Analysis of Natural Gas

---

Analysis of a Natural Gas By Volumetric Chemical Methods and by Mass Spectrometer MARTIN SHEPHERD, National Bureau of Standards, Washington, D . C .

A standard sample of natural gas was analyzed by 50 laboratories in cooperation with Subcommittee D-3-VI1 of the American Society for Testing Materials. The conventional chemical methods were used by 30 laboratories, and the other 20 employed the mass spectrometer. The results are compared in this report.

T

HE task of standardizing the methods of analysis of gaseous fuels has been undertaken by Subcommittee VI1 of Committee D-3 of the American Society for Testing Materials. A Jeries of cooperative analyses of standard gas samples was planned

frequency with which they occur. For example, in the plot for the chemical determination of carbon dioxide (left-hand section of Figure 1, C), it will be seen that one analysis gave 0.6%

t o furnish part of the basic information necessary to the development of these methods. (These samples are not to be confused with the regular standard samples prepared and offered for sale by the Sational Bureau of Standards. They are mivtures prepared especially for these cooperative analyses and are issued to laboratories cooperating with the American Society for Testing hlaterials on this project.) The plan of this development has been outlined in t x o previous reports (5, 6). The analysis of a standard sample of natural gas by the volumetric chemical methods of absorption and combustion (8) and the analysis of this same standard sample by the mass spectrometer (9) have been reported. The preparation of this standard sample of natural gas, the measurement of its specific gravity and heating value, and the instructions for its transfer to the analytical apparatus have been given in detail (‘7,8). The two reports cited also give a general account of the cooperating laboratories and the models of apparatus and specific methods of analysis employed. All the above information is necessary to the evaluation of the data presented in this paper. The purpose of this report is to assemble the analytical data from these two series of cooperative analyses, and briefly to compare and discuss the results from the chemical and physical methods. Since the data show the real state of gas analysis throughout this country with respect to these methods, they are worth more than a moment’s reflection on the part of those who analyze gases or use the analyses.

MANNER OF PRESENTING

ANALYTICAL DATA

j

All the analytical data have been tabulated, but long contemplation of these tables (5, 6) will not reveal what may be seen at tl glance when these same data are presented in the form of frequency-distribution plots. These plots amount to actual pietures of the analytical data and present the greater part of the whole story with most of its implications so clearly that what few remarks seem desirable have mostly been included in the legends of the corresponding plots. Thus, the remarks and the plots remain desirably connected, and the reader may acquire the essential facts for himself without the interposition of superfluous words and gratuitous arithmetic. The frequency-distribution plots are in fact the real meat of the report. In each plot, the results of the chemical analyses have been given in the lower section or in the left-hand section, depending upon the shape of the plot. The results of the analysis with the mass spectrometer are given in the upper section or in the right-hand section. Each circle of a plot represents a value derived from a single determination of the substance whose name appears in the legend. The circles are plotted equidistant on the ordinate corresponding to their value. Thus, the abscissas are values derived from the analyses, and the ordinates indicate the

I

11.. PERCENT

’

I

I5

I

1

5

IO

CARBON DIOXIDE

Figure 1.

.rii

10

PERCEN

Frequency-Distribution Plots for Carbon Dioxide and Oxygen

Carbon dioxide. There is n o disagreement between the chemical determinations and those made by mass sprctrometer. Highest frequency and mean value of both groups are found a t 1.00%. Rsproducibility is equally good. Oxygen. There w x s no significant amount of oxygen in thin sample, but this fact was correctly reported in only 42% of chemical determinations and 8470 of spectrometric determinations (discarding values less than 0.0570.) The error is much more apparent in chemical analysis and is ciused by improperly equilibrated reagents. This type of error c i n be corrected. Larger amounts of oxygen found hy chemical method prohably have as source leakage during trawfer of sample t o analytical apparatus.

