New Instruments and Methods of Measurement

Semiconductor Triodes Without Point Contacts

V. S. Vavilov

1. Introduction

In the course of the last three years, a new practically important field of application of semiconductors has been developing rapidly: the amplification of alternating currents by crystalline triodes. As is known, the first of such triodes were “transistors” with point contacts metal–germanium (see, for example, ¹˒²). A number of substantial shortcomings made the field of application of these triodes rather narrow. In addition, the theory of operation of triodes with point contacts has up to the present time not gone beyond qualitative estimates and considerations only partially confirmed by experiment (for example, ³).

The principal shortcomings of crystalline triodes with point contacts are: a very high noise level, which made it impossible to use them as amplifier elements for radio reception; low useful power, due to the negligible area of the “point” contact; and a maximum operating frequency limited to several megacycles.

Below, the principle of operation is described and data are presented for germanium triodes operating without point contacts and representing, in comparison with those described earlier, a major step forward.

2. Principle of Operation of the New Type of Triode

The schematic diagram of a contactless crystalline triode is shown in Fig. 1. The sharp boundaries, artificially created in the semiconductor, between regions of electron \((n)\) and hole \((p)\) conductivity constitute rectifying elements in electrical circuits between the emitter and the base (base electrode) and between the collector and the base. In triodes of type “A” with point contacts, the corresponding role was played by the metal–semiconductor boundaries.

Unlike crystalline triodes of type “A,” rectification thus takes place inside the crystal, and not at its surface, in the blocking layer near the metallic point. All three connections (leads) from the germanium crystal are non-rectifying ohmic contacts of large area, deposited electrolytically.

Another essential difference between triodes of the “\(n — p — n\)” type and triodes of type “A” is that, first, the flow of current carriers injected into the germanium propagates mainly

Fig. 1.

by diffusion, and not by drift in an external electric field. “Multiplication” of the current at the collector (see \(^{1,2}\)), leading to positive feedback and instability in triodes of type “A,” does not occur in “contactless” triodes.

At the bottom of Fig. 1 are shown the energy zones in a germanium crystal divided by a layer with the opposite type of conductivity. In the left-hand drawing the applied external potentials are equal to zero; in the right-hand drawing the applied bias potentials turn the crystal into a “solid-state amplifier.”\(^{5}\)

A bias corresponding to current in the reverse direction is applied to the rectifying boundary \(J_c\) in this case. This direction is such that electrons in the “collector” region of type \(n\) have a small potential energy and cannot overcome the potential barrier present at the boundary with the base (the region with hole conductivity). Similarly, holes are retained in the central layer. Electrons in the emitter region, however, have the possibility of overcoming the low potential barrier and penetrating into the layer with hole conductivity. Having entered

into the central region, the electrons diffuse to the right; some of them reach the right-hand \(p—n\) boundary \((J_c)\). The flow of electrons through the potential barrier depends on the height of the barrier, which can change at the expense of the emitter bias potential while the base potential remains constant.

If the layer with hole conductivity is very thin, only a small part of the electrons will have time to recombine with holes while passing through it (see \(^{6,7}\)). The electron current will readily pass through the layer. Moreover, if the emitter region has a conductivity greater than that of the base, the number of electrons overcoming the potential barrier from left to right will exceed the number of holes overcoming the corresponding barrier in the opposite direction. Let us recall that in Fig. 1 the energies of the holes are plotted downward. The greater part of the current through the left-hand boundary \(J_e\) will be carried by electrons.

In this case the processes occurring in a semiconductor crystal with an interlayer having the opposite-to-the-main sign of the current carriers are, to a high degree, analogous to what takes place in a vacuum triode. The emitter corresponds to the cathode, the base to the region near the control grid, and the collector to the anode. In the case of a successfully chosen triode geometry and the required ratio of conductivities, the controlled electron current penetrating through the base may considerably exceed the signal current that changes the height of the potential barrier. The current amplification factor in such a triode is very high. In view of this, a crystal triode of the \(n—p—n\) type can be used as a triode with a grounded cathode. Here, in contrast to Fig. 1, the emitter is grounded, and the signal is applied to electrode \(b\).

