Recent Advances in the Production and Measurement of High Vacuum*

S. Dushman, Schenectady, U.S.A.

Introduction

Interest in high vacuum was awakened approximately 18 years ago, together with the development of research on electron emission and the application of electron tubes in radio engineering and other fields. About 10 years ago the author published in the journal General Electric Review a series of articles devoted to this question, which soon afterward were issued as a complete book. In these and subsequent works** the questions of producing high vacuum and methods of measuring extremely low pressures were treated in great detail. The need for a detailed exposition of this subject was connected with the extraordinarily broad interest in it both on the part of workers in the field

* S. Dushman, Journ. Franklin Inst., 211, 689, 1931; translated by N. A. Shishakov.

** On vacuum technique the following books have appeared: L. Dunoyer, La technique du vide, Soc. Franc. de Phys., 1924; English translation, Vacuum Practice, D. Van Nostrand and Co., N. Y., 1926; Russian translation, Tekhnika vysokogo vakuuma, GNTI, 1931. F. H. Newman, The Production and Measurement of Low Pressures, D. Van Nostrand and Co., N. Y., 1925. G. W. C. Kaye, High Vacua, Longman, Green and Co., 1927. S. Dushman, High Vacuum, Gen. El. Rev., 1922; German translation, Berthold und Reimann, Julius Springer, Berlin, 1926. A. Goetz, Physik und Technik des Hochvakuums, Vieweg und Sohn, Braunschweig, 1926.

pure science, and also on the part of persons working in the field of the manufacture of various vacuum apparatuses.

From this point of view it seems desirable to consider in this article, first, the successes that have been achieved in recent years, and then the results of a whole series of investigations which, in connection with all this, are of special interest.

The directions along which, over the last 10 years, the greatest successes have been achieved may be described separately in the following three chapters.

1. The use of condensation pumps with especially high pumping speeds.

2. The use of getters or absorbing agents for improving the vacuum in apparatus after a certain preliminary pumping.

3. Increased attention to the question of treating the metallic parts that are introduced inside vacuum apparatuses.

A considerable part of the material available here on these questions has been borrowed from laboratory reports and from discussions with various members of our research laboratory. The names of individual persons who took part in the investigations on these questions are mentioned separately in each particular case, and here the author wishes to take the opportunity to express his gratitude to all these members of our organization, especially to M. Andrews and G. H. Payne, for permission to use their instructions in this article.

General considerations concerning vacuum pumps

Before touching on the question of the development of high-speed pumps, it should be recalled that the chief requirements for any vacuum installation are the following:

1. The attainment of an extremely low pressure.

2. A high speed, which, as far as possible over the widest limits, must be independent of the prelimi-

...of the ultimate rarefaction, and also, in the case of condensation pumps, on the temperature of the boiling vessel.

Degree of Rarefaction Obtained

In modern vacuum technology it has become possible to regard as an axiom that a good pump can evacuate to pressures below \(10^{-5}\) mm of mercury, and in all probability even to \(10^{-6}\) mm. As the scientific unit of pressure, called the bar, the pressure of 1 dyne per \(1\ \mathrm{cm}^2\) is accepted. For technical purposes it is more convenient to use the micron, i.e. \(10^{-3}\) mm of mercury.* Consequently, the basic requirement for a high-vacuum installation is that with its aid it should be possible to reduce the pressure to approximately \(10^{-3}\ \mu\) of mercury. Practically all types of mercury condensation pumps satisfy this condition. In the case of single-stage pumps, such as those described below, for good operation it is required that the preliminary pump produce a moderate rarefaction (approximately \(10^{-3}\) mm); with two- and three-stage pumps this preliminary rarefaction may reach 10 mm and even more.

It is interesting to note that even at such a comparatively high vacuum of \(10^{-3}\ \mu\), in \(1\ \mathrm{cm}^3\) at \(0^\circ\mathrm{C}\) there are \(3.56 \cdot 10^{10}\) molecules. This clearly shows how far the fulfillment of this condition is from what is called a perfect vacuum. On the other hand, this limiting pressure amounts to only a \(10^{-9}\) part of atmospheric pressure.

Pumping Speed

The speed \(S\) of a vacuum installation is defined by the following relation:

\[ \frac{dp}{dt}=\frac{S}{V}(p-p_0), \tag{1} \]

* Since 750 mm of mercury is, to an accuracy of \(1/10\,000\), equal to \(10^6\) bars, we have:

\[ 1\ \text{bar}=0.75\cdot 10^{-3}\ \text{mm}, \]

\[ 1\ \mu=1.333\ \text{bars}. \]

where \(p\) denotes the pressure at the given moment of time, \(V\) the volume of the evacuated system, and \(p_0\) the lowest attainable pressure. From (1) the following expression follows:

\[ S=\frac{V}{(t_2-t_1)}\log_e\left(\frac{p_1-p_0}{p_2-p_0}\right) \tag{2a} \]

or

\[ \frac{p_2}{p_1}=e^{-\frac{S(t_2-t_1)}{V}}, \tag{2b} \]

if \(p_0\) is very small in comparison with the pressures \(p_1\) and \(p_2\), respectively, at the times \(t_1\) and \(t_2\). Thus \(S\) is expressed as volume per unit time, and by a speed of \(1\,000\ \mathrm{cm^3/sec}\) it is meant that, under these conditions, the pressure in a volume of \(1\,000\ \mathrm{cm^3}\) decreases each second to \(1/e\) of its initial value, i.e. to a value of 0.3678. As the investigations of M. Knudsen have shown, a tube offers resistance to the motion of a gas, which may be denoted by \(1/F\), where \(F\) is the conductance*. At very low pressures (when the mean free path has approximately the same magnitude as the diameter of the tube), \(F\) is determined (for a tube with a circular cross-section) by means of the relation:

\[ \frac{1}{F}=\left(\frac{2.394L}{D^3}+\frac{8.184}{D^2}\right)\sqrt{\rho_1}. \tag{3} \]

In this equation \(L\) and \(D\) denote, respectively, the length and diameter of the tube, and \(\rho_1\) is the density at a pressure of 1 bar, which at the absolute temperature \(T\) is determined from the relation:

\[ \rho_1=\frac{M}{83.15\cdot 10^6 T}, \tag{4} \]

where \(M\) is the molecular weight of the gas.

The term with \(D^3\) in equation (3) shows the strong influence of the length, and the term with \(D^2\) the influence of the ends. It is evident that for small values of \(D\) and large lengths \(L\), the quantity \(F\) varies approximately proportionally to \(D^3\) and inversely proportionally to \(L\). The curves in Fig. 1 illustrate equation (3) for the case

* M. Knudsen, Ann. d. Phys. 28, 75, 999, 1909.

air at 20°C;* here, on a logarithmic scale, \(F\) is given as a function of \(L\) for the values of \(D\) indicated at the bottom of the drawing. The straight-line part of each curve shows the effect

Fig. 1. Rate of water flow through tubes at 20°C in cm³/s at 1 bar pressure.

* This drawing was made by J. X. Peignon in connection with his investigations on high-speed pumps.

of the term with \(L/D^3\) in equation (3), while the curved part represents the influence of the term containing \(D^2\).

It follows from equation (3) that, for identical tube dimensions, the outflow velocity of hydrogen as compared with air will be \(\sqrt{29/2}\), or approximately 3.8.

Owing to this resistance to the outflow of gas, exerted by the connecting tubes, the true pumping speed \(E\) will depend both on the speed of the pump itself \(S\) and on the conductance \(F\); moreover, the relation between these three quantities will have the same form as the expression for the series connection of electrical resistances, i.e.

\[ \frac{1}{E}=\frac{1}{S}+\frac{1}{F} \tag{5} \]

or

\[ E=\frac{SF}{S+F}. \]

Table I gives the actual values of \(E\) that can be obtained for various values of \(F\) in the case of pumps having speeds \(S=5000\) and \(20000\ \mathrm{cm^3/sec}\). As these figures show, the influence of a large value of \(S\) becomes more and more insignificant as \(F\) decreases, so that for small values of \(F\) the quantity \(E\) becomes independent of \(S\) and approaches the value of \(F\).

Thus, for a more advantageous use of the maximum speed of a pump, it is very important that \(F\) be as large as possible. On the other hand, these considerations lead to the obvious conclusion that in those cases where the dimensions of the connecting tubes, for whatever reason, are not made especially large, it will be entirely useless to employ pumps with a very high speed. Thus, for example, for \(F=500\) (\(L=28\ \mathrm{cm}\), \(D=1\ \mathrm{cm}\)) a pump having speed \(S=5000\) will be almost as good as a pump for which \(S=20000\). This circumstance requires special attention in those cases where traps with liquid air are used between the pump and the system being evacuated.

RECENT ADVANCES IN THE FIELD OF HIGH VACUUM

TABLE 4

Pumping speeds for various values of \(F\)

\(F\)	\(E\) for \(S = 5\,000\)	\(E\) for \(S = 20\,000\)
\(\infty\)	5 000	20 000
20 000	4 000	10 000
10 000	3 303	6 667
5 000	2 500	4 000
2 500	1 667	2 222
1 000	833	952
500	454	488

Method of measuring the speed of a pump*

The most commonly used method for measuring the speed of a pump is based on equation (3). A capillary tube, calibrated on the basis of this equation, is connected between the pump and a reservoir at constant pressure \(P_2\) (which can be determined by means of a McLeod manometer). The pressure \(P_1\) directly at the opening to the pump is determined by means of an ionization manometer or some other manometer capable of measuring very low pressures. The speed \(S\) is determined by means of the relation:

\[ S = K\left(\frac{P_2 - P_1}{P_1}\right), \]

where \(K\) is a constant for each capillary. Fig. 2 shows a diagram of the apparatus used by J. H. Neill in his investigations of the speeds of mercury condensation pumps of various designs, which are described in the following paragraph.

Mercury condensation pumps**

1. Gaede diffusion pump

In 1915 Gaede published a description of a diffusion pump.*** The operation of this pump can best be understood from consideration of Fig. 3.

* See Kay’s book, Ch. X.

* A more detailed analysis of previous investigations and a description of various pumps can be found in the books on vacuum technique named in the footnote at the beginning of the article, and also in the following papers: A. Gellert, Naturwiss., 7, 988, 1919. W. Gaede, Z. techn. Phys.*, 4, 337, 1923.

* W. Gaede, Ann. d. Phys., 46, 357—392, 1915.

