Research On Mechanism And Modeling Of GaN Power Device For Microwave Applications

In the current electronic information industry,RF power devices occupy an important position.The third generation wide-bandgap material-Gallium Nitride,which is following Silicon and GaAs material,is one of the most important material.It has been used for radar,communication,power electronics,and light device for a long time.It is shown,in particular,a huge advantage for the most popular application such as 5G,Internet of Thing?IoT?,wireless charging,drone,and so on.GaN high electron mobility transistor?HEMT?is based on GaN material,and it is an active device.So it is very necessary for research.We still have some to do to make it technology mature at the moment,so we need to go on to investigate it.One of the most important aspects is the modeling of GaN HEMT,which is very important for GaN HEMT device research.We need to investigate the mechanism of the device and propose a method for improving.It will very helpful and necessary for device design and optimization.Besides,the GaN HEMT model is the most important part of the RF power device design,which is necessary for RF power device design and optimization of the circuits.As the GaN HEMT is used for more high applications,the modeling for low frequency is not suitable.Moreover,it has a more complex self-heating effect,trapping effects,and high voltage case,the investigation on the modeling of GaN HEMT is a hot topic.In this paper,we start with the research of the small-signal model of GaN HEMT.We investigate the self-heating effect,trapping effect,and gate-capacitance model.The main content and ideas contain:1.Research on the small-signal equivalent circuit model for GaN HEMT.For the high-frequency dispersion of W-band small-signal GaN HEM,we propose a small-signal model for W-band GaN HEMT.Due to that the W-band small-signal model contains a lot of parasitic parameters,the reported methods can't be used to extract so many parasitic parameters.Or the reported methods can be used to extract the parameter by using the time-consuming calculation.All the methods are not efficient methods to finish the parameter extraction.So we proposed a full-wave simulation parameter extraction for the proposed small-signal model.With this method,the multi-values phenomenon of the parameters can be avoided.The huge calculation can be reduced.This method is a physics-based parasitic parameter extraction method,which can be used for arbitrary geometry and complex structure.Up to 110 GHz,the accuracy could be increased by 30%at high frequency.2.Research on the self-heating effect of GaN HEMT.We analyze the thermal distribution of the device.Based on the thermal distribution of the devices,we obtain the analytic expression of each layer.Then we do a lot of work to simplify the analytic expressions to propose a universal thermal resistance model.The universal thermal resistance model contains all key parameters of the GaN HEMT devices,for example,the gate length,gate width,the thickness of the activity layer,the thickness of the substrate,the space between two adjacent gates,the space between two adjacent source metals,the thermal conductivity of different layers.We also pay much attention to the via-holes of the GaN HEMT devices.We analyze how the via-holes influence the dissipation of the device.We added this influence into the universal thermal resistance model.We show the verification of the measured current of different size of the GaN HEMT device.This proposed universal thermal resistance model is the first one that can be embedded into the large-signal model of the GaN HEMT with all possible sizes.The results show the errors are less than 5%.3.Research on the trapping-effect of the GaN HEMT device.In this dissertation,we aim to improve the completeness of knowledge of the trapping effect.We point out that the buffer related trapping effect happens at the buffer under the gate,not at the drain access region.When we applied some value of the drain voltage,the hot electron will be traped in the buffer.It will cause a large negative potential in the buffer,which will induce the threshold voltage shift.Based on the proposed mechanism,we proposed a threshold voltage model,which is based on the trapping effect.In this dissertation,we will deal with the dispersion of the S parameter,especially for the S22.We can see the dispersion phenomenon in the MIT MVSG model and UNIK ASM-HEMT.In those two models,the S parameter has a large difference between the measured data and simulated data,especially for S₂₂.In this dissertation,we explain the reason why there is a difference between the measured data and simulated data.In our proposed threshold voltage model,the dispersion of the S₂₂ has been well dealing with.The accuracy could be increased by 20%at low frequency?<5GHz?.4.Research on the gate capacitance of the GaN HEMT device.In the dissertation,we proposed a gate capacitance model based on physics.It contains two parts,one is the intrinsic capacitances.We usually obtain those intrinsic capacitances from the terminal capacitances,which are mature.Another part is the fringe capacitance,we pay more attention to those parts.For the intrinsic capacitances,we modify the problem,which is contained in the MIT MVSG model and UNIK ASM-HEMT.The problem is that the two popular models neglect the saturation phenomenon of the drain current,which will induce the difference between the simulated and measured data.For the fringe capacitance,we used the map method to calculate the fringe capacitance.We also consider the influence of the bias-dependent depletion region.In the end,the outputs of the large-signal model are given.We can conclude that our gate capacitance model is very accurate based on the load-pull measurement results.In the end,the proposed innovations are used in X band MMIC,which is based on6-inch GaN fabrication.The proposed model employed the step-by-step parasitic parameters method.It employed the universal thermal resistance model,threshold voltage model,and gate-capacitance model.The verifications show that the proposed model can obtain accurate MMIC output results.The error for output power and gain is below 3.5%.The dissertation work is very useful for high-power GaN MMIC design and optimization...