GB2095474A - Semiconductor light emitting devices - Google Patents

Abstract

A semiconductor light emitting device such as a laser or LED has a current-confining channel which is narrower (S'3) near the surface and wider (S4) where it meets the active layer 114. The channel may be trapezoidal in section, but preferably comprises a narrower upper channel portion 136 and a wider lower channel portion 138. The channel is defined by high-resistivity (e.g. proton bombarded) regions 132 and the upper channel portion may be further defined by means of a V-groove 134.1. Additionally the V-groove may be filled with semiconductor material. It is found that such devices have low capacitance and low minimum modulation current. In the case of lasers the spontaneous emission power is reduced. <IMAGE>

Description

SPECIFICATION Semiconductor light emitting devices This invention relates to light emitting semiconductor devices, such as lasers and LEDs, and more particularly to the confinement of current flow in these devices.

Nearly two decades ago light emitting semiconductor devices, especially those having a planar p-n junction in a monocrystalline semiconductor body, utilized broad area electrical contacts on opposite major surfaces of the body to apply forward bias voltage and pumping current to the junction. In an LED the resulting radiative recombination of holes and electrons in the active region in the vicinity of the junction generated spontaneous radiation. Primarily, one fundamental modification converted the LED to a laser: a cavity resonator was formed on the semiconductor body by a pair of parallel cleaved crystal facets orthogonal to the junction.When the pumping current exceeded the lasing threshold the radiation changed from spontaneous radiation, which in the LED was emitted from the active region essentially isotropically, to stimulated radiation, which in the laser was emitted as a collimated beam parallel to the junction and along the resonator axis. Of course, other design considerations played a role in making the advance from LED to laser, but these matters are not discussed here inasmuch as our purpose at this point is merely to state the now well-known kinship between p-n junction lasers and LEDs.

The broad area contacts (typically about 100 cm wide) on these devices caused the pumping current density at the p-n junction to be relatively low which therefore meant that relatively high currents (typically hundreds of mA in lasers) were required to achieve desirable radiation power levels. High currents in turn heated the semiconductor body and necessitated coupling the device to a suitable heatsink and/or operation of the device at cryogenic temperatures. The basic solution to this problem was then, and is today, to reduce the area of the p-n junction which has to be pumped so that for a given current density the amount of pumping current required is proportionately lower.One implementation of this solution is to constrain the pimping current to flow in a relatively narrow channel (typically about 12 jim wide) from a major surface of the semiconductor body through the active region.

On of the earliest structures for constraining current to flow in such a channel was the stripe geometry contact first proposed for semiconductor lasers by R.A. Furnanage and D.K. Wilson (U.S. Patent 3,363,195 issued on January 9, 1968). The stripe geometry reduced the threshold current for lasing (compared to lasers with broad area contacts) and limits the spatial width of the output beam. Since that early proposal numerous laser configurations have been devised to implement the stripe geometry concept: (1) the oxide stripe laser; (2) the proton bombarded laser; (3) the mesa stripe laser; (4) the reverse-biased p-n junction isolation laser; (5) rib-waveguide lasers; and (6) buried heterostructures of various types.

The most commonly used configuration for the past eleven years, however, has been the proton bombarded, GaAs-AIGaAs double heterostructure (DH) laser described, for example, by H.C. Casey, Jr. and M.B. Panish in Heterostructure Lasers, Part B, pp.207-210, Academic Press, Inc., New York, (1978). Despite its various shortcomings, lasers of this type have regularly exhibited projected lifetimes in excess of 100,000 hours and a number have axceeded 1,000,000 hours (based on accelerated aging tests). Long lifetimes have also been projected in DH LEDs employing different contact geometries (e.g., dot-shapes or annular rings) but similar proton bombardment to delineate the current channel.

Several of the shortcomings of porton bombarded DH lasers are discussed by R.W. Dixon et al in The Bell System Technical Journal, Vol.59, No.6, pp.975-985(1980). They explored experimentally the optical non-linearity (presence of "kinks" in the light-current (L-l) characteristics) and the threshold current distribution of AlGaAs, proton-bombardment-delineated, stripe geometry DH lasers as a function of stripe width (5,8, and 12 jim) in cases in which the protons did and did not penetrate the active layer.They demonstrated that shallow proton bombardment with adequately narrow stripes (e.g., 5 jim) can result in satisfactory optical linearity (kinks are driven to non-obtrusive, high current levels) without the threshold penalty that has been associated with narrow-stripe lasers in which the protons penetrate the active layer.

