Characterization Methods

The commonly used methods of characterization that a p pear to merit investigation are (a) the critical solution tem- perature with aniline-i. e., aniline point, (a) the viscosity index, ( c ) the viscosity-gravity constant, (d ) the Universal Oil Products (U. 0. P.) boiling point and gravity characteri- zation factor, and ( e ) hydrogen content. All of these methods have been more or less generally used, and data are available in the literature correlated on the basis of each.

The U. 0. P. characterization factor (10) is defined as the ratio of the cube root of the average boiling point in degrees Rankine to the specific gravity a t 60” F. Its values range from 12.5-13.0 for purely paraffinic stocks to 10 or less for aromatic materials. This ratio was derived as the empirical equation of a curve relating average boiling point to specific gravity for a series of cuts taken from a petroleum considered to be of fairly uniform character throughout its boiling range. Such an analysis was made on several types of stock, and the simple cubic equation was adopted as sufficiently accurate.

Values of the characterization factor are somewhat erratic for individual pure compounds and isomers of the same hydro- carbon group. However, when applied to the complex mix- tures of petroleum, these irregularities are to some extent averaged out, and the factor has proved extremely useful in dealing with fractions of both natural and cracked stocks.

It is an advantage of the characterization factor that i t is applicable to the entire range of fractions from light gasolines to residues. However, determination of correct boiling points of heavy stocks is difficult, requiring some type of high-vacuum distillation such as that described by Watson and Wirth (11)

The viscosity-gravity constant proposed by Hill and Coates (6) has the disadvantage of being defined in terms of Saybolt Universal viscosities, which limits its application to a relatively narrow range of stocks of lubricating oil viscosities. Although it is very useful when applied to relatively paraffinic lubricating oil stocks as proposed by the authors, its values lose significance when applied to highly aromatic stocks such as are formed in the cracking process. Because of this re- stricted applicability both as regards parafhicity and vis. cosity, i t cannot serve as a general method of characterization applicable to all stocks. However, the principle of characteri- zation by a function of gravity and viscosity appears sound.

Because of the usefulness and simplicity of the U.O.P. characterization factor, i t was considered desirable to retain the numerical values of this scale and develop empirical charts expressing the characterization factor as a function of vis- cosity and gravity, viscosity index, aniline point, and hydro- gen content as well as boiling point and gravity. By express- ing viscosities in a fundamental unit such as the centistoke, the viscosity-gravity method becomes applicable to all stocks. If desired, a mathematical viscosity-gravity constant similar to that of Hill and Coates could be developed in these terms. However, i t seems better to avoid confusion merely by ex- pressing the characterization factor empirically as a function of viscosity and gravity rather than to introduce a new scale.

Average Boiling Point In dealing with petroleum fractions of wide boiling ranges,

it is difficult to develop an entirely satisfactory method for obtaining the proper average boiling point from the data ordinarily available.

The ideal method for calculating the average boiling point should be such that, when other average properties are con- stant, the average boiling point is independent of width of boiling range. The averaging of boiling point data on a weight or volume basis does not satisfactorily meet this requirement. As width of boiling range is increased, the average boiling

Characterization of Petroleum

Fractions

K. M. WATSON, E. F. NELSON, AND GEORGE B. MURPHY

Universal Oil Products Company, Riverside, Ill.

Empirically developed charts are pre- sented interrelating the Universal Oil Prod- ucts characterization factor with specific gravities, boiling points, viscosities, aniline points, viscosity index, and hydrogen con- tents of petroleum fractions. If for any particular stock two of these properties are known, the others may be more or less satisfactorily approximated. The most satisfactory correlations are based on deter- mination of specific gravity and either aver- age boiling point or viscosity at 210’ F. These determinations permit prediction of the other properties with fair approxima- tion for use in engineering problems.

N ATTEMPTING any general correlation of the properties of a petroleum fraction, it I is necessary that some method be available

for quantitatively expressing the general type of hydrocarbons of which it is composed. This property is sometimes termed the “paraf i ic i ty” of the fraction, a paraffinic composition representing one extreme of the classification and aromatic the other.

