FIELD OF THE INVENTION

The invention is in the field of semiconductor light sources and light amplifiers. In particular, it concerns a superluminescent light emitting diode (SLED), and an amplifier chip, as well as an' apparatus (tunable laser, optical amplifier) containing the amplifier chip.

BACKGROUND OF THE INVENTION

Semiconductor light sources and amplifiers usually include a semiconductor chip with a semiconductor heterostructure and a current injector for injecting an electrical current into the heterostructure. By the injected current, energy levels above the ground state are populated. When the injected charge carriers go back to the ground state (usually by recombination), light particles of an according energy are generated. In Light emitting diodes (LEDs), this happens by way of spontaneous emission, in laser devices and optical amplifiers by stimulated emission. A special class of light sources are the superluminescent light emitting diodes (SLEDs) that include, like laser diodes, a waveguiding structure and are optoelectronic edge emitting devices generating light mainly by way of amplified spontaneous emission. This means that the spontaneously-emitted photons are amplified by means of stimulated emission inside a waveguide and directed towards an aperture. In contrast to lasers, however, SLEDs comprise means for suppressing optical feedback, so that no resonance occurs. SLEDs are neither lasers nor LED devices but a special class of light sources.

For some applications of SLEDs, optical amplifiers, and gain elements for tunable lasers, it is desired that the devices show a broadband characteristics and are capable of generating (or amplifying) radiation in a broad part of the optical spectrum.

Of special interest are light sources that emit in the visible part of the optical spectrum, especially the shorter wavelengths including green, blue and violet light. Currently, semiconductor based light sources in the shorter wavelength range are often based on GaN material systems. With such light sources, violet and blue light has successfully been generated. However, it has been proven to be difficult to produce, by a semiconductor light source, light in the green part of the optical spectrum. This phenomenon is sometimes addressed as the "green gap" in semiconductor light source production. It is mainly due to material properties of the InGaN material system used for obtaining transitions in this wavelength region: For higher indium concentrations, the Indium tends to segregate.

SUMMARY OF THE INVENTION

In accordance with a first aspect of the invention, an optical device is presented, the optical device being a superluminescent light emitting diode or amplifier chip, the optical device comprising a semiconductor quantum well heterostructure embedded in a cladding structure, and a current injector for injecting charge carriers in the heterostructure, so that radiation reaching the heterostructure is amplifiable by stimulated emission. The heterostructure defines a first and a second discrete energy level in a conduction band, the first and second energy levels spaced apart from each other by more than a thermal excitation energy kT _Λt at a working temperature T _w of the optical device, with essentially no energy levels between the first and the second energy level, and the higher one of the first and second energy levels being spaced from a conduction energy band of the cladding structure by more than the thermal excitation energy at the working temperature, the optical device further comprising feedback suppressing means.

As a consequence words, the two energy level pairs of a conduction band state and a valence band state, between which a transition is possible, have a difference in transition energy corresponding to at least kT _w . (i.e. the difference between the conduction band levels plus a possible difference between the valence band states to which a transition is quantum mechanically allowed. In many embodiments, due to the large hole specific mass, this energy difference between states in the valence band is small.).

The term amplifier chip herein refers to both, a chip of a semiconductor optical amplifier (SOA) that is used to amplify radiation incident on it, and to a so-called gain chip, namely a chip used as gain element in a laser, where the laser cavity is not (only) defined by the chip itself but includes light path portions outside of the chip (so-called external cavity semiconductor lasers); this includes ring lasers.

The first and second energy levels may comprise one or a plurality of quantum mechanical states that may be degenerate or approximately degenerate, i.e. in a neighborhood the first and/or second levels may optionally include a plurality of sub- levels, or the first and/or second energy levels may constitute a sub-band. Between the first and second energy levels and between the higher one of the energy levels and the (lowest) conduction band of the surrounding cladding structure, there are no available states. This, of course, does not exclude the possible presence of a comparably small number of impurity states in the regions in-between, the spectral weight of which is low compared to the spectral weight of the first and second energy levels, and is preferably negligible.

