BLUE LIGHT EMITTING DIODE FORMED IN
SILICON CARBIDE
FIELD OF THE INVENTION
The present invention relates to the structure and manufacture of light emitting diodes and in particular relates to light emitting diodes formed in silicon carbide which emit blue light.
BACKGROUND OF THE INVENTION
Light emitting diodes, commonly referred to as "LED's", are semiconductor devices which convert electrical power into emitted light. j5
As is known to those familiar with atomic and molecular structure and the electronic transitions of electrons according to the theories and rules of quantum mechanics, when electrons make transitions between their allowed energy levels in atoms or molecules, these transi- 20 tions are always accompanied by a gain or loss of specific quanta of energy. Specifically, raising electrons to higher energy levels absorbs energy and the movement of electrons from higher energy levels to lower ones generates energy. Generally, the energy given off when 25 electrons make such "downward" transitions is in the form of vibrational energy, often observed as heat, or light energy—i.e. photons—which, if in the visible spectrum, can be detected by the human eye.
In a light emitting diode, the generation or "injec- 30 tion" of a current of either electrons or holes across the diode junction followed by the recombination of the injected carriers with either holes or electrons encourages such electronic transitions and is likewise accompanied by either vibrational energy or light, or both. In 35 general, transitions in direct band gap materials generate mostly light while transitions in indirect materials generate mostly heat and some light. Direct band gap materials are defined as those in which the minimum in the conduction band corresponds to the maxima in the valence band, at the same momentum. Correspondingly, in indirect band gap materials, the respective minima and maxima do not coincide at the same momentum.
As is further known to those familiar with electronic transitions, the wavelength of the light which is generated is directly related to the size of the electronic transition. The nature of electromagnetic radiation is such that smaller electronic transitions within the visible 50 range tend to emit longer wavelengths of light, toward the red portion of the visible spectrum, and that larger energy transitions tend to emit shorter wavelengths of light, toward the violet portion. Furthermore, such transitions are always specifically characteristic of the 55 materials in which they occur so that the entire field of spectroscopy is based upon the premise that atoms, molecules and substances can be identified by the characteristic manner in which they react with the electromagnetic spectrum, including visible, ultraviolet and 60 infrared light. Accordingly, the colors which any given semiconductor material can generate are limited and in particular, it has been difficult to date to successfully produce LED's which emit characteristic blue colors. Because blue is one of the primary colors, the lack of 65 such consistently available efficient blue LED's raises problems in a number of technological fields. Absent available blue light, the colors that can be produced or
imaged using LED's are limited to red and green and those colors which can be formed therefrom.
In order to produce blue light, a semiconductor material must have a band gap larger than 2.6 electron volts (eV). As is known to those familiar with semiconductor materials, the band gap represents the energy difference between the conduction band and the valance band of the particular semiconductor material. At the present time, commercially available visible light emitting diodes based on materials such as gallium phosphide (GaP) or gallium arsenide (GaAs) are not suitable for producing blue light because the band gaps are on the order of about 2.26 eV or less. Instead, a blue light emitting solid state diode must be formed from a relatively large gap semiconductor such as gallium nitride (GaN), zinc sulfide (ZnS), zinc selenide (ZnSe) and alpha silicon carbide (also characterized as "hexagonal" or "6H" silicon carbide). Accordingly, a number of investigators have attempted to produce blue light emitting diodes using alpha silicon carbide.
Silicon carbide offers a number of advantages as a potential semiconductor material for blue light emitting diodes. In particular silicon carbide can be readily doped both p and n type. In addition to its wide band gap, silicon carbide also has a high thermal conductivity, a high saturated electron drift velocity, and a high breakdown electric field. To date, however, silicon carbide has not reached the full commercial position in the manufacture of electronic devices, including light emitting diodes, that would be expected on the basis of its excellent semiconductor properties and its potential for producing blue LED's. This is generally the result of the difficulties encountered in working with silicon carbide: high process temperatures are often required, good starting materials can be difficult to obtain, certain doping techniques have heretofore been difficult to accomplish, and perhaps most importantly, silicon carbide crystallizes in over 150 polytypes, many of which are separated by very small thermodynamic differences.
