GaN-based high electron mobility transistors (HEMTs) have important applications in wireless communications, phased array radar and other fields because of the high two-dimensional electron gas (2DEG) density and mobility^{[1, 2]}. The device structure of a GaN-based HEMT grown along the c-axis has two polarities: metal-polarity and nitrogen-polarity (N-polarity). Among them, metal-polarity HEMTs are widely used and studied. However, compared with metal-polarity HEMTs, N-polarity HEMTs have more advantages in high-frequency applications. The device structure of a N-polarity HEMT differs from that of a metal-polarity HEMTs, and its 2DEG channel is located on the GaN side near the gate. Generally, to realize high-frequency operation, the scaling of a HEMT structure is required. For a N-polarity HEMT, the 2DEG density and mobility hardly change with device dimension scaling^[3−6]. Therefore, a high output power can be maintained at a high operating frequency. However, for a metal-polarity HEMT, the 2DEG channel is located on the GaN side below the AlGaN barrier. Its 2DEG density decreases rapidly with device dimension scaling, which makes it difficult to realize high output power at a high operating frequency. In addition, a N-polarity HEMT enables a low Ohmic contact resistance and a stronger confinement of 2DEG in the channel^{[7, 8]}. In 2011, Brown et al. reported the fabrication of high-performance N-polarity HEMTs with a output power of 1.5 W/mm and a maximum power-added efficiency of 24.2% at 94 GHz^[9]. In 2018, the output power of the N-polarity HEMT reaches 7.94 W/mm at 94 GHz, which is about two times the output power of the metal-polarity HEMT in the W-band at that time^[10]. However, the Schottky barrier height (SBH) of N-polarity GaN is low and thus the reverse leakage is relatively high, which affects the performance of N-polarity HEMTs^[11−14]. This is mainly because of the higher background electron concentration and the reversal polarization field of N-polarity GaN compared to that of metal-polarity GaN^{[13, 15−18]}. Researchers currently solve this problem mainly by introducing an insulating layer between the gate metal and N-polarity GaN to form a metal–insulator–semiconductor (MIS) structure^[18−20]. However, this method increases the distance from the gate to the channel, and thus reduces the gate control capability of N-polarity HEMTs. Meanwhile, this method makes the device process more complex.

In this work, we studied the effect of the annealing process of the Schottky contact electrode (Ni/Au) on the electrical characteristics of N-polarity GaN Schottky barrier diodes (SBDs). In an N-polarity GaN-based SBD, the properties of Schottky contact to N-polar GaN can be well evaluated. Our results show that with the optimized annealing process the electrical characteristics of N-polarity SBD can be improved effectively. An SBD with a higher SBH and rectification ratio, and a lower ideal factor can be achieved by annealing. Moreover, our analysis suggests that the improvement of device performance after annealing is mainly owing to the reduction of the density of the interface state between the Schottky contact metals and the N-polarity GaN, and the increase of barrier height for the electron emission from the trap state at low reverse bias. This work provides a simple and feasible way to obtain high-quality Schottky contacts on N-polarity GaN for N-polarity GaN HEMTs.

All N-polarity GaN films are grown on sapphire substrates using an Aixtron CCS 3 × 2ˊˊ FT metal organic chemical vapor deposition (MOCVD) system. Prior to the growth, the substrate is thermally cleaned at 1060 ℃ for 5 min in the H₂ atmosphere. Then, the growth begins with a low-temperature GaN nucleation layer grown at 570 ℃. After that the growth temperature is ramped up to 1060 ℃ to grow a 2-μm-thick high-temperature GaN film. After the growth, we performed wet etching experiments to verify the lattice polarity of GaN^{[21, 22]}. Figs. 1(a) and 1(b) show the surface scanning electron microscopy (SEM) image of the GaN film before and after etching with 8 mol/L KOH at 40 ℃ for 5 min. It is obvious that the obtained GaN film can be etched by KOH, indicating that its lattice polarity is N-polarity. The crystalline quality and surface morphology of the N-polarity GaN can be found in our previous work^[23]. The crystalline quality of the N-polarity GaN in our work is comparable to that of metal-polarity GaN, as shown in Table 1. But the background electron concentration (N_D) of N-polarity GaN is higher than that of metal-polarity GaN (Table 1) due to the stronger bonding of oxygen atoms on the N-polarity surface^[17].

