INDUSTRIAL AND ENeINEERlNG CHEMISTRY PUBLISHED B Y

C H E M I C A L SOCIETY

WALTER J. M U R ? H Y , EDITOR

Laboratory Tests to Predict the Performance of Metals under Service Conditions

D. W. SAWYER AND R. B. MEARS, Chemical Mebllurgy Division, Aluminum Research Laboratories, New Kensington, PI.

To predict the performance of metals under service conditions, a laboratory test must give results that correlate directly with service results. The most certain method of designing a suitable taboratory tat i s to simulate the conditions of service as closely as possible. Attempts to accelerate the laboratory test by intensifying one of the facton encountered in service often load to misleading results. Several examples of laboratory tests illustrating these points are described.

HERE are several different reasons why engineers often T desire to predict, on the h i s of laboratory tests, the behavior of some specific metal or alloy under definite conditions of serv- ice (7). Such predictions are particularly desirable when equipment is required for some new chemical or process about which no service background has been built up. /Laboratory tests are also helpful in determining whether some untried metal or alloy offers promise of being more suitable than the material previously used. Laboratory tests are often useful in determin- ing the cause of attack which has been encountered in service and in developing methods to prevent or alleviate this attack.

These u~e9 of laboratory corrosion tests are fairly well known. They are concernec with the effect of the product under consid- eration en various metals or alloys under the conditions of serv- ice. However, often it is of importance to determine the effect of the metal or alloy on the properties of the product being proc- essed. &ch teats have not been described so frequently. In the present paper, tests of both types are described.

In the past, much emphasis has been placed on accelerated corrosion tests. All engineers would like to be able to evaluate the relative behavior of various metals or alloys in a matter of minutes instead of weeks or months. However, technical people are now beginning to realize that results from tests “accelerated” by altering certain factors encountered in service will generally lead to false conclusions. I t is becoming axiomatic that the more closely the laboratory test conditions approach the conditions of service, the more dependable the results will be. The teats de- scribed below illustrate the validity of this axiom.

GASOLINE STORAGE TESTS

Airplane gas tanks made of aluminum alloys have been widely used, and in general, have given very eatisfactory service. How- ever, in tanks of a particular design, corrosion was encountered after about one year’s service. In order to determine whether the newer fuels being used in these tanks were responsible for this attack, a series of tests was conducted (9) in which strips of several aluminum alloys were exposed in glasa bottles to the vari- ous fuels in both the presence and absence of liquid distilled water.

In no case were the specimens of any of the aluminum alloys appreciably attacked after one year’s exposure. Evidently the

fuels themGlves, even in the presence of water, were inert to aluminum. Some other factor or factors must have caused the attack under service conditions.

Open boxes were constructed of aluminum alloys. These boxes included torch-welded, spob welded, and riveted joints to simulate the construction of actual gasoline tanks and were equipped with cast fittings like those used in the actual tanks.

Since the previous test had indicated that the various fuels, even in the presence of distilled water, had no action it was de- cided to use a typical leaded aircraft fuel plus a 3.5% sodium chloride solution in distilled water. The gasoline layer waa changed every week and the sodium chloride solution was changed every month.

After one year it was apparent that corrosion had develo ed, but the type of attack obtained was not similar to that wkch had occurred in service. Therefore, the test was not considered to be dependable.

I n the meantime, with the cooperation of the aircraft operators, it was found that an aqueous phase collected in the bottom of the particular tanks in servioe and that these tanks were designed so that it waa impoesible to remove this entrapped liquid com- pletely. An attempt wag made to collect samples of thiB watery layer.

Draining8 from individual tanks after each flight gave from a few dro s to a tables onful of this liquid. These drainings were accumurated until a R u t 5 gallons were obtained. A new &Et was started using small enclosed aluminum alloy boxes provided with all the features of the service tanks. The boxes wem equipped with cast fittings and breather vents just as are the large tanks. A small portion of the tank drainings was laced in each small tank and then these tanks mere partially B e d with leaded aircraft fuel.

The tanks were installed in a delivery truck so that they could be agitated in a manner somewhat similar to that under service conditions. In addition, the gasoline layer was changed every week and the aqueous layer every month.

