Indium segregation measured in InGaN quantum well layer.

The indium segregation in InGaN well layer is confirmed by a nondestructive combined method of experiment and numerical simulation, which is beyond the traditional method. The pre-deposited indium atoms before InGaN well layer growth are first carried out to prevent indium atoms exchange between the subsurface layer and the surface layer, which results from the indium segregation. The uniform spatial distribution of indium content is achieved in each InGaN well layer, as long as indium pre-deposition is sufficient. According to the consistency of the experiment and numerical simulation, the indium content increases from 16% along the growth direction and saturates at 19% in the upper interface, which cannot be determined precisely by the traditional method.

The segregation phenomenon exists generally in kinds of compound semiconductor systems12345678. Due to the coherency strain and different miscibility between different materials, the atoms segregate into the upper layer, which will lead to the component nonuniform distribution of binary or ternary alloy56789. This phenomenon changes the energy levels, which further influences the optical and electrical performances of semiconductor devices. Therefore, many techniques are performed to investigate it, such as secondary ion mass spectroscopy (SIMS)4, transmission electron microscope (TEM)510, X-ray photo-emission spectroscopy (XPS)11, reflection high-energy electron diffraction (RHEED)12, photoluminescence (PL)13141516 and so on. Consequently, several segregation models are performed to analyze the atom content distribution1718192021.

In recent years, light emitting diodes (LEDs) and laser diodes (LDs) with InGaN/GaN multi quantum wells (MQWs) active region have been developed rapidly and used widely in varies of applications22232425262728. In the LED or LD structure, the optoelectronic properties are very sensitive to the indium content distribution in the InGaN well layer, and the physical mechanism is still in controversy. In the meantime, the indium segregation phenomenon in InGaN well layer has been studied for several years293031323334. Due to the indium segregation, the nonuniform distribution of indium content will seriously change the potential profile of InGaN/GaN MQWs. These results reduce the overlap between the electron and hole wave functions in the InGaN well layer and hence decrease the internal quantum efficiency29. Therefore, it is necessary to realize the indium segregation extent exactly. However, as the researchers want to precisely measure the indium content profile in the InGaN well layer, the study on indium segregation runs into difficulties due to the slight change of indium content and lack of effective measurement method.

In this work, a novel design is performed to determine the indium segregation quantitatively in the InGaN well layer. This method can be nondestructive to ascertain the indium distribution profile in InGaN well layer. We first obtain the uniform spatial distribution of indium content by using indium pre-deposition (IP) before InGaN well layer growth. And then the indium content value at the upper interface of InGaN well layer is ascertained by the numerical simulation. After that, the indium content distribution in InGaN well layer is observed by electroluminescence (EL) experimental and theoretical simulation results.

Results

In our design, the InGaN well thickness of conventional LED A is 2.5 nm and the growth time is 2 min. LEDs B-D are grown under the same conditions. Nevertheless, IP before each InGaN well layer growth is introduced and the time of pre-deposition is 1.5, 2, 2.5 min respectively. A growth interruption time of 30 seconds is also performed after the growth of GaN barrier layer. The structure diagram of all the LEDs is shown in Fig. 1(a). Figure 1(b) shows (0002) plane diffraction obtained with ω/2θ scans along the growth direction for the MQWs. More details of high resolution X-ray diffraction (HRXRD) data analysis can be found in Ref. 35. Compared to LED A, the Indium content increases slightly in the InGaN well layer for LEDs B–D due to IP effect, while the thicknesses of GaN barrier and InGaN well layers for all the LEDs are almost the same. Moreover, The FWHM values of the InGaN “+1st” diffraction peak for the LEDs A–D are 267, 223, 201, 204 arcsec, respectively, and the values for LEDs B–D are much smaller than that of LED A, which suggests that the uniformity of indium content distribution is improved due to the indium pre-deposition. Besides, It is also noted that the high series satellite peaks of LEDs B–D is more distinct than that of LED A, which shows that the structure properties are significantly improved with IP process. These results indicate that no InN or island is formed. It is because the indium pre-deposition is performed at a temperature of 730°C, higher than the InN growth temperature which is usually under 550°C.

Figure 1

(a) The structure diagram of all the LEDs and (b) (0002) plane ω/2θ HRXRD spectra of InGaN/GaN MQWs with different IP time.

