Highly reliable GaN HEMTs have entered the commercial RF market over the last decade due to the large band gap, high saturation electron mobility, high temperature operation, and the high breakdown electric field of GaN material [1-3]. However, under high frequencies or high tempera-  ture reverse bias (HTRB), the device which has mature technological progresses is still subject to trap induced limitations. The HTRB stresses are key industrial items prone to reveal the epitaxial quality of materials. Recent works have reported that when the device works at high temperatures and negative gate bias, a significant recoverable negative V_th shift (ΔV_th) was observed [4-5].  Additionally, this effect was ascribed to the deple- tion of trap states located at the dielectric/GaN interface and/or in the gate insulator. Fixed charges in insulator films [7-8] , border traps [9] and interface states at the insulators/AlGaN interfaces are involved in the V_th shift mechanism. It is worth noting that it has not been clear which interface trap state change causes the threshold voltage drift, and how interface states affect device performance.
This paper presents degradation of AlGaN/GaN high electron mobility transistors with time in HTRB experiments at the same bias. The experimental results show that drain-to-source current of AlGaN/GaN HEMTs after HTRB stress was obviously lower than that of the fresh devices. The maximum change of drain-to source current was up to 188 mA. The threshold voltage drifted forward and gate-lag characteristic became worse for the AlGaN/GaN HEMTs after HTRB stress. The dependence of threshold voltage (V_th), capacitance-voltage (C-V), double pulse output curve and gate leakage current (I_gd, I_gs) on HTRB time was revealed at different time intervals. By analyzing the capacitance-voltage characteristic curves, the constant capacitance of forward bias is usually related to the SiN_x capacitance, while the reverse bias depends on the total capacitance (C_total) of the SiN_x layer and AlGaN [10], our experiments show that it is the SiN_x layer that causes the change in capacitance.
The device under test (DUT) is AlGaN/GaN HEMT, as shown in Fig. 1. The V_ds and V_gs bias waveform during the experiments are given in figure. 2. In the past, many experiments were conducted with the same interval time, but the degradation process of the device is often a non-linear process, and the step length setting is very important. Too long step length may miss the time point of device mutation, while too short step length makes the results are affected by the test conditions, making the results inaccurate. In this experiment, we roughly determined the mutation range of the device through preliminary exploration. As we can see, to get a better view of results, the experiment step time was set to 10h, 30h, 60h, 100h, 200h, bias stress test (V_gstress = −10 V, V_dstress = 60 V), the experiment temperature was set to 150 ℃, in order to determine the failure time of the device, we used a computer program to control the applied stress and monitor the real-time current.
All virgin and stressed devices have been extensively characterized. For the purpose of clarity, we compare the original device with the device after 100 h and 300 h stress in Fig. 3 to indicate the changes of output transfer and characteristics. It can be seen from the figure that the threshold voltage drifted forward before and after the stress, and the current collapsed worsen obviously, indicating that the device had a certain degree of degradation. However, the threshold voltage was almost unchanged from 100 h to 300 h, while the output current collapsed worsen. This phenomenon indicates that it is possible that AlGaN/GaN interface captures charge, leading to the reduction of channel carriers density, or it is possible that the increase of electron state density in SiN_x/AlGaN or SiN_x leads to the depletion of 2DEG to a certain extent and the decrease of output current. To shed a light on this, we did other tests.
Figure 4 shows the gate leakage current of the device at different stress stages. The gate source current decreases by 0.041 mA and the gate leakage current decreases by 0.0436 mA before and after the stress. Figure 5 is the gate delay curve before and after the stress. The device needs 30% more time to reach the maximum current. Figure 6 shows C-V characteristic curves after stress, we notice that the stretch-out effect of the curve decreases with time. The Fermi level (E_F) is located far below the valence band maximum of AlGaN at the Al₂O₃/AlGaN interface, which makes electron occupation of interface states no longer a function of the gate bias, leading to absence of stretch-out behavior in the C-V curve [10].
Pulsed drain current versus drain voltage (I_D- V_D) measurements, carried out starting from several quiescent bias points in the off-state, may be used to quantitatively evaluate the current collapse, and to characterize a specific device technology in terms of sensitivity to trapping processes. The pulsed IV waveforms examined can provide the capability of isothermal (constant temperature ) and isodynamic (constant  trap occupancy) transistor measurements for more accurate RF characterization.
The difference on pulsed output characteristics in Fig. 7 is revealed for various quiescent conditions of V_gs-q. Similar plots have largely been reported in the literature related to GaN HEMT structures, as in [6], where (co)doping of the buffer is investigated as the cause for such signature. The current under pulse measurement is lower than that under DC test, and the current collapse appears. The static operating point is (V_gs-q, V_ds-q)=(-4V, 0V), and the device is in the cut-off state.The Schottky gate /AlGaN interface state and the ungate AlGaN surface state can capture electrons and cause the current collapse phenomenon
 

high temperature reverse bias stress,interface trap state,electron state density