Accordingly, the goal of controlling the growth of single crystals or monocrystalline thin films of silicon carbide which are of a sufficient quality to make electronic devices such as diodes practical, useful and commercially viable, has eluded researchers in spite of years of diligent effort, much of which is reflected in both the patent and nonpatent literature.
Recently, however, a number of developments have been accomplished which offer the ability to grow large single crystals of device quality silicon carbide, thin films of device quality silicon carbide, and to introduce dopants to silicon carbide, as required in the manufacture of LED's and other electronic devices. These development are the subject of co-pending patent applications which have been assigned to the common assignee of the present invention and which are incorporated herein by reference. These include Davis et al, "Growth of Beta-Sic Thin Films and Semiconductor Devices Fabricated Thereon," Ser. No. 113,921, Filed Oct. 26, 1987; Davis et al, "Homoepitaxial Growth of Alpha-Sic Thin Films and Semiconductor Devices Fabricated Thereon," Ser. No. 113,573, Filed Oct. 26,1987; Davis et al, "Sublimation of Silicon Carbide to Produce Large, Device Quality Single Crystals of Silicon Carbide," Ser. No. 113,565, Filed Oct. 26, 1987; and Edmond et al, "Implantation and Electrical Activation of Dopants Into Monocrystalline Silicon Carbide," Ser. No. 113,561, Filed Oct. 26, 1987.
3
The alpha polytype of silicon carbide has a band gap of 2.9 electron volts at room temperature. This band gap is large enough so that any color in the visible spectrum should be available providing the appropriate transition can be made. Because the transition is 2.9 eV 5 in pure silicon carbide, however, a full band gap transition produces light of 424-428 nanometers (nm) wavelength, which has a characteristic violet color. Therefore silicon carbide typically must be doped to provide an additional acceptor level in the crystal to which 10 electrons can move from the conduction band of silicon carbide. For example, if silicon carbide is doped with aluminum, the aluminum dopant will form an acceptor level which is approximately 2.7 eV below the conduction band. As a result, electrons making the transition 15 from the conduction band of silicon carbide to the aluminum dopant acceptor level will emit blue light of approximately 455-460 nanometers.
As set forth earlier, because light is emitted by electrons in transition between energy level, the goal in 20 producing a light from a semiconductor device is promoting or otherwise encouraging such transitions. A diode, which reduced to its basic structure represents a p-n junction, such a method for encouraging the transitions. When holes or electrons are injected across the 25 p-n junction, they will recombine with one another, and a number of the recombination events will include the movement of electrons from conduction or donor bands to valance or acceptor bands and emit the desired light.
Because the overall goal in producing LED's is to 30 obtain as much emitted light as possible the underlying related goals are to be able to inject as much current as possible across the p-n junction, to have the greatest possible dopant population in the emitting layer, to have the greatest possible efficiency in producing recombina- 35 tion events, and to have a physical structure, including transparency, which enhances the visible light obtained from the diode.
In this regard, the flow of current in a diode can be thought of either as the flow of electrons from n to p or 40 the flow of holes from p to n. To obtain various hues of blue emitting devices, both modes of injection are necessary. In most instances, however, it is desireable for most of the current going through the diode to be ntype current; i.e. the flow of electrons across the junc- 45 tion and into the p-type material. This results in a higher electron population on the p side of the junction and a resulting greater number of recombinations, transitions, and photons emitted.
Presently available blue LED's, however, tend to 50 operate on a less favorable basis. One particular commercially available device uses the higher p current in order to attempt to get the desired number of recombinations and in particular uses a p-f-n junction in which, as explained more fully hereinafter, the "+" designa- 55 tion represents a generally greater population of active dopant in the particular material. Such a device works predominantly on hole injection to get the recombination which results in the 480 nanometer emission.
As stated earlier, however, the full band gap in silicon 60 carbide is approximately 2.9 eV, and a transition across this band gap will produce a violet photon rather than a blue one. The most efficient transition in silicon carbide, however, is between an impurity band of nitrogen (donor) below the conduction band and an impurity band 65 of aluminum (acceptor) above the valence band so that electrons and holes which recombine upon injection make a transition between the doped nitrogen and
4
doped aluminum bands and emit the 475-480 nanometer photon which has a more characteristic blue color. Therefore, the predominant carrier flow or injection—whether electrons or holes—must be made into the compensated layer, whether p or n. As a result, in order to use hole current to produce blue light, the portion of the diode which is n-type must be doped with both donor (nitrogen) and acceptor (aluminum) dopants, a technique and structure known as "compensation. " Therefore, in order to have a compensated ntype material, a greater number of n-type dopant atoms must be present in the material than p-type dopant atoms.