Quantitative characterization or specification of paraffi- icity is particularly useful in developing general methods for predicting difficultly measurable physical properties of petro- leum and for forecasting behavior in decomposition both as regards yield of products and rates of reaction. For this reason the following study was undertaken to compare the more satisfactory methods of characterization in common use and to develop quantitative relationships between them. These results have permitted more general correlation of prop- erties than was previously possible and also allow transla- tion of experimental results from one system of expression to another.

1460

7 8 /l.&CL'A14 W>/&TS,' cR:T/C)S' kEhPELATUM3 i AND

point calculated on either a weight or volume basis increases even though the specific gravity, average molecular weight, and chemical type are constant. The average on a volume basis shows much less variation than on a weight basis and is preferable for that reason. The 50 per cent temperature in the Engler distillation is frequently a close approximation to the average boiling point on a volume basis if the distillation curve is symmetrical.

In an attempt to improve correlations for wide-boiling mixtures, Watson and Nelson (IO) suggested a molal average boiling point, weighting the distillation temperatures on a molal basis. This average would be obtained by averaging the ordinates under a curve relating moles of distillate to tempera- ture. It was found that this molal average boiling point is less affected by width of boiling range than a volumetrically weighted average. A convenient curve was derived for esti- mating the molal average boiling point from Engler distilla- tion data.

Experience with very wide boiling-range fractions has shown that, although the curve proposed by Watson and Nelson gives a fair approximation to correct molal average boiling points, the corrections indicated on very wide boil- ing mixtures are somewhat too great for the ideal average described above. Apparently this ideal average is intermedi- ate between the volumetric and molal values. The molal average boiling point is satisfactory for Engler slopes of 3 or less. Where the Engler curve slope is 8.0, better correlations are obtained with a correction of approximately 60" F. rather than the 90" F. indicated for the molal average boiling point from the curve of Watson and Eelson. Intermediate cor- rections are proportionately reduced.

It is important that no decomposition shall take place in the distillation from which the average boiling point is calculated by the above method. For higher boiling fractions a vacuum distillation converted to an atmospheric pressure basis is re- quired. If decomposition is suspected in the last part of the distillation, the best approximation is obtained by taking the 50 per cent temperature as the volumetric average boiling point and basing the correction on the slope from 10 to 50 per cent.

In all the following correlations, "average boiling point" refers to the average obtained in the above manner. For close- cut fractions with slopes of 1.5 or less, this correction may

CHARACTER/Z~T/ON FA croRs

1461

1462 INDUSTRIAL AND ENGINEERING CHE-MISTRY VOL. 27, NO. 12

Stock

Midcontinent dist.

Midcontinent reduced crude

Smackover dist.

Dist. from residue in cracking Midcontinent gas oil

Pennsylvania dist.

Pennsylvania lube oil

Cracked residuum

TABLE I. PROPERTIES OF PETROLEUM FRACTIONS Slope -Characterisation Factors from:-

of Viscos- Viscos- Visoos-

Grav- ing from Aniline -Viscosity- - coaity gray- gray- grav- gray- cosity Aniline i ty Point 10-90% Point 100' F. 122' F. 210' F. Index ity i ty i ty ity index point

Av. Engler Boiling i ty at i ty a t i ty a t Boil- Curve, Vis- point- 210° F.- 122'F.- 100'F.- Via-

' A . P . I . O F . %/" F . C. -Centislokes-

47.2 353 0.77 58.3 1,099 . . . . 0.608 . . . 11.8 11.8 . . . 11.7 . . . 11.8 40.4 443 0.56 66.7 1.95 . , . . 0.88 11.8 11.75 11.75 . I , 11.8

30.4 647 0.79 86.3 10.45 t ,050 2.55 . . . 11.8 11.9 11.85 11.8 11.9 28.3 719 0.79 94.7 17.9a 12.10 3.69 90 11.9 11.9 11.9 11.85 12:i 11.95 26.7 783 0.71 98.6 44 27.2a 6.04 85.1 12.0 12.0 12.0 12.0 12 1 12.0 24.9 852 0.33 104 2 115a 60a 10.57 79.5 12.2 12.05 12.05 12 05 12 0 12.1

