Reverse Bias: Leakage Current, Breakdown Mechanisms

Every diode and every BJT/MOSFET junction spends part of its life reverse biased, and the assumption "reverse bias = no current" is only approximately true and only up to a voltage limit. Leakage current sets the noise floor and power budget for photodiodes, sample-and-holds, and battery-powered sensing front ends; breakdown voltage sets the absolute ceiling on how much voltage a junction can block before it fails or is deliberately used as a voltage reference. Getting both wrong is a common root cause of field failures in power supplies, ESD-exposed I/O, and precision analog front ends.

What Happens Physically Under Reverse Bias

Applying reverse bias (p-side negative relative to n-side) widens the depletion region and increases the built-in electric field. Majority carriers are pushed away from the junction, so the diffusion current that dominates forward bias collapses. What remains is a small drift current made of minority carriers that happen to wander into the depletion region and get swept across by the field.

The ideal Shockley diode equation predicts this saturation current directly:

I = I₀·(exp(V/(n·VT)) − 1)

For V negative and |V| >> VT (VT ≈ 25.85 mV at 300 K), the exponential term vanishes and I → −I₀. In an ideal diode, reverse current is therefore a constant, voltage-independent −I₀, called the reverse saturation current. In real silicon devices, measured leakage is orders of magnitude larger than this ideal I₀ and does depend on voltage — which is the first practical correction to make to the textbook model.

Sources of Real Leakage Current

Ideal diffusion current (I₀) is only one contributor. Real reverse leakage, I_R, sums several mechanisms:

Two rules of thumb matter for design:

• Temperature: leakage from diffusion and generation roughly doubles every 8–10°C for silicon, because both are proportional to some power of the intrinsic carrier concentration ni, and ni ∝ exp(−Eg/2kT). A junction leaking 10 nA at 25°C can leak several µA at 100°C.
• Voltage: generation current in the depletion region scales with depletion width, and depletion width scales with √V for a step junction, so leakage creeps up gradually with reverse voltage even well below breakdown — it is not perfectly flat as the ideal equation suggests.

Worked Example: Leakage Budget for a Photodiode Front End

A photodiode is reverse biased at 5 V and specified with I₀ = 2 nA reverse saturation current at 25°C, dominated by diffusion/generation (ignore surface leakage for this estimate). Estimate leakage at 65°C.

Step 1 — apply the doubling rule. ΔT = 40°C. Using a 9°C doubling time:

n_doublings = 40 / 9 ≈ 4.44
Factor = 2^4.44 ≈ 21.8

Step 2 — scale the leakage.

I_leak(65°C) ≈ 2 nA × 21.8 ≈ 43.6 nA

Step 3 — check against the transimpedance stage. If the photodiode feeds a transimpedance amplifier with Rf = 1 MΩ, this leakage alone produces an output offset of:

V_offset = I_leak × Rf = 43.6 nA × 1 MΩ ≈ 43.6 mV

Verification: at 25°C the same calculation gives 2 nA × 1 MΩ = 2 mV, so the offset has grown by the same ×21.8 factor as the current — consistent with a linear R and a temperature-scaled current source, confirming the arithmetic. A 43.6 mV drift at the output of a TIA meant to resolve microvolt-level signals is a real design problem, and it's purely thermal leakage, with no light incident — exactly the failure mode that pushes designers toward smaller-area photodiodes or guard-ring layouts to cut surface leakage.

Breakdown Mechanisms

As reverse voltage increases, the electric field in the depletion region rises until one of two mechanisms triggers a sharp increase in current — breakdown. Which one dominates depends on doping concentration and hence depletion width.

Avalanche breakdown (typically Vz > ~5–6 V, lightly to moderately doped junctions): the depletion region is wide, so an accelerated carrier gains enough kinetic energy over its mean free path to impact-ionize a lattice atom, creating a new electron-hole pair. Each new carrier is accelerated too, and can generate more pairs — a multiplying chain reaction. Current rises steeply, limited only by external resistance. Avalanche breakdown voltage has a positive temperature coefficient: higher temperature means more lattice scattering, shorter mean free path, so carriers need a higher field (higher voltage) to gain enough energy between collisions — Vz increases with T.

Zener (tunneling) breakdown (typically Vz < ~5–6 V, heavily doped junctions): the depletion region is very thin (heavy doping compresses it) and the field is extremely high (>~10⁶ V/cm) at modest voltage. Electrons tunnel directly from the valence band on the p-side to the conduction band on the n-side through the thin barrier — a purely quantum-mechanical effect, no impact ionization needed. Zener breakdown voltage has a negative temperature coefficient: higher temperature narrows the bandgap slightly, making tunneling easier at a lower field, so Vz decreases with T.

Real "Zener diodes" rated near 5–6 V often show both mechanisms simultaneously, which is why that voltage region has the flattest (near-zero) temperature coefficient — a useful property exploited deliberately in some voltage reference designs.

Practical Design Implications

• Derate breakdown voltage with margin. Vz shifts with temperature (in either direction depending on mechanism), and manufacturing spread on breakdown voltage is typically several percent — never design a rail to sit within a few percent of a device's rated VBR.
• Reverse leakage sets minimum bias current in current-mode circuits. In precision integrators, sample-and-holds, and photodiode TIAs, leakage directly adds offset/drift; choose devices with guard rings or smaller junction area where sub-nA leakage matters.
• Self-heating can create thermal runaway in avalanche devices. Because avalanche current is not inherently self-limiting, a reverse-biased junction dissipating I·V without adequate external limiting resistance can destructively escalate; always budget power dissipation, not just voltage, when a device operates in breakdown (e.g., TVS diodes, Zener regulators).
• Junction area and edge design shape leakage more than the ideal I₀ equation suggests. Perimeter/surface leakage often dominates ideal diffusion current in small-geometry devices, so datasheet leakage at your actual operating voltage and temperature — not the ideal Shockley prediction — is the number to design against.
• ESD and transient overvoltage exploit avalanche breakdown deliberately. TVS diodes and avalanche-mode protection devices are sized so avalanche onset clamps a transient safely below the protected circuit's damage threshold, with the clamping energy absorbed as I·V·t in the device.

Key Takeaways

• Reverse bias current is not zero: it's the sum of ideal diffusion (I₀), depletion-region generation current, surface/perimeter leakage, and (in thin/heavily-doped junctions) tunneling — real leakage exceeds the ideal Shockley prediction.
• Leakage roughly doubles every 8–10°C in silicon, tracking ni²; always check leakage-driven offset/drift at your worst-case operating temperature, not just at 25°C.
• Avalanche breakdown (wide depletion, impact ionization, Vz typically > ~6 V) has a positive temperature coefficient; Zener/tunneling breakdown (narrow depletion, quantum tunneling, Vz typically < ~6 V) has a negative one.
• Near Vz ≈ 5–6 V both mechanisms coexist, giving the flattest temperature coefficient — the basis for low-drift Zener references.
• Design margin on breakdown voltage must account for temperature coefficient direction, unit-to-unit spread, and power dissipation in breakdown, not just the nominal rated voltage.