| Sand not packed, per cent. voids | 44½ |
| Sand shaken to refusal, per cent. voids | 35 |
| Sand saturated with water, per cent. voids | 37½ |
Another series of tests made by Mr. H. P. Boardman, using Chicago sand having 34 to 40 per cent. voids, showed the following results:
| Water added, per cent. | 2 | 4 | 6 | 8 | 10 |
| Resulting per cent. increase | 17.6 | 22 | 19.5 | 16.6 | 15.6 |
Mr. Wm. B. Fuller found by tests that a dry sand, having 34 per cent. voids, shrunk 9.6 per cent. in volume upon thorough tamping until it had 27 per cent. voids. The same sand moistened with 6 per cent. water and loose had 44 per cent. voids, which was reduced to 31 per cent. by ramming. The same sand saturated with water had 33 per cent. voids and by thorough ramming its volume was reduced 8½ per cent. until the sand had only 26¼ per cent. voids. Further experiments might be quoted and will be found recorded in several general treatises on concrete, but these are enough to demonstrate conclusively that any theory of the quantity of cement in mortar to be correct must take into account the effect of moisture on the voids in sand.
The effect of the size and the shape of the component grains on the amount of voids in sand is considerable. Feret's experiments are conclusive on these points, and they alone will be followed here. Taking for convenience three sizes of sand Feret mixed them in all the varying proportions possible with a total of 10 parts; there were 66 mixtures. The sizes used were: Large (L), sand composed of grains passing a sieve of 5 meshes per linear inch and retained on a sieve of 15 meshes per linear inch; medium (M), sand passing a sieve of 15 meshes and retained on a sieve of 50 meshes per linear inch, and fine (F), sand passing a 50-mesh sieve. With a dry sand whose grains have a specific gravity of 2.65, the weight of a cubic yard of either the fine, or the medium, or the large size, was 2,190 lbs., which is equivalent to 51 per cent. voids. The greatest weight of mixture, 2,840 lbs. per cu. yd., was an L6M0F4 mixture, that is, one composed of six parts large, no parts medium and 4 parts fine; this mixture was the densest of the 66 mixtures made, having 36 per cent. voids. It will be noted that the common opinion that the densest mixture is obtained by a mixture of gradually increasing sizes of grains is incorrect; there must be enough difference in the size of the grains to provide voids so large that the smaller grains will enter them and not wedge the larger grains apart. Turning now to the shape of the grains, the tests showed that rounded grains give less voids than angular grains. Using sand having a composition of L5M3F2 Feret got the following results:
| —Per cent. Voids— | ||
| Kind of Grains. | Shaken. | Unshaken. |
| Natural sand, rounded grains | 25.6 | 35.9 |
| Crushed quartzite, angular grains | 27.4 | 42.1 |
| Crushed shells, flat grains | 31.8 | 44.3 |
| Residue of quartzite, flat grains | 34.6 | 47.5 |
The sand was shaken until no further settlement occurred. It is plain from these data on the effect of size and shape of grains on voids why it is that discrepancies exist in the published data on voids in dry sand. An idea of the wide variation in the granulometric composition of different sands is given by Table II. Table III shows the voids as determined for sands from different localities in the United States.
Table II.—Showing Granulometric Compositions of Different Sands.
| Held by a Sieve. | A | B | C | E |
| No. 10 | 35.3% | |||
| No. 20 | 32.1 | 12.8% | 4.2% | 11% |
| No. 30 | 14.6 | 49.0 | 12.5 | 14 |
| No. 40 | ... | ... | 44.4 | ... |
| No. 50 | 9.6 | 29.3 | ... | 53 |
| No. 100 | 4.9 | 5.7 | ... | ... |
| No. 200 | 2.0 | 2.3 | ... | ... |
| —— | —— | —— | —— | |
| Voids | 33% | 39% | 41.7% | 31% |
Note.—A, is a "fine gravel" (containing 8% clay) used at Philadelphia. B, Delaware River sand. C, St. Mary's River sand. D, Green River, Ky., sand, "clean and sharp."