On the other hand, lasers with such narrow stripes have exhibited a statistically meaningful, although not demonstrably fundamental, decrease in lifetime. In addition, failure of the protons to penetrate the active layer increased device capacitance thus reducing speed of response and also increased lateral current spreading thus increasing spontaneous emission. In digital systems, the latter implies a higher modulation current to achieve a predetermined extinction ratio or a lower extinction ratio for a predetermined modulation current.

According to the present invention there is provided a semiconductor light emitting device comprising a semiconductor body having an active region in which optical radiation is generated when current flows therethrough and means within the body for constraining the current to flow from a surface of the body in a channel through the active region, the channel being narrower near the said surface and wider near the active region.

Some embodiments of the invention will now be described by way of example with reference to the accompanying drawings, in which: Figure 1 is an isometric view of a light emitting semiconductor device in accordance with our invention; Figure 2 is an end view of another semiconductor light emitting device in accordance with our invention; Figure 3 is an end view of another semiconductor light emitting device in accordance with our invention; Figure 4 is an end view of a mask structure for fabricating a light emitting device of the type shown in Figure 1 or Figure 2; Figure 5 and 6 depict end views of alternative masks for fabricating devices according to our invention by means of proton bombardment; Figure 7 is an isometric view of another semiconductor laser or LED in accordance with our invention;; Figure 8 is a cross-sectional view of another laser or LED in accordance with our invention; and Figure 9 is#a cross-sectional view of another laser or LED in accordance with our invention.

With reference now to Figure 1, there is shown a semiconductor light emitting device (laser or LED) comprising a semiconductor body 11 which includes an intermediate region 14. Region 14, which may have one or more layers, includes an active region which emits radiation 22 when pumping current is applied thereto. Electrode contacts 16 and 18 on body 11 are provided, along with a voltage source 20, to supply the pumping current. In addition, body 11 includes contraining means 32 which causes the pumping current to flow in a relatively narrow channel 36 from the top contact 16 through the active region after which the current may spread out to the bottom contact 18.

Before discussing our invention in detail, it will be helpful to discuss first the general attributes of a preferred configuration of a semiconductor light emitting device known as a double heterostructure (DH). As shown in Figures 1,2, and 3, a DH comprises first and second relatively wide bandgap, opposite conductivity type, semiconductor cladding layers 10 and 12, respectively, and, essentially latticed matched thereto, intermediate region 14 which is between and contiguous with the cladding layers. The intermediate region 14 includes a narrower bandgap active layer, here shown to be coextensive with the region 14, capable of emitting radiation when the cladding layers are forward baised. From the standpoint of quantum efficiency, it is well known that the active layer is preferably a direct bandgap semiconductor.Layers 10, 12, and 14 can be made of materials selected from a number of systems; for example, GaAs-AIGaAs or GaAsSb-AIGaAs for operation a short wavelengths in the 0.7-0.9 jim range approximately, and InP-lnGaAsP or InP-AIGalnAs for operation at wavelengths longer than about 1 jim (e.g., 1.1-1.6 jim).

Voltage source 20 forward biases the cladding layers and thereby injects carriers into the active layer.

These carriers recombine to generate spontaneous radiation in the case of an LED and predominantly stimulated radiation in the case of a laser. In either case, however, the radiation has a wavelength corresponding to the bandgap of the active layer material. Moreover, in the case of a laser or edge-emitting LED, as shown in Figure 1, the radiation 22 is emitted in the form of a beam along axis 23. In the laser the beam is collimated, and axis 23 extends perpendicular to a pair of resonator mirrows 24 and 26 formed illustratively by cleaved crystal facets or etched surfaces. These mirrors constitute optical feedback means for the stimulated radiation. In other applications, for example integrated optics, diffraction gratings may be employed as a substitute for one or both of the mirrors.

Although the electrode means depicted in the laser or edge-emitting LED of Figure 1 includes broad area contacts 16 and 18, it is well known in the art that these contacts can be patterned to form various geometrical shapes. Thus, in the case of transversely emitting LED, in which the light output is taken perpendicular to the layers, contact 16 is typically a broad area contact, but contact 18 might be an annular contact (not shown) which accommodates an etched hole (not shown) in one side of body 11. Where the bottom portions (e.g., substrate) of body 11 are absorbing, this etched hole is used to couple radiation to an optical fibre (not shown) positioned in the hole.

