Journal of Semiconductors, Volume. 45, Issue 11, 112501(2024)

Investigation of performance-enhanced GaN-based E-mode p-channel MOSFET with pre-ohmic-annealing treatment

Huake Su1, Tao Zhang1、*, Shengrui Xu1、**, Hongchang Tao1, Yibo Wang2, Yuan Gao1, Yue Hao1, and Jincheng Zhang1、***
Author Affiliations
  • 1State Key Laboratory of Wide-bandgap Semiconductor Devices and Integrated Technology, School of Microelectronics, Xidian University, Xi’an 710071, China
  • 2Suzhou Institute of Nano-Tech and Nano-Bionics (SINANO), Chinese Academy of Sciences, Suzhou 215123, China
  • show less

    Pre-ohmic-annealing (POA) treatment of P-GaN/AlN/AlGaN epitaxy under N2 atmosphere was demonstrated to effectively achieve good p-type ohmic contact as well as decreased epitaxy sheet resistance. Ohmic contact resistance (Rc) extracted by transfer length method reduced from 38 to 23 Ω·mm with alleviated contact barrier height from 0.55 to 0.51 eV after POA treatment. X-ray photoelectron spectroscopy and Hall measurement confirmed that POA treatment was able to reduce surface state density and improve the hole concentration of p-GaN. Due to the decreased Rc and improved two-dimensional hole gas (2DHG) density, an outstanding-performance GaN E-mode p-channel MOSFET was successfully realized.

    Keywords

    Introduction

    Wide-bandgap gallium nitride (GaN) is suitable for fabricating optical and electronic devices due to its excellent material characteristics, such as wide bandgap, direct bandgap, and high electron saturation velocity[13]. Nowadays, GaN-based high electron mobility transistors (HEMTs) have achieved rapid development, benefiting from the occurrence of the two-dimensional electron gas (2DEG)[46]. For further application in logical and signal processing, p-channel MOSFETs (PMOSs) were realized using two-dimensional hole gas (2DHG) induced by p-GaN/AlGaN heterojunction, which is integrated into GaN-based complementary metal oxide semiconductor (CMOS)[7, 8]. However, the maximum current of enhancement-mode (E-mode) PMOS based on the p-GaN/AlGaN heterojunction is rather low[9, 10]. Recently, Chowdhury et al. first proposed a self-aligned fabrication process and reported an E-mode PMOS with an on-current density of 45 mA/mm[11, 12]. Additionally, a self-aligned E-mode GaN p-channel FinFET with an operation current of 125 mA/mm has also been realized[13]. In addition, Schottky-gated p-channel field effect transistors (p-FETs) are a promising way to further improve the performance of p-FETs[14]. Furthermore, Raj et al. have demonstrated an E-mode Schottky-gated p-FinFET with current densities of 65 and 13 mA/mm[15, 16]. With the rapid development of p-FETs, most CMOS logical circuits have been reported, including NAND, NOR, and ring oscillators[7, 17].

    However, the performance of GaN p-FETs hardly matches that of GaN HEMTs, which constrains the development of GaN-based logic circuits[9, 18], because p-FETs face some severe challenges. The poor Mg ionization efficiency and high background carrier density in GaN limit the performance of p-FETs[19, 20]. Meanwhile, plasma-etching damage and contamination induced during the fabrication process also deteriorate the characteristics of p-FETs[21, 22].

    The other key issue inhibiting p-FET performance is the poor ohmic contact characteristics of p-type GaN, caused by low hole concentration and high surface state density. Obtaining excellent ohmic contact characteristics is in the spotlight, and several methods have been suggested. Utilizing a p-InGaN capping layer as the contact layer can decrease the contact resistance to below 10 Ω·mm, making it a feasible method for future research[14, 23]. Removing the surface oxides through Mg deposition and reducing the surface states by lowering edge dislocations are two effective ways to achieve prominent ohmic characteristics[2427]. Meanwhile, the choice of high-work-function metals, such as Pt, Pd, Au, Ag, Ni/Au, Cr/Au, and Co/Au to achieve good p-type ohmic contact characteristics is also crucial[2830]. However, the Fermi level pinning effect inhibits the improvement of ohmic contacts when using high work-function metals[31]. Boiling aqua regia or KOH solution prior to metal deposition has been reported to remove the surface oxide layer, thereby reducing surface states and contact barrier height, which in turn reduces the contact resistance (Rc)[32, 33].

