III-Nitride light-emitting devices with improved high-current efficiency

BACKGROUND

FIELD OF THE INVENTION

The present invention relates to semiconductor light-emitting devices.

DESCRIPTION OF RELATED ART

III-Nitride light-emitting devices are based on semiconducting alloys of nitrogen with elements from group III of the periodic table. Examples of such III-Nitride devices include In_xAl_yGa_1-x-yN (0 ≤ x ≤ 1, 0 ≤ y ≤ 1, x + y ≤ 1) based light emitting diodes (LEDs) and laser diodes (LDs). Such III-Nitride light emitting devices are commercially valuable high brightness sources of, for example, ultraviolet, blue, and green light.

In_xAl_yGa_1-x-yN crystals such as those from which In_xAl_yGa_1-x-yN light emitting devices are formed typically adopt the wurtzite crystal structure and are consequently polarized even when not externally biased. The two types of polarization that are significant for III-Nitride light-emitting devices are the spontaneous and the piezoelectric polarizations. The spontaneous polarization arises from differences in the alloy composition of adjacent III-Nitride semiconductor layers. The piezoelectric polarization arises from strain applied to the III-Nitride semiconductor layers by, for example, neighboring layers of different composition (and thus having different lattice constants). The effect of the polarization is to create sheets of fixed charge in the device at interfaces between III-Nitride layers of different alloy composition and to create corresponding electric fields within the layers. These electric fields will be referred to herein as "polarization fields." The signs and magnitudes of the sheet charges and of the electric fields are determined by the crystal orientation, the strain in the layers, and the alloy compositions of the layers.

<figref idref="f0001">Figures 1A and 1B</figref> show conventionally simulated band structure diagrams for a portion of an In_xAl_yGa_1-x-yN light-emitting device including an n-type GaN layer 2 formed over a substrate (not shown), an active region 4 (including barrier layers 6 and quantum well layers 8), a p-type Al_0.2Ga_0.8N electron confinement layer 10, and a p-type GaN contact layer 12. The various layers illustrated are of wurtzite crystal structure with the c-axis of the crystal substantially perpendicular to the layers and directed from layer 2 toward layer 12. The simulations assume that the light-emitting device is driven at a forward current density of about 200 Amps / centimeter² (A/cm²). The horizontal axes in <figref idref="f0001">Figures 1A and 1B</figref> represent position in the device in a direction perpendicular to the layers. The interfaces between the layers are indicated by dashed lines. The vertical axes in <figref idref="f0001">Figures 1A and 1B</figref> represent the energy of the conduction band edge E_c and of the valence band edge E_v in the various layers.

In <figref idref="f0001">Figure 1A</figref>, the simulated band structure accounts for polarization fields in n-layer 2 and active region 4, but neglects the spontaneous and piezoelectric polarization in p-type Al_yGa_1-yN electron confinement layer 10. Hence, this simulation is not physically realistic. By comparison to <figref idref="f0001">Figure 1B</figref>, however, it will serve to illustrate the importance of the polarization field actually present in confinement layer 10 in prior art devices. The tilting of band edges E_c and E_v in the quantum wells and barrier layers of active region 4 as shown in <figref idref="f0001">Figure 1A</figref> has been previously recognized to degrade the performance of such III-Nitride light-emitting devices. See, for example, <patcit id="pcit0001" dnum="US912589A" dnum-type="L"><text>U.S. Patent Application Serial No. 09/912,589</text></patcit>, titled "Light Emitting Diodes with Graded Composition Active Regions," assigned to the assignee of the present invention. See also International Patent Application <patcit id="pcit0002" dnum="WO0141224A2"><text>WO 01/41224 A2</text></patcit>.

In <figref idref="f0001">Figure 1B</figref>, the simulated band structure accounts for spontaneous and piezoelectric polarization fields in p-type Al_yGa_1-yN electron confinement layer 10 as well as for polarization fields in layer 2 and in active region 4. In particular, this simulation includes a negative sheet charge located at the interface of electron confinement layer 10 and active region 4, and a positive sheet charge located at the interface between confinement layer 10 and contact layer 12.

