Background of the Invention The efficiency of planar light-emitting diodes (LEDs) is usually low due to total internal reflection at the semiconductor-air (or epoxy for encapsulated devices) interface. The current generation of high efficiency LEDs are all non-planar using some variation on a cube where the light is extracted from all sides, or non-parallel faces such as tapers or trapezoids where the totally reflected light is redirected towards the normal direction by one or more internal reflections. Certain quasi-planar techniques using textured surfaces are also very promising but rely on substrate removal. While these non-planar solutions can give extremely high external quantum efficiencies >50%, a completely planar technology would be preferable.

Microcavities, i. e. , structures with two closely spaced mirrors, can be used to enhance the light-matter interaction. In particular, absorption and emission of light is enhanced when the light is resonant with one of the modes of the cavity formed.

Microcavities are a planar solution to the problem of light extraction. A Fabry-Pérot microcavity around the emitting region uses interference to enhance the emission in the normal direction and to inhibit it in other directions. As well as increasing the efficiency, microcavities also improve brightness (power per area per

solid angle) and can reduce both linewidth and spectral variation with temperature.

The term Resonant Cavity Light Emitting Diode or RCLED is used interchangeably with Microcavity LED or MCLED.

Microcavities can change the directionality and select certain wavelengths of an emitter placed in a cavity.

For semiconductor systems the extraction of light into the non-semiconductor medium is impeded by the large refractive index of most semiconductors. Notably, at a certain critical angle light undergoes total internal reflection at the semiconductor/non-semiconductor interface. A microcavity which redirects the light into this"escape"cone enhances the emission from such a device.

The amount of light extracted depends on the size of the cavity and on the phase dispersion (if any) of the mirrors used. Current microcavity technology uses either a combination of one metal mirror and one distributed Bragg reflector (DBR) or two DBR's. Two metal mirrors is not an option if the extraction of light is an issue.

The shortest cavity that has a Fabry-Pérot resonance has an optical length of half a wavelength. In practice for most semiconductors the active region has a higher refractive index than the low index layer of the DBR mirrors. This produces a phase change such that the optical field has a node at the centre of a half- wavelength cavity. The shortest practical cavity therefore has an optical length of one wavelength.

A half wavelength low index cavity is superior to a one wavelength high index cavity because (a) the cavity is shorter, and (b) there is no guided mode in this structure-the one wavelength high index cavity is embedded between regions of low refractive index giving

rise to a waveguide. Emission of light into the waveguide mode is an important competing mechanism with the Fabry- Pérot mode.

Most texts on multilayer structures discuss only the use of lk or k/2 layers (or multiples thereof) for cavities and B/4 layers for reflecting structures (DBR's). Other solutions are not widely known, and in particular no other solution has been proposed in for microcavity emitters. In general the properties of the multilayer dielectric structure (mirror reflectivities, cavity finesse, etc.) are degraded when moving away from the Bragg condition for the dielectric mirrors and the mk/2 condition (m, integer) for the Fabry-Pérot microcavity.

Efficiencies attained for Microcavity LED's emitting into air are typically 18% for substrate emitting devices and 14% for surface emitting devices.

For microcavities, the main limitations to the extraction efficiency are: (a) the order of the cavity including the penetration of the light field into the mirrors; (b) the enhancement of the guided mode that goes with such multi-layer stacks; and (c) the leakage of light into the substrate at angles greater than angular bandwidth of the DBR mirrors.

The standard cavity has a high refractive index (low bandgap) layer of lk thickness sandwiched between two DBR mirrors. It would be preferable for optical purposes to use a S/2 low index material that would still have an anti-node at the centre of the cavity. This reduces both the cavity order and the coupling to guided modes. However, this option is not compatible with the demands of electrical injection that requires a low band gap active layer (high refractive index). In addition, for the GaAs/AlAs system it is incompatible with high internal quantum efficiency due to the level of defects

associated with AlAs and the problems associated with the planarity of thick AlAs layers.

The state of the art in microcavities is illustrated by US Patents 5,226, 053 (Cho et al) and 5,493, 577 (Choquette/Sandia).

Patent 5,226, 053 concerns a resonant cavity light emitting diode (RCLED) with an active layer capable of emitting light spontaneously under forward bias conditions, within a resonant Fabry-Perot cavity formed by a DBR mirror and a low-to-moderate reflectivity mirror. The optical cavity length is a low integer multiple of A/2 where X is the wavelength of spontaneous emission in the active layer.

