100-Million-Volt Induction Electron Accelerator*

W. F. Westendorp and E. E. Charlton.

I. Introduction

Some of the earliest ideas on the method of accelerating electrons by means of a time-varying magnetic field were proposed by Wideröe[^1], Walton[^2], Jassinsky[^3], and Steenbeck[^4]. Before the publication of Kerst’s first paper[^5] on this subject, there were no indications in the literature that an apparatus based on this principle had ever been successfully realized. Later a paper by Steenbeck[^6] was published, describing some of his early experimental work, apparently preceding D. W. Kerst’s work.

The general theory of the induction electron accelerator was described in previously published papers[^4],[^7]–[^9]. Electrons are accelerated by an electric field produced by a time-varying magnetic field. The induction electron accelerator may be compared with an ordinary transformer, in which the high-voltage secondary coil with a large number of turns is replaced by an evacuated glass tube of toroidal form, equipped with an electron gun with a hot cathode. The electrons are injected into the tube tangentially and, under the action of the magnetic field, move inside the tube in a circle, acquiring at each revolution the same increment of energy as if they were moving along a separate turn of wire. In the process of continuous acceleration of the electrons, the magnetic field guides them along a circular orbit in the vacuum tube, and at a selected moment this same field removes the electrons from this circular orbit, forcing them to strike a target and produce X-rays, or else to leave the vacuum as high-energy electrons.

Having obtained, as in the construction of his first accelerator, leave from the University of Illinois, Dr. Kerst arranged

* Journal of Applied Physics, 16, 10, 581 (1945). Translated from the English by E. Burstein.

with the staff of the research laboratory of the General Electric Company on cooperation in the design and manufacture of a more powerful accelerator.

Fig. 1. Front view of the electron accelerator.

The machine then built, of 20 million volts*) presented no unexpected difficulties in its manufacture, and encouraged us to design and build the still larger machine that is the subject of the present article. The latter weighs approximately 130 tons and is about 9 feet high, 6 feet wide, and 15 feet long. Together with capacitors of 24,000 kilovolt-amperes for power-factor correction, it is housed in a special room intended to protect the surrounding buildings from penetrating radiation.

Although in its general features this machine is very similar to the previous 20-million-volt one, its larger dimensions made necessary a number of changes both in design and in manufacture. In particular

*) This machine, immediately after its manufacture, was transferred on loan to the University of Illinois.

Induction Electron Accelerator

The tube differs greatly from the former one, consisting of 16 molded glass sections fastened with toruses.

The front view of the accelerator is shown in Fig. 1; the top view with the upper yoke of the magnet removed and with the lower magnetizing coil and the vacuum tube in place is shown in Fig. 2.

The arrangement of the accelerator in a special building is shown in Fig. 3.

The magnet operates on 60-cycle current and gives, at the orbit, a magnetic-flux density of 4000 gauss. Electrons are injected with a voltage of from 30 to 70 kV and, if they are allowed to remain on the 66-inch circular orbit for the entire quarter-period, they traverse the magnetic flux 250,000 times, receiving on each revolution an average additional energy of about 400 eV. Since, by giving them a field impulse, they can be displaced from the circular orbit at any moment during the quarter-period, they can be given any desired energy from one million up to one hundred million electron-volts.

Fig. 2. Vacuum tube on the magnet.

Labels visible in Fig. 3:

• Condenser room [[unclear: possibly “12 ft below on the 2nd floor”]]
• Control room, 11 × 30 ft
• Accelerator room, 25 × 40 ft
• X-ray beam
• 3 ft 6 in
• 3 ft 6 in

Fig. 3. Floor plan of the building for the electron accelerator.

Since the electrons are led away from their orbit along a spiral, they strike the target, producing X-rays. In this case the apparatus can be used as a very compliant and easily controlled source of X-rays of any desired voltage up to 100 mil-

...million volts. The output, measured by a small thimble ionization chamber placed at the center of the X-ray beam, rises rapidly with voltage. Under some conditions it was about 2600 roentgens per minute at 100 MV, falling to 7% of this value at 20 MV.

The half-width of the X-ray beam is about 12° at 20 million volts and 2° at 100 million volts.

The machine can operate continuously at full load even in the hottest weather.

II. MAGNET

The core of the magnet, as shown in Figs. 4 and 5, consists of two parts of the yoke, separated by two spacers, two pole pieces,

Fig. 4. Magnetic circuit. The wooden casing, air ducts, and foundation are not shown.

1. Upper part of the yoke. 2. Suspension of the pole piece. 3. Central disks. 4. Pole piece. 5. Spacer. 6. Redwood ring. 7. Lower part of the yoke.

and two central disks. All these parts are made of enameled steel sheets (4.5% silicon) 0.014 inch thick, joined into plates or, in the case of the pole pieces and central disks, into sectors. Each plate or sector consists of a large number of sheets fastened together with varnish polymerized during drying in the appropriate jig or mold.

