JP2017139247A - Epitaxial wafer, semiconductor light-emitting element, light-emitting device, and method of producing epitaxial wafer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an epitaxial wafer capable of substantially increasing emission output.SOLUTION: An epitaxial wafer of the present invention includes: a GaN substrate having, as a main surface, a surface with an off-angle of 0° or more and 10° or less against an m-plane; an n-type conductive layer formed on a main surface on one side of the GaN substrate; and a light emitting layer formed on a main surface on one side of the n-type conductive layer. The PL peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and the PL half-value width of the light emitting layer satisfies a conditional expression (1). Δl≤L×0.4-150 (1) L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)SELECTED DRAWING: Figure 17

Description

  The present invention relates to an epitaxial wafer, a semiconductor light emitting element, a light emitting device, and an epitaxial wafer manufacturing method.

  Gallium nitride (GaN) -based materials are currently widely used to produce light-emitting elements such as semiconductor light-emitting elements (LEDs) and semiconductor lasers (LDs). In a conventional blue light emitting device, an n-type layer and a p-type layer are formed mainly in a c (+) direction ([0001] axis direction) using a sapphire substrate, and an InGaN-based quantum well is formed between the n-type layer and the p-type layer. What produced the light emitting layer has been used. The light-emitting element thus manufactured is applied to, for example, a white light source, and it is required to solve the following problems in order to realize a more efficient light source.

  In the light emitting device using the sapphire substrate, the generation of threading dislocations is unavoidable due to the mismatch of the lattice constant between the substrate and the GaN-based material, and the other is the property of the material along the c-axis direction. Since an internal electric field is generated, there is a problem that the overlapping of wave functions of electrons and holes contributing to light emission is small, and the light emission efficiency is lowered. For this reason, the light emitting element using a sapphire substrate has the situation that the thickness of a quantum well layer must be produced as thin as about 2 nm (nonpatent literature 1).

  The effect of the internal electric field is larger as the In composition is higher, and therefore the luminous efficiency is lowered particularly in the long wave region above blue. Furthermore, the droop phenomenon is also a serious problem and causes a decrease in efficiency in the high implantation region. Although the cause of the droop phenomenon is actively discussed, “Auger recombination due to increased carrier density”, “spillover phenomenon in which carriers pass through without being trapped in quantum well”, “overflow of carriers outside MQW region” "Is considered as the main cause (Non-Patent Document 2).

  In order to solve the above problems, a GaN substrate having no lattice constant difference is used as a substrate, a plane orientation in which no piezoelectric field is generated, and as a result, a thick quantum well layer can be formed. It has been considered effective to use the plane (10-10) plane.

  For example, according to Patent Document 1, an InGaN quantum well layer having a thickness exceeding 10 nm on the m-plane can be produced, and an LED having a low droop characteristic is actually reported.

  Further, Patent Document 2 has a study result on increasing the light emission efficiency of the InGaN quantum well layer on the m-plane GaN substrate having a purple wavelength (407 nm). In this case, it is the complex of oxygen and group III (gallium) present on the slip surface of the crystal that causes the decrease in the light emission efficiency in the m-plane. It is said that it is effective to make the vicinity of the surface a low In composition region. The derivation of growth conditions for this purpose is complicated. An important point is that it is desirable that the growth condition of the InGaN quantum well layer is an ultra-high V / III ratio (10000 to 30000) (see paragraph 0101 of Patent Document 2).

  On the other hand, in Patent Document 1, it is known that the light emission characteristics at a short wavelength (purple) of 400 nm can be obtained as a good light emission output of 120% or more on the m-plane as compared with the c-plane, and the droop is very small. That is, the potential of high luminous efficiency on the m-plane is high, and there is an expectation that an LED having high efficiency and low droop characteristics that could not be realized in the past can be obtained.

  From the above, attempts have been widely made to produce high-efficiency light emitting LEDs using m-plane GaN substrates in the long wave region from purple to blue wavelengths and beyond.

  However, on an m-plane GaN substrate that can eliminate the influence of the piezoelectric field and reduce threading dislocations by a self-supporting substrate, the light emission efficiency in the long wave region should improve, but contrary to expectations, the output increases as the wavelength increases. A problem of abrupt reduction has been reported (Non-Patent Document 3). In Non-Patent Document 3, the light emission output becomes maximum when the light emission wavelength is 400 nm, and the light emission output sharply decreases to about 420 nm. At 440 nm, only light output less than half of that at 400 nm is obtained. This phenomenon is the same even if the quantum well layer thickness is changed.

JP 2010-123803 A International Publication No. 2013/042297

A. Chakraborty et al., Japanese Journal of Applied Physics Vol. 44, No. 5, 2005, pp. L 173-L 175. J. Piprek, Phys.Status Solidi A 207, No. 10, 2010, pp.2217-2225. H. Yamada et al., Appl. Phys. Express 1, 2008, 041101.

  As described in Non-Patent Document 2, in an LED on an m-plane GaN having an InGaN quantum well layer, the emission output is maximized at an emission wavelength of 400 nm, and then rapidly increases as the In composition in the InGaN quantum well layer increases. There was a problem that the light emission output decreased.

  The present inventors have also developed a growth temperature, a growth pressure, a growth rate, and the like when an LED using an m-plane GaN free-standing substrate is formed, such as a substrate specification such as an off-angle, and each layer structure including an InGaN multiple quantum well layer. We investigated the structural dependency such as the condition dependency, the thickness of the InGaN quantum well layer, the thickness of the barrier layer when the multiple quantum well layer was formed, and the composition of the barrier layer. However, as in Non-Patent Document 2, the light emission output becomes maximum at a light emission wavelength of 400 nm, and then the light emission output rapidly decreases. As a result, in the blue region (near the wavelength of 450 nm), the light emission is lower than that of the polar substrate. Only output was obtained.

  As for the off-angle of the m-plane GaN substrate, a substrate having an off-angle of about ± 5 ° in each axial direction was investigated. As a result, the same wavelength dependence of light emission efficiency was exhibited from the 0 ° substrate to the off-angle substrate gradually inclined by about −5 ° in the c-axis (−) direction. In addition, in m-plane GaN substrates having other off-angles (for example, a substrate inclined in the a-axis direction, a substrate inclined in the c-axis (+) direction, or a substrate inclined in both), a violet wavelength is included, Only lower emission characteristics were obtained.

  The present invention has been made in view of such circumstances, and a main object of the present invention is to provide an epitaxial wafer, a semiconductor light emitting element, a light emitting device, and a method for manufacturing the epitaxial wafer that can greatly improve the light emission output.

  Hereinafter, when the off angle is referred to in this specification, the inclination angle in the c (+) direction is indicated. Further, as the accuracy of the off-angle numerical value, since the influence on the crystal quality is substantially small after the decimal point and there is a variation in the plane, the effective number is set to one digit. For example, when it has an off angle of −5.2 ° in the c (+) direction and 0.1 ° in the a-axis direction, it is described as a −5 ° off angle substrate. Similarly, when the off angle is small, for example, when the off angle is + 0.24 ° in the c direction and + 0.13 ° in the a-axis direction, the substrate is described as a 0 ° substrate.

In order to achieve the above object, an epitaxial wafer of the present invention includes a GaN substrate having a main surface having an off angle of 0 ° or more and 10 ° or less with respect to an m-plane, and a main surface on one side of the GaN substrate. An n-type conductive layer formed thereon, and a light-emitting layer formed on a main surface on one side of the n-type conductive layer, and a PL peak wavelength of the light-emitting layer is 410 nm or more and 460 nm or less, The PL half width of the light emitting layer satisfies the conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)

  The epitaxial wafer of the present invention includes a GaN substrate whose main surface is a surface having an off angle of 0 ° or more and 10 ° or less with respect to the m-plane, and an n formed on the main surface on one side of the GaN substrate. A light emitting layer formed on a main surface on one side of the n-type conductive layer, and the PL peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and the room temperature of the light emitting layer is 25. The PL lifetime of the light emitting layer at ° C. is 1.3 nsec or more and 20 nsec or less.

  Furthermore, the epitaxial wafer of the present invention includes a GaN substrate whose main surface is a surface having an off angle of 0 ° or more and 10 ° or less with respect to the m-plane, and an n formed on the main surface on one side of the GaN substrate. A light emitting layer formed on a main surface on one side of the n-type conductive layer, and a PL peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and excitation light of the light emitting layer The fluctuation of the PL peak wavelength when the intensity is changed 1000 times is from 0 nm to 10 nm.

  Further, in the epitaxial wafer of the present invention, the light emitting layer has a quantum well structure, and an interfacial strain buffer layer is formed on at least one side between the quantum well layer and the adjacent barrier layer.

In order to achieve the above object, the semiconductor light-emitting device of the present invention is formed on a main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. An n-type conductive layer made of the formed GaN-based semiconductor layer, a light emitting layer formed on one main surface of the n-type conductive layer, and a p formed on one main surface of the light emitting layer A light emitting peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and a PL half width of the light emitting layer satisfies the conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)

  The semiconductor light-emitting device of the present invention includes a GaN-based semiconductor layer formed on a main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. An n-type conductive layer, a light-emitting layer formed on one main surface of the n-type conductive layer, and a p-type conductive layer formed on one main surface of the light-emitting layer, The emission peak wavelength of the light emitting layer is 410 nm to 460 nm, and the PL life of the light emitting layer at room temperature of 25 ° C. is 1.3 nsec to 20 nsec.

  Furthermore, the semiconductor light emitting device of the present invention includes a GaN-based semiconductor layer formed on a main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. An n-type conductive layer, a light-emitting layer formed on one main surface of the n-type conductive layer, and a p-type conductive layer formed on one main surface of the light-emitting layer, The emission peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and the fluctuation of the emission peak wavelength of 1 mA or more and 350 mA or less of the light emitting layer is 6 nm or less.

Furthermore, in order to achieve the above object, the light emitting device of the present invention is formed on the main surface on one side of the GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. An n-type conductive layer made of a GaN-based semiconductor layer, a light emitting layer formed on one main surface of the n-type conductive layer, and a p-type formed on one main surface of the light emitting layer A PL peak wavelength of the light emitting layer is not less than 410 nm and not more than 460 nm, and a PL half value width of the light emitting layer is a semiconductor light emitting element that satisfies the conditional expression (1), and the semiconductor light emitting element is A wavelength converting substance that absorbs at least part of the emitted light and converts it into light having a longer wavelength.
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)

  The light-emitting device of the present invention comprises a GaN-based semiconductor layer formed on the main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. an n-type conductive layer, a light-emitting layer formed on one main surface of the n-type conductive layer, and a p-type conductive layer formed on one main surface of the light-emitting layer, The emission peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and the PL life of the light emitting layer at room temperature of 25 ° C. of the light emitting layer is 1.3 nsec or more and 20 nsec or less, and the semiconductor light emitting element A wavelength converting substance that absorbs at least part of the emitted light and converts it into light having a longer wavelength.

  Furthermore, the light-emitting device of the present invention comprises a GaN-based semiconductor layer formed on the main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. an n-type conductive layer, a light-emitting layer formed on one main surface of the n-type conductive layer, and a p-type conductive layer formed on one main surface of the light-emitting layer, The emission peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less, and the fluctuation of the emission peak wavelength of 1 mA or more and 350 mA or less of the light emitting layer is 6 nm or less, and at least light emitted from the semiconductor light emitting element A wavelength converting substance that absorbs part of the light and converts it into light having a longer wavelength.