“A jet of steam moves through the tube \(AB\), in which a porous diaphragm \(C\) is fixed. The vessel being evacuated is connected at \(E\). Water vapor diffuses through the capillaries in the diaphragm into the trap \(D\), where it is condensed by means of some cooling agent; at the same time air diffuses through the diaphragm in the opposite direction into the tube \(AB\), from which it is rapidly carried away by the jet of steam. As a result, the pressure in \(E\) decreases and eventually reaches a very small value.”

Fig. 2. Diagram of measuring pump speeds.

Fig. 3. Diagram illustrating the principle of operation of a diffusion pump.

The pumping speed as a function of the vapor pressure and the dimensions of the capillaries was derived by Gaede from consideration of the simple case illustrated in Fig. 4. Here \(AB\), as in Fig. 3, represents a jet of vapor, and \(EF\) is a simple capillary tube, which at the point \(E\) is cooled by means of the jacket \(K\). As a result, all vapor molecules diffusing into the side tube condense on the wall at \(E\).

The mathematical investigation of the diffusion process led Gaede to the following relation for the volume \(V\) of gas which diffuses each second through the opening \(F\) into the vapor:

\[ V=\frac{1}{L}\cdot\frac{\pi r^{3}}{2\eta}\,e^{-rP/1520D\eta_{2}}, \tag{7} \]

where \(L\) is the length of the tube \(EF\),
\(r\) is its radius,
\(\eta_{1}\) is the coefficient of viscosity of the gas,
\(\eta_{2}\) is the coefficient of viscosity of the vapor,
\(D\) is the coefficient of diffusion of the gas,
\(P\) is the vapor pressure near \(F\).

Fig. 4. Diffusion through a capillary into a vapor jet.

Fig. 5. Pumping speed by diffusion through a capillary.

Fig. 5 shows the dependence of \(V\) on \(r\); here one can see that for large values of \(r\) the volume \(V=0\), and that when \(r\) is decreased to a certain value a sharp maximum is obtained. In addition, equation (7) shows that the speed depends on the vapor pressure \(P\), and, in order to decrease the factor with the exponent approximately to unity, \(r\) must vary inversely with \(P\). Thus, for large values of the vapor pressure, \(r\) must be extremely small. However, the equation also shows that the pumping speed does not depend

on the pressure of the residual gas. In this respect Gaede’s diffusion pump represents a definite advance in comparison with pumps of the mechanical type, which had been used up to that time.

The very first form of the diffusion pump, designed by Gaede, is shown in Fig. 6. “The porous diaphragm has been replaced by a steel cylinder \(C\) with a narrow slit \(S\), the width of which can be varied by means of the screw \(H\). The cylinder is immersed at its lower end in the mercury ring \(G\), which forms a seal between the regions of low and high pressure. The mercury in \(A\) is heated, and its vapor passes through the slit in the steel cylinder in the direction indicated by the arrow. Air or another gas from the vessel being evacuated (connected at \(F\)) diffuses into this mercury vapor at \(S\) and then exits through \(E\) into the backing-vacuum pump (connected at \(V\)). The mercury vapor emerging through \(S\) condenses on the walls of the glass, owing to the cooling jacket \(K_1K_2\). The opening \(V\) serves for connection to the backing pump and is used to evacuate the system until the pressure becomes sufficiently low for the diffusion pump to begin working well. As soon as this point is reached, the mercury automatically separates the two spaces, after which evacuation continues already with the aid of the diffusion pump.”

Fig. 6. The first Gaede diffusion pump.

With the aid of this pump, operating at a backing vacuum of about \(0.05\) mm, one can obtain pumping speeds

pumping speed from 50 to 200 cm³/sec, depending on the width of the slit \(S\) and the temperature of the mercury (indicated by thermometer \(T\) in Fig. 6).

2. Langmuir’s Condensation Pump

As a result of a whole series of investigations carried out by Langmuir on the condensation, evaporation, and reflection of molecules on various surfaces, he found that the speed of a mercury pump can be considerably increased if the gas molecules diffusing into the pump from the vessel being evacuated can be given an additional velocity by collisions with high-speed mercury atoms in the vapor jet, and if, at the same time, these atoms can condense immediately as soon as they strike the walls, so that they cannot diffuse back into the system. The old form of the glass condensation pump based on this principle is shown in Fig. 7. The following description of the construction and operation of this pump is taken by us from Langmuir’s article.*

Fig. 7. Langmuir condensation pump. Original form.

“For the pump to operate well, it is very essential that the end of the nozzle \(L\) be placed below the level at which the water stands in the cooler \(J\). In other words, the drain tube \(K\) must be placed somewhat higher than the lower end of the nozzle, as is indicated in the figure. The other dimensions of the pump play a comparatively—

* I. Langmuir, Phys. Rev. 8, 149, 1916; Gen. El. Rev., 1060, 1916.
* I. Langmuir, Journ. Frankl. Inst. 182*, 719, 1916.

plays a considerably less important role. The distance between \(L\) and \(D\) must be sufficiently large so that, against the stream of mercury vapor, no appreciable amount of gas can diffuse back; moreover, a sufficiently large condensation surface must be present.

The pump may be made of any size. Some of the pumps constructed by us had tube \(B\) and nozzle \(L\) about 30 mm in diameter; in other pumps the diameter of this tube was about 5 mm, and the length of the entire pump was only about 90 mm. The larger the pump, the greater the pumping speed that can be obtained.

When the pump is in operation, the boiler with mercury \(A\) is heated either by gas or electrically, so that the mercury evaporates at a moderate rate. A thermometer placed in contact with tube \(B\) under thermal insulation, when the pump was working well, usually indicated a temperature of 100 to \(120^\circ\)C. Under these conditions the mercury in boiler \(A\) evaporates from the surface at a sufficiently high rate. No bubbles are formed in such a case, so that the possibility of shocks during heating is almost always eliminated.

In contrast to the Gaede diffusion pump, there are here no critical data for regulating the temperature. In the case of a pump with electric heating, where nozzle \(L\) is about 20 mm in diameter, the pump begins to operate well at an energy expenditure of about 220 W. The pumping speed remains practically constant up to an increase of the electrical energy to 550 W.

But the forepressure at which the pump operates depends on the quantity and speed of the mercury vapor passing through the nozzle. Thus, for example, in the above-mentioned case with heating at 220 W the pump could not operate with a preliminary vacuum worse than 50 bar, whereas at 550 W a preliminary vacuum up to 800 bar had absolutely no effect on the operation of the pump.

For technical work a metallic form of the condensation pump was constructed, which is shown schematically in Fig. 8.

“A metal cylinder \(A\) has two openings \(B\) and \(C\), of which opening \(B\) connects it with the backing-vacuum pump, and opening \(C\) with the receiver. Inside the cylinder is a funnel-shaped tube \(F\), which rests on the bottom of cylinder \(A\). At the top of the cylinder is suspended a cup \(E\), turned bottom upward, above tube \(F\). A water jacket \(J\) surrounds the walls of cylinder \(A\) from the level \(B\) up to a certain level above the lower edge of cup \(E\).

Mercury is placed in the cylinder at \(D\), as is seen from the figure. When the bottom of the cylinder is heated, the mercury evaporates. Its vapors pass through \(F\), are deflected by cup \(E\), and are directed downward toward the outer cooled wall of cylinder \(A\). The gas entering at \(C\) passes downward between \(A\) and \(E\); then in \(F\) it meets the jet of mercury vapor and from there is carried downward along the walls of \(A\) into tube \(B\). The mercury, which condenses on the walls \(A\), falls along the lower part of funnel \(F\) and returns back to \(D\) through small openings located at the place where the funnel touches the bottom of the cylinder.”

Fig. 8. Diagram of the construction of a nonmetallic condensation pump.

Fig. 9. Modern single-stage metal condensation pump.

Fig. 9 shows a modern metal conden-

suction pump constructed in this laboratory by MacGroom. It has a speed of 5,000–6,000 cm³/sec and can operate well at a fore-vacuum of 100 μ.

3. Multistage pumps

Many works have been devoted to the description of various designs of single-stage mercury pumps by Gaede and Langmuir.* However, the circumstance that these pumps require relatively low fore-pressures forced one to follow the path of further development and, above all, the path of overcoming these difficulties.

It is well known that the steam ejector used in industry can reduce the pressure in a system by a fairly considerable amount. Langmuir pointed out that this action of the ejector is based on a principle different from the principle of operation of the condensation pump. “Gas is drawn into the ejector,” he writes, “by the jet because the pressure in it is lower than the pressure of the gas in the system being pumped out.” If one works with mercury vapor passing through a narrow nozzle, then, on the basis of this principle, it is quite possible to obtain pressures low enough for the operation of the pumps of the types described above. Hence follows the comparatively simple conclusion that both these stages can be combined in one pump.

In his article (1923)** Gaede described a pump, which is shown here in Fig. 10. This pump is made entirely of steel, and all its parts can be easily disassembled. Mercury vapor passes from the boiler to the central tube and condenses on the walls of the other tube surrounding it, cooled by a jacket with running water. The opening at the top connects the pump with the system being evacuated, while the ring filled with mercury separates, as indicated, the region of high vacuum from the fore-pump. Ex—

* A review of the newest designs is given in the article by Hickman and Sanford (K. Hickman and C. R. Sanford, Rev. of Sci. Instruments 2, 140, 1930).

** See the first footnote on p. 675.

high vacuum is obtained by means of a jet of mercury vapor passing through the inverted cup 1; nozzles 2 and 3 provide the preliminary vacuum.

Gaede finds that, when measuring the pumping speed near orifice 1 at a pressure of \(10^{-4}\) mm and at a fore-pressure of 20 mm, the speed obtained for air is 60 l/sec, and for hydrogen—100 l/sec. It must be said, however, that if the speed is measured at a time when some receiver is connected to the wide orifice (and it is precisely this that corresponds to the practical conditions of using the pump), the additional resistance of the tube between orifice 1 lowers the speed to 15,000 cm\(^3\)/sec for air.

Here one should also mention two papers by Molthan,* in which the theory of the action of slits and nozzles in Gaede’s diffusion pump is analyzed in great detail.

As in the case of single-stage pumps, a large number of different designs of two- and three-stage mercury condensation pumps have been described in the literature. It may perhaps be said that almost every laboratory has its own designs of condensation pumps, each of which in most cases has certain improvements over all other types of pumps. This remark is especially applicable to pumps made of quartz glass Pyrex (Corning G702P

Fig. 10. Gaede’s three-stage mercury pump.