Using the at least two energy levels/level pairs allows achieving a broadening of the spectral emission to more than for the for example about 40 meV (this value strongly depends on the material system) that may be achieved in conventional superluminescent light sources by the broadening of the optical emission spectra of a single optical transition. The approach according to the invention makes possible to considerably increase the broadening of the spectral emission up to approximately 100 meV and more.

Embodiments of the device according to the first aspect include devices with a heterostructure in which the transition energies of the energy level pairs, (or, if the valence band is narrow, the energy levels of the conduction band) are spaced apart by at least 30 meV. preferably by between 35-1000 meV or between 35-500 meV. especially preferred 40-150 meV.

Due to the approach according to the invention, such spacings of higher than for example 40 meV, 50 meV or 60 meV or much higher are possible without there being a gap between different peaks resulting from two different transitions. This is because a plurality of energy levels in the valence bands are available for transitions from both energy levels in the conduction band. This is a major advantage over prior art approaches in which the spectrum is broadened by there being two different, longitudinally spaced regions from which two different transitions originate, fhese cannot be spaced further away than for example about 30 meV, in order not to produce a "dip ^" in the optical spectrum. The approach according to the invention is especially favorable for an optical device or optical apparatus suitable of generating light in the visible part of the optical spectrum. Especially, concepts of the invention may make a semiconductor light source possible that emits blue and/or green light of different wavelengths and that may even be tunable between these different wavelengths.

According to an other aspect of the invention, therefore, a semiconductor optical device is provided, the optical device preferably being a superluminescent light emitting diode or amplifier chip and comprising feedback suppressing means, the optical device comprising a semiconductor heterostructure including at least one quantum well layer between two barrier layers, the quantum well layer being of a quantum well semiconductor material having a bandgap of between 1.9 eV and 3.5 eV, for example between 1.9 eV and 3.O eV, especially preferred between 2 eV and 2.6 eV, and both of the two barrier layers being made of a barrier layer material or of barrier layer materials having a bandgap of at least about 0.25 eV higher than the bandgap of the quantum well semiconductor material, preferably about 0.4 eV higher or more.

According to an even further aspect of the invention, an optical device is provided, the optical device preferably being a superluminescent light emitting diode or amplifier chip and comprising feedback suppressing means, the optical device comprising a semiconductor heterostructure including at least one quantum well layer, and a cladding in which the heterostructure is embedded, the quantum well layer being of a semiconductor material having a bandgap of between 1.9 eV and 3.5 eV, for example between 1.9 eV and 3.0 eV, especially preferred between 2 eV and 2.6 eV, and the cladding comprising a semiconductor cladding material or semiconductor cladding materials having a bandgap of at least about 0.5 eV higher than the bandgap of the quantum well semiconductor material, preferably at least about 0.6 eV higher or even at least about 0.8 eV higher. In these aspects, the quantum well material may be a material including Gallium and Nitrogen, i.e. a GaN based compound, such as In _x GaN with x between 0 and 0.35.

Especially preferred are a barrier layer material or barrier layer materials having a bandgap of at least about 0.25 eV higher than the bandgap of the quantum-well layer material, and the cladding includes semiconductor cladding material or semiconductor cladding materials having a bandgap of at least about 0.25 eV higher than the bandgap of the barrier layer material.

According to yet another aspect of the invention, an optical device preferably being a superluminescent light emitting diode or amplifier chip and comprising feedback suppressing means is provided, the optical device comprising a semiconductor quantum well heterostructure embedded in a cladding structure, and a current injector for injecting charge carriers in the heterostructure, the current injector comprising at least one current injector segment operable to inject charge carriers into an according heterostructure segment by means of application of an electrical voltage between two electrodes, the current injector segment and the according heterostructure segment being adapted to each other so as to enable the current injector segment to cause radiation of two distinct optical transitions to be emitted by the optical device by stimulated emission, the two optical transitions being spaced from each other by more than the value ICT _R , where Tn is the room temperature and preferably by more than the value kT _» , T _H being the working temperature of the device.