  Furthermore, in order to achieve the above-mentioned object, the epitaxial wafer manufacturing method of the present invention includes a main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. A first step of growing an n-type conductive layer made of a GaN-based semiconductor layer thereon; a second step of growing a light emitting layer on a main surface on one side of the n-type conductive layer grown in the first step; In the second step, the group V source and the group III are set so that the V / III ratio, which is the ratio of the mole supply amount of the group V raw material to the group feed amount of the group III raw material, is 500 or more and 4000 or less. Supply raw materials.

  ADVANTAGE OF THE INVENTION According to this invention, the manufacturing method of the epitaxial wafer, semiconductor light-emitting device, light-emitting device, and epitaxial wafer which can improve the light emission output significantly is realizable.

FIG. 1 is a graph showing the results of evaluating the wavelength dependence of the light output of a semiconductor light emitting device fabricated under conventional conditions. FIG. 2 is a graph showing the PL mapping result in the substrate surface of the MQW structure manufactured under the conventional conditions. FIG. 3 is a graph showing the result of overwriting the PL spectrum at the position shown in FIG. 2 separately for the case where all the polarization components are received and the case where the polarization is separated. FIG. 4 is a graph showing a result of fitting the PL spectrum separated by polarization in FIG. 3 with a two-component Gaussian peak. FIG. 5 is a graph showing the results of plotting the PL peak wavelengths obtained in FIG. 4 with the position of X in FIG. 3 as the horizontal axis. FIG. 6 is a graph showing the XRD diffraction results at each position in FIG. 2 and the position dependency of the angle of the 0th-order satellite peak. FIG. 7 is a photograph showing a cross-sectional TEM image of an MQW structure fabricated under conventional conditions. FIG. 8 is a graph showing the result of overwriting the EL emission spectrum when the current value of the injection current to the m-plane semiconductor light emitting device fabricated under the conventional conditions is changed. FIG. 9 is a graph showing the result of fitting the EL emission spectrum of FIG. 8 with a two-component Gaussian peak. FIG. 10 is a graph showing the results of PL emission spectra of Sample B and Sample C used in the positron annihilation experiment. FIG. 11 is a graph showing SW plots of Sample B and Sample C obtained in the positron annihilation experiment. FIG. 12 is a cross-sectional view of the epitaxial wafer according to the present embodiment. FIG. 13 is a cross-sectional view of the light emitting layer according to this embodiment. FIG. 14 is a schematic view of the semiconductor light emitting device according to this embodiment, FIG. 14 (a) is a top view, and FIG. 14 (b) is a cross-sectional view taken along the line XX in FIG. 14 (a). It is. FIG. 15 is a schematic diagram of a light emitting device according to this embodiment. FIG. 16 is a flowchart of an epitaxial wafer manufacturing method according to this embodiment. FIG. 17 is a graph showing the wavelength dependence of the total radiant flux when 350 mA is applied to the semiconductor light emitting devices of Example 1, Example 2, and Comparative Example 1. FIG. 18 is a graph showing the PL peak wavelength dependence and PL lifetime of the PL half-value width obtained from the semiconductor light emitting devices of Example 1, Example 2, and Comparative Example 1. FIG. 19 is a graph showing the EL emission spectra, the current value dependency of the EL emission peak wavelength, and the current value dependency of the total radiant flux in the semiconductor light emitting devices of Example 1 and Comparative Example 1. FIG. 20 is a graph showing the relationship between the PL peak wavelength and the PL half width measured by changing the growth conditions of the LED structure and the MQW structure. FIG. 21 is a graph showing the relationship between the PL peak wavelength and the PL lifetime measured by changing the growth conditions of the LED structure and the MQW structure. FIG. 22 shows the dependency of the PL peak wavelength on the excitation light intensity (relative value). FIG. 23 is a graph showing the relationship between the excitation light intensity at room temperature and the PL lifetime when the excitation light intensity of the MQW structure of Example 6 is changed. FIG. 24 is CL spectrum half-value width mapping data obtained from the MQW structure of Example 7-1. FIG. 25 is CL spectral wavelength mapping data obtained from the MQW structure of Example 7-1. FIG. 26 is mapping data of the CL spectrum half width obtained from the MQW structure of Comparative Example 7-1. FIG. 27 is CL spectral wavelength mapping data obtained from the MQW structure of Comparative Example 7-1. FIG. 28 is a CL emission spectrum in the portions shown as 1, 2, and 3 in FIG. FIG. 28 is a CL emission spectrum in the portions shown as 1, 2, and 3 in FIG. FIG. 30 shows the surface morphology by AFM on the InGaN single layer of Example 8-1, where the left side is a surface shape image and the right side is a phase image. FIG. 31 shows the surface morphology by AFM on the InGaN single layer of Comparative Example 8-1, where the left side is a surface shape image and the right side is a phase image. FIG. 32 is a graph showing the temperature dependence of the PL lifetime of the MQW structure of Example 9.

(Background)
As a main motivation for the present inventors to reach the present disclosure, the following points will be described in which the emission spectrum from the InGaN quantum well layer on the non-polar m-plane GaN substrate has a plurality of peaks. . The target InGaN quantum well layer can emit light with a wavelength of 400 nm to 460 nm, and the In composition is generally in the range of 5% to 20%.

  First, the present inventors fabricated a semiconductor light emitting device using an InGaN quantum well layer on an m-plane GaN substrate under conventional conditions, that is, normally used high V / III ratio conditions, and evaluated the wavelength dependence of optical output. The results are shown in FIG. Here, a high V / III ratio condition of 8340 to 15480 is used as the V / III ratio during the growth of the InGaN quantum well layer. The layer structure and growth conditions are as shown in Tables 1 and 2. The InGaN quantum well layer and the GaN barrier layer were laminated by repeating three periods. The substrate, chip structure, mounting form, and evaluation conditions were the same as in Example 1 described later.

  There are various methods for controlling the emission wavelength from the InGaN quantum well layer. Here, the In mole supply ratio in the group III during the growth of the InGaN quantum well layer ((mol supply per minute of TMI) Amount) / (mol supply amount of TMI per minute + mol supply amount of TMG per minute)) and controlling the growth temperature of the light emitting layer.

  The examples and comparative examples shown in this specification are a part of the results of the entire experiment conducted by the present inventors. However, under either condition, the growth temperature is decreased and the In molar supply ratio is increased. The emission wavelength could be lengthened even using the above conditions. That is, there are a plurality of combinations of the growth temperature and the In mole supply ratio for realizing a certain wavelength. However, when the In mole supply ratio is in the range of 50% or more, the light emission efficiency and the light emitting layer quality described below are governed by the wavelength. That is, no matter what combination of growth temperature and In mole supply ratio is used, it is confirmed that the characteristics and quality are the same as long as the other conditions and wavelengths are the same.

  Here, as can be seen from FIG. 1, the light emission output decreases rapidly as the wave length increases. This tendency was exactly the same as that of Non-Patent Document 3 described above.

  As described in the problem, the improvement of the LED light emission output on the m-plane has not been observed so far by changing the growth conditions or changing the substrate type such as the off angle. Therefore, the present inventors consider that the quality of the light-emitting layer itself decreases as the In composition increases, and in order to observe the quality of the light-emitting layer directly and accurately, the growth is performed before the p-type conductive layer. Based on the stopped MQW structure, the cause of quality deterioration and improvement methods were examined.

  Although the off-angle of the m-plane GaN substrate was examined from 0 ° to around 5 °, the quality deterioration situation was almost the same. Here, an LED on an m-plane GaN substrate having an off angle of 0 ° will be described. The layer structure examined here is shown in Table 3. The InGaN quantum well layer and the GaN barrier layer were laminated by repeating three periods. The V / III ratio in the InGaN quantum well layer was as high as 8340.

  After the growth was completed, PL mapping measurement was performed on the entire substrate. In the PL mapping measurement, a 325 nm He—Cd laser was used as an excitation light source, and the entire surface of the substrate was evaluated at a pitch of 0.5 mm. FIG. 2 shows a mapping result of the PL peak wavelength, the PL peak intensity, and the PL half width.

  As can be seen from FIG. 2, the PL peak wavelength becomes longer from the lower left to the upper right of the drawing, and at the same time, the PL peak intensity also decreases. The PL half-value width once spreads as the wave length increases, and is the widest in the vicinity of 440 nm and further narrowed in the vicinity of the upper right end of the drawing.

In addition, the PL emission spectrum obtained here has two intensity peaks, or even if there is only one intensity peak, the spectrum shape is not bilaterally symmetric and bulges on one side (sometimes called a shoulder). There were cases of distorted shapes that could be seen.
These abnormal spectral shapes are due to the fact that there are at least two intensity peaks, and this is due to the superposition of two or more types of light emission. Here, this phenomenon is referred to as multi-peak of light emission.

  In order to investigate the details of the mapping result, the PL emission spectrum was evaluated again using another optical system at the point indicated by “X” on the substrate. As the excitation light here, a picosecond pulse laser with a wavelength of 385 nm was used, and the spot diameter was 100 μmφ. The detailed conditions of the excitation light are the same as those used in the PL lifetime measurement. The measurement points are displayed in the order of X1, X2,..., X10 on the short wave side (from the lower side in FIG. 2) on the substrate.

  The evaluation of the PL emission spectrum described above is based on the component (E // a) in which the electric field vector is polarized in the a-axis direction by inserting a polarizing filter on the light receiving side of the PL emission spectrum when all the emission is received (no polarization). ), And the spectrum was measured under three conditions: only the component (E // c) whose electric field vector is polarized in the c-axis direction.

  The purpose of the evaluation by polarization separation is to examine whether the multi-peak of the PL emission spectrum is due to the band structure. The valence band of the InGaN quantum well layer on the m-plane GaN substrate is split by the influence of in-plane anisotropic strain, and the light emission derived from the A band of the lowest energy level, which is the ground level (long wave side) E // a-polarized light, and light (B-band) derived from a higher level is known to be E // c-polarized light. Therefore, if the multi-peak is due to the band structure, it is a superposition of two single peaks having different polarizations, and can be separated into single peaks by polarization separation.

  FIG. 3 shows a PL emission spectrum measured under these three conditions for each polarization condition, and the PL emission spectrum at each position is overwritten.

  As can be seen from FIG. 3, both the E // a component and the E // c component are not single peaks but multi-peaks. That is, the cause of these multi-peaks is not due to the band structure of the valence band. Note that the scale of the vertical axis is appropriately enlarged and reduced for easy viewing, and is different for each graph.

  Next, in order to examine what component these multi-peaks are composed of, the superposition of the two-component Gaussian peaks on the PL spectra of the E // a component and E // c component of sample A, respectively. Was used to fit each PL peak intensity. The results are shown in FIG. The vertical axis is PL peak intensity, and the horizontal axis is energy display (photon energy (eV)).

As can be seen from FIG. 4, when the PL emission spectrum is composed of two components, the shapes of the peaks of the E // a component and the E // c component can be accurately reproduced. That is, these PL emission spectra are a superposition of the light emission mode (E _high ) on the short wave side and the light emission mode (E _low ) on the long wave side.

  Furthermore, what plotted the peak of each PL emission spectrum which carried out peak separation in FIG. 4 for every position is shown in FIG.

  As can be seen from FIG. 5, both the E // a component and the E // c component show the peaks of the original PL emission spectrum (open squares □ and Δ) on the short wavelength side of the PL emission spectrum. And the peak is close (meaning that the peak on the short wave side is the main), and the spectrum and the peak on the long wave side are closer as the wave length increases (the light emission on the long wave side is the main).

  The point at which the peak position moves from the short wave side to the long wave side differs between the E // a component and the E // c component, but both are around 430 nm.

  Furthermore, the peak wavelength of the original PL emission spectrum shows a jump in the PL peak wavelength from X6 to X7. In order to investigate whether the In composition changes correspondingly, (300 ) Reflective XRD diffraction evaluation was performed.