A comparison of <figref idref="f0001">Figures 1A and 1B</figref> shows that the effect of the sheet charges around Al_yGa_1-yN electron confinement layer 10 in <figref idref="f0001">Figure 1B</figref> is to reduce the magnitude of the potential energy barrier 14 that inhibits electrons from diffusing out of active region 4 into electron confinement layer 10. This reduction in electron confinement due to the polarization field in electron confinement layer 10 causes the quantum efficiency (ratio of photons out to electrons in) of the prior art device illustrated in <figref idref="f0001">Figure 1B</figref> to be lower than the quantum efficiency of the hypothetical device (having no polarization field in layer 10) illustrated in <figref idref="f0001">Figure 1A</figref>, particularly at high drive current densities.

The difference in quantum efficiencies between the two devices is apparent in <figref idref="f0002">Figure 1C</figref>, in which the quantum efficiency 16 of the prior art device of <figref idref="f0001">Figure 1B</figref> and the quantum efficiency 18 of the hypothetical device of <figref idref="f0001">Figure 1A</figref> are plotted as a function of forward drive current density. These quantum efficiencies were simulated by conventional techniques. As <figref idref="f0002">Figure 1C</figref> indicates, at a forward current density of about 400 A/cm² the hypothetical device of <figref idref="f0001">Figure 1A</figref> in which electron barrier layer 10 has no polarization field is three times as efficient as the prior art device of <figref idref="f0001">Figure 1B</figref> in which electron barrier layer 10 is polarized.

Since the tilting of the band edges in the active region is substantially the same for the devices of <figref idref="f0001">Figures 1A and 1B</figref>, the drop in quantum efficiency of the prior art device shown in <figref idref="f0001">Figure 1B</figref> at high drive current densities must be primarily due to the reduction in height of potential barrier 14 by the polarization field in electron confinement layer 10. The inventors believe that, in contrast to the effects of polarization fields in the active region, the decrease in electron confinement and hence in quantum efficiency due to the polarization field in the electron confinement layer has not been previously recognized. Moreover, polarization fields in n-type hole confinement layers may similarly reduce hole confinement and hence quantum efficiency. The inventors believe that the latter problem also has gone unrecognized.

<patcit id="pcit0003" dnum="US2003085409A1"><text>US 2003/085409 A1</text></patcit> discloses a III-nitride light emitting device and <patcit id="pcit0004" dnum="US651513B1"><text>US 651513 B1</text></patcit> discloses an optical emitter with an active region sandwiched between cladding layers, with the active region and the cladding layers between n-type and p-type contact layers.

What is needed is a III-Nitride light-emitting device having improved confinement of electrons and holes to its active region.

SUMMARY

The invention is defined in claim 1 and preferred embodiments of the invention are defined in the dependent claims.

The light-emitting semiconductor device comprises a III-Nitride active region and a III-Nitride layer formed proximate to the active region and may have a thickness that exceeds a critical thickness for relaxation of strain in the III-Nitride layer. The III-Nitride layer may be a carrier confinement layer, for example. The piezoelectric polarization in the III-Nitride layer may be reduced as a consequence of its thickness. The reduced polarization may result in an advantageous increase in the efficiency with which the device emits light.

According to the invention, a light-emitting semiconductor device comprises a III-Nitride light emitting layer, an In_xAl_yGa_1-x-yN (0.115 ≤ x ≤ 0.15, 0.2 ≤ y ≤ 0.35, x + y ≤ 1), and a spacer layer interposing the light emitting layer and the In_xAl_yGa_1-x-yN layer. The spacer layer is formed of GaN, InGaN or AlGaN. The spacer layer may advantageously space the In_xAl_yGa_1-x-yN layer and any contaminants therein apart from the light emitting layer. The composition of the III-Nitride layer may be advantageously selected to determine a strength of an electric field in the III-Nitride layer and thereby increase the efficiency with which the device emits light.