    In this research, pre-ohmic-annealing (POA) treatment under N2 atmosphere was proposed to optimize the p-type ohmic contact characteristics with decreased surface barrier and increased 2DHG density. The influence on contact characteristics by wet treatment and POA treatment under different atmospheres was also investigated. Due to the reduction of oxygen (O) impurities on the surface and within the material by POA and wet treatment, the contact barrier is reduced, and the hole concentration is increased, resulting in improved contact characteristics. The effect of decreased on-resistance and increased current density in GaN PMOS through POA treatment under an N2 atmosphere has been proven effective, showing great potential for fabricating high-performance CMOS logic circuits.

    Experiment and calculation methods

    The epilayer for device fabrication was grown on a 6-inch silicon substrate with a low-temperature AlN and AlN/GaN superlattice acting as the nucleation layer by metal−organic chemical vapor deposition (MOCVD), which consists of a 4.5-μm GaN buffer layer, a 300-nm GaN channel, a 1-nm AlN interlayer, a 15-nm AlGaN barrier layer with Al component of 0.15, a 1.5-nm AlN and a 70-nm p-GaN with Mg concentration of 3 × 1019 cm−3 from bottom to top, as is shown in Fig. 1. The 1.5-nm AlN spacer between p-GaN layer and AlGaN barrier layer is able to lift the valance band of p-GaN near AlGaN due to higher polarization intensity, which effectively improves the Mg ionization rate. After growth, the epilayer was annealed under nitrogen atmosphere with temperature dropping from 850 to 750 °C in 20 min to activate the p-GaN layer. The carrier density of holes and the sheet resistance were extracted to be 3.94 × 1013 cm−2 and 24 kΩ/□, respectively, using contact-Hall measurement. In addition, after the p-GaN activation annealing, the epilayer will be exposed to air for a long time before device fabrication, which may form a natural oxide layer, then compensate a part of holes and lift surface potential.

    (Color online) Cross-sectional schematic of the fabricated GaN E-Mode PMOS.

    Figure 1.(Color online) Cross-sectional schematic of the fabricated GaN E-Mode PMOS.

    Device fabrication commenced with mesa isolation using an inductively coupled plasma (ICP) process with a depth of 200 nm. The epitaxy was then immersed in a monoethanolamine solution at 70 °C for 10 min to remove organic contaminants. The POA process was employed for the optimization of p-type ohmic contact. The sample treated with POA process in N2 atmosphere at 500 °C was named sample B, while the sample treated with POA process in O2 atmosphere at 500 °C was named sample C. The control sample without POA treatment was also prepared and named sample A. Subsequently, all samples were treated with buffered oxide etchant (BOE) for 2 min, which was effective to remove the surface oxidation and decreased the interface barrier. Additionally, a bilayer Ni (30 nm)/Au (50 nm) was deposited using electron beam evaporation (EBE), followed by rapid thermal annealing at 550 °C for 5 min in an O2 ambient to form the ohmic contact. The gate recess region was achieved using a low-damaged ICP process with a low etch rate of 1.3 nm/min, and a total depth of 60 nm was obtained, with 10 nm of p-GaN retained as conduction channel. Afterward, all samples were annealed in nitrogen at 450 °C for 5 min and immersed in TMAH at 80 °C for 10 min to eliminate the etching damages. A 15 nm Al2O3 layer was deposited as both the gate dielectric and the passivation layer in the active region by plasma-enhanced atomic layer deposition, using trimethylaluminum and H2O as the sources of Al and O. Each cycle consists of a 0.1-s pulse followed by an 8-s purge for TMAl, and a 0.1-s pulse followed by a 6-s purge for H2O, resulting in a deposition rate of 0.098 nm/cycle. Finally, the gate electrodes were formed using a metal stack of Ni/Au (30/50 nm). The fabricated GaN PMOSs featured a 2-μm LGS, 2-μm LGD, 4-μm LG and 50-μm WG.

    Results and discussions

    Fig. 2 illustrates the ohmic contact characteristics of various wet treatments before metal deposition. Compared to the sample without wet treatment, the samples with BOE and hot KOH treatment exhibit higher current density, as depicted in Fig. 2(a). Due to the removal of surface oxidation, the measured current density increases from 18.7 mA/mm for the untreated sample to 23.6 and 19.4 mA/mm for samples treated with BOE and hot KOH, respectively, at voltage of −5 V. The Schottky barrier height (SBH) of various treatments at different pad spacing were calculated by thermionic emission theory[31, 34], which can be described as:

    (Color online) (a) I−V curves, (b) SBHs calculated by thermionic emission theory at various pad spacing, (c) Rc and (d) ρc extracted from TLM at various voltage of different wet treatments.