Then, we fabricated SBD devices based on the prepared N-polarity GaN films. Prior to the device process, all N-polarity GaN films were ultrasonically cleaned using acetone, ethanol and deionized water. After that, the SBD devices were prepared by a standard lithography process. The schematic diagram of a N-polarity SBD device is shown in Fig. 2(a). The Ohmic contact metals of SBD devices were Ti/Al/Ni/Au (20/100/40/50 nm) stack metals deposited by electronic beam evaporation. To form Ohmic contacts, contact metals were annealed at 770 ℃ in N₂ atmosphere for 15 s. After the Ohmic contact process is finished, Ni/Au (20/40 nm) Schottky contact metals were deposited. To investigate the effect of the annealing process of Schottky contact metals on the device performance, Schottky contact metals under the same deposition conditions were annealed with different times and temperatures. Fig. 2(b) shows the surface optical microscope (OM) image of an SBD device, where the diameter of the Schottky contact electrode is 140 μm. After the devices were prepared, current−voltage (I−V) characteristic and capacitance−voltage (C−V) characteristic curves of these devices were recorded by a keysight B2902A precision-type source meter and a keysight B1500A semiconductor device analyzer, respectively. Noting that the SBH, rectification ratio, ideal factor, and the density of interface state between Schottky contact metals and GaN of an SBD device can be calculated and analyzed by its I−V characteristic curve.

Firstly, we fabricated four N-polarity GaN SBD devices with various annealing times (30 , 60 , 120 , and 240 s) but a constant annealing temperature of 450 ℃. In addition, we also fabricated a N-polarity GaN SBD device without annealing for comparison. The current density−voltage (J−V) curves of these devices are shown in Fig. 3. The relationship between J and I is:

1J=IA,

where A is the area of the Schottky contact electrode. Fig. 3(a) shows the J−V curves in linear scale for these SBDs with different annealing times, from which the turn-on voltages (V_on) of SBD devices with different annealing times can be extracted. The results show that with the increase of annealing time, the V_on of SBD devices first increases and then decreases. The device with an annealing time of 60 s has the highest V_on of 0.78 V. Fig. 3(b) shows the J−V curves in semilog scale, which provides a clearer view of the reverse leakage current density of each device. We can see that the SBD device without annealing has the highest leakage current density of 73 A/cm² at −5 V. After annealing, the reverse leakage current density of the SBD device decreases significantly. At an annealing time of 60 s, the SBD device has the lowest reverse leakage current density of 0.34 A/cm² at −5 V.

The I−V characteristic curve of a SBD device can be analyzed with the standard thermionic emission theory^{[24, 25]}:

2I=Is(eqV/nkT−1).

At a forward bias V > 3kT/q, Eq. (1) can be simplified as:

3I=IseqV/nkT.

Taking the natural logarithm on both sides of the equation:

4lnI=lnIs+qnkTV,

where q is the absolute value of the electron charge, n is the ideal factor, k is the Boltzmann's constant, T is the test temperature. And I_s is the reverse saturation current, it can be expressed as follows^{[24, 25]}:

5Is=AA*T2e−ϕB/kT,

where ϕB is the SBH and A* is Richard's constant. The expression of A* is as follows^[26]:

6A*=4πqm*k2h3,

where m* is the electron effective mass of GaN (m* = 0.22 × m₀^[27], m₀ is the free electron mass), and h is Planck's constant. Fig. 3(c) shows the lnI−V curve of the SBD with an annealing time of 60 s. The lnI−V curve is fitted with Eq. (4), as plotted with red dotted line shown in Fig. 3(c). It can be seen from Eq. (4) that the slope of the dotted line is q/nkT. Therefore, an ideal factor of 1.54 is obtained. The intercept of the dotted line from the longitudinal axis is lnI_s. As a result, the SBH value of 0.77 eV is calculated with Eq. (5).