When all these precautions were followed, test remlts simulat- ing those of service were obtained. Once a dependable test waa available, it was possible to evaluate various alloys, types of con- struction, and other factors and thus to find a solution to the problem.

This is probably an extreme case, but it will illustrate the necessity for duplicating the conditions of service if a dependable laboratory test is to be obtained.

Therefore, a new test was run.

MUTUAL EFFECT OF GASOLINE AND METALS

In selecting a material for the construction of storage tanks or shipping containers for fuels, it is important to select one which has no deleterious effect on the fuel. There are several published references to the effect of metals on the rate of gum formation in fuels contacting them, but most of them tests were conducted a t elevated temperatures (11, 16). It is by no means certain that

1

2 ' " 3 U S T R I A L A N D E N G I N E E R I N G C H E M I S T R Y Vol. 17, No. 1 - Ti lines

ithr

Alun

3.110 ""D

GaS0li"e Initial Blanks' 2S-I/.Hb Steel Copper steel i D . / l O O

mi. Mp. pm 100 m1. Straight run 0 . 2 Cracked 3 . 2 Blended 2 . 3 Sfriiighf run. leaded 0.6 Cracked leaded 2 . 5 Blended: leaded 1.8

7 . 6

0.8 5 . 6 3 . 2

0 . 4 7 .2 5 . 2 0 . 6 4 . 6 3 . 2

0 . 2 0 . 2 0 . 2 6 . 2 1525.8 9 . 2 2 . 2 3 1 . 2 1 . 4 1.0 2 . 0 ... 5 . 2 1724.0 6 . 4 2 . 4 144 .0 3 . 4

* Blanks stored in contact with glass. * Commereislly pYre aluminum.

Table II. Induction Period of Garoliner before and after Storage for Five Months

F i n d A1"mi""m

S l l O Y Stainleas Gasoline Initial Blanks" 2S-'/,€Ib Steel Copper steel

Iloliis Holrrs Hours Hours Hour Hours Cracked g.0 2 .75 3.0 3 . 5 0 .25 3 . 2 5 Blended 4 . 0 4 . 2 5 4 . 5 0 . 5 4 . 5 Cracked.1eaded 3.6 4:O 4 . 0 4 . 0 0 . 5 3 . 5 Blended,leaded 4 .75 5 .25 5 . 5 6 . 5 0 . 5 5 .25

Blanks stored in contact with g1a.s. I , Commerclrlly pure alumiaum.

the relative behavior of the different metals would he the same a t high temperatures as a t room temperatures, yet storage oontainers are used a t room temperatures. For this reason, a lest WBS run under conditions simulating those of storage.

Some of the results are given in Tables I, 11, and 111. These results indicate that cracked or blended fuels are definitely af- fected by contact with copper. It can he Seen in Table I that the effect of copper on gum formation is most pronounced. The ef- fect of copper on the oxidation induction period is also definite, sa shown in Table 11. Although the effect of copper on 'the oc- tane numhers of the fuels was appreciable (Table III), the mag- nitude WBS le= pronounced than in the ease of the gum formation or oxidation induction periods of the fuels. No definite changes in the other measured properties were detected, so these are not reported here.

The gasoline stored in contact with the other metals WBS not affected, as judged by these property measurements. However, the gasoline stared in contact with low-carbon steel became 6lled with finely divided rust particles.

Figure 1 shows the appearance of the steel specimen a t the conclusion of the test. The heavy rust layer which formed he- tween the coil turns is e l w l y shown.

Figure 2 illustrates the aluminum and copper specimens at the

he discoloration of the gasoline in con- :aused by heavy gum formation.

aluminum specimens showed no evi- ed no more alteration in properties of with glass alone,

MUTUAL EFFECT OF OILS AND METALS

A somewhat similar study is being made of the mutual effect of lubricating oils and metals, employing B method of testing similar to the proposed A.S.T.M. method for studying the oxidation characteristies of steam-turbine oils (5).

