JP6649324B2 - Nitride semiconductor ultraviolet light emitting device and method of manufacturing nitride semiconductor ultraviolet light emitting device - Google Patents

Description

  The present invention relates to a nitride semiconductor light emitting device having an AlGaN-based active layer formed on a substrate made of sapphire, and more particularly to a nitride semiconductor ultraviolet light emitting device having a peak emission wavelength in an ultraviolet region.

  Conventionally, a nitride semiconductor blue light-emitting device using a GaN-based nitride semiconductor for an active layer has been widely used. However, a nitride semiconductor ultraviolet light emitting device using an AlGaN-based nitride semiconductor having an emission wavelength shorter than that for the active layer has not yet been widely used.

  One reason for this is that, in an AlGaN-based nitride semiconductor mainly composed of a mixed crystal of AlN and GaN, the bonding force between Al and N in AlN is extremely large as compared with the bonding force between Ga and N in GaN. It is difficult to grow a high-quality crystal due to circumstances. In particular, even if an already established method of growing a GaN-based nitride semiconductor is adopted as a method of growing an AlGaN-based nitride semiconductor, a high-quality crystal comparable to GaN cannot be produced. It becomes a problem.

Therefore, various methods for improving the crystallinity of an AlGaN-based nitride semiconductor have been proposed in Patent Documents 1 to 4, for example. Specifically, in Patent Literature 1, by forming a buffer layer on a sapphire substrate from a plurality of paired layers in which AlN and GaN are alternately stacked, cracks generated in AlN are suppressed from extending upward. Has been proposed. Further, in Patent Document 2, when a buffer layer made of AlN is formed on a sapphire substrate, a period for supplying a pulse of NH ₃ as a source gas of N is provided to locally reduce the growth rate of the AlN layer. A method for reducing threading dislocations has been proposed. Patent Document 3 proposes a method of reducing threading dislocations by covering an AlGaN island nucleus formed on a sapphire substrate with an AlGaN buffer layer having a higher Al composition ratio than the nucleus. I have. Patent Document 4 proposes a method for improving the crystallinity of an AlGaN-based nitride semiconductor formed above a sapphire substrate by optimizing the off-angle of the substrate.

JP 2006-66641 A JP 2009-54780 A JP 2013-222746 A WO 2013/021464 pamphlet

  The methods proposed in Patent Documents 1 to 4 described above all improve the crystallinity of the nitride semiconductor layer formed thereon by optimizing the buffer layer formed on the substrate or the substrate surface. It is intended.

  Certainly, by optimizing the substrate and the buffer layer, which are the starting points of the crystal growth, it is possible to expect improvement in the crystallinity of the nitride semiconductor layer thereabove. However, only the defects such as cracks and threading dislocations, which propagate from the base to the respective layers above, such as cracks and threading dislocations, which can be expected to be improved by this method, occur in the entire device. Therefore, when these methods are adopted, the active layer in which light emission occurs is not always optimized. Therefore, in the nitride semiconductor ultraviolet light emitting device obtained by adopting the methods proposed in Patent Documents 1 to 4, the light output is not always improved, which causes a problem.

  Therefore, an object of the present invention is to provide a nitride semiconductor ultraviolet light emitting device having an active layer with good light output.

In order to achieve the above object, the present invention provides a substrate made of sapphire having a surface on which a multi-step terrace is formed by being inclined with respect to a (0001) plane, and an AlN layer formed on the surface of the substrate And a light-emitting portion formed on the surface of the base portion and including an active layer having an AlGaN-based semiconductor layer, wherein at least the AlN layer of the base portion and the light-emitting portion The active layer and each layer between them are formed by step flow growth in which two-dimensional growth is performed by growing side surfaces of a multi-stage terrace, and the active layer is formed of Al _X Ga _1-X N (0 <X < 1) having a quantum well structure including at least one well layer, wherein the average roughness in a 25 μm square region on the surface of the active layer is not less than the thickness of the well layer and not more than 10 nm. To provide a nitride semiconductor ultraviolet light emitting device characterized Rukoto.

  According to this nitride semiconductor ultraviolet light emitting device, the segregation of Ga in the active layer (particularly, the well layer) is performed by the step flow growth such that the average roughness on the surface of the active layer is not less than the thickness of the well layer and not more than 10 nm. , The light output of the active layer can be increased.

  Further, in the nitride semiconductor ultraviolet light emitting device having the above characteristics, it is preferable that the average roughness in a 25 μm square region is 3 nm or more on the surface of the active layer included in the light emitting portion.

  According to this nitride semiconductor ultraviolet light emitting device, Ga segregation sufficiently occurs in the active layer (particularly, the well layer). Therefore, the light output of the active layer can be sufficiently increased.

  In the nitride semiconductor ultraviolet light emitting device having the above characteristics, it is preferable that the average roughness in a 25 μm square region is 6 nm or less on the surface of the active layer included in the light emitting portion.

  According to this nitride semiconductor ultraviolet light emitting device, the light output of the active layer can be significantly increased.

  In the nitride semiconductor ultraviolet light emitting device having the above characteristics, an average roughness in a 25 μm square region on a surface of a layer included in the light emitting portion and formed immediately before the active layer has a thickness of the well layer. It is preferable that the thickness is not less than 10 nm. Further, in the nitride semiconductor ultraviolet light emitting device having the above characteristics, the difference between the average roughness in the 25 μm square region on each surface of the active layer and the layer formed immediately before the active layer included in the light emitting part is obtained. It is preferable that the ratio obtained by dividing the absolute value by the average roughness in a 25 μm square area on the surface of the active layer is 10% or less.

  According to these nitride semiconductor ultraviolet light emitting devices, the active layer is formed by uniform growth while maintaining the average roughness on the growth surface, so that segregation of Ga in the active layer (particularly, well layer) is ensured. Can be caused. Therefore, the light output of the active layer can be increased.