  Here, the positions of X2, X4, X6, X7, X8, X9 and X10 were measured. The result is shown in FIG. Further, FIG. 6 shows the position dependency of the angle of the zeroth-order peak. As can be seen from FIG. 6, the change of the satellite peak reflecting the In composition from X6 to X7 and X8 is gradual. That is, the jump in the PL peak wavelength is not due to a change in the In composition.

  Furthermore, in order to investigate the occurrence of such wavelength fluctuations and multi-peaks due to crystal defects such as slip planes, stacking faults, and threading dislocations as seen in Patent Document 2, cross-sectional HAADF-TEM evaluation was performed. Carried out. The results are shown in FIG. As can be seen from FIG. 7, no defects were observed in the cross section of the InGaN quantum well layer, the interface of the quantum well structure was flat, the In fluctuation was not noticeable, and no major problem was found.

  Furthermore, the present inventors investigated what kind of emission spectrum can be obtained with the emission characteristics when current is injected into an actual light emitting element. An emission spectrum by current injection is called an EL emission spectrum. Here, a substrate on which a p-type conductive layer was formed in the same structure as that used in PL peak wavelength fluctuation (described later) was fabricated, and a semiconductor light emitting device was fabricated and evaluated. The results are shown in FIGS.

  As shown in FIG. 8, the shape of the EL emission spectrum changes according to the injection current. Moreover, the shape can be fitted well by superimposing two Gaussian peaks, like the PL emission spectrum. As shown in FIG. 9, long wave side light emission is main at low carrier density (low current value), but short wave side light emission becomes main as the carrier density increases. That is, when the carrier density is high, the relative light emission efficiency of long wave side light emission is low.

  In an attempt to investigate the cause of the quality deterioration of the InGaN quantum well layer, the present inventors further evaluated the state and type of point defects by positron annihilation experiments using samples having different emission wavelengths. Evaluation of point defects by positron annihilation experiments and estimation of defect species using SW plots are described in, for example, Uedono et al., Journal of Crystal Growth 311, 2009, pp. 3075-3079.

  For the point defect evaluation, it was necessary to increase the overall film thickness in order to increase the sensitivity of the evaluation. This time, the sample produced for this purpose was formed by alternately laminating 25 cycles of InGaN quantum well layers 4 nm and GaN barrier layers 4 nm on an m-plane GaN free-standing substrate. Two types of samples, sample B and sample C, were prepared. Sample B and sample C differ only in the In composition. In the sample B, the In composition is about 10%, and in the sample C, it is about 20%.

  The growth conditions here are similar to those shown in Table 2, and the difference from Table 2 is that the thickness and period number of the InGaN quantum well layer and the GaN barrier layer are changed. The V / III ratio during the growth of the InGaN quantum well layer uses a high V / III condition of 8340 to 18080.

  Sample B has an emission wavelength of 393 nm, and Sample C has an emission wavelength of 420 nm. FIG. 10 shows PL emission spectra when sample B and sample C are evaluated using He—Cd laser excitation light. Although the structure of the sample A is different from that of the sample A described above, the PL of the sample B is unimodal and has a narrow half-value width, and the sample C has a multi-peak due to the PL spectrum. Sample B reproduces the good quality when the In composition is low, and sample C reproduces the situation when the multi-peak occurs when the In composition is high and the quality is lowered.

FIG. 11 shows the SW plot obtained by the positron annihilation experiment. In the SW plot, the size of the point defect and the defect type can be shown. As can be seen from FIG. 11, sample B has good quality with almost no point defects, while sample C has point defects, and the type of defects is V _Ga − (V _{N 3} , that is, the possibility of a complex of Ga vacancies and N vacancies has been suggested.

Further, although in FIG. 11 being displayed simultaneously defect type point associated with the oxygen, (e.g., V _GA O _N, etc.) from the plot position of the sample C, it can be seen that the relationship is smaller than the oxygen-related defects .

  In general, it is known that the content of oxygen is large in the epitaxially grown film of the m-plane GaN substrate, but it is not considered that the oxygen concentration is extremely different between sample B and sample C having slightly different wavelengths. Therefore, it is considered that the defects are not caused by oxygen.

  In summary, the emission from the InGaN quantum well layer on the m-plane GaN substrate is an overlap of the emission spectra of two components for both PL and EL, and the influence of band structure, In composition variation, lattice defects and oxygen impurities. I can't explain that.

  From the dependence of the EL emission spectrum on the current value, the light emission on the long wave side is the main light emission under the low carrier density condition. However, when the carrier density is increased, the efficiency is significantly reduced.

  In addition, in the situation where long-wave side light emission occurs, the point defect density increases. Suppressing the long-wave side light emission is essential for improving the light emission characteristics. At wavelengths from violet to blue-violet, the half-value width of the PL emission spectrum broadens due to the presence of long-wave side emission, so narrowing the PL half-value width by improving growth conditions and structure can be an indicator of quality improvement. Recognize.

  Furthermore, the PL lifetime is also an indicator of the quality of the light emitting layer. The PL lifetime at room temperature is governed by non-radiative recombination and represents the ease with which the carriers (electrons and holes) in the crystal disappear due to defects. Although the PL lifetime varies greatly depending on the polarity of the crystal, the influence of the polarity can be eliminated on the m-plane GaN substrate. Therefore, it can be said that the longer the PL lifetime is, the better the quality of the light emitting layer is.

In addition to the above, the dependency of the PL peak wavelength on the excitation light intensity is also an indicator of the emission quality. As can be seen from the EL emission characteristics, since the emission efficiency of the long wave side is low on the high injection side, as the current value increases, the emission shifts from the long wave side emission to the short wave side peak. Strength dependency is increased. The same phenomenon occurs in the PL emission spectrum measurement, and the carrier density at which such a wavelength jump is observed is approximately in the range of 1 × 10 ¹⁶ cm ⁻³ to 5 × 10 ²⁰ cm ⁻³ . Suppressing the change amount of the PL peak wavelength when the excitation light intensity is changed 1000 times as a relative value within this range is a good index for improving the light emitting layer quality.

  The above-described examination of the PL half-width, PL lifetime, and PL peak wavelength dependence on the excitation light intensity is possible with the structure of the light emitting element, but also with the MQW structure without the p-type conductive layer. In view of the above, the inventors of the present invention mainly performed the above-described investigation on the MQW structure, except for the light emission characteristics of the light emitting element.

  In order to improve the optical quality of the quantum well layer, the following viewpoints were studied. Conventionally, in a c-plane GaN substrate mainly used, a high V / III ratio condition is usually used because of concern about crystallinity deterioration due to nitrogen loss. Actually, from the atomic model, when N atoms are attached to the crystal surface in the c (+) plane, it is expected that they are attached to one dangling bond of Ga, and are unstable and easily removed. Therefore, it is considered that the high V / III ratio condition is functioning effectively.

  On the other hand, the c (−) plane shows a reverse atom bond structure. That is, the nitrogen atoms are bonded by three dangling bonds from the surface Ga, and are more stable. On the other hand, the group III atom attached to the surface is attached by one dangling bond, and is considered to be more unstable. When a similar study is performed on an m-plane GaN substrate, the situation varies depending on the off angle. For example, when the surface has an inclined surface on the c (−) side, there is a step consisting of the c (−) surface on the surface, and the nitrogen atoms at the step ends are bonded to Ga atoms by three dangling bonds. . Therefore, it is considered that the stability of the nitrogen atom is good, and conversely, the stability of the group III atom is poor.

  In the case of an m-plane GaN substrate having an off angle of 0 ° or an m-plane GaN substrate having an off angle on the c (+) side, facets are likely to be generated on the surface, and as a result, at the step end on the c (+) side. In addition, the step end on the c (−) side also appears on the surface. As a result, the same phenomenon as in the case of having an off angle on the c (−) side is observed.

  From this point of view, the m-plane GaN substrate does not need to increase the V / III ratio as conventionally performed with the c-plane GaN substrate, and conversely, it is considered that the crystallinity is improved by lowering the V / III ratio. By lowering the V / III ratio, the migration of group III atoms is improved and good step flow growth can be expected. As a result, it is considered that crystal defects that can be non-radiative recombination centers such as vacancies and emission peaks on the long wave side can be suppressed.

  However, in step flow growth, the growth surface becomes flat, and as a result, the interface becomes extremely steep. Therefore, at the interface between the InGaN quantum well layer and the GaN barrier layer, the composition changes abruptly and the lattice constants differ from each other, so that strain accumulates locally. As a result, there is a concern that the migration of surface atoms is hindered by the influence of local strain during the growth of the InGaN quantum well layer.

  In order to avoid the influence of local strain, it is effective to provide an interface strain buffer layer between the quantum well layer and the barrier layer. The interface strain control layer can be realized by a layer having a lattice constant intermediate between the quantum well layer and the barrier layer.

  As described above, step flow growth can be realized by lowering the V / III ratio, and further, by introducing an interfacial strain buffer layer, it is possible to realize a structure with better surface flatness and small local strain. As a result, it is possible to realize a good light emitting layer with a narrow PL half-value width and few non-radiative recombination centers.

(Configuration of epitaxial wafer according to this embodiment)
DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail based on examples with reference to the drawings and tables. In addition, this invention is not limited to the content demonstrated below, In the range which does not change the summary, it can change arbitrarily and can implement. The embodiments and examples schematically illustrate the present invention, and are partially emphasized, enlarged, reduced, or omitted to deepen the understanding. May not be an accurate representation of. Further, various numerical values and quantities used in the embodiments and examples are only examples, and can be variously changed as necessary.

  FIG. 12 is a cross-sectional view of the epitaxial wafer according to the present embodiment. Epitaxial wafer 1 is formed on a GaN substrate (m-plane GaN substrate) whose main surface is a surface having an off angle of 0 ° or more and 10 ° or less with respect to the m-plane, and a main surface on one side of the GaN substrate. An n-type conductive layer, a light emitting layer formed on one main surface of the n-type conductive layer, and a p-type conductive layer formed on one main surface of the light emitting layer. Note that “on the main surface” does not necessarily mean that it is directly above, but it means that it may be positioned above the main surface.

  Specifically, as shown in FIG. 12, the epitaxial wafer 1 includes an m-plane GaN substrate 2, a first undoped GaN layer 3, an n-type GaN contact layer (n-type conductive layer) 4, an AlGaN layer 5, a second It has an undoped GaN layer 6, a light emitting layer 7, a p-type AlGaN cladding layer 8, a p-type GaN contact layer 9 and a p-type InGaN contact layer 10. The m-plane GaN substrate 2 and the layers 3 to 10 are stacked in the order described above.

  The m-plane GaN substrate 2 may be a 0 ° GaN substrate or a GaN substrate with an off angle. The off angle is usually within 10 °, preferably within 6 °. The angle formed between the thickness direction of each layer formed on the m-plane GaN substrate 2 and the m-axis of the GaN-based semiconductor constituting each layer is equal to the off-angle of the m-plane GaN substrate 2. In addition, the m-plane GaN substrate 2 is more preferably a self-standing substrate generally having good crystallinity.

The first undoped GaN layer 3 is produced, for example, using TMG (trimethyl gallium) or NH ₃ (ammonia) as a raw material. The thickness of the first undoped GaN layer 3 is, for example, 1 to 1000 nm, preferably 2 to 20 nm. The growth temperature of the first undoped GaN layer 3 is usually about 900 ° C. to 1100 ° C. The first undoped GaN layer 3 functions to stabilize the surface of the m-plane GaN substrate 2 and improve the quality of each layer on the upper surface.

