FRICTION OF PURE METALS AND THE INFLUENCE OF ABSORBED GASES. TEMPERATURE COEFFICIENT OF FRICTION¹

As a result of numerous studies of electron emission from the surface of a metal, of electron diffraction, and also of optical investigations, it has been firmly established that only a freshly cleaved surface—but not a surface cleaned in air—is covered by an oxide film, an adsorption film of water vapor and of various gases; in a word, by a film of contamination; its thickness may be only a few molecules. Therefore, in experiments connected with surface phenomena, one is dealing not with the clean surface of a metal, but with a film of contamination on that surface. It must also be borne in mind that the lubricant is always located on top of the contamination film and usually displaces the latter. Naturally, contamination films have a great qualitative and quantitative influence on friction, since the latter is a process taking place on the surfaces of bodies.

The contamination film—usually very thin—is very difficult to remove, and for this reason there are very few studies of friction in vacuum with purely metallic surfaces. Holm and co-workers obtained a very clean metal surface by heating in a high vacuum and found that, under these conditions, metals adhere on contact; moreover, in the absence of inert gases, as well as of hydrogen or nitrogen, this adhesion did not disappear, but ceased upon the introduction of oxygen. Further, Holm and co-workers (see the bibliography in the review article by D. Penner, Zhurnal tekhnicheskoi fiziki, 10, 265–273, 1940) measured the magnitude of static friction by the oscillation method and found that, when metal surfaces are degassed, i.e., freed from the contamination film, the force of static friction is large—much greater than for ordinary unlubricated surfaces.

The authors measured the coefficients of dynamic (kinetic) friction between thoroughly degassed metals and established that it is 20 times greater than for ordinary unlubricated surfaces. The role of gases then introduced into the vacuum was that some of them reduced the frictional force, while others had very little effect in this respect or had no effect at all. In the case of friction of purely metallic surfaces, temperature had a variable and irreproducible influence even with the slightest contamination of the metals; if, however, the surfaces were cleaned as thoroughly as possible, the results proved reproducible and reversible; in this case it was established that the coefficient of friction slowly but steadily decreased as the temperature rose.

The authors calculated the value of the coefficient of dynamic (kinetic) friction from the negative acceleration (deceleration) of the sliding body, as the second derivative of the distance traveled with respect to time, from the equation

\[ \frac{d^2 S}{dt^2} = -\mu g, \]

where \(\mu\) is the coefficient of friction, and \(g\) is the acceleration due to gravity. The apparatus for this purpose consisted of a glass curved vacuum tube, shown in Fig. 1. The sliding body was a hollow cylinder \(C\), mounted in bearings, which could slide along rods made of various metals passed through it. On the left a spring with a forked lower end is visible; this spring was drawn back (each time in the same way) to the left by means of an electromagnet \(M\): when current was then switched on in the latter, the electromagnet attracted a piece of soft iron embedded in the Pyrex-glass spring \(S\). When the current was switched off, the spring struck the cylinder \(C\), which had previously been advanced (if necessary) and rotated about its axis by means of small pieces of soft iron \(I\), fused into the glass and controlled from outside the entire apparatus by manipulations with a hand electromagnet.

By means of a scale placed parallel to the horizontal rod along which the cylinder \(C\) slid, it was possible to read off accurately the distance \(S\) traversed by the cylinder. Knowing the time \(t\) of this path, one could determine the deceleration of the cylinder due to friction, and from this, as indicated above, calculate the coefficient of friction, taking the impulse received by the cylinder \(C\) to be constant. The horizontal rod—the lower stationary friction surface—was suspended on nickel loops \(X\) and \(Y\); these loops allowed the rod to expand thermally. A current passed through the rod from contacts \(X\) and \(Y\), heating the rod and (under high-vacuum conditions) degassing it. The cylinder \(C\) was degassed when it was placed on the rails \(R\) on the right: in this position it was heated in vacuum by a stream of electrons from a tungsten filament \(F\), heated by a current. Friction was measured after the surfaces of the degassed metals had cooled. The vacuum inside the apparatus was produced by two oil pumps, and during measurement of the friction force the vacuum was maintained constant at \(10^{-6}\) mm.

Fig. 1. Apparatus for studying friction

To introduce a definite gas into the apparatus, three-way stopcocks were used. The cylinder \(C\) was illuminated from the side, and its shadow fell into the slit of a camera in which a photographic film moved; in this way a path–time curve was obtained (time marks were obtained by directing onto the film a beam, parallel to the cylinder, that passed through a rotating disk with divisions, driven by a synchronous motor: in this way a series of lines marking intervals of 0.003 sec. was recorded on the film). The path–time curve had the form of a parabola, and the deceleration could be calculated for any instant of time by graphical differentiation. Thus the mean coefficient of dynamic (kinetic) friction was derived over the entire distance (the entire time interval) of sliding of the cylinder \(C\).