The conductivity type of the active layer is not critical. It may be n-type, p-type, intrinsic or compensated since in typical modes of operation under forward bias the number of injected carriers may exceed the doping level of the active layer. In addition, the intermediate region 14 may include a plurality of layers which constitute an active region, e.g. contiguous p-type and n-type layers of the same bandgap forming a p-n homojunction or of different bandgaps forming a p-n heterojunction. Furthermore, the heterostructure may have a configuration other than the simple double heterostructure for example a separate confinement heterostructure as described in U.S. Patent 3,691,476 or a strip buried heterostructure of the type described in U.S. Patent 4,190,813.

For CW laser operation at room temperature, the thickness of the active layer is preferably between approximately k12 and 1 .jim, where h is the wavelength of the radiation as measured in the semiconductor.

For operation at low thresholds, the thickness is typically 0.12 to 0.20 jim. However, for LED operation a thicker active layer, typically 2 to 3 jim, is suitable. In either case, for room temperature operation the laser or LED is typically bonded to a suitable heat sink, not shown.

In practice, the layers of a double heterostructure are typically grown by an epitaxial process such as liquid phase epitaxy (LPE), molecular beam epitaxy (MBE), or metallo-organic chemical vapour deposition (MO-CVD). Epitaxial growth takes place on a single crystal substrate 28 which may include a buffer layer (not shown) between the substrate 28 and the first cladding layer 10. Also, as shown in Figures 1 and 3, a contact-facilitating layer 30 is optionally included between the second cladding layer 12 and the top contact 16. The opposite contact 18 is formed on the bottom of substrate 28.

As mentioned previously, in order to constrain the pumping current generated by source 20 to flow in a relatively narrow channel 36 through the active region, high resistivity zones 32 are formed in the semiconductor layers, illustratively in layers 10,12, 14, and 30, by means well known in the art. Techniques for forming zones 32 include, for example, proton bombardment, oxygen bombardment, or suitable etching and regrowth of high resistivity material. Illustratively, the zones 32 have a resistivity of the order of 105-105 Q-cm, whereas channel 36 has a resistivity of only 0.1 #-cm so that typical ratios of resistivity are in the range of 106 : 1 to 107 : 1.

Trapezoidal channel structures As shown in Figure 1, current constraining means 32 forms a relatively high conductivity current flow channel 36 which is narrower (width S,) at its top near major surface 44 and wider (width S2) at its bottom near the active regon (i.e., layer 14). Constraining means 32 comprises laterally separate high resistivity regions 32.1 and 32.2. which bound the channel 36 along its oblique sides 36.1. Although these sides are depicted as straight lines, in practice a linear relationship is not necessary and may not result from actual processing techniques.

We have found that the above shape of the current channel has important effects on device performance.

The narrower channel width at the top increases the current density and thereby the power at which kinks occur. The depth of the high resistivity regions, which preferably extend through the active region 14, affects device capacitance and the amount of spontaneous emission generated in lasers. These matters will be discussed in greater detail later.

Alternatively, as shown in Figure 2, the channel 36 formed by high resistivity regions 32.1 and 32.2 need not reach major surface 44. However, in order that the device resistance be not too high, a dopant may be diffused or otherwise introduced into the surface 44 so as to create a highly conducting diffusion front 45 which penetrates channel 36. In this case, the width S, at the top of the channel 36 is defined by the intersection of the front 45 and the oblique sides 36.1.

Moreover although the channel 36 depicted in Figure 1 is in the shape of a trapezoidal prism extending parallel to axis 23, in the case of a transversely-emitting LED channel 36 might have the shape of a right conical frustum having its axis perpendicular to the layers.

Hi-lo structures )In the device shown in Figure 3, current constraining means 32 has a bi-level or stepped configuration forming a pair of coupled channels 36a and 36b. In particular, means 32 includes first means 32.1 a-32.2a defining a relatively narrow upper channel 36a and second means 32.1 b-32.2b defining a relatively wider lower channel 36b. The constraining means 32 as shown comprises high resistivity regions 32.1-32.2 which bound relatively high conductivity channels 36a and 36b. The regions 32 include (1) upper zones 32.1 a and 32.2a and lower zones 32.1 b and 32.2b. The upper zones are separated by a relatively narrow S1 and extend from the upper major surface 44 of body 11 to a depth d1 short of the active region, thereby defining the narrow upper channel 36a.In contrast, the lower zones are separated by a relatively wider distance S2 > S and extend from the depth d1 into or through the active region, thereby defining the wider lower channel 36b.