The n-type GaN contact layer 4 is doped with n-type impurities such as Si (silicon) and Ge (germanium), for example. The n-type GaN contact layer 4 is produced using, for example, SiH ₄ (silane), TMG, and NH ₃ as raw materials. The thickness of the n-type GaN contact layer 4 is, for example, 1 to 6 μm, preferably 2 to 4 μm, and the growth temperature is the same as that of the first undoped GaN layer 3. The n-type impurity concentration is, for example, 2 × 10 ¹⁸ to 2 × 10 ¹⁹ cm ⁻³ , preferably 5 × 10 ¹⁸ cm ^{−3 to} 1 × 10 ¹⁹ cm ⁻³ .

The AlGaN layer 5 is produced, for example, using TMG, TMA (trimethylaluminum) NH ₃ as a raw material. The thickness of the AlGaN layer 5 is, for example, 5 to 100 nm, preferably 5 to 20 nm.

  The AlGaN layer 5 has an effect of reducing crystal defects that are likely to be generated particularly by MOCVD growth on an m-plane GaN substrate having an off angle of 2 ° or more, and works to improve the quality of the light emitting layer formed thereon. is there.

The second undoped GaN layer 6 is produced using, for example, TMG and NH ₃ as raw materials. The thickness of the second undoped GaN layer 6 is, for example, 20 to 1000 nm, preferably 50 to 200 nm.

  The first undoped GaN layer 3, the AlGaN layer 5, and the second undoped GaN layer 6 may be omitted depending on circumstances. That is, the n-type GaN contact layer 4 can be provided immediately above the m-plane GaN substrate 2, and the light emitting layer 7 can be provided directly above the n-type GaN contact layer 4. The AlGaN layer 5 may be undoped or doped.

The light emitting layer 7 may be a single layer made of InGaN or InAlGaN, but is preferably a multiple quantum well layer (MQW) having a structure in which quantum well layers and barrier layers are alternately stacked. The quantum well layer is preferably formed of a GaN-based semiconductor containing In, such as InGaN or InAlGaN. In the case where the light emitting layer 7 is an InGaN quantum well layer / GaN barrier layer, the InGaN quantum well layer is produced using, for example, TMI (trimethylindium), TMG, and NH ₃ as raw materials. The GaN barrier layer is produced using TMG and NH ₃ as raw materials, for example. The thickness of the quantum well layer is, for example, 2 to 15 nm, preferably 3 to 10 nm. The thickness of the barrier layer is, for example, 2 to 30 nm, preferably 4 to 20 nm. In addition, the number of repetition periods of the quantum well layer and the barrier layer is usually in a range from 2 periods to 12 periods.

When producing the quantum well layer and the barrier layer of the light emitting layer 7, TEG (triethyl gallium) or a mixed gas of TMG and TEG may be used as a gallium raw material instead of TMG.
The barrier layer may be a GaN-based semiconductor having a band gap energy larger than that of the quantum well layer, and a GaN layer or an InGaN layer having an In composition smaller than the quantum well layer can be used.

The p-type AlGaN cladding layer 8 is made of, for example, Al _y Ga _1-y N (preferably 0.04 ≦ y ≦ 0) having a larger band gap energy than both the light emitting layer 7 and the p-type GaN contact layer 9. .2). The p-type AlGaN cladding layer 8 is doped with p-type impurities such as Mg (magnesium) and Zn (zinc), for example. The thickness of the p-type AlGaN cladding layer 8 is, for example, 10 to 200 nm, preferably 10 nm to 50 nm. The p-type impurity concentration is, for example, 1 × 10 ^{19 to} 5 × 10 ²⁰ cm ⁻³ .

  The p-type AlGaN cladding layer 8 can be omitted. That is, it is possible to provide the p-type GaN contact layer 9 immediately above the light emitting layer 7.

The p-type GaN contact layer 9 is doped with p-type impurities such as Mg and Zn. The thickness of the p-type GaN contact layer 9 is, for example, 40 to 200 nm. The p-type impurity concentration is, for example, 1 × 10 ^{19 to} 5 × 10 ²⁰ cm ⁻³ , and the impurity concentration is intentionally changed inside. Al may be mixed into the p-type GaN contact layer 9 to form a p-type Al _x Ga _1-x N (preferably 0.01 ≦ x ≦ 0.05) contact layer.

The p-type InGaN contact layer 10 is formed of, for example, In _x Ga _1-x N (preferably 0.01 ≦ x ≦ 0.05) and is doped with a p-type impurity such as Mg or Zn. The thickness of the p-type InGaN contact layer 10 is, for example, 1 to 20 nm, preferably 10 nm or less, and particularly preferably 5 nm or less. The composition of the p-type InGaN contact layer 10 is preferably determined such that its band gap energy is larger than the band gap energy of the light emitting layer 7 (the band gap energy of the well layer when the active layer is MQW). .

  The p-type InGaN contact layer 10 can be omitted. That is, the epitaxial growth can be terminated at the p-type GaN contact layer 9.

  Moreover, you may add the layer which is not described above. Specifically, an n-type layer may be added between the second undoped GaN layer 6 and the light emitting layer 7. Alternatively, an undoped or p-type layer with a low Mg concentration may be added between the light emitting layer and the p-type AlGaN cladding layer 8. In some cases, reliability can be improved by adding such a layer.

  As described above, the configuration of the p-type conductive layer, the n-type conductive layer, and the light-emitting layer is indispensable for manufacturing LED elements and laser elements. In order to improve or confirm the quality, it is more desirable that there is no p-type conductive layer. This is because when there is a p-type conductive layer, it is necessary to irradiate excitation light through the p-type conductive layer when performing photoluminescence measurement (hereinafter referred to as PL measurement), depending on the structure of the p-type conductive layer. This is because the excitation light is attenuated. Also, since photoluminescence light (hereinafter referred to as PL light) generated from the light emitting layer is received through the p-type conductive layer, it is attenuated depending on the structure of the p-type conductive layer. is there.

  In the present specification, in view of the above, the experimental study is mainly conducted with a structure without a p-type conductive layer. There, the crystal growth is terminated in the multiple quantum well layer (MQW layer). Hereinafter, a structure in which the p-type conductive layer is not formed is referred to as an MQW structure, and a structure in which the p-type conductive layer is formed is referred to as an LED structure.

  However, in this specification, the intensity of PL light affected by attenuation in the p-type conductive layer is not a problem, and only the shape of the PL emission spectrum is a problem, so that the MQW structure and the LED structure can be compared with each other. .

In addition to this configuration, the PL peak wavelength of the light emitting layer 7 is not less than 410 nm and not more than 460 nm, and the PL half width Δl of the light emitting layer 7 satisfies the following conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)

Since the shape of the PL emission spectrum depends on the excitation light intensity, it is necessary to use weak excitation conditions that reflect the optical quality more sensitively. Here, a He—Cd laser (wavelength: 325 nm) that is easy to use is used, the radiant flux on the substrate is adjusted to 0.7 millijoule / second (0.7 mW), and the irradiation size is about 0.1 mmφ. As evaluated. Since it is 8.9 W / cm ² per area and the carrier density generated in the InGaN quantum well layer in the light emitting layer is considered to be approximately 5 × 10 ¹⁷ cm ⁻³ or less, it is desirable because it is a weak excitation condition. .

  It is necessary to evaluate the PL emission spectrum with an excitation light intensity equal to or lower than the excitation light intensity. The PL emission spectrum was obtained by dispersing PL light with a spectroscope and measuring with a light receiving device. The PL half width of the light emitting layer 7 refers to the full width at half maximum of the PL emission spectrum.

  For reference, an example of the relationship between the PL peak wavelength satisfying the conditional expression (1) and the PL half width is shown. For example, when the PL peak wavelength is 420 nm, the PL half-value width that satisfies Equation (1) is 18 nm or less. When the PL peak wavelength is 430 nm, the PL half-value width satisfying the expression (1) is 22 nm or less. Furthermore, when the PL peak wavelength is 440 nm, the PL half width satisfying the expression (1) is 26 nm or less.

  Moreover, the PL peak wavelength of the light emitting layer 7 is 410 nm or more and 460 nm or less, and the PL lifetime by the time-resolved PL measurement of the light emitting layer 7 at room temperature is 1.3 nsec or more and 20 nsec or less. The PL life is preferably 1.5 nsec or more and 15 nsec or less, more preferably 1.7 nsec or more and 10 nsec or less.

  Here, the time-resolved PL measurement was performed at room temperature using a time-correlated single photon coefficient method. As the light source, a tunable pulse laser constituted by a mode-locked Ti: sapphire laser and a harmonic generation crystal was used. The pulse repetition frequency was 80 MHz and the pulse width was 2 ps.

  In order to directly evaluate the optical quality of a quantum well layer having a single layer or a plurality of layers inherent in the active layer structure, time-resolved PL measurement by selective excitation is effective. In order to selectively excite the quantum well layer, in the structure of the present embodiment, the energy of the band gap of the InGaN quantum well layer is larger, and the other layers (here, GaN and AlGaN) are used. It is necessary to select excitation light having energy smaller than the band gap. Therefore, the wavelength of the tunable pulse laser was set to 385 nm.

  Subsequently, after the pulse energy was adjusted by the ultraviolet ND filter, the sample attached to the stage was irradiated. PL light from the sample was guided to a photomultiplier after being dispersed by a spectroscope through a condenser lens.

The pulse energy applied to the sample was determined by measuring the power with a power meter and dividing by the repetition frequency. The laser beam diameter was φ0.1 mm at the sample position. Thereby, the pulse energy density per unit area is 1.6 μJ / cm ² , and the excess carrier density to be excited is estimated to be approximately 1 × 10 ¹⁷ cm ⁻³ .

  This condition is a sufficiently low excitation condition for evaluating the light emission quality of the quantum well layer, and in order to evaluate the light emission quality, it is necessary to use an excitation condition equal to or less than this.

  Here, first, the PL emission spectrum was measured, and then the peak wavelength of the PL emission spectrum was selected with a spectroscope to perform time-resolved PL measurement. Subsequently, the PL life was obtained from the transient response (decay curve) with respect to time of the PL intensity after pulse excitation. The time from the maximum intensity to 1 / e of the maximum intensity in the PL intensity decay curve is defined as the PL life. In general, the decay curve in the time-resolved PL measurement often does not have a single exponential function, but here, the PL life is defined as described above. Actually, the attenuation curve obtained in this example showed a curve close to a single exponential function.

Furthermore, the PL peak wavelength of the light emitting layer 7 is 410 nm or more and 460 nm or less, and the light emitting layer 7 has a wavelength variation of 0 nm or more and 10 nm or less when the excitation intensity is changed due to the excitation density dependency. Note that the wavelength variation when the excitation intensity is changed due to the excitation density dependency is preferably 0 nm to 8 nm, more preferably 0 nm to 6 nm. The wavelength variation here is that the excitation light intensity is increased from 1 to 1000 times when the carrier density generated in the InGaN quantum well layer is in the range of approximately 1 × 10 ¹⁶ cm ⁻³ to 5 × 10 ²⁰ cm ^−3. This refers to the change in the PL peak wavelength when it is varied. In the measurement, the excitation light is condensed with an objective lens to obtain a high energy density, and the energy density (relative value) is changed to a value of 1 to 1000 using an ND filter.