635

V O L U M E 19, NO. 9

636 Table I.

Comparison of Most Probable Measured Spectrometric and Chemical Values of Chemical Values Corrected to Spectrometric Basis, and of Spectrometric Values Translated to Chemical Basis

Group 3 Chemical -4nalysis Corrected io; Group 4 CsHs and CaHs Found by Spectrometer Spectrometric Analysis Translated . Method to Chemical Basis b y Calculating Group 1 B C3Ha a n d CsHa as CHa and CZHI Analysisby Mass Method Method SBS Method Method Gas Spectrometer ' A B B analyses A B iklole % Mole % Mole % Methane 77.6 79.6 79.0 77.9 74.5 74.7 Ethane 14.9(?)a 12.9 14.0 14.7 21.0 20.8 Nitrogen 3.4 3.1 3.0 3.5 3.4 3.4 Propylene 0.2 ... ... 0.2 0.2 0.2 .. .. Propane 2.8 ... ,.. 2.8 2.8 2.8 Carbon dioxide 1.0 i:o 1.0 1.0 1.0 1.0 1.0 i:o i:o Total 99.9 99.6 100.0 100.1 99.6 100.0 100.1 99.9 99.9 0 This value is somewhat questionable. Mode a n d median indicate 14.7 as a better choice, and best chemical value corrected for C3 hydrocarbons supports lower value. Group 2 Analysis b y Chemical Methods' A1 et h od B RIethod Method NBS A B B analyses Mole % 76.5 76.1 75.0 19.0 19.9 20.6 3.1 3.0 3.5

+

+

carbon dioxide, five analyses gave 0.77,, twelve analyses gave the spectrometric values. Finally the spectrometric values can be translated to the chemical basis (see legend of the frequency0.8%, and so on. distribution plot for methane, Figure 4, for the manner of these In addition to the constituents shown in the frequency-distribution plots, the mass spectrometer occasionally found gases not corrections). This is done, in Table I. actually present in the sample. The great sensitivity and anaThe chemical values' with the designation A are those calculytical resolving power of this instrument will sometimes serve lated from total contraction and carbon dioxide produced, while to detect small amounts of gases previously sampled which have those marked B were calculated by accounting for oxygen consumed in addition. Groups 1 and 3 should be in agreement, but not been thoroughly desorbed and discarded. Xor do certain agreement is not attained except in the case of NBS analyses chemical reactions within the spectrometer seem altogether imby method B. Groups 2 and 4 should be in agreement, but again possible. this is achieved only in the case of the NBS analyses. However, There was no carbon monoxide in this sample. This fact was the analyses by the B method show promise in comparison with the A method. established by tests w-ith the National Bureau of Standards carbon monoxide indicating tubes ( 5 ) . Carbon monoxide was inErrors which appeared in the frequency-distribution plots of dicated in nine spectrometric determinations, in amounts from (8) have been corrected in the present set of plots. These correc0.2 to 3.3%. This error usually results from confusion with nitions have not altered the picture in general or in significant detrogen in reading the spectrogram; but when oxygen is present, tail, but are nevertheless made to straighten the record. carbon monoxide can be generated within the spectrometer. In the chemical analysis of a natural gas, carbon monoxide is not CONCLUSIONS sought. If the chemical combustion data are computed accordThe mass spectrometer has in general given a better reproduciing to the formula used for a water gas, however, carbon monbility than that obtained by the chemical methods. This is not oxide may be indicated. For example, five chemical determinations, resulting from such an incorrect calculation, indicated 0.5y0of carbon monoxide. There was no hydrogen in this sample. Its absence was established (within * O . O O 1 ~ O ) by separation at the temperature of liquid hydrogen, using the apparatus and 0. methods previously described (6). Hydrogen was indicated in 11 of the 118 determinations by the mass spectrometer, always in small amounts and with random distribution. 0.. 0. *, Hydrogen is not sought in the chemical analysis of a natural gas; but if the combustion formulas suitable for a water gas are used by mistake, hydrogen may be indicated. As much 1 I I I I &s 0.501, was thus indicated in five chemical determina' C I I 1 1 . 1 I I tions. The chemical methods ignore butanes, and it is questionable whether the spectrometer should have indicated + these gases. Butanes were reported in 15 of 118 deter0.. i.. 0 . . 6 VPLUES TFF SCALE minations, in small amounts ranging from 0.02 to 0.170, and with random distribution. It is doubtful if as much as 0.05% of butanes was present. One laboratory analyzed a fraction of this sample condensed a t low temperature to .........r..........i**o*t**o*o~moo concentrate C1 hydrocarbons, and reported none. B. 4 Butenes were reported in only 6 of the 118 spectrometric analyses, and the amounts indicated were small (0.03 Figure 2. Frequency-Distribution Plot for Nitrogen to 0.1%). I t is very doubtful if any butenes a t all were presAnalyses by chemical methods yielded values ranging from 0.1 t o 6.0%, ent in this sample. with n o sharply defined maximum frequency. Mean of all values between 1.5 and 5.070, inclusive, is 3.1%. Analysis i n a n all-glass apparatus at NaBefore laying aside the frequency-distribution plots, a list tional Bureau of Standards gave value 3.5 * 0.170. Determinations with of the most probable measured spectrometric values can be mass spectrometer yielded values from 1.6 to 4.9 %, with greatest frequency a t 3.4%. Mean of all determinations is 3.470, and mean of values from compared with the most probably measured chemical 0.3. Of 118 spectrometric determinations 3.0 t o 4.0, incluuive, i s 3.5 made, 114 reported nitrogen and 4 confuued nitrogen outright with carbon values, and the chemical values, corrected for the presence monoxide. While the maaa apectmmeter has demonstrated a better reof 2.8y0of CsHs and 0.2% of CIHB, can be compared with producibility, reaultm obtained by both methods are uncomfortablydiverse.

