IGBT: Structure, Characteristics, and Applications

Motor drives, traction inverters, welding equipment, and induction heating systems all need a switch that combines MOSFET-like gate control with bipolar-like conduction efficiency at high voltage and current. Silicon power MOSFETs run out of steam above a few hundred volts because RDS(on) rises steeply with breakdown voltage, and bipolar power transistors need large, slow base drive current. The Insulated Gate Bipolar Transistor (IGBT) fills that gap: it switches with a MOSFET gate but conducts through a conductivity-modulated bipolar path, which made it the default silicon device for 600 V–6.5 kV switching, from a few amperes to several kiloamperes per module. Silicon carbide MOSFETs now compete with it from roughly 650 V to 1700 V.

Structure: MOSFET Gate Driving a Bipolar Output Stage

An IGBT is a four-layer (n+ p n− p+), three-terminal device that merges a MOSFET input structure with a PNP bipolar output structure on the same die.

Emitter (E)          Gate (G)
   |                    |
 [n+]---[p-well]---[poly gate over oxide]
        |     \___channel___/
        |
      [n- drift region]   <- sets blocking voltage
        |
      [p+ substrate/collector]
        |
      Collector (C)

• Gate: insulated by a thin oxide, exactly like a MOSFET — draws no steady-state DC current, only capacitive charge during switching.
• n+ emitter region and p-well: form the MOSFET channel at the surface of the p-well, which inverts when VGE exceeds the threshold VGE(th).
• n- drift region: wide, lightly doped layer that supports the blocking voltage; its thickness and doping set the device's voltage rating, just as in a power MOSFET.
• p+ collector: injects holes (minority carriers) into the drift region during conduction. The injected charge floods the drift region and drastically lowers its resistance — conductivity modulation — which is why the on-state drop stays low even at high current density.

Functionally, the IGBT is a MOSFET supplying the base current of a wide-base PNP transistor. In equivalent-circuit terms:

IGBT collector C  ── PNP emitter (p+ collector layer)
PNP base          ── n- drift region ── MOSFET drain
PNP collector     ── p-well ── IGBT emitter E
MOSFET source     ── IGBT emitter E
MOSFET gate       ── IGBT gate G

An unavoidable byproduct of this structure is a parasitic NPN formed by the n+ emitter, p-well, and n- drift region, which combines with the main PNP to form a parasitic thyristor. If hole current flowing through the p-well resistance under the n+ emitter forward-biases that NPN, the thyristor latches and the gate loses control of the current — this is latch-up. Modern IGBTs suppress it with a heavily doped p+ region under the emitter and a tight emitter–body short, so it does not occur within the datasheet's safe operating area (RBSOA and short-circuit ratings); excessive current or very fast turn-off outside those limits can still trigger it.

Static I-V Characteristics

The output characteristic (IC vs VCE for fixed VGE) looks like a MOSFET family of curves shifted right by a diode-like offset of about 0.7 V, from the p+ collector junction:

• Cutoff: VGE < VGE(th) (typically 4.5–6.5 V) — collector current is essentially zero up to the rated blocking voltage.
• Active region (MOSFET-style naming calls it "saturation", the opposite of BJT usage; see BJT operating regions): at high VCE, IC is set mainly by VGE, through the MOSFET's transconductance multiplied by the PNP's gain. Switching designs pass through this region only during transitions — and during a short circuit, where it limits current to several times the rated value.
• On-state (fully on): with VGE = 15 V, VCE(sat) at rated current is typically 1.5–3.5 V (roughly 1.5–2.2 V at 25 °C for modern 1200 V trench field-stop parts). It behaves like a knee voltage plus a resistance, VCE ≈ VCE0 + rCE·IC, rather than the pure IC·RDS(on) of a MOSFET. Conduction loss therefore grows far more slowly with current than a MOSFET's I²·RDS(on), which is the IGBT's advantage over silicon MOSFETs at high blocking voltage.

