BAE clears Phase 1 of DARPA's THREADS, and the GaN power ceiling now looks thermal

BAE Systems' FAST Labs has completed Phase 1 of DARPA's Technologies for Heat Removal in Electronics at the Device Scale program, known as THREADS, and has been selected to continue into Phase 2 ^[1][2]. The interesting part is not the selection but the number attached to it: DARPA says Phase 1 performers reached approximately a fivefold increase in RF power density over today's state of the art while holding the reliability needed for operational use ^[3].

That gain did not come from a better semiconductor. It came from getting heat out of one. GaN is already the material of choice for high-performance radar, electronic warfare and communications, and it is a wide-bandgap device that tolerates higher power densities and frequencies than the technologies it replaced, which is why it sits inside modern AESA radars ^[4][5]. DARPA has said GaN already provides more than a fivefold improvement in power density over earlier transistor technologies, and that considerably more is theoretically available if the heat problem can be solved ^[6][7]. In other words, the thermal engineering in Phase 1 bought a power-density improvement of roughly the same order as the original move to GaN ^[1].

The framing of THREADS makes the diagnosis explicit. Push more power through an RF transistor and it makes more heat; if that heat cannot be removed, performance and reliability degrade, and the part ends up running below its theoretical power limit ^[8]. DARPA's response was to attack the problem inside the device rather than bolting on larger conventional cooling, with two stated targets: cut thermal resistance within the transistor without giving up electrical performance, and pull heat away from the hot regions without degrading RF performance ^[9][10]. The current program description sets an eightfold reduction in thermal resistance and power densities up to 81 W/mm for X-band transistor and power-amplifier test devices ^[11].

The systems-level claim deserves more caution than it usually gets. DARPA has previously estimated that solving the thermal limitation could increase radar range by two to three times, and its recent update puts Phase 1 at roughly a doubling of range, the low end of that band ^[12][13][2]. Radar range depends on antenna characteristics, frequency, target size, atmospheric conditions and signal processing, so no radar automatically doubles because a transistor got cooler ^[14]. The figure is best read as an indication of how much leverage sits at the die, not as a delivered capability.

BAE's work runs out of its Microelectronics Center in Nashua, New Hampshire, which already develops and manufactures GaN and gallium-arsenide integrated circuits for defense customers, with Modern Microsystems and researchers from Penn State, Stanford, Notre Dame and the University of Texas at Dallas ^[15][16]. The lineage is long: DARPA's earlier Near Junction Thermal Transport program demonstrated a GaN-on-diamond transistor, using diamond's high thermal conductivity to cut junction-region temperature ^[17].

Watch two things in Phase 2, which moves from research into development and validation ^[18]. First, whether the 8x thermal resistance target and 81 W/mm show up in X-band test devices rather than in structures chosen to flatter the cooling scheme ^[11]. Second, whether reliability survives, since the stated goal is more power in the same footprint without becoming less reliable or thermally constrained ^[19].