. . . 35.3 541 0.61 76.6 4.03 9.1" 1.43 . . . 11.8 11.85 ii:& 11.85 . . . 11.88

23.3 862 3.1 . . . 400

33.1 450 0.68 53.4 1.968 27.8 601 0.70 65.5 5.66 22.1 705 0.54 75.3 20 4 21.0 758 0.41 82.2 42.7

17.1 584 0.73 34.4 5.46 12.6 634 0.64 38.6 8.25 7.1 679 0.50 42.0 23.4 4.8 763 0.83 43.8 153.2 5.4 730 1.26 , . . 72.5 21.4 455 1.2 . . , . . . . 14.2 530 0.71 . . . 3.46 8.7 600 0.65 . . , 8 6 3.2 670 0.60 . , . 28.0 2.1 785 1.25 , . , 315

43.4 472 0.50 81.4 2.24 35.4 727 0.33 127.0 11.63 31.9 809 0.74 . . . 37a

30.3 774 1.04 104.2 40.7 29.3 873 3.6 118 28.0 880 4.0 l22:2 179.1

205 25 5 90 12.0 12 2 12.1 12 05 12 I

. . . . 0.885 . . . 11.3 11.3 . . . 11.3 . . 11.35

. . . 1.67 . . 11.5 11.5 11.5 11.4 12.ga 3.47 15 11.5 11.45 11'4 11.4 11:3 11.4 245 5 29 27 11.5 11.5 11 . 5 11.5 11.4 11.52

. . . . 1.39 . . . 10.7 10.55 . . . 10.6 . . . 10.4 . . . . 1.99 10.5 10.45 10.4 10.5 13.50 3.18 -160' 10.2 10.25 10:25 10.2 10:3,5 10.45

. . . . 7.49 -166 1 0 . 8 10.3 . , . 10.35 10.3 10.3 62.5a 5.7 -78.5 10.2 10.3 10.3 10.3 10.7 . . ,

. . . . 0.85 . . . 10.5 10.4 . . . I . . . . .

. . . . 1.25 . . . 10.25 10.25 l0:25 . . , . . . 6.0a 2.01 10.1 10.15 io:i5 10.15 . . . . . . 15a 3.05 -3iO' 9.9 9.9 9.95 9.95 9.9 . . _

102a 9.8 -212 10.2 10.2 10.15 10.2 10.15 . . , . . . . 0.985 . . . 12.1 12.1 . . . 12.1 . . . 12.15 . . . . 2.83 . . . 12.5 12.4 . . . 12 35 . . . 12.8 23.8O 6.04 118 12.5 12.45 12.5 12.4 12.6 . , , 25" 5.98 91.5 12.3 12.3 12.35 12.3 12.2 12.1 66.0 12.0 110.0 12.45 12.5 12.5 12.5 12.5 93.OU 16.15 104 12.5 12.5 12.5 12.5 12.4 12.4

10.3 773 . . . . . 480a 210 . . . . . . 10.7 . . . 10.9 10.9 . , . 820a 340 . . . 10.6 . . . 10.8 10.8 8.5 774 2.6 912 4:5 . . . . . . . 320005 290" . . , 10.5 , , , , . . , . .

. . . 480a 17.05 . . . . . . 9.8 9.8 -3.9 . . , . . . . . . . . . .

. . . . . . -4.3 760 516 . . . 1750" 550a 29.0 -110 9.5 9.85 9.8 10'5

0 Estimated by plotting known da ta on A . S. T. M. Viscosity Chart D341-32T.

generally be neglected, and the 50 per cent temperature taken as the average boiling point.

Experimental Work A large number of cuts of widely different types of stocks,

average boiling points, and widths of boiling range were col- lected to cover as completely ab possible the range encountered in commercial technology. Where possible the average boil- ing point, specific gravity, viscosities a t 100" F., 122" F., and 210" F., and aniline point were determined on each stock. Boiling points of the more volatile stocks were determined by Engler distillations a t atmosperic pressure. Where there was possibility of decomposition, the high-vacuum Engler distil- lation of Watson and Wirth (11) was used.