Fig. 4 shows the calculated ideal factor, SBH, and rectification ratio (I_on/I_off) at ±5 V for SBD devices with different annealing times, where the rectification ratio provides a better representation of the reverse leakage current. After annealing, the SBH of SBD devices is increased, and the ideal factor is reduced, as shown in Fig. 4(a). The SBD device has a higher SBH and a lower ideal factor at an annealing time of 60 and 120 s. In addition, as can be seen from Fig. 4(b), the rectification ratios of SBD devices are improved after annealing, and the SBD with an annealing time of 60 s has the largest rectification ratio of 7700. This means that the reverse current leakage of a GaN SBD is greatly suppressed. In addition, the ideal factor greater than one may be due to the presence of natural oxides on the surface of GaN or the inhomogeneous of the Schottky barrier^[28].

The difference of the SBH values between the SBDs with an annealing time of 60 and 120 s is small (0.01 eV), however, the difference in rectification ratios between these two devices is large (5700). This indicates that the reduction of reverse leakage current is not completely dependent on the rise of the SBH. This is because the reverse leakage current is also related to the interface states between the contact metals and GaN surface, and the SBH can also be affected by Fermi level pinning effect^[29]. The above analysis results show that Schottky contacts with good electrical characteristics can be achieved on N-polarity GaN by annealing at 450 ℃ for 60 s.

Based on the optimized annealing time of 60 s, we fabricated another four N-polarity GaN SBD devices with various annealing temperatures (300, 375, 525, and 600 ℃). Fig. 5(a) shows the linear scale J−V curves of these SBD devices. The J−V curve of the SBD device with an annealing temperature of 450 ℃ is also plotted in Fig. 5(a). We can see that the V_on of the SBD device increases and then decreases with the annealing temperature, and the V_on is the largest when the annealing temperature is 450 ℃. Fig. 5(b) shows the semilog scale J−V curves of these SBD devices. and it can be clearly observed that the reverse leakage current density of the device decreases and then increases with the annealing temperatures. At the annealing temperature of 450 ℃, the SBD has the lowest reverse leakage current density. Further increasing the annealing temperature to 600 ℃, the device reverse leakage current density increases dramatically and is higher than that of the SBD device without annealing. This is mainly because the high annealing temperature enhances the interaction between GaN and the contact metals^{[30, 31]}.

Fig. 6 shows the calculated ideal factor, SBH, and rectification ratio at ±5 V for SBD devices with different annealing temperatures. We can see that the SBD device with an annealing temperature of 450 ℃ has the best electrical characteristics, it has an SBH of 0.77 eV and a rectification ratio of 7700 at ±5 V, while it has an ideal factor of 1.54. At −1 V, its leakage current density is 1.9 × 10⁻⁵ A/cm², as shown in Fig. 5(b). The performance of this SBD device is comparable to that of recently reported N-polarity GaN SBD with a SiN interlayer between the Schottky contact metals and N-polarity GaN (SBH: 0.8 eV; reverse leakage current density at −1 V: ~10⁻⁵ A/cm²)^[18]. This indicates that the electrical properties of the Schottky contacts to N-polarity GaN can be improved effectively by optimizing the annealing process.