Instead of the wire eoii used in the A.S.T.M. test, sheet specimens (10 X 15 em. in size) of several different metals bent in the farm of square tubes have been substituted. The surface area of these tubes was similar to that of the wire samples or- dinarily employed-that is, 1 sq. em. of metal per ml. of oil. Sheet specimens were used instead of wires, since wires of small diameter are affected to B much greater extent in certain environ- ments than are flat surfaces (le). Moisture -'as supplied to the oils, since in service moisture will generally be present. In tests being run st 90" C., 20 ml. of distilled water were added to each s a p : ? of 100 ml. of oil. In tests run a t 120' C . the oxygen, which was bubbled through the oil samples, was saturated with distilled water.

The tests were run in glass tubes 600 nun. long and 45 mm. in diameter, each equipped with an individual condenser. The volume of ail used in each tube was 300 ml. Oxygen was bubbled

Figure 1. Steel Specimen after Five Months' Exposure to Gasoline and

Distilled Water N o k hewv tun iwer which h s d brlnen the

.Oil t Y r n l

~

Tablolll. Octane Numbomof Garoliner beforeand after Storage for Five Months with Metals

Final Alum$- ""nl stam- alloy less

Gaaaline Initla1 Blanks- 2S-V.Hb Steel Copper ateel Straight r u 57.0 60.6 51 .1 50 .2 61 .1 Craoked 76.0 7312 72 .2 73.1 67 .6 7 3 . 0 Blended 68.3 63.8 6 4 . 3 63 .2 64 .8 Straightrun.1eade.d 7 5 . 3 73:4 72 .9 72 .7 7 1 . 9 C,+e$.+$ed 85.; 83.3 83.3. 83.1. 15.i !a:$ I)le"(lea.LeS(le(l 01.0 SU.3 I J . , OL.' 1 1 . 1 0 Y . O

a Blanks stored in oonlact with glass. b Commercially pure aluminum.

A N A L Y T I C A L E D I T I O N 3

slight and rhe oil exposed in contact with aluminum alloy 5 2 s l/&l was aEected no more than the oil exposed in the absence of any metal.

The tests being run at 90' C. are giving qualitatively similar results. The oil in contact with copper was definitely affected after 8 days, and marked sludging occurred (Figure 3). No sludging developed in the case of oil exposed to the other metal specimens. After degreasing, the appearance of some of these other specimens at the conclusion of the test is shown in Figure 4. The steel was definitely rusted and the tinplate was attacked near one edge. The zinc showed a few shallow corroded grooves and the other two specimens (aluminum alloy 14ST and stainless steel) showed only a mild surface staining.

The interpretation of these test results in terms of service iS not known. However, qualitatively similar results were ohtnined by Hunter and eo-workers (10) under conditions more nearly simulating those of service. It would he expected that where oils are used a t elevated temperature, in the presence of air and moisture, these test results would correlate qualitativdy with service performance.

CATHODIC PROTECTION TESTS

January, 1945

Fisure 2. SDocinrns at Conclusion of Test

Table IV. Oxidation of a Solvent-Refined Pennsylvania Lubri- cating Oil at 120" C. in Presence of Moistare and Various Metals

Hours Neutralization Metal Oxidized Number

Unoddised oil NO metal Aluminum alloy 14s-TO Aluminum alloy 2S-Ob Aluminum alloy 52S-1/zHc Copper Zi,X Low-carbon steel stain1es. steei Tinolate

1544 1344 1344 1344 117

1344 412

1344 1344

0.05 0.78 1.35 1.08 0.80 2.70 1.05 2.13 1.23 1.48

a Aluminum slloy. nomind oompoaition: 4.4% Cu. 0.8% Si, 0.8% Mn.

b Commercisily pwe al?mioum. 0.4% Mg.

Aluminum aiioy. nommal cbmpoaition: 2.5% Mg. 0.26% Cr.

the conclusion of the test, other properties of the oil samples. such as interfacial tension (du Noiiy method), steam emulsion number (e), A.S.T.M. color number. (l), ,and viscosity (S) were also determined and the ehanees in weiaht or ameaxance

~ - .. of the metal samples were noted.

At the present time, one series of tests bas been completed, in which a solvent-refined Pennsylvania lubricating oil was used and the tests were run at 120' C. (Table IV).

