WO1996011502A1

WO1996011502A1 - WAVELENGTH TUNING OF GaN-BASED LIGHT EMITTING DIODES, LIGHT EMITTING DIODE ARRAYS AND DISPLAYS BY INTRODUCTION OF DEEP DONORS

Info

Publication number: WO1996011502A1
Application number: PCT/IB1995/000595
Authority: WO
Inventors: Samuel C. Strite
Original assignee: International Business Machines Corporation
Priority date: 1994-10-11
Filing date: 1995-07-31
Publication date: 1996-04-18

Abstract

Light emitting diodes with an active region which comprises Galliium Nitride (GaN) being doped with acceptors (A) and deep donors (DD) having a binding energy in said Gallium Nitride host material of at least 50meV and leading to an impurity to impurity transition of increased wavelength.

Description

DESCRIPTION

Wavelength Tuning of GaN-based Light Emitting Diodes,

Light Emitting Diode Arrays and Displays by

Introduction of Deep Donors

TECHNICAL FIELD

The present invention concerns a method for tuning the wavelength of discrete gallium nitride (GaN) semiconductor light emitting diodes (LEDs), arrays of such diodes, lamps and displays comprised of these LEDs.

BACKGROUND OF THE INVENTION

Light emitting diodes have many important functions. Different applications call for LEDs that emit light within a narrow wavelength band ranging from ultraviolet, blue, into the green and longer wavelengths. Furthermore, LEDs should be bright to ensure better viewability, and efficient to lower power consumption. The light emission should be uniform and have a small divergence angle so that the light can be better collimated or collected. Another aspect is manufacturability and cost if one is planning to enter the mass-market. Discrete LEDs, as well as LED arrays and displays must be simple for mass production at low cost. Ideally, multicolor arrays and displays should be made from LEDs of one material system to permit monolithic integration onto a single chip.

In the majority of LED technologies, the active region is placed between semiconductor cladding layers, one being doped p-type and the other n-type. The optical transitions are induced by injecting electrons and holes into the active layer via an applied bias across the cladding layers. An important and sometimes restrictive premise of this approach is the existence of proper cladding materials which can be doped p- as well as n-type and can serve as substrates for the fabrication of high quality active region.

The earliest LEDs were made of direct bandgap binary semiconductor homojunctions. The wavelength of the light emitted was given by the semiconductor bandgap. Severe limitations on this kind of LED were the inability to tune and the paucity of direct bandgap semiconductors at useful wavelengths, especially in the visible range.

By adding impurities to indirect, wider bandgap semiconductor materials, visible light at low efficiency could be realized. Typical examples are Gallium Phosphide (GaP) doped with Nitrogen (N) emitting green light, and GaP doped with Zinc Oxide (ZnO) leading to emission of red light and SiC doped with AIN gives blue light.

Until recently, the majority of direct wide bandgap materials could not be grown both p- and n-doped. Therefore, LEDs based on conventional p-n-junctions for carrier injection into the active region were not feasible. To circumvent this inconvenience, unipolar MlS-type (metal insulator semiconductor) diodes were successfully applied.

In MlS-type LEDs, the active layer is made insulating by deep impurity levels and sandwiched between an n-type conductive semiconductor layer and a metal contact. By applying an appropriate bias between the metal and conductive semiconductor layer, electrons are injected into the active layer. Once in the active layer, the injected electrons radiatively recombine with holes al the impurity sites, which are refreshed by the counter electrode or impact ionization. Such structures show typical diode-like nonlinear current-voltage characteristics including a threshold voltage and an exponential increase of the injected current as a function of the applied bias. MIS LEDs emitting all visible colors have been realized in the GaN system. Examples of these have been published in:

• "Violet luminescence of Mg-doped GaN" by H. P. Maruska et al., Applied Physics Letters, Vol. 22, No. 6, pp. 303-305, 1973;

• "Blue-Green Numeric Display Using Electroluminescent GaN" by J. I. Pankove, RCA Review, Vol. 34, pp. 336-343, 1973;

• "Electric properties of GaN: Zn MlS-type light emitting diode" by M. R. H. Khan et al., Physica B 185, pp. 480-484, 1993;

• "GaN electroluminescent devices: preparation and studies" by G. Jacob et al.. Journal of Luminescence, Vol. 17, pp. 263-282, 1978;

• EP-0-579 897 A1 : "Light-emitting device of gallium nitride compound semiconductor".

In these documents, a common substrate for GaN is used, namely sapphire. On the sapphire substrate, a thick (several 100μm) layer of n-type GaN was grown, often unintentionally doped. On top of the n-GaN layer, the active layer of insulating GaN was formed. The insulating nature was realized by the incorporation of deep acceptors such as Zn, Cd or Mg during growth which compensate intrinsic donors and thus reduce the conductivity. Metals such as In, Ni, Ag, or Al served as metal contacts to the insulating active layer. As the sapphire substrate is insulating, special effort is necessary to apply a bias to the MIS-diode. For making a contact to the n-GaN layer, either side contacts at the edges of the substrate are formed, or the n-GaN layer is made accessible from above by etching contact holes through the insulating GaN active layer. Further improvement in growth led to semiconductor alloys. The wavelengths of devices based on semiconductor alloys can be tuned within a limited range by variation of the mole fraction of the alloy components. By introduction of quantum well structures the efficiency has been further improved . The variation of the alloy composition is limited, however, by lattice matching.

