Course instructor: Abhijeet L. Sangle
(alsangle@iitb.ac.in, +91 22 2159 6742)
Last lecture ended at the ideal diode equation – the current from carrier diffusion alone.
Real junctions add several effects on top of this ideal picture:
Assumed abrupt depletion approximation, Maxwell–Boltzmann statistics, low injection, and constant total current through the structure.
Under these assumptions the current is carried entirely by minority-carrier diffusion outside the space-charge region.
This ideal current varies as
Real diodes deviate from this at low forward bias, at high forward bias, and under reverse bias – we quantify each next.
The ideal diode equation derived previously relied on strict assumptions:
We model real diodes by breaking these assumptions:
We will analyze Generation and Recombination in the space-charge region, and the impact of High-Level Injection.
Under reverse bias, mobile carriers are rapidly swept out of the space-charge region, meaning \( n \approx 0 \) and \( p \approx 0 \).
However, thermal energy continually generates electron-hole pairs via mid-gap defect states (traps).
Because the region is depleted, these newly generated pairs do not recombine. The strong electric field sweeps them apart.
This continuous generation and sweeping action creates an additional current component: the reverse-biased generation current (\(J_{gen}\)).
From the Shockley-Read-Hall (SRH) recombination theory, setting \(n=p=0\) and assuming a trap at midgap (\(n'=p'=n_i\)):
Defining an effective lifetime \(\tau_0 = (\tau_{p0} + \tau_{n0})/2\), the generation rate is \(G = \frac{n_i}{2\tau_0}\).
The generation current density is the integral of \(G\) over the space charge width \(W\):
Since \(W \propto \sqrt{V_{bi} + V_R}\), \(J_{gen}\) is not constant, but increases slightly with reverse bias.
The total reverse current combines ideal saturation (\(J_s\)) and generation (\(J_{gen}\)).
Silicon (Room Temp):
\(n_i\) is small (\(\sim 1.5 \times 10^{10} \text{ cm}^{-3}\)). Thus, \(J_{gen}\) is typically 3-4 orders of magnitude larger than \(J_s\). Generation dominates!
Germanium (Room Temp):
\(n_i\) is much larger (\(\sim 2.4 \times 10^{13} \text{ cm}^{-3}\)). Here, \(n_i^2\) overtakes \(n_i\), so \(J_s\) dominates.
Under forward bias, electrons and holes are injected into the space-charge region from opposite sides.
Because the region now has a high concentration of both carriers, some fall into mid-gap traps and recombine before crossing the region completely.
This lost charge must be replenished by the external circuit, creating an extra current component: the forward-bias recombination current (\(J_{rec}\)).
It is most significant at low forward bias, where the ideal diffusion current is still small.
Maximum recombination occurs at the metallurgical junction, where the product \(np\) is maximized and \(n = p = n_i \exp(eV_a/2kT)\).
Substituting this into the SRH equation, the maximum recombination rate is:
Integrating this maximum rate over the effective width \(x'\) of the depletion region gives the current density:
The total forward current is the sum of diffusion and recombination currents:
We often approximate this total current with a single exponential using the ideality factor \(n\):
The ideal diode theory assumed low-level injection: injected minority carriers are much fewer than the majority carriers (\(\delta p \ll n_0\)).
At very high forward bias, the injected minority carrier concentration approaches or even exceeds the equilibrium majority carrier concentration.
To maintain space-charge neutrality in the bulk region, the majority carrier concentration must also increase significantly to match the excess minority carriers!
Under high-level injection, the excess carriers dominate, so \(n \approx p\).
Using the fundamental relationship \(np = n_i^2 \exp(V_a/V_t)\), we get:
Since the diode current is proportional to the carrier concentration, the current becomes limited by this new relation:
Notice that the slope on a \(\ln(I)\) vs \(V_a\) plot returns to \(e/2kT\), making the ideality factor \(n \approx 2\) once again.
The real diode I-V curve exhibits three distinct regions in forward bias:
*Note: At the extreme high end, series resistance causes the curve to flatten further.
In many linear amplifier circuits, semiconductor devices are biased with a steady DC voltage, and a small sinusoidal signal is superimposed on it.
This allows us to analyze the device using a linear Small-Signal Equivalent Circuit.
The total applied voltage is:
Because the signal is "small", we can use local linear approximations (derivatives) of the non-linear I-V curve.
The Core Concept:
The ideal current-voltage relationship is:
The small-signal incremental conductance is the slope of the dc current-voltage curve evaluated at the quiescent point (\(V_0\)):
Assuming \(V_0 \gg kT/e\), the \(-1\) term is negligible, yielding:
The diffusion resistance is the inverse: \( r_d = \frac{V_t}{I_{DQ}} \).
The reciprocal of the incremental conductance is the incremental resistance (or diffusion resistance):
When an ac voltage \(v_1(t)\) is superimposed on the dc bias, the minority carrier concentration at the space-charge edge also fluctuates.
For holes injected into the n-region:
Let's look at the hole concentration \(p_n(x,t)\) in the n-region.