  In the nitride semiconductor ultraviolet light emitting device having the above characteristics, an average width of a terrace in a tilt direction of the substrate is 0.3 μm or more and 1 μm on a surface of the AlN layer included in the base portion when viewed from above. The following is preferable. In the nitride semiconductor ultraviolet light emitting device having the above characteristics, it is preferable that an average height of a step formed by a terrace is 8 nm or more and 14 nm or less on the surface of the AlN layer included in the base portion.

  According to these nitride semiconductor ultraviolet light emitting devices, by forming the light emitting portion on the surface of the underlying portion having the AlN layer, Ga segregation occurs due to the step flow growth as described above, so that the light output is good. Can be obtained.

  In the nitride semiconductor ultraviolet light emitting device having the above characteristics, the frequency distribution of the height in a 25 μm square region on the surface of the active layer included in the light emitting portion is convex downward as the height increases from 0. It is preferable that the curve becomes monotonically increasing while taking a maximum value while changing from the curve to an upwardly convex curve, and then becomes monotonically decreasing while changing from an upwardly convex curve to a downwardly convex curve.

  According to this nitride semiconductor ultraviolet light emitting device, since the influence of hillocks is small and the step flow growth is dominant, Ga segregation occurs and an active layer with good light output can be obtained.

  In the nitride semiconductor ultraviolet light emitting device having the above characteristics, it is preferable that the peak emission wavelength is 230 nm or more and 340 nm or less.

  According to this nitride semiconductor ultraviolet light emitting device, Ga segregation is sufficiently generated in the active layer (particularly, the well layer) to increase the light output, and at the same time, the shape of the light emission spectrum is broken or the light emission intensity is reduced. And other problems can be prevented.

  According to the nitride semiconductor ultraviolet light emitting device having the above characteristics, Ga in the active layer (particularly, the well layer) is formed by the step flow growth such that the average roughness on the surface of the active layer is not less than the thickness of the well layer and not more than 10 nm. By causing the segregation, the light output of the active layer can be increased. Therefore, it is possible to obtain a nitride semiconductor ultraviolet light emitting device having an active layer with a good light output.

FIG. 1 is a cross-sectional view of a principal part schematically showing an example of a structure of a light emitting device according to an embodiment of the present invention. The perspective view which showed typically the state of the surface of the off-substrate expanded to the atomic level. FIG. 4 is a plan view schematically showing a light emitting element viewed from a p-electrode and an n-electrode side. 9 is an AFM image showing the state of the surface of the active layer of Samples 1 to 5. 9 is a histogram showing a frequency distribution of heights on the surfaces of the active layers of Samples 1 to 5. FIG. 4 is a diagram showing a comparison between respective surface states of an n-type cladding layer and an active layer in Sample 1. FIG. 9 is a diagram showing a comparison between respective surface states of an n-type cladding layer and an active layer in Sample 3. FIG. 9 is a diagram showing a comparison between respective surface states of an n-type cladding layer and an active layer in Sample 4. 7 is a graph showing light emission characteristics (light output) of Samples 1 to 5. 7 is a graph showing emission characteristics (peak emission wavelength, half width) of Samples 1 to 5. 6 is an AFM image showing a state of a surface of an AlN layer in a base portion of Sample 1; 6 is an AFM image showing a state of a surface of an AlN layer in a base portion of a sample A. 7 is an AFM image showing a state of a surface of an AlN layer in a base portion of Sample 2. 9 is an AFM image showing a state of a surface of an AlN layer in a base portion of Sample 3. 9 is an AFM image showing the state of the surface of the AlN layer in the base portion of Sample 4. 9 is an AFM image showing a state of a surface of an AlN layer in a base portion of Sample 5. FIG. 4 is a spectrum diagram showing emission spectra of a plurality of samples having different AlN mole fractions (peak emission wavelengths) of well layers. 5 is a graph showing the relationship between the peak emission wavelength and the half width of each sample.

  An embodiment of a nitride semiconductor ultraviolet light emitting device (hereinafter, simply referred to as “light emitting device”) according to the present invention will be described with reference to the drawings. The embodiment described below is merely one embodiment in which the light emitting element according to the present invention is implemented as a light emitting diode, and the present invention is not limited to the following embodiment. For example, the light-emitting device according to the present invention can be implemented as another light-emitting device such as a semiconductor laser, or can be implemented as a light-emitting diode having a different configuration from the following embodiments.

<Structural example of light emitting element>
First, an example of a structure of a light emitting device according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a cross-sectional view schematically showing a main part of an example of the structure of a light emitting device according to an embodiment of the present invention. In FIG. 1, in order to facilitate understanding of the description, the main part is emphasized and the invention is schematically shown, and thus the dimensional ratio of each part is not always the same as the actual element. Absent.

  As shown in FIG. 1, the light emitting element 1 according to the embodiment of the present invention includes a base portion 10 and a surface of the base portion 10 (an upper surface in FIG. 1. A lower surface in FIG. The light-emitting portion 20 is formed on the “back surface”; the same applies hereinafter), and the p-electrode 30 and the n-electrode 40 for supplying power to the light-emitting element 1 (particularly, the light-emitting portion 20). The base part 10 is a part corresponding to a base for forming the light emitting part 20. The light emitting section 20 is a portion configured by various layers necessary for light emission.

  The underlayer 10 includes a substrate 11 made of sapphire, and an AlN layer 12 made of AlN formed on the surface of the substrate 11. The AlN layer 12 is obtained by epitaxially growing AlN on the surface of the substrate 11 at a high temperature of about 1150 to 1300 ° C. In the light-emitting element 1 shown in FIG. 1, the underlayer 10 is illustrated as if only the AlN layer 12 was provided on the substrate 11, but if necessary, layers other than the AlN layer 12 may be provided. Is also good. For example, an AlGaN layer made of AlGaN or n-type AlGaN may be added on the AlN layer 12 (below the light emitting unit 20).

