Device structures for the modulation-doped high-electron-mobility transistor (MODFET or HEMT), the metal Schottky-barrier gate transistor (MESFET), and the permeable-base transistor (PBT) With 0.1- to 0.25-µm gate lengths have been examined for their millimeter-wave performance. In particular, their unity-current-gain frequency (f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">T</inf> ), maximum oscillation frequency (f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">max</inf> ), and the stability of power gains are Compared. It is shown that in field-effect transistors with gate lengths below 0.25 µm, the high aspect ratio design approach, involving the ratio of the gate length to its associated depletion depth, needs to be extended to include the gate-to-drain separation in an effective gate-length concept. Using the charge-control approach in conjunction with the carrier velocity saturation effect, it is shown that the overall transit time delay across the total effective gate length results in an effective gate capacitance that is significantly larger than that due to the physical size of the gate itself. It is shown with specific examples that a successful design of these ultrasubmicrometer gate structures depends on the choice of the gate-to-drain separation (L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GD</inf> ) and the separation (d <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">o</inf> ) between the gate and the channel sheet Charge, with an appropriate choice of doping concentration consistent with the desired drain voltage operation. The designs have been based on practically obtainable values of the gate and source series resistances. When the actual gate length is much smaller than L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GD</inf> , as in the case of PBT's, it is shown that the stable power gain margin in these devices is determined by the square of the aspect ratio (L <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">GD</inf> /d <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">o</inf> ). The results show that an f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">max</inf> of approximately 350 GHz can be achieved from a MODFET structure with an effective gate length of 0.14 µm. In a PBT structure an f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">max</inf> exceeding 500 GHz can be achieved with an effective gate length of 0.15 µm, and in a similar MESFET structure an f <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">max</inf> of approximately 450 GHz can be achieved. The device design approach presented clearly indicates the possibility of achieving FET'S operating beyond 200 GHz with useful power gains.
Read full abstract