Devices of new shorter wavelength direct bandgap materials opened up the visible wavelengths to high efficiency. Examples are InGaP and AIGaAs which emit in the red, ZnSe based devices emitting in the green-blue wavelength range, InGaN p-n junction devices emitting in the violet, and InAIGaP emitting orange and yellow light.

Differing approaches aimed at lengthening the emission of GaN based LEDs have been reported in the above-mentioned documents. Each approach relies on the introduction of an acceptor impurity which has a large binding energy ( > 100 meV), such as Zn, Cd, Mg, Hg, Li, Be, Ca and other elements.

The acceptors compensate any free electrons, making the GaN material insulating. Furthermore, the dominant electroluminescent transition, which occurs between the conduction and valence bands in non-compensated material, is channelled through the acceptor energy level causing the wavelength to be shifted by an amount equal to the acceptor binding energy. Experiments have shown that the binding energy of the dominant conduction band to acceptor level transition depends on the dopant element, its concentration and/or the GaN growth conditions.

Based on this concept. GaN MIS LEDs have been fabricated with peak emission wavelengths in the violet, blue, green, yellow, orange and red parts of the spectrum, altogether spanning the entire visible spectrum, including the three primary colors. The state of the art is reviewed by Strife and Morkoc in "GaN, AIN, InN: A Review", Journal of Vacuum Science and Technology, B10, p.1237, 1992. GaN MIS LEDs have demonstrated quantum efficiencies as large as 0.5% and 0.1 % in the green-yellow and blue parts of the spectrum respectively. Threshold voltages for these devices have ranged from 4V for the blue, 5V for the green, and 10V for the yellow. Much of this work is in excess of 20 years old, and was achieved in material that was considerably inferior to what is grown today. Therefore, the values quoted above in no way represent the limits of GaN based MIS LED technology. Work on GaN MIS LEDs was largely discontinued with the development of improved crystal growth and processing techniques which realized high quality undoped and p-type GaN.

The establishment of p-type doping processes for GaN led to the immediate shifting of focus to p-n junction type GaN LEDs. The earliest examples of GaN p-n junction LEDs were homojunction devices as descπbed by I. Akasaki et al. in "Photoluminescence of Mg-doped p-type GaN and Electroluminescence of GaN p-n junction LED", Journal of Luminescence, Vol. 48 & 49, p. 666 - 670, 1991 and I. Akasaki et al. in "Growth of GaN and AIGaN for UV/blue p-n junction diodes", Journal of Crystal Growth. Vol. 128, p. 379 - 383, 1993. While this was the first reported electroluminescence from a conventional p-n diode LED structure in GaN, these LEDs are largely unusable because their emission was limited to the uv (conduction band to valence band transition) and violet (conduction band to Mg acceptor transition) spectra. Akasaki and Amano in "Perspective of the UV/Blue Light Emitting Devices Based on GaN and Related Compounds", Optoelectronics-Devices and Technologies, Vol. 7, No. 1 , p. 49 - 56, June 1992, reported the extension of GaN based LEDs to heterostructures with the realization of an AIGaN/GaN p-n junction LED. but this structure continued to emit at short wavelengths.

Nakamura et al. in "High-power InGaN/GaN double-heterostructure violet light emitting diodes", Applied Physics Letters, Vol. 62, No. 19, pp. 2390 - 2392, 10 May 1993, reported the first GaN-based LEDs which incorporated InGaN in the active region for the purpose of increasing the emission wavelength. However, InGaN alone was not sufficient to extend GaN to the technologically important blue wavelengths. The conduction to valence band transition in these LEDs produced violet light. Blue LEDs were realized, as described above, by introducing Zn into the InGaN active region, to further lengthen the emission by means of substituting a conduction band to acceptor transition for the conduction band to valence band transition. By adding yet more In into the active region, Nakamura et al. Journal of Applied Physics, "High-brightness InGaN/AIGaN double-heterostructure blue-green-light-emitting diodes", Vol. 76, No. 12, pp 8189 - 8191 , 15 December 1994, were able to further extend GaN based LEDs to 500 nm, but at the possible expense of device reliability. The strained nature of InGaN grown on GaN might make 500nm a practical limit. Further red-shifting of the emission wavelength must come from a technique other than further In introduction. Finally, reproducing the same InN mole fraction run after run introduces additional process complexity and is particularly undesirable for multicolor array applications. That is, changing the alloy composition is the most difficult method of tuning the emission wavelength. One method, previously known in the art is to replace Zn with a deeper acceptor level. Cd was introduced into InGaN for this purpose by Nakamura et al. Japanese Journal of Applied Physics, Vol. 32, L338, 1993, but for a given In mole fraction, the photoluminescence wavelength corresponding to Cd was no longer than that of Zn. As a result, GaN based LEDs have been limited by the necessity of using the Zn acceptor, and by the inability to incorporate more In into the active region without degrading device performance, and workers have been discouraged from pursuing alternative acceptors.