Visible labels in the figure: “mercury enters through this opening”; “high vacuum”; “diameter 28 mm”; “speed 12000 cm\(^3\)/sec”; “preliminary vacuum”; “100 cm\(^3\) Hg.”

* W. Molthan, Z. techn. Phys. 7, 377, 452, 1926.

or of legs, which can withstand considerably higher temperatures of the boiler than was possible with lead or soda glasses.

A two-stage pump made of Pyrex glass was described by Kurth (E. K. Kurth). One variation of this design, devised by Ruggles (W. A. Ruggles) in our laboratory, is shown in Fig. 11. A great advantage of this pump is that its manufacture presents no special difficulties, and also that its pumping speed, reaching 3000–4000 cm³/sec, is sufficiently high for all laboratory purposes.

Fig. 11. Two-stage condensation glass pump.

The boiler with mercury, designated by the letter M, is heated, as is seen from the drawing, by a standard radiant furnace, and the mercury vapors pass upward through tube D, protected by some heat-insulating material J. After passing through small openings, the mercury condenses on the surrounding walls, cooled by the water jacket C, and acts here in the same way on the principle of a condensation pump; a higher fore-vacuum pressure is obtained at the nozzle opening B. The backing pump is connected at K, and the receiver—at F. Around the upper part of jacket C there is wound tape S, thereby ensuring flexibility of the joint.

G. K. Payne in our laboratory was for a long time interested in the question of the most expedient design of a high-speed metal pump, which is desirable for pumping out powerful vacuum tubes on an industrial scale, and also when working with large...

mercury arc rectifiers.* Table II gives the results of observations on various types of pumps obtained by De Groot and Ockley (C. T. De Groot and R. F. Ockley), who worked jointly with Payne. In the first column are given the names of pumps of various designs, while the other columns show the minimum pressures obtained with the aid of the pumps, and the preliminary rarefactions \(p_1\) required for good operation. This table also shows the energy expended on heating under optimum working conditions, and the resulting pumping speed \(S\). Since fluctuations in the pumping speed of these pumps may reach 50%, the quantities given in the last column should be regarded as a rough approximation to the mean values from the results of a large number of observations.

TABLE II

No.	Type of pump	Energy for heating	\(p_1\), mm	\(p_0\), \(\mu\)	\(S\), cm\(^3\)/sec
1	Langmuir (single-stage)	300	up to 0.10 0.01	up to 0.0005 0.0002	up to 5 000 6 000
2	Cleveland (two-stage)	365	up to 1.0 5.0	up to 0.0004 0.0008	up to 600 1 000
3	Cleveland (another form)	365	up to 2.5 8.0	up to 0.0004 0.0008	up to 3 000 4 000
4	Payne (two-stage)	550	up to 1.0 3.0	up to 0.0003 0.0007	up to 12 000 15 500
5	Payne (another form)	550	200	0.0003	60 000

Pump No. 1 is the original single-stage metal Langmuir pump. Pump No. 2 belongs to the type of pumps manufactured by the Lamp Department of the General Electric Company. By changing the design—

* For an illustration of this application see the article by Hull and Brown (A. W. Hull and H. D. Brown, A. I. E. E. Proc., Jan., 26—30, 1931).

nozzle considerably increased the pumping speed, as is shown in No. 3. Pump No. 4 represents the first construction of Payne’s pump, and pump No. 5 the same construction, but with a somewhat modified nozzle shape. The latter type of pump is shown in Fig. 12.

Fig. 12. Two-stage mercury condensation pump. Payne’s design.

The high-vacuum side is at the very top, while the fore-vacuum pump is connected to the opening of the spiral tube on the right-hand side. As may be seen from the next-to-last column of the table, the minimum pressures in these measurements reach about \(0.0002\)—\(0.0008\,\mu\). With the aid of metal pumps it is extremely difficult to obtain lower pressures, owing to the unavoidable slight “leaks” at the joints and the evolution of gases by the metal parts of the pumps. On the other hand, when using glass pumps, to which the receiver could be directly sealed on, Ruggles succeeded in obtaining pressures of the order of \(0.0001\,\mu\).

Of course, obtaining such low pressures was always possible only when extreme precautions were taken, chiefly for the purpose of removing all sources of vapors in the receiver. This result is achieved by preliminary heating of the tubes to the maximum temperature that the glass can withstand without softening. At the same time, in order to remove adsorbed water or other vapors from the walls, it was necessary to heat in a burner flame, or to treat by other methods, such as an induction coil, also those glass parts of the apparatus that lead from the tube to the liquid-air trap. In a high-vacuum system, stopcocks are entirely undesirable not only because they dimin—

...reduce the pumping speed, and also because of their tendency to produce leaks despite all precautions.

Not so long ago, when annealing evacuated apparatus, it was necessary to make narrow constrictions on the tubes, which greatly reduced the pumping speed. At present it has become entirely possible to seal off much wider constrictions; moreover, the internal diameter of such constrictions, in the case of large cathode lamps, now reaches 6 mm.

In the case of the metal parts used in tubes, it was formerly customary to bombard these parts by means of electrons from heated cathodes. In doing this, it was necessary to apply very high voltages to the anode, while the electron current, in order to keep the filament from being destroyed, had to be maintained at a comparatively small value, except for brief moments of intense bombardment. About ten years ago such treatment of metal parts—at least in the initial stages of pumping—had to be replaced by heating with high-frequency currents from a portable oscillator of 5–10 kW. With this method, not only is the removal of gases from the metal parts considerably faster, but the filament is also protected from the chemical action of water vapor and from the results of bombardment by positive ions—two reasons that, in the former pumping methods, often caused premature damage to the lamps. In addition, during bombardment at voltages of 50,000 V and higher, puncture of the glass or the formation of arcs between the electrodes often occurs, especially in cases where large quantities of gases are evolved. If treatment by means of high voltage is carried out at the stage when most of the gas has already been removed beforehand by high-frequency currents, these possible causes of lamp damage can also be avoided.

In one of the following sections, the question of the nature and quantities of the various gases in the metal parts of tubes, as well as further details concerning the removal of such gases, will be considered in greater detail.

In modern pumping methods, one more of their features deserves attention. With the earlier types of condensation pumps, which require a preliminary rarefaction of the order of 0.1 mm and even 0.01 mm, it was always necessary to use, as a preliminary pump, an oil pump, which in turn had to operate with another preliminary pump, giving a vacuum of 2 to 5 mm of mercury. The constant noise from these oil pumps, of course, did not in the least contribute to quiet work in a laboratory or in a factory. Thanks to modern two-stage pumps, operating at a preliminary rarefaction of 1–2 mm, it is now entirely possible to carry out the operation of a large number of these pumps on a central fore-vacuum, i.e. on a single common line, which in turn is connected with a large oil pump located at a considerable distance from the room for pumping apparatus.

Fig. 13. Diagram of a typical installation for pumping.

One modification of this method, which enjoys great attention in many factories, consists in placing a large reservoir between the mercury and oil pumps. In this case the oil pump can operate for a comparatively short time, necessary only in order to lower the pressure below 1 mm, after which this reservoir alone can already cope with the gas that is being pumped out by the mercury pump. An installation of this type is shown in Fig. 13.

The tube being evacuated, \(B\), is connected by means of a wide tube (internal diameter 25 mm), through the liquid-air trap \(L\), to the pump \(P\). Between the pump and the reservoir \(R\) there is inserted the condenser \(C\), shown enlarged in the upper left part of the figure, and the trap \(T_1\). The other two traps, \(T_2\) and \(T_3\), are necessary in order to protect the vacuum system and the fore-vacuum line from clogging by oil. The ionization manometer, also shown enlarged in the upper left part of the figure, is an essential component of the vacuum apparatus, serving for the determination of very low pressures, whereas the McLeod manometer, for determining the pressure on the rough side of the mercury pump, can be connected only between \(C\) and \(K\).

Condensation pumps with organic liquids

In 1928 Burch* at the Metropolitan-Vickers Electrical Company found that certain high-boiling petroleum products could be used with great success in condensation pumps. One such product, according to Burch’s description, has at \(118^\circ\)C a vapor pressure of about 1 bar† and, in addition, can be heated without decomposition to a temperature corresponding to a pressure of 100 bar. Using this liquid in a condensation pump, Burch obtained a vacuum of the order of \(10^{-4}\) bar, without making use of the liquid-air trap customarily employed in such cases. As a result of these investigations, the Metropolitan-Vickers firm constructed a metal pump in which these petroleum hydrocarbons were used instead of mercury.

The idea of replacing mercury by organic liquids was used by Hickman and Sanford** in the Kodak laboratories for the construction of small glass pumps in which, instead of the paraffin hydrocarbons prepared by Burch, \(n\)-dibutyl

* C. R. Burch, Nature 122, 729, 1928.
** Loc. cit.

ethyl phthalate or butyl benzyl ester of the same acid. Fig. 14 shows a comparatively simple form of a glass pump constructed for work with these liquids. Tube \(A\) leads to the receiver, \(C\)—to the pump, and \(B\)—to the preliminary-vacuum pump. Since the vapors of the organic liquid leave the jet at temperatures \(35\text{–}100^\circ\) higher than in the case of mercury vapor, quite moderate cooling is sufficient; thus, for tubes \(1.5\ \mathrm{cm}\) in diameter or less, Hickman and Sanford find “that copper wire wound in a spiral around the tube produces excellent

Fig. 14. Glass pump (Hickman design) with organic liquid.

Fig. 15. Hickman condensation pump with organic liquid.

cooling.” A photograph of a pump furnished with such a spiral with loops is shown in Fig. 15.

Table III, borrowed from the article of these investigators, contains the operating characteristics of pumps containing various organic vapors, and, for comparison with them—

results obtained with mercury. The last two columns show the energy for heating and the optimum speed under satisfactory operation of the pumps.