Figure 3. Sludge

Another series of tests at 90' C. has hien started.

In the test run a t 120' C., it was found that copper definitely increased the neutralization number of the oil after 117 hours' exposure, so this sample was removed from test (see Table IV). The oil in contact with the low-carbon steel sample r$n for 472 hours, after which time ita nentralin;ation number was markedly increased; it was therefore removed from the test. The other samples were continued in test for 1344 hours. A t the end of this extended period of exposure, i t was found that the neutralization numbers of the oils had been aEected somewhat more in the casea of the samples exposed in con- tact with tinplate, aluminum alloy 1&T, or stain- less steel than for the remaining samples. The &ect of zinc and aluminum alloy 2SO was very

The use of solution potential measure- menta to predict whether cathodic protec- tion (8, 14) can he successfully applied furnishes an example of another. type of laboratory test. If two dissimilar metals are coupled together and immersed in an electrolyte, in general, au electric current will flow between them. The metal from which positive current leaves to enter the electrolyte is the anode and the other metal is the cathode. Normally attack of the anodic metal is stimulated by such a contsct while attack of the cathodic metal is reduced. This reduction in attack of the cathodic metal caused by current flowing to i t from the solution is termed cathodic protection. The mechanism of this protection has been discussed in pre- vious papers (8,13).

In the classical electromotive series, aluminum is listed as having a more anodic solution potential than zinc. However, in many natural waters, including sea water, zinc is auodic to aluminum. There- fore, in such waters zinc attachments CBO

be used to protect aluminum chemical equipment csthodically. Furthermore, it should be possible to determine under what conditions cathodic Drotection of

Figure 4. Effect of a Solvent-Refined Pennsylvania Lvbricatin Oil upon Metal Specimens Oxidized at 90' C. in Presence of w9 ator

Tind.1. 21°C COPWI AI~mImum alloy 14S-T SIBinirr *el LOW-rubo" rt..l

4 I N D U S T R I A L A N D E N G I N E E R I N G C H E M I S T R Y Vol. 17, No. 1

in predicting the feasibility of cathodic protection under service

In addition to proper potential relationships, suitable spacing Jf zinc attachments is required in order for protection to be 100%

Table V. Potential Difference betweed Zinc and Aluminum in conditions. Various Solutions

pH Temp. Potential Difference a " T i " , ,

Composition of Solution P . r " I ,

effective. Location and distribution of the zinc attachments a& much more difficult to determine by Iaboratory tests. At present these points are generally determined by engineering judgment based on the analysis of the water, the potentials of zinc and aluminum, anll the construction of the part to be protected.

SUMMARY

To predict the performance of metals under service conditions, laboratory test must give results that correlate directly with em- ice results, and conditions of service must be simulated as closely as possible. Attempts to accelerate the laboratory teat by in- tensifying one of the factors encountered in service often lead to misleading results.

The development of a test of aircraft fuel tanks is described. Results'obtained

0 . 2 2 (Zinc anodic) 0 . 1 5 (Zinc anodic) 0 . 0 5 Zinc anodic) 0 . 0 2 [Zinc anodic)

0 . 2 2 (Zinc anodic) 0 . 1 5 (Zinc anodic) 0 . 0 5 Zinc anodic) 0 02 {Zinc anodic) 0 . 1 6 (Aluminum anodic) 0 . 3 7 (Alumjnum anodic) 0 . 5 8 (Aluminum anodic)

6 . 5 77 7 . 5 77 8 . 5 77 9 . 0 77 9 . 8 77

10 .7 77 11 .2 77

0 . 1 6 (Aluminum anodic) 0 . 3 7 (Aluminum anodic) 0 . 5 8 (Aluminum anodic) 0 .30 (Zinc anodic) 0 . 3 1 (Zinc anodic) 0 . 2 8 (Zinc anodic)

3 . 7 77 2 . 6 77 1 . 6 77 6 . 0 77

10 .8 77 3 . 4 77

0 . 3 2 (Zinc anodic) 0 . 2 3 (Aluminum anodic) 0 .19 (Zinc anodic) 0 . 1 5 (Zinc anodic) 0 . 2 2 (Aluminum anodic) 0 . 0 5 (Zinc anodic)

5 . 6 77 1 0 . 7 77 2 . 8 71 6 . 4 122 6 . 9 158 7 . 3 186

0 .18 (Zinc anodjc) 0.16 Zinc anodic) 0.16 !zinc anodic)

77 0 . 1 (Zinc anodic) from simple tests which neglected various service factora were inconclusive; it was necessary toduplicate the conditionsot serv- ice in order to obtain dependable results.