  The light emitting section 20 includes an n-type cladding layer 21 made of n-type AlGaN, an active layer 22, a p-type cladding layer 23 made of p-type AlGaN, and a p-type GaN And a p-type contact layer 24 configured. The active layer 22, the p-type cladding layer 23, and the p-type contact layer 24 formed above the n-type cladding layer 21 are partially removed by reactive ion etching or the like, so that the n-type cladding layer 21 is removed. Some surfaces are exposed. Therefore, above the partial region (first region R1) on the n-type cladding layer 21, the region from the active layer 22 to the p-type contact layer 24 is formed.

  The p-electrode 30 is made of, for example, Ni / Au and is formed on the surface of the p-type contact layer 24. The n-electrode 40 is made of, for example, Ti / Al / Ti / Au, and is formed on a part of the surface of a region other than the first region R1 (the second region R2) of the n-type cladding layer 21.

  In the light emitting element 1, an off-substrate whose surface is inclined at a small angle with respect to the (0001) plane of sapphire is used as the substrate 11. Here, the state of the surface of the off-substrate will be described with reference to the drawings. FIG. 2 is a perspective view schematically showing the state of the surface of the off-substrate enlarged to the atomic level. In FIG. 2, in order to facilitate understanding of the description, the essential parts are emphasized and the invention is schematically illustrated, and the dimensional ratio of each part is not always the same as the actual element. Absent.

  As shown in FIG. 2, a multi-stage terrace T is formed on the surface of the substrate 11 which is an off-substrate. This is because, when a bulk single crystal of sapphire is cut out at an angle slightly inclined with respect to the (0001) plane (that is, off angle θ), the (0001) plane appears along the cutting direction. . Note that the magnitude of the off angle θ and the direction in which the off angle is provided (specifically, the direction in which the (0001) plane is inclined, for example, the m-axis direction or the a-axis direction) are determined as desired in each layer on the substrate 11. It may be arbitrarily determined as long as growth is realized.

The active layer 22 _has a Al _{X Ga 1-X} N comprising at least one quantum well structure of the well layer 22a composed of (0 <X <1). The thickness of the well layer 22a is such that a quantum size effect (quantum confinement effect) is exhibited, and is, for example, 10 nm or less. In a typical quantum well structure, the well layer 22a is sandwiched between barrier layers 22b having a larger band gap than the well layer 22a. For example, when the well layer 22a is made of AlGaN, the barrier layer 22b is made of AlGaN or n-type AlGaN having an AlN mole fraction larger than that of the well layer 22a. The quantum well structure provided in the active layer 22 may be a single quantum well structure composed of only one quantum well structure, or a multiple quantum well structure in which a plurality of quantum well structures are stacked. Good. Further, for example, the thickness of the well layer 22a is 2 nm or more and 3 nm or less, and the thickness of the barrier layer 22b is 6 nm or more and 8 nm or less.

  Further, the active layer 22 includes an electron block layer 22c made of p-type AlGaN having an AlN mole fraction larger than that of the well layer 22a and the barrier layer 22b at an interface (outermost surface) in contact with the p-type cladding layer 23. . The electron blocking layer 22 c is a layer that suppresses electrons injected into the active layer 22 from entering the p-type cladding layer 23. For example, the film thickness of the electron block layer 22c is 15 nm or more and 30 nm or less, and is typically 20 nm.

  In the active layer 22, the barrier layer 22b and the electron block layer 22c are not essential components. However, by providing the barrier layer 22b and the electron blocking layer 22c, many electrons and holes can be confined in the well layer 22a, and the electrons and holes can be efficiently recombined (that is, emitted light). Is preferred.

  AlGaN constituting each layer described above is formed by a well-known epitaxial growth method such as a metal organic compound vapor phase epitaxy (MOVPE) method or a molecular beam epitaxy (MBE) method. In each of the above-described layers, the n-type layer is doped with, for example, Si as a donor impurity. In each of the above-mentioned layers, for example, Mg is added to the p-type layer as an acceptor impurity. In each of the above-described layers, a layer composed of AlN and AlGaN whose conductivity type is not specified is an undoped layer to which no impurity is added. Further, AlGaN, AlN and GaN constituting each layer described above may partially or entirely contain other elements as long as their properties do not greatly change (for example, even if AlGaN contains a trace amount of In). Or AlN may contain a small amount of Ga).

  The AlN molar fraction in the n-type cladding layer 21, the barrier layer 22b, and the p-type cladding layer 23 is, for example, 30% or more and 80% or less (more preferably 50% or more and 80% or less, and still more preferably 55% or more and 80% or less). The AlN molar fraction of the well layer 22a is, for example, 5% or more and 80% or less (more preferably, 5% or more and 60% or less). However, the AlN mole fraction of the well layer 22a is smaller than the AlN mole fraction of the n-type cladding layer 21, the barrier layer 22b, and the p-type cladding layer 23 (conversely, the GaN mole fraction of the well layer 22a). Are larger than these layers). When an AlGaN layer is added to the underlayer 10, the AlN mole fraction in the AlGaN layer may be set to a size within the same range as the n-type cladding layer 21, the barrier layer 22b, and the p-type cladding layer 23. .

  For example, the light emitting element 1 has a peak emission wavelength of 230 nm or more and 350 nm or less. Further, for example, the light emitting element 1 is a back emission type light emitting element in which light emitted from the active layer 22 is extracted from the substrate 11 side. When an AlGaN layer is added to the underlayer 10, the AlN mole fraction of the AlGaN layer is set to be larger than the AlN mole fraction of the well layer 22a. In this case, the AlN mole fraction of the AlGaN layer may be set to be the same as the AlN mole fraction of the n-type cladding layer 21, or may be larger than the AlN mole fraction of the n-type cladding layer 21. May be set as follows.