The EL spectra are measured for all the LEDs under 1 mA injection current. Figure 2 reveals that the major emission peak wavelengths for LEDs A–D are 458.8, 467.4, 474.2, 474.1 nm, respectively. According to the variation of peak wavelength, it is observed that the peak wavelength red-shift (~15.4 nm) with the increase of the IP time for the LEDs A–D. This EL red-shift is consistent with our former HRXRD analysis of IP effect. During the IP process, most of the deposited indium is taken away by the gas flow in the reactor, while part of indium remains on the wafer and generates an “indium floating layer”1920. The “indium floating layer” could prevent atoms exchange between the surface layer and subsurface layer, which reduces the effect of indium segregation. As longer pre-deposition time is performed, more indium atoms adsorb on the surface of GaN layer and these atoms are always exchange into the upper layer due to the lower energy of the surface in the following InGaN growth. As a result, more indium atoms incorporate into the InGaN films leading to the red-shift of emission peaks in EL measurement. We also note that though the IP time of LED C and D are 2 min and 2.5 min, respectively, their EL peak wavelengths are nearly the same. The unchanged wavelength with increasing IP time provides evidence for the saturation of indium content in the InGaN well layers. By 2-min IP time, the number of indium atoms is enough to totally occupy the low-energy states on the GaN surface layer. Consequently, the indium atoms exchange between the subsurface and surface layers keeps dynamic balance and the segregation is completely suppressed in the following InGaN layer growth. As a result, the spatial distribution of indium content is uniform in the each InGaN well layer, as long as IP is sufficient. It is another difference with InGaAs system, the excess indium pre-deposition will lead to a higher indium content distribution at the lower part of InGaAs layers.

Figure 2

Room-temperature EL spectra (1 mA) for LEDs A–D.

The major emission peak wavelengths for LEDs A–D are 458.8, 467.4, 474.2, 474.1 nm, respectively.The two green lines show a redshift of ~15.4 nm, as the IP time increases.

Now we have shown that the indium content reaches saturation in the condition of 2 min IP time. The further question is how does indium content distribute along growth direction in InGaN well due to the segregation for LED A.

It has been demonstrated that the indium content varies exponentially along the growth direction in InGaAs materials. According to exponential model, the indium content in the th monolayer in the InGaAs materials is given in the form;

In our case, we hereby build the model about the indium segregation of the conventional LED structure in the following form: where y is the indium content, x is the distance from the lower interface, and a is a constant. In addition, c is a coefficient about the indium segregation length. Because the indium content of the upper interface of InGaN well makes not much difference to the lower interface and the thickness of InGaN well is very small, the above equation can be simplified as follows: where y0 is the indium content at lower interface of InGaN well layer and y1 is the slope. So the exponential model has been replaced by the linear one. The schematic indium content profiles of the MQWs for LED A and LED C are shown in Fig. 3(a) and Fig. 3(b), respectively. The indium content in the conventional structure increases along with the growth direction of InGaN well layer due to the indium segregation, while the indium content is constant in the structure with IP process.

Figure 3

The indium content distribution charts of the MQWs in the active region for LED A and LED C.

a): the indium content increases along the growth direction for LED A; b) the indium content is consistent for LED C. In the meanwhile, the indium composition at the upper layer of sample A is the same with indium content of InGaN well layer of LED C.

During the InGaN growth in MOCVD system, the indium incorporation efficiency is poor and most of the indium atoms are desorbed at growth temperature (730°C). For example, in our experiment, the ratio of Ga/In is usually about 1:1, but the indium content in the InGaN well layer is much smaller than 0.5. The excess indium has the similar effect of IP, filling the low-energy states on the surface. In our design, the growth time of indium well layer is also 2 min. As mentioned above, 2-min IP time provides enough indium to balance the atom exchange, which means that there should be enough indium atoms to make the upper interface of InGaN well layer saturated, even without IP (LED A). As indicated by the red dotted line in Fig. 3, we can conclude that the indium content difference of the upper interface of InGaN well layer for LED A and LED C is slight, since the indium contents of upper surfaces in both samples are saturated.