The approach according to this aspect of the invention is in contrast to prior art approaches. Broad emission spectra of prior art devices, such as amplifiers or multi- segment broadband SLEDs, were obtained by having several segments of different characteristics in series, each segment being individually contacted. While often a backside contact or an extensive frontside contact serves as common first electrode (it may be referred to as "ground" contact), the second electrode of the different segments is distinct, resulting in a total of at least three electrodes. The distinct, non- parallel electrodes are necessary because only then it is possible to have different "pumping levels", that is different current densities in the two segments. By this in the prior art approaches, the Amplified Spontaneous Emission spectra are shifted towards higher wavelength values and lower wavelength values, respectively.

The approach according to this aspect of the invention makes a - set-up with only one current injector segment possible, but does not exclude the use of several segments, at least one of which comprises two distinct optical transitions.

The distinct transitions will preferably be so that their emission spectra overlap. In the approach according to this aspect, the distinct optical transitions preferably have comparable energy density levels, resulting in a in a contiguous, broad, non-Gaussian optical spectral characteristic of the device.

According to an even further aspect of the invention, an optical device being a superluminescent light emitting diode or amplifier chip is provided, the optical device comprising feedback suppressing means, and comprising an active quantum well heterostructure embedded in a cladding structure, the active quantum well heterostructure comprising at least one quantum well, and the cladding structure comprising on at least one side of the heterostructure, heterogeneous graded index semiconductor layers in which the energy gap increases as a function of a distance to the active heterostructure. Preferably, the materials are moreover chosen so that the index of refraction for the radiation generated in the active heterostructure decreases in the graded index semiconductor layers as a function of the distance to the active heterostructure In an embodiment, heterogeneous graded index semiconductor layers are present on two opposite sides of the heterostructure.

Graded index semiconductor layers (sometimes referred to as graded index carrier confinement (GRICC) layers) may for example be realized by grading the chemical molefraction of compounds such as ternary or quaternary material compounds. This allows to have a graded profile of the conduction and valence bands with a decrease in the energy band-gap from the external cladding layers towards the active region(s). The GRICC layers improve the carrier transport from the bulk part of the device, where the high doping level guarantees high carrier densities, towards the active region as carriers lose potential energy, moving therefore to a more preferred energy condition while descending through the GRICC layer structure.

In the embodiments with graded index layers, the above-specified preferred conditions for the bandgap of cladding semiconductor material hold for the Outer ^" portion of the cladding, thus of portions being further away from the active region.

By the preferred grading of the index of refraction in the above-specified manner, moreover the following advantage may be achieved: In the optical device, photons generated close to one active section of the waveguide are amplified while traveling along the entire length the waveguide toward the other end of the active section, giving rise to most of the amplified spontaneous emission that can be detected at this second end. During this travel along the waveguide the number of photons increases due to the stimulated emission of other photons, causing the population of carriers in the excited state(s) to decrease. This in turns reduces the material gain, which is required for generating other photons. As a result, the maximum amount of photons that can be generated at a specific place in the active section is limited by the number of carriers populating that section. When the number of photons is much larger than the number of carriers, the section cannot contribute to the photons population. This is commonly referred to as gain saturation effect in semiconductor optical amplifiers (SOAs) and is a shared property of both SOAs and SLEDs. The gain saturation effect can be partially compensated with a drastic reduction of the photon overlap with the active region, namely the optical confinement factor. The graded index approach helps to drastically reduce the confinement factor. This is done to ensure that the active region(s) can sustain a photons generation rate compatible with the rate of photons traveling along the longitudinal direction, reducing the overlap with the active region(s). Further, the GRICC structure helps increasing the carrier transport properties.

Especially preferred are embodiments of the invention that combine two, three four or all five aspects of the invention. All mathematically possible combinations of aspects are physically realizable.