  FIG. 13 is a cross-sectional view of the light emitting layer 7 according to this embodiment. The light emitting layer 7 preferably includes an InGaN layer. As shown in FIG. 2, the light emitting layer 7 is preferably composed of a multiple quantum well layer in which quantum well layers 7A and barrier layers 7B are alternately stacked, and at least one of the quantum well layers 7A and the barrier layers 7B. An interfacial strain buffer layer 7C that has an intermediate lattice constant between the quantum well layer 7A and the barrier layer 7B and controls strain between the quantum well layer 7A and the barrier layer 7B is provided at the side interface. The interfacial strain buffer layer 7C is preferably formed of a GaN-based semiconductor containing In, such as InGaN or InAlGaN. The In composition of the interfacial strain buffer layer 7C is a composition between the quantum well layer 7A and the barrier layer 7B, and the thickness is, for example, 0.1 to 3 nm, preferably 0.5 nm to 2 nm. Moreover, even if it is a layer of a single composition, a composition may change in steps.

  The interfacial strain buffer layer 7C can be omitted. That is, the light emitting layer 7 may be only a multiple quantum well layer in which the quantum well layers 7A and the barrier layers 7B are alternately stacked.

Furthermore, the dark spot density of the m-plane GaN substrate 2 is 2 × 10 ⁸ cm ⁻² or less. More preferably, it is 2 × 10 ⁷ cm ⁻² or less, and further preferably 2 × 10 ⁶ cm ⁻² or less. Here, the dark spot density of the m-plane GaN substrate 2 refers to the density at which lattice defects such as misfit dislocations in the crystal are observed as dark spots in the cathodoluminescence image. Such a low dark spot density can be realized by using a GaN free-standing substrate instead of a heterogeneous substrate.

  Furthermore, the main surface on the other side of the m-plane GaN substrate 2 is roughened. Here, the roughening of the m-plane GaN substrate 2 is an uneven surface shape that can improve the light extraction efficiency, and the shape can be imparted using surface treatment or dry etching of the substrate. The surface roughness of the m-plane GaN substrate 2 is preferably such that the ratio of the flat surface having the original plane orientation is 50% or less.

(Configuration of Semiconductor Light-Emitting Element According to this Embodiment)
FIG. 14 is a schematic diagram of the semiconductor light emitting device 20 according to this embodiment. 14A is a top view, and FIG. 14B is a cross-sectional view taken along the line XX in FIG. 14A.

  The semiconductor light emitting element 20 is formed by performing processes such as etching, electrode formation, element separation, etc. on the epitaxial wafer 1 described above. The semiconductor light emitting device 20 includes an n-type conductive layer composed of a GaN-based semiconductor layer whose main surface is a surface having an off angle of 0 ° or more and 10 ° or less with respect to the m-plane, and a main surface on one side of the n-type conductive layer. A light emitting layer formed on the surface and a p-type conductive layer formed on the main surface on one side of the light emitting layer.

And the light emission peak wavelength of the said light emitting layer is 410 nm or more and 460 nm or less, and the PL half value width of the said light emitting layer is the range which satisfy | fills the following conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm)

  Moreover, the emission peak wavelength of the light emitting layer 7 is 410 nm or more and 460 nm or less, and PL lifetime by the time-resolved PL measurement of the light emitting layer 7 at room temperature is 1.3 nsec or more and 20 nsec or less.

  Furthermore, the semiconductor light emitting device 20 according to this embodiment has a small wavelength variation due to current injection. For example, the wavelength fluctuation due to current injection up to 350 mA at an emission peak wavelength of 420 nm is 6 nm or less, preferably less than the resolution (about 1 nm) of the measurement system.

  Specifically, as shown in FIG. 14B, the semiconductor light emitting device 20 includes an m-plane GaN substrate 2, a first undoped GaN layer 3, an n-type GaN contact layer (n-type conductive layer) 4, and an AlGaN layer 5. , Second undoped GaN layer 6, light-emitting layer 7, p-type AlGaN cladding layer 8, p-type GaN contact layer (p-type conductive layer) 9, p-type InGaN contact layer 10, n-side metal electrode 11, p-side contact electrode 12 and a p-side metal electrode 13. The m-plane GaN substrate 2 may be removed by peeling or the like.

  As shown in FIG. 14B, the n-side metal electrode 11 is formed on a partly exposed surface of the n-type GaN contact layer 4. The p-side contact electrode 12 is formed on the upper surface of the p-type InGaN contact layer 10 as shown in FIG. The p-side contact electrode 12 may be, for example, ITO, nickel, platinum, titanium, silver, tungsten, chromium, or an alloy containing these metals. The p-side metal electrode 13 is formed as a pad electrode on a part of the p-side contact electrode 12.

  The semiconductor light emitting device 20 according to the present embodiment can exhibit high output performance at 410 nm or more by current injection. The relationship between the emission peak wavelength and the emission output when 350 mA is injected is as follows. For example, when the emission peak wavelength is about 420 nm, the emission output is 490 mW or more. When the emission peak wavelength is about 430 nm, the emission output is 370 mW or more. Furthermore, when the emission peak wavelength is about 440 nm, the emission output is 350 mW or more.

  In the semiconductor light emitting device 20 according to the present embodiment, the case where the n-side contact electrode 11 is formed on the partially exposed surface of the n-type GaN contact layer 4 has been described. However, the present invention is not limited to this, for example, The n-side contact electrode 11 may be formed on the other main surface of the m-plane GaN substrate 2 and can be applied to various other semiconductor light emitting devices.

(Configuration of Light-Emitting Device According to this Embodiment)
FIG. 15 is a cross-sectional view of the light emitting device 30 according to the present embodiment. The light emitting device 30 includes the semiconductor light emitting element 20 described above and a wavelength conversion material that absorbs at least part of the light emitted from the semiconductor light emitting element 20 and converts it into light having a longer wavelength. Specifically, the light emitting device 30 includes a semiconductor light emitting element 20, a package 21 that houses the semiconductor light emitting element 20, a translucent material 22, and a wavelength conversion unit 23.

  The semiconductor light emitting element 20 has the same configuration as the semiconductor light emitting element 20 according to the present embodiment, and is, for example, a semiconductor light emitting element that emits blue-violet light to blue light having an emission peak wavelength of 410 nm or more and 460 nm or less.

  The package 21 is a known package. For example, various types of packages other than a type in which a heat-resistant resin such as polyamide resin, epoxy resin, or silicone resin is integrally formed with a lead frame can be applied. As the translucent material 22, for example, a silicone resin or glass can be used. Further, the translucent material 22 may be omitted, and the space between the semiconductor light emitting element 20 and the wavelength conversion unit 23 may be a cavity.

  The wavelength conversion unit 23 contains, for example, a yellow phosphor, and converts a part of blue-violet light to blue light emitted from the semiconductor light emitting element 20 into yellow light. And the wavelength conversion part 23 discharge | releases toward the exterior white light which a surface blue light and yellow light mix.

  Applications of the light emitting device 30 according to the present embodiment include illumination, a display, a backlight of a liquid crystal display device, an indicator, and the like, but are not limited thereto.

(Epitaxial wafer manufacturing method according to this embodiment)
FIG. 16 is a flowchart of the method for manufacturing the epitaxial wafer 1 according to this embodiment. In the manufacturing method of epitaxial wafer 1 according to the present embodiment, an n-type conductive layer is formed on one main surface of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane. A first step of growing, and a second step of growing a light emitting layer on a main surface on one side of the n-type conductive layer grown in the first step.

  Specifically, in the manufacturing method of the epitaxial wafer 1, first, the first undoped GaN layer 3, the n-type GaN contact layer 4, the AlGaN layer 5, and the second undoped are formed on the main surface on one side of the m-plane GaN substrate 2. The GaN layer 6 is laminated and grown under a predetermined condition by metal organic chemical vapor deposition (MOCVD) (step SP1).

  Subsequently, in the manufacturing method of the epitaxial wafer 1, the group V material and the group III material are supplied on the main surface on one side of the second undoped GaN layer 6 so that the V / III ratio is 500 or more and 4000 or less. The repetitive structure of the grown quantum well layer 7A and barrier layer 7B interfacial strain buffer layer 7C is laminated and grown under a predetermined condition by metal organic vapor phase epitaxy to form the light emitting layer 7 (step SP2). ).

  Here, in the manufacturing method of the epitaxial wafer 1, it is preferable to grow a quantum well layer made of an InGaN layer as the quantum well layer 7A. Furthermore, it is preferable to supply the group III raw material so that the ratio of the supply amount of the indium raw material to the total supply amount of the group III raw material is 50% or more and 90% or less during the growth of the quantum well layer composed of the InGaN layer. The ratio of the supply amount of the indium raw material here is preferably 70% or more and 90% or less, and more preferably 80% or more and 90% or less.

  Furthermore, as a growth condition of the light emitting layer 7, it is desirable to grow at a relatively low growth rate of 1 nm / min to 8 nm / min. The growth rate here is preferably 1 nm / min or more and 7 nm / min or less.

  Subsequently, in the method for manufacturing the epitaxial wafer 1, the p-type AlGaN cladding layer 8, the p-type GaN contact layer 9, and the p-type InGaN contact layer 10 are formed on the main surface on one side of the light emitting layer 7 by the metal organic chemical vapor deposition method. Are stacked and grown under predetermined conditions (step SP3).

  The results of experiments conducted by the present inventors will be described below. However, the present invention is not limited by the method used in these experiments and the structure of the sample.

Example 1
The epitaxial wafer and semiconductor light emitting device of Example 1 have the same configuration as the epitaxial wafer 1 and semiconductor light emitting device 20 shown in FIGS. The epitaxial wafer and semiconductor light emitting device of Example 1 were produced according to the following procedure.

(Epitaxial growth)
First, an m-plane GaN substrate having a length × width × thickness of 8 mm × 20 mm × 330 μm was prepared in the MOVPE apparatus. This m-plane GaN substrate had a carrier concentration in the range of 1.0 × 10 ¹⁷ cm ^{−3 to} 5.0 × 10 ¹⁷ cm ⁻³ and an off angle in the + c direction of −5 °. A semiconductor laminate was epitaxially grown on the polished front surface of the prepared m-plane GaN substrate using the atmospheric pressure MOVPE method. That is, on the main surface on one side of the m-plane GaN substrate, the first undoped GaN layer, the n-type GaN contact layer, the AlGaN layer, the second undoped GaN layer, the light emitting layer, the p-type AlGaN cladding layer, and the p-type GaN A contact layer and a p-type InGaN contact layer were sequentially epitaxially grown.

The first undoped GaN layer was grown to a thickness of 10 nm using a substrate temperature of 900 ° C. and TMG and NH ₃ as raw materials. The growth rate at this time is 15 nm / min. Further, the n-type GaN contact layer was grown at a substrate temperature of 900 ° C., using TMG, NH ₃ , and silane as raw materials to a Si concentration of about 7 × 10 ¹⁸ cm ⁻³ and a thickness of 1500 nm. The growth rate at this time was 17 nm / min. The AlGaN layer was grown to a thickness of 6 nm using a substrate temperature of 900 ° C. and using TMG, TMA, and NH ₃ as raw materials. The second undoped GaN layer was grown to a thickness of 100 nm using a substrate temperature of 815 ° C. and TMG and NH ₃ as raw materials.

For the light emitting layer, TMG, TMI, and NH ₃ are used as raw materials, and four GaN barrier layers and three InGaN quantum well layers are alternately grown so that the lowermost layer and the uppermost layer are barrier layers. Was formed. An interfacial strain buffer layer was inserted between the GaN barrier layer and the InGaN well layer. A waiting time of 5 seconds was provided between the growth of the quantum well layer and the interface strain buffer layer. The growth temperature was constant between the interface strain buffer layer and the InGaN quantum well layer, and was in the range of 750 ° C. to 770 ° C. The growth temperature of the GaN barrier layer was in the range of 790 ° C. to 810 ° C., and a temperature 40 ° C. higher than the growth temperature of the InGaN quantum well layer was adopted. When changing the growth temperature between the interface strain control layer and the barrier layer, a waiting time of 2 minutes was provided. Here, a plurality of structures having different wavelengths were prepared by adjusting the temperature. The thickness of the InGaN quantum well layer was 3.6 nm, the thickness of the GaN barrier layer was 18 nm, the thickness of the interface strain buffer layer was 1 nm, and the In composition was about 4%. No impurities were added to the light emitting layer.

