US20070045638A1

Movatterモバイル変換

Info

Publication number: US20070045638A1
Application number: US11/211,921
Authority: US
Inventors: Yu-Chen Shen; Nathan Gardner; Satoshi Watanabe; Michael Krames; Gerd Mueller
Original assignee: Lumileds LLC
Current assignee: Lumileds LLC
Priority date: 2005-08-24
Filing date: 2005-08-24
Publication date: 2007-03-01
Also published as: RU2008110934A; CN101297410B; RU2412505C2; WO2007023419A1; US20070145384A1; US20090261361A1; BRPI0615190A2; US8847252B2; KR20080040770A; CN101297410A; BRPI0615190B1; US20170117440A9; TWI438921B; EP1922766B1; EP1922766A1; US20160093770A1; JP2007067418A; TW200717871A; JP2013102240A; US9640724B2

Abstract

A III-nitride light emitting layer is disposed between an n-type region and a p-type region. The light emitting layer is a doped thick layer. In some embodiments, the light emitting layer is sandwiched between two doped spacer layers.

Description

BACKGROUND

1. Field of Invention

This invention relates to the light emitting region of a semiconductor light emitting device.

2. Description of Related Art

Semiconductor light-emitting devices including light emitting diodes (LEDs), resonant cavity light emitting diodes (RCLEDs), vertical cavity laser diodes (VCSELs), and edge emitting lasers are among the most efficient light sources currently available. Materials systems currently of interest in the manufacture of high-brightness light emitting devices capable of operation across the visible spectrum include Group III-V semiconductors, particularly binary, ternary, and quaternary alloys of gallium, aluminum, indium, and nitrogen, also referred to as III-nitride materials. Typically, III-nitride light emitting devices are fabricated by epitaxially growing a stack of semiconductor layers of different compositions and dopant concentrations on a sapphire, silicon carbide, III-nitride, or other suitable substrate by metal-organic chemical vapor deposition (MOCVD), molecular beam epitaxy (MBE), or other epitaxial techniques. The stack often includes one or more n-type layers doped with, for example, Si, formed over the substrate, a light emitting or active region formed over the n-type layer or layers, and one or more p-type layers doped with, for example, Mg, formed over the active region. III-nitride devices formed on conductive substrates may have the p- and n-contacts formed on opposite sides of the device. Often, III-nitride devices are fabricated on insulating substrates, such as sapphire, with both contacts on the same side of the device. Such devices are mounted so light is extracted either through the contacts (known as an epitaxy-up device) or through a surface of the device opposite the contacts (known as a flip chip device).

U.S. Pat. No. 5,747,832 teaches a “light emitting gallium nitride-based compound semiconductor device of a double-heterostructure. The double-heterostructure includes a light-emitting layer formed of a low-resistivity In_xGa_1−xN (0<x<1) compound semiconductor doped with p-type and/or n-type impurity.” See U.S. Pat. No. 5,747,832, abstract. Specifically,column 5 lines45-50 recite “[i]n the present invention, the light-emittinglayer18 preferably has a thickness within a range such that the light-emitting device of the present invention provides a practical relative light intensity of 90% or more. In more detail, the light-emittinglayer18 preferably has a thickness of 10 Å to 0.5 μm, and more preferably 0.01 to 0.2 μm.”Column 10 lines44-49 teach “[i]_nthe third embodiment, the n-type impurity doped in In_xGa_1−xN of the light-emittinglayer18 is preferably silicon (Si). The concentration of the n-type impurity is preferably 1×10¹⁷/cm³to 1×10²¹/cm³from the viewpoint of the light emission characteristics, and more preferably 1×10¹⁸/cm³to 1×10²⁰/cm³.”

Commercial III-nitride devices with InGaN light emitting layers often have multiple quantum well light emitting layers less than 50 Å and typically doped to less than about 1×10¹⁸cm⁻³, as these quantum well designs can improve performance, especially in poor quality epitaxial material, at low drive current. At higher drive currents desirable for lighting, these devices suffer decreasing efficiency with increasing current density. Needed in the art are devices that exhibit high efficiency at high current density.