The plates are 7 inches thick and are separated from one another, as can be seen in Figs. 2 and 6, by numerous narrow wooden slats arranged so as to allow effective forced-air cooling. The plates are compressed between casing boards by electrically insulated steel bolts, as shown in Figs. 1, 2, and 5.

INDUCTION ELECTRON ACCELERATOR

Fig. 5. View of the magnetic circuit assembly. External air ducts are not shown.
1. Large plate. 2. Small plate. 3. Maple board. 4. Armature. 5. Yoke tie rods. 6. Insert tie rods. 7. Insert. 8. Plywood. 9. Rubber shock absorber.

Fig. 6. Cross section of the magnet, showing the forced-air cooling system.
1. Air duct and attachment. 2. Inlet air ducts. 3. Wooden spacers. 4. Magnetic circuit (silicon steel). 5. Machine foundation.

Pole pieces.

The sheets from which the sectors of the pole pieces were made were stamped with the profile shown in Fig. 7. With plates of different lengths (differing by 2 inches) it was possible to assemble, in a special clamp, a sector tapering toward one end by approximately 4 degrees. Before assembly, a thin film of lacquer was applied to each sheet, and the form was screwed down in order to obtain a strong wedge-shaped plate. When dried at 150°, the lacquer polymerized, after which the sector could be regarded as one solid piece.

Fig. 7. Profile of a pole-piece sector.

The rectangular slots at the ends serve for placing and holding in position the sectors of the lower pole piece and for holding the sectors of the upper pole piece on the upper yoke. For this purpose the assembled sectors are encircled by a large split, carefully machined wooden ring (10 in Fig. 8), consisting of a large number of pieces of redwood. In a similar manner, at the center, a textolite ring (6 in Fig. 8) is placed in the rectangular slots. A steel crosspiece clamps the textolite ring to the upper pole piece, while a brass rod screwed into the hub of the crosspiece supports the center of the pole piece. Between the plates forming the upper yoke pass steel strips connected to the wooden ring and supporting the rim of the pole piece. The lower pole piece is fastened to the lower yoke by means of a textolite crosspiece, which allows cooling air to circulate through the center.

A sheet of pressed wood, \(1/16\) inch thick, separates each pole piece from the yoke to which it is attached, in order to prevent a short circuit between the plates.

The shape of the pole pieces was found with the aid of solid steel models at scales of 1:8 and 1:4. Measurements on these solid

Fig. 8. Central part of the machine in enlarged view.

1. Magnetizing coil. 2. Power input winding. 3. Central disk.
2. Gasket of pressed wood. 5. Textolite gasket. 6. Central suspension.
3. Centering crosspiece. 8. Separating ring. 9. Retaining ring.
4. Red-wood ring. 11. Vacuum tube.

models were carried out with excitation by direct current, reversal of whose direction produces a ballistic deflection of the galvanometer in the measuring circuit. The shape was chosen so that the magnetic-field intensity in the median plane between the poles varied inversely proportionally to the radius to the power \( \frac{3}{4} \). This relation ensures focusing of the electrons after the first revolution into a narrow beam, the greater part of which passes the electron gun. How well this law of decrease of the intensity was realized is shown in Fig. 9, which depicts the logarithm of the field intensity as a function of the logarithm of the radius.

Fig. 9. Logarithmic plot of the dependence of intensity on radius.

1. Measurement data. 2. Straight line with slope \( -\frac{3}{4} \).

Central disks.

The central disks are intended to ensure fulfillment of the requirement necessary for the motion of the electron in a circular orbit, namely, that the magnetic

the flux through the orbit was equal to the product of twice the magnetic-flux density at the orbit by the area of the orbit.

The magnetic-field strength outside the central disks or inside the vacuum tube must be sufficiently uniform along a line parallel to the axis. This is achieved by using two central disks, arranged so that the size of the upper and lower air gaps is equal to half the central gap.

With such an arrangement, a simple and easily manufactured profile of the pole pieces (Fig. 7) gives the desired field distribution.

The plates of the central disks were stamped to a suitable width, corresponding to the thickness of the disks, and strips of various lengths were cut from the stamped plates at intervals of 1 inch, in order to obtain 15-degree sectors.

The central disks are supported and separated from one another by rings, and from the pole pieces by textolite bars, as shown in Fig. 2. They are placed at such angles to the radius of the pole piece that the cooling air is directed through the gaps at an almost constant velocity.

Yoke.

The upper and lower parts of the yoke (Figs. 4 and 5) each consist of six wide plates and four narrower ones, two on each side. Such an arrangement gives a rough approximation to a semielliptical cross-section, collecting the magnetic flux through the round pole pieces with less variation in magnetic-flux density and less transverse flux than with a simple rectangular cross-section of the yoke. The upper part of the yoke was assembled on a temporary support, and the upper pole piece and the magnetizing coil were fastened to it. This 60-ton unit was subsequently brought by crane into its final position.