BRIEF DESCRIPTION OF THE DRAWINGS

<ul id="ul0001" list-style="none" compact="compact"><li><figref idref="f0001">Figure 1A</figref> shows a simulated band structure for a portion of a hypothetical III-Nitride light-emitting device including an unpolarized electron confinement layer.</li><li><figref idref="f0001">Figure 1B</figref> shows a simulated band structure for a portion of a prior art III-Nitride light-emitting device including a polarized electron confinement layer.</li><li><figref idref="f0002">Figure 1C</figref> compares simulated quantum efficiencies for the devices of <figref idref="f0001">Figures 1A and 1B</figref> as a function of drive current density.</li><li><figref idref="f0003">Figure 2</figref> is a schematic diagram of a III-Nitride light-emitting device according to one embodiment.</li><li><figref idref="f0004">Figure 3A</figref> shows a simulated band structure for a portion of a III-Nitride light-emitting device including a relaxed electron confinement layer according to one embodiment.</li><li><figref idref="f0005">Figure 3B</figref> compares simulated quantum efficiencies for the prior art device of <figref idref="f0001">Figures 1B</figref> and the device of <figref idref="f0004">Figure 3A</figref>.</li><li><figref idref="f0004">Figure 4</figref> shows the magnitude, in units of MV/cm, of the electric field across a layer of In_xAl_yGa_1-x-yN grown on GaN as a function of composition.</li><li><figref idref="f0006">Fig. 5</figref> shows the magnitude, in units of C/cm², of the net polarization in a layer of In_xAl_yGa_1-x-yN grown on GaN as a function of composition, and the energy gap, in units of electron volts (eV), of the In_xAl_yGa_1-x-yN layer.</li><li><figref idref="f0007">Figure 6A</figref> shows a simulated band structure for a portion of a III-Nitride light-emitting device including a quaternary In_xAl_yGa_1-x-yN electron confinement layer according to one embodiment.</li><li><figref idref="f0007">Figure 6B</figref> compares simulated quantum efficiencies for the prior art device of <figref idref="f0001">Figures 1B</figref> and the device of <figref idref="f0007">Figure 6A</figref>.</li><li><figref idref="f0008">Figure 7A</figref> shows a simulated band structure for a portion of a III-Nitride light-emitting device including a quaternary In_xAl_yGa_1-x-yN hole confinement layer according to one embodiment.</li><li><figref idref="f0008">Figure 7B</figref> compares simulated quantum efficiencies for the prior art device of <figref idref="f0001">Figures 1B</figref> and the device of <figref idref="f0008">Figure 7A</figref>.</li><li><figref idref="f0009">Figs. 8 and 9</figref> are a plan view and a cross sectional view of a large junction light emitting device according to embodiments of the invention.</li><li><figref idref="f0010">Fig. 10</figref> is an exploded view of a packaged light emitting device. It should be noted that the dimensions in the figures are not necessarily to scale. Like reference numbers in the various figures denote like parts in the various embodiments.</li></ul>

DETAILED DESCRIPTION

Several embodiments are disclosed in which polarization fields are reduced in electron and/or hole confinement layers in III-Nitride light-emitting devices. Confinement of electrons and holes in the active regions of these devices may thereby be improved. This can result in devices with improved quantum efficiency. The III-Nitride light-emitting devices described may be, for example, light-emitting diodes or laser diodes.

In one embodiment (<figref idref="f0003">Figure 2</figref>), a III-N based LED 20 includes a sapphire substrate 22, a nucleation layer 24, an optional n-type III-Nitride contact layer, an n-type III-Nitride hole confinement layer 26, a III-Nitride active region 28, a p-type III-Nitride electron confinement layer 30, and an optional p-type III-Nitride contact layer 32. Active region 28 includes one or more quantum well layers and one or more barrier layers. Active region 28 may be spaced apart from electron confinement layer 30, hole confinement layer 26, or both by optional spacer layers 27 and 29. N-type ohmic electrode 34 may be disposed on n-type contact layer 23 or n-type hole confinement layer 26. P-type ohmic electrode 36 may be disposed on p-type contact layer 32 or p-type electron confinement layer 30. Application of a suitable forward bias across ohmic electrodes 34 and 36 results in emission of light from active region 28.

The various semiconductor layers in light-emitting device 20 may be deposited by, for example, conventional metal-organic chemical vapor deposition (MOCVD) techniques. Conventional photolithographic, etching, and deposition techniques may be used to form electrodes 34 and 36.