SUMMARY

In accordance with embodiments of the invention, a III-nitride light emitting layer is disposed between an n-type region and a p-type region. The light emitting layer is doped to a dopant concentration between 6×10⁸cm⁻³and 5×10¹⁹cm⁻³, and has a thickness between 50 Å and 250 Å. In some embodiments, the light emitting layer is sandwiched between and in direct contact with two n-type spacer layers and one or both of the spacer layers is doped to a dopant concentration between 6×10¹⁸cm⁻³and 5×10¹⁹cm³.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 illustrates a III-nitride light emitting device.

FIG. 2 is a plot of external quantum efficiency as a function of current density for a device as shown inFIG. 1 and a device according to embodiments of the present invention.

FIG. 3 illustrates a III-nitride light emitting device according to embodiments of the invention.

FIG. 4 is a plot of decay time as a function of silicon doping level for several InGaN films.

FIGS. 5 and 6 are a plan view and a cross sectional view of a large junction flip chip light emitting device.

FIG. 7 illustrates a thin film light emitting device.

FIG. 8 is an exploded view of a packaged light emitting device.

FIGS. 9A and 9B illustrate relative internal quantum efficiency as a function of silicon doping level in the light emitting layer and first and second spacer layers for devices according to embodiments of the invention emitting light at 450 nm and 400 nm, respectively.

FIG. 10 illustrates relative internal quantum efficiency as a function of light emitting layer thickness for several devices.

FIG. 11 illustrates injection efficiency and internal quantum efficiency as a function of blocking layer composition for several simulated devices.

FIG. 12 illustrates relative internal quantum efficiency as a function of blocking layer composition observed in actual devices.

FIGS. 13A-13K illustrate portions of the conduction band for devices including grading in the light emitting region.

DETAILED DESCRIPTION

FIG. 1 illustrates a common III-nitride light emitting device. An n-type region11 is grown over asapphire substrate10. Anactive region12, including multiple thin quantum well layers separated by barrier layers, is grown over n-type region11, followed by aGaN spacer layer13, a p-type AlGaN layer14, and a p-type contact layer15.

FIG. 2 illustrates the external quantum efficiency of a device such as the device ofFIG. 1 (triangles onFIG. 2) and a device according to an embodiment of the present invention (circles onFIG. 2). The external quantum efficiency is the internal quantum efficiency, defined as the flux of photons produced divided by the flux of carriers supplied, multiplied by the extraction efficiency. For a given lamp design, the extraction efficiency is constant, thus the extraction efficiency is the same for both devices shown inFIG. 2. As the current density applied to the device ofFIG. 1 increases, the external quantum efficiency of the device initially increases, then decreases, as illustrated inFIG. 2. As the current density increases past zero, the external quantum efficiency increases, reaching a peak at a current density of about 10 A/cm². As current density increases beyond 10 A/cm², the external quantum efficiency drops quickly, then the decrease in external quantum efficiency slows at high current density, for example beyond 200 A/cm². Embodiments of the invention are designed to reduce or reverse the drop in quantum efficiency at high current density.

In accordance with embodiments of the invention, III-nitride light emitting devices include a thick double heterostructure light emitting region that is highly doped. Though the embodiments below describe devices where the light emitting layer is doped n-type with Si, it is to be understood that in other embodiments, other dopant species including p-type dopant species may be used. The thick double heterostructure light emitting region may reduce charge carrier density, and doping in and adjacent to the light emitting region may improve the material quality of the light emitting region, both of which may reduce the number of carriers lost to nonradiative recombination. Embodiments of the invention are designed to reduce or reverse the drop in quantum efficiency at high current density observed inFIG. 2 for the device ofFIG. 1.

FIG. 3 illustrates a light emitting device according to embodiments of the invention. Alight emitting region35 is sandwiched between an n-type region31 and a p-type region39.Light emitting region35 may be spaced apart from n-type region31 and p-type region39 by optional first and

second spacer layers

33 and37. The device may include an optional preparation layer32 disposed between n-type region31 andfirst spacer33, and/or anoptional blocking layer38, disposed betweensecond spacer37 and p-type region39.