Inserts.

The inserts are constructed more simply than the yoke, because their steel sheets have the full size of the insert. Here, too, six wide plates and two narrower ones on each side were used, in accordance with the cross-section of the yoke. On each insert, along the edge nearer to the pole piece, a gasket of pressed wood 6 inches wide and 0.045 inch thick is fastened. The upper yoke rests on these two gaskets and on textolite gaskets under the upper central disk.

III. ELASTIC MOUNTING OF THE MAGNET.

To reduce the transmission of vibrations from the magnetic circuit to the concrete floor, and to avoid possible resonant vibrations of the floor,

and also so that the magnet can oscillate about its center of gravity, the entire 130-ton assembly is mounted on shock absorbers made of natural rubber.

IV. MAGNETIZING COILS.

Through the magnetizing coils (Figs. 1, 2, and 8) flows the current that creates the main magnetic flux through the orbit and the guiding field in the orbit. For operation at 100 MV, 40,000 ampere-turns are required. This current is distributed over 80 turns of wire, forty in the upper and forty in the lower coil. At 100 MV, the voltage per turn is about 600 V, and the apparent power accordingly reaches 24,000 kVA. This apparent power is supplied, at an effective voltage of 24,000 V, from banks of capacitors located in the upper room of the building. From the capacitors the current flows to the coils through cables that are at a voltage of 12,000 V with respect to ground. For safety, the connection is made permanent, without any switches or other disconnecting devices.

The cable for the magnetizing coils consists of 61 strands of copper wire, 0.102 inch in diameter, braided with two layers of varnished cambric 0.010 inch thick. The individual wires are insulated from one another to prevent the occurrence of parasitic currents in the copper due to the leakage magnetic flux, which is perpendicular to the cable. Four layers of varnished cambric provide insulation at 600 V between turns of one layer. Five turns are laid in a layer, so that the maximum voltage between layers of turns is 6000 V. Adjacent layers are separated by textolite spacers 1¼ inches thick.

Both coils are fastened on radial textolite strips. The lower coil is spaced 1¼ inches from the yoke. The strips under the upper coil are suspended on brass rods running from the upper yoke. It is important to provide the proper insulation between the input windings, the magnetizing coils, and the steel plates *).

*) After the article had been written, an accidental discharge of a large capacitor onto the core of the magnet occurred, leading to complete cessation of the emission of X-rays. Investigation showed that the magnetic field at various points along the tube was no longer in phase, and that this was caused by damage to the insulation between individual steel sheets in some plates, due to the high-voltage discharge. The magnetic flux produced by excessive Foucault currents in these plates was not in phase with the magnetic flux caused by the principal magnetizing current. This defect was corrected, and the emission of X-rays was restored, by neutralizing the magnetic fluxes produced by the Foucault currents by means of an auxiliary magnetic flux, created by a current flowing through one turn of copper wire wound on each damaged plate. This incident once again emphasized the importance of ensuring

V. CAPACITOR BANK.

The capacitor bank of 24,000 kVA, supplying the reactive power to the magnetizing coils, consists of approximately one thousand capacitors arranged in sixteen frames made of angle iron. In each frame, more than sixty capacitors are connected in parallel. Eight frames on each side of the room are connected in series, forming two banks of 24,000 V and 500 amperes, each of which is connected with the sections of the magnetizing coils as described above. All the frames are mounted on standard 15,000-volt bus-bar insulators placed directly on the floor. The power loss in the capacitor room amounts to \( \frac{1}{3}\% \) of the reactive power, i.e. 80 kW. This heat is removed by forced air circulation (9000 cu. ft. per minute), the air temperature rising by \(16^\circ\) C. Above the frames and between them there are arc suppressors, directing the air between the capacitors, which are spaced \(3\frac{1}{2}\) inches apart.

VI. POSITION OF THE ELECTRON ORBIT.

According to calculation, the electron orbit was to have a radius of 33 inches with a magnetic-flux density on the orbit of 4000 gauss at 100 MV.

The true position of the orbit was determined by two different methods. The first method was described earlier\(^7\). It uses a set of three coils with diameters comparable with the diameter of the orbit; the method is based on the fact that the electric gradient has a minimum on the orbit.

The second method is based on the principal requirement that the magnetic flux through the orbit be equal to the magnetic-flux density multiplied by twice the area of the orbit. In this method the voltage produced in one turn of 66-inch diameter was applied in opposition to the voltage produced in two coils of 1000 turns each, with an effective diameter of \(66\sqrt{1000}\) inches. A cathode oscillograph served as the indicator.

A zero reading of the oscillograph was to be obtained if the orbit had a diameter of 66 inches and if the small coils, with their centers, were located at the ends of this diameter.

