Measurement-based Investigation of the DC and RF Transconductance for Various HEMT Technologies in High-and Low-temperature Conditions

: The purpose of this experimental study is to evaluate quantitatively the impact of the temperature on the behavior of various high electron-mobility transistor (HEMT) technologies through the analysis of the DC and RF transconductance. The experimental data are reported for six different HEMT devices, in order to develop a comparative analysis based on various technologies, including gallium arsenide (GaAs) and gallium nitride (GaN) materials, matched and pseudomorphic HEMTs, single- (S-H) and double-heterojunction (D-H) HEMTs, and both virgin and multi-layer devices. The reported findings show that the impact of the ambient temperature on the HEMT behavior strongly depend on the tested technology and operating conditions. As a matter of fact, a higher temperature can lead to increased or degraded transconductance, depending on the device technologies and bias point. In the GaAs-based devices, an operating bias condition at which the DC and RF transconductance are temperature insensitive can be defined, owing to two-opposite temperature-dependent effects counteracting with each other. at an experimental investigation of the behavior of various HEMT technologies in high-and low-temperature conditions. Six different HEMT devices studied: 0.5 (2 × 100) S 0.5 μm × (2 × 75) μm S - H HEMT, 0.5 μm × (2 × 100) D -H EMT, 0.5 μm × (2 × 100) μm multi -layer D- H GaAs pHEMT, 0.25 μm × (2 × 100) HEMT, and 0.15 μm × (4 × GaN HEMT . The devices under test (DUTs) are characterized in terms of both DC and scattering (S-) parameter measurements at three different ambient temperatures (T a ): -40°C, 25°C, and 150°C. The experimental data are used to investigate the effects of the temperature on the DC and RF values of the transconductance, which is a key figure of merit for determining low-noise and high-power performance of the HEMT devices that are widely used in the design of both low-noise [32-35] and high-power amplifiers. The analysis is focused on the impact of the temperature on the DC transconductance (g m ) that is determined from the measured DC transcharacteristics, the RF transconductance that is estimated as the real part of the measured short-circuit forward transfer admittance (Y 21 ) at low low-frequency, and the intrinsic RF transconductance (g mo ) of the small-signal equivalent-circuit model that is extracted from the S-parameter measurements. The experiments show that the temperature effects can strongly depend on the selected technology and bias condition. It should be underlined that in the four investigated GaAs-based devices there is an operating bias point at which the transconductance is temperature insensitive, the transconductance zero temperature


Introduction
The high electron-mobility transistor (HEMT) was introduced in the early 1980s [1,2] and, over the years, has continuously evolved to meet the steadily increasing demanding requirements of the everexpanding microwave applications. The evolution of this type of transistor has been accomplished through the use of different materials, such as gallium arsenide (GaAs) and gallium nitride (GaN), different structure designs, such as conventional lattice-matched HEMTs and pseudomorphic HEMTs (pHEMTs), multiple two-dimensional electron gas (2-DEG) channels, such as using a double heterojunction (D-H) rather than the conventional single heterojunction (S-H) in order to have two channels, different geometrical parameters, such as gate length (Lg) and width (W), and different operating conditions, such as bias point and frequency band. To improve device performance, HEMT devices are typically based on using the multi-finger layout, which consists of placing many gate fingers in parallel [3][4][5]. Therefore, the total gate width of multi-finger devices is given by the product of the number of fingers and their length. To develop highly integrated monolithic microwave integrated circuits (MMICs), HEMT devices might be realized using the multi-layer three-dimensional (3-D) technology [6][7][8].
Many studies have been devoted to the analysis of the temperature effects on the performance of the HEMT devices, focusing on both GaAs [9][10][11][12][13][14][15][16][17][18] and GaN [18][19][20][21][22][23][24][25][26][27][28][29][30][31] semiconductor technologies. The present article is aimed at an experimental investigation of the behavior of various HEMT technologies in highand low-temperature conditions. Six different HEMT devices are studied: 0.5 μm × (2 × 100) μm S-H GaAs HEMT, 0.5 μm × (2 × 75) μm S-H GaAs HEMT, 0.5 μm × (2 × 100) μm virgin D-H GaAs pHEMT, 0.5 μm × (2 × 100) μm multi-layer D-H GaAs pHEMT, 0.25 μm × (2 × 100) μm GaN HEMT, and 0.15 μm × (4 × 50) μm GaN HEMT. The devices under test (DUTs) are characterized in terms of both DC and scattering (S-) parameter measurements at three different ambient temperatures (Ta): -40°C, 25°C, and 150°C. The experimental data are used to investigate the effects of the temperature on the DC and RF values of the transconductance, which is a key figure of merit for determining low-noise and high-power performance of the HEMT devices that are widely used in the design of both low-noise [32][33][34][35] and high-power [35][36][37][38] amplifiers. The analysis is focused on the impact of the temperature on the DC transconductance (gm) that is determined from the measured DC transcharacteristics, the RF transconductance that is estimated as the real part of the measured short-circuit forward transfer admittance (Y21) at low low-frequency, and the intrinsic RF transconductance (gmo) of the small-signal equivalent-circuit model that is extracted from the S-parameter measurements. The experiments show that the temperature effects can strongly depend on the selected technology and bias condition. It should be underlined that in the four investigated GaAsbased devices there is an operating bias point at which the transconductance is temperature insensitive, the so called transconductance zero temperature coefficient (GZTC) point [17], whereas a significant degradation of the transconductance is observed for the two investigated GaN-based HEMTs at all studied values of the gate-source voltage.
The article is organized with the following sections. Section II describes the investigated devices and the experimental characterization, Section III reports and discusses the achieved experimental findings, and Section IV draws the main conclusions. The two S-H devices investigated in this work are AlGaAs/GaAs HEMTs grown by molecular beam epitaxy (MBE) on a 600-μm-thick semi-insulating undoped GaAs substrate. Both devices have the same gate length of 0.5 μm but differ in the gate widths, which are 200 μm and 150 μm. The interdigitated layout of both devices is based on the parallel connection of two fingers but the two devices differ in the finger length: 100 μm and 75 μm. Further fabrication details of these two S-H GaAs HEMTs can be found elsewhere [39].