    Figure 2.(Color online) (a) I−V curves, (b) SBHs calculated by thermionic emission theory at various pad spacing, (c) Rc and (d) ρc extracted from TLM at various voltage of different wet treatments.

    • Table 1. 2DHG and Rsh measured by contact-Hall as well as Rsh extracted by TLM at 10 V of the three samples.

      Table 1. 2DHG and Rsh measured by contact-Hall as well as Rsh extracted by TLM at 10 V of the three samples.

      Sample2DHG (1013 cm−2)Rsh (kΩ/□)Rsh extracted by TLM (kΩ/□)
      A3.942432.9
      B4.5321.528.6
      C3.2225.433.4

    I=ISATexp(qVnkT)[1exp(qVkT)],

    where V is the applied voltage, q is the charge of electron, n is the ideality factor, k is the Boltzmann constant, and T is the Kelvin temperature. By fitting the linear relationship between I/[1exp(qV/kT)] and V, the reverse saturation current ISAT was obtained. Then the SBH φBcan be calculated by:

    ISAT=AA*T2exp(qφBkT),

    where A is the contact area and A* is the effective Richardson’s constant. Fig. 2(b) shows the wet treatment is effective for reducing the contact barrier. Figs. 2(c) and 2(d) show the extracted Rc and specific contact resistance (ρc) at extracted voltage from 3 to 10 V. It is obvious that the wet treatments are effective in reducing the Rc. The Rc decreases from 90 to 43 Ω·mm after the BOE treatment and to 81 Ω·mm after treatment with hot KOH solution. However, the treatment with hot KOH solution slightly improves the ohmic contact characteristics. The BOE treatment is more effective and convenient, and was adopted after the POA treatment to achieve better contact characteristics.

    The ohmic contact characteristics of samples A−C are evaluated using the linear transfer length method (TLM). IV curves of the three samples with 3-μm spacing are displayed in Fig. 3(a), indicating the increased conduction current of sample B, while sample C exhibits degradation. It is reported that high density surface states lead to the non-linear behavior of IV curves[26, 27, 34]. To analyze the impact of the contact barrier, Rc and sheet resistance (Rsh) of three samples are extracted from 3 to 10 V, which are shown in Figs. 3(b) and 3(c) (the inset in Fig. 3(b) exhibits the fitted curves), respectively. The impact of the contact barrier decreases with the increasing extraction voltage, which leads to the decrease in Rc and Rsh. The extracted Rc are 38/23/44 Ω·mm for sample A/B/C, respectively, at an extraction voltage of 10 V. Rc of sample B is lower than that of sample A, while Rc of sample C is higher than that of sample A, which implies that sample B exhibits better ohmic contact characteristics.

    (Color online) (a) I−V curves of three samples with 3 μm spacing, (b) the Rc and (c) Rsh extracted from TLM at various voltage for three samples, (d) SBHs calculated by thermionic emission theory at various pad spacing. The inset of (b) shows the fitted curves for the three samples extracted at 10 V.

    Figure 3.(Color online) (a) I−V curves of three samples with 3 μm spacing, (b) the Rc and (c) Rsh extracted from TLM at various voltage for three samples, (d) SBHs calculated by thermionic emission theory at various pad spacing. The inset of (b) shows the fitted curves for the three samples extracted at 10 V.

    To further investigate the influence of SBH on ohmic contact characteristics, thermionic emission theory is also used to calculate the SBH. The calculated SBHs at various pad spacings of the three samples are shown in Fig. 3(d). Sample C shows the largest SBH, while sample B has the lowest SBH, regardless of the pad spacing. The mean value of SBHs is 0.55 ± 0.005, 0.51 ± 0.005 and 0.58 ± 0.005 eV for sample A, B, and C, respectively, which is consistent with the non-linear phenomenon observed on Figs. 3(a)−3(c).

    Fig. 4 investigates the influence of POA and BOE treatment on surface morphologies. The root mean square roughness (RMS) for samples without treatment, treated by POA in N2 atmosphere and BOE (sample B), treated by POA in O2 atmosphere and BOE (sample C), and treated only by BOE (sample A) is 0.293, 0.324, 0.304, and 0.321 nm, respectively. No significant deterioration in surface morphologies was observed after annealing or BOE treatment, indicating that surface morphologies have a negligible influence on contact characteristics.