If, however, the orbit deviates from the desired position, then by changing the position of the small coils one can obtain a zero reading—

electrical insulation in the magnetic core in future installations. It is hardly likely that the original insulation between layers can be sufficiently reliable at the large voltage gradients that may arise in an accidental discharge of the capacitors onto the core, so that in such installations the core must be effectively protected from the capacitor circuit by a strong dielectric or by a grounded conductor.

oscillograph reading, and from the radius at which the small coils give a zero indication one can find the true diameter of the orbit. The calculation showed that this method makes it possible to determine the orbit within \(\pm \frac{1}{16}\) inch for a radius of 33 inches. This second method served as a check on the three-coil method.

If, during the first assembly of the machine, the orbit deviates in diameter from the required one, an adjustment is made: the upper yoke is raised, and the separating shims on the inserts and on the upper central disk are replaced. When their thickness is increased, the magnetic flux through the orbit decreases more than the field on the orbit, and therefore the orbit decreases. The required increase in the separating gap can be calculated for each machine as a function of the error in the orbit radius. For the present machine such a calculation showed that, if the orbit radius is 0.100 inch greater than normal, the upper pole piece must be raised by 0.0035 inch. After the correct position of the orbit has been ensured, the upper yoke is removed, and the tube is taken out of its assembly ring and placed in its final position, as shown in Fig. 2.

During operation of the machine, contraction of the orbit leads to further complications. If the primary voltmeter and the megavoltmeter are calibrated as functions of the values of \(H\) and \(R\) on the normal orbit, which determine the electron energy, then a correction must be introduced into the electron energy after contraction of the orbit, since \(H\) and \(R\) at the target differ from the values of \(H\) and \(R\) on the normal orbit. The correction factor is the product of two other factors: one expressing the strengthening of the field upon contraction of the orbit (1.0238), and another indicating the change in the product \(HR\) due solely to the change in position (0.9785). The product is a negligibly small correction factor (1.0023).

VII. VACUUM TUBE AND PUMPING SYSTEM

The assembled glass vacuum tube (Figs. 2 and 8) has the form of a torus with an outside diameter of 74 inches and an inside diameter of 58 inches, with an almost elliptical cross-section whose horizontal axis is 8 inches and whose vertical axis is \(4 \frac{7}{8}\) inches. It consists of 16 sectors of molded Pyrex glass*) with a minimum thickness of \(\frac{1}{4}\) inch, subjected to heat treatment and hardening.

Preliminary experiments with individual unhardened sectors with wall thickness slightly less than \(\frac{1}{4}\) inch showed that after

*) For the molding and hardening of the curved glass sectors we are indebted to “Corning Glass Works.”

of the air pumping, they do not withstand the external atmospheric pressure. But if the sectors are made somewhat thicker and tempered, they withstand a pressure difference of 70 pounds per sq. inch with a sufficient safety factor.

The technique used for preparing these 16 sectors and ensuring the impermeability of the vacuum was developed so well that the large number of sectors did not prove to be an obstacle.

Each glass sector was smoothly ground on both ends with an angle between the ends of 22.5 degrees. To allow for a certain distortion during tempering, each sector was left sufficiently long so that, after tempering, it would be possible to grind off \( \frac{1}{64} \) inch from each end.

To prevent charging of the inner surface of the tube, which would lead to distortion of the electron orbit, it was necessary to make the inner surface of the sectors conducting. For this purpose the inner surface was treated with a sand-blasting apparatus and then silvered. A process of chemical silvering was used, and the use of solutions of low concentration at a definite temperature permitted precise control of the resistance.

For assembling the tube, the sectors were placed on a precisely made reinforced steel ring with a flat, ground upper surface. To remove the glass tube from the steel ring and set it in place, mounting supports and retaining clamps were used. The flat ends of the glass sectors were placed opposite one another, and the fitting was checked with thickness gauges 0.5 thousandth of an inch thick. After this, red sealing-wax paint was applied externally at each joint and an all-round heating was carried out. On the last, sixteenth, sector it was necessary to change one angle slightly in order to eliminate the total assembly error. The ends of three special sectors at the butt joints with brass and glass were ground at a right angle to facilitate joining. For ease of disassembly these joints were covered with a putty made of pine resin and shellac. Fig. 10 shows the details of the target device.

The connection to the pump is rigid and consists of a \(3 \frac{1}{2}\)-inch glass elbow tube leading from the tube to a liquid-air trap. This trap is of the concentric type, with an inner tube \(3 \frac{1}{2}\) inches and an outer tube \(5 \frac{1}{2}\) inches in diameter. Between the liquid-air trap and the vacuum tube a standard glass ionization manometer is connected. The liquid-air trap is connected to a mercury pump having a pumping speed of 80 liters per second. In front of the diffusion pump there is a vacuum valve, in which leakage outward is prevented by a metalli-

rical corrugated tube. This valve is always closed if the vacuum tube is not operating. It connects with a rotary fore-vacuum pump having an automatic centrifugal valve-switch, which prevents oil or air from entering the vacuum system when the motor stops.

