1. Introduction

In the past decade, gallium nitride (GaN) has been widely used in fabricating ultraviolet, blue and green light-emitting diodes (LEDs) or laser diodes (LDs) because of the merits of its direct and wide bandgap (3.4 eV), high carrier mobility, and good thermal and chemical stability [1–3]. Although GaN/Si heterostructures were also deemed to be promising candidates for making integrated high-speed or high-power photoelectronic devices [4–6], the practical course was badly baffled by the inferior interface quality resulted from the large lattice mismatch between the two semiconductors [7], because an inferior interface quality would bring serious damage to both the inner quantum efficiency (IQE) and the performances of as-constructed devices. To reduce the lattice mismatch and thereby improve the interface quality, two main approaches have been developed in the past years [8, 9]. One was the usage of an intermediate layer to accommodate most of the lattice mismatch and enhance the metallurgical compatibility between GaN and silicon substrates. This was usually realized by introducing a specially designed multilayer with a lattice constant gradient or an N-ion implanted substrate surface with partially distorted crystal lattices [10, 11]. The other was the adoption of nanoheteroepitaxy method, which was often carried out by utilizing various nanopatterned substrates to release the interfacial stress, reduce the dislocation density and improve the IQE [12–14]. For example, the GaN-Si interfacial residual stress could be greatly reduced by growing GaN film on a silicon nanopore array [15], and good rectifying properties could be obtained by growing p-GaN nanowires on n-Si crystal wafers [16]. These results greatly promoted the confidence for making GaN/Si-based LEDs or LDs with high efficiency and broad emission band range.

Encouraged by these experiments and utilizing silicon nanoporous pillar array (Si-NPA) [17] as functional substrates, we have constructed a GaN/Si nanoheterostructure array (GaN/Si-NPA) by growing GaN nanograins onto Si-NPA, in which an effective yellow or infrared (NIR) electroluminescence (EL) tuned by the applied voltages was obtained [18]. This indicates that GaN/Si-NPA might be a promising material system for fabricating practical GaN/Si-based LEDs. According to the basic theory of luminescence, the adjustability of the EL wavelength inferred that there might have different radiative recombination paths in GaN/Si-NPA, such as the band-band transition or the transitions relating with the high-density defects formed in GaN or at the interfaces. Clearly, the co-existence of multi-recombination paths would produce strong effect on the EL qualities, both the EL intensity and monochromaticity. On the other hand, thermal treatments have been proved to be an effect approach to promote the EL properties of a semiconductor heterojunction through improving the interfacial quality and changing the electronic structures. For instance, the carrier concentration of n-ZnO/p-Si could be changed through annealing treatment and the J-V curve as well as the EL properties could be adjusted notably [19]. The EL intensity and peak position of n-ZnO nanorods/p-GaN LED could be tuned through controlling the concentration and sorts of the defect states by performing annealing treatments at different temperatures and in different atmospheres [20]. As a result, a systematic study of the annealing effect on the EL properties is necessary for both clarifying the luminescent mechanism and promoting the emission qualities of GaN/Si-NPA.

In this paper, two GaN/Si LEDs were prepared based on as-grown and annealed GaN/Si-NPA. The structural and physical properties, including the X-ray diffraction (XRD) patterns, surface morphologies, current density-voltage (J-V) curves, EL and photoluminescence (PL) spectra, were measured and comparatively studied. Based on the experimental results, the EL mechanisms of the LEDs were put forward through building up the corresponding electronics structures. Our results might indicate a novel approach for designing and fabricating high-performance LEDs based directly on GaN/Si nanoheterostructures.