Devices Under Test and Measurement Process
The two investigated D-H pHEMTs are based on using the lattice mismatched AlGaAs/InGaAs/GaAs system grown by MBE on a 600-μm-thick semi-insulating undoped GaAs substrate. The two devices have a gate length of 0.5 µm and a gate width of 200 µm. The interdigitated layout consists of two fingers, each being 100-µm long. These two pHEMTs differ only because one of the two devices is made utilizing the multi-layer 3-D MMIC technology. The complete MMIC integration method consists of opening a Si3N4 window using buffered oxide etch (HF, hydrofluoric acid) and lithography on a wafer containing pre-fabricated pHEMTs and other passive components in a vertical plane. On top of the pre-fabricated pHEMTs, a conductor layer is placed to allow other passive components to be connected together; while providing also a track for probing the devices. Further fabrication details of these two D-H GaAs pHEMTs can be found elsewhere [40].
The two GaN HEMTs are based on an Al0.253Ga0.747N/GaN heterostructure grown by metal organic chemical vapor deposition (MOCVD) on a 400-μm-thick SiC substrate. The two devices differ in both gate length and width: 0.25 μm × (2 × 100) μm and 0.15 μm × (4 × 50) μm. The GaN devices were fabricated at the University of Lille, France and further fabrication details can be found elsewhere [41].   Figure 2 illustrates the flow diagram of the temperature-dependent on-wafer measurement process and the subsequent analysis of the impact of the ambient temperature on the DC and RF transconductance. The microwave experiments consist of DC and S-parameters measured from 45 MHz to 50 GHz at -40°C, 25°C, and 150°C. The analysis is performed using the DC characteristics and the Sparameters measured in the saturation region: Vgs was varied from -1 V to 0.8 V at Vds = 3 V for the S-H GaAs HEMTs, Vgs was varied from -1.5 V to 0.6 V at Vds = 3 V for the D-H GaAs pHEMTs, and Vgs was varied from -8 V to -3 V at Vds = 15 V for the GaN HEMTs. The analysis is focused on using the experimental data for developing an accurate study of the DC and RF transconductance.
To ensure that the data were free of human error, the device parameters were measured with a thermal probe station coupled to an HP4142B DC source and an HP8510C vector network analyzer (VNA) by using Keysight's IC-CAP, which is a powerful commercial software for DC and RF semiconductor device characterization and modelling. After the sample reached uniform steady-state temperature, DC and frequency-dependent measurements were carried out at each temperature.