Is i I I i :...l I I ..... ..... ...... ........ ........... .......... .+.. ..................... ..........................

....

........... .+ ............ ............... ................ ................. .................. ..................... .......................... ............................. .................................. f

.

632

S E P T E M B E R1947

.

lytical picture. However, Rith but two models of spectrometer, a definite individuality was shown (see frequencydistribution plots of 9). This individuality was partly subnioiged, since there were only three of one kind I compared with eighteen of the other. In the case of the * I m i r.~.....r...i spectrometric analysis, there was a definite standardizaI I I I I I 1 I l tion that seemed lacking in the chemical approach. Had there been as many kinds of spectrometers and methods of using them as there were models of chemical apparatus and methods of using them, the results from j .../....‘.a. a . the spectrometric and chemical methods might have not i.... a differed so greatly with respect to reproducibility. The .*.............a ..)a a .* present advantage of spectroscopic standardization ..!....*....t....t.. ..e.... should be officially captured and maintained; a tentative ?q.S 75.0 765 7&0 76.5 77.0 77.5 780 76.5 79.0 PERCENT YETWANE standard is actually being prepared by Subcommittee A.S.T.M. D-3-VII. The chemical methods should be acFigure 3. Frequency-Distribution Plot for Rlethane corded the advantage of standardization which they SO Amounts of methane found by chemical analysis and mass spectrometer evidently lack and need. naturally will not agree. Chemical analysis does not disclose true composiNany of the explanations of the relative incoherence of tion of a mixture containing more t h a n two hydrocarbons of a related series. I n analysis of this sample, CsHa was calculated as 2C2H6 - CH4; C3He either as some of the chemical analyses have been given in pubCHL- 5/3C3He and C2Hs i- 7/3C3Hs, or CHI - 2/3CsHs and C I H ~ -t5/3C3Hs, depending on whether CHI and CzHs were calculated from total contraction and lished reports dealing with errors in volumetric gas carbon dioxide produced, or from these two plus oxygen consumed. If most analysis. The list is long and will not be given here, but probable values indicated by mass spectrometer for C3Ha and C3Ha are accepted, net errors i n chemical analyses would be: CHI, 3.1% too low and C9H6, errors which are not usually taken into account have been 6.1% too high when computed from total contraction and carbon dioxide produced; or, CHL,2.9% too low and CzHs, 5.9% too high (when oxygen condiscussed in a series of papers dealing specifically with eumed is included). Corrections do not bring chemical and spectrometric the analyses .of pure gases (2). There has previously been methods into full agreement. For chemical method, a n exceedingly wide distribution is noted. Actually, determinations ranged from 69.0 t o 85.5, a no especial R idespread enthusiasm for using available inepread of 16.5 YO,compared with a spread of 4.6 70(75.4 to 79.0) for spectrometric determinations. There is no tendency toward a marked maximum frequency formation or acquiring additional knoxTledge to improve i n either aeries. [Maxima appear when plotted i n intervals of 0.5, and such analytical procedures; but the experiment of a countryplots indicate proper selection of valuee t o be included i n mean (8, footnote 3).] Mean of chemical results taken from 73.5 t o 80.0 (94% of all values) is 76.5. u ide cooperative analysis clearly demonstrated the need Mean of spectrometric results taken from 76.7 t o 78.4 (80970 of all values) is 77.6 0.5. “Corrected” chemical mean is 79.6. 