Rather than thin quantum well layers separated by barrier layers, light emittingregion35 may include one or more thick light emitting layers, for example thicker than 50 angstroms. In some embodiments,light emitting region35 includes a single, thick light emitting layer with a thickness between 50 and 600 angstroms, more preferably between 100 and 250 angstroms. The upper limit on thickness is due to current growth techniques which result in poor material quality as the thickness of the light emitting layer increases beyond 600 angstroms, for example at thicknesses above 1000 angstroms. Poor material quality typically results in reduced internal quantum efficiency. As growth techniques improve, growth of devices with thicker light emitting layers without reduced internal quantum efficiency may be possible and thus within the scope of embodiments of the invention.

The optimal thickness may depend on the number of defects within the light emitting layer. In general, as the number of defects increases, the optimal thickness of the light emitting layer decreases. In addition, defects may be centers for nonradiative recombination, thus it is desirable to reduce the number of defects as much as possible. A comment defect in III-nitride materials is a threading dislocation. The concentration of threading dislocations is measured per unit area. The concentration of threading dislocations in the light emitting region is preferably limited to less than 10⁹cm⁻², more preferably limited to less than 10⁸cm⁻², more preferably limited to less than 10⁷cm⁻², and more preferably limited to less than 10⁶cm⁻². Achieving the above-described threading dislocation concentrations may require growth techniques such as epitaxial lateral overgrowth, hydride vapor phase epitaxy, and growth on freestanding GaN substrates. Epitaxial lateral overgrowth involves selective growth of GaN over openings in a mask layer formed on GaN layer grown on a conventional growth substrate such as sapphire. The coalescence of the selectively-grown GaN may enable the growth of a flat GaN surface over the entire growth substrate. Layers grown subsequent to the selectively-grown GaN layer may exhibit low defect densities. Epitaxial lateral overgrowth is described in more detail in Mukai et al., “Ultraviolet InGaN and GaN Single-Quantum Well-Structure Light-Emitting Diodes Grown on Epitaxial Laterally Overgrown GaN Substrates,” Jpn. J. Appl. Phys. Vol. 38 (1999) p. 5735, which is incorporated herein by reference. Hydride vapor phase epitaxial growth of freestanding GaN substrates is described in more detail in Motoki et al., “Preparation of Large Freestanding GaN Substrates by Hydride Vapor Phase Epitaxy Using GaAs as a Starting Substrate,” Jpn. J. Appl. Phys. Vol. 40 (2001) p. L140, which is incorporated herein by reference.

In addition to being thick, the light emitting layers of light emittingregion35 region are doped, for example doped n-type with Si. In some embodiments, Si is used as the dopant because Si may provide other improvements to the material, such as a rough surface that may improve light extraction from the device or relieve strain in the light emitting layer.FIG. 4 is a plot of decay time as a function of silicon doping level for several InGaN films. To gather the data illustrated inFIG. 4, InGaN films doped to the levels illustrated were probed with a laser at low excitation intensities. The lifetimes of the carriers, i.e. the length of time before a carrier is consumed by a defect, were measured. A longer lifetime indicates better material quality in the film; a shorter lifetime indicates worse material quality. As illustrated inFIG. 4, InGaN films doped to between 6×10¹⁸and 2×10⁹cm⁻³have the longest lifetimes, indicating that nonradiative recombination is slowest in these films.FIG. 4 demonstrates that the rate of nonradiative recombination may be influenced by Si-doping levels in the light emitting layers.

FIG. 9A illustrates the internal quantum efficiency at 330 A/cm²as a function of silicon doping level for devices with 96 angstrom thick In_0.16Ga_0.84N light emitting layers which emit light at a peak wavelength of about 450 nm.FIG. 9B illustrates the internal quantum efficiency at 330 Å/cm²as a function of silicon doping level for devices with 96 angstrom thick In_0.08Ga_0.92N light emitting layers which emit light at a peak wavelength of about 400 nm. As inFIG. 4,FIGS. 9A and 9B illustrate that devices with InGaN light emitting layers with silicon doping levels between about 6×10¹⁸and 3×10¹⁹cm⁻³have the highest internal quantum efficiency. In particular,FIG. 9A illustrates a peak in internal quantum efficiency at 330 A/cm²at about 2×10⁹cm³for a device emitting light at a peak wavelength of about 450 nm.FIG. 9B illustrates a peak in internal quantum efficiency at about 8×10¹⁸cm⁻³for a device emitting light at a peak wavelength of about 400 nm. As the doping level in the light emitting layer drops below 6×10¹⁸cm⁻³or increases above 3×10¹⁹cm⁻³, the internal quantum efficiency drops. As the silicon doping level increases beyond 3×10¹⁹cm⁻³, the material quality becomes poor.