For all aspects of the invention, preferred embodiments include one or more of the following features:

Choosing a thickness of the active (quantum well) layer(s) plus the barrier layers in such a way that the overall thickness of the active region (comprising the active layer(s) and the barrier layers) is still comparable to the electron de Broglie wavelength, and the structure quantum well(s) plus barriers still gives a structure with reduced density of states. In such a situation, the solution of Schrδdinger's equation outside the quantum well is higher in energy compared to the bottom of the conduction band, respectively lower compared the top of the valence band of the barrier layers. This helps to ensure that the energy level difference between the at least two discrete, bound states and the higher energy unbound states is larger than the thermal excitation energy. In addition or as an alternative, the structure is chosen such that between the barrier layers immediately adjacent to the quantum well layer(s) and cladding layers bordering the barrier layers there is a band gap jump (i.e. the band gap of cladding portions immediately adjacent the barrier layers is substantially higher than the band gap of the barrier layers.)

The layer structure that includes the active heterostructure (active region) and the cladding will preferably include a p-doped side and an n-doped side, where the active region need not be purposefully doped at all. According to a preferred embodiment, preferably the p-doped side comprises close to the active heterostrucrure a high bandgap barrier layer that has an increased bandgap to ensure that electrons that have not recombined in the active region may not travel to the p-doped side but are reflected by the barrier layer.

The device is configured so that in a normal operation regime, the carrier density in the active heterostructure exceeds about 2.5 10 cm ^" . This is especially achievable in combination with a feedback suppressing means, which ensures that a surplus of carriers does not recombine due to stimulated emission brought about by photons circulating in a resonant cavity. Especially advantageous is a realization of this in a GaN-based optical device. This is because the high carrier density could reduce, if not completely screen, the piezo-electric field effects in GaN-based edge emitting devices. This in turn would allow growing quantum well layers with comparably large thickness of more than 3 nm, for example even more than 4 nm, 5 nm or 6 nm without having the problem of the piezoelectric field caused reduced electron-hole wavefunction overlap. The larger thickness of the quantum well layers in turn makes lower transition energies possible, thus allows to produce longer wavelength light. For all embodiments, the feedback suppressing means may comprise one or a combination of the following elements:

Tilting the waveguide on a semiconductor chip and thereby causing radiation reflected by end facets not to be reflected back into the waveguide but away from it.

As an alternative or in addition to tilting the waveguide, treating an end facet of the waveguide to be non-perpendicular to the waveguide direction.

Applying low power-reflectivity coatings at the ends of the cavity.

Using an absorber section where optical device comprises a heterostructure (that may be identical with the heterostructure in the gain section) but that is unbiased or reverse biased, as for example described in US patent 7,1 19,373 and references cited therein.

A broadening or narrowing of the waveguide towards at least one of the end facets.

- A bending or sharp bent of the waveguide.

Other means that result in the suppression of Fabry-Perot modes and ultimately the reduction of the coherence length of the emitted light. An optical apparatus (external cavity, for example tunable laser, optical amplifier, low coherence light source apparatus etc.) according to the invention comprises an optical device according to any one of the aspects of the invention and further comprises optic components (mirrors, lenses, prisms transparent plates, waveguides, active optic components etc.) arranged so as to direct radiation onto the optical device and to direct at least portions of radiation amplified by the optical device onto an optical apparatus optical output. The apparatus further may comprise, as known in the art, wirings, contacts controls etc. for operating the apparatus.

BRIEF DESCRIPTION OF THE DRAWINGS

The foregoing and other features and advantages of the invention will be further described in the following more particular description of preferred embodiments of the invention as illustrated in the accompanying drawings. The drawings are all schematical and not to scale. In the drawings, same reference numerals refer to same or corresponding elements.