2. Experimental details

Si-NPA was prepared by hydrothermally etching (111) oriented, boron-doped single crystal Si (sc-Si) wafers in a solution of hydrofluoric acid containing ferric nitrate [17]. A thin layer of platinum (~3 nm), which acted as catalyst in the subsequent GaN growing process, was pre-deposited on freshly prepared Si-NPA samples by a magnetron sputtering technique. Using high-purity metal Ga (99.999%, 0.8 g) and NH₃ gas (99.999%, introduced with a flow rate of 20 sccm) as the sources for the two elements, GaN were grown on Si-NPA by a chemical vapor deposition (CVD) method. The deposition was carried out in a vacuum tube furnace equipped with multichannel gas inlets and a gas mixing chamber at 1050 °C for 20 min. Here two kinds of GaN/Si-NPA were prepared, one was the as-grown sample and the other was annealed at 800 °C for 3 hours in nitrogen atmosphere afterwards. Layers of indium tin oxide (ITO, ~100 nm) acting as top electrode and Al (~500 nm) acting as back electrode were deposited by magnetron sputtering and vacuum evaporation methods, respectively. As-constructed LEDs have a device structure of ITO/n-GaN/p-Si-NPA/sc-Si/Al. For the convenience of narration, the two LEDs here were named as as-grown LED and annealed LED, respectively. The two LEDs were both annealed at 300 °C for 1 hour in Ar atmosphere to realize ohmic contact between the electrodes and the semiconductors. The active areas of the diodes were specified as 10 mm × 10 mm. The surface morphology and the crystal structure of GaN/Si-NPA were characterized by a field emission scanning electron microscope (FESEM, JSM 6700F) and an X-ray diffractometer (Panalytical X' Pert Pro). The electrical and luminescent properties of the devices were measured at room temperature through an electrical group system consisted of Sourcemeter-2400 (Keithley) and a fluorescence spectrometer (Spex Fluorolog-3), respectively.

3. Results and discussion

The XRD patterns of as-grown and annealed GaN/Si-NPA are shown in Part A of Fig. 1(a) , in which all the diffraction peaks were indexed to crystalline hexagonal wurtzite GaN (JCPDS card: No. 50-0792). The obvious difference between the two curves is the reduction of the full width at half maximum (FWHM) for all the corresponding diffraction peaks after annealing treatment, as could be seen more obviously in Part B of Fig. 1(a). The typical cross-sectional FESEM image of as-grown GaN/Si-NPA is given in Fig. 1(b), in which GaN layers characterized by two different morphologies were observed. The upper layer was composed of two kinds of quasi one-dimensional GaN nanostructures, straight nanowires with an average diameter of ~30 nm and pencil-like nanorods with an average diameter of ~300 nm. Both the nanowires and the nanorods were well separated and nearly aligned locally perpendicular to the substrate surface, with an average length of ~1.5 μm. Between the nanowire/nanorod layer and Si-NPA substrate was a granular layer consisted of large quantities of GaN nanocrystallites (nc-GaN). The layer thickness and the average grain size were ~150 nm and ~20 nm, respectively. No apparent morphological variation was found by comparing the FESEM images of the samples before and after annealing treatment. Therefore, it was reasonable to think that the reduction of the FWHM of the diffraction peaks observed in Fig. 1(a) should result from the growing up of nc-GaN, which might have been formed in the GaN granular layer. This indicates that the crystallinity of as-deposited nc-GaN might have been greatly improved after the annealing treatment. As a consequence, the density of crystal defects should have been largely reduced.

 

Fig. 1 (a) Part A: the XRD patterns of as-grown and annealed GaN/Si-NPA; Part B: the comparison of the FWHM variation for all the corresponding XRD peaks before and after annealing treatment. (b) The cross-sectional FESEM image of as-grown GaN/Si-NPA.

Download Full Size | PDF

The dark J-V curves of as-grown and annealed LEDs measured at room temperature are depicted in Fig. 2 . The inset of Fig. 2 shows the schematic structure of the LEDs. Both of the J-V curves exhibited rectifying characteristic. Because the contact between Si-NPA and sc-Si has been proved to be ohmic [21], the observed rectification behaviors confirmed the formation of heterojunctions for both as-grown and annealed GaN/Si-NPA. But all the junction parameters for annealed LEDs, including the turn-on voltage, breakdown reverse voltage and leakage current density, have changed largely compared with those for as-grown ones. For example, the turn-on voltage (for obtaining a current density of 1 mA/cm²) increased from ~1.6 V to ~3.9 V, and the leakage current density (at an applied voltage of −4 V) reduced from ~3.2 mA/cm² to ~0.04 mA/cm². The rectiðcation ratios for the two LEDs were calculated to be ~9 (at ± 3.9 V) and ~36 (at ± 4.8 V), respectively. According to the basic theory of heterojunctions [22], the leakage current density of a heterojunction was generally attributed to the defect-mediated tunneling effect caused by a high defect or trap concentration at the interface. Therefore, the distinct reduction of the leakage current density for the annealed LED might indicate an improvement of interfacial quality and a decrease of the defect state density, just as what occurred in the annealing process of ZnO nanorods/Si heterojunctions [23].