During the growth of the light emitting layer, the NH ₃ flow rate was kept constant at 2.8 L / min. The supply molar flow rates of TMI and TMG in the InGaN quantum well layer are 61 μmol / min and 14 μmol / min, respectively, and the V / III ratio, which is the ratio of the supply molar flow rate of the group V raw material and the group III raw material, is 1670. Met. The supply molar flow rates of TMI and TMG in the InGaN interfacial strain buffer layer were 10 μmol / min and 14 μmol / min, respectively, and the V / III ratio was 5160. The feed molar flow rate of TMG at the GaN barrier layer was 14 μmol / min and the V / III ratio was 8930.

The p-type AlGaN cladding layer was grown to a thickness of 40 nm using TMG, TMA, biscyclopentadienyl magnesium and NH ₃ as raw materials at a substrate temperature of 900 ° C. The p-type AlGaN cladding layer is doped with Mg, and the Mg concentration is about 2 × 10 ¹⁹ cm ⁻³ . The flow rates of TMG and TMA were adjusted so that the crystal composition was Al _0.1 Ga _0.9 N.

The p-type GaN contact layer was grown to a thickness of 40 nm using a substrate temperature of 900 ° C. and using TMG, TMA, NH ₃ , and biscyclopentadienyl magnesium as raw materials. The p-type GaN contact layer is doped with Mg, and the Mg concentration is about 7 × 10 ¹⁹ cm ⁻³ . Immediately after the growth of the p-type GaN contact layer was completed, the supply of NH ₃ was stopped, the substrate heating was also stopped, and the substrate was cooled to 820 ° C.

The p-type InGaN contact layer was grown to a thickness of 1 nm using a substrate temperature of 820 ° C. and using TMG, TMI, NH ₃ , and biscyclopentadienyl magnesium as raw materials. The p-type InGaN contact layer is doped with Mg, but the Mg concentration is unknown because of the surface layer.

  All of the above crystal growth was performed using nitrogen as a carrier gas. As a result, the epitaxial wafer of Example 1 was prepared.

  Immediately after crystal growth was completed, the heating of the epitaxial wafer was stopped and cooled, and the flow of nitrogen as a carrier gas was continued until the epitaxial wafer was cooled.

  Table 4 shows the structure of each layer of the epitaxial wafer of Example 1, the supply amount of ammonia, the supply amount of raw materials, the V / III ratio, the substrate temperature, the thickness, and the dopant.

(Formation of mesa and p-side electrode)
An ITO layer having a thickness of 210 nm was formed as a p-type contact layer on the entire surface of the epitaxial wafer obtained by the above procedure on the p-type InGaN contact layer. Thereafter, heat treatment was performed at 520 ° C. for 20 minutes in an air atmosphere using a quartz heating furnace. Thereafter, etching was performed by an RIE dry etching apparatus until the n-type GaN contact layer was reached to form a mesa.

(Formation of n-side electrode)
Next, an n-side metal electrode was formed on the surface of the n-type GaN contact layer 2 partially exposed by mesa formation. The n-side metal electrode was a laminated film including an Al layer (thickness 100 nm) and an Au layer (thickness 300 nm) in this order. The n-side metal electrode was patterned by a normal lift-off method.

(Formation of p-side electrode)
Next, a p-side metal electrode including a Ti—W layer (thickness: 108 nm) and an Au layer (thickness: 300 nm) in this order is formed as a pad electrode on the n-side contact electrode and the p-side contact electrode, and then heat treatment is performed. Using a furnace, heat treatment (alloy treatment) was performed at 500 ° C. for 1 minute in a nitrogen atmosphere.

(Roughening of the main surface on the other side of the substrate)
In order to roughen the main surface on the other side of the m-plane GaN substrate, a circular SiO ₂ mask having a thickness of 1.0 μm and a diameter of 2.0 μm is used, and the distance between centers of adjacent masks is set to 6.0 μm. A triangular lattice was produced. Thereafter, dry etching was performed using chlorine gas to form a protrusion having a height of 4.0 μm.

  Finally, the epitaxial wafer was divided into 550 μm × 550 μm squares using a diamond scriber to produce the semiconductor light emitting device of Example 1.

(Example 2)
The epitaxial wafer and semiconductor light emitting device of Example 2 were prepared in the same manner as the semiconductor light emitting device of Example 1 except that an InGaN interface strain buffer layer was not inserted between the GaN barrier layer and the InGaN well layer.

(Comparative Example 1)
The epitaxial wafer and the semiconductor light emitting device of Comparative Example 1 mainly have a higher ammonia flow rate during the growth of the InGaN quantum well layer to increase the V / III ratio of the light emitting layer, and the InGaN between the GaN barrier layer and the InGaN well layer. It was produced in the same manner as the semiconductor light emitting device of Example 1 except that no interfacial strain control layer was inserted.

The light emitting layer of Comparative Example 1 is composed of four layers of GaN barrier layers and three layers of InGaN quantum well layers, using TMG, TMI, and NH ₃ as raw materials so that the lowermost layer and the uppermost layer are barrier layers. It was formed by growing alternately. The growth temperature was 790 to 810 ° C. for the GaN barrier layer and 750 to 770 ° C. for the InGaN quantum well layer. The growth temperature of the GaN barrier layer was 40 ° C. higher than the growth temperature of the InGaN quantum well layer. The thickness of the InGaN quantum well layer was 3.6 nm, and the thickness of the GaN barrier layer was 18 nm. No impurities were added to the light emitting layer.

During the growth of the light emitting layer, the NH ₃ flow rate was kept constant at 14 L / min. The supply molar flow rates of TMI and TMG in the InGaN quantum well layer were 61 μmol / min and 14 μmol / min, respectively, and the V / III ratio was 8340. The feed molar flow rate of TMG in the GaN barrier layer was 14 μmol / min and the V / III ratio was 43520.

  The surface of the epitaxial wafer after completion of the epitaxy in Example 1 and Example 2 was flat as observed with an optical microscope. Moreover, even in the evaluation with a fluorescent microscope using a mercury lamp, the field of view was uniform and no special structure was found. From the above results, in Example 1 and Example 2, no crystal defects are observed, and the in-plane In concentration is considered to be relatively uniform.

(Evaluation)
FIG. 17 shows the wavelength dependence of the total radiant flux when 350 mA is applied to the semiconductor light emitting devices of Example 1, Example 2, and Comparative Example 1.

  As can be seen from FIG. 17, the V / III ratio of the quantum well layers of Example 1 and Example 2 is 1670 in common, and the TMI molar supply amount ratio in the group III raw material is 81%. The light emission output of the semiconductor light emitting devices of Example 1 and Example 2 is higher than the light emission output of the semiconductor light emitting device of Comparative Example 1 by 100 mW or more when the emission peak wavelength is the same.

  FIG. 18 shows the PL peak dependence of the PL half-value width obtained from the semiconductor light emitting devices of Example 1, Example 2, and Comparative Example 1.

  As can be seen from FIG. 18, the PL half-value widths of the semiconductor light-emitting elements of Example 1 and Example 2 are similar to the PL half-value width of the semiconductor light-emitting element of Comparative Example 1 when the emission peak wavelength is the same. Remarkably small. In addition, the PL lifetime of the semiconductor light emitting devices of Example 1 and Example 2 is significantly longer than the PL lifetime of the semiconductor light emitting device of Comparative Example 1 when the emission peak wavelength is the same.

  FIG. 19 shows the EL emission spectrum, the current value dependency of the EL emission peak wavelength, and the current value dependency of the total radiant flux in the semiconductor light emitting devices of Example 1 and Comparative Example 1.

  The V / III ratio of the quantum well layers of Example 1 and Example 2 is 1670 in common, and the TMI molar supply ratio in the Group III raw material is 81%. As can be seen from FIG. 19, in Example 1, the EL emission spectrum is unimodal, while in Comparative Example 1, multi-peak emission having a shoulder on the short wave side is shown. In Example 1, the emission peak wavelength was not changed by current injection from 1 mW to 350 mW, whereas in Comparative Example 1, the wavelength was shortened by 10 nm or more as current injection increased.

  In Comparative Example 1 in which long-wave side light emission with low efficiency occurs, the EL light emission peak is strongly influenced by long-wave side light emission in the low current region, but as the injection current value increases, the short-wave side light emission becomes relatively strong. The EL emission peak is strongly influenced by short-wave side emission. The wavelength is rapidly shortened by current injection.

  The EL emission spectrum of Comparative Example 1 is a situation where the emission peak is changing from the long wave side to the short wave side, and a bulge is seen on the short wave side of the spectrum shape. However, since the long-wave side light emission is reduced in Example 1 with improved crystallinity, the EL spectrum shape is almost unimodal, and the light emission mode remains single even if the current value is changed. is there. As a result, the current density dependency of the wavelength becomes extremely small.

(Example 3)
In Example 3, the relationship between the PL peak wavelength and the PL half width measured by changing the growth conditions of the LED structure and the MQW structure is plotted in FIG.

  Example 3-1 is the semiconductor light emitting device of Example 1, and Example 3-2 is the semiconductor light emitting device of Example 2. Example 3-3 is a similar structure of the semiconductor light emitting device of Example 1 and is an LED structure formed up to the p-type conductive layer, and Example 3-4 is a similar structure of the semiconductor light emitting device of Example 1, It is an MQW structure in which a p-type conductive layer is not formed and a multiple quantum well layer is formed.

  The m-plane GaN substrate used in the LED structure of Example 3-3 and the MQW structure of Example 3-4 is formed on a 2-inch m-plane GaN substrate, and the small rectangular m-plane GaN of Example 1 is used. The method of creating the substrate is different from that of the substrate (described later). The V / III ratio during the growth of the InGaN quantum well layer of the LED structure of Example 3-3 and the MQW structure of Example 3-4 is 1670 as in Example 1.

Example 3-5 is an MQW structure in which the NH ₃ flow rate during the growth of the light emitting layer is changed from the structure of the semiconductor light emitting device of Example 1, the flow rate is 5.6 L, and the V / III ratio is 3340. Example 3-6 is a similar structure of the semiconductor light emitting device of Example 2, has an MQW structure, further reduces the TMI molar supply amount to 74%, and has a V / III ratio during the growth of the InGaN quantum well layer. Is 2280.

  The PL half-value widths of the MQW structures of Example 3-3 and Example 3-6 are all increased according to the wavelength as compared with the PL half-value widths of the semiconductor light emitting devices of Example 1 and Example 2. Although it is narrow overall.

Comparative Example 3-1 is a similar structure of the semiconductor light emitting device of Comparative Example 1, has an MQW structure, NH ₃ flow rate during growth of the InGaN quantum well layer is 14 L, and V / III ratio is 15480. Comparative Example 3-2 is the MQW structure shown in Table 2 above. The NH ₃ flow rate during the growth of the InGaN quantum well layer is 14 L, and the V / III ratio is 8340. The off-angle of the m-plane GaN substrate used in Comparative Example 3-2 is 0 °, and accordingly, the substrate growth conditions are different from those of the substrate having an off-angle of −5 °.

  Comparative Example 3-3 has a similar structure to the semiconductor light emitting device of Example 1, has an MQW structure, reduces the TMI supply amount during the growth of the InGaN quantum well layer to ¼ that of Example 1, and increases the V / III ratio. Increased to 4240

  The PL half-value widths of the MQW structures of Comparative Examples 3-1 to 3-3 are all clearly wider than the PL half-value widths of the semiconductor light emitting devices of Example 1 and Example 2. .