- Fig. 1 is a schematic view of an SLED according to the invention;

Fig. 2 is a cross section of a layer structure of a first embodiment of the SLED according to the invention;

Fig. 3 is a cross section of a layer structure of a second embodiment of the SLED according to the invention:

- Fig. 4 shows a variant of index guiding concept; Figs. 5 and 5a shows the energy band gap profile of a cut near the active region in the direction transverse to the direction of light emission of two different embodiments; a possible profile for both p-type and n-type dopant is in both cases also shown;

Fig. 6 shows the refractive profile of a cut near the active region in the direction transverse to the direction of light emission; a possible profile for the optical mode in the transverse direction is also shown;

Fig. 7 shows a cross section of a layer structure with a waveguide according to an embodiment of the invention;

- Fig. 8 shows an energy band gap profile of a cut near the active region in the direction transverse to the direction of light emission of the embodiment of Fig. 7; and

Fig. 9 and Fig. 10 show top views of a one-segment and a multi-segment optical device, respectively.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

The SLED 1 , a basic layer structure of which is illustrated in Figure 1, comprises a semiconductor quantum well heterostructure forming an optically active zone 2 between a first cladding layer 3 and a second cladding layer 4. As will be explained in more detail further below, the cladding layers may in themselves be inhomogeneous and comprise a sub-layer structure, and are preferably partially or fully made of a graded-index semiconductor compound to form GRICC layers.

For example, the first cladding layer 3 may be made of a semiconductor material of a first conductivity type (for example an n-doped semiconductor), and the second cladding layer may be of a second conductivity type (such as a p-doped semiconductor).

The structure may be provided on a substrate (not shown) of a suitable material. The substrate may be conducting (for example n-type conducting), in which case a first contact electrode may be arranged on a bottom side of the substrate. The substrate may alternatively be electrically insulating, in which case a contact layer may be arranged between the first cladding layer 3 and the substrate, and a contact electrode (not shown) contacting the contact layer may be located at a lateral distance to the waveguide. Ways of contacting a heterostructure by appropriate electrodes are known in the art and will not be described in any further detail here.

The optically active zone 2 may comprise one or more optically active regions, as will be explained in more detail referring to the following figures. The optically active zone is capable of producing light upon injection of an electric current in a vertical direction. More concretely, the layer structure includes an at least partially horizontal waveguide so that light traveling along the waveguide is produced and amplified, by means of stimulated emission.

The embodiment of Figure 2 comprises a single active layer 15. The optically active layer 15 has a smaller bandgap than the surrounding barrier layers 7 (and, if at least one optically active layer is directly adjacent a cladding layer, also than the cladding layer).

The energy structure relationship between the cladding layers 3, 4, the barrier layers 7 and the active layer 15 is such that more than one radiative transition can be excited in the optically active layer. The optically active layer is a quantum well layer. This is the case if the de Broglie wavelength of the charge carriers is comparable to the layer thickness or even larger than it. More concretely, the thickness of the active layer may for example be less than 10 nm, less than 7 nm or even less than 5 nm or less than 4 nm.

The optically active layer 15 has a smaller bandgap than the barrier layers 7 and the cladding layers. More concretely, the bandgaps of the barrier layers (and cladding layers) and of the quantum well active layer, and the thickness of the quantum well layer are chosen so that at least two discrete energy levels may exist in the conduction band of the quantum well layer, the two discrete energy levels being spaced apart by more than the thermal excitation energy, and the upper one of these two energy levels being spaced apart from the lowermost barrier layer/cladding layer conduction band state by more than a thermal excitation energy.

The optically active zone 2 in the structure according to Figure 3 includes a plurality of (i.e. at least two) optically active transitions. The optically active regions in Fig. 2 are formed by two distinct optically active layers 5, 6 spaced from each other in a vertical direction. Between the optically active layers 5, 6 and potentially also adjacent the optically active layers, there are barrier layers. The optically active layers have a smaller bandgap than the surrounding barrier layers and than the cladding layers. Also in this embodiment, the optically active layers may be quantum well layers. The thickness of the active layers 5, 6 may for example be less than 3 nm or less than 4 nm. The thickness of the two active layers 5, 6 may be equal or unequal. The thickness of the barrier layer 7 between the quantum well layers may be of the order of magnitude of the quantum well layer thickness, for example between 5 nm and 10 nm. It may also be higher than the thickness of the quantum well layers themselves. If the barrier layer 7 is not thicker than about an (exponential decay) penetration depth of the bound state wavefunction in the quantum well layer, the two quantum well layers will together form a common, coupled system.