Channels 36a and 36b may have the approximate shape of parallelepipeds extending perpendicular to the plane of the paper, as in a laser or edge-emitting LED; or in a transversely-emitting LED may form cylinders extending transverse to the layers.

When the high-resistivity regions 32 were fabricated by proton bombardment in GaAs-AIGaAs lasers, this Hi-Lo structure exhibited several advantages. First, the narrow upper channel 36a increased the current density in the active region and thereby caused kinks to be shifted to satisfactorily high current levels out of the range of typical laser operation compared to wide (e.g., 12 jim) stripe geometry DH lasers. Second, this feature also resulted in more unformlydistributed and lower lasing threshold lasers, providing higher yields.

Third, because the wider lower channel 36b reduced lateral current diffusion and spreading, less spontaneous radiation was emitted outside the resonator of the laser, thereby allowing for lower minimum modulation currents for predetermined extinction ratios in digital applications. Fourth, the latter feature resulted in reduced device capacitance for both lasers and LEDs, thereby permitting high speed of operation (i.e., higher pulse repetition rates in digital applications).

To reduce device capacitance the proton bombardment should penetrate the p-n junction which, in a conventional DH, is located at one of the interfaces between active layer 14 and cladding layers 20 and 12.

However, to reduce spontaneous emission, the protons preferably penetrate through the active region where recombination occurs.

Fabrication of trapezoidal channels As shown in Figure 4, one way to fabricate a trapezoidal channel of the type shown in Figure 1 is to epitaxially grow a removable semiconductor layer on major surface 44 and by well-known photolithography and preferential etching techniques to pattern the layer to form inverted trapezoidal openings 54 which expose portions of surface 44. Between the openings, the remaining segments 52 of the removable layer form trapezoidal attenuation masks. For a Group Ill-V compound semiconductor layer, the oblique side walls 56 of the remaining segments correspond to (1 1 1A) crystallographic planes which make an angle of about 55 degrees with a (100)-oriented surface 44.

Alternatively, the openings in the removable layer may be etched as inverted trapezoids so that the remaining segments 52 are trapezoids. In either case, therefore, the trapezoids and inverted trapezoids are complementary.

Bombardment of the masked surface 44 with particles 50 (e.g. protons, oxygen) results in deepest penetration between the segments, no penetration under the central (thickest) parts of the segments, and gradually decreasing penetration under the oblique sides of the segments. Of course, a thinner mask segment would allow some penetration under the central parts of the segments, a technique which would be useful in realizing the channel configuration of Figure 2.

After bombardment is completed and before metallization to form electrical contacts, the attenuation masks are removed. To this end, it is preferable that the material of mask 52 be different from that portion of body 11 adjacent surface 44 so that stop-etch procedures can be advantageously employed. For example, surface 44 is typically GaAs in which case mask 52 could be AlGaAs and a well-known HF etchant or iodine etchant (e.g., 1139 KI, 65g 12, 1 00cc H2O) could be used as a stop-etch to remove mask 52. Plasma stop-etching may also be used as a substitute for wet-chemical procedures. Finally, it should be noted that a buffered peroxide solution is also a preferential etchant and can be used to etch the openings which form mask segments 52.

Formation of the removable layer also gives a fringe benefit related to the cleanliness of the epitaxial growth process. When liquid phase epitaxy is used to fabricate the semiconductor layers of these devices, the last grown layer typically gets contaminated from various sources, especially globules of the molten metal (e.g., Ga) used as the source solutions. Consequently, this last layer, which is usually the cap or contact-facilitating layer 30 (Figures 1-3), has to be cleaned by etching, a step which requires careful control since layer 30 is typically very thin (e.g., 0.5 jim). In the process described here, however, the last-grown layer is the attenuation mask which can be much thicker (e.g., 3.0 jim) and can be readily removed by stop-etch techniques as mentioned above.