Viscosities were determined by Saybolt Universal, Furol, and Thermo viscometers on the stocks to which each WRS best suited. The calibrations of the A. S. T. M. were used for con- verting the Furol and Universal results to centistokes, and that of Fortsch and Wilson (5) was used for the Thermo in- strument. Checks were made on a number of samples using modified Ostwald pipets. These results were in satisfactory agreement with those of the industrial instruments.

Aniline points were determined by adding aniline in 0.5-cc. increments to 5 cc. of the oil until a maximum solution tempera- ture was reached.

In addition to the experimentally determined results, consideration was given to data from the literature, including those of FitzSimons and Bahlke (W, Hill and Coates (6) , and FitzSinions and Thiele (4) on viscosities, gravities, and boil- ing points, and Sachanen and Tilicheyev (7) on aniline points.

Results The experimental results on a typical list of stocks are sum-

marized in Table I. Column 9 gives characterization factors

determined from boiling point and gravity. These values were taken from Figure 1 which is a combined plot of the characterization factor and the molecular weight and critical temperature relationships of Watson and Kelson ( I O ) . The A. P. I. gravity and average boiling point fix the point repre-

3

DECEMBER, 1935 INDUSTRIAL AND ENGINEERING CHEMISTRY 1463

senting a stock on the chart. By interpolation between the curves, the characterization factor, molecular weight, and critical temperature may be estimated.

Figures 2, 3, and 4, relating viscosity and gravity as ordi- nates and abscissas with lines of constant boiling point and characterization factor, were obtained by plotting all the data collected, both experimentally and from the literature, and drawing the best average curves by cross interpolation be- tween these points. These charts relate characterization factor, viscosity, gravity, and boiling point so that, if any two properties are known, either of the other two may he esti- mated, From them it is possible to estimate viscosities of light stocks from boiling point data or approximate boiling points of heavy stock from viscosity data as well as to de- termine characterization factors from either boiling point or viscosity and gravity.

It was found that the best correlations are obtained with viscosities a t the highest possible temperature. Particularly in the case of heavy viscous stocks the relationship between boiling point and viscosity a t the lower temperatures is affected by the width of boiling range. This influence is greatly minimized at the higher temperature. For example, if the width of boiling range of a viscous stock is increased as by blending naphtha with it, its viscosity a t 100" F will be lower than predicted from Figure 4, whereas a t 210" E'. it will agree with the result from Figure 2. This corresponds to the effect of reducing the viscosity index of an oil by widening its boil- ing range or to the well-known "cutting action" of small amounts of light blending stock in reducing the viscosity of a fuel oil at low temperatures. For this reason considerable error may result from use of the high viscosity ranges of Fig- ures 3 and 4. These charts apply only to stocks of average width of boiling range. Figure 2 is more general in its appli- cability.

The upper curve in Figure 5 presents the average relation- ship between characterization factor and viscosity index ( 2 ) . A. pointed out above, the viscosity index is not a good method of characterization becaube of its dependence on width of boiling range. Stocks of wide boiling ranges will have vis- cosity indexes lying above the curve of Figure 5 . A further disadvantage to the use of viscosity index as a general method of cliaracterization lies in the limited viscosity range of stocks to which it is applicable as now defined. It was attempted to extend this definition to cover a wider range by a nomographic extrapolation of the original data (Q), but this method does not seem to warrant much extension beyond its original purpoae of evaluating lubricating oils.

The lower section of Figure 5 is a correlation between aniline point, c h a r a c t e r i z a t i o n factor, and average boiling point. This correlation is r a t h e r rough, in part be- cause of the difficulty of determining aniline points of heavy, dark-colored oils. There is also indication that aniline point is influenced by width of boiling range and the data, in general, were more erratic than the other properties measured.

