Gate and drain current transients following the application of a step/pulse reverse bias to the gate of AlGaN/GaN HEMTs can be described by a superposition of 2-3 extended exponents with characteristic times of the main process having activation energy of 0.7-0.8 eV. The drain current relaxation is a consequence of charge trapping and movement in the AlGaN barrier. A qualitative model explains this process by movement of a step of charge trapped on deep states in the barrier and movement of the step toward the interface by multiple emission/capture hops. The process is assumed to be confined to channels in AlGaN (presumably dislocations) surrounded by potential barriers, preventing effective injection of the carriers into the channels with forward bias pulses. The model qualitatively explains the often reported appearance of hole-trap-like signals in deep level studies of AlGaN/GaN HEMT structures and the persistent changes in capacitance and current of these structures occurring upon pulsing and illumination. One of the most serious problems with III-Nitrides-based high electron mobility transistors (HEMTs) is the phase delay of the drain current upon pulsed change of the gate bias, so called "gate-lag". [1][2][3] This phenomenon causes increase of the switching time and long-term drift of the threshold voltage and the drain current. The generally accepted view is the gate-lag is caused by trapping of electrons flowing from the gate at high reverse voltage and subsequent capture of electrons by deep centers under the gate. Both the capture and emission process from these traps can be studied by analysis of gate and drain transients as a function of pulsing conditions and temperature by using current or capacitance deep level transient spectroscopy (CDLTS or DLTS) 2-5 or by direct measurements of individual transients.1,6-9These measurements provide the concentration and position of the traps involved and their capture cross sections. This is complicated by the presence of multiple traps, strong non-uniform electrical and strain fields, and the contribution of the band-like states related to extended defects and to surface or interface states. As a result, the drain and gate current relaxation curves are non-exponential. A common approach is the deconvolution of the actual relaxations into a sum of independent exponential relaxations and fitting the pre-exponential coefficients to experiment.6 Various methods to improve the convergence and reproducibility of the procedure exist. 7,8 Another popular approach is to build the derivative of the relaxation curve on the logarithm of time, find the peaks in such plots, and trace the peak position variation with temperature to extract the activation energy and capture cross section of the traps.2,3,9 The number of peaks in either approach is quite limited (typically, 2-3) in either method, 9 but the peaks are generally broad which requires invoking a large number of exponents to attain good fit in the basic method, 6 while providing no knowledge on the physical nature o...