270 higher t h a n spectrofor correction and standardization, and the stiff commetric mean. Chemical analysea performed a t National Bureau of Standards, petition offered by the mass spectrometer has added a using all-glass apparatus and taking account of oxygen consumed in calculations, gave a mean of 74.9 * 0.2 for methane. Corrected as indicated above. healthy stimulus in this direction. t h i s hecomes 77.8 0.2, i n agreement with spectrometric mean. The mass spectrometer is relatively so new and has proved to be so extraordinarily useful that for the most true with respect to the selected chemical method, which yielded part it has simply been used, and not too much attention has been . results in agreement with the known or most probable values given to measuring exactly how good it is and in detail what it will and d l not do. Accordingly, the literature is meager and with the best spectrometric values, BS noted in the legends of the frequency-distribution plots. But, on the whole, the and offers few explanations for the relative success or the OCchemical determinations were, by comparison with the spectroeasional shortcomings of this instrument. metric determinations, inconveniently various. The explanaThe analytical data reported in this cooperative analysis distions of this fact will probably neatly equal the commentators close that in any one laboratory, different computers reading, in number. There were seventeen models of chemical apparatus and two models of the mass spectrometer used in the cooperative analysis. Three methods were employed in the chemical analysis, with many significant variations of actual procedure (the total number a. being unknown). There was e. essentially one method employed in the spectrometric analysis, with few variations 1 6 VALUES of procedure, none of which OFF SC&L€ appeared particularly significant. K i t h only minor ex1 1 ceptions the spectrometric analyses were conducted as pre140 145 150 155 160 I 205 210 215 221 scribed by the makers of the spectrometers. These inFigure 4. Frequency-Distribution Plot f o r E t h a n e struments have been availChemical values have little regard for Gaussian pattern. with a wide distribution from 12.5 to 24%! a able for a relatively short epread of 11.5% as compared with 2.3% for spectrometric values. Chemical plot shows no maximum d l e tribution, although a maximum of 18.5 is indicated when interval 0.5% is plotted ( 5 ) . Chemical mean time, evidently insufficient for computed from values between 17.0 and 21.5%, inclusive (8970 of all determinations), is 19.0. Spectrometric mean computed from values between 14.4 and 15.6, inclusive (80% of all determination.), is 14.9 i the development of excellent 0.3, and greatest frequency appears a t 14.7%. Chemical mean, corrected as indicated i n legend for Figure 3, is 12.9970, not i n agreement with spectrometric result. Chemical analyses performed a t National Bureau but diverse ideas such as of Standards gave value 20.6 0.2 for ethane. Again corrected am indicated, this become. 14.7 * 0.2 sometimcs complicate an anawhich agrees with spectrometric value. I

........................ .... .... ................ ................. ........ ....

.

I:....... : !: ............ ....................... I : :I: I . .... ............ .............................. f ....

..~..../....*...

f

...

L

-p-

..... ...... ........... ................

i I:

+

638

V O L U M E 19, NO. 9

interpreting, and computing the same spectrogram produced results in almost exact agreement. Disagreements did not exceed 0.1% even for components present in the largest amounts. On