Another example representing the state of the art is given in "Candela-class high-brightness InGaN/AIGaN double-heterostructure blue-light-emitting diodes" by S. Nakamura et al., Applied Physics Letters, Vol. 64, No. 13, pp. 1687-1689, 1994. The vertical layer structure of the LED disclosed in this article consists of a stack of GaN/AIGaN/lnGaN layers grown on sapphire. The active layer consists of Zn doped InGaN sandwiched between p- and n-doped AIGaN layers, the sandwich forming a double-heterostructure. The Zn doping leads to optical transitions whose energy is related to the energy 6/11502 PCMB95/00595

of Zn-related impurity states. The bandgap diagram of this structure is illustrated in Figures 1A. To tune this device, the InGaN alloy having a smaller bandgap energy than GaN was introduced into the active region.

It was further observed that the device efficiency could be improved by codoping Si into the active region as well, by S. Nakamura in the article "InGaN/AIGaN blue-light-emitting diodes", Journal of Vacuum Science and Technology, A13(3), pp. 705 - 710, May/June 1995, and as illustrated in Figure 1B. However, it is to be noted that Si is a very shallow donor in GaN and that it has a negligible effect on the device wavelength.

In summary, GaN and the InGaN alloy is deemed to be the material of preference for fabricating discrete blue LEDs, LED arrays emitting light at any wavelength spanning the entire visible spectrum, and may even supplant some more mature semiconductors in the discrete LED markets for green, orange and yellow operation. Indeed, S. Nakamura et al. report such discrete LEDs with quantum well structures based on higher IN mole fraction InGaN in "High-Brightness InGaN Blue, Green, and Yellow Light-Emitting Diodes with Quantum Well Structures", Japanese Journal of Applied Physics, Vol. 34, pp. L797 - L799, 1995. Blue, green, and yellow light emission were achieved by S. Nakamura et al. by making use of different Indium mole fractions in quantum well structure LEDs. While GaN has been extended yet further into the visible by further increasing In content, this approach still suffers from potential reliability and reproducibility problems as discussed above.

It becomes obvious from the above, that GaN is a special semiconductor material in need of unique solutions to achieve improved brightness, and to consequentially exploit the inherent properties offered by the InAIGaN material system. In particular its extension from the uv and violet into the longer wavelength range towards green, and the fine-tuning of the emission wavelength so as to match the human eye's sensitivity, calls for new approaches. In addition, concepts to integrate different such LEDs with multicolor capability over the entire visible spectrum on a single substrate and using such LEDs for display applications are barely developed.

While all of the primary colors have been demonstrated in GaN LED structures, it is currently not clear that acceptors exist which correspond exactly to those wavelengths at which the human eye best perceives each color or is most sensitive. Furthermore, the currently known GaN-based diodes do not emit light at exactly the three most desirable wavelengths for a full color display. It would be a fortunate coincidence if indeed three ideal acceptors exist for a particular InGaN alloy which at the same time would lead to a sufficiently bright light emission. Clearly, an additional invention is required to expand the color gamut of GaN LEDs beyond what is readily available from known acceptor impurity transitions.

Two major challenges exist in adapting known approaches into workable discrete LEDs, LED arrays, and LED displays. The first is fabricating GaN diodes of sufficiently high brightness at the desired wavelengths which is extremely important for use in battery powered display applications, for example. The second is finding a suitable combination of GaN alloy composition and the respective acceptor impurities to produce certain, predefined colors, e.g. the three different wavelengths which appear to the human eye as the pure primary colors. The latter criteria is of particular interest in the case of multicolor LED arrays and LED displays. As discussed above, it is further desirable to achieve optimal colors in non-alloyed GaN to decrease crystal growth complexity and eliminate strained layers.

The concept of monolithic integration of an array of multicolor LEDs and details concerning the manufacturing of GaN diodes and arrays are addressed and claimed in the copending PCT patent application with application number PCT/EP 94/03346, which was filed on 1 1 October 1994, and the copending PCT patent application with application number PCT/IB 95/00367, which was "filed on 17 May 1995. These patent applications are incorporated by means of reference into the present description and their priority is herewith claimed.

It is an object of the present invention to provide means for red-shifting the emission wavelength of GaN-based LEDs and for achieving GaN-based LEDs emitting light at technologically important wavelengths.

It is an object pf the present invention to provide efficient and bright GaN LEDs for both discrete and array applications.

It is an object of the present invention to provide GaN LEDs emitting light at an exactly predefined wavelength matching the human eye's sensitivities to both color and brightness.

It is another object of the present invention to provide appropriate structures enabling such LEDs.

It is a further object to provide multicolor LED arrays or lamps for displays incorporating efficient and bright LEDs spanning a wide color gamut.

SUMMARY OF THE INVENTION

The above-mentioned objects have been achieved by provision of a structure as claimed in claim 1 .