It was p o i n t e d out by Sweeney and Voorhees (8, that the hydrogen content of a p e t r o l e u m fraction may be estimated as a func- tion of average boiling point and gravity. The curves of F i g u r e 6 r e p r e s e n t a

similar relationship, plotting hydrogen content against char- acterization factor for materials of constant boiling points. These curves were based on pure compound data and on the results of Rorgstrom, Norton, and Lewis ( 1 ) on fuel oils. The extrapolation of the available data by this method of plotting leads to results for lom-boiling materials of low characterization factor which are considerably different from thoje of the plot of Sweeney and Voorhees. Their plot is more consistent with the properties of benzene and its low- boiling homologs, but Figure 6 appears to agree somewhat better with measurements on highly cracked stocks. More data are necessary in the regions of disagreement before an entirely reliable plot can be determined. Fortunately, for the types of stocks ordinarily encountered, the agreement is good.

To serve as an indication of the accuracy of agreement be- tween characterization factors determined by various meth-

ods, values are inc luded in Table I for comparison with each other. Column 9 g i v e s characterization factors based on boi l ing point a n d g r a v i t y f rom Figure 1. Columns 10, 11, and 12 give factors based on gravity and v i s c o s i t y a t 210" F., 122" F., and 100" F. from Figures 2, 3, a n d 4, r e s p e c t i v e l y . Columns 13 and 14 give characterization fa c t o r s estimated from v i s c o s i t y indexes and aniline points through the relationships of F i g u r e s 5 and 6. The agreement with the origi- nal bo i l ing point-gravity factors is fair in all cases

1464 INDUSTRIAL AND ENGINEERING CHEMISTRY VOL. 17, NO. 12

with the worst deviations in the results based on aniline point (2)

The relationships proposed above are merely approxima- (4) tions and direct determination of a property is preferable to its estimation from other properties. However, it is believed (5 )

(6) that these charts are of sufficient accuracy to be useful in engineering problems where complete data are not available. (7)

and low-temperature viscosities. (3)

Acknowledgment (8)

AcknowIedgment is due J. L. Wien, J. E. Westenberg, and J. 0. Iverson for assistance in experimental work and cor- relations presented in this paper.

Literature Cited

(9) (10)

(11)

Davis, Lapeyrous, and Dean, Oil Gus J . , 30, No. 46, 92 (1932). FitzSimons, 0.. and Bahlke, W. H., Proc. Am. Petroleum Inat.,

FitzSimons, O., and Thiele, E. W., IND. EKG. CHEM., Anal. Ed.,

Fortsch, A. P., and Wilson, R. E., IND. ENQ. CHEM., 17, 291

Hill, J. B., and Coates, H. B., Ihid. , 20, 641 (1928). Sachanen and Tilicheyev, "Chemistry and Technology of

Sweeney, W. J., and Voorhees, A., IND. EKG. CHEM., 26, 197

Watson, K. M., Oil Gus J. , 33, No. 11, 34 (1934). Watson, K. M., and Nelson, E. F., IND. ENG. CHEM., 25, 880-7

Watson, K. M., and Kirth, C., IND. ENG. CHEM., Anal. Ed., 7,

11, No. 1, 70 (Jan. 2, 1930).

7, 11 (1935).

(1925).

Cracking," New York, Chemical Catalog Co., 1932.

(1934).

(1933).

72 (1936). RECEIVED M a y 6, 1935. Presented before the Division of Petroleum Chem- istry a t the 89th Meeting of the American Chemical Society, New York, N. Y., April 22 to 26, 1935.

(1) Borgstrom, P., Norton, R. D., and Lewis, 0. I., J . Am. SOC. Naval Engrs., 46, 173 (1934).

A New Low-Melting Alloy SIDNEY J. FRENCH, Colgate University, Hamilton, N. Y

HE fusible quaternary eutectic alloy usually called "Lipowitz eutectic alloy" is Q T commonly given the following percentage

composition: bismuth, 50; lead, 27; tin, 13; and cadmium, 10. The alloy melts sharply a t 72' C. and freezes at 70" C. No other alloy composed of these four metals melts below 72' C. although some references give the melting point of Wood's metal (bismuth, 50; lead, 25; tin, 12.5; and cadmium, 12.5) as low as 60' C. The addition of indium gives quinternary alloys having melting ranges below those of the quaternary alloys.