Fabrication of Hi-Lo structures A number of fabrication techniques can be employed to fabricate our Hi-Lo structure. As mentioned previously, the high resistivity regions 32 can be formed by proton bombardment, oxygen bombardment, or etching and regrowth of high resistivity material. For purposes of explanation, however, we assume that these regions are formed by proton bombardment.

One straightforward technique would entail two proton bombardment steps and two masks. In the first step a proton attenuation mask S1 wide and protons of energy E1 (e.g., 150 keV) would be used to delineate narrow upper channel 36a. In the second step a proton attenuation mask S2 wide and protons of energy E2 > E1 (e.g., E2 = 300 keV) would be used to delineate wide lower channel 36b.

Delineation of the channels 36a and 36b in a single proton bombardment step is also possible. To do so a compound attenuation mask having higher proton attenuation in the centre and lower attenuation on the sides can be used. Two versions of this type of mask are depicted in Figures 5 and 6. In each case a thick metal pad 40 of width Sa is formed on top of a plateau 42 which in turn is formed on the major surface 44 nearest active region 14. Pad 40 essentially totally attenuates the protons 50 so that no proton damage occurs in the narrow channel 36a, and plateau 42 only partially attenuates the protons 50 so that damaged zones 32.1 a and 32.2a extend to a depth d, short of the active region.Outside the plateau 42, the mask provides virtually no attenuation either in Figure 5 (because the mask does not extend that far) or in Figure 6 (because the mask is very thin there). Thus, outside the plateau 42 proton damaged zones 32.1 b and 32.2b extend to a depth d2 and penetrate the active region 14. Preferably, as shown, these damaged zones 32.1 b and 32.2b extend through the active region 14. Illustratively, in Figures 5 and 6 pad 40 comprises plated Au.

Plateau 42 in Figure 5 comprises layers of Au (42.1), Pd or Pt (42.2), and Ti (42.3) and in Figure 6 comprises a mesa of SiO2 (42.4) overlayed with Ti-Pt layers (42.5).

The following examples illustrate in more detail how masks of this type were used to fabricate light emitting devices. In each of the two examples, the semiconductor body 11 comprised a (100)-oriented, n-GaAs substrate 28 on which were grown by standard LPE the following epitaxial layers: an n-GaAs buffer layer (not shown); an n-Al3#Ga64As cladding layer 10 about 1.5 jim thick; a p-Al.08Ga.92As active layer 14 about 0.15 jim thick; a p-AI 36Ga 64As cladding layer 12 about 1.5 jim thick; and a highly doped p-GaAs cap layer 30 about 0.5 jim thick. The completed wafer (body 11 plus epitaxial layers) was processed as follows to fabricate light emitting devices, particularly, lasers.

Example I To fabricate lasers using the compound attenuation mask 40-42 of Figure 5, a lift-off photoresist mask was deposited on surface 44, and standard photolithographic techniques were used to open an elongated stripe window 12 jim or 18 jim wide perpendicular to the {1 1 0} cleavage planes. Ti, Pd, and Au layers 42.3, 42.2, and 42.1 were sequentially deposited using a vacuum E-gun system. The deposition rate was controlled by a commercially available monitoring system so that the Ti, Pd, and Au layers had thicknesses of 1000 Angstroms, 1500 Angstroms, and 5000 Angstroms, respectively. The total thickness of 0.75 jim for plateau 42 was selected to provide 50 percent reduction in the penetration depth of 300 keV protons 50. The stripe geometry plateau 42 was then formed by well-known etching procedures to lift-off the photoresist mask.

Next, the pad 40 was formed also in the shape of a 5 jim wide stripe by electroplating Au to a thickness of about 1-2 jim using standard photolithographic procedures. The Au pad 40 provided a substantially complete barrier to the high energy (300 keV; dosage 3x1 015cm#2) protons 50, thus forming narrow upper channel 36a of width Sa = 5 jim and wider lower channel 36b of width S2 = 12 jim or 18 jim. Between S, and S2 the plateau 42 provided only partial attenuation so that protons penetrated to a depth d, = 1.5 jim.

Outside S2 no attenuation mask was present, and protons penetrated to a depth d2 = 2.8 jim and hence extended through the active layer 14.