  Table 5 shows changes in structure or growth conditions from the semiconductor light emitting devices of Example 1 and Example 2 (Table 4).

  As can be seen from FIG. 20, both the example and the comparative example have wider half-value widths as the wavelength becomes longer, but the examples and the comparative example have clearly different distributions, and conditional expression (1) It is divided into two with the line of.

  In Example 3-1 to Example 3-6, the wavelength dependence of the PL half-value width was almost the same.

The method for producing the 2-inch m-plane GaN substrate used in Example 3-3 and Example 3-4 is described below. A 2-inch m-plane GaN substrate was manufactured by the following procedure.
(I) A primary GaN crystal was grown by HVPE using a GaN template on C-plane sapphire with a mask pattern formed on the main surface as a seed, and a -c plane substrate (primary substrate) was cut out from the primary GaN crystal.
(Ii) A secondary GaN crystal was grown by an ammonothermal method using the primary substrate as a seed, and an m-plane GaN substrate (secondary substrate) was cut out from the secondary GaN crystal.
(Iii) Using the secondary substrate as a seed, a tertiary GaN crystal was grown by an ammonothermal method, and an m-plane GaN substrate (tertiary substrate) was cut out from the tertiary GaN crystal.
(Iv) A target 2 inch m-plane GaN substrate (quaternary substrate) was prepared by the HVPE method using the tertiary substrate as a seed.

Example 4
In Example 4, the relationship between the PL peak wavelength and the PL lifetime measured by changing the growth conditions of the LED structure and the MQW structure is plotted in FIG.

  Example 4-1 is the semiconductor light emitting device of Example 1, and Example 4-2 has a similar structure to the semiconductor light emitting device of Example 2 and has an MQW structure. Example 4-3 and Example 4-6 are the same as the LED structure and MQW structure of Example 3-3 to Example 3-6. Comparative example 4-1 and comparative example 4-2 are the same as the MQW structure of comparative example 3-1 and comparative example 3-2.

  Comparative Example 4-3 is a similar structure of the semiconductor light emitting device of Example 2 and has an MQW structure. The TMG flow rate during the growth of the InGaN quantum well layer is increased to increase the growth rate to 8.4 nm / min. By adjusting the time, the thicknesses of the quantum well layers are made equal, the V / III ratio is 2740, and the TMI molar supply ratio is 2740.

  As can be seen from FIG. 21, Examples 4-1 to 4-6 all show that the PL life is sufficiently longer than 1.3 nsec and the optical quality is good. On the other hand, Comparative Example 4-3 has a narrow PL half-value width, but has a short PL life. When the TMI molar supply ratio is low, the optical quality deteriorates. Comparative Example 4-1 to Comparative Example 4-3 have a short PL life and poor optical quality.

(Example 5)
In Example 5, the excitation intensity dependence of the PL peak wavelength measured by changing the growth conditions of the LED structure and the MQW structure is plotted in FIG.

  Example 5-1 and Example 5-2 are the same as the MQW structure of Example 3-6, and adopt different portions of the in-plane PL peak wavelength. Example 5-3 is the same as the MQW structure of Example 3-4. Comparative Example 5-1 is the same as the MQW structure of Comparative Example 3-2.

  As can be seen from FIG. 22, in Example 5-1 to Example 5-3, the PL peak wavelength has almost no excitation intensity dependency, whereas in Comparative Example 5-1, as the PL peak intensity increases, the PL peak increases. The wavelength fluctuates greatly.

(Example 6)
In Example 6, the relationship between the excitation light intensity at room temperature and the PL lifetime when the excitation light intensity of the MQW structure is changed is plotted in FIG.

  Example 6-1 is the same as the MQW structure of Example 4-4. Comparative Example 6-1 is the same as the MQW structure of Comparative Example 4-2.

The measurement is performed using substantially the same conditions and method as the time-resolved PL measurement described above, and an ND filter is used to investigate the behavior of the PL lifetime in a wide range of weak excitation intensity, and the pulse energy density per unit area is measured. Was changed from 1.6 nJ / cm ² to 1.6 μJ / cm ² .

Comparative Example 6-1 shows a short PL life of about 1 nsec over the entire measurement range, whereas Example 6-1 has a maximum value of 0.16 μJ / cm ² and lower excitation light intensity. There is a tendency for the PL life to slightly decrease. However, Example 6-1 still showed a long PL life of 2 nsec or longer. It was shown that the optical quality of Example 6-1 was good because it exhibited such a long PL lifetime even under weak excitation conditions where the influence of the non-radiative recombination process was strong against the excited excess carriers. .

(Example 7)
In Example 7-1, the cathode luminescence (CL) spectrum mapping of the MQW structure having the MQW structure of the low V / III ratio similar to that of Example 3-4 and the uppermost barrier layer being a GaN barrier layer is shown in FIG. And plotted in FIG.

  FIG. 24 is a diagram showing the half width of the CL spectrum obtained at each point in the plane, and FIG. 25 is a diagram showing the peak wavelength of the CL spectrum obtained at each point in the plane. The measurement was performed at room temperature using an SEM-CL apparatus, the acceleration voltage was 5 kV, the beam current value was 1 nA, the SEM observation magnification was 20000 times, the measurement interval was 50 nm pitch, and 80 × 80 points were measured.

  Comparative Example 7-1 has a structure with a high V / III ratio similar to that of Comparative Example 3-1, but uses a substrate with a 0 ° OFF angle and a structure in which the quantum well layer is a single layer. CL mapping was plotted in FIGS. 26 and 27.

  FIG. 26 is a diagram showing the half width of the CL spectrum obtained at each point in the plane, and FIG. 27 is a diagram showing the peak wavelength of the CL spectrum obtained at each point in the plane. The measurement was performed using the same conditions and method as in Example 6.

  As can be seen from FIGS. 24 to 27, in Example 7-1, the in-plane peak wavelength is uniform and the CL half-value width is narrow and uniform, whereas in Comparative Example 7-1, the peak wavelength distribution is uniform. Large, full width at half maximum, and partly has a very wide half width. Further, the half-value width distribution correlates with the wavelength distribution, but the half-value width distribution is larger. From the above, in Example 6-1 a uniform and sharp spectrum was obtained even in a very small region, whereas in Comparative Example 6-2, the wavelength was also non-uniform, but more than half of that. The value range is non-uniform on a submicron scale, and there is a very wide area in some areas.

  FIG. 28 is a CL emission spectrum obtained at points 1 to 3 illustrated in FIG. As can be seen from FIG. 28, Comparative Example 7-1 shows a substantially symmetrical Gaussian emission spectrum at the point where the half-value width is narrow, whereas the point where the half-value width is wide shows a shoulder on the long wave side on the spectrum. It was found that the full width at half maximum was widened. From the above, it was found that the reason why the full width at half maximum was increased in Comparative Example 7-1 was that a low energy (long wave) emission peak appeared locally.

  FIG. 29 is a CL emission spectrum obtained at points 1 to 3 illustrated in FIG. As can be seen from FIG. 29, in Example 7-1, a highly uniform emission characteristic showing a Gaussian-shaped emission spectrum that is narrower than the narrowest spectrum of Comparative Example 7-1 at any point is obtained. It was.

(Example 8)
Example 8-1 has a low V / III ratio MQW structure similar to that of Example 3-4, and after the first InGaN quantum well layer was grown, the subsequent GaN barrier layer was not grown, and was left as it was. The one with lowered temperature is used. This surface image is shown in FIG. The measurement was performed on 2 μm square using an AFM apparatus. The left side in the figure is an AFM image, and the right side is a phase image.

  Comparative Example 8-1 has a high V / III ratio MQW structure similar to Comparative Example 3-1, and after growing an InGaN quantum well layer as in Example 7-1, a subsequent GaN barrier layer was grown. FIG. 31 shows a surface image when the temperature is lowered as it is. The measurement was performed on 2 μm square using an AFM apparatus. The left side in the figure is an AFM image, and the right side is a phase image.

  As can be seen from FIG. 30 and FIG. 31, in Example 8-1, a large number of convex portions were generated on the surface, and the base was bunched in steps, whereas in Comparative Example 8-1, the surface was flat. It was flat at the atomic level.

  The convex portion of Example 7-1 is considered to be a solidified group III metal droplet mainly composed of In. Since the V / III ratio is low, the incorporation of In is reduced, and what remains as metal on the surface seems to have solidified as the substrate temperature decreases, but was it a droplet during growth or the entire surface? It is unclear whether it was covered. It has been found that these droplets re-evaporate and disappear by raising the substrate temperature during the growth of the barrier layer and providing a growth waiting time. Therefore, these droplets have little influence on actual device characteristics.

Example 9
In Example 9, the temperature dependence of the PL lifetime of the MQW structure is plotted in FIG.

  Example 9-1 has an MQW structure with a low V / III ratio similar to that of Example 3-4, and this substrate was put into a cryostat, and the sample temperature dependence of the PL lifetime was measured. The temperature was adjusted from 2.3K to 300K using a refrigerator and a heater. The PL life measurement conditions are the same as described above.

  Comparative Example 9-2 had an MQW structure with a high V / III ratio similar to that of Comparative Example 3-1, and the same measurement as in Example 8-1 was performed.

  As can be seen from FIG. 32, in Example 8-1, the PL life was 0.77 nsec at a very low temperature (2.3 K), but the PL life increased as the sample temperature increased, and the PL life was 300 K (room temperature). 1.87 nsec.

  On the other hand, in Comparative Example 9-1, the PL life at a very low temperature is 0.62 nsec, which is shorter than that in Example 8-1, but the PL life hardly changes with increasing temperature, and the PL life is also at room temperature. Is as short as 0.79 nsec.

  Considering the above, it is considered that the influence of the non-radiative recombination process is small at extremely low temperatures, and the PL lifetime is mainly determined by the radiative recombination lifetime. Therefore, although the PL lifetime is extremely short, the probability of excess carriers at the bottom of the Γ point in the k space decreases due to the influence of phonon scattering as the temperature rises, so that the emission lifetime is prolonged.

  On the other hand, the influence of the non-radiative recombination process, which had a small effect at extremely low temperatures, increases with increasing temperature. As a result, as the sample temperature rises, the dominant factor of PL lifetime gradually shifts to the non-radiative recombination process.

  In Example 8-1, since the influence of the non-radiative recombination process was small, it was considered that the PL life was extended mainly due to the influence of the luminescent recombination process with an increase in temperature, whereas in Comparative Example 8-2, It is thought that the influence of the non-radiative recombination process increases with the temperature rise, and the PL lifetime becomes short and constant.

(Discussion)
The reason why the output of Example 1 and Example 2 was higher than that of Comparative Example 1 was that the quality of the InGaN quantum well layer was improved by changing to a low V / III ratio condition during the growth of the InGaN quantum well layer. . The light emission on the long wave side was suppressed by the low V / III ratio growth condition, the multi-peaks seen in the emission spectrum became a single peak, and the luminous efficiency was dramatically improved.

  In particular, in the wavelength dependence of the conventional light emission output, the light emission output suddenly decreased when the wavelength exceeded the purple wavelength (405 nm). As can be seen from the result of 420 nm in Example 1, in the present invention, the light emission output is 405 nm to 420 nm. It is noteworthy that there is no reduction in output at all.

  Although a sufficiently high light emission output can be obtained from Example 2 without the interface buffer layer, the wavelength dependency of the light emission output of Example 1 is interpolated and compared with Example 2 at the same wavelength. It is estimated that the higher output is obtained. This can be said to be a structure suitable for higher output due to the presence of the interface strain buffer layer.