In the shown embodiment, the bandgaps of the different quantum well layers are not equal. More concretely, at least one of the layer thickness (accounting for the levels of energy subbands) and of the semiconductor material composition (accounting for the 3D "bulk' ^" energy band structure) is different between two different layers. In Fig. 3, a first quantum well layer 5 is illustrated to be wider than a second quantum well layer 6.

Also in this system bandgaps of the barrier layers (and cladding layers) and of the quantum well active layers, and the thickness of the quantum well layers are chosen so that at least two discrete energy levels may exist in the conduction band of the heterostructure, the energy levels being spaced apart by more than the thermal excitation energy, and the upper one of these two energy levels being spaced apart from the lowermost barrier layer/cladding layer conduction band state by more than a thermal excitation energy.

A combination of the embodiments presented in Figure 2 and Figure 3 is also possible to further increase the spectral emission of the device. Figure 4 illustrates the principle of an index guided waveguide. The optically active zone 2 may be formed as in any one of the previous embodiments or as in any other embodiment of the invention. The structure comprises a ridge 41 defining a waveguide with a waveguide direction out of the picture plane. If the second cladding layer 4 is not too thick, such a ridge will laterally confine light and cause a light beam to propagate along the waveguide. The ridge may be made of a same material as the second cladding layer, or may be different therefrom, with an index of refraction that is approximately equal.

In addition or as an alternative to a ridge, the structure may comprise a buried heterostructure, i.e. at least one laterally structured layer not being the topmost layer of the structure. More in general, the skilled person will know for example from

SLED or laser technology several ways to make a waveguide in a semiconductor heterostructure, the waveguide defining a waveguide direction, and the invention is not confined to any particular one of these methods but may be carried out using a variety of waveguiding mechanisms or combinations thereof.

Figure 5 shows an example of a simplified energy band gap (thick line) of the semiconductor compounds used, with emphasis on the graded index layers where the band gap changes as a function of its transverse position. This is for example done by a gradual or stepped characteristics of a material parameter (such as the percentage of a constituent of the semiconductor material, such as Al or In), or of a growth parameter during manufacturing of the preferably epitaxially grown layers, that reflects in a gradually or steppedly adapted growth characteristics.

Also shown in Fig. 5 are the doping levels of the active regions for both carrier specimens. The p-type doping profile (solid line; 51 ) and the n-type doping profile (dashed line; 52) in the depicted embodiment show sharp variation of the doping level. However the doping level can also be changed gradually within the cladding layer. Advantageously, the doping level outside the GRICC layers region is much higher compared to a design without the GRICC layers, allowing a better supply of carriers to the active region 2. For example, the doping is such that the density of impurities in the region outside the GRICC layers is at least 1 10 impurities/cm .

Figure 6 shows a refractive index profile for a device of the embodiment of Fig. 5 with GRICC layers 50 in the p-type cladding 4 and also in the n-type cladding 3, where the large refractive index of the layers far away from the active region helps decreasing the overlap of the optical mode 53 on the active region 2 and in particular of the active layer 15 or alternatively, referring to embodiments as the one shown in Fig. 3, of the active layers 5 and 6 (not shown in Fig. 6).

Figure 5a depicts a variant of the embodiment shown in Fig. 5. In contrast to the latter embodiment, the embodiment of Fig. 5a comprises one GRlCC structure only, whereas at the other side, a homogeneous n-type cladding 3 is directly adjacent the active region 2.