Patent 5,493, 577 discloses a resonant cavity light emitting device that can be a VCLED or a VCSEL (vertical-cavity surface-emitting laser) having, between mirror stacks, an active region that is preferably from one to a few wavelengths of the light to be generated in the device, and which has a control layer proximate the active region.

In all of the prior devices, it would be desirable to have a X/2 optical cavity length but this is incompatible with the demands of electrical injection, and cannot be achieved in a practical way.

Summary of the Invention The invention, as set out in the claims, has an optically active region made up of two B/8 low refractive index layers surrounding a A/4 thick high refractive index layer. This gives most of the advantages of the X/2 cavity from an optical standpoint while remaining compatible with the needs of electrical injection and growth.

In this context, "optically active"region means: for a light-emitting device, the region from where the majority of light is emitted; for a detector, the region

in which the majority of light is absorbed; and, for a non-linear optical element, the region wherein the material has non-linear optical properties as a function of the light sensitivity.

If the active regions as defined were to be much longer than lambda the microcavity would lose most of its advantages.

B (lambda) is the device operational wavelength.

The optical thickness of each layer is n*L where n is the refractive index and L the thickness.

The invention provides a multi-layer dielectric structure where the optical mode is localised due to two layers of low refractive index of optical thickness less than one quarter wavelength surrounding a high refractive index medium of optical thickness less than one-half wavelength where there is an optically active material centred in the high refractive index layer and the combined optical thickness of the three layers is less than or equal to three/quarters of a wavelength.

This type of cavity has several unique properties that make it especially useful for microcavity light emitting diodes. The effective cavity length is reduced.

There is anti-guiding around the active region. Finally, metal contacts that would normally shadow the active region actually suppress emission in the normal direction allowing the light to escape from underneath the contacts.

This cavity also has advantages for lasers, including vertical cavity surface emitting lasers (VCSEL), because the threshold condition depends on how large a fraction of the spontaneous emission is going into the lasing mode. In this device this fraction is improved for the same reasons as given for the microcavity light emitting diodes.

The innovative step is to take the X/2 low index layer and replace the central portion with a high index layer of, e. g., k/4 high index material, leaving two adjacent layers of A/8 low index material. It is not obvious that this structure will still produce a Fabry- Pérot-like resonance, but it does.

The structure according to the invention is referred to as a"phase-shifted cavity"because the Fabry-Perot mode is created by a pair of layers which shift the phase to give a Fabry-Pérot mode spatially centred between the two phase shift layers, there being no obvious physical cavity. It has the following important qualities: - It has a Fabry-Pérot mode.

- The order of the cavity (as described by how the wavelength of the Fabry-Pérot resonance moves with respect to changes in cavity length) is close unity, i. e. that of a B/2 cavity.

- The extraction efficiency as a microcavity is improved because the cavity is shorter.

- The guiding of the guided mode parallel to the layers is reduced and thus the coupling or the spontaneous emission to guided mode is reduced.

- The band gap of semiconductor materials follows <BR> <BR> the opposite tendency of the refractive index, i. e. , low index = high band gap. This makes the structure attractive for the electrical confinement of carriers in the active region.

Because the cavity is shorter it is more sensitive to changes in the surrounding layers, this can be used to change the optical properties of the structure with and without metal mirrors. This will be explained in more detail hereinafter.

Preferably, the refractory index of the three layers of the optically active zone is graded at their interfaces, in which case the interface of the layers of graded refractory index can be space defined as being where the value of the refractory index is (n2 + nl)/2 or (n3 + nl) /2, respectively, where nl, n2 and n3 is the average value of the refractive index in the respective layers.

With structures according to the invention, record external quantum efficiencies (defined as the ratio of number of photons emitted by the device to the number of electrons/holes injected into the device) of 19% have been achieved for surface emitting devices and 29% external quantum efficiency has been achieved for MCLEDs with AlOx mirrors.

The inventive multi-layer structure can be used in conjunction with other reflecting surfaces to enhance the interaction of light coming from or going to the exterior of the structure and the optically active region in the centre of the dielectric structure.

The inventive multi-layer structure can be used in any type of semiconductor microcavity. It is fully compatible with standard fabrication techniques and the limitations of electrical injection. Its immediate application is therefore in ALL light emitting devices, in particular in planar light emitting diodes and vertical cavity surface emitting lasers. Another application is for photo-detectors. A further application is for non-linear optical elements, i. e. in which the optically active region is made of a material with non- linear properties, e. g. the refractive index and/or the absorption coefficient varies with light intensity, and whose behaviour depends on such non-linear optical properties.