To reveal the influence mechanism of annealing process on the electrical characteristics of SBD devices, we studied the interface characteristics between the Schottky contact metals and N-polarity GaN on I−V curves. The density of interface state (N_SS) between the Schottky contact metals and N-polarity GaN can be calculated according to Card and Rhoderick’s method^[32]:

7NSS=1q{εiδ[n(V)−1]−εsWD},

where ε_i is the relative permittivity of the thin oxide dielectric layer between GaN and Schottky contact metals, δ is the thickness of this dielectric layer. The values of ε_i and δ do not affect the comparative analysis of the density of interface state. Here, ε_i = 7ε₀, δ = 1 nm are adopted for calculation^[28]. ε₀ is the vacuum permittivity. ε_s is the permittivity of the GaN. W_D is the space charge region width on the GaN side. W_D is defined as follows:

8WD=2εs(ϕB/q−V−kT/q−Vn)qND,

9Vn=kTqlnNCND,

where N_C is the conduction band state density of GaN. n (V) is the voltage-dependent ideal factor function and it can be defined as follows^[33]:

10n(V)=qVkTlnIIs.

Then, the energy distribution of the interface states can be estimated with n(V) and the effective SBH ϕe^[34]:

11EC−ESS=q(ϕe/q−V),

12ϕe=ϕB+[1−n(V)]qV,

where E_SS is the energy of the interface states and E_C−E_SS is the position of the interface state energy at the interface relative to the conduction band. Based on the above equations, we obtain the relationship curves between N_SS and E_C−E_SS for SBD devices with different annealing process, as shown in Figs. 7(a) and 7(b). It can be observed from Fig. 7(a) that the density of interface state of SBD device without annealing are higher, between 2 × 10¹³ eV⁻¹∙cm⁻² and 4 × 10¹³ eV⁻¹∙cm⁻². Meanwhile, the interface state densities are decreased after annealing, as shown in Fig. 7(a). However, the density of interface state (between 4 × 10¹³ eV⁻¹∙cm⁻² and 1.2 × 10¹⁴ eV⁻¹∙cm⁻²) of the device with an annealing temperature of 600 ℃ is higher than that of the device without annealing, as shown in Fig. 7(b). At an optimized annealing process (450 ℃ for 60 s), the density of interface state is reduced to a relatively low value, and thus the SBD device has a lower ideal factor and a reverse leakage current density as well as a higher SBH.

Moreover, we analyzed the current leakage mechanism of N-polarity GaN SBD devices, because the SBH and the current leakage of an SBD is also strongly dependent on its current leakage mechanism. The current leakage mechanism of the SBD devices can be investigated by Poole−Frenkel emission (PFE) model and Fowler−Nordheim (FN) tunneling model. Among them, PFE model is a temperature-dependent leakage current mechanism at a weak electric field, and its leakage current I_PF can be defined as follows^[35]:

13IPF=Zξexp[−q(ϕt/q−qξ/πεsh)kT],

where Z is a constant, ϕt is the barrier height for the electron emission from the trap state and ε_sh = 5.8ε₀ is the permittivity of the semiconductor at high frequency. The electric field (ξ) of the Schottky contact can be calculated from the results of C−V measurement^[36]:

14ξ=CεsA(Vbi−Vn),

where V_bi is built-in voltage. The measured C−V curves under reverse bias for SBD devices without and with annealing at 450 ℃ for 60 s are shown in Fig. 7(c). The values of V_bi for SBD devices without and with annealing can be extracted from the 1/C²−V curve^[12], which are 0.78 and 1.26 V, respectively, as shown in the inset of Fig. 7(c). In addition, it can be found that the C of the SBD device without annealing shows an unusual decay at the voltage range from −2 to −5 V, and decreases to negative values at the voltage around –4 V, which may be due to the high density of N_SS^{[37, 38]}. Based on the I−V and C−V measurement results, we obtain the ln(I_PF/ξ)−ξ^0.5 curves for these two SBD devices in the low reverse bias range from 0 to −2 V, as shown in Fig. 7(d). We can see that ln(I_PF/ξ) and ξ^0.5 for both SBD devices are approximately linearly related. The fitted slopes of the ln(I_PF/ξ)−ξ^0.5 curves for the SBD devices without and with annealing are 0.01 and 0.013, respectively, which are close to the theoretical value of 0.012 calculated based on Eq. (13). This suggests that the PFE is the main current leakage mechanism at low reverse bias. Fig. 7(e) shows the measured lnI−1/T curves for the SBD devices without and with annealing at 450 ℃ for 60 s at temperatures ranging from 150 to 250 K at −1.3 V. We can see that lnI is linearly related to 1/T, which is consistent with Eq. (13). The fitted ϕt of the SBD devices without and with annealing are 0.13 and 0.42 eV, respectively. Therefore, it is suggested that the optimized annealing process raises ϕt, which leads to the reduction in the leakage current.