The inventors have recognized that the piezoelectric polarization in electron confinement layer 30 can be advantageously reduced by growing this layer to a thickness exceeding its critical thickness for relaxation. The critical thickness is the thickness beyond which sufficient dislocations form in electron confinement layer 30 to substantially relax the strain induced in it by the difference in lattice dimensions between it and the layers underneath it. A layer that is thicker than its critical thickness is typically referred to in the art as a "relaxed layer." The critical thickness for relaxation depends upon the particular growth conditions of the layer and upon the alloy composition of the layer. For example, the critical thickness of a layer of Al_yGa_1-yN with y=0.2 is 700 Å to 900 Å, according to Shen et al. and Hiroki et al., reported in <nplcit id="ncit0001" npl-type="s"><text>B. Shen, T. Someya, Y. Arakawa, "Dependence of strain relaxation of the AlxGa1-xN barrier on transport properties of the two-dimensional electron gas in modulation-doped AlxGa1-xN/GaN heterostructure," Applied Physics Letters, vol. 76, pp. 2746-2748 (2000</text></nplcit>) and in <nplcit id="ncit0002" npl-type="s"><text>M. Hiroki, N. Maeda, N. Kobayashi, "Metalorganic vapor phase epitaxy growth of AlGaN/GaN heterostructures on sapphire substrates," Journal of Crystal Growth, vol .237, pp. 956-960 (2002</text></nplcit>). The critical thickness of a layer of Al_yGa_1-yN with y=0.4 is approximately 400 Å, according to Hiroki et al.

In some implementations, electron confinement layer 30 is grown to a thickness beyond its critical thickness and is consequently a relaxed layer. In such implementations, the magnitude of the sheet charges at either side of electron confinement layer 30 is reduced to that caused by the spontaneous polarization alone. The resulting magnitude is generally about half of the value that includes the effect of strain (i.e., the piezoelectric polarization).

In one implementation (<figref idref="f0004">Figure 3A</figref>), for example, electron confinement layer 30 is formed from p-type Al_yGa_1-yN (y=0.2) and has a thickness of 800 Å (which is greater than its critical thickness of about 700 Å), contact layer 32 is formed from p-type GaN, spacer layer 29 is formed from n-type GaN, and hole confinement layer 26 is formed from n-type GaN. In this implementation, active region 28 includes In_xGa_1-xN (x=0.12) quantum well layers separated by GaN barrier layers. The detailed structure of active region 28 is not critical, however. For example, other implementations may include either more or fewer such quantum well and barrier layers. In the implementation shown in <figref idref="f0004">Figure 3A</figref>, each layer other than electron confinement layer 30 is grown to a thickness less than its critical thickness.

The simulated band structure shown in <figref idref="f0004">Figure 3A</figref> accounts for the polarization fields in all of the layers shown and assumes that the light-emitting device is driven at a forward current density of about 200 A/cm². The magnitude of the sheet charge on either side of electron confinement layer 30 has been reduced to the value due to spontaneous polarization alone. Consequently, electron confinement layer 30 in this implementation provides a larger potential barrier 14 to diffusion of electrons out of the active region than is provided by the strained prior art electron confinement layer 10 shown in <figref idref="f0001">Figure 1B</figref>.

In <figref idref="f0005">Figure 3B</figref>, the quantum efficiency 54 of the novel device illustrated in <figref idref="f0004">Figure 3A</figref> and the quantum efficiency 16 of the prior art device illustrated in <figref idref="f0001">Figure 1B</figref> are plotted as a function of forward drive current density. These quantum efficiencies were simulated by conventional techniques. Comparison of the two quantum efficiency curves demonstrates that even a partial reduction in the magnitude of the polarization sheet charge in the electron confinement layer (achieved, in the device of <figref idref="f0004">Figure 3A</figref>, by allowing electron confinement layer 30 to relax) can result in substantial improvement in the high-current quantum efficiency.

In other implementations, hole confinement layer 26 is grown to a thickness beyond its critical thickness and is therefore a relaxed layer. In such implementations, the magnitude of the sheet charges at either side of hole confinement layer 26 is reduced to that caused by the spontaneous polarization alone. Consequently, the confinement of holes to the active region is advantageously improved compared to prior art devices not utilizing relaxed III-Nitride hole confinement layers.