The circles inFIG. 2 illustrate the external quantum efficiency as a function of current density for a device according to embodiments of the invention. In the device illustrated by the circles inFIG. 2, a 130 angstrom thick In_0.12Ga_0.88N light emitting layer is doped with Si to a concentration of 10¹⁹cm⁻³. This device emits light at about 430 nm. In contrast to the device ofFIG. 1, illustrated inFIG. 2 by triangles, in the device according to the embodiments of the invention, the external quantum efficiency improves as the current density increases, then levels off at an external quantum efficiency of about 26% at a current density of about 250 A/cm². At the same current density, the device ofFIG. 1 has an external quantum efficiency of only about 18%, which drops as the current density increases.

In some embodiments, device performance significantly improves only when the optimal thicknesses according to embodiments of the invention and the optimal silicon doping levels according to embodiments of the invention are implemented together, as illustrated inFIG. 10, which is a plot of internal quantum efficiency at 330 A/cm²as a function of light emitting layer thickness for several devices. The diamonds inFIG. 10 represent devices with a light emitting layer with a threading dislocation density of 1.5×10⁹cm⁻²doped to 10¹⁸cm⁻³, a doping level below the optimal silicon doping ranges described above. The squares inFIG. 10 represent a device with a light emitting layer with a threading dislocation density of 4×10⁸cm⁻²doped to 10¹⁹cm⁻³, a doping level within the optimal silicon doping ranges described above.

As illustrated by the diamonds inFIG. 10, for a light emitting layer doped to less than the optimal doping levels described above, the internal quantum efficiency drops as the thickness of the light emitting layer increases into the optimal thickness ranges described above. For example, the internal quantum efficiency for a device with a light emitting layer doped to 10¹⁸cm⁻³drops from a peak at a light emitting layer thickness less than 50 angstroms, to zero at a light emitting layer thickness of about 130 angstroms. In contrast, as illustrated by the squares inFIG. 10, at the optimal doping levels described above, at a light emitting layer thickness between about 80 angstroms and about 230 angstroms, the internal quantum efficiency is above the peak internal quantum efficiency observed for the light emitting layer doped to only 10¹⁸cm⁻³.

FIG. 10 also illustrates that even if the light emitting layer is doped to the optimal doping levels described above, the internal quantum efficiency of the device suffers if the light emitting layer thickness is outside the optimal thickness range described above. For example, the squares inFIG. 10 illustrate that at 30 angstroms, a common thickness for thin quantum wells in a device such as illustrated inFIG. 1 and a thickness below the optimal light emitting layer thicknesses described above, a device with a light emitting layer doped to an optimal doping level of 10¹⁹cm⁻³still demonstrates very low internal quantum efficiency. Thus, in some embodiments both the light emitting layer thickness and the light emitting layer dopant concentration must be within the optimal ranges described above in order to realize improvements in internal quantum efficiency.

In some embodiments, silicon-doped first and second spacer layers33 and37 are combined with the optimal light emitting region thicknesses and doping levels described above. As illustrated in Table 1 below, the internal quantum efficiency of a thick, optimally silicon-doped light emitting layer may be further boosted by doping to the same doping ranges the spacer layers directly adjacent to the light emitting layer. The spacer layers may be, for example, between about 20 and about 1000 angstroms thick, and are usually about 100 angstroms thick.

TABLE 1


Relative Internal Quantum Efficiency at 330 A/cm²for
64 Å InGaN Light Emitting Layers

Spacer Layer Doping

	10¹⁸cm⁻³	10¹⁹cm⁻³

Light Emitting	10¹⁹cm⁻³	31%	41%
Layer Doping
10¹⁸cm⁻³	9.4%	39%

The data illustrated inFIGS. 2, 4, and10 and in Table 1 are for devices with InGaN light emitting layers with 12% InN, which emit light at about 430 nm. Increasing the wavelength of emitted light requires increasing the amount of InN in the light emitting layer. In general, as the amount of InN in a layer increases, the material quality of the layer deteriorates. Accordingly, devices with more InN in the light emitting layers may require greater dopant concentrations in the light emitting layers, in order to achieve the improvements in efficiency, as illustrated inFIGS. 2, 9A, and10. For example, in devices with InGaN light emitting layers with 16% InN, which emit light at about 450 nm, the optimal silicon doping level may be, for example, 1×10¹⁹cm⁻³to 5×10¹⁹cm⁻³as illustrated inFIG. 9A, instead of 6×10¹⁸cm⁻³to 10¹⁹cm⁻³as illustrated inFIG. 9B for a device emitting light at 400 nm.