By doping the ternary InGaN active region with appropriate acceptor donor pairs additional flexibility in tuning the LED wavelength is gained and the wide bandgap GaN host material system is extended to even longer wavelengths. The present doping approach leads to a new and additional degree of freedom in the design of LEDs, LED arrays, and LED displays. It not only offers additional flexibility in tuning the LED wavelength, but it also permits the GaN host material system to be extended into the green and longer wavelengths where it will outperform GaP, ZnSe and InGaAIP based devices.

In addition, the present doping approach facilitates improved LED arrays and displays where the optimal color gamut is a strong function of the three component wavelengths and thus leads to improved full color devices. It is also relevant to discrete LEDs because of the steep wavelength dependence of the eye sensitivity. A shift of 10-20 nm can make a mediocre LED appear much brighter.

The key for achieving the three optimum color components lies in finding the perfect match of impurities and semiconductor bandgap. However, it is inconvenient to modify the bandgap by variation of the alloy mole fraction. One would like to realize all three colors simply in GaN which is much simpler to reproduceably grow than say Alo isGao syN. I DESCRIPTION OF THE DRAWINGS

The invention is described in detail below with reference to the following schematic drawings:

FIG. 1A shows a bandgap diagram of a GaN heterostructure LED having a Zn doped InGaN active region, as known in the art.

FIG. 1B shows a bandgap diagram of another GaN heterostructure LED having a InGaN active region doped with Zn and Si, as known in the art.

FIG. 1 C shows a bandgap diagram of a GaN heterostructure LED having a InGaN active region with deep donor/acceptor pairs, in s accordance with the present invention, so as to emit light at a red-shifted wavelength with respect to the light emitted by the

LEDs of Figures 1A and 1B.

FIG. 1D shows a bandgap diagram of another GaN heterostructure LED 0 having a InGaN active region with deep donor/acceptor pairs, in accordance with the present invention, so as to emit light at a red-shifted wavelength with respect to the light emitted by the

LEDs of Figures 1A and 1B.

5 FIG. 2 shows a schematic cross section of a simple, discrete MlS-type

GaN LED with an active region comprising deep donor/acceptor pairs, according to the present invention.

FIG. 3 shows a schematic cross section of a discrete p-i-n type GaN 0 LED with an active region comprising deep donor/acceptor pairs, according to the present invention. FIG. 4 shows a schematic cross section of another discrete p-i-n type

FIG. 5 shows a schematic cross section of a discrete MlS-type GaN

LED on a conductive substrate with an active region comprising deep donor/acceptor pairs, according to the present invention.

FIG. 6 shows a schematic cross section of an array element with three vertical MlS-type LEDs on a non-conductive substrate, the active regions of which are doped with deep donor/acceptor pairs, according to the present invention.

GENERAL DESCRIPTION

In Figure 1 A, a GaN double heterostructure with an InGaN active region doped with deep Zn acceptors is shown. Due to the addition of In into the GaN compound and the introduction of Zn as acceptor the wavelength is shifted to a certain, but limited extent towards longer wavelengths. The result of the delicate balance between the InGaN mole fraction and the Zn acceptors is a pure blue of 450nm. However, if Zn were only slightly deeper, corresponding to 460nm at this InGaN mole fraction, the LED would appear markedly brighter.

The advantages of the codoping approach which is described above are increased efficiency when Si is introduced along with Zn in the InGaN active region This approach is schematically illustrated in Figure 1 B. The emission wavelength λ₂ of the structure illustrated in Figure 1 B is about the same as the one of Figure 1A λ It is to be noted that the shift in emission wavelengths due to the Si is in the range of a few meV only, which is not perceivable by the human eye. Such a shift of the emission wavelength is thus not of technical importance. Si is introduced purely to increase the efficiency by replacing the less efficient conduction band to Zn transition with the favored Si-Zn donor acceptor transition. In fact, as pointed out above, the inventive concept of introducing a deep donor impurity to further lengthen the wavelength is entirely unknown and heretofore undesired.

In Figure 1C, it is illustrated how longer wavelength operation of GaN based devices can be achieved according to the present invention. By introduction of a deep donor (DD), i.e. a donor whose level is substantially deeper than the one of the known shallow donors (SD), e.g. Si or Ge, the energy gap between the donor (DD) and acceptor (A) pair is reduced leading to an increased wavelength λ₃ λ₂, λ,. Choosing suitable deep donors, the principle emission wavelength can be shifted by more than 50meV, which is easily perceived by the human eye as both a color and intensity change.

A very important advantage of the present invention is the flexibility gained for color tuning of GaN-based LEDs. Tuning is no longer limited to the choice of an acceptor, but rather the choice of the deep donor acceptor pair. This is perhaps most clearly illustrated by a contrived example. Let us assume that of the 80 or more elements readily available for GaN doping experiments that it is determined that 10 form suitable donors and another 10 are suitable acceptors, each of the 20 having their own characteristic binding energy.