Source of Water pH CI 80, aolid. Temp. Potential Difference In two testa the effect of the metal on P.p.m. P.p.m. P.p.m. F. Poll ita environment was of more importance

7 . 9 8 91 169 122 Zn anodic to AI 0 . 2 3 8 . 1 8 91 169 168 Zn anodic to AI 0 . 2 6 itself. Copper was found to accelerate

Memphis. Tenn.. well water 8 . 2 7 . 2 . . 8 91 . . leg 'E E: ::$!: i t ::g the decomposition of cracked fuels a t room 7 . 4 . . : ; ::; E; t;$i; t: temperature and the decomposition of a 7 . 8 . . .. 8 . 6 .. .. . . . 186 Al anodic to Zn 0.01 lubricating oil a t higher temperatures. 8 . 6 14 BO 318 122 Zn anodic to AI 0.40 Aluminum d o 5 and stainless steel were 8 . 2 14 90 318 158 A1 anodic to Zn 0 . 0 8 genedly inert in these tests. Other 8 . 7 14 90 318 186 A l a n o d i c t o Z n 0 . 1

Kewauner, Wia.. tap water 8.0 84 261 667 77 Zn anodic to AI 0.aO metalaandalloyshadadefhiteeffectunder 8 . 1 34 261 687 122 Znanodicto AI 0 . 4 1 7 . 8 34 261 687 I58 AI anodic to Zn 0.08 s ~ c s c conditions.

Altoona, Pa., synthetic tap water 4 . 8 5 61 98 77 Zn anod/c to AI 0.60 Measurement of solution potentiala of 4 . 4 5 61 98 122 Zn anodic to AI 0 .40 4 . 1 5 61 @8 158 Zn anodic to AI 0. M) zinc and aluminum has been found to be a

useful .laboratory test for predicting the 4 . 3 6 61 98 185 Zn anodic to A1 0.46

quwcent conditlons) 7 . 3 9,000 1,200 18.630 77 Zn anodic to A1 0.29 f&bility of using athodic protection for bubbled through nolution) 7 . 3 9,000 1,200 18.5sO 77 Zn anodic to AI ' 0.26 duminum chemical equipment. A corre-

lation between service reeulta and poten- tial measurements ia given.

Table VI. Potential Difference between Zinc and Aluminum in Industrial Wahn Composition of Water

Total

New Kensington. Pa., tap water 7 . 8 8 91 169 77 Zn anodic to AI 0 .27 t h , ~ the c o d o n mktance of the

Xlanitowac, Wis.. tap water 8 . 4 14 90 318 77 Zn anodie to AI 0 . 5 3

Gowanus Canal, Brooklyn (under

Gowanus Canal, Brooklyn (air

this kind is feasible, by measuring the solution potentials of zinc ACKNOWLEDQMENTS and aluminum under conditions similar to those of service.

In Tables V and VI the potentials of zinc and aluminum in a wide variety of natural waters and dilute salt solutions are given. In most cases, zinc is anodic to aluminum in nearly neutral or in definitely acid solutions, but reverses in potential in alkaline eolutions. Usually the potential difference is less favorable for cathodic Drotection in hot solution than

The fuels wed in the studiea of the mutual effect of gasoline and met& were mpplied through the courtesy of D. R. Stevens of Mellon Institute, who also grrve advice and aasistance in con- ducting the teats.