Example II To simplify the fabrication procedure of Example I, the Ti-Pd-Au plateau 42 was replaced as shown in Figure 6 with dielectric stripe 42.4 (e.g., SlO2 or Si3N4) overlayed with Ti-Pt layer 42.5. This compound mask was made by depositing about 1.0-1.2 jim of SiO2 on surface 44 using standard vapour phase techniques.

This thickness was again chosen to provide 50 percent attenuation to the 300 keV protons 50. Next, the SiO2 layer was photolithographically delineated and etched in standard buffered HF etchant to form stripes 12 jim or 18 jim wide perpendicular to the (1 10) cleavage planes. After removing the photolithography mask, the SlO2 stripe 42.4 and the surface 44 were covered with 1000 Angstoms of Ti and then 1500 Angstroms of Pt by standard evaporation procedures. Finally, pad 40 was formed in the shape of a 5 jim stripe 1-2 jim thick using standard photolithography and electroplating techniques.As before, the masked wafers were subjected to 300 keV protons in a dosage of 3x 1015cm-2 to form simultaneously the narrow upper channel 36a and the wider lower channel 36b. In this case layer 42.5 reduced the proton energy so that d2 decreased to about 2.3 jim.

In both Examples I and II, after proton bombardment was completed, the compound masks 40-42 were removed from surface 44 by means of an HF etchant. This step also prepared the surface 44 for subsequent metallization to form standard p-metal contacts.

Experimental results--Hi-Lo structures In order to provide a standard for comparison, one half of each wafer in Examples I and II was processed into control lasers having 5 jim wide stripes with shallow proton bombardment (150 keV). Each remaining half wafer was processed as above into Hi-Lo lasers using compound masks 40-42 of three types: Type (1) - 5 jim wide Au pad 40 on 18 jim wide SiO2/Ti-Pt plateau 42 (Example II); Type (2) - 5 jim wide Au pad 40 on 12 jim wide SiO2/Ti-Pt plateau 42 (Example II); and Type (3) - 5 jim wide Au pad 40 on 18 jim wide Ti-Pd-Au plateau 40 (Example I).

Comparisons set forth in the table below were based on a number of parameters: spontaneous emission power SL at 50 mA of drive current; slope ASL of spontaneous emission portion of the L-l curve; capacitance C measured at 1 MHz (average C is listed below); and minimum modulation current MMI which is defined as the difference in current between upper and lower light power levels P2 and P" respectively, which yield a light intensity extinction ratio ER between ON and OFF states when the laser is pulsed (the median MMI is listed below for ER = 15:1, P2 = 2.5 mW, and P, = 0.167 mW).

LASER TYPE Parameter Control Type(1) Control Type(2) Control Type(3) Sv(mW) 0.200 0.151 0.142 0.042 0.071 0.041 AS#(mW) 0.29 0.28 0.20 0.07 0.12 0.06 MMI (mA) 59 45 76.5 24 70.5 26 C(pf) 83 21 115 35 54 12 In addition to the data shown in the table, we found that 90 percent of the Type (2) lasers had MMls within a specified 30 mA MMI with a statistical variance 20 = 3 mA. Similarly, 75 percent of the type (3) lasers had MMI within 30 mA whereas none of the corresponding control lasers did. These results imply improved device yield.

Note that the Type (2) lasers, which have S2 = 12 jim wide stripes, exhibit the largest decrease in 5L and the highest yield for an MMI S 30 mA, but these advantages alone do not necessarily dictate the use of this stripe width. Consideration should be given to the impact on the light power output level Pk at which kinks occur. In general, we found that kink formation occurred at higher Pk in the control lasers than in Hi-Lo lasers, but the latter were still well within specifications (i.e., Pk 3 3 mW). Type (1) lasers showed little change in Pk.

However, the Type (2) lasers, which utilized the narrowest attenuation masks (S2 = 12 jim), showed a marked reduction of about 50 percent in Pk compared to corresponding control lasers. In contrast, Type (3) lasers, which had S2 = 18 jim, had a smaller reduction of about 35 percent in Pk. This data suggests that it may be advantageous for the width S2 to be between 12 jim and 18 jim.

In an alternative form of device the means defining the narrow upper channel 36a is realised by means of a groove etched into the upper surface 44. Such a groove in combination with a wider lower channel 36b is expected to have features and advantages comparable to those described above. The details of such structure are discussed in connection with Figures 7 to 9.