The above examples describe optimal thicknesses and doping levels for each of thefirst spacer layer33,light emitting region35, andsecond spacer layer37. In various embodiments one or more of

regions

33,35, and37 may not be intentionally doped or may be doped to a level below the optimal doping ranges given above. For example, all three of

regions

33,35, and37 may be optimally doped;spacer layer33 andlight emitting region35 may be optimally doped andspacer layer37 may not be intentionally doped or may be doped to a level below the optimal range;spacer layer37 andlight emitting region35 may be optimally doped andspacer layer33 may not be intentionally doped or may be doped to a level below the optimal range; or both spacer layers33 and37 may be optimally doped andlight emitting region35 may not be intentionally doped or may be doped to a level below the optimal range.

In some embodiments, the internal quantum efficiency of the device may be further improved by including an optionalcurrent blocking layer38, as illustrated inFIG. 3. Blockinglayer38 confines current within the light emitting layer, and in some embodiments is an aluminum-containing p-type layer, often p-type AlGaN or p-type AlInGaN. The internal quantum efficiency in a device is a function of the product of the current injection efficiency and the radiative recombination efficiency. The current injection efficiency is the ratio of the amount of current that recombines in the light emitting layer to the amount of current supplied to the device. The radiative recombination efficiency is the ratio of the amount of current that recombines in the light emitting layer and emits light (in contrast to current that recombines in the light emitting layer for example in a crystal defect, and does not emit light) to the total amount of current that recombines in the light emitting layer.

FIG. 12 illustrates the internal quantum efficiency as a function of composition of AlN in blockinglayer38, as observed in actual devices that emit light at 430 nm. As illustrated inFIG. 12, as the composition of AlN in blocking layer increases above 0%, the internal quantum efficiency improves to a peak in internal quantum efficiency at an AlN composition of about 20%. Accordingly, in some embodiments of the device, the AlN composition in an AlGaN blocking layer is greater than 8% and less than 30%, preferably greater than 15% and less than 25%. The reduction in internal quantum efficiency as the AlN composition increases over 20% may be due to incorporation of contaminants during growth of high AlN composition layers.

The AlN compositions described above may be generalized to desirable band gaps for blockinglayer38, as illustrated below in Table 2. The data in Table 2 are calculated by the equation E_g,Al_x_Ga_1−x_N=E_g,GaN·(1−x)+E_g,AlN·x−bx(1−x) where E_g,GaNis the band gap of GaN, 3.4 eV, E_g,AlNis the band gap of AlN, 6.2 eV, and b is a bowing parameter, 1 eV at room temperature.

TABLE 2


Band Gap for Al_xGa_l-xN Blocking Layers

	x, Composition of AlN	E_g,AlGaN(eV)

	0.05	3.49
	0.08	3.55
	0.12	3.63
	0.15	3.69
	0.17	3.73
	0.20	3.80
	0.25	3.89

As illustrated in Table 2 AlN compositions between 8% and 25% correspond to band gaps between 3.55 and 3.89 eV. Accordingly, in embodiments of the invention, blocking layer may be a layer of any composition with a band gap greater than 3.5 eV. Since the upper limit on AlN composition illustrated inFIG. 12 is likely due to materials problems particular to current growth techniques for AlGaN layers and not due to band gap, the upper limit to the band gap of an AlGaN blocking layer is that of an AlN blocking layer, assuming these material problems are resolved.

Blockinglayer38 must be thick enough so charge carriers cannot tunnel through blockinglayer38, generally greater than 10 Å thick. In some embodiments, blockinglayer38 is between 10 and 1000 Å thick, more preferably between 100 and 500 Å thick. In some embodiments, blockinglayer38 may be part of or the entire p-type region39; for example, blockinglayer38 may be a layer on which an electrical contact to the p-type side of the light emitting layer is formed.

