After the realization of p-type conduction in Mg-doped GaN by Akasaki and Amanoet al. with low-energy electron beam irradiation (IEEBI) treatment in 1989^[1] and thermal annealing in N₂ ambient by Nakamuraet al. in 1992^[2], the studies on improving the electrical properties of GaN-based p-type layers have been carried out successively^[3-10]. Hall-effect measurements are commonly used to evaluate electrical properties. A good ohmic contact is generally required to obtain accurate values during the measurements. High work function metal such as Pd^[11-13] or Ni^[14,15] deposited in a high vacuum chamber is essential and special contact layers are usually adopted. It is time-consuming and much costly for the additional annealing of metal contact or the removal of contact layers for accurate measurements.

In this paper, a suitable contacting scheme consisting of p-InGaN and p⁺-GaN was grown on GaN-based p-type layers. After the annealing of p-type layers, indium was used as the contact metal. The electrical properties could be evaluated quickly and accurately with such simple procedures. The use of the contact structure was proved experimentally to have no influence on the measurement of p-(Al)GaN layers.

The GaN-based p-type layers and special contact layer were grown onc-plane free-standing (FS-) GaN substrates at atmospheric pressure in Taiyo Nippon Sanso (TNSC) horizontal MOCVD reactor (SR-4326KS). 1μm-thick unintentionally doped GaN was grown on FS-GaN substrate at 1000 °C, followed by 500 nm Mg-doped p-GaN or p-Al_0.14Ga_0.86N (2.5 nm)/GaN (2.5 nm) superlattices (SLs) and 22 nm heavily doped p⁺-GaN grown at 850 °C. After that, temperature was lowered to 740 °C to grow 2 nm p-In_0.2Ga_0.8N. Trimethylgallium (TMGa), trimethylaluminum (TMAl), trimethylindium (TMIn), ammonia (NH₃) and bis-cyclopentadienyl magnesium (Cp₂Mg) were used as the precursors for Ga, Al, In, N, and Mg, respectively. Hydrogen (H₂) was used as a carrier gas when growing (Al)GaN layers while nitrogen (N₂) was used for InGaN layers. The Mg doping level in bulk p-type layers were less than 2 × 10¹⁹ cm⁻³. Due to some delay in raising the doping level, 22 nm thick p⁺-GaN was grown to raise the Mg concentration up to ~ 1 × 10²⁰ cm⁻³. The Mg doping concentration could be made ×11 times higher in 22 nm thickness. This rate was confirmed by secondary ion mass spectroscopy (SIMS) using another sample at a deeper position apart from the surface to avoid the measurement error due to surface contamination. The doping level of Mg in the top 2 nm p-InGaN layer was estimated to be 4 × 10¹⁹ cm⁻³ by the Mg and Group III mole flow ratio and SIMS results of separately grown thicker InGaN layers.

V-pits would form during the growth below the temperature around 850 °C through threading dislocations (TDs) for Mg-doped p-GaN layers^[16]. FS-GaN substrates with low TDs were therefore used in this experiment to measure the electrical properties avoiding the V-pits formation. Common sapphire substrates can also be used when p-type growth temperature is higher than 900 °C which is also confirmed by our separate study.

Samples were annealed in a tube furnace in dry air at ~ 500 °C for 15 min to activate Mg in p-type layers. Then they were cut into 5 × 5 mm² squares andϕ 0.2 mm indium balls were pressed near the corners of the squares (diameters of the indium platelet became ~ 0.6 mm after pressing). The samples were heated up to 300 °C for several seconds on a hot plate to make close contact between indium dots and contact layer. The diameters of the pressed indium platelets shrunk a little to ~ 0.4 mm after heating. Resistivity and Hall-effect measurements with van der Pauw geometry were performed by Accent HL8800 to obtain sheet resistivity (R_sh), resistivity (ρ), Hall mobility (μ) and hole concentration (p) at room temperature. The magnetic field was 0.388 T and the currents were 100μA during the measurements.

Inductively coupled plasma (ICP) etching was carried out for selected Hall samples in an Apex SLR ICP system at 6.0 mTorr. The flow rate of boron trichloride (BCl₃) was 25 sccm and RF power was 55 W. Hall effects were measured again after ICP etching to check the influence of contact layer on the electrical properties of bulk p-type layers.

LinearI–V curves are essential to obtain accurate results from Hall-effect measurements. Therefore, annealing temperature (T_A) ~ 500 °C was chosen to avoid the degradation of p-InGaN contact layer to make a good ohmic contact. Different annealing temperatures from 440 to 520 °C were checked for p-GaN samples.I–V curves from different electrodes were also examined during the measurements. As theI–V curve shows in the inset ofFig. 1, ΔV is defined as the linear intercept of bulk p-type layers withx (voltage) axis. It is obvious that smaller ΔV refers to better linearity. It should be noted that each data point inFig. 1 is the average of 4 samples annealed at the specific temperature. Judging from theρ and ΔV fromFig. 1, 460 °C was almost enough to get good ohmic contacts while ~ 500 °C was necessary to further remove H from the bulk p-layers. Although not shown here, it was also confirmed with other samples that the ΔV increased with increasing the annealing temperature above 520 °C, which is probably caused by the oxidization of top InGaN layer in an oxygen containing ambient. As a result, the annealing temperature was chosen to be 510 °C to acquire both good contacts and sufficient activation of p-layers.

After optimizing the annealing temperature, the electrical properties of p-GaN and p-AlGaN/GaN SLs can be easily measured. 8 samples were taken from each one quarter of 2 inch wafer.Fig. 2 shows the results for p-GaN and p-AlGaN/GaN SLs, respectively.

Good linearity was confirmed inI–V curves between all the electrodes and typical examples are shown inFig. 2.F factors were 1.00 ± 0.01 in the measurements showing good symmetries. The solid lines are the average value of 5 samples inFigs. 2(b) and2(c) and 7 samples inFigs. 2(e) and2(f). The standard deviations ofρ,μ,p andR_sh were calculated to be ~ 3% for p-GaN samples and ~ 4% for p-AlGaN/GaN SLs samples. Not all the square samples could be measured because of the leakage due to pits or through side walls, the faults during cleavage of squares or the peeling off of soft In dots. Still more than 5 out of 8 samples showed consistent results. This is more than enough to get accurate data for p-type layers.

It’s necessary to check the influence of the contact layer on the electrical properties of bulk p-type layers. Therefore, step etching was applied for further analysis. During the etching procedure, not only the surface layer will be removed but also some etching damage will penetrate into the surface region (Fig. 3). The damage was mainly caused by the formation of nitrogen vacancy (V_N) during ICP etching^[17,18]. V_N is a donor defect and can compensate a hole and also hazards to p-type conduction by forming Mg–V_N complexes in GaN-based materials. Different etching method leads to different etching damage, so it is meaningful to determine the etching damage thickness. It is obvious that relatively weaker etching conditions as we chose is preferred in this experiment.

Etching depth (d_E) was measured by AFM to determine ICP etching rate. Since the p-AlGaN/GaN SLs samples have an average Al content of only ~ 7%, the etching speed should be similar for p-GaN and p-AlGaN/GaN SLs samples. As shown inFig. 4, linear fitting (broken line) can be expected from these data points. There is an offset ford_Ewhich may refer to unstable etching condition in the initial stage or due to the existence of oxidized layer at the surface.

The influence of contact layers was then carefully investigated. Average Mg concentration measured by SIMS was 1.3 × 10¹⁹ cm^–3 for bulk p-AlGaN/GaN SLs, which is quite uniform except the very initial stage and near the surface as shown inFig. 5(a). Since Hall measurements are used for the evaluation of uniform layers, the thickness of actual conductive layer (d_C) with a uniform hole concentration should be determined carefully. Owing to the delay of doping,d_C is obviously smaller than Mg doped layer thickness (d_Mg). The period thickness of p-AlGaN/GaN SLs was measured by high-resolution X-ray diffraction (HR-XRD) and fitted to be 4.99 nm (Fig. 7). So, the total thickness of p-AlGaN/GaN SLs was 499 nm, which was similar to our designed thickness. Since the contact layer was 24 nm-thick,d_Mg was calculated to be 523 nm.

The band diagram near the fresh p-InGaN/p⁺-GaN surface is simulated inFig. 6(b). The surface depletion width is mainly determined by the acceptor concentration and the barrier height at the surface. Owing to the strong spontaneous and piezoelectric polarization effect and lowed barrier height of InGaN, the valence band is tilted upward and the depletion width is 5 nm. Assuming the Fermi-level stabilization energy, the valence band at the surface is located at –2.22 eV for In_0.2Ga_0.8N^[19], the electric field in p-In_0.2Ga_0.8N layer is calculated to be 3.4 MV/cm^[20,21] and the conduction-to-valence-band offset ratio of InGaN/GaN interface is assumed asΔEC:ΔEV=60:40^[22] in the simulation. The SIMS data for the surface region is usually not so reliable due to surface contamination, thus a designed profile is assumed instead as shown inFig. 5(b). As the Mg concentration near the surface increases to approach ~ 1 × 10²⁰ cm^–3, the hole concentration tends to decrease due to self-compensation when Mg concentration is above 4 × 10¹⁹ cm^–3, which has been examined through our separate study. Here it should be pointed out that the highly doped layer near the surface does not necessarily imply higher conduction, but its resistivity tends to stay nearly the same or even becomes higher due to self-compensation. However, it still has a prominent effect to reduce the depletion width at the surface, facilitating the ohmic contact. The equivalent depth where uniform conduction starts (dCs) is calculate to be 5 nm considering the expected hole distribution near the surface. Same method is applied to determine the equivalent depth where uniform conduction ends near the starting position of Mg-doping (dCe), anddCe is calculated to be 498 nm. As a result, the effective conductive layer thickness with a uniform hole concentration before etching (dC0) is 493 nm.

4 samples marked as A1 to A4 were selected from previous 7 p-AlGaN/GaN SLs samples. The ICP etching time was varied as 6, 14, 22, 30 min for A1, A2, A3 and A4, respectively. Then,d_Ecould be calculated easily through the fitting curve inFig. 4. As we know,ρ can not be measured accurately if the thickness is not certain. However,R_sh which can be measured accurately without the knowledge of layer thickness.R_sh will definitely increase due to the decrease ofd_Cafter etching. The change ofR_sh before and after different etching time could be fitted to give the thickness of non-conductive layer near the etched surface (d_NC) consisting of a damage layer and an associated depletion layer, as can be written in Eqs. (1) and (2).

1Rsh=dC0dC⋅Rsh0,

2dC=dCe−dE−dNC,

whereRsh0 is the average sheet resistivity before etching.

Fig. 6(a) shows theR_sh after etching. The solid line, broken line and dotted line are the fitted ones usingR_sh expressed by Eqs. (1) and (2) withd_NC set to 0, 28 and 50 nm, respectively. It should be noted thatd_NC could increase with increasing etching time. However, a constant value ofd_NC = 28 nm also gives us a reasonable fitting inFig. 6(a) and is used in the subsequent calculation. The band diagram of etched surface when contact layers are completely removed is also simulated inFig. 6(c). Here, p-Al_0.14Ga_0.86N/GaN SLs is simplified as p-Al_0.07Ga_0.93N single layer. The valence band at the surface is positioned at –2.56 eV for Al_0.07Ga_0.93N^[19]. The depletion width is estimated to be 15 nm so that the etching damage depth is regarded as 13 nm. Then,d_C can be calculated properly and the electrical properties can be determined accurately in the Hall measurements using thesed_Cvalues. InFig. 6(d), measuredR_sh values including both non-etched and etched samples are plotted against estimatedd_C.R_sh is inversely proportional tod_C, which is indicative of assumed uniform doping. This result should justify our treatment to estimated_C.d_Mg implies the designed thickness of Mg-doped layer. The effective thicknessd_C is smaller thand_Mg because of the etching depth, surface and sub/epi interface depletion width, Mg doping-delay, damaged width, and the p⁺ layer carrier profile.

P-GaN samples were also analyzed using the same method and results are added inFig. 8 together with p-AlGaN/GaN SLs. The solid and broken lines and the data points at 0 nm inFig. 8 are the average values before etching. It is clear that the measured values ofρ,μ andp remain almost identical before and after etching. The contacting scheme is experimentally proven to be suitable and practical. The electrical properties of AlGaN/GaN SLs are summarized as shown inTable 1.

Since it is obvious that 2 nm p-InGaN is quite thin and exists within the depletion width, this p-InGaN/p⁺-GaN hetero-interface will have no significant effect on the bulk electrical properties while improving the contact properties. This contacting scheme to Pd-based metal has also been studied in our previous work^[23], exhibiting a relatively low specific contact resistivity of 4.9 × 10⁻⁵ Ω·cm² @J = 3.4 kA/cm².

The determined difference betweend_C andd_Mg is ~ 30 nm for unetched sample in this study. For convenience, this value is recommended for similar p-type structures in practice. If the sample structure is known,d_C can be calculated by the method described in this paper, and one can certainly improve the accuracy.

Quick and accurate Hall-effect measurements for GaN-based p-type layers were established by using suitable p-In_0.2Ga_0.8N (2 nm)/p⁺-GaN (22 nm) contacting scheme with In balls after optimizing annealing condition. The electrical properties of both p-GaN and p-AlGaN/GaN SLs samples were evaluated and the standard deviation of each value was ~ 3% and ~ 4%, respectively. The thickness of non-conductive surface layer was estimated to be ~ 28 nm for the etched samples and the etching damage was estimated to be ~ 13 nm. Identical values could be obtained before and after the removal of contact layer when reasonable conducive layer thickness was used.