  The thickness of each AlGaN constituting each layer except the active layer 22 in the light emitting section 20 is, for example, 2000 nm or more and 4000 nm or less for the n-type cladding layer 21, 500 nm or more and 600 nm or less for the p-type cladding layer 23, and the p-type contact layer 24. Is 100 nm or more and 300 nm or less. In the underlayer 10, for example, the thickness of the AlN layer 12 is 1500 nm or more and 4000 nm or less. In addition, when an AlGaN layer is added to the underlayer 10, the thickness of the AlGaN layer is, for example, 200 nm or more and 300 nm or less.

  Next, the p-electrode 30 and the n-electrode 40 will be described with reference to the drawings. FIG. 3 is a plan view schematically showing the light emitting element viewed from the p-electrode and n-electrode sides.

As shown in FIG. 3, the p-electrode 30 is formed on almost the entire surface of the first region R1, and the n-electrode 40 is formed on almost the entire surface of the second region R2. The chip size of the light emitting element 1 is 800 μm in each of the vertical and horizontal directions, and the area of the first region R1 in which the p-electrode 30 is formed is about 168000 μm ² .

  Each of the p-electrode 30 and the n-electrode 40 is formed, for example, by forming a photoresist to be an inverted pattern of the electrode (a pattern covering a surface other than the position where the electrode is formed) and then forming a multilayer metal by an electron beam evaporation method or the like. It is formed by depositing a film and removing the photoresist and the multilayer metal film on the photoresist by lift-off. After forming one or both of the p-electrode 30 and the n-electrode 40, a heat treatment may be performed as necessary by RTA (instantaneous thermal annealing) or the like.

<Conditions for Obtaining Active Layer with Good Light Output>
[Surface roughness of active layer]
According to ordinary knowledge in the field of semiconductor light emitting devices, in order to obtain an active layer 22 having a good light output, it is necessary to introduce defects such as dislocations and cracks that hinder light emission during crystal growth of the active layer 22. Suppressed growth, that is, two-dimensional growth (growth in which layers are carefully stacked one by one) while maintaining the flatness of the growth surface should be performed. In particular, in order to uniformly grow the extremely thin well layer 22a having a thickness of about several nm while suppressing the introduction of defects, the roughness on the growth surface should be at least smaller than the thickness of the well layer 22a. Should be suppressed.

  However, as a result of intensive studies, the applicant of the present application has found that, during the crystal growth of the active layer 22 (particularly, the well layer 22a), the light output is good by performing the growth while maintaining the growth surface with a large roughness. Active layer 22 was obtained. Hereinafter, this point will be described with reference to the drawings.

  The roughness of the growth surface during crystal growth of the active layer 22 can be expressed using the average roughness of the surface of the active layer 22 after growth. The average roughness is, for example, a height measured by an atomic force microscope (AFM: Atomic Force Microscope) (relative height with a predetermined height in a measurement region of an object to be measured being 0). , Can be calculated by the following equation (1). In the following formula (1), Z (i) is the height of each point measured by AFM, Ze is the average value of height Z (i), and Ra is the average roughness.

  FIG. 4 is an AFM image showing the state of the surface of the active layer of Samples 1 to 5. In the AFM images shown in FIG. 4 and other drawings, a brighter (whiter) area indicates a higher area. FIG. 4 also shows the average roughness (hereinafter simply referred to as “average roughness”) in the illustrated 25 μm square area. FIG. 5 is a histogram showing the frequency distribution of the height on the surface of the active layer of Samples 1 to 5, and the frequency of the height obtained for each of the 25 μm square regions of Samples 1 to 5 shown in FIG. The distribution is shown.

  Here, the 25 μm square area for calculating the average roughness and the frequency distribution is the upper limit size at which the average roughness and the height can be calculated using the AFM, and is approximately constant regardless of the location in the chip. Is an area large enough to obtain the value of. Note that the numerical value of 25 μm has no special significance, and any other numerical value may be used as long as the region is as wide as 30 μm or 20 μm.

  In samples 1 to 5 shown in FIGS. 4 and 5, the influence of the terrace T (see FIG. 2) formed on the surface of the substrate 11, which is an off-substrate, is passed down to the surface of the active layer 22. Specifically, at least the AlN layer 12 of the underlayer 10, the active layer 22 of the light emitting section 20, and each of the layers (in this case, the n-type cladding layer 21) are formed on the side surfaces of the multi-step terrace. And two-dimensionally grown by step flow growth. For reference, FIGS. 11 to 16, which will be described later, show the state of the surface of the AlN layer 12 of the underlayer 10. 6 to 8, which will be described later, show the state of the surface of the n-type cladding layer 21 of the light emitting unit 20.

  Further, Samples 1 to 5 are manufactured in the order of Sample 1, Sample 2, Sample 3, Sample 4, and Sample 5 under conditions where the side surface of the terrace is easy to grow selectively. That is, among Samples 1 to 5, Sample 1 is manufactured under the condition that the side surface of the terrace is most difficult to grow selectively, and Sample 5 is manufactured under the condition that the side surface of the terrace is most likely to grow selectively. The conditions under which the side surface of the terrace is likely to be selectively grown include, for example, a condition in which the off angle of the substrate 11 is large within a certain range (for example, from 0 ° to several degrees), or a condition in which the terrace is easily exposed. Growth rate (specifically, for example, the growth rate is achieved by appropriately setting various conditions such as a growth temperature, a supply amount of a raw material and a carrier gas, and a flow rate).