Also, the energy gap of the n-type cladding is different from the energy gap of the p- type cladding. Claddings comprising high-bandgap layers of different energy gaps may also be used if both sides of the heterostructure comprise a GRICC structure.

Further variants would be possible, for example the combination of one GRICC structure with a stepped structure on the other side, a GRICC structure including a step, etc. Figure 7 schematically shows an example of a layer structure that follows the principles explained with respect to Figures 1 , 2, and 4-6 above and that is especially suited for the preferred embodiment of an optical device that produces and/or amplifies light in the visible range of the optical spectrum, more concretely blue or green or violet light. The skilled person will readily understand that the concept and teaching of the invention can be applied to structures with different layer compositions and layer thicknesses than the one given in the example.

The structure comprises a GaN substrate 61 being a crystal of a thickness between 100 μm and 300 μm. On top of the substrate, a first high bandgap layer 63 is arranged. The first high bandgap layer is a heavily n-doped 1 μm Ao _. iGaN layer. Atop the first high bandgap layer 63, the GRICC structure 62 incorporating the heterostructure with the active layers is grown. The structure 62 is explained in more detail with respect to Fig. 8 below. Adjacent the GRICC structure 62 is a high bandgap layer, being heavily p-doped. In the shown embodiment, the second high bandgap layer is interrupted by an electron barrier layer 65 with an even higher bandgap and thus comprises a thin first sub-layer 64 and a thick second sub-layer 66. The layer sequence atop the structure 62 is thus:

• 20 run Al _0. i GaN sub-layer 64

• 5-10 nm AIo ₂ GaN barrier layer 65

• heavily p-doped 1 μm Al ₀ i GaN sub-layer 66.

The function of the electron barrier layer 65 is to keep electrons coming from the n- doped side (right side in the figure) that have not recombined with holes in the active region from progressing to the p-doped side and there recombining with holes. By the electron barrier layer 65. the electrons are reflected back towards the active region.

The layer system is structured to form a waveguide as already shown referring to Fig. 4. In the depicted example, the waveguide is formed by the 1 μm AIo iGaN sub- layer 66 being at least partially etched away in a region away from the waveguide.

The current is injected into the active region via electrodes. A first electrode 71 is arranged on the backside of the substrate. The second electrode 72 is arranged on the cladding structure, here on the 1 μm AIo jGaN sub-layer 66. In order to ensure a good coupling, the device comprises a 20 nm GaN interface layer 67 between the AIo i GaN and the electrode 72.

The electrodes 71 , 72 may be metal electrodes as conventionally used for semiconductor radiation sources, for example of a Ti/Pt/Au sequence.

Figure 8 depicts the bottom of the conduction band in the GRICC structure and the region surrounding it. The bottom of the conduction band at the same time reflects the bandgap structure. In accordance with the well known "60/40 rule' ^" (or "70/30 rule"), often energy differences of the bottom of the conduction band amount to between 60% and 70% of the according bandgap differences.

The graded index layers 50 are of Al _x GaN with x varying between 0 (at the interfaces to the barrier layers 7 and 0.1 (at the interfaces to the high bandgap layers 63m 64. The barrier layers 7 in the depicted embodiment are not of GaN but of Ino o ₄ GaN. thus a material with a lower bandgap than GaN. This leads to the depicted jump in the bandgap at the interfaces between the graded index layers 50 and the barrier layers 7. The trilayer comprising the two barrier layers 7 and the active layer 15 thus in itself confines discrete states - in the depicted embodiment, it is one discrete state 81 - if the total thickness of the trilayer is not substantially greater than the de Broglie wavelength and preferably smaller than the de Broglie wavelength. For example the total thickness of the trilayer may be chosen not to exceed 20 nm. This "miniband" structure thus helps in ensuring that the distance between the higher energy state 82.2 in the conduction band and the named discrete energy state 81 exceeds the thermal excitation energy to minimize the probability of carrier thermal escaping out of the quantum wel 1.