In some embodiments, the internal quantum efficiency of the device may be further improved by including an optional preparation layer32, as illustrated inFIG. 3. Preparation layer32 may be a smoothing structure as described in U.S. Pat. No. 6,635,904, “Indium Gallium Nitride Smoothing Structures For III-nitride Devices,” granted Oct. 21, 2003, and incorporated herein by reference. Preparation layer32 is formed over n-type region31. The preparation layer may be an n-type layer located beneath the light emitting layer, within 5000 angstroms of the light emitting layer. The preparation layer can have a thickness ranging from about 200 angstroms to several microns. Preparation layer32 has a lower indium composition thanlight emitting region35; for example, preparation layer32 may be an InGaN layer containing 2-12% InN, and more preferably containing 2-6% InN. In some embodiments, preparation layer32 may be part of n-type region31; for example, preparation layer32 may be a layer on which an electrical contact to the n-type side of the light emitting layer is formed.

In addition to the decrease in efficiency at high current density, the device ofFIG. 1 may also exhibit a peak wavelength that shifts as the current density applied to the device increases. Device designs according to embodiments of the invention may desirably reduce or eliminate the shift in the peak wavelength as the current density increases, as illustrated in Table 3. The devices in Table 3 labeled “FIG. 3 Device” have 130 angstrom thick light emitting layers doped with Si to a concentration of 1×10¹⁹cm⁻³that emit light at a peak wavelength of about 430 nm, 200 angstrom thick GaN first spacer layers doped with Si to a concentration of 1×10¹⁹cm⁻³, 100 angstrom thick GaN second spacer layers doped with Si to a concentration of 1×10¹⁹cm⁻³, and 210 angstrom thick Al_0.16Ga_0.84N blocking layers.

TABLE 3


Peak Wavelength Shift for Devices According to Embodiments of the
Invention and Devices According toFIG. 1:

	Wavelength shift,	Wavelength Shift,
Current Density Change	FIG. 1 Device	FIG. 3 Device

From 20 to 930 A/cm²	8 nm	3 nm
From 20 to 400 A/cm²	6 nm	2 nm
From 20 to 200 A/cm²	4nm	1 nm

Though in the above examples each device includes only a single light emitting layer, some embodiments of the invention include multiple light emitting layers separated by barriers. In addition, though the above examples use silicon as the dopant in the light emitting region and surrounding layers, in some embodiments other suitable dopants may be used in addition to or instead of silicon, such as other group IV elements such as germanium and tin, group VI elements such as oxygen, selenium, tellurium, and sulfur, group III elements such as aluminum or boron, and p-type dopants such as magnesium. Finally, though the examples above describe devices with InGaN light emitting layers that typically emit light in the near-UV through infrared range, in other embodiments the light emitting layer or spacer layers may be GaN, AlGaN, or AlInGaN, and the devices may emit UV through red light.

Though in the examples described above each doped layer or region (such as the light emitting layer or spacer layers) is uniformly doped, in other embodiments one or more doped layers or regions may be partially doped, or the doping may be graded. Alternatively or in addition, the composition of one or more layers described above may be graded. As used herein, the term “graded” when describing the composition or dopant concentration in a layer or layers in a device is meant to encompass any structure that achieves a change in composition and/or dopant concentration in any manner other than a single step in composition and/or dopant concentration. In one example, doping in one or both of the spacer layers is graded. In another example, the InN composition in the light emitting layer is graded. Each graded layer may be a stack of sublayers, each of the sublayers having a different dopant concentration or composition than either sublayer adjacent to it. If the sublayers are of resolvable thickness, the graded layer is a step-graded layer. In the limit where the thickness of individual sublayers approaches zero, the graded layer is a continuously-graded region. The sublayers making up each graded layer can be arranged to form a variety of profiles in composition and/or dopant concentration versus thickness, including, but not limited to, linear grades, parabolic grades, and power-law grades. Also, graded layers are not limited to a single grading profile, but may include portions with different grading profiles and one or more portions with substantially constant composition and/or dopant concentration regions.