E. M. Kipp of the Aluminum W r c h Lsboratoriea cooper- ated in planning and conducting these tests.

in cooler solutions. For the past 5 years, wherever the a p

plication of cathodic protection to alum- Table VII. Examples d Cathodic Protection of Aluminum by Zinc in Industry hum equipment was considered, the solu- tion potentiala of zinc and aluminum have been measured in the particular solution or water in question at the service. tem- perature and the decision aa to whether cathodic protection by zinc attachments was feasible. haa been based on these meeeuremente. The resulta of measurements of solution

potentials and of behavior under d c e conditions are given in Table VII. I t will be noted that in all casea where zinc was found to be definitely anodic to du- minwn, the use of zinc attachments under service conditions haa been beneficial. Thus, the laboratory measurement of solu- tion potentiala has proved highly effective

Compoaition of Coolins Potential Water Difference

P a m . P.p.m. P.p.m. O F. Vol: Yaarr

Service Application pH C1 804 solid. Total Tpmp. Zn to Anodid AI Period Rwulta Protection

Shell and f.ube condenrer, r e c l a i m e d r a t e r throqh tuba

Ammonra de hlegmator, spray pontwater

Ammonia de hlesmator, spray ontaater

Heat excIanger AI t u b a cast iron shell and headers for coolinc

4 . 1 14 5 63 120 0.6

0 .39

0 .28

0

2 Complete

Part i i

Partial

Partial

6 . 7 208

6 . 7 20%

8 . 2 19

1,169

1,159

57

2,087

2.067

347

77

160

60

2

2

... gasen from Solvay proccu with water

Drinking water cooler, aluminum tobw

Jacketed tank for recrya- tallisms tartaric acid witb water aa coolant

Condenser containing aluminum tuba, steel .bell. and well water

7 . 8 8

7 . 3 9,000

7 . 2 ..

91

1 3 0 0

169

18,630

40

77

77

0.25

0 .29

0 .44

3 . 5

2

7 moa.

Complete

Complete

Complete ...

January, 1945 A N A L Y T I C A L E D I T I O N 5

LITERATURE CITED (10) Hunter, B. F., Ambrose. H. .I.. and Powers. K. P.. Poiuer. 83. Am. SOC. Testing Materials, Method D155-39T. Ib id . , Method D157-36. Ibid., Method D445-42T. Ibid., Method D663-42T. A m . SOC. Testing Materials Proc., 43, 275 (1943). Am. SOC. Testing Materials, "Standards on Petroleum Prod-

ucts", 1940. Borgmann, C. W., and Mears, R. B., "Principles of Corrosion

Testing", A.S.T.M. Symposium on Corrosion Testing Pro- cedure, 1937.

Brown, R. H., and Mears, R. B., Trans. Electrochem. SOC., 81, 465-83 (1942).

Dix, E. H., Jr., and Mears, R. B. ,S .A.E.Jourd,46 ,215 (1940).

97-9 (1939). (11) Mardles, E. W. J., Proc. World Petroleum Congress. 2, 59 (1933). (12) Mears, R. B., Belt Lab. Record, 11, 141 (1933). (13) Mears. R. B., and Brown, R. H., Trans. Electrochem. Soc., 74,

(14) Mears, R. B. , and Fahrney, H. J.. Trans. A m . Inst. C h m . Engra.,

(15) Story, L. G. , Provine, R. W., and Bennett, H. T., IND. ENG.

519 (1938).

37, 911 (1941).

CHEM., 21, 1079 (1929).

PREBEATED before the Division of Petroleum Chemistry, Sympoaium on Bench Scale Techniques, a t the 108th Meeting of the A U ~ R I C A N CEI~MICAL SOCIETY. New York, N. Y.

Ternary Mixtures of' A Quantitative

Three Isomeric Heptanes Method of Analysis

VERLE A. MILLER, Research Laboratories Division, General Motors Corporation, Detroit, Mich.

This paper describes a method based on a refinement of the solution temperpture of the hydrocarbon mixture in diethyl phthalate and nitrobenzene, b y which the composition of a mixture containing P,4- and P,%dimethylpentane with 4,4,3-trimethyIbutane may be determined quantitatively, the first two components within 396 and the last component within 0.3%. The entire analysis requires approximately one hour.

RrlCTIONAI, distillation, supplemented with curves for F other pertinent physical data, is entirely suitable for the analysis of paraffin mixtures up to and including the hexanes, but these methods alone are not adequate for the analysis of certain mixtures of the heptanes (see Table I).