 

Fig. 2 The room-temperature J-V curves of as-grown and annealed GaN/Si-NPA. Inset: the schematic diagram of the LEDs.

Download Full Size | PDF

For clarifying the underlying transportation mechanism of the variation, the log-log plot of the J-V data is presented in Fig. 3 . It was found that both the curves for as-grown and annealed LEDs could be fitted by two straight lines. For as-grown LED, the J-V curve exhibits firstly a linear relation at a low forward voltage region (V < 0.9 V, region I). This indicates that the transportation of the carriers obeying the Ohmic law. With the applied voltage increased over 0.9 V (region II), the J-V curve exhibits an exponential relationship (J~V^3.3), which infers a typical space charge limited current (SCLC) mechanism [24]. The SCLC mechanism was usually observed in wide bandgap p-n diodes, such as ZnO/Si [25, 26] and ZnO/SiC [27]. As for annealed LED, the J-V curve also exhibits a linear relation before the inflection point of ~1.9 V (region I′), but the current density is about three orders of magnitude lower than that of as-grown device. With the applied voltage increased beyond ~1.9 V (region II′), the transportation mechanism also transferred to the SCLC model, but with a relationship of J~V¹⁰. Clearly, the exponent varied largely from ~3.3 for as-grown LED to ~10 for annealed LED, and the increment of the exponent in SCLC model indicated a narrowed distribution of the localized states and a lowered defect state density in the annealed LED [28].

 

Fig. 3 The room-temperature log-log plots for the forward J-V of as-grown and annealed GaN/Si-NPA.

Download Full Size | PDF

The EL spectra of the two LEDs are presented in Fig. 4(a) . Under an applied forward bias of 10 V, the current densities for as-grown and annealed LEDs were ~205 mA/cm² and ~185 mA/cm², respectively, and both devices exhibited efficient EL with relatively good monochromaticity, a yellow band peaked at ~567 nm and with a FWHM of ~23.5 nm for as-grown LED, and a NIR band peaked at ~830 nm and with a FWHM of ~18.5 nm for annealed LED. To clarify the origins of the EL for the two LEDs, the PL spectra of as-grown and annealed GaN/Si-NPA were measured, as is given in Fig. 4(b). Under the excitation of an ultraviolet with a wavelength of 320 nm (using a 300 W Xe lamp as the light source and operated with a slit width of 3 nm), both samples showed a strong ultraviolet PL band centered at 366 nm. These bands were due to the near-band-edge (NBE) emission of crystal GaN. A weak but distinct yellow PL band peaked at ~560 nm was observed in as-grown GaN/Si-NPA, but it almost disappeared in the annealed samples. The yellow PL band was usually attributed to the radiative recombination related with the deep-level defects in undoped GaN [29–31], such as gallium vacancies (V_Ga), which has a relatively small formation energy and low migration barrier [29, 30, 32]. Therefore it is rational to deduce that the yellow PL and EL band for as-grown GaN/Si-NPA have the same origin, i. e., both of them originate from a defect-related radiative recombination process. The presentation and disappearance of the yellow PL band before and after annealing treatment reflected a reduction of defect density in GaN.

 

Fig. 4 The room-temperature (a) EL spectra of as-drown and annealed GaN/Si-NPA LEDs operating with a dc voltage of 10 V, and (b) PL spectra of as-grown and annealed GaN/Si-NPA.