The trap is at all times surrounded by liquid air. If, for experimental work, it is necessary to change the electron gun or the target, then thoroughly dried air is admitted into the tube through a branch pipe placed between the trap and the tube. After the tube is again closed and the pump is switched on, the pressure falls within an hour to 0.01 micron, and the tube can operate satisfactorily. By the end of the second hour the pressure usually falls to 0.001 micron or even less, and can without difficulty be maintained at this level during prolonged operation of the accelerator.

Fig. 10. Target and tube suspension.

1. Textolite suspensions. 2. Tungsten target (wire of diameter 0.100 inch). 3. Ring of redwood. 4. Rubber. 5. Textolite post. 6. Connection. 7. Conductors of the chain shortening the orbit.

Because of the very rigid connection of the tube with the pump, there were at first some fears that excessively strong oscillations would be transmitted from the magnet to the glass. This, however, was prevented by a fairly complicated system of elastic mounting of the tube. It consists of 19 flexible suspensions, each of which (Fig. 10) is made of two textolite strips \( \frac{1}{8} \times 2 \frac{1}{2} \times 11 \) inches, placed horizontally one on top of the other, the ends of which rest on half-inch cubes of soft rubber. One cube is located at the central disks; the other two are mounted on a textolite post screwed to a ring of redwood encircling the strip end piece.

This suspension bends under the weight of the tube by \( \frac{1}{4} \) inch and has its own natural frequency of oscillation of about 5 cycles per second. The rubber cubes, in which holes have been made, are fitted onto pins that do not allow the cubes to shift, but do not prevent them from bending. Thus the rubber serves to provide flexibility in the horizontal direction. The elastic suspension of the tube is so effective that, when current is flowing in the machine, a hand placed on the tube does not feel oscillations.

VIII. ELECTRICAL CIRCUITS

Fig. 11 shows the main power circuits. A sufficiently powerful three-phase motor with a generator was used, so that a 200 kW single-phase load could be taken from it without fear of overheating the poles due to eddy currents. The advantage of using a motor-generator set over direct connection to the supply line lies in its independence from voltage fluctuations in the line and in the accuracy of voltage regulation by means of a regulator acting on the generator field winding. The motor is synchronous, so that the carefully maintained frequency of the supply line is transmitted to the generator output. It was found, however, that the still-existing small fluctuations of the supply-line frequency caused corresponding but relatively large fluctuations of the accelerator voltage if the generator was directly connected to the input winding of the accelerator. The explanation of this effect lies in the noticeable fluctuation of the load power factor with frequency and the corresponding voltage drop caused by the synchronous reactance of the generator. When the frequency is equal to 60 cycles per second and the required number of capacitors is connected in the capacitor bank, the input power factor is unity. But when the frequency increases by 0.1 cycle per second, the apparent power of the capacitor bank increases by \(\frac{0.1}{60}\times 24\,000\), i.e., 40 kVA, while the apparent power of the magnetizing coils decreases by 40 kVA, which results in a load on the generator with a leading phase component of 80 kVA, corresponding to 33 amperes at 2400 volts. With synchronous reac-

Fig. 11. Main power circuits.

1. Motor-generator, 1200 kVA, 2400 V.
2. Stabilizing resistance.
3. Capacitor connected in parallel.
4. Equalizing transformer.
5. Input power winding of the upper pole (4 turns).
6. Input power winding of the lower pole (4 turns).
7. Outer section of the upper magnetizing coil (20 turns).
8. Corresponding inner section.
9. Outer section of the lower magnetizing coil.
10. Corresponding inner section.
11. Capacitor bank rated at 24,000 kVA. Each symbol represents a group with capacitors rated at 30,000 kVA.

tance of the generator at 9 Ω, this additional current tends to raise the output voltage per vector component by 300 V and, if the change in frequency occurs sufficiently rapidly, the voltage regulator is not able to correct the voltage. Obviously, the solution is to connect in series a capacitor neutralizing the synchronous reactance. Experience has shown that this solution is quite satisfactory in preliminary work, when the accelerator operated at 50 MW and four turns of the upper output winding were connected in series with four turns of the lower winding, while the generator operated at full voltage. When, however, the input power windings were connected in parallel for operation at 100 MV, the system with series-connected capacitors became unstable, since low-frequency power oscillations occurred throughout the entire system, from the motor and up to the tuned circuit with its 24,000 kVA. It was found that these oscillations can be completely eliminated by means of an 8-ohm resistance, dissipating from 50 to 80 kW depending on the tuning. The tuning changes from day to day and during operation because of the temperature coefficient of the capacitors. Although this coefficient is only \(-0.0005\) per degree Celsius, one may expect changes of 1% when the temperature changes by 20° from the starting temperature on a cold day to the final temperature on a hot day. This change corresponds to a reduction in the reactive power of the capacitors by 240 kVA. The circuit is tuned once and for all for the mean operating temperature so that, at that temperature, the power factor for the input power is equal to unity. In this case the circuit can operate over the entire temperature range without retuning.