FIGS. 13A-13K illustrate several grading schemes for thelight emitting region35 ofFIG. 3.FIGS. 13A-13K illustrate a portion of the conduction band of energy band diagrams includingfirst spacer33,light emitting region35, andsecond spacer layer37. In a device with an InGaNlight emitting region35, the larger the band gap, i.e. the higher the level shown on each figure, the less InN is present. Thus, takingFIG. 13A as an example,first spacer layer33 is GaN or InGaN with no InN or a low composition of InN. The InN composition is increased and held constant in a first portion of InGaNlight emitting layer35, then the InN composition is graded to zero or a low composition at the interface withsecond spacer layer37. InFIGS. 13G-13K, there are one or morelocal maxima130 in the band gap withinlight emitting region35. In an InGaNlight emitting region35, these local maxima in band gap represent regions with lower InN composition than the surrounding regions of light emittingregion35. In some embodiments, the difference in band gap between these local maxima and the regions of light emittingregion35 surrounding them is small enough that the maxima do not have electronic states, meaning there is no quantum confinement in the regions between them, thus the regions between them are not quantum wells.

The semiconductor structure illustrated inFIG. 3 may be included in any configuration of a light emitting device.FIGS. 5 and 6 illustrate a flip chip device incorporating the structure ofFIG. 3.FIG. 7 illustrates a thin film device incorporating the structure ofFIG. 3.

FIG. 5 is a plan view of a large junction device (i.e. an area greater than or equal to one square millimeter).FIG. 6 is a cross section of the device shown inFIG. 5, taken along the axis indicated.FIGS. 5 and 6 also illustrate an arrangement of contacts that may be used with the semiconductor structure illustrated inFIG. 3. The device ofFIGS. 5 and 6 is described in more detail in U.S. Pat. No. 6,828,586, which is incorporated herein by this reference. The entire semiconductor structure illustrated inFIG. 3 and described above in various examples is represented onFIG. 6 asepitaxial structure110, grown on agrowth substrate10 which remains a part of the finished device. Multiple vias are formed in which n-type contacts114 make electrical contact to n-type region31 ofFIG. 3. P-type contacts112 are formed on the remaining portions of p-type region39 ofFIG. 3. The individual n-type contacts114 formed in the vias are electrically connected byconductive regions118. The device may be flipped relative to the orientation illustrated inFIGS. 5 and 6 and mounted on a mount (not shown) contact-side down such that light is extracted from the device throughsubstrate10. N-type contacts114 andconductive regions118 make electrical contact to the mount by n-type connection region124. Underneath n-type connection region124, the p-type contacts112 are isolated from n-type contacts114,conductive regions118, and n-type connection region124 bydielectric116. P-type contacts112 make electrical contact to the mount by p-type connection region122. Underneath p-type connection region122, n-type contacts114 andconductive regions118 are isolated from p-type connection region122 bydielectric120.

FIG. 7 is a cross sectional view of a thin film device, a device from which the growth substrate is removed. The device illustrated inFIG. 7 may be formed by growing thesemiconductor structure57 ofFIG. 3 on aconventional growth substrate58, bonding the device layers to ahost substrate70, then removinggrowth substrate58. For example, n-type region31 is grown oversubstrate58. N-type region31 may include optional preparation layers such as buffer layers or nucleation layers, and optional release layers designed to facilitate release of the growth substrate or thinning of the epitaxial layers after substrate removal.Light emitting region35 is grown over n-type region31, followed by p-type region39.Light emitting region35 may be sandwiched between optional first and second spacer layers33 and37. One ormore metal layers72, including, for example, ohmic contact layers, reflective layers, barrier layers, and bonding layers, are deposited over p-type region39.

The device layers are then bonded to ahost substrate70 via the exposed surface of metal layers72. One or more bonding layers (not shown), typically metal, may serve as compliant materials for thermo-compression or eutectic bonding between the epitaxial structure and the host substrate. Examples of suitable bonding layer metals include gold and silver.Host substrate70 provides mechanical support to the epitaxial layers after the growth substrate is removed, and provides electrical contact to p-type region39.Host substrate70 is generally selected to be electrically conductive (i.e. less than about 0.1 Ωcm), to be thermally conductive, to have a coefficient of thermal expansion (CTE) matched to that of the epitaxial layers, and to be flat enough (i.e. with an root mean square roughness less than about 10 nm) to form a strong wafer bond. Suitable materials include, for example, metals such as Cu, Mo, Cu/Mo, and Cu/W; semiconductors with metal contacts, such as Si with ohmic contacts and GaAs with ohmic contacts including, for example, one or more of Pd, Ge, Ti, Au, Ni, Ag; and ceramics such as AlN, compressed diamond, or diamond layers grown by chemical vapor deposition.