the other hand, within one laboratory two spectrograms taken with the same sample on the same spectrometer usually differed by amounts ranging from 0.1 to 0.8%. Disagreements of 0.4 or 0.570 were fairly frequent, and the average best work yielded checks to 0.2 or 0.3%. (In a few cases, checks to 0.1% or better were reported.) When the probable error of the spectrometer's gasometric system is considered, it is difficult to see how closer checks can be expected, as matters now stand. The sample is measured on a closed-end mercury-in-glass manometer of about 5-mm. bore. Approximately 35 to 40 mm. of pressure is estimated to the nearest 0.1 mm. The reading error alonescan be 0.1 mm. or 0.2 to 0.30/,. Everything else in this extremely complex system must then be perfect or compensating if checks to this degree are to be regularly expected. (This seems an unnecessary handicap to the volumetric gas analyst who regularly estimates 100 ml. to the nearest 0.02 ml., and with an accuracy of 0.05 ml. or better when a11 volumetric errors have been accounted for. Such results are obtained with ordinary volumetric apparatus; special apparatus is capable of greater accuracy.) A source of real astonishment to the volumetric gas analyst is derived from the fact that the gas pressure in the analyzing tube of the mass spectrometer is of the order of magnitude of mm. of mercury, considered a reasonable vacuum not so many years ago. The remarkable sensitivity of the spectrometer to minute amounts of gases may be the source

--

-

J-

. . .. ... . .. . . .. .. .... .. .... . .... . . .... .. .... .... .. ... ... ..... .. .... ..... ... . ... ..... ... ..- -+-1:

e.

m.

* * a .*e ..a

..e

0 . 4

*.a e..

...e * .a

4

a

4 4

a

e.... m....

4

I _

PERCENT

Figure. 5.

PROPANE

PERCENT

ETHYLENE

. .. .. ... ... . ...... .... .... ...... ...... ... ... ... 0

e.

e.

0..

0.. 0..

PERCENT PROPYLENE

Frequency-Distribution Plots for Propane, Ethylene, and Propylene

Table 11. Estimate of Probable Acourac) Determination (Constituent)

Crude 90% Range

70

%

Methane, chemical Ethane, chemical h-itrogen, chemical Methane, AlS Sitrogen, hIS Propane AlS Carbon dioxide cheniieal Carbon dioxide, AIS Ethylene, hfS Oxygen, chemical Oxygen, RlS Propylene, .\IS

25 5 4.5

1 (0.1) 1 (0.1) 1 (0.1) l(O.1) l(O.1) 0.1 0.1 0.1

'

2.3 1.8 0.8 0.7 0.6 0.5 0.4

0.23 0.18

Report t o Shepherds Refereeb

... ...

0.01

7% 0.5 -3 0 . 5 -3 0.5 -3 0 . 2 -1.5 0.2 - 1 . 5 0 . 1 -0.5 0.1 -0.5 0 . 1 -0.4 0 05-0.3 0 05-0.3 0 02-0.15

0.02-0.1

Parenthetic values refer t o unclear statements in preceding text. b First value is best accuracy justifiable, second IS magnitude of least difference between individual determina,ions worth serious thought a

of its greatest confusion, for changes of composition of the sample by sorption and desorption in the measuring manifold and other parts are vivid possibilities. The reproducibility achieved among laboratories does not justify reporting methane, ethane, or nitrogen to the optimistic 0.01To, or the more conservative 0.1% conventionally accepted as representing probability; and this is true for both the chemical and the spectrometric methods. When the amount of the hydrocarbon is of the order of 10% or greater, it is realistic t o think in terms of whole per cent if the results obtained by one spectrometer (or chemical apparatus) selected a t random are to be compared with the results obtained by another spectrometer (or chemical apparatus) again selected a t random. In the case of the spectrometer, amounts of propane around 3y0 can conscientiously be reported to the nearest 0.1%; and propylene present to the extent of 0.270 can probably be estimated to 0.01%. Carbon dioxide of the order of 1% can be reported to 0.1% with either method. This is about how the matter now stands, and if the. needs of engineering and research are evaluated with equal realism, it may be that these results will often be adequate. But since the results can be improved, this seems the sensible thing t o do. I t hm been noted that the standard mixture dealt with is a simple one, and that less concordant results may be expected from more complicated mixtures of this type. This is true. It has also been claimed that accuracies of the order of 0.01% are possible. The data have in general failed to demonstrate this.

Propane.

Chemical analysis does not account for propane.

.

Spec-

rrometric results are spread over a range of 1.2 %, with a mean value of 2.7; 78% of determinates lie within 0.4% with a mean of 2.8 * 0.06.

Greatest frequency is a t 2.8. iently similar.