LED devices with conventional structure (LED A) and IP structure (LED C) are fabricated and characterized by EL spectra measurement. Besides, theoretical simulations of two LED structures are also carried out by Crosslight software. For LED C, the indium is uniformly distributed, so the square quantum well structure is adopted in the simulation where the indium content is the only unknown. The indium content can be obtained by fitting the simulated EL spectral to the experimental one, which is 19% (Fig. 4(a)). Based on above analysis, the indium content at upper interface of InGaN/GaN MQWs is also 19% for LED A. Another simulation for LED A is carried out. Triangle-shaped quantum well structure is used because the indium content is assumed to increase with InGaN layer thickness. Since the indium content at upper interface is known as 19%, the indium content at lower interface can be fitted similarly by comparing the simulated and experimental results, as shown in Fig. 4(b). When the indium content at the lower interface in InGaN well layer is 16%, the calculated spectrum agrees with that of the experiment. As a result, the influence of indium segregation on the content distribution can be obtained. For the conventional InGaN blue LED structures in our experiment, the indium content increases from 16% to 19% along the growth direction because of the indium segregation effect.

Figure 4

The calculated and experimental EL spectra (1 mA) of the LED A and LED C. The calculated EL spectra with different indium distribution are shown as color line and the experimental EL spectra is shown as black line.

Based on the data consistency of the experiment and numerical simulation, the indium content is ~19%, while the indium content increases from ~16% along the growth direction and saturates at ~19% in the upper interface of InGaN well layer.

Discussion

In the light of the above, we have obtained the indium content profile in the InGaN well layer. It is noteworthy to mention here that due to the typical InGaN well thickness (~2.5 nm) and the slight indium content change in the InGaN well layer, the indium segregation cannot be accurately measured by the conventional method, such as TEM, SIMS, and so on363738. Due to indium segregation, the nonuniform components distribution will lead to the local polarization effect and then influence the properties of optoelectronic devices. The new method is nondestructive to present a better understanding of indium segregation, which should be taken into consideration in the epitaxial growth process.

In conclusion, we have first developed a new method to determine the indium segregation in InGaN well layer. The effect of IP prior to the InGaN well growth on the indium content profile is demonstrated. As IP is sufficient, the uniform spatial indium content profile is realized in each InGaN well layer. Based on EL experiment and theoretical simulation results, the indium content increases from 16% along the growth direction and saturates at 19% in the upper interface of InGaN well layer. This method established in our work can describe the segregation behavior, which will enhance the better understanding of state densities, energy band and polarization effect in optoelectronic devices or microelectronic devices.

Methods

All growths are performed on 2-in (0001) sapphire substrates by low-pressure metal organic chemical vapor deposition (MOCVD). The precursors are trimethylgallium (TMGa), triethylgallium (TEGa), trimethylindium (TMIn), and ammonia (NH3), respectively. Prior to the growth of GaN nucleation layers (NLs), the substrates are exposed to hydrogen (H2) at 1100°C to desorb surface contaminants for 12 min. The temperature is lowered to 550°C and 25 nm thick GaN NLs are deposited using TMGa and NH3. These NLs are subsequently annealed in NH3 and H2, following the growth of 2.5 μm thick planar GaN layers at 1100°C. Then 3 μm-thick Si doped n type GaN layers are deposited at 1100°C, followed by the five periods of InGaN/GaN MQWs. The growth temperature of GaN barrier and InGaN well layer are 830°C and 730°C, respectively. Finally the 200 nm thick Mg-doped p-type GaN layer is deposited and the samples are annealed for 20 min to activate Mg ion in the nitrogen atmosphere at 700°C. In the whole deposition process, H2 is used as an ambient gas for GaN layer, while N2 is used as an ambient gas for the growth of InGaN/GaN MQWs to increase the indium incorporation efficiency. The rate of the indium flow is 93 μmol/min. Compared to LED A, LEDs B–D are distinguished by adopting IP time. For LEDs B–D, prior to the growth of each InGaN QW layer, indium atoms are deposited and the deposition time is 1.5, 2, and 2.5 min, respectively. A growth interruption time of 30 seconds is also performed after the growth of GaN barrier layer.

Then the epitaxial chips are fabricated from these structure using standard mesa etching and contact fabrication techniques. The fabrication of the LED chips with a device area of 300*300 μm2 is described in detail elsewhere. The LED chips are then mounted on a Cu cold stage. Then the EL spectra of all samples under 1 mA current injection levels are obtained in pulsed mode with a duty cycle of 10% and a pulse length of 20 ms. At the same time, the simulation is carried out to calculate the EL spectra according to the new model by the Crosslight software.

Author Contributions

H.C. and Y.J. conceived and directed the research. Z.D. carried out all the wafers growth experiments, chips fabrication and characterization test. In the meanwhile, Z.D. and L.C. carried out the simulation by the Crosslight software. In the end, Z.D. wrote the main manuscript text and all the authors reviewed the manuscript.