    (Color online) Surface morphologies for samples without treatment, treated by POA in N2 atmosphere and BOE (sample B), treated by POA in O2 atmosphere and BOE (sample C), and treated only by BOE (sample A).

    Figure 4.(Color online) Surface morphologies for samples without treatment, treated by POA in N2 atmosphere and BOE (sample B), treated by POA in O2 atmosphere and BOE (sample C), and treated only by BOE (sample A).

    XPS measurement is used to reveal the effect of POA treatment on surface states. Fig. 5(a) displays the XPS spectra of Ga 2p, showing peaks located at 1117.2, 1116.9, and 1116.9 eV for sample A, B, and C, respectively. Due to the natural oxidation and impurities introduced during epitaxy, many Ga−O bonds form at the surface of the three samples. The XPS intensity ratios of Ga−O to Ga−N are calculated to be 1.53, 1.21, and 2.55 for sample A, B, and C by deconvoluting the Ga 2p spectra into GaN and GaO spectra. Additionally, N 1s spectra presented in Fig. 5(b) also demonstrate that the N 1s peak of sample B is intensified when compared to that of sample A. It indicates that the N atomic concentration of sample B after POA treatment in N2 atmosphere is substantially higher than that of sample A[32]. Whereas, sample C treated after POA in an O2 atmosphere results in the introduction of more O atoms at the surface, while the number of N atoms decreases. Therefore, the SBH would be elevated due to more GaOx generated on the surface, degrading the ohmic contact characteristics[22, 33]. Fig. 5(c) presents the measured valence band maximum (VBM) of these three samples for the surface barrier states identification[24]. The VBM binding energies of sample A, B, and C are 1.20, 0.57, and 1.58 eV, accordingly. VBM of sample B is higher than that of sample A by 0.63 eV, whereas VBM of sample C is lower than that of sample A by 0.38 eV. The higher VBM indicates a smaller SBH for holes, and a brief schematic illustration of this is shown in Fig. 5(d).

    (Color online) (a) XPS spectra of Ga 2p and (b) N 1s, (c) VBM spectra and (d) surface band barrier schematic of three samples.

    Figure 5.(Color online) (a) XPS spectra of Ga 2p and (b) N 1s, (c) VBM spectra and (d) surface band barrier schematic of three samples.

    Unlike surface treatment by wet solution, POA is an effective way to reduce the density of O impurities present in the deeper regions below the surface. As shown in Fig. 6(a), the ratio of Ga−O to Ga−N obtained by deconvoluting the Ga 3d spectra exhibits the lowest value for the samples treated with POA in an N₂ atmosphere and BOE, compared to the samples without treatment or those treated only with BOE. The O 1s spectra shown in Fig. 6(b) also demonstrate lower intensity for the sample treated with POA in an N2 atmosphere and BOE, and higher intensity for the sample treated with POA in an O2 atmosphere and BOE, compared to other samples. Meanwhile, the samples treated with BOE for 2 and 5 min exhibit similar intensities in the O 1s spectra, indicating saturation in the removal of oxidation by BOE. To eliminate the effect of BOE treatment, the surface etching process was performed using Ar+ plasma with an approximate etching depth of 8 nm. The similar trend in the O 1s spectra after etching, as shown in Fig. 6(c), indicates the influence of POA treatment in depth. Since O tends to substitute at the N site, acting as a donor or forming Mg−O complexes, the O impurities will compensate the holes, resulting in a low hole concentration[35, 36]. Therefore, the POA treatment in an N2 atmosphere shows the ability to disrupt the equilibrium of O-related complexes due to their low binding energies[36]. Then, O impurities will spread to the surface, especially along dislocations, owing to the strong oxygen adsorption on the surface[3739]. When combined with BOE treatment, the major O impurities near the surface and within the material will be removed, as shown in Fig. 6(d). Therefore, the treatment with POA in N₂ and BOE not only eliminates the surface barrier but also increases the hole concentration, as confirmed by the Hall measurement in Table 1. Conversely, annealing in an O2 atmosphere produces a contrary effect. The O-rich atmosphere will saturate the surface with oxygen, thereby increasing the probability of its spreading into the bulk and compensating for the holes[39]. The O 1s spectra shown in Figs. 6(b) and 6(c), along with the Hall results exhibited in table 1, support the speculation. The increased O impurities on the surface and within the materials lead to a higher contact barrier height, aggravating the contact characteristics.