These results are b e a m i n g a n v e n -

Ethylene. Chemical analysis for unsaturated hydrocarbons is discussed below. Spectrometric analysis produced oud result t h a i ethylene was indicated i n 51 out of 118 determinations (and by 18 of the 21 instruments). This uncertainty seems interesting, 43 % voting for, 57% against. There was no such uncertainty concerning propylene. There was no positive correlation between low propylene and indicated presence of ethylene. Greatest frequency for rthylene appears a t 0.1, and mean of all plotted determinations in 0.2%.

Propylene. I n t h e analysis of a natural gas by ehemical methods, unsaturated hydrocarbons are not alwaya determined, and, if they Chemical determination of this group is usually i n error because of difficulty in equilibratina usual reagents with respect to other constituents of gas sample. Uowever, modified volumetricprocedures are capable of yielding good results for unsaturated group as a whole. A series of chrmical analyaes made a t hatianal Bureau of Standards, employing apparatus described ( 4 , Figure 2, part 2) and general procedures described in (3) gave 0.19 * 0.04% for total unsaturated compounds in this sample, in evcellent agreement with propylene found by mass spectrometer. Whilr only half of spectrometric analyse. appeared to separate ethllene as a constituent of this sample, identiticatinn of propylene was almost unanimous and result8 are remarkably consistent. Propylene waa reported in 114 of 118 determinations, with greatest frequenrv a t 0.2970, a mean of 0.2 i 0.04. and a total apread of only 0.3970. These resulta attest remarkable analytical power of spectrometer i n this instance. arc, cannot be individually separated.

'

S E P T E M B E R 1947

639

1. . ...... *

I

b

.

.

e

e.

_ .

I

j

..... ......... ............ ................ .............:~ ............ ............... a .e

e..

.I

).a.

7 VILUE: OFF S C I

I

I

’

spectrometer’s approach is occasionally too enthusiastic, and an additional compound or two are thrown in for generous measure, the errors may be largely amended by aduptiiig a standard procedure designed properly to equilibrate the spectrometer with respect to the sample whosc composition is desired. In the absence of possible chemical interference, the disposition to indicate compounds that were not in the sample may be overcome. But, even though the chemical methods of absorption and combustion will not resolve such mixtures with respect to all the constituents which may be present, nevertheless, chemically derived analytical data have been useful for many years and will continue to be even more useful when the methods are standtu-dized. The specific gravity and heating value can be calculated from a good chemical analysis about as well as from an analysis by the mass spectrometer, and the combustion characteristics of a fuel gas can be directly measured by the cliemical methods. If such things are desired there seems no justification for an investment in apparatus which nould cost about 50 times tls niuch in one case as in the other, exclusive of personnel requirements. ACKKOWLEDGMENT

Figure 6.

Frequency-Distribution Plot for Calculated Specific Gravity

Measured specific gravity of this sample was 0.6820 * 0.UU005 (made by Carroll Creitz, 1). Specific gravitl calculated from chemical analyses ranged from 0.630 to 0.710, h u t 889%of all values reported lie between 0.657 and 0.685, with a mean of 0.672. (This mean is also value of greatest frequency derived from plotting these results to interval 0.002.) Specific gravity calculated from chemiral analyaes made a t National Rureau of Standards is 0.681 * 0.001, which is i n agreement with known value. Spectrometric analyses kielded calculated specific graviti-m from 0.673 to 0.701, b u t 3 high values can he eliminated to givr a mean of 0.681 * 0.004. Distribution is even, with no apparent maximum frequency. Mean i8 i n agreement with known value, and t h e reprodurihility is better t h a n achieved by chemical methods.

The author is very grateful to Churchill Eisenhart and Celia Martin of the bureau’s Statistical Engineering Section for correcting errors in the frequency-distribution plots, and to Rlarthada V. Kilday and Jean Doyle for assistance in computing specific gravities and heating values. He is also grateful to John W. Tukey of Princeton for permission to include his analysis of the