The injector and the electrical circuit for orbit contraction are arranged, as in the accelerator, for 20 million volts \(^{10, 11}\).

IX. COOLING SYSTEM.

In an annex above the machine room there is a ten-horsepower blower, delivering 8000 cubic feet of air per minute at a pressure of 1 inch of water for dissipating the hundred-kilowatt losses in the iron of the magnet. Before entering the room, the air passes through filters. Three rows of thermostatically controlled openings determine the ratio of recirculation to outside inflow and exhaust. The low-pressure edge of the annex is connected by an air duct with the top of the machine. Here it receives air from the external system of air ducts located around the magnetic circuit of silicon steel. Part of this “external” duct system is also formed by the foundation of the machine. Fig. 6 shows a transverse section of the machine through one of the cooling slots, \( \frac{3}{4} \) inch wide, between the yoke plates and the inserts. Rectangular air--

air intake channels pass from the front to the rear part of the machine and are visible in Fig. 1, where they are shown as dark openings. In Fig. 6 one can easily trace the path of the air between the light wooden spacers placed between the steel plates. The lower side of the upper yoke and the corresponding sides of the inserts and of the lower yoke are completely covered with wooden sheathing, so that the air is driven along them, as indicated by the arrows. Eight-inch openings in the center of the upper and lower yokes and in the pole tips allow the passage of air from the external space at the vacuum tube through the upper and lower gaps into the main air channel. The magnetizing coils do not receive forced air cooling, since quarter-inch vertical slots between the layers provide sufficient cooling by convection.

Three parts of the vacuum tube—the points of connection with the vacuum system, the electron gun, and the target—receive additional cooling.

The condenser room has its own separate air-cooling system, located in the upper superstructure, with temperature-controlled openings and with air intake and exhaust to the outside. The five-horsepower blower of this system can deliver 9000 cu. ft. per minute at a pressure of \( \frac{1}{2} \) inch of water. With 80-kilowatt losses in the condensers, which corresponds to \( \frac{1}{3}\% \) of 24,000 kVA when operating at 100 MV, the air temperature under steady conditions rises by 16° C.

X. THE BUILDING.

The need for protection from X-rays, the presence of a concentrated floor load, and the special design features of the special crane required the construction of a special building for the machine. The floor plan of the building (Fig. 3) shows that the wall thickness of the accelerator room was not less than 36 inches. This thickness of the concrete walls was maintained up to a height of 15 feet, where the walls have a setback, passing above into brickwork and glass. Since the X-ray beam is relatively narrow, and the scattered radiation from the chamber is blocked by the upper yoke, the indicated height was considered sufficient. The setbacks of the concrete wall were used for the rails of a sixty-ton crane. The floor of the machine room is reinforced concrete, 16 inches thick. The massive walls of the building and the nature of the soil made it necessary to drive piles to prevent settlement. A motor-operated steel door, 1 inch thick, leading into the machine room, provides chiefly the stopping of primary and scattered electrons.

As indicated on the floor plan, in that part of the wall which is exposed to X-rays there is a section of unsecured concrete blocks. If it is desired to use the X-ray beam outside the building, part of this section can be removed.

In the control room there are panels with control handles, switches, and measuring instruments for the main power circuit, the injector circuit, the circuit for changing the orbit, and the ionization manometer. Other instruments, such as, for example, the measuring instruments of the direct-reading ionization chamber and the Geiger–Müller counter, are also located within the operator’s field of view. The megavoltmeter gives a direct reading of the electron energy at the moment of orbit contraction.

The blowers for the machine and condenser rooms are started from the control room. A vane switch, controlled by the air flow, provides interlocking, permitting the main power to be supplied only after the blower of the condenser room has been started. The blower of the machine room can be started either manually, by pressing a button, or automatically by means of a relay which operates at half the output voltage of the generator. On all doors to the machine and condenser rooms there are interlocking switches. In addition, for the safety of personnel, the door to the condenser room is locked.

The condenser room is located above the control room and the corridor. In the ceiling on each side openings have been left connecting the condenser room with the blower and with the filter system in the annex. Similar openings are present in the ceiling of the machine room for connecting the cooling system of the machine with the annex.

The noise level in the room for the accelerator during operation is very high—between 110 and 120 decibels—partly owing to the properties of the completely enclosed room with concrete walls and windows made of glass structural brick.