  Conditions under which the side surface of the terrace is likely to be selectively grown may differ depending on the type and structure of the film forming apparatus. Therefore, the conditions may be specified by actually preparing some samples in the film forming apparatus. What is important is that the active layer 22 is formed by the step flow growth, not the conditions under which there are countless combinations depending on the film forming apparatus.

  Samples 1 to 5 were produced under different conditions as described above, and the surface state of the active layer 22, particularly the average roughness, was different. Specifically, the average roughness of the surface of each active layer 22 of Samples 1 to 5 was 10.9 nm for Sample 1, 5.08 nm for Sample 2, 3.66 nm for Sample 3, 3.94 nm for Sample 4, and Sample 5 is 5.94 nm.

  Further, in Samples 1 to 5, since the layers from the AlN layer 12 of the underlayer 10 to the active layer 22 of the light emitting unit 20 are continuously formed by the step flow growth as described above, the surface of the active layer 22 is formed. The average roughness at Specifically, the average roughness of the surface of the active layer 22 of each of the samples 1 to 5 is equal to or greater than the thickness of the well layer 22a, and is equal to or greater than 3 nm. As described above, growing the active layer 22 so that the average roughness on the surface of the active layer 22 exceeds the thickness of the well layer 22a is contrary to ordinary knowledge in the field of semiconductor light emitting devices. .

  As shown in FIG. 4, in sample 1, simultaneously with the step flow growth, three-dimensional growth in which nuclei generated at random positions on the terrace grow to form hexagonal hillocks (hillocks) H has occurred. . Further, as shown in FIG. 5, the frequency distribution of the height of the surface of the active layer 22 of the sample 1 is like a normal distribution (from a downwardly convex curve to an upwardly convex curve as the height increases from 0). Since the curve does not monotonically increase while taking inflections and takes a maximum value, and then changes from an upwardly convex curve to a downwardly convex curve and monotonically decreases, the height does not fluctuate. Be regular. Therefore, in the sample 1, the influence of the hillock H, which causes the height to fluctuate irregularly, is large, and the influence of the terrace, which causes the height to fluctuate irregularly, is small. Growth can be said to be dominant.

  As shown in FIG. 4, hillocks H are also present on the surface of the active layer 22 of the sample 2 as in the case of the sample 1. However, in the sample 2, the area occupied by the hillocks H is significantly reduced as compared with the sample 1. Further, as shown in FIG. 5, since the frequency distribution of the height on the surface of the active layer 22 of the sample 2 is different from the sample 1 in a normal distribution, the height variation becomes somewhat regular. I have. Therefore, in the sample 2, the influence of the hillock H, which causes the height to fluctuate irregularly, is small, and the influence of the terrace, which causes the height to fluctuate irregularly, is large. Growth can be said to be dominant.

  Further, as shown in FIG. 4, in Samples 3 to 5, no large hillock H was generated, and traces of step flow growth (triangular facets having a multi-stage shape) were observed over the entire surface. Further, as shown in FIG. 5, the frequency distribution of the height on the surface of the active layer 22 of the samples 3 to 5 is like a normal distribution, and the fluctuation of the height is regular. Therefore, in Samples 3 to 5, the influence of hillock H, which is a cause of irregularly varying height, is extremely small, and the influence of terrace, which is a cause of irregularly varying height, is extremely large. It can be said that step flow growth is much more dominant than step flow growth.

  Incidentally, the average roughness of the surface of the active layer 22 of the sample 1 is 10.9 nm, which is significantly larger than those of the other samples 2 to 5. This is because in Sample 1, three-dimensional growth is dominant, and the effect of hillock H is large. As described above, in the sample in which three-dimensional growth is dominant, hillocks H are inevitably included in a 25 μm square area for obtaining the average roughness set on the surface of the active layer 22. Can be greater than 10 nm.

  The average roughness of the surface of the active layer 22 of Sample 2 is 5.08 nm, which is larger than that of Samples 3 and 4. This is because although the sample 2 is dominated by the step flow growth, it is affected by the hillock H to some extent.

  The average roughness of the surface of the active layer 22 of Sample 5 is 5.94 nm, which is larger than that of Samples 3 and 4. This is not due to the effect of the hillock H, but because the step of the terrace is large in the sample 5, the average roughness on the surface of the active layer 22 is thereby increased.

  FIG. 6 to FIG. 8 are diagrams showing the states of the surfaces of the n-type cladding layer and the active layer in each of Sample 1, Sample 3, and Sample 4 in comparison. As shown in FIG. 1, the n-type cladding layer 21 is a layer formed immediately before the active layer 22. That is, FIGS. 6 to 8 are diagrams showing the states of the front surface and the back surface of the active layer 22 in comparison.

  FIGS. 6A to 8A are AFM images showing a state in a 25 μm square region on each surface of the n-type cladding layer 21 and the active layer 22. FIGS. 6B to 8B show the frequency distribution of the height on the surface of each of the n-type cladding layer 21 and the active layer 22 shown in FIGS. 6A to 8A. It is a histogram. FIGS. 6C to 8C show three-dimensional regions smaller than the respective regions of the n-type cladding layer 21 and the active layer 22 shown in FIGS. 6A to 8A. It is an AFM image displayed as a bird's eye view. FIGS. 6A to 8A also show the average roughness of each surface of the n-type cladding layer 21 and the active layer 22 of the samples 1, 3, and 4.

  As shown in FIG. 6A, in Sample 1, the average roughness on the surface of the n-cladding layer 21 is 10.3 nm, and the average roughness on the surface of the active layer 22 is 10.9 nm. Further, as shown in FIG. 7A, in Sample 3, the average roughness on the surface of the n-cladding layer 21 is 3.78 nm, and the average roughness on the surface of the active layer 22 is 3.66 nm. Further, as shown in FIG. 8A, in Sample 4, the average roughness on the surface of the n-cladding layer 21 is 3.83 nm, and the average roughness on the surface of the active layer 22 is 3.94 nm.