With reference now to Figure 7, there is shown a semiconductor light emitting device (laser or LED) analogous to that of Figures 1-6 comprising a semiconductor body 111 which includes an intermediate region 114. Region 114, which may include one or more layers, includes an active region which emits predominantly stimulated radiation 122 in the case of a laser or spontaneous radiation in the case of an LED, when pumping current is applied thereto. Electrode contacts 116 and 118 on body 111 are provided along with a voltage source 120, to supply the pumping current In addition, body 111 includes constraining means 132-134 which cause the pumping current to flow in a relatively narrow channel 136-138 from the top contact 116 through the active region after which the current may spread out to bottom contact 118.

As shown in Figure 7, current constraining means comprises V-groove first means 134 which extends from major surface 144 to a depth d5 short of the active region, thereby defining relatively narrow upper channel 136; and further comprises laterally separate, high resistivity regions 132 which bound lower wider channel 138, and which extend from at least depth ds approximately, to the active region (i.e., into or through the active region). As shown, separated regions 132 illustratively extend from surface 144 and preferably through the active region. V-groove 134 is positioned within the space between regions 132. However, it is not essential that the high resistivity regions 132 actually reach all the way to the major surface 144.In fact, for contacting purposes it may be advantageous to have a high conductivity layer interposed between regions 132 and contact 116 as described in U.S. Patent 4,124,826.

With reference to the DH of Figure 7, the V-groove 134 has a width S3 at major surface 144 and a depth d5 where it penetrates second cladding layer 112, thereby defining upper channel 136 as having essentially the same width. In contrast, the high resistivity regions 132 are separated by a wider distance S4 > S3 and extend from surface 144 to a depth d6 > d5 into and preferably through the active region, thereby defining the wider lower channel 138 of the width S4.

Alternatively, as shown in Figure 8, the upper channel 136 can be further restricted by additional high resistivity regions 132.1 which bound a portion of the oblique sides 134.1 of V-groove 134, thus defining the upper channel width 53' of Figure 8 as being less than S3 of Figure 7. In practice, the regions 132 and 132.1 can be fabricated (e.g., by proton bombardment) to depths of d6 and d7, respectively (d6 > d7); and then the V-groove 134 can be etched to a depth d5 so as to penetrate the regions 132.1 (d7 < d5 < d6).

These V-groove configurations are expected to exhibit several advantages. First, the narrow upper channel 136 increases the current density in the active region and thereby causes kinks in lasers to be shifted to higher current levels out of the range of typical operation. Second, this feature should also result in lower and more uniformly distributed lasing thresholds, providing higher device yields. Third, because the wider lower channel 138 reduces lateral current diffusion and spreading, less spontaneous radiation is emitted outside the resonator of the later, thereby allowing for a lower modulation current for a predetermined extinction ratio in digital applications. Fourth, the latter feature results in reduced device capacitance for both lasers and LEDs, thereby permitting higher speed of operation (i.e., higher pulse repetition rates in digital applications).

Other arrangements can be devised in accordance with these principles by those skilled in the art without departing from the spirit and scope of the invention. In particular, the V-groove 34 of Figure 7 or Figure 8 may be refilled with semiconductor material, resulting in devices (especially lasers) with several useful characteristics as discussed below. Moreover, although the groove has been described as a V-groove, its precise geometrical shape is not critical. A V-groove results when Ill-V semiconductors are subject to certain etchants which preferentially etch crystallographic planes, but V-grooves or rectangular grooves might result from other etchants or other processes (e.g., ion beam milling or plasma etching).

As shown in Figure 9, the V-groove has been filled with semiconductor material 134' and, depending on the procedure used to effect refilling, this may or may not result in formation of layers 134.2 which are adjacent the V-groove and/or major surface 144. Moreover, depending on the material of cladding layer 112 and the type of procedures used, material 134' may or may not be epitaxial (i.e. monocrystalline).

Several embodiments result depending on the relative size of the bandgaps Eg of the DH layers relative to that of V-groove material 134'. Case l: Eg (134') > Eg (114); that is, the V-groove material 134' has a larger bandgap than the active layer 114. As a consequence, laser radiation penetrating V-groove material 134' experiences reduced absorption as compared with Figure 7. Case II: Eg (134') > Eg (112) > Eg (114); in addition, cladding layers 110 and 112 and the active layer 114 all have the same conductivity type, and V-groove material 134' and cladding layer 112 have opposite conductivity types. This configuration is a form of isotype laser in which the p-n junction is located along oblique surfaces 134.1. In this case, the V-groove material 134' is preferably monocrystalline. Case Ill: Eg(l 12) > Eg (134') > Eg (114); that is, the V-groove material 134' has a lower bandgap than cladding layer 112 but a higher bandgap than active layer 114. As a consequence, the refractive indicesn have the relationship n (114) > n (134') > n (112) sothatthe laser radiation would be refractive index guided along the V-groove.

