The current fell when the forward voltage rose
An ordinary forward-biased diode usually carries more current when its voltage is increased. Esaki’s heavily doped germanium junction produced a more surprising trace: current rose to a peak, fell through a range of increasing voltage, and then rose again. The declining branch was reproducible. It was therefore something to explain about the junction, not simply a defective point on a graph.
Leo Esaki investigated the effect in 1957; his short paper, received that October, appeared in Physical Review in January 1958. The essential experimental object was a narrow junction between heavily doped regions. The measurement was an electrical current-voltage characteristic, with temperature and junction preparation relevant to its shape. To understand what the curve established, imagine moving the voltage setting slowly and recording the resulting current after the circuit settles. The peak and valley are features of that entire trace. Merely reporting a large reverse current would have left several conventional leakage mechanisms in play.
Sources: [4][1]
Heavy doping changed both distance and occupation
Doping supplies donors or acceptors that alter the electron and hole populations. In an ordinary junction, a depletion region separates the conducting reservoirs. Heavy doping makes that region narrow and its internal electric field strong. A quantum electron has a wavefunction that extends into a classically forbidden region, so a sufficiently thin barrier can have appreciable transmission even when the electron lacks the classical energy to pass over it.
A rectangular-barrier estimate makes transmission fall exponentially as barrier thickness increases. Barrier width is a length, the effective mass describes the electron’s response in a material, and the barrier energy must exceed the incident energy in this approximation. This estimate retains the leading exponential dependence and omits matching prefactors. For an illustrative excess barrier of 0.2 electronvolt and effective mass one tenth of the free-electron mass, adding two nanometres reduces that exponential factor by about eighteen times. Those chosen numbers are a teaching calculation, not an extraction from Esaki’s sample.
Heavy doping does something else just as important: it moves the carrier reservoirs into the degenerate regime, where the positions of the chemical potential relative to the bands strongly affect which states are occupied. A thin barrier supplies transmission; the reservoir occupations decide which possible transitions can actually carry a net current.
Sources: [1]
A destination must be available at the right energy
An elastic tunneling event connects states at the same total energy on the two sides. An electron needs an occupied initial state and an unoccupied final state. The current is the imbalance between forward and reverse transitions. A useful schematic expression sums transmission, the densities of available states and the difference in their occupations over energy. The two occupation functions are Fermi distributions; the energy argument on each side must include its bias-dependent band alignment.
This sum over allowed transitions organizes the reasoning, not a complete quantitative germanium tunnel-diode calculation. Momentum selection, crystal structure and phonon-assisted transitions also matter; germanium has an indirect band structure. A drawing that merely slides two bands past each other suppresses those details. Its value is to show why “a higher voltage helps electrons through a barrier” is incomplete. Bias simultaneously changes transmission conditions and the overlap between occupied states in one reservoir and empty states in the other. It can improve one ingredient while reducing the other.
At zero applied bias, available states can support tunnelling in both directions, but equilibrium occupations make the opposite currents cancel. A thin barrier alone therefore does not supply a steady net current. Bias changes that balance and the alignment of available states. The voltage of the current peak is an alignment property of the junction, not a direct reading of the rectangular barrier height in the illustrative barrier estimate.
Sources: [1]
The peak, valley and later rise tell different stories
Near the beginning of forward bias, useful occupied-to-empty overlap grows and tunneling current increases. Beyond the peak, that overlap is reduced, so the tunneling contribution falls. At larger forward voltages, ordinary carrier injection and diffusion can become substantial, producing the later rise. The measured curve adds these channels together; its valley does not generally mark a perfectly forbidden transition with exactly zero current.
Temperature tests help separate the ingredients. Thermal broadening changes the sharpness of occupations, and phonons can assist transitions. Defects and imperfect interfaces can create additional routes. A peak-to-valley ratio consequently describes a specified device under specified conditions. It should not be treated as a universal number set only by the fact that tunneling exists.
This is also why a temperature-dependent current alone would have been weaker evidence than the organized pattern of a narrow, heavily doped junction and its characteristic peak, falling region and valley. The interpretation explains where along the curve a mechanism should dominate and how competing processes can fill in the expected minimum. It turns a puzzling electrical trace into a test of the electronic states on both sides.
Sources: [1]
Negative slope does not mean negative consumed power
The falling branch has negative differential resistance: a small increase of voltage gives a decrease of current. Differential resistance is the inverse slope of the local current-voltage curve. It differs from the static ratio of voltage to current, and it differs again from the total electrical power delivered to the device.
For a simple invented example, suppose the current falls from one milliampere at 0.1 volt to half a milliampere at 0.2 volt. The finite-difference resistance across that interval is minus two hundred ohms, but voltage and current remain positive at both endpoints. The device still consumes positive average electrical power. Biased at an appropriate operating point, such a descending characteristic can supply small-signal gain or counteract losses in an oscillating circuit, with the bias source providing the energy.
The surrounding circuit matters. Its load line can intersect several branches, and a voltage sweep may jump between operating points instead of tracing every point smoothly. Series resistance and measurement dynamics therefore belong in an interpretation of the data. An observed jump is not automatically an extra intrinsic feature of the tunneling spectrum.
Sources: [1]
The discovery made occupations part of device design
The 1973 award contribution shared by Esaki and Giaever joined two different uses of tunneling in solids. Esaki’s semiconductor junction connected a nonmonotonic device current to degenerate carrier reservoirs. Giaever’s metal-insulator-superconductor junction used a different electronic system to investigate superconducting excitation structure. Their common quantum process did not make their specimens or interpretations interchangeable.
The discovery connected a quantum transition with a device current. A thin junction enables transmission, overlap between occupied and empty states determines net transfer, and competing conduction processes take over at higher bias. Together they produce measurable peaks and valleys and make band alignment a usable form of electronic control.
Sources: [1][2][3]