TABLE III

Liquid used in the pump	Vapor pressure at 25°C (mm Hg)	Vapor pressure at 0°C (mm Hg)	Immobile air: heating energy (W)	Immobile air: optimum speed (cm³/sec)	Immobile air: pressure for capillary pump speed (in mm)	With fan: heating energy (W)	With fan: optimum speed (cm³/sec)	With fan: pressure for capillary pump speed (in mm)	Conditions under which changing the cooling in the pump has no effect on pump operation: heating energy (W)	Conditions under which changing the cooling in the pump has no effect on pump operation: optimum speed
Mercury	\(2{,}3 \cdot 10^{-8}\)	\(2{,}4 \cdot 10^{-4}\)	40	7 000	0,15	60	8 000	0,20	45	6 500
Medical paraffin (middle fraction)	\(10^{-5}\)	\(10^{-6}\)	35	2 000	0,07	50	4 000	0,10	—	—
\(N\)-dibutyl phthalate	\(7{,}8 \cdot 10^{-5}\)	\(9{,}5 \cdot 10^{-6}\)	32	6 500	0,07	45	7 000	0,09	35	6 000
Butylbenzyl phthalate	\(6{,}2 \cdot 10^{-6}\)	\(2{,}6 \cdot 10^{-7}\)	50	9 500	0,07	70	10 000	0,08	55	9 000
Fractionated oil from a rotary pump	\(2{,}9\text{--}4{,}0 \cdot 10^{-6}\)	\(1{,}7\text{--}2{,}1 \cdot 10^{-7}\)			0,6			0,07	65	4 500

Hickman and Sanford draw the following conclusions concerning the advantages of using phthalic-acid esters in condensation pumps.

“Phthalic-acid esters prove to be very advantageous owing to their good pumping speed, constancy of action, the ease with which they can be obtained in a pure state, and the low elasticity of their vapor. Of these esters, butylbenzyl ether is the most important member because of its high boiling point, but it is not easy to ob-

obtain. Therefore, it is necessary to recommend n-dibutyl ether, which, for use in diffusion pumps, can quite satisfactorily replace mercury.

For many purposes no cooling trap at all is required; in other cases a mixture of salt with ice or a mixture of solid carbon dioxide with acetone gives quite sufficient cooling. The optimum operating conditions for pumps with phthalic-acid esters are now well known, and, provided only that these conditions are observed, small pumps filled with these materials can function perfectly without any supervision.

Thus condensation pumps with organic liquids, such as those described above, should probably find application in many practical cases, especially where liquid air is not available.

Substitutes for liquid air

In this connection one should mention the investigations of Hughes and Poindexter* on the possibility of using sodium and potassium as traps for mercury and water vapors, and also for carbon dioxide, whereby the necessity of using liquid air is eliminated. The vacuum which can be obtained by this method, insofar as this can be judged from the readings of an ionization manometer, is practically the same as the vacuum obtained with the aid of liquid air.

Whereas the latter is unquestionably the most effective means for removing condensable vapors, solid carbon dioxide (“dry ice”) proves to be an excellent means for condensing mercury vapors, since the vapor pressure of the latter at −78°C (the temperature at which solid carbon dioxide has a vapor pressure of 1 atm) is \(3 \cdot 10^{-6}\ \mu\). Since, however, at this temperature the vapor pressure of water is approximately \(1\ \mu\), under these conditions it is absolutely necessary to heat—

* A. L. Hughes and F. E. Poindexter, Phil. Mag. 1, 423, 1925; F. E. Poindexter, Journ. Opt. Soc. Am. 9, 629, 1924.

...to heat the entire system, including the condensation pump, at least to a temperature of 360° abs. If this proves impossible, then the rate at which the pressure in the system decreases will be limited by the rate at which water vapor can evaporate from the walls of the connecting tubes; as a result, as the author has had occasion to observe, a very considerable time may be required to attain a very good vacuum.

In lamp practice, it is customary to use \(P_2O_5\) for removing water vapor. Comparatively recently it has been possible to obtain other dehydrating substances,* which are just as effective as phosphoric anhydride and, in addition, more convenient to handle.

Of course, all these substances should find useful application, for example in cases of such investigations where it is necessary to keep the vacuum system completely dry for a prolonged time.

The Use of Getters in Vacuum Technology

By a getter is meant a certain reagent which is placed in a vacuum apparatus in order to improve the vacuum after this apparatus has been disconnected from the pump. The first such getter to enter technical practice was phosphorus, proposed by Malignani in 1894 for use in lamps with a carbon filament. Subsequently Soddy (F. Soddy) proposed, as a method for removing residual gases, the use of calcium evaporation, since in this way it is possible to reduce pressures to values considerably lower than those given by oil pumps.

The broad development of the production of cathode lamps for radio would have been impossible had getters such as magnesium, calcium, and other reagents not begun to be used; these not only consume the residual gases present in

* For example, anhydrous magnesium perchlorate (anhydrone) and anhydrous barium perchlorate. See Yoe, Mc. Gahay and Smith, Journ. Ind. Eng. Chem. 20, 656, 1928; Smith, ibid. 19, 411, 1927; Yoe, Chem. News 130, 340, 1924.

lamps after evacuation, but also maintain the vacuum at the level required for good operation of the lamps throughout their entire lifetime. In devices containing incandescent cathodes, residual gases can cause harm in two ways. First, if the pressure is sufficiently high, so that the mean free path of the electrons has the same order of magnitude as the distance between anode and cathode, the gas molecules are ionized by collisions with the electrons. As a result there is a partial or even complete destruction of the space charge, making it impossible to regulate the amount of space current by means of the grid. Secondly, the electron emission from the cathode is reduced either owing to the formation of adsorbed monatomic layers of gas, or owing to positive-ion bombardment of the cathode surface and the subsequent removal (by sputtering) of the material that is active for electron emission.*

In order to eliminate these destructive effects, generally speaking it is very important that the pressure in a cathode tube have an extremely low value. For nitrogen at a temperature of 25° C, the mean free path of electrons at a pressure of \(10^{-3}\) mm is equal to 42.5 cm, and consequently the collision frequency over 1 cm of path will be about 0.002. Assuming that ionization occurs as a result of each collision, we see that one positive ion is formed over 1 cm of path by 425 electrons. Since, however, one positive nitrogen ion neutralizes the space effect produced by 227 electrons, it follows that the residual pressure under these conditions must be at least 10 times lower, i.e. it must be about \(10^{-5}\) mm, in order to prevent the destruction of the space charge.

Electron emission from thoriated tungsten is extraordinarily sensitive to traces of residual oxygen;

* For a detailed consideration of the question of the influence of gases on emission, see the author’s review: “Thermionic Emissions,” Rev. of Modern Phys. 2, 381, 1930.

thereby it is very easy to show that above \(2\cdot 10^{-6}\) mm of pressure only 1 sec. of time is required in order completely to cover a thorium film with oxygen molecules. Thus, in order to obtain a lifetime of several thousand hours, which is normal for such filaments, the pressure of the residual oxygen must be less than \(10^{-12}\) mm of mercury, or approximately \(10^{-15}\) atm.

The theory of the action of various getters in destroying residual gases has never been sufficiently clear. In some cases this action may be regarded as purely chemical: between the volatilizing metal and the residual gases chemical compounds are formed. In other cases the disappearance of the gas is strongly promoted by the presence of a potential between the anode and cathode. It must be supposed that the gas molecules are ionized or are brought into an active state and under such conditions react more readily with the getter. As a result, the getter deposited on the walls of the bulb can extract residual gases by means of repeated adsorption. Gas molecules striking a clean metallic surface form on it a monomolecular adsorbed gas layer, which, owing to the extremely small rate of reverse evaporation, is stable over a very wide range of temperatures. Thus it is necessary to distinguish three different types of gas disappearance: 1) chemical, 2) electrochemical, and 3) adsorption.

Reactions between tungsten at high temperatures and residual oxygen and nitrogen are characteristic reactions of the chemical type. In his investigations of electron emission from thoriated tungsten, Langmuir obtained the necessary high vacuum by means of a tungsten filament heated to temperatures of about \(2900^\circ\mathrm{K}\), the bulb being immersed in liquid air. Subsequent investigations with an ionization manometer showed that the residual pressure obtained by this method is \(10^{-4}\) or fewer bar.* Oxygen reacts with

* The details of these phenomena are described in Langmuir’s papers: Journ. Am. Chem. Soc. 37, 1139, 1915; Journ. Ind. Eng. Chem. 1, 345, 1915.

with tungsten on the surface, and the resulting WO₃ volatilizes and condenses on the walls. Nitrogen and carbon monoxide react with tungsten atoms in the gaseous phase, forming the compounds WN₂ and WCO, while hydrogen dissociates and is retained by the walls in the atomic state.

The removal of gases at high pressures with the aid of calcium, by Soddy’s method, is a reaction of the same type as the reaction between tungsten and nitrogen; but at low pressures of the residual gases the latter are removed either by adsorption on a metallic film condensing on the walls, or by an electrochemical process.

At a pressure of \(10^{-8}\) mm and at a temperature of \(25^\circ\)C the mean free path for a nitrogen molecule is 7.5 cm, i.e. it is of the same order as the dimensions of common vacuum apparatus. Consequently, most molecules fly directly to the walls of the bulbs, and the frequency of collisions in space becomes small in comparison with the speed with which the molecules strike the walls. And indeed, a simple calculation shows that the ratio between the rate of incidence of molecules on unit area and the collision frequency per unit volume is proportional to \(L/D\), where \(D\) denotes the distance between the walls, and \(L\) is the mean free path. Thus, at pressures below \(10^{-8}\) mm, the rate of disappearance due to collisions between molecules in space becomes extremely small, and the observed disappearance of the gas, generally speaking, must be ascribed to adsorption of the residual gas molecules by the deposit on the glass walls.

Thus, while adsorption constitutes the most effective method for improving the vacuum at small initial pressures, electrochemical reactions also play a certain role. In the manufacture of cathode lamps this type of reaction is known as “high-voltage training”; it is very widely used to improve the vacuum immediately after the lamps have been sealed off. To reduce the pressure to a certain extent by means of bombardment

... bombardment by electrons of comparatively high velocity (100–200 V) is quite possible even without any getter. That the ions obtained in this way, owing to collisions between electrons and gas molecules, fly to the negatively charged walls seems altogether indisputable. The removal of gases in incandescent lamps with the aid of phosphorus was investigated both by the author with collaborators in the laboratories of the General Electric Company in America, and by N. R. Campbell in the laboratories of the same company in England.*

In this field it has been possible to accumulate a very large number of observations; however, the explanations of these phenomena, generally speaking, still remain controversial.

Although activated charcoal at liquid-air temperatures is very effective from the point of view of reducing gas pressure, its use in apparatus evacuated by pumps is wholly impractical. In this connection it is interesting to note that in some of the most recent experiments, which at the author’s suggestion were carried out by Becom (U. S. Bacom), the pressure in an ionization manometer connected to a tube containing well-degassed charcoal and immersed in liquid air was less than \(5\cdot 10^{-9}\) mm. This value represents the highest vacuum that we are able to measure. The initial pressure during pumping in this case was about \(10^{-5}\) mm.