Some implementations may include both a relaxed hole confinement layer 26 and a relaxed electron confinement layer 30.

In contrast to the present disclosure, one of ordinary skill in the art typically avoids growing any layer near the active region in a III-Nitride light-emitting device to a thickness beyond the layer's critical thickness, as the resulting dislocations are generally expected to reduce the quantum efficiency of the device by forming nonradiative recombination centers for electrons and holes injected into the device. Also, as reported by Hiroki et al. and Shen et al., the electron mobility in or near layers that are relaxed tends to be much lower. The inventors have recognized that despite the possibility of nonradiative recombination at dislocations, the use of a relaxed confinement layer in a III-Nitride light-emitting device may be advantageous.

The inventors have also recognized that the polarization fields in electron confinement layer 30 and/or in hole confinement layer 26 may be advantageously reduced, eliminated, or changed in sign by forming these layers from quaternary In_xAl_yGa_1-x-yN alloys of particular compositions rather than from Al_yGa_1-yN alloys as used in the prior art. (Quaternary In_xAl_yGa_1-x-yN alloys include all three of In, Al, and Ga. That is, x &gt; 0, y &gt; 0, and x + y &lt; 1.) An additional advantage to the use of quaternary In_xAl_yGa_1-x-yN confinement layers is that they may be lattice-matched to GaN layers such as, for example, GaN barrier layers in the active region. Such lattice matching facilitates the formation of nitride semiconductor layers of improved crystallinity.

<figref idref="f0004">Figure 4</figref> illustrates the magnitude, in units of megavolts / centimeter (MV/cm), of the total polarization field (piezoelectric polarization + spontaneous polarization) across a layer of quaternary In_xAl_yGa_1-x-yN grown on GaN. On the vertical axis is the mole fraction of aluminum in the layer, and on the horizontal axis is the mole fraction of indium in the layer. The lines with numbers in them are isometric lines of field magnitude. Along the vertical axis is represented the case of the zero indium mole fraction (i.e., the case of Al_yGa_1-yN). At any particular point on the vertical axis, as one moves horizontally to the right, the indium mole fraction increases, and the magnitude of the field decreases, eventually switching sign. The sign change in the total polarization field occurs because the sign of the piezoelectric polarization changes as the strain in the In_xAl_yGa_1-x-yN layer varies from compressive to tensile with increasing mole fraction of indium. The total polarization is zero for compositions in which the spontaneous and piezoelectric polarizations cancel each other. (<figref idref="f0004">Figure 4</figref> is taken from<nplcit id="ncit0003" npl-type="s"><text> V. Fiorentini et al., Physical Review B, vol. 60, pp. 8849-8858, 1999</text></nplcit>, in which it is identified as <figref idref="f0003">figure 2</figref>.)

The height of the potential barrier to carrier diffusion provided by an In_xAl_yGa_1-x-yN confinement layer is affected by both the net polarization in the layer and the bandgap of the In_xAl_yGa_1-x-yN material, which decreases with increasing indium mole fraction. <figref idref="f0006">Figure 5</figref> shows the variation with composition of the energy gap of an In_xAl_yGa_1-x-yN confinement layer grown on a GaN layer, and the variation with composition of the net polarization in the In_xAl_yGa_1-x-yN confinement layer. The net polarization in each layer is the sum of the spontaneous polarization and the piezoelectric polarization. A difference in net polarization between two layers results in a sheet charge at the interface between the two layers. The larger the difference in net polarization, the larger the sheet charge.