Experimental Results and Discussion
The DC transconductance at three different ambient temperatures (Ta = -40°C, 25°C, and 150°C) for the studied S-H GaAs HEMTs, D-H GaAs pHEMTs, and GaN HEMTs are given in Figures 3-5, respectively. As can be observed in Figure 3, the DC transconductance of the two studied S-H GaAs HEMTs at Vds = 3 V may increase or decrease depending on the selected value of the input voltage. This is due to two temperature-dependent effects contributing in opposite ways to the resultant behavior of the DC transconductance with increasing temperature: the degradation of the electron transport properties and the shift of the threshold voltage (VTH) towards more negative values. The former effect is predominant at higher Vgs, whereas the latter effect is predominant at lower Vgs. It should be mentioned that the shift of VTH can be compensated by the bias circuitry by biasing the device at fixed output voltage and current rather than at fixed output and input voltages [17]. It is worth pointing out that both devices exhibit the zero temperature coefficient point in the DC transconductance behaviour.   Similarly to what has been observed for the S-H GaAs HEMTs, the two studied D-H GaAs pHEMTs at Vds = 3 V show a GZTC point in the DC transconductance that is achieved at Vgs = -0.8 V with Vds = 3 V (see Figure 4). This result indicates that the GZTC point is the same for virgin and multi-layer devices and is shifted towards more negative values of Vgs for the D-H GaAs pHEMTs with respect to those observed for the S-H GaAs HEMTs (see Figures 3 and 4). Likewise to the S-H GaAs HEMTs, both virgin and multilayer D-H GaAs pHEMTs show a peak in gm that decreases with increasing temperature. This peak occurs at about Vgs = -0.3 V. By selecting the bias condition of Vds = 3 V and Vgs = -0.3 V, we obtain the following values of gm for the two devices at the three different ambient temperatures: 63 mS (-40°C), 57 mS (25°C), and 50 mS (150°C) for the virgin device and 59 mS (-40°C), 54 mS (25°C), and 46 mS (150°C) for the multilayer device. This denotes that gm is slightly reduced after using the multilayer 3-D MMIC technology [42]. Figure 5 shows that, in case of the two studied GaN HEMTs, the DC transconductance is significantly degraded with rising temperature, due to the degradation of the carrier transport characteristics that is predominant temperature-dependent effect. The two devices show a peak in gm that decreases with increasing temperature and occurs at about Vgs = -4.8 V. By selecting the bias condition Vds = 15 V and Vgs = -4.8 V, we obtain the following values of gm for the two devices at the three different ambient temperatures: 59 mS (-40°C), 51 mS (25°C), and 33 mS (150°C) for the longer device and 61 mS (-40°C), 53 mS (25°C), and 38 mS (150°C) for the shorter device. As expected, the values of gm are higher for the device with a shorter gate length.
At frequencies low enough to neglect the reactive components, Y21 can be defined in terms of the equivalent-circuit parameters (ECPs) (see Fig. 6) as follows [43]: This equation shows that the RF transconductance can be estimated as the low-frequency Y21, which should be purely real and correspond to the DC transconductance when the low-frequency dispersion effects are negligible. To develop the present analysis, we consider only the real part of the measured Y21 that is evaluated at 45 MHz, which is the lowest frequency point of our measurements. Figures 7-9 reports the impact of the temperature on the performance of the six studied devices by focusing on Re(Y21) at 45 MHz. Similarly to what has been observed for the DC transconductance, the low-frequency Re(Y21) also exhibits the GZTC point in case of the GaAs HEMTs and pHEMTs (see Figures  7 and 8), whereas the low-frequency Re(Y21) gets degraded for all Vgs values by increasing the temperature in case of the GaN HEMTs (see Figure 9).
The equivalent-circuit model in Figure 6 was used to represent the measured S-parameters of the six studied devices. The ECPs were obtained using the well-known "cold" pinch-off methods, which has been widely and successfully applied to HEMT technology in recent years [44][45][46][47][48][49][50]. The intrinsic RF transconductance values are then extracted at the "hot" bias conditions of interest. The achieved values of the intrinsic RF transconductance for the six studied devices are presented in Figures 10-12. Similarly to what has been observed for the DC and RF transconductance (see , the intrinsic RF transconductance also exhibits the GZTC point in case of the GaAs HEMTs and pHEMTs (see Figures 10  and 11), whereas the intrinsic RF gm gets degraded for all Vgs values by increasing the temperature in case of the GaN HEMTs (see Figure 12). Table 1 shows that, by considering a fixed bias point, the intrinsic RF gm for the studied devices is found to be greater than the Re(Y21) at 45 MHz and gm, which is due to the impact of the extrinsic resistances (see equation (1)). Table II reports the values of VgsGZTC, which is the gate-source voltage at which the transconductance is temperature insensitive, for the four studied S-H GaAs HEMTs and D-H GaAs pHEMTs at Vds = 3 V. The achieved values of VgsGZTC are roughly the same by using gm, Re(Y21) at 45 MHz, and gmo.

Conclusion
To gain a comprehensive overview of the temperature effects on the HEMT performance, an experimental study was developed focusing on the behavior of both DC and RF transconductance for various HEMT devices in high-and low-temperature conditions. It was shown that the temperature effects can strongly depend on the chosen device technology and selected bias condition. However, in GaAs-based devices, an operating bias condition at which the DC and RF transconductance are temperature insensitive can be defined, as a result of two-opposite temperature-dependent effects counteracting with each other: the degradation of the electron transport properties and the shift of the threshold voltage towards more negative values. On the other hand, the GaN-based devices show a significant reduction of both DC and RF transconductance at higher temperatures for all studied values of the gate-source voltage, as a result of the degradation of the electron transport properties that is the predominant temperature-dependent effect.