Brief Description of Drawings In the accompanying schematic drawings, given by way of example: Fig. 1 is a diagram of refractive index vs distance of a device according to the invention; Fig. 2 is a similar diagram which schematically illustrates a first implementation of a device according to the invention; Fig. 3 schematically illustrates a second implementation of a device according to the invention; Figs. 4a, 4b and 4c show cavity structures, namely Fig. 4a a lk cavity, Fig. 4b a B/2 cavity, and Fig. 4c a S/8 phase-shifted cavity according to the invention; Fig. 5a shows the optical field of the Fabry- Pérot mode as a function of position for a lk high index cavity (dashed line), a B/2 low index cavity (dotted line) and the B/8 phase-shift cavity (solid line); Fig. 5b shows the optical field of the guided mode as a function of position for a 1X high index cavity (dashed line), a A/2 high index cavity (dotted line) and the B/8 phase-shift cavity (solid line); Figs. 6a, 6b and 6c show three substrate emitting microcavities, namely Fig. 6a with a single Bragg pair plus lS cavity; Fig. 6b same as Fig. 6a but without Bragg pair; and Fig. 6c a B/8 phase-shifted cavity according to the invention; Fig. 7 shows the absorption in a gold mirror for a structure with: (a) a lk cavity and a Bragg mirror, (b) a lk cavity and (c) a phase-shifted cavity; Fig. 8 shows the extraction efficiency from a substrate emitting structure with a metal gold mirror for a structure with (a) a 1X cavity and a Bragg mirror, (b) a lS cavity and (c) a phase-shifted cavity;

Fig. 9 (a) is a diagram of a demonstration device showing refractive index versus distance for a surface emitting device; Fig. 9 (b) is a close up of the emitting region of Fig. 9 (a) including the doping profile; and Fig 10 is a graph of the external quantum efficiency versus current density for a 400 x 400 ym square MCLED according to Figs 9 (a) and 9 (b).

Detailed Description In this detailed description, all numerical values quoted for refractive indices of materials are their approximate values at 980 nm.

One implementation of the invention is for planar light-emitting devices, wherein the active region is placed between two mirrors so that the light is coupled out of or into the active region. One of the two mirrors could be the air/semiconductor interface. One of the mirrors could be a non-semiconductor dielectric such as AlOxide Fig. 1 schematically represents a device according to the invention. In Fig. 1, the refractory index and thickness of the two layers of low refractory index is given by nl, Ll and n3, L3 respectively, the refractory index and thickness of the central layer is given by n2, L2, the refractory index of materials adjacent the outside of said two layers is given by nO (adjacent nl) and n4 (adjacent (n3). Moreover, the following relationships are to be observed: nl < nO, n2; n3 < n2, n4; and in most preferred embodiments the optical thickness (nl*L1 ; n2*L2; n3*L3) of each layer is in the range: /16<nl*Ll </4 ; B/8 < n2*L2 < B/2 ;

B/16 < n3*L3 < B/4.

In all cases, the combined optical thickness (nl*Ll + n2*L2 + n3*L3) of the three layers making up the optically active zone is given by: nl*L1 + n2*L2 + n3*L3 < B3/4.

More specifically, in preferred embodiments, the optical thickness: nl*Ll is close to or equal toR/8 ; n2*L2 is close to or equal to k/4 ; and n3*L3 is close to or equal to B/8 ; whereby the combined optical thickness (nl*Ll + n2*L2 + n3*L3) of the three layers making up the optically active zone is close to or equal to A/2.

Fig. 2 illustrates an embodiment wherein the active region is placed between air on one side, the air/semiconductor interface forming a mirror, and a distributed Bragg Reflector DBR composed of AlAs/GaAs on the other side.

The vertical axis corresponds to the variation in band gap of the semiconductor material. High band gap implies low refractive index. In this example the high band gap material is taken to be AlAs with refractive index n2. 9 and the low band gap semiconductor is GaAs (n=3. 5). The lengths are with respect to the operating wavelength (lambda) and are denoted in terms of optical length. Physical lengths L are given by L = lambda/n.

Such layered structures are easily grown by epitaxial methods such as molecular beam epitaxy (MBE) or metal- organic chemical vapour deposition (MOCVD) and their variants.

