially all the variance lies in t h e analysis, while compound 4 indicates that most of the variance lies in the sample preparation (weighing error). If we average the variance components across the 4 compounds, we find that one fourth of the variance is due to sample preparation and three fourths of the variance is due to analysis. Therefore, the procedure that we have established for routine molecular weight analysis involves two sample preparations with each to be analyzed in triplicate. With a RSD of 7.370, t h e error which will be observed from the average of these six analysis will be 3.0%. I t must be remembered that the instrumentation used includes a Perkin-Elmer DSC-1B calorimeter with a maximum sensitivity of 40 pcal/s (5). Newer instrumentation, the Perkin-Elmer DSC-2 for example, is available with a factor of ten greater sensitivity ( 5 ) . Since most of the error in this type of analysis arises from instrumental noise, it a p p e m that one could obtain a dramatic increase in precision, and thereby in accuracy, by using a more sensitive calorimeter.
LITERATURE CITED (1) J. Zynger, Anal. Chem., 47, 1380 (1975). (2) F. D. Rossini, "Chemical Thermodynamics". University of Notre Dame, Notre Dame, Ind., 1949, pp 89-1 11. (3) R. A. Archer, W. 8. Blanchard, W. A. Day, D. W. Johnson, E. R . Lavagnino. c. w. Ryan, and J. E. Baldwin, J. Org. Chem., 42, 2277 (1977). (4) E. L. Crow, F. A. Davis, and M. W. Maxfield. "Statistics Manual", Dover Publications, New York, 1960. (5) Perkin-Elmer Instrument manuals for the model DSC-1B and the model DSC-2.
RECEIVED for review June 8, 1979. Resubmitted April 7, 1980. Accepted April 7, 1980.