According to the invention electron confinement layer 30 (<figref idref="f0003">Figure 2</figref>) is formed from a quaternary In_xAl_yGa_1-x-yN alloy having a mole fraction of indium selected to determine the strength of the polarization field in the layer and thereby increase the potential barrier 14 to electron diffusion provided by the layer. In such implementations, the sheet charges at either side of electron confinement layer 30 may be eliminated, reduced in magnitude, or changed in sign. The confinement of electrons to active region 28 and the quantum efficiency of light-emitting device 20 may be thereby improved. Suitable indium and aluminum mole fractions include to those in the range of 0.115 - 0.15and 0.2 - 0.35, respectively. Optimal compositions of the quaternary In_xAl_yGa_1-x-yN confinement layer are determined by maximizing the confinement potential. The confinement potential is increased by a reduction in the sheet charge, and by an increase in the energy gap. Compositions of In_xAl_yGa_1-x-yN which are advantageous for use as electron confinement layers are any of those compositions which have an energy gap greater than approximately 3.5 eV. In addition, the composition of the In_xAl_yGa_1-x-yN electron confinement layer may be selected such that the difference between the net polarization in the electron confinement layer and the net polarization in an adjacent layer (a quantum well of the active region or spacer layer disposed between the electron confinement layer and the active region) is less than about 0.02 C/cm², more preferably less than about 0.01 C/cm². Compared to an AlGaN confinement layer with the same energy gap, a quaternary In_xAl_yGa_1-x-yN confinement layer with any amount of indium will be advantageous.

In one implementation, for reference and not being part of the invention, for example, the confinement layer 30 is formed of In_xAl_yGa_1-x-yN where x = 0.106 and y = 0.3. In this case, the energy gap of the In_xAl_yGa_1-x-yN layer is the same as the energy gap of an Al_yGa_1-yN layer where y=0.2, but the polarization-induced sheet charge is smaller. The band structure of this implementation is illustrated in <figref idref="f0007">Figure 6A</figref>. In <figref idref="f0007">Figure 6B</figref>, the quantum efficiency of the novel device 60 illustrated in <figref idref="f0007">Figure 6A</figref> and the quantum efficiency 16 of the prior art device illustrated in <figref idref="f0001">Figure 1B</figref> are plotted as a function of forward drive current density. These quantum efficiencies were simulated by conventional techniques. Comparison of the two quantum efficiency curves demonstrates that even a partial reduction in the difference between the net polarization sheet in the electron confinement layer and the net polarization in an adjacent layer (achieved, in the device of <figref idref="f0007">Figure 6A</figref>, by selecting an optimal composition of In_xAl_yGa_1-x-yN which has the same energy gap as the Al_yGa_1-yN confinement layer illustrated in the prior art device of <figref idref="f0001">Figure 1B</figref>) can result in substantial improvement in the high-current quantum efficiency.

In another implementation, according to the invention, the confinement layer 30 is formed of In_xAl_yGa_1-x-yN where x = 0.115 and y = 0.2. In this case, the energy gap of the In_xAl_yGa_1-x-yN layer is the same as the energy gap of an Al_yGa_1-yN layer where y=0.06, but the difference in net polarization between the In_xAl_yGa_1-x-yN layer and an adjacent layer is nearly zero. The effect of the small energy gap of this layer on the confinement potential is offset by the effect of the small polarization sheet charge. The band structure of this implementation is illustrated in <figref idref="f0008">Figure 7A</figref>. In <figref idref="f0008">Figure 7B</figref>, the quantum efficiency of the novel device 70 illustrated in <figref idref="f0008">Figure 7A</figref> and the quantum efficiency 16 of the prior art device illustrated in <figref idref="f0001">Figure 1B</figref> are plotted as a function of forward drive current density. These quantum efficiencies were simulated by conventional techniques. Comparison of the two quantum efficiency curves demonstrates that a partial reduction in the magnitude of the polarization sheet charge in the electron confinement layer can result in substantial improvement in the high-current quantum efficiency even in cases where the energy gap of the confinement layer is smaller than that typically used for prior art devices.

Similarly, in some implementations hole confinement layer 26 (<figref idref="f0003">Figure 2</figref>) is formed from a quaternary In_xAl_yGa_1-x-yN alloy having a mole fraction of indium selected to determine the strength of the polarization field in the layer and thereby increase the potential barrier to hole diffusion provided by the layer. In such implementations, the sheet charges at either side of electron confinement layer 26 may be eliminated, reduced in magnitude, or changed in sign. The confinement of holes to active region 28 and the quantum efficiency of light-emitting device 20 may be thereby improved. It is particularly important for a hole confinement layer that the polarization sheet charge be small. Consequently, the optimal composition of a quaternary In_xAl_yGa_1-x-yN hole confinement layer is one with an energy gap greater than 3.5 eV, and one where the difference between the net polarization in the hole confinement layer and the net polarization in an adjacent layer is less than 0.02 C/cm², more preferably less than about 0.01 C/cm².