  As described above, in each of Sample 1, Sample 3, and Sample 4, the average roughness of each surface of the n-cladding layer 21 and the active layer 22 is substantially equal. In particular, the ratio obtained by dividing the absolute value of the difference in the average roughness on the respective surfaces of the active layer 22 and the n-cladding layer 21 by the average roughness on the surface of the active layer 22 is within 10%. This is because the active layer 22 is formed by uniform growth while maintaining the average roughness on the surface. Although only the sample 1, the sample 3 and the sample 4 are illustrated here, the same applies to the sample 2 and the sample 5.

  As shown in FIGS. 7B and 8B, in each of Samples 3 and 4, the frequency distribution of the height on each surface of the n-cladding layer 21 and the active layer 22 is both normal distribution. It is. This indicates that the active layer 22 was formed while maintaining step flow growth from the beginning to the end. Although only the samples 3 and 4 are illustrated here, the same applies to the samples 2 and 5. On the other hand, such a tendency is not observed in Sample 1 in which three-dimensional growth is dominant.

  On the other hand, as shown in FIGS. 7C and 8C, in each of Samples 3 and 4, the terrace (the multi-stage triangular facet) on the surface of the active layer 22 is formed by the n-cladding layer 21. The corners are rounded compared to the terrace on the surface. This phenomenon will be described with reference to FIGS. 9 and 10 described later. Although only the samples 3 and 4 are illustrated here, the same applies to the samples 2 and 5. On the other hand, such a tendency is not clearly observed in Sample 1 in which three-dimensional growth is dominant.

  9 and 10 are graphs showing the light emission characteristics of Samples 1 to 5. Specifically, FIG. 9 is a graph showing peak emission wavelengths and half widths of Samples 1 to 5, and FIG. 10 is a graph showing optical outputs of Samples 1 to 5. Note that FIGS. 9A and 10A are graphs in which the vertical axis represents light emission characteristics and the horizontal axis represents sample numbers. FIGS. 9B and 10B are graphs in which the vertical axis represents the light emission characteristics and the horizontal axis represents the average roughness on the surface of the active layer 22.

  As shown in FIG. 9, in Samples 2 to 5 in which the step flow growth is dominant as compared with Sample 1 in which the three-dimensional growth is dominant, the peak emission wavelength shifts to the longer wavelength side, and the half width is increased. Becomes larger.

  Regarding this result, considering that AlGaN constituting the active layer 22 (particularly the well layer 22a) is a mixed crystal of AlN and GaN, and GaN has a longer emission wavelength than AlN, the active layer 22 (particularly the well layer 22a) It is presumed that Ga contained in AlGaN constituting 22a) segregates, and a current also flows through the segregated region to emit light. That is, it is presumed that Ga is segregated in the active layer 22 (particularly, the well layer 22a) by performing the step flow growth such that the average roughness on the surface of the active layer 22 exceeds the thickness of the well layer 22a. Is done.

  Further, as described above, the rounded terrace on the surface of the active layer 22 shown in FIGS. 7C and 8C also indicates the segregation of Ga. Specifically, when the active layer 22 (particularly, the well layer 22a having a large GaN mole fraction) is formed by step flow growth, Ga that is easily migrated as compared with Al is placed on the side surface of the terrace and in the next stage. It is presumed that the terrace was rounded by gathering at the boundary with the surface of the terrace to smooth the steps.

  Then, as shown in FIG. 10, the three-dimensional growth is more dominant than the step flow growth (the average roughness on the surface of the active layer 22 is larger than 10 nm). The target samples 2 to 5 have higher light outputs. This indicates that the light output of the active layer 22 can be increased by segregating Ga in the active layer 22 (particularly, the well layer 22a).

  As described above, Ga segregation occurs in the active layer 22 (particularly, the well layer 22a) by the step flow growth such that the average roughness on the surface of the active layer 22 is equal to or more than the thickness of the well layer 22a and equal to or less than 10 nm. By doing so, the light output of the active layer 22 can be increased. Therefore, it is possible to obtain the light emitting element 1 having the active layer 22 having a good light output.

  Further, as shown in FIG. 10, in Samples 2 to 5 in which the average roughness on the surface of the active layer 22 is 3 nm or more, the segregation of Ga sufficiently occurs in the active layer 22 (particularly, the well layer 22a). Therefore, the light output of the active layer 22 can be sufficiently increased.

  As shown in FIG. 10, as compared to Sample 5 in which the average roughness on the surface of the active layer 22 is about 6 nm, Samples 2 to 4 in which the average roughness on the surface of the active layer 22 is less than The light output increases remarkably (for example, 1.5 times or more). Therefore, by setting the average roughness on the surface of the active layer 22 to 6 nm or less (preferably 5.5 nm or less), the light output of the active layer 22 can be significantly increased.

  As shown in FIGS. 7 and 8, in the light emitting section 20, the average roughness of each surface of the active layer 22 and the n-type cladding layer 21 formed immediately before the active layer 22 is substantially equal. Therefore, not only the active layer 22 but also the n-type cladding layer 21 has an average surface roughness equal to or more than the thickness of the well layer 22a (more preferably, 3 nm or more) and 10 nm or less (further, 6 nm or less). In this case, since the active layer 22 is formed by uniform growth while maintaining the average roughness on the growth surface, Ga segregation can be reliably generated in the active layer 22 (particularly, the well layer 22a). Therefore, the light output of the active layer 22 can be increased.

[AlN layer of base part]
Next, the state of the surface of the AlN layer 12 in the underlayer 10 will be described with reference to the drawings. 11 to 16 are AFM images showing the state of the surface of the AlN layer in the base portions of Samples 1 to 5 and Sample A. Note that Sample A is a sample manufactured under conditions intermediate between Sample 1 and Sample 2.