Download Full Size | PDF

The mechanism of the EL could be explained through analyzing the energy diagram of GaN/Si-NPA (Fig. 5 ). As has been reported, the electron affinity ψ of GaN and Si-NPA was ~4.1 eV and ~3.6 eV, and their bandgap E_g were ~3.4 eV and ~2.0 eV, respectively [33–35]. The barrier heights at the interfaces for the conduction bands and valence bands, ΔE_C and ΔE_V, were calculated to be ~0.5 eV and ~1.9 eV, respectively. As a result, the interfacial band off-set ΔE = E_g1(Si-NPA) - ΔE_C = E_g2(GaN) - ΔE_V = 1.5 eV, which is equal to the difference of their quasi-Fermi levels after applying bias. In terms of the depletion region of a heterojuction, the distribution of the depletion region is proportional to the built-in field but inversely proportional to the doping concentration. The low defect density for annealed GaN/Si-NPA would surely lead to a low doping concentration. Compared with as-grown LED, the built-in field for annealed LED is wider, which will result in a larger turn-on voltage (Fig. 2 and Fig. 5). When a larger forward bias was applied, the yellow EL band peaked at ~567 nm (~2.2 eV) observed in as-grown LED most probably originated from the radiative recombination of the deep-level defect states, such as V_Ga, in GaN (Fig. 5(a)), just as what occurred in the PL process. The V_Ga, which would form deep acceptor level in GaN, can accept the electrons transited from conduction band and give yellow emission under the excitation of electric field [32]. Furthermore, the low energy barrier (ΔE’ = 2.2 - 1.5 = 0.7 eV) between the valence bands of Si-NPA and deep acceptor levels will be favorable to the injection of holes for realizing the yellow emission. On the other hand, for annealed LED, the NIR EL peaked at ~830 nm (1.5 eV) is attributed to interfacial transition between electrons in the conduction band of GaN and holes in the valence band of Si-NPA for the higher ΔE_V (1.9 eV), as is shown in Fig. 5(b) [18]. As-grown LED exhibits a high leakage current density (high defect density), in which nonradiative recombination would dominate the interfacial recombination. So only the yellow EL originating from the GaN defect states could be observed. After the annealing treatment, both the defect density in GaN and at the interface would be reduced notably, thus the radiative recombination would mainly occur through band-band transition at the interface. In addition, the ΔE (1.5 eV) is much lower than the energy bandgap of Si-NPA (~2.0 eV) and GaN (~3.4 eV), so the transition probability at the interface would be much higher than the NBE transition within either Si-NPA or GaN according to quantum theory [36]. As a result, the annealing process could effectively tune the EL of GaN/Si-NPA from yellow band to NIR band.

 

Fig. 5 The mechanism illustration of the yellow EL from as-grown (a) and NIR EL from annealed GaN/Si-NPA LEDs based on the energy band diagram.

Download Full Size | PDF

Just as discussed above, the origins of the yellow and NIR luminescence from as-grown and annealed GaN/Si-NPA LEDs were attributed to the defect-related radiative transition in GaN and the rediative recombination at the interface of GaN/Si, respectively. Clearly, for the LEDs based on two semiconductors with large lattice mismatch, the promotion on the relatively low IQE is crucial for its practical device application. In addition to further improvement of the material quality of GaN/Si-NPA through optimizing the preparing conditions, such as controlling the microstructure and surface chemical status of Si-NPA, changing the CVD preparing and post-treating parameters, adopting different LED fabrication arts or procedures, some recently developed approaches could also be used for references. The representative demonstrations include the fabrication of nonpolar InGaN quantum well (QW) LEDs [37] or InGaN QW LEDs with large optical matrix elements [38] and surface plasmon coupling [39, 40], through which the radiative recombination rate was greatly improved. This might indicate a promising path for achieving high IQE in the LEDs and might be utilized in preparing GaN/Si-NPA-based LEDs.

4. Conclusions

In conclusion, a GaN/Si nanoheterostructure array was prepared by depositing n-GaN on p-Si-NPA. Two prototype GaN/Si LEDs with a device structure of ITO/n-GaN/p-Si-NPA/sc-Si/Al were prepared based on as-grown and annealed GaN/Si-NPA, from which sharp yellow and NIR EL bands were obtained correspondingly. The origins for the yellow and NIR EL were attributed to the defect-related radiative recombination in GaN and the interfacial band-band rediative recombination of GaN/Si-NPA, respectively. Our research might have provided a doable approach for fabricating visible and NIR LEDs directly based on GaN/Si heterostructures.