.. l l :.... ;1 :.. ’ ’ I ........... . .... ...... .................... ...... 0

e

I

A conservative statistical estimate of probable accuracy has been given by John W. Tukey of Princeton. His estimate of these data is worth considerable reflection, and is reproduced with his permission (Table 11). I n explanation he states: “The following table compares the author’s suggestions with those of the referee. The latter have been given in the two forms ‘best accuracy of statement justified’ and ‘smallest individual difference worthy of serious thought.’ I t is a matter of individual opinion whether these should all be doubled or all halved, but the relative relations seem fairly well settled by the frequency plots. .” The parenthetical values as noted by Tukey 111 the table are only the expression of a convention that has long been without factual basis. Their only meaning is that they mean nothing. I n comparing the other column of data there are two viewpoints to be considered. One is the viewpoilit of the analyst who would derive the best expression for probable accuracy by comparing results from the whole group of spectrometers; and Dr. Tukey’s “best accuracy justifiable” seems a proper anm-er in this case. The other viewpoint is the harder and more realistic one of the nian who must make actual use of the data from one spectrometer-he k n o m not which-selected a t random. He will probably think in terms of the author’s whole per cent, or Tukey’s even greater “magnitude of least difference between individual determinations worth serious thought.” The question thtlr1 remains-is this too realistic? The final choice between the chemical and spectrometric rnethods will depend upon the use to be made of the analytical data, and the amount of these data regularly required. The spectrometer can be justified if the analytical load is great enough, fur as an apparatus alone it produces so many analyses that a sizable crew of interpreters and computers may be required to keep up with it. If the actual need is for complete composition of a complex mixture like a natural gas, the mass spectrometer is capable of giving it and the rhemical methods are not. If the

..

.a.

I

1

I ~

I

I

I

I

I

I .

. .1

*.e

a**..

I. I

I

~

I I

I I 1 1 1

1 .

1

Jl*:i::l VILUESFF SCALE

* STU

Figure 7.

PER

CUBIC

FOOT

Frequency-Distribution Plot for Calculated Heating Value

Heating value calculated from chemical methods varied f n , r t , IUS7 t o 11.53 H.t.u. per cuhic foot, enough to initiate arguments hetween buyers and sellers. Mean, taken from 1086 t o 1113 (859’0 of al. values) is 1101 B.t.u. per cubic foot. Greatest frequency is noted a t 1104 B.t.u. tihen values are plotted t o 2 rather t h a n 1 l3.t.u. ‘\leasured heating; value is 1103 t 4 B.t.u. per cubic foot [average of two sets r i f measurements with Junker’s calorimeter a t National 8iireau of Standards. made hy J. H. Eiseman and R. S. Jessup. An imcount of other meaaurements is given i n (8). Value 1103 given in ( X ) would he changed to 1105 if made to conform to requirements of i . S . l . . \ f . tentative method of test for calorific value of gaseous fuels hg water-flow calorimeter; A.S.T.M. designation D9001-6T.1 Value calculated from XBS analyses previously discussed is 1109 i 1.5 B.t.u. per cuhic foot. Values calculated from analyses by mass 3pectrometer are higher. Spread is over 42 B.t.u. RPaximum frequency appears a t 1111. mean of all values is 1111 and mean of value8 from 1100 to 1121, inclusive (S5% of all determinations), is 1112 * 4.4. Thus general chemical value agrees with measured value on t h e low side, N B S chemical value agrees with measured value on the high side, NBS chemical value agrees with mean spectrometric value, hut spectrometric value requires all persuasiveness of generous and - deviations to hring i t into agreement with measurad \a1ne.

+

V O L U M E 19, NO. 9

640 frequency plots. The “SBS analyses” by chemical methods, cited in text and legends of plots, were performed by Shuford Schuhmann. LITERATURE CITED

( 1 ) Natl. Bur. Standards, Misc. P u b . M177 (1947). (2) Natl. Bur. Standards, Research Papers 625 (1933) ; 661 (1934) ; 715 (1934); 962 (1937); 1113 (1938); 1396 (1941). (3)Ibid., 1175 (1939).

(4) I b i d . , 1382 (1941). (5) Shepherd, Martin, ANAL. cHEbr.9 19;77 (1947) (6) Shepherd, Martin, Bur. Standards J . Research, 2, 1145 (1929); Research Paper 75. (7) Shepherd, Martin, J . Research N a t l . Bur. Standards, 36, 313 (1946) ; Research P a p e r 1704. ( 8 ) Shepherd, Martin, J. Research iyatl. Bur. Standards, 38, 19 (1947) ; Research Paper 1759. (9) Shepherd, Martin, J . Research N a t l . Bur. Standards, 38, 491 (1947); Research Paper 1789.