  Although the multi-peak feature has been described in the background section, the original light emission is light emission on the short wave side, and light emission on the long wave side is derivative. However, when the excitation density is low, carriers flow into the emission level on the long wave side with a small band gap, and the emission on the long wave side becomes the main.

  In the long wave side light emission, the wavelength difference from the original light emission wavelength is determined corresponding to the off angle of the substrate. Moreover, by emitting light in both modes, the full width at half maximum of the PL and EL emission spectra increases.

  In addition, since the emission efficiency of the long wave side light emission is relatively lowered with respect to the light emission from the original InGaN quantum well due to the increase of the carrier density (current), the influence of the original light emission wavelength is increased by the current injection. As a result, the emission wavelength is shortened.

  Regarding the emission wavelength of Comparative Example 1, what is seen as the shoulder on the short wave side is the emission from the original InGaN quantum well, which is substantially the same as the peak wavelength of the spectrum of Example 1. Therefore, Comparative Example 1 and Example 1 have the same peak wavelength when high current is injected.

  In such a situation where a specific long-wave side emission spectrum is generated, the PL lifetime corresponding to the increase of the point defect density and the increase of the non-radiative recombination centers is decreased, and the light emission efficiency is decreased.

  In summary, the generation of long-wave side light emission increases the half-value width of the emission spectrum, decreases the emission efficiency, increases the point defect density and non-radiative recombination centers and decreases the PL lifetime, and changes in peak wavelength when the carrier density is changed. It has to do with everything.

  These various physical properties only represent one essence from another angle. Accordingly, the light emission quality can be reflected by any index.

  In the results of Example 3, the PL half-value widths of Comparative Examples 3-1 and 3-2 are wide. Among them, Comparative Example 3-1 has a wider half width. This is because, as described above, when half-wave side light emission occurs, the full width at half maximum widens, but the energy difference from the peak wavelength of the original light emission spectrum differs depending on the off angle of the substrate. In the 0 ° OFF substrate of Comparative Example 3-1, the long wave side light emission is far from the original light emission of the InGaN quantum well, and in the Comparative Example 3-2, the long wave side light emission is closer to the original light emission.

  However, in any case, when the half-value width is wider than the straight line in the figure, it can be said that multi-peaks have occurred. During PL evaluation, for example, by setting the wavelength resolution low, depending on the measurement conditions, the multi-peaks may not be seen separately. However, if the half-value width is greater than this straight line, multi-peaks will occur. And the characteristics deteriorate.

  From Example 3-1 and Example 3-2, it is clear that the PL half-value width can be narrowed by lowering the V / III ratio during the growth of the InGaN quantum well layer. In Example 3-3 and Example 3-4, a 2 ″ substrate having a different manufacturing method is used as the substrate. However, there is no change in the wavelength dependency of the half width, in other words, the manufacturing method of the substrate as long as it is an m-plane. It can be said that it does not depend on.

Further, in Example 3-5, NH ₃ is slightly increased and a slightly higher V / III ratio of 3340 is adopted. However, although the PL half-value width is slightly increased there, a sufficiently narrow spectrum is obtained. Has been obtained. However, it is known that when the V / III ratio is further increased, long-wave side light emission occurs rapidly, and the half-value width of the PL spectrum increases.

On the contrary, an experiment was conducted on the lower limit where NH ₃ was reduced to lower the V / III ratio. In this case, the incorporation of In into the crystal deteriorates, and it is difficult to increase the wave length and In droplets are easily formed on the substrate surface. However, even if In droplets were generated, they could be easily removed with an HCl solution, and the overall crystallinity was good.

  Further, in Example 3-6, the In mole supply ratio was lowered, but a large difference was not seen in this range (74-81%), and the half width remained narrow.

  As for the results of Example 4, a long PL life was obtained in all of Examples 4-1 to 4-6 having a low V / III ratio. The PL lifetime at room temperature is governed by non-radiative recombination, and the longer the PL lifetime, the smaller the effect of defects. In Comparative Example 4-1 and Comparative Example 4-2 having the same structure as Comparative Example 1 and Comparative Example 2, the occurrence of multi-peaks is observed. In such a situation, the quality deteriorates without exception. Is also shortened.

  On the other hand, Comparative Example 4-3 increases the growth rate although it is as low as 2740 when only the V / III ratio is observed. The V / III ratio dominates the In-plane quality of the m-plane, but if it is cleared, all other conditions are not allowed, and the growth rate and In mole supply ratio are within a certain range. It can be said that it is necessary to use an appropriate value.

  The contents shown in Example 5 are the same as the current value dependency and mechanism of the peak wavelength of the EL emission spectrum described in Example 1. The deterioration in the quality of the m-plane which is a problem here is accompanied by the generation of long-wave side light emission. When this light emission occurs, this long-wave side light emission becomes the main at a low carrier concentration. When the carrier density becomes high, the main peak wavelength shifts to the light emission on the short wave side, which is the original emission center, so that the dependency of the emission peak wavelength on the carrier density increases.

  In Example 5-1 and Example 5-2, in which only the original light emission from the InGaN quantum well is seen with good quality, the emission wavelength is single and the wavelength change is small even if the PL excitation intensity is changed. In Comparative Example 5-1, in which side light emission occurs, the long wave side light emission becomes main on the low excitation side, and shifts to the original short wave side light emission by increasing the excitation intensity, so the wavelength varies.

  In Example 7-1 in which only the original light emission from the InGaN quantum well was observed with good quality, the emission spectrum from the original InGaN was obtained in the entire in-plane region, and both the wavelength and the half width were uniform. However, in Comparative Example 7-1 in which long-wave side light emission occurred, long-wave side peaks were partially generated. For this reason, a region having a wide half-value width is locally observed. The origin of these well-written long-wave emissions is not clear, but it is clear that this situation is closely related to the situation where optical quality is degraded.

  From Example 8-1 and Comparative Example 8-1, it was found that the surface morphology was flatter in Comparative Example 8-1 in which the optical quality was lowered. That is, the occurrence of the long wave peak is considered to be an abnormality at the atomic level regardless of the surface roughness and cannot be detected by the AFM.

  From Example 9-1 and Comparative Example 9-1, the temperature characteristics of the PL life reflecting the optical quality are obtained. In general, it is thought that in-plane diffusion of carriers is likely to occur as the sample temperature increases, and the influence of non-radiative recombination centers increases. In Comparative Example 8-1, the influence of the non-radiative recombination center is strongly observed. On the other hand, in Example 8-1, the PL lifetime is affected by the influence of phonon scattering as the temperature rises, which is an ideal physical phenomenon. There was a situation that became longer.

DESCRIPTION OF SYMBOLS 1 ... Epitaxial wafer, 2 ... m-plane GaN substrate, 3 ... 1st undoped GaN layer, 4 ... n-type GaN contact layer, 5 ... AlGaN layer, 6 ... 2nd undoped GaN layer, 7 ...... Light emitting layer, 7A.Quantum well layer, 7B..Barrier layer, 7C..Interfacial strain buffer layer, 8 .... p-type AlGaN cladding layer, 9 .... p-type GaN contact layer, 10 .... p-type InGaN Contact layer, 11 ... n-side metal electrode, 12 ... p-side contact electrode, 13 ... p-side metal electrode, 20 ... semiconductor light emitting element, 21 ... package, 22 ... translucent material, 23 ... Wavelength converter, 30 ... Light emitting device

Claims (16)

a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane;
An n-type conductive layer formed on the main surface on one side of the GaN substrate;
A light emitting layer formed on the main surface on one side of the n-type conductive layer;
With
The PL peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less,
An epitaxial wafer characterized in that a PL half-value width of the light emitting layer satisfies a conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm) 2. The epitaxial wafer according to claim 1, wherein a PL lifetime of the light emitting layer at a room temperature of 25 ° C. is 1.3 nsec or more and 20 nsec or less. 3. The epitaxial wafer according to claim 1, wherein a fluctuation in the PL peak wavelength when the excitation light intensity of the light emitting layer is changed 1000 times is 0 nm or more and 10 nm or less. The epitaxial wafer according to any one of claims 1 to 3, wherein the light emitting layer includes an InGaN layer. The light emitting layer comprises a multiple quantum well layer in which quantum well layers and barrier layers are alternately stacked,
An interface strain buffer that has an intermediate lattice constant between the quantum well layer and the barrier layer at an interface on at least one side between the quantum well layer and the barrier layer, and buffers strain between the quantum well layer and the barrier layer. The epitaxial wafer according to claim 1, further comprising: a layer. The epitaxial wafer according to claim 1, wherein a dark spot density of the GaN substrate is 2 × 10 ⁸ cm ⁻² or less. The epitaxial wafer according to claim 1, wherein a main surface on the other side of the GaN substrate is roughened. an n-type conductive layer made of a GaN-based semiconductor layer formed on one main surface of a GaN substrate having a main surface with an off angle of 0 ° to 10 ° with respect to the m-plane;
A light emitting layer formed on the main surface on one side of the n-type conductive layer;
A p-type conductive layer formed on the main surface on one side of the light emitting layer;
With
The emission peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less,
The PL half width of the light emitting layer satisfies the conditional expression (1).
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm) an n-type conductive layer made of a GaN-based semiconductor layer formed on one main surface of a GaN substrate having a main surface with an off angle of 0 ° to 10 ° with respect to the m-plane;
A light emitting layer formed on the main surface on one side of the n-type conductive layer;
A p-type conductive layer formed on the main surface on one side of the light emitting layer;
Have
The PL peak wavelength of the light emitting layer is 410 nm or more and 460 nm or less,
The PL half width of the light emitting layer is a semiconductor light emitting element that satisfies the conditional expression (1),
A wavelength converting substance that absorbs at least part of the light emitted by the semiconductor light emitting element and converts it into light having a longer wavelength;
A light emitting device comprising:
Δl ≦ L × 0.4−150 (1)
L: PL peak wavelength (unit: nm) Δl: PL half-value width (unit: nm) a first step of growing an n-type conductive layer made of a GaN-based semiconductor layer on a main surface on one side of a GaN substrate whose main surface is a surface having an off angle of 0 ° to 10 ° with respect to the m-plane;
A second step of growing a light emitting layer on a main surface on one side of the n-type conductive layer grown in the first step;
With
In the second step, the Group V material and the Group III material are supplied so that the V / III ratio, which is the ratio of the mole supply amount of the Group V material to the mole supply amount of the Group III material, is 500 or more and 4000 or less. An epitaxial wafer manufacturing method characterized by the above. In the second step, a multiple quantum well layer in which quantum well layers and barrier layers are alternately stacked is grown as the light emitting layer, and the V / III ratio is 500 or more and 4000 or less when the quantum well layer is grown. The method for producing an epitaxial wafer according to claim 10, wherein the Group V raw material and the Group III raw material are supplied so that The epitaxial wafer manufacturing method according to claim 11, wherein in the second step, a quantum well layer made of an InGaN layer is grown as the quantum well layer. In the second step, when the quantum well layer composed of the InGaN layer is grown, the group III material is added such that the ratio of the supply amount of the indium material to the total supply amount of the group III material is 50% or more and 90% or less. The method for producing an epitaxial wafer according to claim 12, wherein the epitaxial wafer is supplied. The epitaxial wafer manufacturing method according to claim 10, wherein, in the second step, the quantum well layer is grown at a growth rate of 1 nm / min or more and 8 nm / min. The epitaxial wafer manufacturing method according to any one of claims 10 to 14, further comprising a third step of heat-treating the light emitting layer in the atmosphere. The epitaxial wafer manufacturing method according to claim 10, wherein, in the first and second steps, the n-type conductive layer and the light emitting layer are grown by MOCVD.