The dashed line represents the steady-state DC distribution.
The oscillating solid line represents the total concentration swinging with the AC voltage.
The shaded region represents the AC charge \(\Delta Q\) that must be physically moved in and out of the neutral region during each cycle.
The hole concentration at \(x=0\) is driven by the total voltage \(V_0 + v_1(t)\):
If the AC signal is small, \(v_1(t) \ll V_t = kT/e\). We can expand the exponential using a Taylor series, retaining only linear terms:
Assuming a sinusoidal signal \(v_1(t) = \hat{V}_1 e^{j\omega t}\):
We use the time-dependent diffusion equation (E-field = 0 in neutral region):
We expect a solution of the form: \(\delta p_n(x,t) = \delta p_0(x) + p_1(x)e^{j\omega t}\).
Substituting this in, the AC component must satisfy:
Let \(C_p^2 = (1 + j\omega\tau_{p0})/L_p^2\). The general AC solution is \(p_1(x) = K_1 e^{-C_p x} + K_2 e^{C_p x}\).
For a long diode, \(K_2 = 0\), and matching our boundary condition at \(x=0\), we get:
The AC hole diffusion current phasor at \(x=0\) is found by taking the derivative of \(p_1(x)\):
Substituting \(C_p\) and multiplying by Area (\(A\)), the total AC hole current phasor is:
Doing the same for electrons gives \(\hat{I}_n\). The Admittance \(Y = (\hat{I}_p + \hat{I}_n) / \hat{V}_1\):
The admittance expression is complex and difficult to synthesize into simple circuit elements.
If we assume the frequency is not too high, \(\omega \tau_{p0} \ll 1\). We can expand the square root using Taylor series:
Substituting this back into \(Y\), we can separate it into real and imaginary parts:
Where \(g_d\) is the diffusion conductance and \(C_d\) is the diffusion capacitance.
Extracting the imaginary part, we find the diffusion capacitance is:
To complete the small-signal model, we must add two more physical realities:
The total admittance of the actual junction is \((g_d + j\omega C_d + j\omega C_j)\), placed in series with \(r_s\).
The total voltage applied to the diode is the voltage across the junction \(V_a\) PLUS the drop across the series resistance \(r_s\):
In most diodes, \(r_s\) is negligible. However, at high currents, the \(I \cdot r_s\) drop becomes significant.
The actual measured current will be less than the ideal exponential model predicts for a given applied voltage \(V_{app}\).
This causes the I-V curve to bend horizontally at high currents.
The stored minority charge that gives rise to \(C_d\) also affects how quickly a diode can be switched from forward to reverse bias.
Consider a diode carrying forward current \(I_F\), suddenly switched at \(t=0\) to a reverse bias.
The stored charge can't disappear instantly: it must first be removed (by reverse current and recombination) before the junction can support the reverse voltage.
This delay shows up directly in the diode current waveform.
Immediately after switching, the diode current jumps to \(-I_R\) and stays there while stored charge is being swept out.
This interval is the storage time \(t_s\), after which the excess minority-carrier concentration at the junction edge reaches zero:
After \(t_s\), the current decays back toward \(-I_R\)'s steady leakage value.
Switching on is faster than switching off, but it isn't instantaneous either.
Two things must happen:
Until both are complete, the terminal voltage and current are still settling toward their final dc values – this sets an upper limit on switching speed.
If both p and n regions are doped extremely heavily (degenerately), the Fermi level moves into the conduction band on the n-side and into the valence band on the p-side.
The depletion width becomes extremely narrow (tens of Å).
At equilibrium, the Fermi level is constant across the junction. Filled states in the n-conduction band face filled states in the p-valence band, and empty states face empty states, resulting in zero net tunneling current.
Applying a forward bias aligns filled n-states opposite to empty p-states, allowing quantum mechanical tunneling.
At small forward bias, filled states line up with empty states – tunneling current rises sharply to a peak current (\(I_p\)) at voltage \(V_p\).
As bias increases further, the bands slide out of alignment and current falls to a valley current (\(I_v\)) at voltage \(V_v\).
This falling region is a region of negative differential resistance, useful for oscillators and fast switching.
At still higher forward bias, normal diffusion current takes over and the curve rises again.
Under reverse bias, filled valence-band states on the p-side now line up with empty conduction-band states on the n-side.
Tunneling can occur immediately, with no threshold voltage, and the reverse current rises rapidly and monotonically with \(|V_R|\).
This large, smoothly increasing reverse current makes the tunnel diode useful in some microwave and switching applications.
Generation current (reverse) and recombination current (forward) arise from carrier statistics inside the depletion region itself.
The ideality factor n (1 to 2) captures which mechanism dominates; high injection bends the curve at large forward bias.
The small-signal model adds a diffusion resistance \(r_d\) and diffusion capacitance \(C_d\), set by dc bias.
Switching is not instantaneous: stored minority charge must be removed (\(t_s\)) or built up first.
Degenerate doping turns the pn junction into a tunnel diode with a negative differential resistance region.