The aluminum in quaternary III-Nitride confinement layers may promote contamination of these layers with oxygen. Such oxygen contaminants may enhance nonradiative recombination of electrons and holes in the quaternary III-Nitride layers and thereby reduce the quantum efficiency of a light-emitting device in which they are included. Hence, it may be advantageous to space quaternary III-Nitride confinement layers apart from light-emitting layers (e.g., quantum wells) in the active region of the device. Such spacing apart is accomplished, for example, with spacer layers 27, 29 included in the active region. Spacer layers 27 and 29 may have a thickness between about 50 angstroms and about 200 angstroms. A common thickness is about 100 angstroms. Spacer layers 27 and 29 are usually n-type layers, though in some embodiments spacer layer 29 may be p-type. The spacer layers are GaN or ternary nitride alloys of aluminum, indium, and gallium. For example, in the implementation illustrated in <figref idref="f0007">Figure 6A</figref>, a GaN spacer layer 29 is disposed between electron confinement layer 30 and the quantum well layers 28. Similarly, in the implementation illustrated in <figref idref="f0008">Figure 7A</figref>, a GaN spacer layer 29 is disposed between electron confinement layer 30 and the quantum well layers 28. In other implementations, though, quaternary III-Nitride confinement layers may be in contact with light-emitting layers.

<figref idref="f0009">Fig. 8</figref> is a plan view of a large junction device (i.e. an area greater than or equal to 0.2 square millimeters). <figref idref="f0009">Fig. 9</figref> is a cross section of the device shown in <figref idref="f0009">Fig. 8</figref>, taken along axis DD. <figref idref="f0009">Figs. 8 and 9</figref> illustrate an arrangement of contacts that may be used with any of the epitaxial structures 40 illustrated in <figref idref="f0003">Figs. 2</figref>, <figref idref="f0004">3A</figref>, <figref idref="f0007">6A</figref>, <figref idref="f0008">7A</figref>, and described above. The active region of epitaxial structure 40 is separated into four regions separated by three trenches in which n-contacts 34 are formed. Each region is connected to a submount by four p-submount connections 44 formed on p-contact 36. N-contact 34 surrounds the four active regions. N-contact 34 is connected to a submount by six n-submount connections 46. The n- and p-contacts may be electrically isolated by an insulating layer 42.

<figref idref="f0010">Fig. 10</figref> is an exploded view of a packaged light emitting device. A heat-sinking slug 100 is placed into an insert-molded leadframe. The insert-molded leadframe is, for example, a filled plastic material 105 molded around a metal frame 106 that provides an electrical path. Slug 100 may include an optional reflector cup 102. The light emitting device die 104, which may be any of the devices described above, is mounted directly or indirectly via a thermally conducting submount 103 to slug 100. A cover 108, which may be an optical lens, may be added.

While the present invention is illustrated with particular embodiments, the invention is intended to include all variations and modifications falling within the scope of the appended claims. Referring again to <figref idref="f0003">Figure 2</figref>, for example, substrate 22 may be formed from materials other than sapphire, such as SiC. Electrode 36 may be transparent or semitransparent to light emitted by active region 28. Alternatively, electrode 36 may be highly reflective to light emitted by active region 28, and light-emitting device 20 may be mounted as a flip-chip with contacts 34 and 36 facing a submount.

Moreover, III-Nitride electron and hole confinement layers having reduced polarization fields as disclosed herein may be formed in other III-Nitride light-emitting devices such as, for example, those described in <patcit id="pcit0005" dnum="US6133589A"><text>U. S. Patent 6,133,589</text></patcit> and in <patcit id="pcit0006" dnum="US6486499B"><text>U.S. Patent 6,486,499</text></patcit>, both of which are assigned to the assignee of the present invention. In addition, light-emitting devices in accordance with the present invention may include layers not shown in the illustrated embodiments and may omit illustrated layers such as contact layer 32 (<figref idref="f0003">Figure 2</figref>), for example. The composition of the various III-Nitride layers in the inventive light-emitting devices (such as the contact, quantum well, and barrier layers, for example) may vary from the example compositions disclosed.