Using the conventional approach, to realize a given color, one is restricted to only those 10 acceptor impurities along with the freedom to tune via In mole fraction. The conventional approach becomes even more restrictive if one would like to realize a monolithic full color array as disclosed in the above-referenced co-pending patent applications. In this case, one must fix a single InGaN mole fraction, and then be fortunate enough to find 3 acceptors from the pool of only 10 in total, which correspond to red, green and blue at that InGaN concentration. While this is improbable, it is even a more remote possibility that 3 acceptors can be found which together correspond precisely to the wavelengths offering the widest color gamut. Using the inventive approach herein disclosed, one has at his disposal 100 possible pairs (10 deep donors X 10 acceptors) with which to experiment. It is much more likely that at a fixed InGaN concentration that three pairs can be identified which correspond not only to the three primary colors, but to an optimized color gamut as well. This concept can be seen in the following. One may wish to make an LED which is as bright as possible, and therefore desires an emission wavelength of 550 nm, corresponding to the maximum of the human eye's response. If an acceptor can be identified in a given InGaN mole fraction to emit light at 500 nm, then the designer must choose a deep donor whose binding energy corresponds to the desired 50nm shift in wavelength. If no such deep donor exists, then it should be possible to substitute an acceptor that emits light at 480nm together with a deeper donor known to have a binding energy which would yield the desired 70nm shift.

Another approach, in accordance with the present invention, is to introduce an acceptor (A) and deep donor (DD) pair precisely defining the emission wavelength and to provide shallow donors (SD) at the same time in the active region. This approach is illustrated in Figure 1 D. The shallow donor (SD), which is preferably Si, continues to assist efficiency by transporting carriers in the impurity rather than conduction band. From there the electrons are transferred to the deep donor (DD) level before recombining with the acceptor. The shallow donor, e.g. Si, does not noticeably contribute to the emission wavelength but increases the efficiency of light emission.

We stress that Zn acceptors are known to be efficient luminescent centers in GaN. It may be desirable to maintain Zn as the acceptor impurity, and seek a different tuning mechanism to extend to longer wavelengths. Nakamura has realized blue-green electroluminescence at 500 nm by increasing the In mole fraction, but it is not clear that such large mole fraction devices are reliable. The introduction of a deep donor impurity can accomplish the desired wavelength tuning, without the difficulties and attendant reliability questions of increased In content. Examples of elements known to be deeper donors than Si in GaN are: O, S, Se, and Ti. The binding energy of O, for example, has been estimated to be roughly 80meV, as described by Chung and Gerschenzon in "The influence of oxygen on the electrical and optical properties of GaN crystals grown by metalorganic vapor phase epitaxy", Japanese Journal of Applied Physics, Vol. 72, p. 651 - 659, 1992. Other promising materials are the elements of groups 1A through VINA and VIB. In the present context, we deem a donor to be deep if it has a binding energy in excess of 50meV. This corresponds to an easily visible shift of the LED emission wavelength at detected by the human eye, and at a constant device quantum efficiency, a corresponding shift in the brightness as a result of the eye's sharply peaked sensitivity from blue to green. (Note; to go from orange to red, the eye becomes less sensitive again).

Many acceptors are known in GaN, all of which are deep as defined above. The most important of the acceptors are Mg, Zn, Hg, Cd, Li, Be, C, P, Ag and the elements of the groups IIA, IIB and IA of the periodic table, but other elements are also known to introduce deep acceptor levels into GaN.

The introduction of deep donors is not obvious, because investigators have pursued the modification of the In mole fraction in the InGaN active region to shift the emission towards longer wavelengths, and secondly, investigations of GaAs being doped with donor/acceptor pairs led to light emission at drastically reduced efficiency, i.e. there was a prejudice against this codoping approach. Wavelength tuning by variation of the In mole fraction is not optimal because InN is lattice mismatched with GaN. High In content InGaN layers are more likely to have dislocation formation and motion leading to device degradation and failure. Substituting a deeper donor for Si, according to the present approach, permits longer wavelength operation with no additional In, so that the device will perform the same at 550nm as at 450nm. In addition, it is deemed to be easier to chose the right deep donor acceptor pairs instead of modifying the ternary material composition during the epitaxy. For sake of convenience, the basic concepts of the present invention is now described in connection with different embodiments. Some of these embodiments are shown in the drawings by means of schematic cross-sectional views, whereas others are described in connection with a band diagram of the layers close to the active region.

In the following, discrete LEDs and an example of a multicolor LED array, based on the material system (Al_xGa, _ _x)_yln, _ _yN (0< x, y <1 ), are shown. This material systems has the advantage of offering multicolor capabilities with an extreme spectral width compared with materials having a narrower energy gap.

In Figure 2 a layered MlS-type semiconductor LED 20, based on (Al_xGaι _ _x)_yln, __yN, is shown This MlS-type GaN LED 20 mainly consists of a layered structure 22, 23 of semiconductor layers grown on top of a planar substrate 21. Several substrates are suitable for devices based on crystalline (Al_xGa^ _xJ_yln^ _yN, for example sapphire, SiC, Si, ZnO or AIGalnN. Sapphire, which is insulating, is the substrate traditionally used.