As stated earlier, one commercially available LED uses a p+-n junction of this type to get the 480 nanometer recombination and resulting photon. Such LED's are formed by using a p-type substrate and growing a p+ layer on top by liquid phase epitaxy (LPE) with aluminum (Al) as the p-type dopant. Following the addition of the p+ layer during LPE, nitrogen gas may be bubbled into the LPE melt. With the aluminum dopants still in place, the result is a compensated n-type layer. By using this growth technique, one is essentially limited to this device configuration.
There are a number of problems and limitations, however, associated with the use of liquid phase epitaxy to form a p+-n junction. First, it requires the use of a p substrate. Generally, such a substrate has a rather high resistivity because the mobility of holes is only one-sixth of the mobility of electrons and because less than 2% of the acceptor atoms are ionized at room temperature. This results in a higher resistance in forward bias for a diode, which as known to those familiar with such devices, is a less desireable diode characteristic.
One "cure" for this problem is to increase the hole concentration in the p-type substrate. The addition of the extra dopant, however, literally makes the crystal opaque and reduces the emitted light that can be observed. Thus, the problem is a trade off between desirable high transparency and undesirable high resistivity. Adding more p-type dopant to desirably lower the resistivity correspondingly and undesirably lowers the transparency. Alternatively, maintaining a desirable transparency correspondingly and undesirably results in high resistivity.
Yet another attempt to avoid the problem is to put both contacts for the diode on the face of the diode in order to avoid using the substrate as a conductor; see, for example, U.S. Pat. No. 4,531,142. This is an extremely difficult manufacturing technique, however, as reflected in the lower availability and high cost of such diodes.
In addition to using a highly transparent substrate, the light output can be increased by increasing the current injected into the compensated layer. Here the attempt is to increase the p concentration in the p region, which requires increasing the p-type dopant in the epitaxial layer. There are, however, limitations to how high the p concentration can made. In particular, every dopant atom present does not automatically result in an ionized carrier (hole or electron) being present. Generally speaking, the amount of ionized carriers is directly proportional to the number of dopant atoms, but is inversely and exponentially proportional to the ionization (activation) energy of the dopant atom. For example, the ionization energy of aluminum is on the order of 210-220 millielectron volts (meV), while that of nitrogen is only 70-80 meV. Therefore, it is much easier to
5
raise the concentration of ionized n-type dopant atoms than it is to raise the concentration of ionized p-type dopant atoms using nitrogen and aluminum respectively.
Those familiar with such transitions will be aware that the ionization of the materials is thermally generated; i.e. the number of dopant atoms ionized depends upon the temperature as well as the ionization energy. For example, at room temperature with a doping level of 1X1019 atoms/cm3, only approximately 1% of aluminum carrier-atoms are ionized while approximately 22% of nitrogen carrier atoms are ionized. Therefore, for the same number of dopant atoms, the population of ionized n-type dopant ions will be many times as great as that of p-type dopant atoms." As "a" "corresponding result, adding more p-type dopant—usually aluminum—to lower the resistivity likewise lowers the transparency. Furthermore, obtaining a satisfactory p+ layer at room temperature is difficult, as the upper limit of ionized carrier concentration is about 1 to 2xlOI8cm_3 at room temperature. As another disadvantage, the manufacturing technique will always be Umited by the solid solubility of aluminum in silicon carbide.
Furthermore, LPE processes tend to facilitate the addition of n layers to p layers rather than vice versa because nitrogen gas can be introduced into the melt to add the dopant to the resulting epitaxial layer. A typical acceptor atom such as aluminum, however, is much more difficult to add to an epitaxial layer in LPE processes than is nitrogen. Accordingly, adding a p-type epitaxial layer to an n-type substrate is generally regarded as an unavailable process for forming diodes in silicon carbide. The reason is that instead of being able to introduce nitrogen gas into the melt to form the n-type layer, aluminum would have to be introduced as the last step, a process which is much more difficult, if not impossible, compared to adding nitrogen gas.