Both in vacuum triodes and in crystal triodes of the type described, control is effected by the interaction of two kinds of electron flow. The conduction current in the metal of the grid of a vacuum triode controls the flow of electrons between the grid wires, caused by thermionic emission from the cathode. In the crystal triode, the flow of holes in the base changes the potential difference between the emitter and the base and controls the flow of electrons through the layer having hole conductivity.

Before proceeding to concrete data on the operation of the new type of crystal triodes, it should be pointed out that germanium crystals with a layer of conductivity of the opposite sign can also be used as a kind of “photomultiplier.” In this case the electrode connected to the base plays no role. As is known, photoconductivity occurs in germanium in the visible and infrared regions \(^{1,8}\). If light falls on the surface of the crystal near one of the \(n—p\) boundaries, the electrons and holes \(^{8,9}\) released by it are separated by the electric field, as a result of which a hole current flows into the base.

These holes, accumulating in the central layer, charge it positively and thereby lower the potential energy of the electrons in it. As a result, a larger number of electrons overcome the potential barrier and flow into the collector region.

After the illumination of the crystal ceases, the accumulated charge rapidly disappears as a result of the diffusion of holes beyond the limits of the \(p\) region and their recombination with electrons, and also through recombination in the layer itself.

If the layer with \(p\)-conductivity is very thin, and the electron density in the emitter region is large, a considerable number of electrons (in comparison with the number of holes penetrating to the emitter and recombining) will be able to overcome the potential barrier and pass to the collector. The current amplification, as shown in \(^{5}\), is proportional to the ratio of the conductivities of the two layers and inversely proportional to the thickness of the hole layer.

3. EXPERIMENTAL SAMPLES OF CONTACTLESS TRIODES

A triode of the “\(n—p—n\)” type is a single-crystal germanium plate, about 5 mm long and with a cross-section of the order of a square millimeter, with three leads (wires) mechanically firmly attached to the surface of the germanium \(^{10}\). The triode is enclosed in a strong plastic casing protecting the fragile surface of the crystal. The most interesting question, which according to the literature remains unclear, is the method by which the interlayer with hole conductivity is created. It is known (see, for example, \(^{11,12}\)) that germanium can be made a hole semiconductor by nuclear bombardment with alpha particles or fast neutrons. Nuclear reactions in the body of the crystal lead to the formation of “impurity centers” (atoms of a neighboring element in the periodic system), which, by capturing electrons from the filled energy band, cause the appearance of charge-carrying “holes.” In the case of sharply bounded beams of charged particles of sufficient energy, obtaining regions of the crystal of the required configuration with the changed sign of conductivity appears quite possible. Another method for obtaining sharp \(n—p\) boundaries lying in a plane is the introduction into the melt, from which a single crystal is grown by slow pulling, of a certain amount of the required impurity (for example, gallium \(^{13}\)). In one of the works devoted to the rectifying properties of \(n—p\) boundaries in germanium \(^{14}\), it is indicated that this method gave good results. Finally, the possibility is not excluded of obtaining the crystals shown in Fig. 1 by fusing together two pieces of “elec-

“tron” germanium, on the surface of one of which a layer of \(p\)-conductivity has been created. The degree of purity of the germanium used for triodes is extremely high: according to\(^6\), the lifetimes of holes in the collector region reached 300–400 microseconds. The experimental triodes were made with different conductivity values; according to the authors’ data, typical values of the conductivity in different parts of the crystal were: in the emitter region—\(100\ \Omega^{-1}\,\text{cm}^{-1}\); in the base—\(1\ \Omega^{-1}\,\text{cm}^{-1}\); in the collector region \(0.1\ \Omega^{-1}\,\text{cm}^{-1}\).