Later it will be described that one can take advantage of the good conductivity of SiC, Si and AIGalnN, because due to this properly, means for applying a bias to a LED, or a particular LED in a LED array can be simplified. Both substrates, sapphire and SiC, are transparent to visible light and permit the LEDs to be designed such that the light generated is preferably emitted through the substrate into the halfspace below the substrate (whereby the term 'below' corresponds to the backside of substrate, i.e. the side not being used for the deposition of the layers of semiconductors). Of course, LEDs and LED arrays based on any of the discrete LEDs herein described can be designed such that the electroluminescent light directed towards the halfspace above the substrate is used. High-quality crystalline layers or- (Al_xGa, _ _x)_yln, _ _yN can be grown by means of epitaxy methods such as metal organic vapor phase epitaxy (MOVPE) or molecular beam epitaxy (MBE). Typical growth conditions are for example described in:

• EP-0-579 897 A1 ,

• EP-0-551 721 A2,

• "GaN, AIN, and InN: A review" by S. Strite et al., Journal of Vacuum Science and Technology, Vol. B 10, pp. 1237-1266, 1992.

From these references, descriptions of standard device processing steps, such as etching processes, doping (AI„Ga, _ _x)_yln, _ _yN p- and n-type during and after crystal growth with a variety of dopants (e. g. Zn, Cd, Si, and Mg), and the formation of metal coatings resulting in either Schottky barriers or ohmic contacts can be taken. Such fabrication steps are considered as known and are hereinafter not discussed in detail.

The LED 20 shown in Figure 2 comprises a conductive n — (Al_xGa, _ _x)_yln, __yN layer 22 on top of the substrate 21 , and a further (Al_xGa, _ _x)_yln, _ _yN layer 23 serving as active layer. This layer 23 is doped with acceptors and deep donors, according to the present invention, in order to precisely define the LED s light emission wavelength. The concentrations are chosen so that the density of acceptors is larger than the density of donors. This insures both that the material is insulating and that holes exist at the impurity sites to promote efficient recombination. The first-grown n - (Al_xGa, __x)_yln, __yN layer 22 is either undoped and its n-type conductivity relies on native defects (i.e. unintentionally-doped), or its conductivity is further increased by n-doping, e.g. by adding shallow donors such as Si or Ge during growth. For fabrication of a MlS-type LED 20 different doping procedures are suitable, depending on the particular application. This doping can be performed either during or after growth of the active layer. For the thickness of the insulating layer 23 in Figure 2, trade-offs exist leading to optimized values. The thickness of this active layer 23 is typically in the range of 20nm - 1μm.

Figure 3 depicts a second layered structure, namely a discrete p-i-n type LED 30. The shown double heterostructure consists of a codoped active layer 33 (Al_xGa, _ _x)N sandwiched by two (AluGa, _ ,,), _ _vln_vN cladding layer 32 and (Al_mGa, _ _m)ι - _nln_nN cladding layer 34, one being p-doped, the other being n-doped. The upper cladding layer is structured so as to provide electrical confinement and a contact via to the lower cladding layer. The mole fractions x, u, v, m and n are chosen such that a heterobarrier occurs at the interfaces between active and cladding layers (e.g. with x = 0, v = 0, u = 0.5 m = 1 , n = 0.5). For n- and p-doping of the cladding layers, Si and Mg, respectively, can be used. The active layer is codoped with a deep donor and acceptor. Because in the pin LED, both electrons and holes are supplied, the deep donor and acceptor concentrations can be equal, or either may be larger than the other.

Examples of doping conditions which can be achieved and allow discrete LEDs, multicolor LED arrays, and LED displays, to be made in accordance with the present invention, are known from the article "Photoluminescence of ion-implanted GaN" by J. I. Pankove et al., Journal of Applied Physics, Vol. 47, No. 12, pp. 5387-5390, 1976, and the other references mentioned above in the context of GaN-based MIS-LEDs. Equivalent data for other members of the family (Al_xGa, __x)ι - _yln_yN can be considered as being a continuous function of x and y.

Another embodiment is now described. The cross-section of a discrete GaN heterostructure LED 40 is given in Figure 4. The LED 40 comprises a sapphire substrate 41 , for example, a buffer layer 42, e.g. made of GaN, a n-type GaN layer 43, a n-type Al_xGa₍, _ _X)N lower cladding layer 44 (preferably with: 0.1 < x < 0.3), an ln_yGa_(l _ _y)N active region 45 (preferably with: 0.01 < y < 0.2) being doped with deep donor acceptor pairs, an upper p-type Al_xGa₍, _ _x)N cladding layer 46 (preferably with: 0.1 < x < 0.3), and a highly p-doped GaN contact layer 47. There are two ohmic contacts 48.1 and 48.2 provided, one contacting the uppermost layer 47 and the other one contacting layer 43. Zn is used as acceptor and O as deep donor in GaN to achieve light emission at about 470nm, vs. 450nm for only Zn. If the substrate is conductive then the second contact can be formed to the bottom of the substrate. The layers can be grown using metal organic chemical vapor deposition (MOCVD) at atmospheric pressure, or other common techniques. In the present example Si is used for n-doping of the layers 43 and 44 and Mg is used for doping the upper cladding layer 46 and the contact layer 47. Ni/Au and Ti/AI are used as the material of contacts 48.1 and 48.2. The thickness of the layers of LED 40 are as follows:

Layer 42: between 10 and 100nm Layer 43: between 1 and 10 μm Layer 44: between 0.1 - 2 μm Layer 45: between 10 and 200nm Layer 46: between 0.1 - 1 μm Layer 47: between 0.1 - 1 μm

Another embodiment of the present invention is based on the schematic band diagram of Figure 1 D. This embodiment is illustrated in Figure 5. The substrate 51 of this LED 50 is conductive (e. g. SiC, Si, InGaAIN). In this case, the doped semiconductor layer 52 in combination with the substrate 51 serve together with the contact 55 as an electrode. Layer 53, serving as active layer, comprises n — (Al_xGa, _ _x)_yln, _ _yN and layer 52 comprises (Al_xGa, _ _x)_yln, _ _yN. The active layer 53 is doped with acceptors and deep donors, according to the present invention, in order to define the LED's light emission wavelength. Additionally, the active layer 53 comprises Si or another shallow donor (< 50meV) in order to increase the efficiency of the device 50. On top of the active layer 53 a top contact 54 is formed, the size of which defines the light emitting area. Based on the above examples of discrete devices, multicolor LED arrays, lamps, and LED displays can be made. To create the required multicolor capability a lateral variation of the doping conditions in the active layer is needed. To achieve this, it is favorable to perform the active layer doping after the growth of the active layer. The large lateral variation of the doping conditions required in case of a multicolor LED array or LED display cannot be easily controlled during growth of the active layer in the present state of the art. An approach to realize a large lateral variation of the doping conditions is described in the copending PCT application PCT/EP 94/03346.

To achieve multicolor capability according to the present invention, the active layer needs to be doped with deep donor/acceptor pairs, and optionally shallow donors, with a large lateral variation of the doping conditions. This codoping with deep donors and acceptors is not only used for adding multicolor capabilities but also to meet the objectives of the present invention. The task of doping semiconductor layers after growth including the control of a lateral variation of the doping conditions is divided into a mask step with subsequent codoping steps, whereby several steps of this kind may be sequentially carried through. During a mask step, the surface of the semiconductor structure to be modified by doping is covered by a mask such that only certain islands on top of the surface are accessible for dopants. During the subsequent doping steps, the masked semiconductor structure is exposed to dopants, e.g. a deep donor and acceptor pair. For doping all methods are adequate which allow for the controlled incorporation of dopants into a defined volume of a semiconductor structure through the surface of this structure. Examples of such doping methods are ion implantation or vapor deposition over a wide range of temperatures. Additional annealing leads to a redistribution of dopants by diffusion within the sample and/or to annealing of defects. The annealing is optional for ion-implantation since in this case, its function is mainly to activate dopants rather than to redistribute them. However, after vapor deposition, annealing is mandatory for the incorporation of dopants into a semiconductor structure. The cross-section of a three-color MIS LED array 60, in accordance with the present invention, is shown in Figure 6. It is assumed that

• the substrate 61 is not conductive, e.g. sapphire;

• the LEDs represent light sources whose shape observed in the direction perpendicular to the substrate 61 corresponds roughly to the size and arrangement of the metal contacts on the active layers 64.x (with x = 1 ,2, ...);

each LED can be biased via one individual metal contact 63.x (x = 1 , 2, ...) on top of the active layers 64.x this contact not being shared with other LEDs, and another, common contact 65 which is shared with other LEDs, on the same row or column.

the electrical current through the active layers 64.x of the LEDs is mainly perpendicular to the substrate 61 since the active layer is thin and its resistivity is high in comparison to the adjacent layers.

The following steps are required to arrive at the device structure shown in Figure 6. Since the layer 64.x of the LEDs shown in Figure 6 are intended to serve as active layers, they must be compensated with appropriate impurity pairs (deep donor/acceptor pairs), as described above, to achieve light emission at wavelengths that perfectly match the color perception of the human eye. Since individual diodes with three different colors are desired, for example if one wants to make a full color LED display, at least three different mask steps with subsequent codoping steps are required, each step defining the doping conditions for the active areas of the entire set of equivalent LEDs in the array. The hatched areas in Figure 6 indicate active regions which have been doped in one of the before-mentioned doping steps. These areas are marked with the symbols D, (i = 1 , 2, and 3) in order to distinguish between regions with different doping conditions. The shape of a particular island D, being characterized by constant doping conditions has not to be identical with the shape of light sources fabricated. Since the shape of the metal contact defines the light pattern emitted by the LEDs, transparent metals (e. g. ITO, i. e. indium tin oxide) or semitransparent thin metal layers or combinations thereof can be used for the metal contacts if it is desired to collect the light out of the metal-contact side. If non-transparent metals are used for the metal contacts, the light of the LEDs could still be collected on the contact side.