Finally, LPE processes for SiC comprise growth of epitaxial layers from a silicon melt in a graphite crucible at temperature in excess of 1600° C. Typically, impurities in the graphite crucible, which is physically consumed during epitaxial growth as part of the LPE pro.cess, become incorporated in the growing epitaxial layers; i.e. the p-f- and n compensated layers. Many of these impurities have energy levels that fall within the SiC band gap and their presence leads additional undesired recombination events, resulting photons, and a consequent broadening of the emission peak. Therefore, using this growth technique, sharp, narrow bandwidth emission has not been demonstrated to be practical. For example, the aforementioned commercially available LED specifies a full width at half maximum (FWHM) bandwidth (also referred to as "spectral half-width") of 90-95 nanometers.
Accordingly, there exists the need for improved techniques for manufacturing, and improved resulting structures of, blue LED's formed in silicon carbide that can operate on the basis of injection of electrons as well as holes, that can therefore achieve higher dopant concentrations, higher purity films, more transparent substrates, better current-voltage characteristics, lower resistance, and that can be used to produce diodes which emit in the 475-480 nanometer range, in the 455-460 nanometer range, and in the 424—428 nanometer range, all with narrow bandwidths.
6
SUMMARY OF'THE INVENTION
Accordingly, the present invention comprises a light emitting diode formed in silicon carbide and that emits 5 visible light having a wavelength of between about 475-480 nanometers, or between about 455-460 nanometers, or between about 424-428 nanometers. The diode comprises a substrate of alpha silicon carbide having a first conductivity type and a first epitaxial 10 layer of alpha silicon carbide upon the substrate having the same conductivity type as the substrate. A second epitaxial layer of alpha silicon carbide is upon the first epitaxial layer, has the opposite conductivity type from the first layer, and forms a p-n junction with the first 15 epitaxial layer. In preferred embodiments, the first and second epitaxial layers have carrier concentrations sufficiently different from one another so that the amount of hole current and electron current that flow across the junction under biased conditions are different from one 20 another and so that the majority of recombination events take place in the desired epitaxial layer.
The foregoing and other objects, advantages and features of the invention, and the manner in which the same are accomplished, will become more readily ap25 parent upon consideration of the following detailed description of the invention taking in conjunction with the accompanying drawings, which illustrate preferred and exemplary embodiments, and wherein:
30 DESCRIPTION OF THE DRAWINGS
FIGS. 1 and 2 are schematic diagrams of diodes according to the present invention that emit at a peak wavelength of between about 455 and 460 nanometers; FIGS. 3 and 4 are schematic diagrams of diodes ac35 cording to the present invention that emit at a peak wavelength of between about 424 and 428 nanometers;
FIGS. 5 through 8 are schematic diagrams of diodes according to the present invention that emit at a peak wavelength of between about 475 and 480 nanometers; 40 FIGS. 9 and 10 are emission spectra for diodes according to the present invention;
FIG. 11 is a plot of current versus voltage characteristics for a diode according to the present invention; and FIG. 12 is a cross-sectional view of a mesa LED 45 structure according to the present invention.
DETAILED DESCRIPTION OF THE
PREFERRED EMBODIMENTS
FIG. 1 shows a light emitting diode generally desig50 nated at 20 formed in silicon carbide according to the present invention and which emits visible light having a wavelength of between about 455-460 nanometers. The diode comprises an n-type substrate of alpha silicon carbide 21. An ohmic contact 22 provides electrical 55 contact to the substrate 21. In particular embodiments of the invention, the ohmic contact 22 to the n-type substrate 21, or to n-type epitaxial layers as described hereinafter, may comprise a metal such as nickel. Similarly, the ohmic contacts to be described hereinafter to 60 both p-type substrates and p-type epitaxial layers may comprise a metal such as aluminum.
In the diode 20, a substantially uncompensated n-type epitaxial layer 23 is upon the n-type substrate 21 and has a carrier concentration greater than the carrier concen65 tration of the n-type substrate. As used herein, the phrase "substantially uncompensated" refers to a crystalline epitaxial structure in which some minor amount of opposite conductivity type dopant may be present,
|