Above all it was found that the coefficient of (dry) friction \(\mu\) is constant over a large range of sliding velocities—from very small values up to \(100\ \mathrm{cm/sec}\). If the constancy of the tension of the spring \(C\) was maintained at each push, then the product of \(\mu\) by the distance \(l\) traversed by the cylinder \(C\) proved constant for a given cylinder and spring (i.e., the magnitude of the impulse—the push) for different surfaces of the rod, and therefore the value of the coefficient of friction was inversely proportional to \(l\): the apparatus was calibrated by measuring the product \(\mu l\), and from this \(\mu\), the coefficient of friction, was determined. The scatter of the values of \(l\) in a series of measurements was up to 5% for small values of \(\mu\); when \(l\) was small, i.e. the coefficient of friction was high, the error was greater.

In the apparatus described, Holm investigated the friction of nickel on tungsten, nickel on nickel, copper on copper, and gold on gold. For the first case, when the metal surfaces were “clean” (in air), \(\mu = 0.3\). When the air had been evacuated from the apparatus, \(\mu\) increased to 0.5. When the rod was heated by a current (and the sliding cylinder was heated by it as well), \(\mu\) increased to 0.7.

Finally, the rod and the cylinder were separated from one another and heated to the maximum attained under the conditions of the experiment: nickel to \(1300—1400^\circ\) K, tungsten to \(1650^\circ\) K; then these bodies were allowed to cool, contact was re-established between them, and the friction was measured quickly: it proved to be very large. When similar experiments were repeated several times, the coefficient of friction continued to increase and after several hours reached a maximum (about 6, Fig. 2).

Fig. 2. Effect of degassing on friction

Sliding: \(a\)—nickel on tungsten, \(b\)—nickel on nickel, \(c\)—copper on copper, \(d\)—gold on gold; \(A\)—pumping, \(B\)—degassing of the wire, \(C\)—degassing of the wire and cylinder, \(D\)—degassing of the plate, \(E\)—degassing of the cylinder and plate

Fig. 3. Effect of oxygen on the friction of pure metals

Sliding: \(a\)—nickel on tungsten, \(b\)—nickel on nickel, \(c\)—copper on copper, \(d\)—gold on gold; \(A\)—admission of oxygen (beginning), \(B\)—after 12 hours; \(B_1\)—state after 5 hours, \(B_2\)—state after 24 hours, \(B_3\)—state after 16 hours, \(B_4\)—state after \(\frac{1}{2}\) hour, \(C\)—visible formation of a film

If the degassed surfaces that had then been tested were left at room temperature under a vacuum of \(10^{-5}—10^{-6}\) mm, the coefficient of friction gradually fell: apparently, contamination of the friction surfaces by residual gas had an effect.

The paper notes that, according to Langmuir and Villars (1931), the oxide film on tungsten is destroyed only when heated to \(1500^\circ\) K, and later (1935) K. Leuter and Roberts found that this film is completely removed only when heated to \(2000^\circ\) C. Bosworth and Rideal (1937) determined that even at \(1600^\circ\) K a tungsten surface is still covered with its oxide, while van Cleave (1938) found that the oxide film is rapidly removed only at \(1750^\circ\) K.

The friction curves for nickel on nickel, copper on copper, and gold on gold, with the same course of the experiment as described above, are shown in Fig. 2.

When, after the experiment described, traces of oxygen were introduced into the apparatus (at a pressure of only 0.3 mm Hg), this quickly caused a decrease in the value of the coefficient of friction (Fig. 3; time is plotted on the abscissa). When the apparatus containing oxygen was left to stand for many hours, the friction decreased still further, and the admission of larger quantities of oxygen did not change the effect. In the case of an admission of pure oxygen, the coefficient of friction did not change. When the residual gas in the apparatus was

hydrogen, the maximum magnitude of the frictional force remained steady. Carefully purified nitrogen also had no effect on the magnitude of the friction, but if there were even traces of oxygen in nitrogen, the result obtained was the same as with oxygen alone. Mercury vapor also reduced the coefficient of friction.

To determine the influence of temperature on friction, the rod was heated by an electric current (its temperature was measured at a distance with an optical pyrometer). Without careful degassing, it was impossible to obtain definite, convergent results for the influence of temperature on friction: it could both increase and decrease with temperature (Fig. 4).

With very careful degassing and a high vacuum, the experimental results became reproducible and reversible (Fig. 5): for all metals except gold, friction decreased with increasing temperature. Gold, apparently, softened with heating, which caused point welding over a larger area. In general, the results on the influence of temperature make it possible to judge that the surfaces were sufficiently clean.

Fig. 4. Influence of temperature on the friction between surfaces of nickel and tungsten, incompletely degassed.

Fig. 5. Influence of temperature on the friction between clean metallic surfaces.

Sliding: a—tungsten on tungsten, b—nickel on nickel, c—copper on copper, d—gold on gold.

In most experiments, after the cylinder had passed along the rod several times under the same conditions, the friction decreased somewhat; this apparently occurred, first, because irregularities on the rubbing surfaces were smoothed out, and secondly (when the decrease in the coefficient was significant), because those traces of gas which were still retained between individual grains and crystallites of the metals and which were released in the process of wear of the metals during friction were being removed.

The most important conclusion of the authors’ work with respect to the phenomena of so-called dry friction (i.e., of unlubricated surfaces) is, in practice, that contamination films are, generally speaking, useful in engineering, since they greatly reduce friction in comparison with pure metals.

M. L. Goldovskii, Moscow

REFERENCES

1. F. P. Bowden and T. P. Hughes, Proc. Roy. Soc., 8, 263, 1939.