The bandgap of Ino O ₄ GaN is 3.21 eV, the bandgap of GaN amounts to 3.42 eV, and the bandgap of Ao |GaN is 3.58 eV

The material of the quantum well 15 in the shown example is Ino _O2 GaN with a bandgap of 2.5 eV. In this example, the first and second discrete states 82.1 , 82.2 in the conduction band account for blue (and violet) light emission.

As a first variant of the depicted embodiment, if the device is to emit green and blue light, the Indium content of the quantum well can be enhanced to about 30%. Then, the bandgap of the barrier layers and/or of the graded index layers and/or of the high bandgap layers may or may not be reduced, if reduced, this may be done by enhancing the In content and/or reducing the Al content.

As a second variant of the depicted embodiment, the barrier layers may be made of GaN instead of In ₀ OaGaN, so that there is no bandgap jump between the barrier layers 7 and the graded index layers 50. In this variant, the "miniband' ^" effect is not present, but the quantum well is deeper. This variant also shows that the borderline between barrier layers on the one hand and cladding layers on the other hand is somewhat arbitrary, since in this the barrier layers continuously merge into the ^" graded index (cladding) layers.

As further variants of the depicted embodiment (with or without the "miniband"), materials accounting for emission at different wavelengths may be used, such as ternary compounds like InGaAs-based materials for the active layers (with In contents in the region between 2% and 40, especially between 2% and 10%), GaAs for the barrier layers, and AlGaAs or AlGaInP compounds for the cladding, all on a GaAs crystal substrate. Then, the emission spectrum will be shifted to longer wavelengths compared to the above embodiment, and the optical device may for example be suited as light source for Red or Near-Infrared (NIR) light, for example from 630 run to 1 150 nm.

As yet other variants of the depicted embodiment (with or without the "miniband"), materials accounting for emission at longer wavelengths may be used, such as quaternary compounds like AlGalnAs-based materials for the active layers (with Al contents in the region between 10% and 50%), AlGaInAs for the barrier layers, and

InP compounds for the cladding, all on a InP crystal substrate. Then, the emission spectrum will be shifted to longer wavelengths compared to the above embodiment, and the optical device may for example be suited as light source in the NIR regime, for example from 1200 nm to 1800 nm.

As yet further variants ZnO based compounts may be used. ZnO has 3.37 eV energy bandgap, thus similar to the 3.39 eV of GaN. ZnO has a hexagonal structure and comparable other properties. In this case the compound ZnMgO would replace AlGaN in the above, and ZnCdO would replace InGaN; otherwise the structures can be made analogous to the above-described approach. A reference describing the ZnO based material systems is: Journal of Cryst. Res. Technol. 39, No. 1 1 (2004), incorporated herein by reference.

Referring to Figures 9 and 10 the principle of the one-segment-more-than-one- transition principle of one aspect of the invention is illustrated. Figure 9 shows a device with one active segment only, and therefore also one second electrode segment 72.1 (the first electrode being, in the manner illustrated in Fig. 7, arranged at the back side). In addition to the one segment the device may optionally further comprise an absorber segment, in which an absorber second electrode segment 72.3 is used to keep the PN junction formed by the heterostructure at zero bias (thus the absorber second electrode segment 72.3 is on ground potential as well), or to reverse bias the PN junction in the absorber segment.

Although according to aspects of the invention one segment is sufficient to produce radiation from more than one radiative transition, this does not exclude the use of multiple segments, for example to even further enhance the bandwidth of the device. Such a configuration is illustrated in Fig. 10, where there are two active segments corresponding to two second electrode segments 72.1, 72.2. Also in a configuration with multiple segments, an absorber segment may optionally be present as feedback suppressing means. A further feedback suppressing means of the embodiments shown in Figures 9 and 10 is the tilt of the waveguide 91 (thus, the waveguide is not perpendicular to the end faces of the device 1).

Various other embodiments may be envisaged without departing from the scope and spirit of the invention.