    (Color online) (a) Ga 3d spectra for samples with different treatments and O 1s spectra for samples (b) without etching and (c) with etching. (d) Treatment schematic of samples with different treatments.

    Figure 6.(Color online) (a) Ga 3d spectra for samples with different treatments and O 1s spectra for samples (b) without etching and (c) with etching. (d) Treatment schematic of samples with different treatments.

    2DHG density and Rsh measured by contact-Hall of the three samples are listed in Table 1, and Rsh extracted by TLM at 10 V is also included. Sample B exhibits higher 2DHG density and smaller Rsh than sample A, while sample C treated with POA process in O2 atmosphere shows an opposite tendency. The reduced O impurities of sample B will raise the valence band and a lot of holes will be released[40]. Moreover, higher temperature treatment may break more Mg−H bonds, thereby improving the ionization efficiency of Mg[19].

    Figs. 7(a) and 7(b) show the transfer characteristics of the three devices on linear scale and semi-logarithmic scales. Sample B shows the superior conduction current at the same gate voltage which benefits from the improved ohmic contact characteristics and the higher 2DHG density. The threshold voltage (VTH) determined by linear extrapolation of the three samples are −3.1, −2.80, and −3.3 V, respectively. The reason for the different VTH among three samples is that a depletion of higher 2DHG density beneath the gate requires a higher voltage[41]. Figs. 7(c)−7(e) are the output characteristics curves of the three samples. When gate voltage is −8 V and drain−source voltage (VDS) is −5 V, the output current density is 1.51, 2.00, and 0.88 mA/mm for sample A, B, and C, respectively. The on-resistance (RON) extracted at VDS = −5 V is 3.32, 2.49, and 5.68 kΩ·mm for sample A, B, and C, respectively.

    (Color online) Transfer characteristics on (a) linear scale and (b) semi-logarithmic scale and (c)−(e) output characteristics of samples A−C, respectively.

    Figure 7.(Color online) Transfer characteristics on (a) linear scale and (b) semi-logarithmic scale and (c)−(e) output characteristics of samples A−C, respectively.

    Conclusion

    In this work, POA treatment in N2 atmosphere is introduced to GaN-based PMOS. The POA treatment in N2 atmosphere enhances ohmic contact characteristics and 2DHG density of devices, resulting in improved transfer and output characteristics. The ohmic contact characteristics extracted by TLM show that POA treatment in N2 atmosphere is able to reduce Rc and SBH, however, POA treatment in O2 atmosphere shows an opposite tendency. XPS measurement illustrates that POA treatment in N2 atmosphere can reduce Ga−O bonds on epitaxy surface and O impurities in p-GaN, which results in the decrease of surface barrier height and enhancement of the density of 2DHG. POA is a prominent method to achieve a better ohmic contact with p-GaN and enhance the performance of GaN-based PMOS.

    [11] [11] ChowdhuryN, XieQ Y, YuanM Y, et al. First demonstration of a self-aligned GaN p-FET. 2019 IEEE International Electron Devices Meeting (IEDM), 2019, 4.6. 1

    [12] N Chowdhury, Q Y Xie, J Niroula et al. Field-induced acceptor ionization in enhancement-mode GaN p-MOSFETs, 5.5. 12(2020).

    [14] K Nomoto, R Chaudhuri, S J Bader et al. GaN/AlN p-channel HFETs with Imax >420 mA/mm and ~20 GHz fT/fMAX, 8.3. 1(2020).

    [15] A Raj, A Krishna, N Hatui et al. GaN/AlGaN superlattice based E-mode p-channel MES-FinFET with regrown contacts and >50 mA/mm on-current, 5.4. 1(2021).

    [23] S J Bader, R Chaudhuri, A Hickman et al. GaN/AlN Schottky-gate p-channel HFETs with InGaN contacts and 100 mA/mm on-current, 4.5. 1(2019).

    Tools

    Get Citation

    Copy Citation Text

    Huake Su, Tao Zhang, Shengrui Xu, Hongchang Tao, Yibo Wang, Yuan Gao, Yue Hao, Jincheng Zhang. Investigation of performance-enhanced GaN-based E-mode p-channel MOSFET with pre-ohmic-annealing treatment[J]. Journal of Semiconductors, 2024, 45(11): 112501

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Research Articles

    Received: May. 10, 2024

    Accepted: --

    Published Online: Dec. 23, 2024

    The Author Email: Zhang Tao (TZhang), Xu Shengrui (TZhang), Zhang Jincheng (JCZhang)

    DOI:10.1088/1674-4926/24050015

    Topics