4. CHARACTERISTICS OF TRIODES

The static characteristics of an experimental crystal triode are shown in Fig. 2. Curves of this kind are obtained in the following way: the corresponding current sources are connected into the emitter and collector circuits; potential differences corresponding to various current values are measured. Currents are considered positive if they flow into the emitter and collector, as shown in Fig. 3. Potentials are considered positive if their signs correspond to Fig. 3. The part of the characteristics corresponding to the normal amplification regime of the triode is shown by solid lines. From Fig. 2 it is clear what currents and potential differences must occur in order for the circuit to operate as an amplifier. For the triode whose characteristics are given in Fig. 2\(^ {10}\), any positive bias potentials on the collector within approximately \(0.1\) to \(35\ \text{V}\) lie in the operating region. The emitter currents on all curves are negative, i.e., the current flows from the emitter into the source connected between the emitter and the base. The collector currents corresponding to the values of \(I_e\) plotted along the abscissa axis are almost equal to \(I_e\). Since \(I_c\) and \(I_e\) are opposite in sign, this means that the greater part of the current flowing into the collector leaves through the emitter, while the current in the base circuit is very small.

Let a constant positive potential be applied to the collector, for example by connecting a battery into the circuit between \(c\) and \(b\) (Fig. 1). If a “negative” current is made to enter the emitter (a battery and a series resistance are also connected), then the collector current can be controlled by the emitter current. In the given sample of the triode the current could vary from \(20\ \mu\text{A}\) to \(5\ \text{mA}\). The choice of the value of the potential and of the collector current is determined by the power that it is necessary to obtain from the triode. The greater this power, the greater the current and the potential difference that must be supplied by the source. Since the efficiency in class \(A\) operation (in the linear part of the characteristic, without cutoff)\(^ {15}\) cannot exceed the theoretical limit of 50%, the power of the amplified signal cannot be greater than half the pow-

of the power delivered by the battery. For example, with the collector operating at 20 V and 2 mA, the output power (for an undistorted sine wave) cannot be more than 20 mW.

From the lower curves of Fig. 2 it is seen what potential differences must be applied in the emitter circuit. The entire range of variation of \(U_e\) corresponds to only 0.4 V; the working region of the characteristics corresponds to only a few millivolts. Thus, at a constant value of \(U_c\), very small changes in \(U_e\) will cause a significant change in the collector current; if, however, the collector current is kept constant, the changes in \(U_c\) will many times exceed the change in emitter potential. Therefore the crystal triode can be used as an amplifier connected between a current source with small

Fig. 2. Static characteristics of a crystal triode of type \(n\)—\(p\)—\(n\).
a) — at comparatively large currents; b) — at very small currents.

resistance and a load having high resistance; according to^10, the voltage gain in this case may reach 10,000.

When amplifying alternating current, a battery with a series resistance may be connected as the source of the emitter “bias” current.

Fig. 3.

If the bias exceeds the potential drop in the emitter circuit, equal to several hundredths of a volt, the current \(I_e\) is almost equal to the battery voltage divided by the series-connected resistance.

5. COEFFICIENT OF USEFUL ACTION

(EFFICIENCY)

In the case of large signal amplitudes at the input and an ohmic load, the operating point will move along a straight line (see the upper curves of Fig. 2, a). The load line in this case corresponds to \(10\,000\) ohms in the external circuit. Since the “collector” characteristics are nearly equally spaced straight lines, the instantaneous values of \(U_c\) may vary within limits almost from 0 to \(2U_c\) (for example, at a collector voltage of 20 V, from 0.1 to 39.5 V, which corresponds to an efficiency of 48.5% out of the 50% theoretically possible). In some specimens of triodes the efficiency may be still greater. The efficiency calculations were based on the assumption of sinusoidal current oscillations in the emitter circuit. However, the emitter resistance depends somewhat on the current \(I_e\); in view of this, the current source in the emitter circuit, in order to obtain high efficiencies and small distortions, must have a large resistance.

When operating in the region of small-amplitude signals, crystal triodes give sufficient gains at much lower currents and voltages than in the example given above. This is explained by the fact that, also in the region of very small potentials and currents, the characteristics remain rectilinear (Fig. 2, b). For operation with small signals, \(U_c\) of about 0.1 V and a col-

collector, slightly exceeding 10 μA. The power expended in this case amounts to only a few microwatts. This power is several thousand times less than the power expended in heating the cathode in a vacuum amplifier tube.