In vacuum apparatus of industrial type it is necessary to use a getter that must have negligible vapor pressure and at the same time must form an active surface on the walls of the bulb for condensation of residual gases. As was already mentioned above, magnesium, calcium, and similar metals are commonly used as getters. A small piece of metal is fastened to the outer side of the anode, and after heating the tube to remove water vapor and carbon dioxide,

* In addition to the previous references, see the articles by Campbell: N. R. Campbell, Phil. Mag. 40, 555, 1920; 41, 685, 1921; 42, 227, 1921; 43, 914, 1922; 48, 553, 1924; Trans. Am. Electrochem. Soc. 44, 87, 1923.

adsorbed on the glass walls, the getter is made to volatilize by heating the anode with a high-frequency current. After this, pumping is continued for some time, necessary for removing the gases given off by the heated metal, after which the tube is sealed off. During conditioning and throughout the subsequent life of the lamp, the pressure of the residual gases rapidly decreases and in the end reaches values of \(10^{-5}\)—\(10^{-6}\) mm of mercury.

Fig. 16. Bulb with an ionization manometer for testing getters (Endryus).

In certain experiments, continued by Kidner (C. A. Kidner) and by the author in 1921,* it was possible to observe that, when calcium was used, the hydrogen pressure in a bulb of \(500\ \mathrm{cm}^3\) capacity decreased within a few minutes from \(1.5\cdot 10^{-3}\) to \(5\cdot 10^{-4}\) mm, and within 2 hours—to \(5\cdot 10^{-7}\) mm. The disappearance of hydrogen was studied chiefly because, in comparison with oxygen, nitrogen, and carbon monoxide, its removal is the most difficult matter.

Most recently, Endryus and Bacon in our laboratory have carried out systematic investigations of the behavior of various gases in tubes sealed off from the pump. Here a 40-watt lamp bulb \(A\) (Fig. 16), containing a getter, was connected with an ionization manometer \(B\) of the type which several years ago was developed by Found and the author.* The anode and the conducting wires of the manometer were made of molybdenum, the filament was tungsten; all metallic parts were completely

* Dushman, High Vacuum, 1922.

** Details of this work will be published shortly.

* Phys. Rev. 17, 7, 1921; 23**, 734, 1924.

dehydrated under simultaneous pumping. Between the pump and the system with the getter there was a liquid-air trap. The getter was slowly introduced in the form of vapors before pumping.

Fig. 17 shows the results obtained with magnesium sputtered by various methods. The upper

Fig. 17. Effect of magnesium on vacuum.

curve shows the change of pressure with time for a bulb without a getter, pumped under the same conditions as the bulb with a getter. The two adjacent curves were obtained for magnesium sputtered with a tungsten spiral without preliminary dehydration. The curve lying between \(0.01\) and \(0.1\ \mu\) represents the result of sputtering a magnesium sample dehydrated by means of a spiral which was heated for 15 min., after which the metal was sputtered. The three lowest

of these curves were obtained with magnesium in molybdenum envelopes, and before melting, thorough degassing was carried out.

The same difference between ordinary and degassed metals is illustrated by the curves in Fig. 18 for calcium. The two upper curves were obtained for metal that had been degassed for 15 min. in a spiral, while the three lower curves

Figure labels visible in the graph: Borax; Calcium; Weeks.

Fig. 18. Effect of calcium on mercury.

were obtained for metal that was evaporated over the course of 2 hours.

The behavior of sodium is of very great interest, as illustrated by Fig. 19. The three upper curves were obtained with pellets of \(\mathrm{NaCl}+\mathrm{Cl}\) in a metallic envelope. When they are heated to a high temperature, sodium is formed, which then condenses on the walls of the flask. The two lowest curves were obtained with metal,

deposited on the walls of the flask by the electrolysis of glass in the usual way. Obviously, in experiments of the latter kind there was much more metal free from gas, since the vacuum obtained was considerably better (0.0001–0.001 μ) than with any other getter.

The series in which getters could be arranged according to their ability to absorb residual gases generally corresponds to chemical activity. Thus, for example, caesium is the most active metal, but its comparatively high vapor pressure and its strong tendency to form positive ions prevent its wide application. Sodium in this respect occupies the neighboring place, but even the fact that the metal has a considerable vapor pressure at a temperature of 200° C is already a reason for disregarding it. Of the alkaline-earth metals, barium is the most active, but it is difficult to make it suitable—

Fig. 19. Effect of sodium on vacuum.

Advances in the Physical Sciences, Vol. XI, issue 5.

applicable because of its easy oxidizability. This difficulty can be overcome to a certain extent if it is used in the form of an alloy with aluminum. On the other hand, calcium does not oxidize so readily and at the same time is just as active as barium. Magnesium reduces oxygen to pressures quite sufficient for the operation of a thoriated filament, but, in comparison with other getters, it is not so active with respect to hydrogen and nitrogen. However, its stability with respect to the oxygen of the air and the comparative ease with which it can be attached to the anode have led to its wide use in the manufacture of cathode lamps.

As another material for gettering, one may recommend an alloy obtained in the production of rare earths and containing 40% cerium; in addition, this alloy contains lanthanum (the greatest amount) and other rare-earth metals. As will be indicated in one of the following paragraphs, the rare earths strongly absorb hydrogen at room temperature, and when heated to a high temperature they react vigorously with oxygen and nitrogen. Whether, however, this material will be as active or as convenient as the alkaline earths cannot yet be said, since at present there are no data on this question in the literature.

Gases in Metals and the Treatment of the Latter in Vacuum

As was indicated above, the development of vacuum apparatus has led to recognition of the important role played by the removal of gases from metallic parts. In some cases this degassing can be carried out by heating with high-frequency currents during the pumping itself, but it often proves impossible or impractical to use this method because the construction of the tubes is unsuitable for this purpose. In such cases, preliminary treatment of the metals in a vacuum furnace, or the use of metals melted in vacuum, is of enormous help.

A review of the extensive literature on the question of gases in metals shows that gases may be present either in the state of solution in the metal or in the form of stable chemical compounds. In addition, a certain quantity of gases is always present on the surface of metals in an adsorbed form. These adsorbed gases, as well as dissolved gases, are easily removed by heating below the melting point, but gases present in the form of oxides or nitrides are considerably more difficult to remove. In the latter case, the only truly effective method proves to be melting in vacuum.

The very important role played by gases in metals, exerting an influence on their physical properties, has also been recognized by metallurgists; moreover, a large part of the investigations on this question falls within the last several years. This problem has been studied chiefly from the standpoint of the origin of gases in metals, with some investigators interested in the solubility and diffusion of gases in metals. Since the conclusions obtained in this work prove to be important also from the standpoint of producing vacuum, it would be of interest to trace briefly the achievements in this field.

Quantity and composition of gases in metals

In connection with his investigations of lamps with tungsten filaments, Langmuir developed a method for analyzing small quantities of gases contained in metals and glass. He found* that, if care is taken to remove water vapor and carbon dioxide from the walls of the glass, the quantity of gas released from a tungsten filament does not exceed ten times the volume of the filament itself. The largest part of the gas is released when the filament is heated to a temperature of 1500° C. The gas consists approximately of 70–80% carbon monoxide; the remaining part consists chiefly of hydrogen and carbon dioxide.

* Langmuir, Journ. Am. Chem. Soc. 35, 105, 1912.
* Langmuir, Trans. A. I. E. E. 32*, 1921, 1913.

Using the very same method, Switzer, in his laboratory, determines the quantity and composition of gases released by wires made of various metals, which are used in incandescent lamps. The metals were taken in the form of filaments having a total volume of about \(0.03\ \mathrm{cm}^3\), so that they could be heated to incandescence by passing an electric current through them. Various samples of so-called untreated* nickel wire give quantities of gases from \(0.005\) to \(0.015\ \mathrm{cm}^3\), consisting of 75–90% carbon monoxide and 10–20% carbon dioxide, with small amounts of hydrogen. Wires similar to these made of German silver, copper, and copper-coated ferro-nickel alloy (which is used for leads in the glass stems of lamps) give quantities of gases from \(0.003\) to \(0.02\ \mathrm{cm}^3\). The composition of these gases is approximately the same as in the case of nickel wires. In all these cases the volume of gas does not exceed the volume of the metal; but since even such small amounts of gas as \(0.010\ \mathrm{cm}^3\) in a bulb of \(500\ \mathrm{cm}^3\) capacity are sufficient to raise the pressure to \(15.2 \cdot 10^{-3}\ \mathrm{mm}\), the importance of removing these gases before sealing-off is quite obvious.

Ryder somewhat modified Langmuir’s method and determined the nature and composition of the gases released by untreated commercial copper when heated in a vacuum.** The following vapors and gases are released here: carbon dioxide, carbon monoxide, water, and nitrogen. For a sample having a volume of \(1.31\ \mathrm{cm}^3\), the total amount of gas released at a temperature of \(750^\circ\mathrm{C}\) was about \(0.2\ \mathrm{cm}^3\).

Gases in steel were investigated by Alleman and Darlington.*** Some typical results obtained in these investigations are shown in Table IV.

* See the book by Dushman mentioned above.
* H. N. Ryder, Journ. Am. Chem. Soc. 40, 1656, 1918; Journ. Franklin Inst. 187, 508, 1919.
* Alleman and Darlington, Journ. Franklin Inst. 185*, 161, 333, 461, 1918. This article contains a detailed survey of earlier work on gases in iron alloys.

TABLE IV

Gases released from steel

No.	Volume of gas per 1 g of metal (in cm³)	Volume of gas per 1 cm³ of metal (in cm³)	Mass, temperature (°C)	CO₂	O₂	CO	H₂	N₂
1	25.2	197	1 468	0.13	2.08	59.8	18.18	19.81
2	18.6	146	1 500	1.20	0	79.8	11.65	7.35
3	8.5	67	1 100	0.68	1.57	26.15	43.40	28.50

Analysis of these three metal samples gave the following figures:

Sample	1 and 2	3
Percent carbon	1.049	0.084
silicon	0.153	0.005
phosphorus	0.028	0.094
sulfur	0.028	0.082
manganese	0.405	0.536

Table V shows the results of analysis of gases obtained from Bessemer steel (0.1% carbon) at different temperatures. The total quantity of gases released was 28.1 cm³ per 1 g, or 220 cm³ per 1 cm³ of metal.