XI. SOME EXPERIMENTAL RESULTS.

Figs. 12 and 13 give absorption curves in steel and lead for various X-ray energies. The X-ray energy was regulated by shifting the phase of the electrical circuit contracting the orbit. The slope of the curves in Figs. 12 and 13 is expressed in Fig. 14 through the coefficient of linear absorption.

Fig. 15 shows the distribution of X-rays for various energy levels. In the interval from 20 to 50 MV the X-ray cone narrows very rapidly; the change from 50 to 100 MV is considerably smaller.

These data were obtained with an ionization chamber having an effective cross section of 0.85°, limiting the resolving power. Therefore, in order to determine the dimensions of the radiation cone at 100 MV, a more sensitive radiographic method was devised.

Fig. 12. Passage of X-rays through steel.

Ordinates are in arbitrary units of intensity, without observing the ratio between the curves. Curves for 100, 50, 40, 30, 20, and 10 megavolts were obtained with an ionization chamber of direct reading. Curves for 5 and 4 megavolts were obtained with a fluorescent screen and a photoamplifier.

Fig. 13. Passage of X-rays through lead.

Ordinates are in arbitrary units of intensity, without observing the ratio between the curves. Curves for 100, 75, 50, 40, 30, 20, and 10 megavolts were obtained with an ionization chamber of direct reading.

Radiographic determination of the half-width of the beam at 100 MV was based on the simultaneous exposure and development of two films, located in the beam respectively at 8 ft and at \(8\sqrt{2}\) ft (11 ft 4 in.) from the target. As the half-width of the beam at 8 ft there was taken that diameter of the image at whose ends the intensity is equal to the maximum intensity on the film located at \(8\sqrt{2}\) ft. This value corresponded to 2 degrees.

Fig. 14. Linear absorption coefficient of X-rays in lead and steel.

The coefficient, in inverse centimeters, was calculated from the slopes of the curves in Figs. 12 and 13.

Fig. 15. Distribution of X-rays.

The data were taken with the center of the chamber 9 ft 1 in. from the target and are plotted as percentages of the maximum intensity for 20, 50, and 100 MV. The chamber had an effective diameter of \(3/8\) in., i.e., \(0.85^\circ\), which limited the resolving power. The diameters of the beam of relative intensity (the half-width of the beam) are equal, as is evident from the graph, to 12.3, 4.2, and 3.7 degrees, respectively.

The intensity of X-rays at the center of the beam is maximal when operating at 100 MV. The output falls if the pulse shortening the orbit is given a phase lead and the energy of the electrons incident on the target is reduced. Fig. 16 shows the variation of the X-ray intensity as a function of their energy. For one of the curves an additional filter of \(1/4\) in. of lead was placed before the ionization chamber; for the other curve this lead was absent, and the only filter was 0.005 in. of aluminum. With such a thickness of the additional lead filter, the intensification factor compensates the absorption factor at 30 MV and above.

Fig. 17 is a radiograph taken with 100-MV X-rays. In this radiograph, obtained with the film at a distance of 20 ft from the target, a watch was photographed, placed on the side of a 4-inch steel plate facing away from the source of rays. The dark spot, вид-

Fig. 16. Logarithmic plot of the intensity of X-rays as a function of voltage.

The data were obtained with a direct-reading ionization chamber at the center of the beam, 150 cm from the target. The effective diameter of the chamber was 9.5 cm, or 3.6°. The intersection point of the two curves at 30 MV was obtained as follows: in front of the chamber a lead pendulum 0.250 dm thick was swung, and the voltage was selected at which the instrument readings remained unchanged.

… on the film was formed owing to the concentration of X-radiation in a cone with a half-width of 2° at 100 MV.

The output of X-rays at 100 MV, measured with a small ionization chamber “Victoreen 100 R,” placed in a \( \frac{1}{4} \)-inch lead

Fig. 17. Radiograph through 4 inches of steel at a distance of 20 feet from the target, obtained with 100-MV X-rays.

The watch was placed facing the source on the far side of the steel plate. The dark spot on the radiograph resulted from the concentration, at 100 MV, of the X-radiation coming from the accelerator in a cone with a half-width of 2°.

...housing at a distance of 550 cm from the target and reduced to the standard distance of 100 cm, amounted to as much as 2600 roentgens per minute.

XII. PULSE OPERATION FOR WORK WITH A WILSON CHAMBER.

With a Wilson chamber the machine must not operate continuously, since only individual short pulses are needed at approximately one-minute intervals.

For this purpose the machine was reconnected as shown in Fig. 18. The capacitor bank is divided in the middle; one half of it can be charged to a positive potential by one kenotron, the other—to a negative potential by a second kenotron, with the aid of one and the same high-voltage transformer. The oil circuit breaker serves to close the circuit, after which an oscillatory discharge begins with a time constant of 0.28 seconds. The oil circuit breaker does not have to open the circuit under current, so that here it was possible to use a standard oil circuit breaker rated at 13.8 kV, although the potential difference may be of the order of 40 kV before switching on.