6. AMPLIFIER CIRCUIT WITH GROUNDED EMITTER

The possible variants of amplification circuits and of connecting crystal triodes in multistage circuits are quite numerous. The most advantageous appears to be the circuit with a grounded emitter (Fig. 4). The power gain in this case reaches 50 dB. The problem of connecting stages is simplified by the fact that the input resistance is somewhat higher than in circuits with a grounded base, while the output resistance is considerably lower (respectively, several hundred ohms and several hundred kilohms). A gain of the order of 30 dB per stage is possible without the use of interstage step-down transformers.

Suppose that the collector is connected through a transformer winding to a DC voltage source (Fig. 4, a). Since \(U_c\) amounts to fractions of a volt, the collector potential will be almost equal to the external voltage. If no DC bias is applied to the base, the base is at a potential with respect to ground whose absolute value is equal to \(U_e\), and the collector current is exactly equal to the emitter current.

Fig. 4. a) one of the circuits of an amplifier with a grounded emitter; b) modification of circuit a for small collector currents; c) modification of circuit a for large collector currents.

In some cases of triode application, when it is necessary to reduce the consumed power, a modification of the circuit is possible (Fig. 4, b). A resistor is connected between the base and ground; since the self-established base potential relative to ground is positive, the current in the base circuit has the necessary direction in order to reduce the collector current. To increase the collector current, the circuit of Fig. 4, c, may be used, in which

between the positive terminal of the batteries and the base. Other variants of circuits for connecting the triode, such as, for example, a cascade with a grounded collector, which are of interest for practical applications, are considered in \(^{10}\).

7. DEPENDENCE OF TRIODE OPERATION ON FREQUENCY AND NOISE LEVEL

From the theory of crystalline triodes of the \(n—p—n\) type it follows \(^{5}\) that the limiting signal frequency at which amplification is possible depends on the geometry of the triode and on the physical properties of the germanium used for its manufacture.

As follows from the principle of operation of the triode, electrons entering the layer with \(p\)-conductivity from the emitter side pass through this layer by diffusion; a certain (very small) time is spent in traversing the corresponding path. If all electrons passed through the \(p\)-layer in one and the same time, this would cause only a delay of the output signal relative to the input signal. In reality, there is always a spread in the times taken by electrons to pass through the layer; if this spread becomes comparable with the period of the signal, the amplification falls. The frequency \(f_{ca}\), at which the diffusion times of the electrons begin to have an effect, is inversely proportional to the square of the thickness of the \(p\)-layer and, consequently, increases rapidly as the layer thickness is reduced. In the experimental triodes described in \(^{5}\) and \(^{10}\), \(f_{ca}\) reaches \(5—20\) mc.

Another cause affecting the operation of crystalline triodes at high frequencies is the capacitance of the \(l_e\) boundary. It can be shown that the capacitance of the emitter—base boundary limits the operating frequency range by the frequency \(f_{ce}\), approximately equal to \(f_{ca}\).

A third independent cause limiting the range of operating frequencies is the capacitance of the base—collector boundary. The \(n\)-type germanium on one side of it and the \(p\)-type germanium on the other behave as the plates of a capacitor. Since the boundary thickness is very small, the equivalent capacitance may reach a considerable value.

If the triode is used as an amplifier in a circuit with a grounded emitter, then, for the specimens described, by selecting optimal circuit parameters it was possible to obtain the following amplification factors: at 11 kc — 50 db, at 100 kc — 41 db, and at 940 kc — 31.4 db. Thus the operation of triodes as amplifiers in the radio-frequency range is entirely possible.