The device layers may be bonded tohost substrate70 on a wafer scale, such that an entire wafer of devices are bonded to a wafer of hosts, then the individual devices are diced after bonding. Alternatively, a wafer of devices may be diced into individual devices, then each device bonded tohost substrate70 on a die scale, as described in more detail in U.S. application Ser. No. 10/977,294, “Package-Integrated Thin-Film LED,” filed Oct. 28, 2004, and incorporated herein by reference.

Host substrate

70 andsemiconductor structure57 are pressed together at elevated temperature and pressure to form a durable bond at the interface betweenhost substrate70 andmetal layers72, for example a durable metal bond formed between metal bonding layers (not shown) at the interface. The temperature and pressure ranges for bonding are limited on the lower end by the strength of the resulting bond, and on the higher end by the stability of the host substrate structure, metallization, and the epitaxial structure. For example, high temperatures and/or high pressures can cause decomposition of the epitaxial layers, delamination of metal contacts, failure of diffusion barriers, or outgassing of the component materials in the epitaxial layers. A suitable temperature range for bonding is, for example, room temperature to about 500° C. A suitable pressure range for bonding is, for example, no pressure applied to about 500 psi.Growth substrate58 is then removed.

In order to remove a sapphire growth substrate, portions of the interface betweensubstrate58 andsemiconductor structure57 are exposed, throughsubstrate58, to a high fluence pulsed ultraviolet laser in a step and repeat pattern. The exposed portions may be isolated by trenches etched through the crystal layers of the device, in order to isolate the shock wave caused by exposure to the laser. The photon energy of the laser is above the band gap of the crystal layer adjacent to the sapphire (GaN in some embodiments), thus the pulse energy is effectively converted to thermal energy within the first 100 nm of epitaxial material adjacent to the sapphire. At sufficiently high fluence (i.e. greater than about 500 mJ/cm²) and a photon energy above the band gap of GaN and below the absorption edge of sapphire (i.e. between about 3.44 and about 6 eV), the temperature within the first 100 nm rises on a nanosecond scale to a temperature greater than 1000° C., high enough for the GaN to dissociate into gallium and nitrogen gasses, releasing the epitaxial layers fromsubstrate58. The resulting structure includessemiconductor structure57 bonded tohost substrate70. In some embodiments, the growth substrate may be removed by other means, such as etching, lapping, or a combination thereof.

After the growth substrate is removed,semiconductor structure57 may be thinned, for example to remove portions of n-type region31 closest tosubstrate58 and of low material quality. The epitaxial layers may be thinned by, for example, chemical mechanical polishing, conventional dry etching, or photoelectrochemical etching (PEC). The top surface of the epitaxial layers may be textured or roughened to increase the amount of light extracted. A contact (not shown) is then formed on the exposed surface of n-type region31. The n-contact may be, for example, a grid. The layers beneath the n-contact may be implanted with, for example, hydrogen to prevent light emission from the portion of light emittingregion35 beneath the n-contact. Secondary optics known in the art such as dichroics or polarizers may be applied onto the emitting surface to provide further gains in brightness or conversion efficiency.

FIG. 8 is an exploded view of a packaged light emitting device, as described in more detail in U.S. Pat. No. 6,274,924. A heat-sinking slug100 is placed into an insert-molded leadframe. The insert-molded leadframe is, for example, a filledplastic material105 molded around ametal frame106 that provides an electrical path.Slug100 may include anoptional reflector cup102. The light emitting device die104, which may be any of the devices described in the embodiments above, is mounted directly or indirectly via a thermally conductingsubmount103 to slug100. Acover108, which may be an optical lens, may be added.

Having described the invention in detail, those skilled in the art will appreciate that, given the present disclosure, modifications may be made to the invention without departing from the spirit of the inventive concept described herein. Therefore, it is not intended that the scope of the invention be limited to the specific embodiments illustrated and described.