The emitting region is denoted by the vertical arrow, which could be the GaAs region itself or a quantum well placed therein.

In Fig. 3, the active region is placed between air on one side, the air/semiconductor interface forming a mirror, and a distributed Bragg Reflector DBR composed of AlOx/GaAs on the other side. The notation used is the same as for Fig. 2 except that the AlAs of the DBR in Fig. 2 has been replaced by AlOxide (AlOx). This can be achieved through standard wet oxidation procedures. AlOx is insulating and has a refractive index of approximately 1.6.

In the case of AlOxide mirrors the light goes into either the surface emitting direction or into the guided mode. Furthermore, the phase change due to metal is such that the surface normal mode is inhibited leaving only the guided mode. Therefore light will propagate sideways out from underneath the metal. This reduces the shadowing effect of metallic contacts and only becomes significant with the short optical length of the novel structure.

The above examples show III-V based semiconductors although the principles are general to all dielectric multi-layer emitters.

The invention will now be further described and compared to generic prior devices, with reference to Figs. 4 to 10 which illustrate the structures under consideration and their comparative performances. In these examples, standard methods of calculating light extraction were employed, using a plane wave expansion of dipole emission and standard electric field transfer matrices. This approach is a complete approach in that all near field and far field terms of the dipole are automatically included.

Generic case First, we examine two generic cavities: Fig. 4a the standard 1 k high index cavity, Fig. 4b the B/2 low index cavity, compared with the A/8 phase-shift cavity according to the invention, Fig. 4c. In these Figures, the dotted line denotes the source position.

The high refractive index layers are taken to be GaAs while the low refractive index layers are AlAs whose refractive indices are shown in the figures.

At normal incidence, all three cavities show a Fabry-Perot mode in transmission. It is surprising to find a Fabry-Pérot mode for the phase-shifted structure as there is no apparent cavity. If the low index layers had zero thickness, the cavity would be 3X/4 in length, as the layer thicknesses increase, we eventually reach B/4 in thickness for each layer. In both cases the structure is just one long DBR with no resonance. For intermediate lengths close to B/8 a virtual B/2 cavity forms. These types of phase layers are commonly used in distributed feedback lasers to create a single mode in the centre of the DFB stop-band. However, for DFB lasers the phase shift layers are placed far apart in order to distribute the optical field throughout the laser structure, while in these microcavities the phase shift layers are acting to concentrate the optical field.

Figure 5 (a) shows the optical field of the Fabry-Pérot mode as a function of position for a 1R high index cavity, a B/2 low index cavity and the inventive A/8 phase-shift cavity. The field has a maximum at the center of the cavities in each case and the penetration of the optical field into the DBR mirrors is apparent.

The field falls off most rapidly for the B/2 low index cavity, but the A/8 phase-shift cavity shows a definite advantage over the lk cavity. A full numerical

calculation shows that in the normal direction the fraction of emission is 20% for the lk cavity, 36% for the X/2 cavity and 28% for the phase shifted cavity. If we allow for the fact that the emitter is in a low refractive index in the X/2 case then the relative extraction efficiency reduces to 28% (a factor of <BR> <BR> n (low) /n (high) =0.82) showing that the phase-shifted cavity keeps the advantage of the X/2 cavity.

The optical fields of the guided modes are compared in Fig. 5 (b). The B/2 low index cavity has no guided mode so shows a X/2 high index cavity for comparison. The lS cavity has the most confined guided mode while the X/8 phase-shift cavity is less confined than the X/2 high index cavity. The confinement factor is significantly lower for the X/8 phase-shift cavity making it a poor candidate as an edge-emitting laser. The guided mode competes with the Fabry-Pérot mode in MCLEDs so the reduced confinement factor gives a second advantage the X/8 phase-shifted cavity. The numerical calculation shows that the percentage of light going into the guided mode reduces from 36% for the lS cavity, to 28% in the A/8 phase-shifted cavity Substrate Emitters with Top Metal Mirror The prior state-of-the-art for substrate emitting MCLEDs is 23% at 980nm. The devices were large, 2mm (diameter), had thinned substrates to reduce free carrier absorption and an anti-reflection coating on the substrate side. The high efficiency relied to some extent on photon recycling. More typical results are 18% for 400, um devices.

For substrate emitting devices a metal mirror can replace the top DBR or the semiconductor/air interface.