Observations show that hydrogen and carbon dioxide are released at a temperature of 1,250°C and below, and that much higher temperatures are required for the removal of nitrogen and oxygen. Another interesting feature of these results consists in the comparatively large volume of gases released in comparison with tungsten, nickel, and copper, which were discussed above.

TABLE V

Gas	At 1,000°C	At 1,250°C	At 1,500°C	At 1,675°C
CO₂	1.08	0.62	0.00	0.00
O₂	2.40	3.07	4.28	6.25
CO	48.90	56.10	18.75	8.42
H₂	21.16	15.08	4.20	1.10
N₂	26.46	25.13	72.77	84.23

Large-scale production and the properties of metals melted in vacuum are analyzed in detail in Rohn’s article.* He indicates that from 1 kg of a metal melted in vacuum—for example iron or nickel—one usually obtains from 2 to 3 l of gas (measured at atmospheric pressure), but sometimes the amount of gas reaches 15 l. This corresponds to a volume of gas exceeding the volume of the metal by 16–120 times. The gas consists for the most part of hydrogen and carbon oxides, whose formation must be ascribed to chemical reactions that are not completed during the production of the metal and that continue while the metal is heated to a high temperature in vacuum. Thus, for example, carbon monoxide and carbon dioxide arise as a result of the reaction between carbon in the metal and residual oxides, and hydrogen as a result of the interaction between water vapor and metals.

As for the temperatures at which the evolution of gases occurs, Rohn’s observations on chromium–nickel alloys show that the largest part of the gas is evolved after the melting of the metal and that the gas then contains a comparatively large percentage of carbon monoxide and small quantities of nitrogen, oxygen, and hydrogen. However, the greater part of the hydrogen is evolved at a temperature below 1000°C.

Incidentally, Rohn points to the rapid progress that has been made in increasing the maximum size of ingots that can be obtained by melting in vacuum. The weights of the maximum chromium–nickel ingots melted and cast in vacuum in 1922, 1924, and 1926 were, respectively, 40, 230, and 1400 kg. No definite data are available concerning the degree of rarefaction at which this process is carried out, but the diagrams in the original work indicate that the pressure during melting was approximately 10 mm.

It is quite natural that the problem of developing exact methods for melting metals in vacuum and determining the nature

* W. Rohn, Z. Metallkunde 21, 12, 1929.

and the composition of the gases released in this process began to be used with very great attention. Since a review of these methods does not fall within the scope of this article, it will be sufficient here to mention some results obtained by various investigators. A very detailed review of the various methods was written by Hessenbrucher.* Tables VI and VII show the results obtained by him for copper and nickel.

TABLE VI

Gases in copper treated at a temperature of 1250°C

Sample	Gas volume in cm³ per 100 g of metal	Gas volume per 1 cm³ of metal	CO₂	H₂	CO	N₂
Pure copper (1)	2,72	0,24	58,2	7,90	19,50	14,40
Pure copper (2)	1,46	0,13	85,7	7,43	—	6,87
Pure copper (3)	1,79	0,16	33,9	28,55	37,55	—
Pure copper (4)	0,78	0,07	61,1	11,11	22,22	5,57
Electrolytic copper	7,97	0,71	10,9	39,52	49,70	—

TABLE VII

Gases in nickel treated at a temperature of 1470°C

Sample	Gas volume in cm³ per 100 g of metal	Gas volume per 1 cm³ of metal	CO₂	CO	H₂	N₂
Nickel single crystal	482	42,9	2,30	90,0	3,50	4,20
Purified nickel	113	10,1	2,92	72,4	24,70
Electrolytic nickel	788	0,7	0,00	21,1	78,9

Gas samples (1) and (3) in Table VI were taken at the beginning of casting, while samples (2) and (4) were taken at the end. Sulfurous gas, as was indicated by Schulze** in the discussion of the causes of the occurrence of gases in metals, apparently arises as a result of the reaction

\[ \mathrm{Cu_2S + 2Cu_2O = SO_2 + 6Cu.} \]

* W. Hessenbrucher, Z. Metallkunde 21, 46, 1929.
* E. H. Schulze, Z. Metallkunde 21*, 7, 1929.

In the case of electrolytic nickel, a large amount of hydrogen must arise from gases occluded during deposition.

Recently, Villachon and Chaudron* reported the results of an investigation of the gases evolved by metals which had previously been melted in a vacuum furnace, in which it was possible to maintain a pressure of 0.02 mm at a temperature of 1700°C and 0.002 mm at 1000°C. The metals were melted in magnesite crucibles; the metals were then rolled into thin sheets and heated in tubes of quartz or Pyrex glass to a temperature of 800°C in the case of nickel and 600°C in the case of copper and iron. The gases evolved consisted of hydrogen and carbon monoxide, the total amount of gas being approximately equal to the volume of the metal.

Fig. 20. Flask with a molybdenum cup for determining gases in metals.

For many purposes, information on the exact composition of the gas is not so important as information on the total volume of gas that can be evolved upon heating to a temperature near the melting point. On the basis of such data it is possible to compare the effects of different methods of preliminary treatment of metals. To obtain

* A. Villachon et G. Chaudron, Revue de Métallurgie, 27, 363, 1930.

of such data, Andrews and her students at the Massachusetts Institute of Technology, during the summer vacation, developed the apparatus shown in Figs. 20 and 21.

The first of these figures shows a 7-inch flask containing a molybdenum cup for heating samples of metal. The gases liberated by the latter are pumped by means of a condensation pump into a reservoir, which is connected with a McLeod manometer. In this way the total amount of gas evolved is measured. By means of a trap with liquid air under the reservoir one can also determine the amount of gas that does not condense at the temperature of liquid air. It consists of hydrogen, nitrogen, and carbon monoxide, while the condensing gases are water vapor and carbonic acid.

Fig. 21. Schematic drawing of an apparatus for determining gases in metals.

The analysis is carried out as follows. The sample is placed in tube \(B\). The entire system is heated for an hour at a temperature of \(450^\circ\text{C}\), and the molybdenum cup, for its degassing, is heated by means of high-frequency currents.

Then the sample is shaken down into the cup by lightly tapping tube \(B\), and the reservoir for collecting the evolved gases is disconnected from the high-vacuum pump. Thereafter heating by means of high frequency is begun again, the temperature of the cup being determined with an optical pyrometer and continued until the time when

while the evolution of gases at the maximum temperature is rendered entirely negligible.

Table VIII shows some typical results obtained with the aid of this method. Since the densities of copper and nickel are almost identical (8, 9), \(100\ \text{cm}^3\) per \(1\ \text{kg}\) of metal corresponds to 0.89 volume of gas per volume of metal. Experience shows that the volume of gas rarely exceeds the volume of the metal; moreover, during treatment at high temperatures in vacuum the gas content may be reduced to 10% of the volume of the metal and even more.

TABLE VIII

Material	Noncondensed gases in \(\text{cm}^3/\text{kg}\)	Condensed gases in \(\text{cm}^3/\text{kg}\)
Nickel wire 0.7 mm, annealed in hydrogen	49.0	3.1
Nickel wire 0.7 mm, annealed in hydrogen	7.0	3.9
Nickel foil 0.176 mm	60.2	8.2
The same, treated in vacuum (1 hour at a temperature of \(1000^\circ\text{C}\))	40.5	5.5
Molybdenum mesh	52.9	8.3
The same, treated in hydrogen (15 min. at a temperature of \(1150^\circ\text{C}\))	10.8	4.0
The same, heated in vacuum (1 hour at a temperature of \(1050^\circ\text{C}\))	9.8	3.7
Nickel wire, annealed in hydrogen, drawn through a diamond	96.0	6.5
Pure copper; softened	12.1	2.5
Pure copper; softened	6.3	2.1
Pure copper; softened	2.5	1.6
Copper, annealed to a temperature of \(1600^\circ\text{C}\) in vacuum	22.0	—
Copper, annealed to a temperature of \(1600^\circ\text{C}\) in vacuum	6.3	3.6
Copper, boiled in vacuum in a molybdenum cup	5.4	3.3

Some time ago Marshall (A. Z. Marshall) in our laboratory was engaged in degassing copper and other metals by boiling them in molybdenum tubes in an extraordinarily good vacuum. A complete description of the ap-

the apparatus and method of analysis will be given by them in the near future; here, in Table IX, only some of the results obtained by them are presented. In the case of copper specimens, “surface gas” corresponds to the gas evolved at temperatures of about \(1\,000^\circ\mathrm{C}\) and below, with the greater part of the internal gas being evolved during the volatilization of the metal.

TABLE IX

Nature of the specimen	Volume of gas in \(\mathrm{mm}^3\) per \(1\ \mathrm{cm}^3\) of metal	Volume of gas in \(\mathrm{mm}^3\) per \(1\ \mathrm{cm}^3\) of metal
	Surface gas	Internal gas
1. Electrolytic copper	5.1	71.0
2. Commercial copper	6.7	40.0
3. Copper melted in vacuum	9.1	86.8
4. The same	6.9	52.0

The gases evolved consist mainly of carbon monoxide, with small quantities of hydrogen and carbon dioxide, except in the case of electrolytic copper, where the greater part of the surface gas consists of hydrogen.

For molybdenum specimens calcined to a temperature of \(2\,000^\circ\mathrm{C}\), the volume of gas evolved varies from 37 to \(75\ \mathrm{mm}^3/\mathrm{cm}^3\), the gas consisting of carbon monoxide and nitrogen. It is interesting to note that specimens treated in vacuum absorb very little gas at atmospheric pressure, provided only that they have not been touched with the fingers. In the latter case, upon heating to a temperature of \(1\,600^\circ\mathrm{C}\), large quantities of carbon monoxide are rapidly evolved.

Sorption of gases by metals

The term sorption was proposed by McBain,* and by it is meant a whole series of related phenomena which often occur simultaneously and which include all cases in which the pressure of a gas decreases owing to metals, charcoal, or other

* J. W. Mc. Bain, Phil. Mag. 18, 916, 1909; Z. Phys. Chem. 68, 471, 1909.

substances. Gases can condense on the surface, and such a phenomenon is called adsorption, or they can penetrate into the metal, forming a solution with the metal as the solvent. The latter process is known under the name absorption. In addition, gases can combine chemically with the metal, as occurs in the case of the sorption of oxygen and hydrogen by alkali metals. The term sorption applies to all these phenomena and quite aptly denotes all cases in which the exact mechanism of the capture of a gas by a metal is unknown.