Preparation of Quinternary Alloys Quinternary alloys containing indium were prepared by adding

successive amounts of indium to the quaternary eutectic alloy. The quaternary alloy was prepared from c. P. metals which were accurately weighed, placed in a hard glass test tube, and heated to 325" C. The molten alloy was stirred constantly while cool- ing. The melting and freezing points of the quaternary eutectic alloy were determined by means of cooling and melting curves. Fifteen grams of the alloy were then placed in a small soft glass test tube and melted. Indium was added and the alloy was heated to 160" C. The alloy was then permitted to cool in an air bath and was constantly stirred during the process to prevent undercooling. The approximate freezing range was thus deter- mined.

Table I sho-xs the freezing ranges of these alloys. Since no changes were made in percentages of the other metals present, the ratios of bismuth to tin to cadmium remained as they were in the quaternary eutectic alloy, The percentages given in Table I are approximate since the alloys were not analyzed.

~

TABLE I. FREEZINQ POINTS OF ALLOYS Lipowita Freezing Lipowitz Freezing Alloy Indium Range -4110~ Indium Range

% % a c. % % 0 c. 100.00 99.01 95.04 97.55 95.48 94.02 93.11 91.24 89.46 57.76

0.00 0.99 1.96 2.95 4.54 5.98 8.89 8.76

10.54 12.26

69.7 6 8 . 0 M 9 . 5 65.5 -68.00 63.00-65.5 61.5 -63.5 67.00-60.5 56.00-59.5 54.00-57.00 52.5 -55.5 50.6 -53.00

86.09 84.50 83.15 51.64 80.19 78.82 77.66 75.00 88.67 50.00

13.91 15.50 16.85 18.36 19.81 21.18 22.34 25.00 33.34 60.00

49.5 -52.00 4 8 . 5 -51.00 48.00-50.00 47.00-48.5 47.00-48.5 47.00-48.5 47.5 -49.00 48.00-50.00 49.00-51.00 58.00-59.00

Each successive alloy was obtained by adding an accurately weighed amount of indium to the previous alloy.

Table I indicates that there is a regular fall in the freezing range as the percentage of indium is increased; the minimum is obtained when the percentage of indium reaches 18 to 20 per cent.

Cooling and Melting Studies Cooling and melting curve studies were made of alloys

containing the following percentages of indium: 28.30, 25, 22.34, 19.25, 18.58, 18.10, 17.79, and 16.50:

The alloy containing 28.30 per cent indium was first prepared by adding the required amount of indium to 12 grams of Lipo- wits alloy which had been prepared as indicated above. The other alloys were prepared in turn by the addition of appropriate amounts of Lipowitz alloy. It was thus unnecessary to remove any of the alloys from the tube in which the determinations were made. A small soft glass test tube was used for the studies. The thermometer, which was compared with a Bureau of Stand- ards certified thermometer, TYas placed with the bulb in the molten alloy and was used to stir the alloy, The test tube was placed in an 800-cc. water bath surrounded, in turn, by an air bath. The water was stirred with a motor stirrer. The bath was cooled by direct contact with the air of the laboratory, the rate of cooling being about 0.3" C. per minute. In determining melting curves, the bath was heated with a small shielded gas flame so that the rate of temperature rise was about 0.3' C. per minute.

Figures 1 and 2 show cooling and melting curves for alloys containing 16.5, 18.10, and 22.34 per cent indium. The percentages of indium given in the figures were not determined by analysis of the alloy but indicate merely the amount of indium added to the quaternary alloy. The alloy containing 18.1 per cent indium froze sharply at 46.7' C. and melted sharply a t 46.9" C. The other alloys showed greater melting- freezing ranges; none had a final solidifying temperature below that contJaining 18.1 per cent indium.

Figure 3 compares the cooling curves of two quinternary alloys in which the percentages of bismuth, lead, tin, and cad- mium were varied and the percentage of indium was kept constant a t 18.1 per cent. The composition of one of these alloys was the same as that shown in Figures 1 and 2; in the other the ratio of bismuth to lead to tin was 4 to 2 to 1. The percentage compositions were as follows: