Low-resistivity vertical current transport across AlInN/GaN interfaces

Effects of n-type doping of Al0.82In0.18N/GaN heterostructures on the conduction band (CB) profile have been investigated. Doping concentrations well above 1019 cm−3 are required to reduce the large barriers in the CB. Experimentally, Si- and Ge donor species are compared for n-type doping during metalorganic vapor phase epitaxy. For Si doping, we find substantial interface resistivity that will strongly contribute to total device resistivity. Doping of AlInN is limited by either the onset of a self-compensation mechanism (Si) or structural degradation of the AlInN (Ge). Only by Ge doping, purely ohmic behavior of periodic AlInN/GaN layer stacks could be realized.

S emiconductor microcavities in the group-III nitride material system find application in laser diodes and light-emitting diodes. 1,2) Lattice-matched Al 0.82 In 0.18 N (LM-AlInN) can be grown on GaN with high quality allowing for a virtually unlimited number of layer pairs as mechanical stress may not be incorporated during growth. 3,4) The refractive index contrast between these layers of about 7% is very attractive for edge emitting laser diodes and vertical cavity semiconductor lasers. 5,6) Yet, the conduction band (CB) offset between LM-AlInN and GaN amounts to 1.0 eV. 7) Also, the wurtzite nature of these materials gives rise to spontaneous polarization charges at each interface. As a consequence, large barriers build up in the CB of AlInN/ GaN heterostructures which increases electrical losses in electrical injection lasers and LEDs. With the conventional growth direction in (0002) positive net polarization charges are located at the lower interface of AlInN/GaN heterostructures while negative charges exist at the upper interface. A comprehensive analysis is required on how impurity doping has to be applied to reduce such barriers.
In GaN bulk material, Si and Ge are known shallow donors with similar activation energies of about 17-19 meV for a doping level near 10 17 cm −3 . 8) Reports about n-type doping of AlInN are scarce since uniform layer growth, as needed to perform reliable Hall-effect measurements, is limited to near lattice-matched composition and to about 100-200 nm thickness. 9) A theoretical calculation of Ge substitutional donors in Al x Ga 1−x N presented by Gordon et al., predicts DX center formation for x Al > 0.52. 10) Blasco et al. however demonstrated n-type conductivity in strained Al x Ga 1−x N:Ge/GaN layers grown by molecular beam epitaxy up to x Al = 0.66 with activation energies between 5 and 40 meV. 11 by the onset of 3D growth which is caused by SiN formation. 12) Still, conductive AlInN/GaN Distributed Bragg Reflector (DBR) structures grown by MOVPE with ohmic characteristics were demonstrated by the application of delta-like Si doping (6.0 × 10 19 cm −3 ) and implementation of graded DBR interfaces. However, graded AlInN/GaN interfaces require a complex growth process and the achieved resistance of 17 Ohms is still relatively high. 13,14) On the other hand, Germanium (Ge) doping of GaN during MOVPE growth is only limited by the onset of self-compensation at around [Ge]∼3-4 × 10 20 cm −3 . 15,16) Ge doping of AlInN has not been reported previously. The generally weak stability of AlInN during MOVPE growth may put limits on the doping with Ge atoms. In addition, Ge doping in MOVPE is known for a memory effect leading to non-uniform incorporation during growth. 17) In this report, we investigate doping and growth of periodic AlInN/ GaN layer stacks that can serve as one-dimensional photonic crystals for mode confinement in optical waveguides and cavities. As the simulation of the CB profile suggests, heavy n-type doping of AlInN and GaN is beneficial for electrical transport. Accordingly, the limits of n-type doping of AlInN by either Si-or Ge-atoms in MOVPE deserve investigation. For both species, very high doping levels may not be achievable due to either growth chemistry or incorporation issues. This report, therefore, explores Si-and Ge doping in AlInN with the goal to achieve low-resistive vertical current transport across AlInN/ GaN heterostructures. For doping concentrations below 10 19 cm −3 large barriers exist in the CB which originate from the upper AlInN/GaN interface as seen in Fig. 1(a). These barriers extend across the whole AlInN layer but also into the upper GaN layer. At the lower GaN/AlInN interface electron accumulation pins the band bending of the heterostructure. At the upper AlInN/GaN interface permanent negative polarization charges and the CB offsets between GaN and AlInN are the contributing physical phenomena to these barriers. A reduction of the barrier at the upper AlInN/GaN interface could be achieved upon heavy doping of the GaN layers up to 10 20 cm −3 . Mostly, the penetration width of the electronic barrier into the upper GaN section gets reduced. Across the AlInN layer, the barrier is still present and height and width would still amount to roughly 1 eV and >20 nm, respectively. As seen in Fig. 1(b), the barrier height and width can be significantly reduced for a doping level well above 1 × 10 19 cm −3 , if applied both to GaN as well as to AlInN layers. For doping levels of 5 × 10 19 cm −3 (AlInN) and 1 × 10 20 cm −3 (GaN), activation energies below 100 meV for the donor species in AlInN would have only little effect on the CB profile [ Fig. 1(c)]. The simulation results suggest the use of Ge as the dopant in MOVPE growth of AlInN/GaN structures, as it offers at least for GaN growth to reach the required high doping concentrations.
For the experimental analysis, AlInN/GaN stack structures have been grown by MOVPE in an AIX200/4 RF-S reactor with standard precursors (TMAl, TMGa, TMIn, and NH 3 ) on sapphire substrates. The sequence of In 0.18 Al 0.82 N and GaN layers was grown at 780°C and 1050°C, respectively, on top of GaN/AlGaN/AlN buffer structures. Details of the buffer growth and the AlInN/GaN growth are already published. 19) Parameters for AlInN growth with nitrogen as the carrier gas and T growth = 780°C have to be very different from standard GaN growth in order to incorporate In into the layer. Furthermore, the growth rate for AlInN is only 0.14 μm h −1 at which smooth layer surface morphologies and best crystallinity are obtained as inferred from X-ray diffraction measurements. The angular position of the AlInN(0002) reflection is taken for calculation of the In composition applying a linear change of the c lattice constant.
Doping of the layers with either Si or Ge atoms is obtained with silane (SiH 4 ) or isobutylgermane (IBGe) as precursors, respectively. One pair of the layer stack comprises 35 nm In 0.18 Al 0.82 N/200 nm GaN. Hall-effect measurements on GaN samples were performed for calibration of the free electron concentrations at corresponding levels. For the AlInN layers, doping levels are difficult to extract by Halleffect measurements for reasons already mentioned. Optimum Ge doping in AlInN was evaluated for the vertical electrical resistance of single and stacked AlInN/GaN layers. In order to determine the onset of structural degradation for Ge doping within AlInN, we assessed structural properties by high-resolution X-ray diffraction and atomic force microscopy. Vertical electrical transport behavior through the AlInN/GaN stack was investigated by recording currentvoltage (I-V ) characteristics on 1 × 1 mm 2 large mesa structures with top metal electrodes of 200 × 200 μm 2 . These mesa structures were processed by dry etching and photolithography methods followed by deposition of a Ti (14 nm)/Al(33 nm)/Ni(10 nm)/Au(60 nm) metal stack by evaporation. In order to extract the interfacial resistances of the Al 0.82 In 0.18 N/GaN stack a modified transmission line measurement (TLM) method was used. Thereby, the current transport through pairs of 200 × 100 μm 2 sized mesa structures with varying distances was measured. Temperature-dependent I-V measurements were done in a liquid-nitrogen cooled cryostat. Electrical four-point-probe measurements were performed to obtain precise resistance values.
From secondary ion mass spectroscopy (SIMS) analysis of a series of Ge-doped AlInN layers sandwiched between steadily Ge-doped GaN layers a linear incorporation behavior is found at T growth = 780°C. In this series, the molar flow rate of IBGe was varied between 0.6 and 6 μmol min −1 , which corresponds to a vapor phase concentration relative to TMIn  respectively. The smaller, more frequent oscillations are due to the finite thickness of the AlInN and their presence proves sharp interfaces between AlInN and GaN. They have been used to confirm the thickness of the AlInN layer. The shift of the AlInN diffraction peak towards larger angles marks a significant loss of In, leading to strained layer growth. This may also lead to the formation of the graded AlInGaN interface seen in the SIMS profiles. Such graded interfaces, as well as increasing surface roughness, causes a damping of the thickness oscillations towards the low-angle side of the diffraction diagram. Regarding the polarization charges, the substitution of In with Ga may lead to a still unstrained quaternary Al 0.82 (Ga x In 1−x ) 0.18 N layer. One would then expect an ever higher spontaneous polarization which would increase the band bending. Compensation of the stronger band bending would require even higher n-type doping levels in the AlInN layers.
Indium desorption from InGaN and InAlN layers due to the presence of hydrogen in the gas phase is commonly known for In containing nitrides. [19][20][21][22] When similar experiments were done with hydrogen-diluted GeH 4 (10% in gas mixture) dopant source we noticed an even more dramatic loss of In. However, AlInN layers doped with a SiH 4 /H 2 mixture (100 ppm SiH 4 ) at a SiH 4 /(TMAl+TMIn) ratio of 10 -5 exhibited virtually no change in composition (not shown), although the amount of hydrogen introduced in the reactor chamber was similar. We conclude that hydrogen released during pyrolysis of the precursors reduces the In content of the AlInN. A size effect due to Ge atoms (similar size as Ga) can also have an effect on the In content but the change of the In content by a few percent is hardly explained by Ge doping levels of around 10 20 cm −3 . Another cause may exist by surface segregation of Ge atoms hindering In atoms to attach to the surface during growth. Through the reduction of the AlInN growth temperature by around 30°C this precursor-related In loss is compensated and single AlInN:Ge layers with unchanged nominal In content with regard to undoped AlInN reference samples have been obtained at an IBGe input molar flow of 0.6 μmol min −1 .
In the following, we compare Si-and Ge-doping of GaN/ AlInN stacks with regard to current transport across the GaN/ AlInN interface. For that purpose, I-V curves were recorded from mesa structures with top-side contacts on the mesa surface and bottom contacts on the n-doped GaN buffer below the stack as shown in the insets of Fig. 4(b). In a first series we varied the type of dopant (Si or Ge) in AlInN and GaN while in a second series the number of layer pairs was changed (1, 3, and 10 pairs). Figure 4   mesas in this measurement had a size of 100 × 200 μm 2 . A reference sample consisting only of GaN:Si (solid black curve) exhibits a linear current-voltage dependence with a total series resistance (including setup resistance) of 10 Ω. Those samples with undoped and lowly doped AlInN are very resistive (>1 kΩ) and exhibit nonlinear I-V curves. While the resistivity across the AlInN/GaN stack reduces significantly with increasing doping levels, even at the optimum SiH 4 flow of 0.1 μmol min −1 the nonlinear trend around 0 V bias evidently proves non-ohmic characteristics which is assigned to residual interface charges between AlInN and GaN. At SiH 4 -flows of 0.2 μmol min −1 , the resistivity and nonlinearity eventually increase again which hints to a self-compensation mechanism as structural degradation is not revealed by X-ray diffraction analysis. The doping series is concluded with a sample where the free electron concentration in the surrounding GaN layers is increased to [n]∼4 × 10 19 cm −3 now using Ge as the dopant. An ohmic I-V characteristic, similar to the reference GaN:Sionly sample, is observed. All these trends are in qualitative agreement with our initial simulations. Please note that the presence of a Ge-related memory effect may lead to autodoping of the AlInN layer when grown successively after a GaN:Ge layer. In Fig. 4(b), I-V curves of the second series consisting of 1, 3, and 10 pairs of AlInN:Ge/GaN:Ge layers are shown. Here, large 1 × 1 mm 2 mesas were processed for the I-V measurements. The free electron concentration level in GaN:Ge is around [n]∼2 × 10 19 cm −3 according to calibration samples while the molar flow of IBGe within the AlInN was set to 0.32 μmol min −1 (0.4% IBGe/ TMIn + TMAl) where the structural quality of the AlInN layer is still preserved. All stacks show ohmic characteristics and low series resistances. While an exact value of the Ge concentration in AlInN remains to be determined, the free electron concentration in AlInN must reach levels above 10 19 cm −3 according to our simulation in order to eliminate the upper barrier in the CB.
There is a non-monotonic increase of the series resistance with the number of layer pairs in this series but for 1 and 3 AlInN/GaN layers in the stack the setup-related resistance dominates the measurement. For quantitative evaluation of the series resistance, TLM measurements were applied to a 10-pair Ge-doped AlInN/GaN stack with two different current transport schemes [ Fig. 4(c)]. In a lateral transport scheme, 100 × 200 μm 2 contacts with a set of increasing spacings were deposited on the surface, while for the vertical transport scheme, the semiconductor material between the contacts was etched down to the n-doped GaN buffer layer. These two TLM measurement configurations allow us to separate the resistivity of the stack structure (as a contact resistance value) from the contact resistivity value of the metal/semiconductor contacts. For the lateral transport, the contact resistance as calculated from the TLM I-V data was assumed to be dominated by the metal/semiconductor contact resistance. TLM analysis of the second configuration (with mesa etching) yields a contact resistance that is the sum of both metal/semiconductor contacts and resistance of the stack structure. With the specific contact resistance of ρ c ∼ 3 × 10 -6 Ω cm 2 a value of 0.15 × 10 -6 Ω cm 2 per AlInN/GaN interface is determined, about two orders of magnitude lower than the contact resistance as deduced from the data given in Ref. 14. Upon increasing the number of layer pairs to 30, as in typical application scenarios such as vertical optical microcavities, we note a significant loss of optical transparency in such Ge-doped structures which requires further optimization (will be published elsewhere).  Temperature-dependent I-V curves were recorded in a liquid-nitrogen cooled cryostat over a temperature range of 77-400 K in 20 K steps. Electrical connections were wirebonded onto respective bond pads on the sample enabling vertical current transport measurements in four-point-probe geometry. As Fig. 5 shows only a weak temperature dependence with a 10% decreasing resistance towards lower temperatures is found. This is similar to metallic behavior and indicates the existence of a degenerate electron gas in the stack.
The impact of n-type doping with either Si or Ge during MOVPE of periodic AlInN/GaN stack structures has been investigated in terms of structural and electrical characteristics. A donor concentration well above 1 × 10 19 cm −3 in both AlInN and GaN is advantageous to reduce polarization charge induced barriers at AlInN/GaN interfaces. According to our experiments, AlInN cannot be doped to yield such high free electron concentration due to a self-compensation effect in the case of Si and due to structural degradation in the case of Ge. Free electron concentrations of [n] > 4 × 10 19 cm −3 in the GaN layers which are easily achievable. Using Ge doping, we find a significant reduction of the series resistance of AlInN/GaN layers with ohmic current transport characteristics even at cryogenic temperatures (77 K). Therefore, Ge doping is favorable for periodic AlInN/GaN stack structures with many such interfaces. As an example, we have demonstrated 10xAlInN/GaN periodic layer stacks with specific contact resistances of 1.5 × 10 -7 Ω cm 2 per interface.