The phenomena of adsorption are extremely important from the standpoint of obtaining very low pressures. As was mentioned above, the behavior of most getters at low pressures is explained on the basis of ideas concerning the surface condensation of residual gases. Langmuir* was the first to show that in all cases of true adsorption the condensed gas forms a monomolecular layer. Considering adsorption as the result of kinetic equilibrium between the rates of condensation and evaporation, Langmuir derives a simple relation between the pressure and the amount of adsorbed gas, and this relation proves to be in good agreement with observations. In this connection the mean life of an adsorbed molecule, i.e. a quantity reciprocal to the rate of evaporation, plays an extremely important role. Thus, for example, for oxygen adsorbed on mica, the lifetime at 90°K is 90,000 times greater than the lifetime of molecules on the surface of liquid oxygen at the same temperature, and in all cases of adsorption the relative lifetimes of adsorbed atoms are thousands and millions of times greater than for liquids at the same temperatures.

A comparatively simple calculation shows that, by means of adsorption alone, it is theoretically possible to obtain a considerable decrease in pressure. If one takes a spherical flask of radius 5 cm and assumes that the molecules

* I. Langmuir, Journ. Am. Chem. Soc. 40, 1361, 1918.

if they have a diameter of \(3\cdot 10^{-8}\) cm, then the number of molecules required to form a monomolecular layer on the inner wall will be \(3.49\cdot 10^{17}\), which corresponds to the number of molecules in the flask at a pressure of \(93\ \mu\). Consequently, this represents the maximum decrease in pressure, provided, of course, that we have an absolutely clean surface before admitting the gas into the flask. In this way the effects arising when alkali and alkaline-earth metals are used are very simply explained.

Langmuir arrived at his conclusions concerning the presence of adsorbed layers on the basis of studies with surface films on liquids, and also on the basis of the chemical and physical behavior of metals in the presence of gases at low pressures; but recently an experimental method (of an entirely new type) has been developed for the study of surface films. Electrons are reflected at the surface, and therefore the study of the characteristics of the reflected electrons can serve for investigating the nature and arrangement of adsorbed atoms on the surface.

That electrons incident on the surface of a metal behave in all respects like waves, with the wavelength here inversely proportional to their velocity, was shown by the investigations of Davisson and Germer, J. P. Thomson, and E. Rupp. Thus electron rays, like X-rays, can be used to study the structure of crystals. It turned out, however, that the most valuable application of this new discovery consists in the investigation of the structure of surface films.

The principle of the method is roughly as follows. A beam of electrons of variable velocity is directed at the surface under study at a definite angle of incidence. What is measured is the intensity of the reflected beam as a function of the accelerating voltage for the electrons. At certain definite voltages the intensity of the reflected beam reaches certain maximum values, and from the magnitude of the voltage, with the aid of

ratio, analogous to Bragg’s original equation for determining the structure of crystals, one can calculate the distance between atoms in the surface layer.

In this way Davisson and Germer* and Rupp** studied the arrangement of gas atoms adsorbed on the surface of nickel and iron. It is obvious that such a method may prove quite applicable to a detailed study of the action of many getters used in vacuum technology.

The absorption of gases by metals has now been investigated very thoroughly, especially thanks to the work of Sieverts*** in Jena. A typical phenomenon of this class is the dissolution of hydrogen in palladium, platinum, and tantalum, and also of oxygen in silver.

One of the most interesting consequences revealed by investigations in this field is that the amount of gas dissolved at constant temperature is proportional to the square root of the pressure. This is confirmed for oxygen in molten silver and for hydrogen in practically all metals with which it forms solutions.

The behavior of hydrogen with various metals is of particularly great importance. Its solubility in silver, copper, nickel, cobalt, and iron increases with temperature both for solid and for molten metals; moreover, near the melting point a rapid increase of this solubility is observed. Fig. 22 shows isobars (solubility curves at atmospheric pressure) for copper, nickel, and iron. Along the ordinates are plotted the amounts of absorbed hydrogen (in milligrams) per 100 g of metal. As will be shown below, hydrogen readily diffuses through these metals. The amount of dissolved—

* Davisson and Germer, Phys. Rev. 31, 307, 1928.

** E. Rupp, Z. Elektrochem. 29, 586, 1929; Ann. d. Phys. 5, 453, 1930. Transl.

*** Sieverts, Z. Metallkunde, 21, 37, 1929. This article contains a review of results obtained with various metals and gases.

gen at constant temperature is proportional to \(p^{1/2}\).

The unusually high solubility of hydrogen in palladium, especially when it is in a highly dispersed state (palladium black), has been known for a long time. At atmospheric pressure and a temperature of \(20^\circ\text{C}\), this metal can absorb from 670 to 800 volumes of hydrogen. With increasing temperature the solubility decreases; however, at very low temperatures the solubility increases considerably.

With other metals, such as Ta, V, Th, Ti, and metals of the rare-earth group, hydrogen likewise forms solutions which, in their properties, prove to be semimetallic. Fig. 23 shows isobars at a pressure of \(1\ \mathit{atm}\) for

Fig. 22. Solubility of hydrogen in nickel, copper, and iron.

Fig. 23. Isobars for hydrogen in various metals.

some of these metals for temperatures from 0 to 1400° C, including palladium. Considering the results of these observations, Sieverts comes to the following interesting conclusion: “In these hydrides, which contain extraordinarily large quantities of hydrogen (1 volume of Ti dissolves 1800 volumes of hydrogen and 1 volume of Th—1700 of its volumes), there are no chemical compounds; rather, as can easily be seen from a comparison of the isotherms at high temperatures, there are solid solutions.”

The relative solubility of hydrogen in various metals at a temperature of 20° C and at atmospheric pressure is shown in Table X.

TABLE X

Ti (1.7)	V (0.6)
Lu (2.0)	Nb (0.7)
La (2.8) Ce (2.8) Hf?	Ta (0.8)
Th 3.2

With the alkali and alkaline-earth metals, hydrogen forms salt-like hydrides, in which hydrogen is present in the form of a negative ion. They are stable at ordinary temperatures (the dissociation pressure for KH at a temperature of 100° C is equal to \(7.76 \cdot 10^{-6}\) mm*), but their dissociation pressure at temperatures above 200° C is already appreciable.

Very little is known concerning the solubility of nitrogen in metals, with the exception only of the case of iron. Sieverts indicates that between temperatures of 900 and 1000° C electrolytic iron absorbs nitrogen very slowly, and in this way 0.06% of gas by weight can be introduced. Finely divided iron practically does not absorb nitrogen at all up to a temperature of 900° C. At this temperature 100 g of iron absorbs about 21 mg of nitrogen. Upon a further increase in temperature the solubility decreases somewhat. When the temperature is lowered to 900° C, absorpti-

* Extrapolated from the measurements of Keyes (F. G. Keyes, Journ. Am. Chem. Soc. 34, 779, 1913) at temperatures above 300° C.

dissolved gas begins to be released back again, so that this process is reversible. The amount of dissolved gas is proportional to \(p^{1/2}\). The rapid increase in solubility at a temperature of \(900^\circ\) C is probably connected with the transition of the metal from the \(\beta\)- to the \(\gamma\)-modification.

For the study of further details concerning the actual solubilities of various gases in more ordinary metals, the reader is referred to the tables at the end of Sieverts’ article.

The strong sorption of gases by metals in a highly divided state has been known for a long time. The best illustration of this fact is the use of platinum and palladium black in vacuum technique and in studies on catalysis. The absorption of hydrogen and carbon monoxide by pyrophoric iron, nickel, and cobalt was investigated by Nikitin.* Hydrogen is very strongly absorbed by these powders at the temperature of liquid air. At a temperature of \(-80^\circ\) C and at room temperature, iron does not absorb hydrogen, but at temperatures of \(380^\circ\) and above the amount of absorbed gas again increases rapidly. Hydrogen dissolved in this way in iron powder can be removed only very slowly by heating to a temperature of \(400^\circ\) C. In the case of carbon monoxide, iron powder absorbs the gas very well at a temperature of \(510^\circ\) C, and in all cases it is observed that the amount of absorbed gas varies with pressure according to Freundlich’s equation:

\[ \alpha = \beta p^{1/n}. \]

Consideration of this question would be incomplete without some remarks on the results of the most recent investigations on the diffusion of gases into metals.

Observations on the diffusion of hydrogen through various metals (Pd, Pt, Fe, Ni, and Cu) are described by Borelius and Lindblom.** They find that the diffusion rate \(m\),

* N. Nikitin, Z. anorg. Chem. 154 130, 1926.

* C. Borelius und S. Lindblom, Ann. d. Phys. 82*, 201, 1926.

as a function of temperature is determined by the relation:

\[ m = Ae^{-Q/RT}, \]

where the heat of diffusion \(Q\) per \(1\) mole has values from \(9\,400\) cal for iron to \(19\,400\) cal for platinum. At constant temperature the diffusion rate is proportional to \(p\).

Similar results on the variation of diffusion rates with pressure and temperature were obtained by Lombard* for the diffusion of hydrogen through nickel, iron, and platinum. It turned out that, if these rates are compared with the diffusion rates of nitrogen, argon, and helium through nickel, the former are by no means small.

This latter result is in agreement with certain observations that have been made on the solubility of rare gases in metals. Although Sieverts concludes that these gases are insoluble in metals, Lombard finds an appreciable solubility.

In conclusion, it should be pointed out how these observations on gases in metals relate to the choice of materials for anodes in cathode lamps. The higher the temperature to which a metal can be heated in vacuum, the greater the degree of degassing must be. Consequently, tungsten, molybdenum, and tantalum are the most desirable metals for constructing anodes intended to operate at high temperatures. In the United States, therefore, there is a tendency to use molybdenum, which, moreover, can be rolled and machined fairly easily. On the other hand, in Europe the use of tantalum is widespread, chiefly because of its ability, in the heated state, to absorb hydrogen and other gases. Therefore, for high-power lamps each of these materials is used in enormous quantities. For low-power cathode lamps and for amplifier tubes, nickel is used as a getter

* V. Lombard, Revue de Métallurgie 26, 343, 1929. This article contains a full bibliography of the earlier investigations.

and magnesium, owing to their much lower cost in comparison with molybdenum and tantalum. In this case, however, the preliminary treatment of nickel in vacuum furnaces is of particular importance, both for reducing the subsequent gas evolution during evacuation of the completed lamp and for absorbing residual gases.

MANOMETERS FOR MEASURING LOW PRESSURES

In the earlier period of the development of vacuum physics, a very large number of different methods for measuring low pressures were proposed. Descriptions of these methods are found in the literature already cited above. The experience of the last decade has led to the rejection of most of these proposals, and at present, for practical purposes, a comparatively small number of types of such manometers are regarded as advantageous.