  FIGS. 11A to 16A are AFM images showing states in a 25 μm square area on the surface of the AlN layer 12. FIGS. 11B to 16B are AFM images displaying an area smaller than the area of the AlN layer 12 shown in FIGS. 11A to 16A. FIGS. 11C to 16C are AFM images in which the area of the AlN layer 12 shown in FIGS. 11B to 16B is displayed as a three-dimensional bird's-eye view. Further, FIGS. 11A to 16A also show the average width of the terraces (hereinafter, simply referred to as “terrace width”) with respect to the tilt direction of the substrate 11 when viewed from above. FIGS. 11B to 16B also show the average height of the terrace (hereinafter, referred to as “terrace height”).

  For example, the terrace width is a predetermined distance (for example, 25 μm) with respect to the tilt direction (the left-right direction in the figure) of the substrate 11 in each of the regions shown in FIGS. 11A to 16A. It can be calculated by randomly selecting a range, measuring the number of terraces included in the range, and dividing the distance by the measured number. Note that the terrace width may be calculated by repeating this calculation a plurality of times (for example, about 10 times) and averaging the obtained values.

  For example, the terrace height is a height measured at the position of the tip of the terrace when measuring the area shown in each of FIGS. 11B to 16B by AFM, and is adjacent to the end. The difference can be calculated by calculating the difference between the height of the next terrace, which is the position, and averaging the difference over a plurality of points (for example, about 50 points).

  As shown in FIGS. 11A to 16B, the terrace width tends to decrease in the order of Sample 1, Sample A, Sample 2, Sample 3, Sample 4, and Sample 5. Also, as shown in FIGS. 11B and 11C to 16B and 16C, sample 1, sample A, sample 2, sample 3, sample 4, and sample 5 are arranged in this order. , Terrace height tends to be large.

  Then, as described above, Samples 2 to 5 (see FIG. 10) having a good optical output of the active layer 22 have a smaller terrace width and a larger terrace height as compared with Sample 1 and Sample A. . Specifically, in samples 2 to 5, the terrace width is 0.3 μm or more and 1 μm or less, and the terrace height is 8 nm or more and 14 nm or less.

  Therefore, by forming the light emitting portion 20 on the surface of the base portion 10 having the AlN layer 12, it is possible to obtain the active layer 22 having good optical output that causes Ga segregation by the step flow growth as described above. .

[AlN mole fraction of well layer]
In Samples 1 to 5 exemplified in the above description, as apparent from the fact that the peak emission wavelength shown in FIG. 9 falls within the range of 265 nm ± 3 nm, the mole fraction of AlN in the well layer 22a is substantially the same. Is the size of However, even in the light-emitting element 1 having the well layer 22a whose AlN molar fraction is significantly different from that of the samples 1 to 5, Ga segregation may occur similarly to the samples 1 to 5.

  Hereinafter, the relationship between the AlN mole fraction (peak emission wavelength) of the well layer 22a and the segregation of Ga will be described with reference to the drawings. FIG. 17 is a spectrum diagram showing emission spectra of a plurality of samples in which the AlN mole fraction (peak emission wavelength) of the well layer is different. FIG. 18 is a graph showing the relationship between the peak emission wavelength and the half width of each sample. 17 and 18 also show, for comparison, a sample in which the well layer 22a is made of GaN. In addition, in FIG. 17, a representative sample is selected from the samples shown in FIG. 18 for the sake of easy viewing of the outline of the emission spectrum and convenience of comparison, and the intensity of the peak emission wavelength in each of the samples is 1 It is standardized so that FIG. 18 also shows an approximate straight line for the measurement results of the peak emission wavelength and the half width of each sample.

  FIG. 17 shows the emission spectrum of each sample having peak emission wavelengths of 265 nm, 285 nm, 320 nm, 335 nm, and 354 nm. Note that the sample having a peak emission wavelength of 265 nm is a sample having the same peak emission wavelength as the above-described samples 1 to 5. A sample having a peak emission wavelength of 354 nm (a thin curve in the figure) is a sample in which the well layer 22a is made of GaN.

  In a sample in which the well layer 22a is made of GaN (a sample having a peak emission wavelength of 354 nm), Ga is not segregated because the well layer 22a is made of a single GaN crystal. In this case, since the variation in the energy of each of the electrons and holes confined in the well layer 22a is reduced, the variation in the energy difference between the recombination electrons and the holes is reduced (that is, the variation in the emission wavelength is reduced). Therefore, the emission spectrum has a small half width. Hereinafter, the half-value width in the emission spectrum of this sample is defined as a half-value width of the emission spectrum when Ga segregation does not occur, and is a reference for determining the presence or absence and degree of Ga segregation (hereinafter, referred to as “reference half-value width”) ).

On the other hand, in each sample in which the well layer 22a is made of Al _X Ga _1-X N (each sample having a peak emission wavelength of 265 nm, 285 nm, 335 nm, and 320 nm), the well layer 22a is a mixed crystal of AlN and GaN. As a result, Ga segregation may occur as described above. In each of the emission spectra of these samples, the half width is larger than the reference half width. This is because the variation in the energy of each of the electrons and the holes confined in the well layer 22a increases due to the segregation of Ga, and the variation in the energy difference between the recombination electrons and the holes increases (that is, light emission). This is because the wavelength variation becomes large).

However, as shown in FIG. 17 and FIG. 18, the magnitude of the half width (ie, the degree of segregation of Ga) in the emission spectrum of each sample is not uniform. Specifically, the half width becomes smaller as the peak emission wavelength becomes smaller. This is the mole fraction X of AlN of _Al X _{Ga 1-X} N constituting the well layer 22a is 1 (i.e., crystals of a single AlN) closer to, because the amount of Ga is less likely to less and less segregated It is.

