MECHANICAL ANALYSES 



77 



density, namely a 6 per cent increase in the calculated 

 particle diameter, and a percentage error of 1.5 per cent 

 in the distribution curve of the samples analyzed. 

 Robinson (1926, 1927), among others, has suggested that 

 because of the difficulties in determination of particle 

 size, the data of mechanical analyses shotld be present- 

 ed in terms of settling velocities. In samples which 

 contain appreciable amounts of sand grades, however, 

 this procedure cannot be followed satisfactorily, and 

 even when the data are presented in terms of particle 

 diameters, considerable uncertainty arises in the join- 

 ing of the two parts of the distribution curves. As point- 

 ed out by Krumbein, a concept of equivalent radius is 

 implied in the mechanical analysis by sieving of irregu- 

 larly shaped sand grains, as the results of sieving are 

 based on the minimum and mean dimensions of the par- 

 ticles. The rate of sedimentation, on the contrary, is 

 dependent on the maximtun and mean dimensions of the 

 particles. 



6. The error due to the concentration of the settling 

 solution is negligible, according to Fisher and Oden 

 (1924), if the concentration is less than 2 per cent. 



7. There are certain errors inherent in the pipette 

 method itself which arise in the withdrawal of the pi- 

 pette sample. These are owing to (1) the inaccuracy in 

 measuring the depth of samples, (2) a certain amount of 

 mixing which accompanies the introduction of the pipette 

 into the suspension, and (3) the fact that the pipette does 

 not withdraw a horizontal stratum of extreme thinness 

 as required by theory, but rather taps a spherical zone. 

 Kohn (1928) and Keen (1931) have shown that the errors 

 due to this last cause are practically negligible. 



In an attempt to determine empirically the magni- 

 tude of the errors in the pipette analyses, duplicate and 

 triplicate analyses were carried out on samples 31 and 

 19, respectively. The results are given in table 25. The 

 two parts of sample 31, a red clay from the south Pacif- 

 ic containing 20 per cent calcium carbonate, remained 

 dispersed throughout the period of the experiment. The 

 deviation in duplicate determinations of the fraction 

 greater than 31.3 microns in particle size was 3.6 per 

 cent, an error probably arising in the initial sampling. 

 For the remaining size grades the average deviation be- 

 tween duplicate determinations was 1.0 per cent and the 

 probable error of a single determination was 0.6 per 

 cent. The average deviation of the cumulative percent- 

 ages was slightly greater. These results may be com- 

 pared with those given by Gripenberg, who obtained an 

 average deviation between duplicate determinations of 

 0.8 per cent. A calculation from the data given by Rit- 

 tenhouse shows a probable error for a single determina- 

 tion of 0.54 per cent. It should be emphasized that the 

 measures of variation or error given above are ex- 

 pressed as percentages of the total amount of material 

 in the sample analyzed, and not of the amount of materi- 

 al in the size grades. Correns and Schott (1932) have 

 published comparative data on mechanical analyses by 

 the method of Atterberg, a modified Oden method, and 

 the pipette method. Although the results of duplicate 

 analyses are presented only graphically by these au- 

 thors, a study of their figures shows that deviations be- 

 tween duplicate analyses by the pipette method are of 

 about the same order of magnitude as those given above. 

 The general agreement between different methods of 

 analyses is also quite good. 



The agreement between the triplicate determinations 

 in sample 19 is very satisfactory for grade sizes be- 



tween 15.6 and 2 microns. Below this point, however, 

 the analyses show a marked discrepancy which probably 

 is owing to differing degrees of dispersion between the 

 three parts analyzed, and to slow coagulation. 



Results. As already stated, the results of the pi- 

 pette analyses are given in table 23 which lists the per- 

 centages in each size grade, and in table 24 which gives 

 the statistical constants for each sample. The analyses 

 may be conveniently placed in two groups, those of clays 

 low in calcium carbonate, and those of Globigerina 

 oozes. Figures 13 to 21 give histograms and cumulative 

 curves in terms of both settling velocities and equivalent 

 diameters for the nine clays analyzed, namely, numbers 

 31, 34, 49, 69, 70, 72, 73, 74, and 77; the figures tor the 

 Globigerina oozes, numbers 19, 21, 23, 40, 43, and 81 

 are grouped in figures 22 to 27. 



The most striking fact shown by the analyses of the 

 clays is the essential uniformity of the particle size dis- 

 tribution of these samples, with the exception of no. 49. 

 The shape of the curve for this sample almost certainly 

 indicates that coagulation occurred during analysis, and 

 this was confirmed by the appearance of stratified layers 

 in the suspension. Sample 31, doubtless owing to the 

 presence of 20 per cent calcium carlx>nate, is somewhat 

 more coarse-grained than the other clays, and sample 

 72, which contains a greater amount of siliceous organic 

 remains than most north Pacific red clays, also has a 

 slightly greater median diameter. The average median 

 diameter of the remaining clays, nos. 34, 69, 70, 73, 74, 

 and 77 is close to 1.05 microns, and the deviations be- 

 tween the percentages found in each size grade are almost 

 within the limits of error of the method of analysis, the 

 deviation of the percentages in nos. 69, 70, 73, 74, and 

 77 from the average for these samples being 1.20. The 

 coefficient of sorting of these clays averages about 2.9, 

 ranging from 3.17 in sample 74 to 2.72 in sample 73. 

 The clays thus show a fair degree of sorting. Theskew- 

 ness is small in every sample except no. 74 which shows 

 a moderate skewness toward the coarse side, perhaps 

 caused by convection currents during analysis. The av- 

 erage cumulative curve for samples 69, 70, 73, 74, and 

 77 is shown in figure 12, compared with results for other 

 marine clays obtained by Galliher (1932), Rubey (1930), 

 Stetson (1933), Trask (1932), and Gripenberg (1934). The 

 general parallelism between all the curves is easily 

 seen, but it will be noticed that the only sediments as 

 fine-grained as these clays are those obtained from the 

 Baltic Sea by Gripenberg. On the other hand, the marine 

 Middle Oligocene clay from Malliss, analyzed by Cor- 

 rens and Schott (1932), is even more fine-grained than 

 any of the Pacific clays, as are many soils. OdSn(1915), 

 in his analysis of a Challenger sample from the Atlantic, 

 pointed out the fine-grained nature of red clay. The 

 curve given by Oden for a Challenger red clay from the 

 south Pacific is very similar to that obtained for Carne - 

 gie sample 49 from the same area, and probably also 

 represents coagulation during or preceding analysis. 



The curves for the Globigerina oozes, samples 19, 

 21, 23, 40, 43, and 81, are much more complex in type. 

 Well -developed double maxima in the sand and clay 

 grades are shown in the histograms of samples 19 and 

 43, and to a lesser extent of the remaining samples, ex- 

 cept no. 40. Double maxima are also strikingly shown in 

 the analyses by the Bureau of Soils' method given below. 

 Oden likewise found curves showing two maxima in his 

 analyses of Globigerina oozes from the Atlantic and Pa- 

 cific. A few analyses, given in table 26, of the calcium 