In addition, we analyze the current leakage mechanism for these two SBD devices at higher reverse bias. The current leakage mechanism at strong electric field can be studied by the temperature-independent FN tunneling model. In this model, the leakage current I_FN can be defined as follows^[39]:

15IFN=Dξ2exp(−Bξ),

where D is a constant, and B can be expressed as^[39]:

16B=8π2m*(ϕeFN)33qh.

Fig. 7(f) shows the ln(I_FN/ξ²)−ξ⁻¹ curves of the SBD device with annealing at 450 ℃ for 60 s. The ln(I_FN/ξ²)−ξ⁻¹ curve of SBD device without annealing is not shown in Fig. 7(f) because its C−V curve shows an unusual decay at bias greater than −2 V, as shown in Fig. 7(c). We can see from Fig. 7(f) that ln(I_FN/ξ²) and ξ⁻¹ are approximately linearly related, which is consistent with Eq. (16). This suggests that the FN tunneling is the main current leakage mechanism at higher reverse bias.

To understand the effect of annealing process on the current leakage of N-polarity SBD devices intuitively, we plot the schematic energy band diagrams of SBD devices without and with annealing at 450 ℃ for 60 s, respectively, as shown in Fig. 8. As we mentioned above, the density of interface states of SBD devices without annealing is higher than that with annealing at 450 ℃ for 60 s. Meanwhile, it is worth noting that the energy of the interface states is close to the conduction band edge, as shown in Figs. 7(a) and 7(b). Therefore, we can infer that these interface states above the Fermi level would be positively charged due to the ionization, and the density of the positive interface charges of the SBD device without annealing is higher than that with annealing. This means that the SBD device without annealing has a lower SBH ϕB compared to the SBD device with annealing. As we marked in Figs. 8(a) and 8(b), the values of SBH ϕB for the SBD devices without and with annealing are 0.51 and 0.77 eV, respectively, according to our calculation results shown in Fig. 4(a) and Fig. 6(a). In addition, the schematic diagrams of current leakage under FN model and PFE model are also plotted in Fig. 8. The barrier height ϕt for the electron emission from the trap state of the SBD device with annealing (0.42 eV) is higher than that of the SBD device without annealing (0.13 eV), according to the fitting results mentioned above. Therefore, from the energy band diagram, we can see intuitively that the SBD device with annealing has a lower reverse leakage current compared with the SBD device without annealing due to the higher SBH and ϕt.

Lastly, to facilitate the understanding of the physical notations in this work, we list some of the main physical notations and their corresponding physical names in Table 2.

In summary, we have grown N-polarity GaN films by MOCVD and fabricated N-polarity GaN SBD devices through the photolithography and metal deposition process. Importantly, we studied the influence of the annealing process of Schottky contact metals on the electrical performance of SBD devices. Our results show that the annealing treatment is conducive to improve the SBH, and reduce the ideal factor and reverse leakage current density of the SBD device. This is mainly because annealing can reduce the density of interface state between Schottky contact metals and the N-polarity GaN and increase the barrier height for the electron emission from the trap state, but too high an annealing temperature will greatly increase the density of interface state and seriously degrade the electrical properties of the SBD device. It is reasonably believed that this work provides a feasible way to realize high-quality Schottky contacts on N-polarity GaN for N-polarity GaN-based HEMTs.