In this regard, as shown in FIG. 18, if the well layer 22 a is made of Al _X Ga ₁ -XN so that the peak emission wavelength becomes 230 nm or more (preferably 240 nm or more, more preferably 250 nm or more), light emission can be obtained. It is possible to make the half width of the spectrum sufficiently larger than the reference half width (broken line in the figure) (that is, to increase the optical output by causing Ga segregation).

On the other hand, as shown in FIGS. 17 and 18, the half width increases as the peak emission wavelength increases. However, _Al X _{Ga 1-X} mole fraction X of AlN of N is 0 constituting the well layer 22a (i.e., a single GaN crystal) approaches the, by segregation excessively becomes large amounts of Ga In addition, problems such as the shape of the emission spectrum being collapsed (for example, two or more peaks can be separated) or the emission intensity decreasing.

In this regard, as shown in FIGS. 17 and 18, the well layer 22a is formed of Al _X Ga ₁ -XN (0 <X <1) so that the peak emission wavelength is 340 nm or less (preferably 335 nm or less). Then, it becomes possible to make the above-mentioned problem hard to occur.

  As described above, by setting the peak emission wavelength to be 230 nm or more and 340 nm or less, Ga segregation is sufficiently generated in the active layer 22 (particularly, the well layer 22a) to increase the light output, and the shape of the emission spectrum is changed. It is possible to make it difficult to cause a problem such as collapse or a decrease in emission intensity.

  INDUSTRIAL APPLICABILITY The present invention can be used for a nitride semiconductor light emitting device in which an AlGaN-based active layer is formed on a substrate made of sapphire, and particularly, a nitride semiconductor having a peak emission wavelength in an ultraviolet region. It is suitable for use in an ultraviolet light emitting element.

1. Light emitting device (nitride semiconductor ultraviolet light emitting device)
Reference Signs List 10 base part 11 substrate 12 AlN layer 20 light emitting part 21 n-type cladding layer 22 active layer 22a well layer 22b barrier layer 22c electron blocking layer 23 p-type cladding layer 24 p-type contact layer 30 p-electrode 40 n-electrode

Claims (10)

A base portion including a substrate made of sapphire having a surface on which a multi-step terrace is formed by being inclined with respect to the (0001) plane; and an AlN layer formed on the surface of the substrate;
A light-emitting unit including an active layer having an AlGaN-based semiconductor layer formed on the surface of the underlayer,
At least the AlN layer of the base portion, the active layer in the light emitting portion and each layer therebetween are epitaxial growth layers having a surface on which a multi-step terrace is formed,
The active layer has a quantum well structure including at least one well layer composed of Al _X Ga _1-X N (0 <X <1) and having a thickness of 2 nm or more and 3 nm or less;
A nitride semiconductor ultraviolet light emitting device, wherein an average roughness in a 25 μm square region on the surface of the active layer is not less than the thickness of the well layer and not more than 10 nm.   2. The nitride semiconductor ultraviolet light emitting device according to claim 1, wherein an average roughness in a 25 μm square area on the surface of the active layer included in the light emitting unit is 3 nm or more. 3.   The nitride semiconductor ultraviolet light emitting device according to claim 1, wherein an average roughness in a 25 μm square area on the surface of the active layer included in the light emitting unit is 6 nm or less. The average roughness in a 25 μm square region on the surface of a layer formed immediately before the active layer included in the light emitting section is equal to or more than the thickness of the well layer and equal to or less than 10 nm. The nitride semiconductor ultraviolet light emitting device according to any one of claims 1 to 3 . The absolute value of the difference between the average roughness in the 25 μm square region on the surface of each of the active layer and the layer formed immediately before the active layer included in the light emitting portion is calculated by calculating the absolute value of the difference between the 25 μm square The nitride semiconductor ultraviolet light emitting device according to any one of claims 1 to 4, wherein a ratio divided by an average roughness in the region is 10% or less. The surface of the AlN layer included in the base portion, in a top view, according to claim 1 to 5 in which the average width of the terraces in the inclination direction of the substrate, characterized in that at 0.3μm or more and 1μm or less The nitride semiconductor ultraviolet light emitting device according to any one of the above items. The surface of the AlN layer included in the base unit, the average height of the step terrace is formed, according to any one of claims 1 to 6, characterized in that at 8nm or more and 14nm or less Nitride semiconductor ultraviolet light emitting device.   On the surface of the active layer included in the light emitting portion, the frequency distribution of the height in a 25 μm square region is changing from a downward convex curve to an upward convex curve as the height increases from 0. The curve according to any one of claims 1 to 7, wherein the curve is monotonically increased and has a maximum value, and then has a curved shape that monotonically decreases while inflection from an upwardly convex curve to a downwardly convex curve. The nitride semiconductor ultraviolet light emitting device according to the above. The nitride semiconductor ultraviolet light emitting device according to any one of claims 1 to 8, wherein a peak emission wavelength is 230 nm or more and 340 nm or less. Forming a base portion including a substrate made of sapphire having a surface on which a multi-stage terrace is formed by being inclined with respect to a (0001) plane; and an AlN layer formed on the surface of the substrate. ,
Forming a light-emitting portion including an active layer having an AlGaN-based semiconductor layer formed on the surface of the base portion,
25 μm square region on the surface of the active layer having a quantum well structure including at least one well layer having a thickness of 2 nm or more and 3 nm or less composed of Al _X Ga _1-X N (0 <X <1) Under the condition that the average roughness of the AlN layer of the base layer, the active layer of the light emitting section, and each layer therebetween is at least equal to the thickness of the well layer and not more than 10 nm, A method for manufacturing a nitride semiconductor ultraviolet light emitting device, characterized in that it is formed by two-dimensional growth by step flow growth.