enables a film of oil of uniform thickness to exist between the surfaces, and the resistances are not vitiated by the collision of projecting portions of the disk with each other. The rounded end of the upper shaft fits into a corresponding depression in the top of the upper disk. This method of connection retains the disk over the proper centre, yet it is allowed to sway enough to correct any irregularity of motion caused by imperfection of construction or wear of the lower disk. To obtain the desired condition of pressure, weights are placed directly upon the upper spindle. The axes of the upper and lower spindles do not lie in the same straight line, but are parallel, being about one-eighth of an inch out of line with each other; such construction, giving a discoid motion, prevents the disk from wearing in rings and assists in the uniform distribution of the oil. An arm is keyed through the lower part of the upper spindle and engages with projections upon the upper disk. Upon this arm which is turned to the arc of a circle, whose development is two and one-half feet, a thin brass wire is wrapped upon this arc and reaches to the dynamometer, so that the tension of the dynamometer is tangential and the leverage is constant for all positions of the upper disk within its range of motion. The dynamometer consists of a simple bar of spring steel fastened at one end, and bent by the pull applied at the other. Its deflection is indicated by a pointer upon a circular dial, the motion of the spring being multiplied about eighty times by a segment and pinion. The whole is enclosed in a steam-gauge case. When completed, the machine was subjected to a long series of tests with the same oil to determine the accuracy of the results, and the best method of procuring them. The operation of the machine under equal conditions with the same oil gives results which are as closely consistent with each other as could be expected from such physical measurements. As an example, four tests of the Downer Oil Co. Light Spindle (Sample No. 7) at 100° Fah. and on different days gave .1145, .1094, .1118,.1094; mean, .1113. Another example (Sample No. 14) Heavy Spindle Oil, made by the same firm, yielded for a coefficient of friction as the result of five different trials .1246, .1195, .1297, .1201, .1221; mean, .1233. Much of the irregularity, slight as it is, is due to the variable speed of the engine. Concurrent results were obtained under equal circumstances, but the coefficient of friction varied, not merely with the lubricants used, but also with the temperature, pressure, and velocity. The results of my own experiments on mediate friction do not agree with the laws of friction as given in works on mechanics, but the coefficient of friction varies in an inverse ratio with the pressure as shown graphically on diagram C. The variation of friction with the temperature is shown graphically on all sheets reporting the friction of each sample of oil. These curves belong to the hyperbolic class of a high degree; but I have not been able to deduce an equation which will answer to the conditions of more than one, because the law of the curves is modified by a constant dependent upon the individual sample of oil used. A little difference in the sample would cause a difference in the line of curve. Reference is made to diagram D showing the coëfficient of friction under equal ranges of temperature and velocity, but with a different series of pressures. 8 7 Curves showing Changes of Coefficient of Friction under Varying Conditions. Coefficient of friction at 100° and 500 revolutions per minute. The ratio of the changing coefficient varies with the temperature at which the range of results is taken. Friction varies with the area, because the adhesiveness of the lubricant is proportional to the area, and the resistance due to this cause is a larger fraction of the total mechanical effect with light, than it is with heavy pressures. The limit of pressure permitting free lubrication varies with the conditions; for constant pressures and slow motion it is believed to be about five hundred pounds per square inch, while for intermittent pressures, like the wrist pin of a locomotive, the pressure amounts to 3000 pounds per square inch. It has been stated that about 4000 foot pounds of frictional resistance per square inch is the maximum limit of safe friction under ordinary circumstances. As the results of this preliminary work indicated that the coëfficient of friction varied with all the circumstances, it was necessary to simulate the conditions of specific practical applications to determine the value of a lubricant for such purposes. It was decided to begin these investigations with spindle oils, and therefore the machine was loaded with five pounds to the square inch and run at about five hundred revolutions per minute; as the oil is then submitted to conditions of attrition corresponding to those met with in extremes of velocity and pressure, in the case of a Sawyer spindle running at 7600 revolutions per minute with a band tension of four pounds, and the results subsequently given refer only to the friction under these conditions, except when definitely stated to the contrary. This particular spindle was selected because, of the five million ring spindles in the United States, about one and a half million are of this manufacture, and in a large number of the remainder, the conditions of lubrication are quite similar. The apparatus is used in the following manner to measure the coefficient of friction of oil. After cleaning with gasoline and wiping carefully with wash leather, the disks are oiled and run for about five hours, being kept cool by a stream of water circulating through the upper disk. From time to time they are taken apart, cleaned, and oiled again. After using any oil, even if the disks are afterwards cleaned, the results with the oil subsequently used give the characteristics of the previous oil, and it is only after thirty-five to forty-five miles of attrition that these results bebecome consistent with each other; each succeeding result, meantime, approaching the final series. This seems to indicate that the friction exists at the surface of the two disks between the film of oil acting as a washer and the globules of oil partially embedded within the pores of the metal. If the dense bronze and steel retain the oil despite attempts to remove it, how much longer must it require to replace the oil in machinery with a new variety whose merits are to be tested. These experiments confirm the wisdom of the increasing use of cast iron for journals, as its porosity enables it to contain and distribute the lubricant. When the disks are ready to test the oil, the apparatus is cooled by the circulation of water; the flow of which is stopped when the machine is started. At every degree of temperature, the corresponding resistance is read on the dynamometer. When the thermometer indicates a temperature of sixty degrees, the counter is thrown in gear, and the time noted. When one hundred and thirty degrees is reached, the counter is thrown out of gear, and the time noted. This not only gives the velocity of the rubbing surfaces, but the number of revolutions required to raise the temperature a stated number of degrees, and is a close criterion of the oil. The coefficient of friction is the ratio of the pressure to the resistance, and is deduced in the following manner. In the friction of annular disks, the portions of the surfaces near the perimeter have a greater leverage than those near the centre. The mean sum of these moments is found by the calculus. Let p radius of any infinitesimal narrow ring or band. Width of band dp 66 2 pd p 66 (1) (2) (3) (4) The expression for the area of an annular disk = ~ (R2 — r2) (5) To express the moment of a ring in terms of an annular surface we divide (4) by (5), which gives |