Claims (22)

1. A semiconductor light emitting device comprising a semiconductor body having an active region in which optical radiation is generated when current flows therethrough and means within the body for constraining the current to flow from a surface of the body in a channel through the active region, the channel being narrow near the said surface and wider near the active region.

2. A device as claimed in claim 1 wherein the channel comprises a relatively narrow upper channel portion which extends from the surface to a depth short of the active region and a wider lower channel portion which extends from the said depth to the active region and the constraining means comprises first means for confining the current to the upper channel portion and second means for confining the current to the lower channel portion.

3. A device as claimed in claim 2 wherein the second means comprises a pair of high resistivity second zones bounding the lower channel portion.

4. A device as claimed in claim 3 wherein the second zones extend through the active region.

5. A device as claimed in any of claims 2 to 4 wherein the first means comprises a pair of high resistivity first zones bounding the upper channel portion;

6. A device as claimed in any of claims 3 to 5 wherein the high resistivity zones are proton bombarded zones.

7. A device as claimed in any of claims 2 to 6 comprising a first cladding layer, a second cladding layer nearer the surface than the first layer and the active region comprising an active layer between the cladding layers and wherein the upper channel portion extends from the surface to the said depth which is located in the second cladding layer, and the lower channel portion extends from the said depth through the active layer.

8. A device as claimed in any of claims 2 to 7 wherein the upper channel portion is about 5 jim wide and the lower channel portion is between 12 and 18 jim wide.

9. A device as claimed in any of the preceding claims having a laser resonator axis along which, in use, radiation propagates and wherein the constraining means defines the channel as an elongated region extending substantailly parallel to the axis.

10. A device as claimed in any of claims 2 to 8 for use as a light emitting diode wherein the constraining means defines the channel portions as cylinders whose axes extend perpendicular to the active region.

11. A device as claimed in any of claims 2 to 7 wherein the first means includes an elongated groove defining the upper extremity of the upper channel portion.

12. A device as claimed in claim 11 wherein the first means comprises a portion of the surface having a V-groove formed therein and a pair of laterally separated high resistivity first regions extending to the oblique sides of the V-groove, the V-groove extending below the first regions.

13. A device as claimed in claim 11 or 12 wherein the first means further includes semiconductor material filling the groove.

14. A device as claimed in claim 13 wherein the material filling the groove has a larger bandgap than the adjacent portions of the body.

15. A device as claimed in claim 13 or claim 14 wherein the material filling the groove has the same conductivity type as the adjacent portions of the body.

16. A device as claimed in claim 13 wherein the material filling the groove has opposite conductivity type to the adjacent portions of the body, thereby forming as p-n junction at the sides of the groove effective to inject carriers into the active region.

17. A device as claimed in claim 13 wherein the material filling the groove has a smaller bandgap than the adjacent portions of the body and a larger bandgap than the active region.

18. A method of forming a device as claimed in claim 1 wherein the channel is of trapezoidal section and is formed by a) epitaxially growing a semiconductor layer on a major surface of a semiconductor body, b) patterning the layer to form a mask having oblique side walls, c) subjecting the body to particle bombardment so as to form high resistivity zones in the region exposed by the mask and under the oblique side walls, and d) removing the mask from the surface.

19. A method as claimed in claim 18 wherein the layer and the portion of the body adjacent the surface are of different semiconductor materials, and in step (d) the masks are removed by a stop-etch procedure.

20. A method as claimed in claim 18 or claim 19 wherein the layer and the body are of Group Ill-V compounds, the surface as a (100) crystal orientation, and in step (b) the openings are formed by exposing the layer to an etchant which preferentially etches along (11 1A) crystallographic planes.

21. A semiconductor light-emitting device substantially as herein described with reference to the accompanying drawings.

22. A method of making a semiconductor light-emitting device substantially as herein described with reference to the accompanying drawings.