After the definition of active regions with constant doping conditions, contacts for applying a bias to each particular LED are realized. The conductive semiconductor layer 62 of the LEDs serves as a common electrode to all LEDs unless device isolation is desired in which case appropriate means for electrical isolation such as etching of isolation trenches or deep compensating implantations are applied. However, in this particular example, it is assumed that each LED is individually addressable by means of one individual contact on top of the compensated regions D, and the common bottom contact. Consequently, the conductive layer 62 can be used as common electrode for an entire row or column. As the substrate is assumed to be non-conductive, a physical contact to the conductive layer 62 must be realized. This can be done by etching a contact hole through the top layers 64.x or using side contacts. However, if the LED array is large, the conductivity of the doped layer might not be sufficient for side contacts to the conductive layer 62. Then, a multitude of contact holes or trenches can be etched through the active layers and an appropriate wiring of conductive material 65 can be patterned to provide a low series resistance for all LEDs. For biasing a particular LED, metal contact areas 63.x of appropriate shape are defined for each LED on top of the active layers 64.x (or D,). Known procedures for the metallurgy and the pattern definition, such as photolithographical steps or printing can be used . For addressing each LED independently, different functional elements might be added .

The multicolor array shown in Figure 6 can be simplified if the substrate 61 is conductive (e. g. SiC, Si, InGaAIN). In this case, the doped semiconductor layer 62 in combination with the substrate 61 would serve as a single common electrode and special contact holes for accessing the doped semiconductor layer can be avoided.

For sake of simplicity, we herein referred to the three primary colors red, green, and blue. However, it is to be noted that almost any combination of wavelengths can be precisely achieved by selection of appropriate deep donor/acceptor pairs.

It is obvious that the above embodiments can be modified in many ways. The colors, the number of different colors, the size, the shape and the arrangements of the diodes, and the arrangement of the metallization are arbitrary.

The LEDs and LED arrays described above and hereinafter claimed are well suited for use in connection with different kind of display assemblies provided that an appropriate electrical wiring and driving circuitry is employed. The present LEDs and multicolor LED arrays could also be used as light source illuminating a a light modulating display, such as a LCD display panel.

By proper choice of the donor acceptor pairs arrays of multicolor LEDs and LED displays can be made having individually optimally tuned color components for improved color gamut.

The doping of GaN with deep donor/acceptor pairs extends the concept, known in the art, of introducing an impurity whose optical transition in an LED is at a longer wavelength than the conventional band to band optical transition. The invention offers additional flexibility in tuning the LED wavelength and permits the GaN material system to be extended to even longer wavelengths. Due to the introduction of impurity pairs into the GaN host material the wavelengths can be extended into the green and longer wavelengths where it will outperform GaP and InGaAIP LEDs. In particular green GaN LEDs can be achieved according to the present invention. Up to now, only the indirect bandgap semiconductor GaP and unreliable ZnSe-based devices are capable of producing pure green. Such LEDs can be fine tuned in the present deep donor/acceptor pair scheme so as to perfectly match the eye's sensitivity at particular wavelengths.

The present embodiments are kept as simple as possible, i.e. only one layer serves as active layer. It is, however, to be understood that, without deviating from the gist of the invention, any kind of active region with stacked layers, or quantum-well active regions can be used instead. Initially, only devices based on crystalline materials were discussed, however, the basic concept can be extended to other states of solid matter, e.g. to polycrystalline or amorphous material.

Claims

1 . Light emitting diode with an active region comprising Gallium Nitride (GaN) which is doped with an acceptor, characterized by the introduction of a deep donor into said active region such that an impurity to impurity transition from the deep donor level to said acceptor takes place, the emission wavelength being at least 50meV longer than in case of the corresponding conduction band to acceptor transitions.

2. The light emitting diode of claim 1 , wherein the acceptor binding energy mainly defines the color of light emission, and the binding energy of said deep donor is about a factor of 2-10 less than the acceptor binding energy.

3. The light emitting diode of claim 1 , wherein said active region comprises Indium Gallium Nitride (InGaN), Aluminum Gallium Nitride (AIGaN), or Indium Aluminum Gallium Nitride (InAIGaN).

4. The light emitting diode of claim 1 , wherein an element of the groups VIB, or MIA - VIIIA of the periodic table of elements is used as deep donor.

5. The light emitting diode of claim 1 , wherein an element of the groups HA, IIB, or IA of the periodic table of elements is used as acceptor.

6. The light emitting diode of claim 4, wherein Oxygen (O), Sulfur (S), Titanium (Ti), or Selenium (Se) is used as deep donor.

7. The light emitting diode of claim 5, wherein Magnesium (Mg), Zinc (Zn), Silver (Ag), Carbon (C), Mercury (Hg), Lithium (Li), Beryllium (Be), Calcium (Ca), or Cadmium (Cd) are used as acceptors. i

8. The light emitting diode of claim 1 , wherein a shallow donor having a binding energy of no more than 40meV, is further included into said active region.

5 9. The light emitting diode of claim 8, wherein said shallow donor is Silicon (Si).

10. Light emitting diode array, and in particular a multicolor light emitting diode array, comprising at least one light emitting diode according to

10 any of the preceding claims.

1 1. Display comprising the light emitting diode array of claim 10.

12. Computer or head-mounted display assembly comprising the display of 15 claim 11 and means for driving the light emitting diodes of said display.

20

25

30