The noise level of “transistors” of type “A” with point contacts, as is known, is approximately 1000 times higher than

in modern vacuum triodes.^{2,3} Triodes of the \(n—p—n\) type have a noise level several orders of magnitude lower. In individual specimens, according to the authors’ data, the noise level in a \(1000\ \text{cps}\) band ranged from \(8\ \text{db}\) to \(25\ \text{db}\), i.e., in some cases it was close to the noise level in modern receiving-amplifying pentodes. As in crystalline triodes of type “A,” the noise level decreases with frequency (approximately by \(11\ \text{db}\) when the frequency is increased by a factor of 10). With a decrease in the thickness of the \(p\)-layer, the noise level also decreases. In the first approximation the observed noise level agrees with the theory of the origin of noise caused by recombination processes.^{16} According to this theory, each volume element is a noise source, the square of the “noise current” being proportional to the square of the deviation of the concentration of “injected” current carriers from its normal value. As applied to triodes of the \(n—p—n\) type, the roughly hundredfold decrease in noise level in comparison with transistors of type “A,” according to this theory, is explained by the difference in the geometry of the triode and in the current density in the semiconductor.

8. CONCLUSION AND FINDINGS

In conclusion, it seems useful to us to summarize the principal properties of the new triodes and to give some additional data on them not mentioned above:

1. A relatively low noise level: in most triodes the noise level does not exceed \(10—20\ \text{db}\) in a \(1000\ \text{cps}\) band.

2. High stability of operation. The input and output resistances of the triodes are always positive, both in circuits with a grounded emitter and in the case of a grounded base or collector. This gives complete freedom in choosing the most convenient connection circuit and makes it possible to vary over a wide range the input and output resistances of the amplifier stages.

3. Large amplification factors, reaching \(100—300\) per stage.

4. The possibility of amplifying signals with a power of at least up to \(2\ \text{W}\).

5. High efficiency. Owing to the fact that the static characteristics are very close to ideal ones, in amplification in class A it is possible in practice to obtain an efficiency of \(48—49\%\) out of the \(50\%\) theoretically possible. In amplification with cutoff (classes B and C) the efficiency can be still higher.

6. Mechanical strength and small dimensions. The germanium crystal of the triode is enclosed in a solid “bead” about \(5\ \text{mm}\) in diameter, into which enter 3 wires firmly fastened to the surface of the crystal.

1. Absence of a microphonic effect under shaking.

2. A limited upper range of operating frequencies. The collector capacitance limits the region of the maximum attainable amplification factors at audio frequencies. However, with a proper choice of the circuit parameters, at the cost of some reduction in gain it is possible to obtain a flat frequency response at least up to 1 MHz.

3. Unusually low power consumption by the supply sources. As an example, in [10] an audio-frequency generator is mentioned which requires 6 μA for its supply at a potential difference of 0.1 V, i.e., a power of 0.6 μW. Let us recall that heating the cathode of an ordinary receiving-amplifying tube consumes a power of a million microwatts or even more.

References

1. “Crystal Detectors,” Soviet Radio, 1950, supplement to the head of Chapter V (A. B. Rzhanov).
2. V. S. Vavilov, UFN 40, 120 (1950).
3. Bardeen and Brattain, Phys. Rev. 75, 1208 (1949).
4. W. Shockley, Bell Syst. Techn. Journal 28, 435 (1949).
5. W. Shockley, M. Sparks, G. Teal, Phys. Rev. 83, 151 (1951).
6. W. Shockley et al., Bell Syst. Techn. Journal 28 (1949).
7. W. Shockley, Pearson and Haynes, Phys. Rev. 81, 835 (1951).
8. F. F. Volkenshtein, Electrical Conductivity of Semiconductors. Gostekhizdat (1947).
9. J. N. Shive, Phys. Rev. 76, 575 (1949).
10. R. Wallace and W. Pietenpol, Proc. Inst. Radio Engineers 39, 753 (1951).
11. Lark-Horovitz et al., Phys. Rev. 82, 763 (1951).
12. K. McKay, Phys. Rev. 82, 329 (1951).
13. Teal, Sparks and Buehler, Phys. Rev. 81, 637 (1951).
14. W. Pietenpol, Phys. Rev. 82, 120 (1951).
15. V. F. Vlasov, Electrovacuum Devices. Svyazizdat (1949).
16. W. Shockley, Electrons and Holes in Semiconductors, N. Y., 1950.