Threshold voltage is the minimum gate voltage needed to turn on a transistor, specifically the point at which current begins flowing freely between its source and drain terminals. In a MOSFET (the type of transistor inside nearly every modern chip), this voltage creates a conductive channel at the surface of the silicon, allowing electrons to move through. It’s one of the most important parameters in semiconductor design because it directly controls how fast a chip can switch and how much power it wastes when idle.
What Physically Happens at the Threshold
A MOSFET has three key terminals: the gate, the source, and the drain. The gate sits on top of a thin insulating layer (the oxide), which sits on top of a silicon body. When no voltage is applied to the gate, no current flows between source and drain because there’s no path for electrons to travel through.
As you increase the gate voltage, it starts pulling charge carriers toward the silicon surface beneath the oxide. At first, only a trickle of current flows through a mechanism called diffusion, where carriers drift randomly due to concentration differences. This is called weak inversion, or subthreshold operation. The transistor isn’t really “on” yet, but it isn’t perfectly off either.
At the threshold voltage, the concentration of mobile electrons at the silicon surface equals the concentration of holes (positive charge carriers) in the bulk material. This marks the transition to strong inversion, where a well-defined conductive channel forms and current flows primarily through electric-field-driven drift rather than random diffusion. The threshold is essentially the crossover point where drift current and diffusion current are equal. Above this voltage, the transistor is fully on and current increases rapidly with additional gate voltage.
What Determines the Threshold Voltage
Several physical properties of the transistor set its threshold voltage. The most influential are the thickness of the gate oxide layer and the doping concentration of the silicon underneath. A thicker oxide means the gate’s electric field has a harder time reaching the silicon surface, which raises the threshold. Heavier doping (adding more impurity atoms to the silicon) also raises it, because more gate voltage is needed to counteract the existing charge distribution.
The relationship can be expressed simply: threshold voltage scales with oxide thickness divided by the oxide’s ability to store charge (its permittivity), multiplied by terms involving the doping level and the intrinsic properties of silicon. In practice, chip manufacturers tune these variables during fabrication to hit a target threshold for each transistor type on a chip.
The Body Effect
The textbook threshold voltage assumes the source terminal and the silicon body are at the same voltage. In real circuits, that’s not always the case. When a voltage difference exists between the source and the body, the threshold shifts upward. This is called the body effect.
The adjusted threshold voltage equals the baseline value plus an extra term that grows with the source-to-body voltage. A device parameter called the body effect coefficient (often written as gamma) controls how sensitive the threshold is to this voltage difference. The body effect matters most in circuits where transistors are stacked in series, since the source of one transistor may sit at a higher voltage than the shared body connection, pushing its threshold higher than expected and slowing the circuit down.
Temperature Shifts the Threshold
Threshold voltage changes with temperature, and the shift is significant enough to affect chip performance. In standard NMOS transistors, the threshold drops by roughly 1.3 millivolts for every degree Celsius the temperature rises. PMOS transistors (the complementary type used alongside NMOS) can shift by about 2 millivolts per degree in the opposite direction.
This means a chip running 80°C above room temperature could see its NMOS thresholds drop by around 100 millivolts. Lower thresholds let more current leak when the transistor is supposed to be off, increasing power consumption. It also makes the transistor switch faster, which can cause timing issues if the design wasn’t built to handle that variation. Designers must account for these shifts across the full operating temperature range of a chip.
How Threshold Voltage Is Measured
Pinning down the exact threshold of a real transistor is trickier than the theory suggests, because there’s no single sharp boundary where the device flips from off to on. The most common industry method is the constant-current approach: you sweep the gate voltage upward while monitoring drain current, and define the threshold as the gate voltage where current reaches a specific reference value. A typical reference point is on the order of 1 microamp, scaled by the transistor’s width-to-length ratio.
Another approach plots the square root of drain current against gate voltage in saturation mode. This produces a roughly straight line, and extending that line to the voltage axis gives the threshold. Both methods yield slightly different numbers, which is why datasheets always specify which extraction method was used. For modern transistors with channel lengths below 100 nanometers, the choice of method can shift the reported threshold by tens of millivolts.
Why Chip Designers Use Multiple Thresholds
Modern processors don’t use a single threshold voltage for every transistor on the die. Instead, chip designers choose from libraries offering low, regular, and high threshold options, often abbreviated LVT, RVT, and HVT. The reason comes down to a fundamental trade-off between speed and power leakage.
A transistor with a low threshold voltage switches faster because less gate voltage is needed to push it into strong inversion. But when it’s supposed to be off, more current leaks through because it’s closer to the conduction threshold at all times. A high threshold transistor is the opposite: slower to switch, but it leaks far less current in its off state. Using a thicker gate oxide raises the threshold and simultaneously reduces both subthreshold leakage and the tunneling current that passes directly through the oxide itself.
The smart approach is to mix them. Critical signal paths that determine maximum clock speed get low-threshold transistors for speed. Everything else, the vast majority of a chip’s billions of transistors, gets regular or high-threshold devices to minimize idle power draw. This mixed-threshold technique can reduce dynamic power consumption by 5% to 30% while maintaining the same timing targets. For leakage specifically, mixing thresholds within individual logic gates (rather than just at the gate level) can cut leakage by an additional 20% compared to simpler dual-threshold approaches.
SRAM memory cells take this a step further in some designs, dynamically switching between low and high threshold modes. During active access, cells operate in a high-performance, low-threshold state. When idle, they shift to a high-threshold state to suppress leakage. This is increasingly important as transistor counts climb into the billions and even tiny per-transistor leakage currents add up to significant wasted power.
Threshold Voltage in Modern Transistors
At leading-edge manufacturing nodes (7, 5, and 3 nanometers), transistors have moved from traditional flat designs to three-dimensional structures called FinFETs, where the channel wraps around a thin silicon fin. The threshold voltage in these devices is controlled by fin geometry, doping, and the work function of the gate metal rather than just oxide thickness and bulk doping.
Exact threshold values at these nodes are proprietary, but manufacturers typically offer at least two threshold options per transistor type at each node. The actual voltage differs slightly between the 7, 5, and 3 nanometer generations as the fins shrink and gate control improves. Choosing between high and low threshold variants doesn’t just affect power and speed. It also influences how resistant the transistor is to errors caused by radiation strikes, which matters for aerospace and automotive applications where reliability is critical.