Attempts have been made by a great many workers, in several fields of research, to develop adequate methods of hydrocarbon analysis. These include ultraviolet and infrared abeorption spectra (3, 19, 2.2, 24, 15) and Raman spectra (9, 14, 15, $8, sa), and more recently the mass spectrograph has been applied to the solution of i;hi3 problem. Rosenbaum, Grosse, and Jacobson (26) in their work on the Raman spectra of the nine isomeric heptanes, report that analysis of close to 50-50 binary mixtures gave results within 5%. However, an attempted analysis of t b ternary mixture of 2,4- and 2,2-dimethylpent,ane with 2,2,3- triwt:thyibutne (the heptane mixture which boils a t about 80" C,;,? even in approximately equal proportions, gave results whirh varied as much as 120/,.

This paper describes a method, based on a refinement of the solutios temperature of the hydrocarbon mixture in diethyl phthalate and nitrobenzene, by which the composition of a mixture containing 2,4- and 2,2-dimethylpentane with 2,2,3- trimethylbutane may be determined quantitatively, the first two components within 3% and the last component within 0.37,. IC order to analyze such a ternary mixture quantitatively, it is only necessary to determine the solution temperature of the unknown sample in diethyl phthalate and nitrobenzene with en accuracy of ==O.0lo C. These temperatures are then used with a series of calibration curves, which were prepared with mixtures of known composition, to determine graphically the percentage of each of the three constituents present in the unknown mixture. The entire analysis requires approximately one hour.

HETORY OF APPLICATION OF CRITICAL SOLUTION TEMPERATURE MEASUREMENTS TO HYDROCARBON ANALYSlS

Chavanne and Simon (6) first observed that the presence of aromatic hydrocarbons decreased the critical solution tempera-

ture of a hydrocarbon mixture in aniline and that the lowering was directly proportional to the weight of aromatics present. Tizard and Marshall (31) introduced the more simple determi- nation of "aniline points", which numerous workers (2, 4, 13, 17, 18, 21, 23) have shown to be of the greatest utility for the determination of the quantity of the various classes of compounds present in gasoline and kerosene distillates. I t has been common practice to employ pure, dry aniline that will give an aniline point of 70" * 0.1" C. for n-heptane and a reproducibility of about 0.1" C. is usually claimed. As compared with freshly distilled aniline, water-saturated aniline may cause a rise of as much as 20" C. in the observed aniline point. Tilitsheyew and Dumskaya (29) studied the method for mixtures of pure aromatic hydro- carbons and mixtures of pure aromatic hydrocarbons with light petroleum distillates. .4ubree (1) used a second solvent, benzyl alcohol, and thus obtained two equations with two un- knowns which he could solve for the aromatic content without removal of these hydrocarbons by chemical means. Erskine (12) first used nitrobenzene instead of aniline for the determi- nation of aromatics and'reported that this critical solution tem- perature decreased with rise in molecular weight instead ot increasing as with aniline.

Considerable work on the critical solution temperature of various pure hydrocarbons in one or more of the three solvents, aniline, benzyl alcohol, or nitrobenzene has been done by Chavanne and Simon (6 ) , Garner (13), Maman (,El), Edgar and Calingaert (fO), and Wibaut, Hoog, and Smittenberg (33). As far as the author has been able to determine there is no mention in the literature of any attempt to use this valuable and easily determined constant for the quantitative determination of individual paraffin compounds which are present in a mixture,

SEARCH FOR OTHER CRITICAL SOLUTION TEMPERATURE SOLVENTS

A preliminary study showed that the spread between the aniline points of pure 2,4- and 2;2-dimethylpentane is only 0.35' C. This agrees well with the value (0.4') obtained by Wibaut, Hoog, and Smittenberg (33) but not with that (1.1") obtained by Edgar and Calingaert (10) (see Table I). In all, 97 compounds were investigated in an attempt to find

solvents which would give a wider spread of solution tempere- tures for these two hydrocarbons. The criteria for such a solvent are: It must be obtainable in a pure state; it should be fairly stable; it must give a solution temperature within a reasonable working range; and it mi=- give a satisfactory end point with B