Semiconductor Properties: Bandgaps, Thermal Limits, and SiC vs GaN

Semiconductor properties are the specific electrical and thermal characteristics—like bandgap energy, negative temperature coefficient of resistance, and charge carrier mobility—that allow a material to act as an insulator at absolute zero but conduct electricity when energized, heated, or doped. These properties are not just textbook trivia; they dictate whether your power supply sags under load, whether your high-frequency inverter switches cleanly, and exactly how much heatsink mass you need to bolt to your chassis to prevent a catastrophic thermal failure.

The Core Semiconductor Properties That Dictate Circuit Behavior

When you select a diode, BJT, or MOSFET, you are really selecting a specific set of semiconductor physics. The three properties that matter most on the workbench are:

• Bandgap Energy (eV): The minimum energy required to excite an electron from the valence band to the conduction band. Think of the bandgap as a tollbooth on a highway: electrons need a specific minimum amount of energy (the toll) to jump from the parking lot into the open highway. Silicon has a bandgap of 1.12 eV, while Silicon Carbide (SiC) is 3.26 eV, allowing SiC to block much higher voltages before breaking down.
• Temperature Coefficient of Resistance: Intrinsic semiconductors have a negative temperature coefficient (NTC)—they conduct better as they get hotter because thermal energy frees more charge carriers. However, heavily doped regions (like a MOSFET channel) exhibit a positive temperature coefficient (PTC) at high currents due to phonon scattering, which is a lifesaver when paralleling devices.
• Carrier Mobility: How fast electrons or holes move through the crystal lattice when an electric field is applied. Higher mobility means faster switching speeds and lower on-resistance.

Worked Numeric Example: Thermal Runaway in Parallel BJTs

Because of the NTC property of the base-emitter junction in Bipolar Junction Transistors (BJTs), paralleling them without emitter resistors is a classic bench mistake. Let us look at the math behind why this fails.

The Setup: You need to switch a 10A load, so you put two 2N3055 BJTs in parallel, assuming they will share 5A each. The base-emitter voltage ($V_{BE}$) required to turn them on is nominally 0.7V.

The Numbers: The $V_{BE}$ of a silicon BJT has a temperature coefficient of roughly -2mV/°C. If Q1 is mounted slightly closer to another heat source and runs just 10°C hotter than Q2, its required $V_{BE}$ drops by 20mV (to 0.68V). Because both bases are tied together at the same drive voltage, Q1 now turns on harder and hogs 8A of the load, while Q2 takes only 2A.

The Result: Q1 dissipates more power ($I^2R$ and $V_{CE} imes I$), gets even hotter, its $V_{BE}$ drops further, and it hogs more current until it exceeds its 115W maximum dissipation and literally desolders itself from the board. This is thermal runaway, driven entirely by the semiconductor's negative temperature coefficient.

Where You Meet This in Practice: Si vs. SiC vs. GaN

Understanding semiconductor physics is what separates a parts-swapper from a power electronics designer. Here is how these properties manifest in modern builds:

Silicon (Si): The Workhorse

Silicon's 1.12 eV bandgap limits its breakdown voltage and operating temperature. For a 12V or 24V DIY buck converter, Si MOSFETs like the IRF3205 are cheap and effective. But push them past 600V or 150°C, and intrinsic carrier generation causes massive leakage currents, leading to thermal destruction.

Silicon Carbide (SiC): The High-Voltage Tank

With a 3.26 eV bandgap, SiC requires vastly more energy to break down. This allows for much thinner drift regions, resulting in lower on-resistance at high voltages (1200V+). You will meet SiC when building high-power solar inverters or EV chargers. Catch: SiC MOSFETs require higher gate drive voltages (often +15V / -3V) compared to standard Si (+10V / 0V) to fully enhance the channel.

Gallium Nitride (GaN): The Speed Demon

GaN devices utilize a High Electron Mobility Transistor (HEMT) structure, creating a 2D electron gas (2DEG) with massive carrier mobility. This results in near-zero gate charge and incredibly fast switching transitions. GaN is why modern 100W USB-C chargers are the size of a matchbox. Catch: The extreme switching speeds of GaN mean PCB layout parasitics (trace inductance) will cause massive voltage ringing if you do not design the loop area tightly.

Bench Scenario Walkthrough: The Melted MOSFET Heat Sink

Let us walk through a real-world failure where ignoring a specific semiconductor property destroyed a prototype.

1. Setup: Building a 24V to 12V synchronous buck converter switching at 500 kHz. The design used an IRF3205 (Silicon N-channel MOSFET) rated for 55V and 110A. The datasheet boasts an incredibly low $R_{DS(on)}$ of 8mΩ at 25°C.
2. Numbers (Conduction vs. Switching): At a 100A load, conduction loss is $I^2 imes R = 100^2 imes 0.008 = 80W$. That is manageable with a decent heatsink. However, the IRF3205 has a total gate charge ($Q_g$) of 120nC. The turn-on and turn-off transition times ($t_r + t_f$) are roughly 100ns. The switching overlap loss is calculated as $P_{sw} = 0.5 imes V_{DS} imes I_D imes (t_r + t_f) imes f_{sw}$. $P_{sw} = 0.5 imes 24V imes 100A imes 100ns imes 500,000Hz = 60W$.
3. Outcome: Total power dissipation was 140W (80W conduction + 60W switching). The TO-220 package thermal resistance to ambient was too high, the junction temperature spiked past 175°C, and the MOSFET shorted out, melting the solder pads.
4. What Went Wrong: The designer looked only at the conduction property ($R_{DS(on)}$) and ignored the dynamic switching property (gate charge and transition time). The fix was swapping the Si MOSFET for an EPC2045 GaN FET. The GaN device has a $Q_g$ of just 8nC and transition times under 10ns. Switching losses dropped to roughly 6W, and the board ran cool to the touch.

Common Confusions: Semiconductors vs. Conductors vs. Insulators

When discussing power semiconductor behavior, hobbyists frequently fall into two traps:

Confusion 1: "Semiconductors just have higher resistance than copper."
False. A conductor like copper has overlapping valence and conduction bands; electrons flow freely regardless of temperature (though copper's resistance increases slightly with heat due to lattice vibrations). A semiconductor has a forbidden energy gap. At absolute zero, it is a perfect insulator. It only conducts when external energy (heat, light, or forward bias voltage) bridges that gap.

Confusion 2: "All semiconductors get more conductive when hot."
This is true for intrinsic (undoped) semiconductors and the base-emitter junction of a BJT. But it is dangerously false for the conductive channel of a power MOSFET. As a MOSFET heats up, increased lattice vibrations (phonons) scatter the electrons in the channel, causing $R_{DS(on)}$ to increase significantly (often by 1.5x to 2x at 100°C compared to 25°C). If you calculate heatsink requirements using the 25°C datasheet value for $R_{DS(on)}$, your circuit will overheat under continuous load.

FAQ: Semiconductor Properties on the Workbench