A New Approach to Analytical Chemistry As Taught at Louisiana State University A. R. CHOPPIN, ARTHUR L. LEROSEN, AND PHILIP W. WEST Coates Chemical Laboratories, Louisiana State University, Baton Rouge, La. The paper describes the efforts at Louisiana State University to supplement and extend classical methods and introduce students to the new physical approaches to the field of analytical chemistry.

I

N ORDER to solve any chemical research problem, suitable analytical methods must be available. The nature of these available methods determines the scope of chemical studies, while the development of new procedures and techniques precedes the development of a field wherein the methods can be applied; i t follows then that a thorough training in analytical chemistry is needed by every worker in branches of this science. I n addition, specialists in the field of analytical chemistry are needed to pioneer the discovery and development of new techniques. The success of newer instrumental and microchemical methods has brought about far-reaching changes in the field of analytical chemistry. New laboratories have been equipped with wide assortments of iastruments, many of which were research curiosities before the war. The idea that an analytical laboratpry would contain many instruments costing from $1000 to $20,000 each would have been considered fantastic ten years ago. However, since this is the situation now existing, the problem confronting industry is to find qualified analytical chemists capable of exploiting newly developed instruments and extending special techniques for the benefit of research and control groups. Quite properly the sources for such men are the colleges and universities. However, since a training in the “classical” methods of analysis would seem to fall short of present-day requirements, a new approach to this phase of the chemistry curriculum is indicated. The problem of academic training for analytical chemistry has been very well summarized by Hallett ( 1 ) . Training for instrumentation specialists has been discussed by Muller ( 2 ) who points out present deficiencies of colleges in this field. The present paper describes the efforts made a t Louisiana State University to supplement and extend classical methods and introduce students to the new physical approaches of the field. EDUCATIONAL PHILOSOPHY

Discussions of the academic training for analytical chemistry must take into consideration both undergraduate and graduate curricula. Certainly the undergraduate studies must give the student a thorough training in the fundamentals of chemistry, mathematics, and physics. Furthermore, undergraduate work should provide an education as well as a training. Any specialization, then, must come through graduate studies, since four years of undergraduate study provide scarcely enough time for the basic subjects. Graduate work, on the other hand, should

not be narrowed to such an extent that the field of specialty monopolizes the attention of the .student. Overemphasis of the specialty a t any time is a short-sighted policy which must ultimately result in stagnation. In this regard it is important to note that modern methods of analysis require a thorough training in physics and physical chemistry. The close relationship between inorganic and analytical chemistry seems generally to be appreciated, while a majority of chemists appear to have overlooked the fact that orgainic chemistry is equally important, since practically all new analytical reagents are organic compounds. Furthermore, well over 50% of all analyses now run are made on organic substances. The Undergraduate Level. Training in analytical chemistry rightfully begins with the course in “qualitative” analysis. Certainly this course is among the most important as far as its value in the fundamental training of the student is concerned. On the other hand, as generally taught it seems to be misnamed. More properly i t might be termed “systematic inorganic chemistry” or “analytical separations.” Its qualitative analysis aspects stem from antiquity and are so cumbersome that they find little practical application in a busy commercial laboratory. In this department, therefore, identification of chemicals as such is taught in the courses on microchemistry and spectroscopy, while the socalled “qualitative analysis” is.taught with emphasis on inorganic chemistry and analytical separations. Quantitative analysis is generally considered as the course in which manipulative skills are to be developed. While everyone seems to agree to this, some schools have tended to neglect the theoretical aspects of the subject. Quantitative analysis a t Louisiana State University is aimed to give the student both laboratory practice and a fundamental knowledge of the principles involved. The course is unique in that only one semester of the “classical” methods of analysis is taught a t the sophomore level, after which a semester of instrumental methods of analysis is taught a t the junior level. The basic course consists of a balanced selection of fundamental operations along with a thorough study of chemical calculations and pertinent chemical theory. Substitution of instrumental for traditional methods in the second half of the course permits the practice and acquiring of skills and a t the same time introduces the student to methods consistent with modern industrial practice. Instrumental methods of analysis deal with the application of physical methods to the analysis