I. Mercury manometers. The simplest of these is the McLeod manometer, in which a given volume \(V\) of gas, whose pressure \(P\) is to be measured, is compressed in a capillary tube to a volume \(v\) and to a pressure \(p\), after which the height of the mercury column is read. If it is desired to obtain a manometer of high sensitivity, then the ratio \(V/v\) must be as large as possible, whence follows the necessity of using large volumes of mercury and capillaries of very small diameter. In this way it is possible to make manometers with a sensitivity of approximately up to \(0.001 \cdot 10^{-3}\) mm of mercury.* Under these conditions, however, the manometer operates very slowly, and, in addition, the mercury has a tendency to adhere in the capillary. In ordinary practice McLeod manometers are used to determine the pressure on the fore side of a high-vacuum pump, and a sensitivity of \(0.1 \cdot 10^{-3}\) mm is quite sufficient. Of course, the greatest drawback of manometers of this type is that they indicate the pressure only of non-condensable gases.

* W. Gaede, Ann. d. Phys. 41, 289, 1913.

At various times various modifications of the McLeod manometer have been proposed. Thus, for example, Pfund* increased the sensitivity of the manometer by introducing a small filament into the capillary and using this combination as a thermal manometer (see below) for measuring the pressure of a compressed gas.

Hickman** described a mercury manometer that records pressures between 0.01 and 5 mm and that may prove useful in some cases in the investigation of low pressures. However, for many purposes this operating range of the manometer is too high.

II. Thermal manometer. Pirani*** proposed a manometer in which use is made of the circumstance that at low pressures the thermal conductivity of a gas is practically a linear function of the pressure. The Pirani manometer consists of a bulb containing tantalum, which can be heated by an electric current. The change in the thermal conductivity of the gas with pressure can be determined with the aid of this manometer by one of the following three methods.

1. The filament is maintained at a constant voltage, and the change in current is measured as a function of the pressure.

2. The resistance, and consequently also the temperature of the filament, are kept constant, and the total power is measured as a function of the pressure.

3. At constant current, the measure of the pressure is the change in the resistance of the filament.

Pirani found that the last method gives the greatest sensitivity. When working with this manometer, as a compensator for changes in the ambient temperature, an identical tube is used, which is pumped out as well as possible.

Hale**** improved the Pirani manometer by taking a thin platinum filament instead of a tantalum one and attaching it to the stem of an ordinary vacuum lamp. All con-

* A. H. Pfund, Phys. Rev. 13, 78, 1921.
** K. C. D. Hickman, Journ. Opt. Soc. 18, 305, 1929.
*** M. Pirani, Verh. d. Deutsch. Phys. Ges. 8, 24, 1906.
**** C. F. Hale, Trans. Am. Electrochem. Soc. 20, 243, 1911.

contacts between the filament and the hooks were welded, and the resistance of the filament was measured by means of a Wheatstone bridge, in which a manometer and a compensator are connected into two arms, while into the other two arms are connected two other resistances, approximately equal to the resistance of the filament in the manometer. For pressures from \(5 \cdot 10^{-3}\) mm and below, the resistance, as experience shows, varies linearly with the pressure. With the aid of this method it is very easy, and works well, to measure pressures down to \(10^{-5}\) mm of mercury.

Campbell* measures the change in voltage across the filament at constant resistance (method 2). If \(v_0\) denotes the voltage at \(p = 0\), and \(v\) the voltage at pressure \(p\), then

\[ \frac{v^2 - v_0^2}{v_0^2} = k \cdot f(p), \]

where \(k\) is a constant for the given manometer, and \(f(p)\) is a quantity determined on the basis of calibration data. Recently Stanley** investigated the influence of various design features on the sensitivity of the instrument. When a loop of very thin platinum wire, operating at a very low temperature (below \(100^\circ\) C), is used, it is quite satisfactorily possible to measure pressures from \(2 \cdot 10^{-5}\) mm to \(4 \cdot 10^{-6}\) mm.

Thompson and Henelly (H. C. Thompson and E. F. Henelly) in our laboratory developed a simple form of recording manometer, which proved very convenient for practical purposes. The measuring tube and the compensator are made of a glass tube 22 mm in diameter; their length is about 70 mm. Four or more standard 25-watt, 115-volt tungsten filaments are welded to the hooks. The total resistance of the filament at room temperature is about 15 ohms. The measuring tube and the compensator are connected into two arms of a Wheatstone bridge (Fig. 24), while the other two arms contain two resistances \(R_1\) and \(R_2\) (each about 14 ohms) and a resistance (about 2 ohms), which can be adjusted in order to obtain

* N. R. Campbell, Proc. Phys. Soc. 33, 287, 1921.
** L. E. Stanley, Proc. Phys. Soc. 41, 194, London 1929.

Fig. 24. Diagram of electrical connections for a thermal manometer.

tometer on the ammeter \(R_3\). A dry battery is used as the constant source of potential.

Fig. 25 shows the current curve (in milliamperes) as a function of pressure (in microns) for a typical experiment. In order to avoid confusion, the drawing has three pressure scales. Curves \(A\), \(B\), and \(C\) refer to dry air at three different battery voltages, and curves \(D\) and \(E\), respectively, to hydrogen and argon. It is obvious that, for pressures below \(40 \cdot 10^{-3}\) mm, the curves run linearly and do not depend on the battery voltage.

If a sensitive milliammeter is used, and if due care is taken in the construction of the tube and the compensator, then with the aid of such a circuit it is possible to measure pressures down to \(10^{-6}\) mm. However, when it is desired to use a comparatively inexpensive type of measuring instrument, the sensitivity is reduced to \(10^{-5}\)—\(10^{-4}\) mm.

III. Thermoelectric mano-

Fig. 25. Relation between current and pressure for a thermal manometer.

etc. Instead of measuring resistance, the temperature of the filament can be measured by means of a thermocouple in contact with this filament.* Rohn observed that, for a range of pressures between \(7.5 \cdot 10^{-2}\) mm and \(7.5 \cdot 10^{-4}\) mm Hg, the e.m.f. is proportional to the logarithm of the pressure, with the sensitivity rapidly decreasing at low pressures.

Rumpf** described a manometer consisting of a platinum spiral, along the axis of which a thermocouple of thin constantan and copper wires is mounted. Within the pressure range from 0.001 to 3 mm Hg the e.m.f. varies approximately linearly with the logarithm of the pressure. At lower pressures the manometer is almost entirely insensitive.

Fig. 26. Thermoelectric manometer.

A manometer based on this same principle was developed in our laboratory by Gordon (Fig. 26). It consists of a platinum-iridium strip \(0.00064 \times 0.0225\) cm, to the center of which a thermocouple made of chromel-nickel strips measuring \(0.00125 \times 0.01\) cm is welded. All this is mounted on a stem with four lead-in wires in a tubular bulb of the type used in UX-199 radio tubes. A direct current of about 30 mA is passed through the platinum strip, and the temperature is measured by means of the thermocouple. The current is supplied by three dry cells connected in series, through a fixed resistance of about 110 ohms (with a very small temperature coefficient) and a variable resistance of 40 ohms. For reading the thermocouple e.m.f. a sensitive millivoltmeter is used. The apparatus must, of course, be calibrated against a standard

* Rohn, Z. Elektrochem., 20, 539, 1914.
** E. Rumpf, Z. techn. Phys., 5, 224, 1926.

to the McLeod manometer for each gas separately. Hydrogen gives the largest deviations, while mercury vapor produces very small deviations for the same changes in pressure.

IV. Ionization manometer. The most satisfactory method, applicable to pressures below \(0.1\,\mu\), consists in measuring the ionization produced in the gas by means of a definite electron current. Electrons emitted by a tungsten or oxide filament are accelerated by applying a positive potential of 100 to 250 V to the neighboring electrode (anode), and the positive ions formed as a result of collisions of the electrons with gas molecules are directed to a third electrode, to which a negative potential (from 10 to 50 V) is applied with respect to the cathode.

A manometer based on this principle was first described by Buckley.* This method was then investigated more fully by Dushman and Found.** The manometer used in this work is shown in Fig. 13. It consists of two tungsten spirals, of which the inner one is used as the cathode and the outer one as the anode, and of a molybdenum cylinder surrounding these spirals, used as a collector for the positive ions. For a given electron current (from 1 to 20 mA) and a constant anode voltage (from 100 to 250 V), the ionization current proves to be proportional to the pressure; moreover, within the initial range of electron currents (from 1 to 20 mA), the ionization varies linearly with the electron current. In the case of this manometer, the lower limit of pressures is determined, first, by the sensitivity of the instrument used to measure the current of positive ions, and, second, by electrical leakage between the collector and the filament. This leakage is eliminated to a significant degree by means of the construction of two feet shown in the figure. With galvano-

* O. E. Buckley, Proc. Nat. Acad. Sciences, 2, 683, 1911.
** S. Dushman and C. G. Found, Phys. Rev. 17, 7, 1921; 23, 734, 1924.

meter, with a sensitivity of \(10^{-7}\) A, and with 10 mA of electron current it is possible to determine pressures down to \(10^{-8}\) mm.

In describing this manometer design it was reported that the ionization current, at constant electron current and at a given pressure, is proportional to the total number of electrons in the molecule. However, Reynolds showed that this result is only an accidental coincidence of the data and that it is not confirmed for other manometer designs.*

For pressures of the order of \(10^{-3}\ \mu\) or higher, an ordinary three-electrode vacuum tube may be used. Observations with manometers of this form, published by Simon,** show the same proportionality of the ionization current to the electron current and to the pressure.

Of course, before using this manometer it is very important to free the metallic parts from gas by the usual methods. Since oxide cathodes are far less subject to damage in the event of accidental ingress of air, they are very often preferred for use in ionization manometers.

For very high sensitivity, the positive-ion current may be amplified by means of a “push-pull” circuit. An instrument of this type for direct readings was described by Found and Reynolds. The positive-ion current may also be used to charge a capacitor. A method of vacuum measurement based on this principle has recently been described by Sarbey.*

* A complete discussion of this will be published shortly.

* C. C. Simon, Z. techn. Phys., 5*, 221, 1924.

* C. C. Found and N. B. Reynolds, Journ. Opt. Soc. Am. and R. S. I., 13**, 217, 1926.

** M. D. Sarbey, Electronics, 594, April 1931.