Fig. 18. Circuit for pulse operation.

1. Magnetizing coils of the upper pole of the accelerator. 2. Magnetizing coils of the lower pole of the accelerator. 3. Capacitor bank. 4. Oil circuit breaker, closed by the force of gravity, opened by air. 5. Kenotrons. 6. Power high-voltage transformer.

A load weighing 250 pounds usually keeps the switch closed (which ensures continuous closure of the circuit). Compressed air, by means of a piston, raises the load, opening the switch. An air valve controlled by a solenoid releases the air at the required moment, and the switch closes. When the switch closes, a relay disconnects the charging circuit in order to avoid unnecessary charging of the kenotrons.

In view of the fact that all the other connections remain untouched, it is not necessary to switch any heavy wires or cables, and the machine can be transferred instantaneously from continuous operation to pulse operation and back by simple switching of a throw-over switch. In continuous operation the oil circuit breaker simply passes the current of the capacitor bank, 1000 amperes, at a frequency of 60 cycles.

For pulsed operation with a Wilson chamber, a number of variants are possible, such as:

1. The pulse circuit is controlled manually or by a motor, while the Wilson chamber is controlled by counters operating in coincidence. The advantage here is that the Wilson chamber is electrically independent of the accelerator and can be analogous to the chamber arrangement used in the study of cosmic rays.

2. The expansion mechanism of the Wilson chamber switches on the pulse circuit, and the various delaying electrical circuits make the chamber operate and provide light for photographic recording. In this case it is important to use some electrical pulse from the accelerator—for example, a pulse that displaces the orbit—in order to supply light at the appropriate time by means of a delaying electrical circuit.

Experience has shown that the injector circuit and the orbit-contracting circuit operate instantaneously both in pulsed and in continuous modes. This is partly due to the fact that the polarity of the charge on the capacitors is chosen so that the accelerating half-cycle is the second, not the first; during the first half-cycle the capacitors in the injector circuit and in the orbit-displacing circuit are charged. If no further precautions are taken, a train consisting of approximately 12 X-ray pulses is generated, of which only the first is needed, while the others may interfere with the operation of the Wilson chamber. Therefore a relay has been installed which interrupts the injector circuit after the first half-cycle, thus permitting only a single admission of electrons and one X-ray pulse.

The pulsed mode of operation is much quieter than the continuous one, since the machine operates for less than one second during a minute. The extremely low intensity of X-rays required for operation of the Wilson chamber, and the minute intervals between pulses, allow the experimenter working with the Wilson chamber to remain near it when this is necessary for visual observation and adjustment.

If, during pulsed operation, personnel are permitted to be present in the same room as the apparatus, then, because of the low noise during pulsed operation and the sudden application of high voltage, direct access to the electronic accelerator must be electrically blocked by switch-operated safety devices.

The authors gratefully express their appreciation to Dr. V. D. Kulidge for his continual interest and the many useful suggestions made in the course of this work. We are also indebted to many of our colleagues in the research laboratory for their valuable assistance during the construction of this new instrument for scientific research.

References

1. R. Wideröe, Ueber ein neues Prinzip zur Herstellung hoher Spannungen, Archiv f. Elektrotechn., 21, 1928, 387—406.
2. E. T. S. Walton, The Production of High-Speed Electrons by Indirect Means, Proc. Camb. Phil. Soc., 25 Pt. IV, October 1929, 469—481.
3. W. W. Jassinsky, Beschleunigung der Elektronen in elektromagnetischem Wechselfeld, Archiv f. Elektrotechn., 30, 590—603, 1936.
4. M. Steenbeck, U. S. Patent 2103.303 (February 28, 1937).
5. D. W. Kerst, Letter to the Editor: “Acceleration of Electrons by Magnetic Induction,” Phys. Rev., 58, 841, 1940.
6. M. Steenbeck, Beschleunigung von Elektronen durch elektrische Wirbelfelder, Naturwiss., 31, 234—239, May 7, 1943.
7. D. W. Kerst, The Acceleration of Electrons by Magnetic Induction, Phys. Rev., 60, 47—53, July 1941.
8. D. W. Kerst and R. Serber, Electronic Orbits in the Induction Accelerator, Phys. Rev., 60, 53—58, July 1941.
9. D. W. Kerst, Abstract. Induction Electron Accelerator, Phys. Rev., 59, 110, 1941.
10. D. W. Kerst, A 20-Million Electron-Volt Betatron or Induction Accelerator, Rev. Sci. Inst., 13, 387—394, 1942.
11. D. W. Kerst, The Betatron, Radiology, 40 (1943), 115—119; H. W. Koch, D. W. Kerst and P. Morrison, Experimental Depth Dose for 5, 10, 15 and 20 Million Volt X-rays, Radiology, 40, 120—127, 1943.