Again three structures are compared (see Figs. 6a, 6b, 6c), each with a top Gold mirror and a bottom DBR. The (complex) refractive index of Gold is taken to be 0.2 +6i at 980 nm. In Fig. 6a the gold mirror is followed by a

Bragg pair and a lk cavity. Fig. 6b is the same as Fig.

6a but no Bragg pair. Fig. 6c shows the phase shifted cavity. In these Figures the substrate is to the right after the Bragg mirror. In each case a GaAs phase match layer of B/6 in thickness is used to adapt the 120 degree phase change from the Gold mirror.

The first two cavities are the standard designs used for substrate emitting microcavities. In particular, Fig. 6b InGaAs based devices give around 18% external quantum efficiency for"normal"sized devices (d = 400ym diameter) and up to 23% for large devices (diameters up to 2 mm). The difference is attributed to recycling of emission in the plane of the device through re-absorption and re-emission in the large devices. One concern with such devices is that the proximity of the gold mirror may lead to losses. This is a reason for adding a Bragg pair between the gold mirror and the cavity (Fig. 6a). In practice, this design of Fig. 6a, shows no improvement over Fig. 6b.

The calculated loss to the gold mirror is shown in Fig. 7, and the Bragg pair helps in reducing the coupling to the gold mirror. However, the phase-shifted cavity behaves very differently. The loss to the gold is reduced by at least a factor of 4 and it shows the opposite dependence on wavelength. This is a general feature of phase-shifted cavities; although the source is much closer to the metal mirror, the mirror loss is reduced. The modes with a strong overlap with the source have a weak overlap with the metal and vice versa.

Fig. 8 shows the calculated extraction efficiency as a function of wavelength. It is assumed that the substrate thickness is small (no absorption losses) and that there is an ideal anti-reflection coating. It can be seen that the A-only structure (b) is marginally better than the-A+ Bragg pair (c). The phase-shifted cavity

(a) is significantly better than the other two structures.

Implementation: Surface emitters with air interface.

The phase-shifted principle can be adapted to any existing microcavity structure with the advantage that the effective order decreases by one. We have examined surface emitting structures where the top mirror is simply the GaAs/air interface. Although the reflectivity is only 30% this is sufficient to give a microcavity effect when combined with a highly reflective bottom mirror. The GaAs/Air interface has the important property that the phase change on reflection is constant with angle (until the critical angle). Figure 6c shows a schematic of the structure.

An interesting property of this structure is that it is so thin that when gold is deposited on it the phase shift due to the metallic reflection is sufficient to inhibit emission in the normal direction. Instead the light is redirected into leaky and guided modes. Normally the light would be absorbed in the metal if this inhibition did not occur. The result is that the light has a tendency to move sideways out from underneath the metal contact. If the light is reabsorbed then photon assisted current spreading occurs, reducing the degree of contact shadowing which is a limiting factor for surface emitting devices.

Devices were fabricated by molecular beam epitaxy following the above design. The semiconductor layer sequence is shown in Figs. 9a and 9b, including the doping profile. The device was grown by molecular beam epitaxy, processed, mounted and bonded on a standard LED mount. The light output was measured by placing the device inside a calibrated integrating sphere. Fig. 9b shows the thickness of the different layers to scale as well as their refractive indices. Following growth, the

structures were processed to have n-contacts and p- contacts. The resulting devices, having a square mesa with a 400ym aperture, were cleaved, and mounted onto standard mounts and wire bonded to the support. The finished devices were placed inside a calibrated integrating sphere and the light output as a function of drive current was measured. Its performance is shown in Fig. 10. The maximum extraction efficiency is 19% and exceeds that from any other similar surface emitting LED.

The maximum efficiency occurs at very low current densities, of the order of 2 A/cm2. The maximum efficiency is close to the calculated extraction efficiency of 25. 5% showing that the internal quantum efficiency is high and that the effect of contact shadowing at low current densities is minimal.

Laboratory demonstrations have also been made with substrate emitting devices and surface emitting devices with where AlOx layers have been used. The latter showed external quantum efficiencies in excess of 28% and surpass current state-of-the-art.

In conclusion, the invention provides a novel microcavity structure which uses a phase shifted cavity to form a Fabry-Perot cavity which behaves like a B/2 high index cavity except that there is a maximum of the optical field in the centre of the cavity instead of the usual node. Using this principle, surface emitting devices were made where the top mirror was just the Air/Semiconductor interface. Record external efficiencies were measured. The low current density operation makes these devices interesting for applications needing large area and low absolute currents.