Key datasheet parameters:

The Turn-Off Tail: The IGBT's Defining Weakness

Conduction relies on stored minority-carrier charge in the n- drift region, and that charge has to disappear before the device is fully off. The gate can cut off the MOSFET channel almost instantly, but the remaining holes can only recombine or be swept out, so a current tail keeps flowing after VCE has already risen. This tail:

• Adds turn-off loss that faster gate drive can't remove: a smaller gate resistor speeds up the first part of turn-off, but the tail is set by carrier lifetime and device design, not by dVGE/dt.
• Limits practical switching frequency to roughly 1–50 kHz for hard-switched designs (vs. hundreds of kHz to MHz for power MOSFETs).
• Is shaped by the device generation. Punch-through (PT), non-punch-through (NPT), and trench/field-stop structures trade VCE(sat) against tail energy. Trench field-stop is the modern standard: an n-type field-stop layer near the collector lets the drift region be much thinner, which lowers both VCE(sat) and the stored charge behind the tail. Manufacturers still offer "low VCE(sat)" and "fast" variants of the same voltage class along that trade-off curve.

Worked Example: Conduction Loss Comparison

Consider one switch in a 1200 V motor-drive leg carrying 50 A during its on-time at a 50% duty cycle (a simplification of the real sinusoidal phase current), at 25 °C. Compare an IGBT modeled as VCE0 = 0.9 V and rCE = 22 mΩ (2.0 V at 50 A) with two 1200 V SiC MOSFETs: an 80 mΩ part, roughly a 30 A class device, and a 25 mΩ part sized for this current.

IGBT:   P_cond = D · (VCE0 + rCE·I) · I
MOSFET: P_cond = D · I² · RDS(on)

Against the undersized 80 mΩ part, the IGBT wins at high current and loses at low current; the crossover is at I = VCE0 / (RDS(on) − rCE) ≈ 15.5 A. That shape — a fixed knee plus a small slope versus a pure resistance — is the real IGBT-versus-MOSFET trade-off.

Against a MOSFET actually sized for the job, the picture changes: the 25 mΩ SiC part loses less at every current in the table, and still does if its RDS(on) rises 50% when hot (≈ 47 W at 50 A, while the IGBT's own VCE(sat) also rises with temperature). Against silicon MOSFETs, whose RDS(on) at 1200 V is far higher, the IGBT's conduction advantage is decisive. Against SiC, the IGBT's remaining advantages are cost, short-circuit ruggedness, and mature high-current modules — which is why it still dominates industrial drives, while a growing share of EV traction inverters has moved to SiC.

Applications and Selection Guidance

• Motor drives and traction inverters (600–1700 V, a few kHz to about 20 kHz): the IGBT's low VCE(sat) at high current and its low cost outweigh its switching loss when frequency stays low; SiC wins where efficiency, size, or higher switching frequency justify its price.
• Welding and induction heating: welding inverters rely on the IGBT's surge and short-circuit robustness; induction cookers typically use IGBTs (often reverse-conducting types) in soft-switched resonant converters at roughly 20–50 kHz, where soft switching keeps the tail loss manageable.
• UPS and solar inverters: IGBT modules (multi-die packages with antiparallel diodes) remain the cost-effective choice from multi-kW to multi-MW, though SiC is spreading in string inverters and high-efficiency UPS designs.
• High-frequency SMPS (>100 kHz): avoid IGBTs — the tail loss per switching event makes them lose badly to MOSFETs or SiC/GaN devices as frequency rises.
• Gate drive: drive it like a MOSFET gate, typically +15 V on and 0 V or −5 to −15 V off. Negative turn-off bias, or an active Miller clamp, guards against dV/dt-induced false turn-on in bridge legs, and the series gate resistor sets dV/dt and di/dt. Add desaturation (DESAT) detection: a shorted IGBT must be turned off within its tSC rating, and a soft (slowed) turn-off limits the overvoltage spike from the large di/dt.

Key Takeaways

• An IGBT merges a MOSFET gate structure with a bipolar PNP output stage, giving voltage-controlled switching with a low, conductivity-modulated on-state drop at high current.
• The on-state drop behaves like a knee plus a small resistance (VCE0 + rCE·I), so loss grows much more slowly with current than a MOSFET's I²·RDS(on) — decisive against silicon MOSFETs at high voltage, much less so against SiC.
• The turn-off current tail, caused by stored minority carriers in the drift region, is the fundamental switching-speed limit and a dominant switching-loss term.
• A parasitic NPN-PNP thyristor exists in every IGBT; modern designs suppress latch-up within the rated SOA, but excessive current or turn-off stress outside it can still trigger it.
• Trench field-stop structures are the modern compromise, lowering both VCE(sat) and tail energy, with "low-loss" and "fast" variants along the remaining trade-off.
• Application choice is a frequency, current, and cost trade: IGBTs for kHz-range, high-power switching; MOSFETs or wide-bandgap devices for high-frequency or efficiency-driven designs.